Public Management

How effective are public managers as they seek to influence how public organizations
deliver policy results? How, and how much, is management related to the perform-
ance of public programs? What aspects of management can be distinguished? Can
their separable contributions to performance be estimated?
The fate of public policies in today’s world lies in the hands of public organizations,
which in turn are often intertwined with others in latticed patterns of governance.
Collectively, these organizations are expected to generate performance in terms of
policy outputs and outcomes. In this book, two award-winning researchers investigate
the effectiveness of management in the public sector. First, they develop a systematic
theory on how effective public managers are in shaping policy results. The remainder
of the book then tests this theory against a wide range of evidence, including a data set
of 1,000 public organizations.

Laurence J. O’Toole, Jr., is the Margaret Hughes and Robert T. Golembiewski

Professor of Public Administration and Distinguished Research Professor at the
University of Georgia, Athens. He also holds an academic appointment as a professor in
the Faculty of Management and Governance at Twente University in the Netherlands.
Kenneth J. Meier is the Charles H. Gregory Chair in Liberal Arts at Texas A&M
University, College Station. He is also Professor of Public Management at Cardiff
Business School.
Public Management
Organizations, Governance, and Performance

Laurence J. O’Toole, Jr.,

and
Kenneth J. Meier
cambridge university press
Cambridge, New York, Melbourne, Madrid, Cape Town, Singapore,
São Paulo, Delhi, Dubai, Tokyo, Mexico City

Cambridge University Press

The Edinburgh Building, Cambridge CB2 8RU, UK

Published in the United States of America by Cambridge University Press, New York

www.cambridge.org
Information on this title: www.cambridge.org/9781107004412

# Laurence J. O’Toole, Jr., and Kenneth J. Meier 2011

This publication is in copyright. Subject to statutory exception

and to the provisions of relevant collective licensing agreements,
no reproduction of any part may take place without
the written permission of Cambridge University Press.

First published 2011

Printed in the United Kingdom at the University Press, Cambridge

A catalogue record for this publication is available from the British Library

Library of Congress Cataloging-in-Publication Data

O’Toole, Laurence J., 1948–

Public management : organizations, governance, and performance / Laurence J. O’Toole,
Kenneth J. Meier.
p. cm.
ISBN 978-1-107-00441-2 (Hardback)
1. Public management. 2. Administrative agencies–Management. I. Meier, Kenneth J., 1950–
II. Title.
JF1351.L68 2011
351–dc22
2010040919
ISBN 978-1-107-00441-2 Hardback

Cambridge University Press has no responsibility for the persistence or

accuracy of URLs for external or third-party internet websites referred to
in this publication, and does not guarantee that any content on such
websites is, or will remain, accurate or appropriate.
 Contents

List of figures page vi

List of tables vii
Preface xi

1 Public management and performance: an evidence-based

perspective 1

2 A model of public management and a source of evidence 23

3 Public management in interdependent settings: networks,

managerial networking, and performance 55

4 Managerial quality and performance 100

5 Internal management and performance: stability, human resources,

and decision making 131

6 Nonlinearities in public management: the roles of managerial

capacity and organizational buffering 182

7 Public management in intergovernmental networks: matching

structural networks and managerial networking 240

8 Public management and performance: what we know,

and what we need to know 267

Glossary 287
References 291
Index 309

v
 Figures

4.1 Linear and nonlinear relationship page 120

4.2 Controlling for quality 121
4.3 Differences by size of central staff 123
5.1 The interaction of management quality with networking and
stability: quartile regression coefficients 146
5.2 The interaction of superintendent stability with teacher stability,
management quality, and networking 146
5.3 The interaction of networking with management quality and stability 147
5.4 The interaction of teacher stability with superintendent stability,
management quality, and networking 147
6.1 The marginal impact of missed school days contingent on
managerial capacity 206
6.2 The marginal impact of Katrina students contingent on
managerial capacity 206

vi
 Tables

3.1 Factor loadings for nodes involved in managerial

networking, 2007 page 60
3.2 Levels of interaction are fairly constant over time 60
3.3 Managerial networking improves environmental support 62
3.4 Mean scores for interactions between different groups
of officers in English local authorities 65
3.5 Factor loadings for managerial networking, 2000 survey data 71
3.6 Management and organizational performance: additive
linear estimation 73
3.7 Management and organizational performance: nonlinear impacts 74
3.8 Management and organizational performance: autoregressive
and nonlinear models 76
3.9 Managerial networking at different levels of educational performance 79
3.10 Managerial networking interactions with resources and constraints 80
3.11 The impact of network interaction on disadvantaged student
indicators 88
3.12 The impact of network interaction on advantaged student indicators 88
3.13 The impact of network interaction on advantaged student indicators 90
3.14 The impact of individual network nodes on disadvantaged
student indicators 90
4.1 Determinants of superintendent salaries 109
4.2 The impact of management on performance: standardized tests 112
4.3 Management quality and other measures of performance 113
4.4 Management quality in an autoregressive specification 114
4.5 The nonlinear relationship between networking and performance 119
4.6 Nonlinear effects of networking, controlling for quality 121
4.7 Nonlinear effects of networking on performance: impacts
of the relative size of central administrative staff 122
5.1 The impact of management on performance: standardized tests II 141
5.2 Management, stability, and other measures of performance 143
vii
viii List of tables

5.3 Management, stability, and performance: autoregressive estimation 143

5.4 Measuring the quality of human capital: factor loadings 154
5.5 The reliability of the human capital measure 155
5.6 The impact of human capital on organizational performance 156
5.7 The impact of human capital on alternative indicators of performance 158
5.8 The autoregressive impact of human capital on alternative
indicators of performance 158
5.9 The impact of a 10 percent or greater budget shock on
organizational performance 165
5.10 The impact of budget shocks on other indicators of performance 166
5.11 Superintendent priorities: what is your primary goal
for improving your district? 167
5.12 The nonbarking dog: the relationship between a budget
shock and instructional expenditures 168
5.13 Redistributing funds to core functions: percentage point
reductions in allocations 169
5.14 Reducing core costs: teachers’ salaries and class size 170
5.15 Reducing core support tasks: instruction, aides, and support staff 171
5.16 Seeking less expensive core personnel 171
6.1 The impact of a 10 percent or greater budget shock on
students’ state examination performance 190
6.2 The impact of a 10 percent or greater budget shock on the
performance of college-bound students 191
6.3 The impact of a 10 percent or greater budget shock on students’
state examination performance 192
6.4 The impact of a 10 percent or greater budget shock on the
performance of college-bound students 194
6.5 Organizational shocks: Hurricanes Katrina and Rita 200
6.6 Environmental shocks and student performance: the impact
of students and missed class days 202
6.7 Administrative capacity can overcome the impact of
environmental shocks 205
6.8 How the interaction of managerial networking and
managerial capacity affects organizational performance 212
6.9 How the interaction of managerial networking and managerial
capacity affects organizational performance: attendance 215
6.10 How the interaction of managerial networking
and managerial capacity affects organizational performance:
college readiness 216
ix List of tables

6.11 The impact of buffering on organizational performance 228

6.12 The impact of buffering on alternative indicators of performance 228
6.13 Buffering: is the relationship linear or reciprocal? 230
6.14 Does the interaction with M2 add explanation to a linear model? 230
6.15 Does buffering interact with resources in a reciprocal manner? 231
7.1 Management and dependence on state aid 253
7.2 Management and dependence on state aid II 253
7.3 Management and dependence on state aid: college aspirations
indicators 254
7.4 Management and state aid: low-end indicators 254
7.5 Management and funding diversity 255
7.6 Management and funding diversity II 255
7.7 Management and funding diversity: college aspirations indicators 256
7.8 Management and funding diversity: low-end indicators 256
7.9 Summary of results: state aid 257
7.10 Summary of results: funding diversity 257
8.1 Intercorrelations of management measures 275
8.2 Practical lessons for managers 276
 Preface

How effective are public managers as they seek to influence the efforts of
public organizations to deliver policy outputs and outcomes? How, and how
much, is management related to public program performance? What aspects
of management can be distinguished, and can their separable contributions
to performance be estimated? How do managers deal with internal operations,
opportunities in the environment, and threats or shocks from outside the
organization? Can the networking behavior of managers and the networked
structures in which many public organizations sit shape policy results – for
good or ill?
In this book we address these salient questions – and more. Whereas in
an earlier volume (Meier and O’Toole 2006) we explored the relationship
between democratic governance systems and public bureaucracy via the
literatures of political science and public administration, and used empirical
analyses to sort through the issues, in this new book we put the politics-and-
administration theme to one side – mostly – to focus on management and
performance. This approach does not mean that we reconstruct some sort
of implicit politics/administration dichotomy. Indeed, the role that public
managers occupy includes some highly political elements, and some of our
work on management and performance demonstrates some explicitly polit-
ical patterns. It turns out, for instance, that managerial interactions with
external stakeholders shape the outcomes of their organizations in ways that
reflect the distribution of power in their settings. It is also the case, further-
more, that maintaining personnel stability within the organizations we have
been studying provides particularly strong benefits for the least advantaged
clientele of the agencies. Public management certainly involves political
themes and potentially controversial outcomes. What is distinctive about
this volume, instead, is that our focus is directly on managers and how they
shape results.
Examining the link between public management and performance might
seem to be covering rather well-trodden ground. Until fairly recently,
xi
xii Preface

however, the contributions of management to performance were either

contestable or very difficult to estimate. One major theoretical approach,
for example, the population ecology view of organizations, has strongly
argued that management does not matter, that organizations succeed or fail
because they are lucky. By “lucky,” population ecology theorists mean that
some organizations have favorable environments with ample resources and
are assigned tasks that are tractable. What might appear to be good manage-
ment, therefore, is an artifact of an organization’s niche. Other analysts,
for instance some who employ public choice as a theoretical lens, might even
see public managers as impediments (“rent seekers”) – or, at best, inefficient
contributors – to the delivery of goods and services to the public.
At what might seem like the other extreme, a major international move-
ment, the new public management (NPM), holds that management is the
key to effective public programs and that, if governments were to adopt
NPM’s set of favored reforms (mostly borrowed from the private sector),
citizens would get better government at a lower cost. The overly strong
versions of this argument can be considered “managerialism”: public man-
agement as the potential magic cure that converts failure into success.
Despite these contesting theoretical approaches, the body of systematic
research on whether and how much management really matters has been
relatively slim – particularly so if one restricts the search to investigations
drawing evidence from large numbers of public organizations and employing
archival rather than perceptual measures of performance.
In recent years, however, the number of studies systematically examining
the links between management and performance has grown. We incorporate
many of these recent findings in this book. In the main, though, this volume
integrates much of our own systematic theoretical and empirical work on
public management and performance conducted over an extended period.
The approach we take allows for the development of relatively clear and
evidence-based answers to the questions of how effective various facets of
management can be – that is, how much difference they make. This book
develops its perspective and findings on such issues in a cumulative and
progressively nuanced fashion over the course of several chapters.
We have opted to distill our reading of the theoretical literature into
a simple, parsimonious theory about how organizations are managed. Our
theory is built around a set of five principles. First, public organizations are
autoregressive (or inertial) systems; they seek to counter the natural ten-
dency toward entropy in the environment. As a result, what organizations
do today will be very similar to what they have done in the past. Second,
xiii Preface

public management can be divided into two broad parts: managing within
the organization and managing the organization’s relationships with the
environment. Both can be expected to have performance implications.
Third, external management can be divided, at least theoretically, into
efforts to exploit opportunities in the environment and efforts to buffer
the organization from threats that the environment might generate.
Fourth, managers use structures, systematic processes, and procedures
to regularize organizational actions. Put succinctly, organizations and their
managers organize. They set up stabilizing routines that embed knowledge
and experience so that cases can be handled quickly and consistently. Finally,
the relationships between variables – that is, management, stabilizing elem-
ents such as structure, and the environment – are nonlinear. In simple terms,
this point means that management is not just another input to program
performance but, rather, that it interacts with a variety of other factors
and can produce large gains in effectiveness relative to the resources that
management consumes.
Our approach is to formalize these principles via a mathematical model
and then to test aspects of it systematically through most of the chapters of
this volume. In particular, we focus on large panel data sets that permit the
inclusion of a wide variety of control variables. In this fashion we can isolate
the independent effects of public management as the key variable, instead of
mistaking as management effects those that are actually due to some other
factor that has been omitted from the analysis.
The most frequently used data set in the book, and perhaps in the field of
public management, is the Texas school district data set, which we have built
and refined for ten years. It has several significant advantages over other
public management data bases, as we explain in this volume. Although Texas
schools constitute the primary data set, we recognize that even a state as
diverse as Texas cannot contain all the relevant organizational and environ-
mental variables. In this book, therefore, we incorporate additional analyses
we have undertaken with data on local police departments, local govern-
ments in the United Kingdom, and state unemployment insurance agencies.
Further, we explain relevant findings garnered in others’ research, as appro-
priate, at several points in the coverage. We make no claim that we have
looked at the universe of organizations and organizational characteristics,
but we have assembled research results from several thousand public
organizations.
The work represented here stems from a decision made about a dozen
years ago to pool our interests in order to develop a long-term research
xiv Preface

agenda centered around the management-and-performance theme. One of

us, O’Toole, had developed theoretical ideas about management and per-
formance, especially in complex institutional settings including networks,
and had worked empirically in a number of fields of public policy. Meier
had also undertaken numerous studies in multiple policy fields, including
public education, and was especially experienced at tapping the advantages
of large-N statistical approaches. This book reflects a merging of these
interests and an effort to speak broadly to the field of public management.
We gratefully acknowledge assistance we have received from others.
We owe a special debt to school district superintendents in Texas, since
they have served multiple times as respondents to our surveys, which are
designed to learn about the management of those public organizations.
These busy public managers have tolerated our questions thus far through
five separate survey efforts (plus one more specialized one) over a ten-year
period, with more currently planned. Four reviewers for Cambridge Univer-
sity Press offered thoughtful assessments of this book and helpful suggestions
for strengthening it. We also thank George Boyne, Alisa Hicklin, and Richard
Walker for permitting our use of some data in this volume that had been
gathered for joint research efforts with these colleagues. We are grateful to
doctoral students and alumni of doctoral programs in the Department
of Public Administration and Policy at the University of Georgia, and the
Department of Political Science at Texas A&M University. We have refined
our ideas and – we think – improved them on the basis of our teaching and
research experiences with these colleagues. The same can be said for many
scholars elsewhere, who have engaged with us regarding some of the specifics
of this research agenda and its progress. Among this broad group, aside
from those already mentioned, we would like to thank in particular Rhys
Andrews, Stuart Bretschneider, Gene Brewer, Amy Kneedler Donahue,
Sergio Fernández, H. George Frederickson, Carolyn Heinrich, J. Edward
Kellough, H. Brinton Milward, David Peterson, David W. Pitts, Hal
G. Rainey, Bob Stein, and Sren Winter. Finally, we pay tribute to the
immeasurable benefit we have derived from the many forms of support
provided by our families: Mary Gilroy O’Toole, Conor Gilroy O’Toole, Katie
Easton O’Toole, and Diane Jones Meier. Needless to say, the weaknesses and
errors that may remain can be lodged firmly at our own feet.
1 Public management and performance:
an evidence-based perspective

Governments around the globe cope with critical issues and thorny policy
challenges: encouraging economic growth, combating climate change, edu-
cating young people, protecting against disease, building and maintaining
infrastructure, planning urban communities, providing social security, and a
great deal more. Talented policy designers, and the contributions of policy
analysts, can render many of these difficult tasks less daunting. Governments
can also learn from each other’s experiences, so that mistakes do not
necessarily have to be repeated in many places before policy learning can
occur (Rose 1993). To convert sensible policy ideas into reliable and effective
streams of programmatic action, however, much more is needed.
Few policies are self-executing.1 Typically, public programs require the
concerted effort of many people, often coordinated via formal organization,
to achieve their intended results. While some policy interventions can avoid
the need for substantial coordination – monetary policies and other govern-
mental efforts to shape market conditions, for instance, rely for much of
their effectiveness on individuals’2 uncoordinated responses to reconfigured
incentives – the great bulk of policies are delivered into the hands of
intended implementers, whose responsibility it is to make policy come alive
in patterns of goal-oriented behavior. Indeed, the promise of democracy in
advanced nations is fundamentally tied to the ability of representative insti-
tutions to deliver regularly on their policy commitments through such
processes of converting public intention into action.
Governments typically face these implementation challenges with regard
to numerous policy objectives and programmatic initiatives. In the United
States, for example, the national government is committed to thousands of
such policy efforts, and several hundred of these are intergovernmental: they
encourage or require subnational governments to be a part of the action
as well through grant-in-aid programs, intergovernmental mandates, and
other such approaches. Subnational governments also develop their own
policy initiatives. In addition to the fifty states, the United States is home to
1
2 Performance: an evidence-based perspective

89,476 local governments – municipalities, counties, special districts,

townships, and school districts (US Census Bureau 2008). Virtually all these
entities engage in efforts to deliver on policy results, and the same pattern is
followed in country after country.
In addition to plenty of people and considerable resources, accomplishing
public purposes also requires public management. At multiple levels in large
organizations, managers coordinate people and resources toward the accom-
plishment of collective purpose; they also tap the interdependent organiza-
tional environment in support of such purpose and to protect the
organization’s efforts from potential disturbances. This is what is meant by
public management. Some individuals, in other words, have to orchestrate
the myriad individuals, routines, resources, and possibilities into a policy-
responsive mosiac – to make, in Paul Appleby’s (1949) memorable phrase,
a “mesh of things.” The concerted efforts of perhaps thousands of people to
move toward complex public objectives does not spontaneously emerge; it
must be organized and induced, and the task is necessarily ongoing. This
book is devoted to a close examination of what public managers do as they
take on such responsibilities, and we do so from a particular perspective: we
are interested in the link between management, on the one hand, and public
program performance, on the other.
Performance is a highly salient notion in recent years among those around
the world who care about public management. It has acquired even more
importance as government agencies and other organizations have struggled
to deliver results under conditions of austerity. The economic winds that
have buffeted programs in many countries have often resulted in budget cuts
just as public service needs and demands have escalated. The term “per-
formance” is often used imprecisely, thus sometimes generating confusion.
We mean by the concept of “performance” the achievements of public
programs and organizations in terms of the outputs and outcomes that they
produce. Performance can be considered to have numerous dimensions, such
as efficiency (the cost per unit of output or of service delivery), effectiveness
(the extent to which policy objectives are being achieved), equity (how fairly
outputs and outcomes are distributed among key targets or stakeholders),
and public satisfaction (Boyne 2003). Accordingly, performance covers a
broad territory – especially when one considers that improvements on a
given criterion (efficiency, for instance) might result in declines on another
(equity, say). In this book, we pay particular attention to performance in
terms of effectiveness, while also taking into account the resources available,
and in certain analyses we address the theme of equity as well.
3 Performance: an evidence-based perspective

The management task is even more challenging than would seem apparent
at first glance. Public organizations – agencies, departments, bureaus,
authorities, and the like – are at the core of the apparatus for policy
implementation. Indeed, the US federal government’s “bureaucracy”3 is
impressively large – hundreds of organizations, approximately 2.7 million
civilian employees. Even though the national civil service has actually
declined in size during the past half-century,4 its scope and reach have
not, because many policy initiatives involve contributions to policy action
from entities outside the national bureaucracy. Indeed, Paul Light (1999)
estimates that in 1996 a “shadow” federal workforce of approximately
12.7 million people beyond those in the civil service were involved in
carrying out national policy – including government contractors and state
and local employees.
Policy implementation is complicated by the fact that many important
public policies and public programs call for the joint efforts of actors in two
or more – sometimes many more – organizations, frequently in more
than one government, and often in the for-profit and nonprofit sectors.
The expression governance is now often used to denote these broadened
patterns of collective action. “Governance,” as the saying goes, often means
more than just governments (Rhodes 1997; Peters and Savoie 2000;
Kooiman 2003). The need for such multiorganizational action in networked
patterns means that the task of public management requires attention to
such interunit coordination along with a focus on internal organizational
responsibilities.
The fate of public policies in today’s world lies in the hands of public
organizations, which in turn are often intertwined with others in latticed
patterns of governance, which collectively are expected to generate perform-
ance: policy outputs and outcomes. Public management, therefore, means
dealing with organizations, governance, and performance. This book exam-
ines the intersection of these three themes and how managers address them.
Even though our focus includes the performance of public programs, this
volume is not another study of performance measurement or performance
management (for instance, Radin 2006 or Moynihan 2008). That is to say,
we do not explore in detail the issues and controversies involved in measur-
ing performance, nor do we systematically investigate how managers use
performance information to help influence what the people in their organ-
izations do. Instead, in our empirical analyses we rely mostly on perform-
ance information that is regularly collected and typically treated as relevant
and important by managers and others. Where appropriate, we also report
4 Performance: an evidence-based perspective

on performance-related findings by other researchers who have been explor-

ing the relationship between public management and performance. In other
words, this volume explores what public managers do, whether and how
their efforts translate into policy results that are treated as relevant by those
interested in the policy field in question, and the extent to which the impacts
of management on results are modest or sizable.

A perspective on public management and performance

In the chapters that follow, we conclude with confidence that public man-
agement makes a difference to performance, and the impact is far from
trivial. At the same time, it is important to avoid the leap to what some have
called “managerialism”: the “seldom-tested assumption that better manage-
ment will prove an effective solvent for a wide range of economic and social
ills” (Pollitt 1990: 1). We steer an evidence-based middle course here (Meier
and O’Toole 2009b). Much like the similar movements in medicine (Guyatt,
Cairns, Churchill, et al. 1992) and in public policy (Heinrich 2007), our
research, grounded in evidence-based public management, has as its object-
ive to assess the conventional wisdoms – what earlier scholars might have
called proverbs (Simon 1946) – so as to separate the wheat from the chaff
and determine what actually works in practice. Specifically, we consider the
theory and literature of public management, look for ways of tapping the
relevant aspects of what managers do, and estimate the effects of public
management on public program performance, while controlling for other
relevant factors – in particular the difficulty of the policy-relevant tasks and
the resources available for their successful achievement.
Evidence-based public management can proceed in a variety of ways,
including the careful analysis of key case studies. We have opted, however,
to employ the approaches and quantitative techniques of the social sciences.
These include formal and precise theories that generate testable hypotheses
and the statistical analysis of organizations over a period of time.
In this fashion, our theoretical and empirical perspectives avoid two
approaches that are sometimes adopted. As mentioned in the preface, at
one extreme are the population ecologists. This approach, best represented
in the research literature on public management by Herbert Kaufman
(1991), holds that public organizations survive and flourish because they
are lucky, not because they or their managers make sound decisions.
Organizations, in this view, are simply at the mercy of their environments.5
5 Themes from the research literature

Some political scientists offer a modified version of this notion by suggesting

that, particularly in the United States, the broader political system imposes
so many constraints on public managers that they are hamstrung in their
efforts and thus mostly consigned to a rather weak role (Wilson 1989).
At the other extreme are the managerialists, those who appear to attribute
virtually all performance to the purportedly heroic efforts of public man-
agers. Some themes of the so-called new public management6 (NPM) seem
to imply a similar notion, since the oft-mentioned refrain is to “let the
managers manage.” In a more tempered fashion, a literature on public sector
leadership suggests that key managers can have dramatic impacts on per-
formance (Doig and Hargrove 1987). Contrary to such assertions, a great
deal of evidence in fields such as management clearly points to limitations
on what can be accomplished by management, especially in the short run,
although there are also good reasons to expect the actions of managers to be
consequential for performance. For this reason, we are careful in this book
to specify some additional likely influences on policy outputs and outcomes,
including features of the environment in which organizations must try to
accomplish their tasks.
Our analysis indicates that managerial influences on public program
performance are multiple, substantively as well as statistically significant,
and yet accompanied by other influences that need to be taken into consid-
eration in any adequate accounting for results. Demonstrating such patterns
and explaining how managers “do their thing” with such effects on results is
the primary task of this book.
To begin exploring the difference that management makes, we start by
reviewing some of the core themes in the research literature on public
management. This earlier work provides valuable signals about how to
approach the subject of management and public program performance.

Themes from the research literature

In one way or another, researchers have explored the subject of public

management for more than a century. Methods and insights have gradually
evolved and become more sophisticated over time, although certainly there
remain plenty of unverified – and doubtlessly invalid – assertions and
assumptions. In a general sense, we can observe that remarkable progress
has been made in researchers’ efforts to build solid empirical findings about
the world of public management – in the United States, in Europe, and in
6 Performance: an evidence-based perspective

many other parts of the world. If anything, the array of issues addressed in
serious studies of public organizations and their management has expanded
in recent years, with notable investigations of such topics as public service
motivation; red tape, its causes and consequences; government contracting
and privatization; the use of discretion by supervisors and front-line
workers; the differences between public and private management; the chal-
lenges of involving stakeholders in public decision making; the adoption of
new public management and other managerial reforms; the development of
interorganizational collaboration and networks for the delivery of public
programs; innovations in public organizations; institutional isomorphism
(the consequence of mimesis, or organizations’ copying or deriving their
institutional forms from other such organizations) in the public sector;
emergency or crisis management; and diversity management. This list,
furthermore, is merely a partial one.
On the specific theme of public sector performance, moreover, consi-
derable important work has been accomplished by researchers. Even
leaving aside the frequently studied issue of whether so-called “pay for
performance” systems produce useful results (many studies raise serious
questions about the notion), public organizational performance has been
approached from a number of angles. Some, for instance, have explored the
meaning and determinants of the US national government’s recent efforts to
assess program performance by means of so-called “PART” (Program Assess-
ment Rating Tool) scores (see, for example, Dull 2006, Gilmour and Lewis
2006, Moynihan 2006, 2008; more about PART scores shortly). Researchers
have sought to understand whether setting performance targets helps to
improve performance (Boyne and Chen 2007). Others have sought to esti-
mate the influence of such diverse factors as organizational goal ambiguity,
institutional design and reputation, and individual characteristics on per-
formance (respectively, Chun and Rainey 2005, Krause and Douglas 2005,
and Kim 2005). Researchers have tried to determine how features of network
structure can shape performance (Provan and Milward 1995; Schalk,
Torenvlied, and Allen 2010).
Some broad meta-analyses of hundreds of studies related to governance
and performance have attempted to develop some generalizations and themes
from the work of many others (Forbes and Lynn 2005; Hill and Lynn 2005).
In addition, there have been efforts to compile the results of a number of
different studies of performance from different empirical contexts (for
example, see the full set of papers in the October 2005 issue of the Journal
of Public Administration Research and Theory, as well as Boyne et al. 2006).
7 Themes from the research literature

It is clear that, although a number of important issues remain to be sorted

out – such as the kinds of samples and data sets that might be most appropri-
ate for studying management and performance – the systematic exploration
of the performance theme is well under way. (We examine the subject of
performance data later in this chapter.)
Research designs and findings have proliferated, and there is by no means
a consensus about how to understand what makes for effective management
(an important criterion of performance) in contemporary systems of gov-
ernment. Still, the broad study of public management has provided a set of
building blocks that we can use to begin a systematic analysis of the
relationships between management and performance. Some of the key
themes and findings are deserving of brief review, since they provide a
grounding for any serious and sustained research program focused on public
management and performance.
Several of the most notable compilations on and reviews of public man-
agement offer general agreement on certain broad points (for example, see
Moore 1995, Ferlie, Lynn, and Pollitt 2005, and Lynn 2006; by far the best
coverage of the empirical and theoretical literature is that by Rainey 2009).
The core literature of the field assumes or, more often, argues for some
distinctiveness to the management of public programs and public organiza-
tions.7 While the generic management literature often assumes otherwise,
and while some proponents of the new public management advocate
designing public sector settings to more closely approximate market-like
ones, in the main the scholars of public management see sufficient distinct-
iveness that it should be investigated in its own right. In this book we do not
make systematic public–private comparisons (but see Meier and O’Toole
forthcoming (b)), although we do treat public organizations and public
management as sufficiently different – even unique – that they should be
explored on their own, and largely on the basis of insights from the literature
on that subject in particular.8
Second, the literature is in agreement that public management matters –
and, in particular, that it makes a difference in public program performance.
A fair amount of this literature consists of case studies. Indeed, numerous
richly textured case studies and comparative case studies of public manage-
ment all make persuasive arguments that management matters. Many focus
on management in and of individual agencies (for instance, Doig and
Hargrove 1987, Behn 1991, Ban 1995, Riccucci 1995, 2005, and Holzer and
Callahan 1998),9 while others emphasize management in more multiorga-
nizational, networked settings (Gage and Mandell 1990; Provan and
8 Performance: an evidence-based perspective

Milward 1995; Klijn 1996). More recently, several studies have sought to
cover numerous public organizations and/or governments in making and
seeking to validate this claim (examples include Donahue et al. 2004, Boyne
and Walker 2006, Brewer 2006, and Walker and Boyne 2006; see also some of
the literature referenced earlier in this chapter); and some of these assert, but
do not really test, the effects of management on results (for instance, the
products of the Government Performance Project (GPP): see Ingraham,
Joyce, and Donahue 2003, and Ingraham 2007). Given the broad agreement
that public management matters (a conclusion that this book also validates),
the important question guiding the analyses here is this: how effective is
public management at generating performance?
Third, the case study literature, as well as the literature based upon larger-N
studies of many organizations, is mostly in agreement that public man-
agement is not a simple function but, rather, encompasses multiple aspects.
Different scholars offer different lists of functions. Ingraham, Joyce, and
Donahue (2003) treat financial management, human resources management
(HRM), information technology management, and capital management as
the central elements of the managerial function,10 while, in their early work
together, Boyne and Walker (2006; Walker and Boyne 2006) concentrate on
both managerial strategy process and strategy content. A number of case
study authors and others focus in particular on “leadership” by managers
(for example, Doig and Hargrove 1987, Behn 1991, and Terry 2002). Studies
emphasizing the external/network-related aspects of public management
highlight in particular the brokering, framing, exchange-related functions –
above all, collaboration. For instance, Agranoff and McGuire (2003) identify
a number of vertical and horizontal aspects of collaborative activity in which
local government managers engage, including information seeking, adjust-
ment seeking, policy making and strategy making, resource exchange, and
project-based work. Rainey (2009) draws numerous structural and proced-
ural dimensions into public managers’ purview, while also noting the
assumed importance of variables such as organizational environment, tech-
nology and tasks, goals, and culture. So, although there is no clear consensus
on the preeminent or essential functions of public management – no agreed-
upon “POSDCORB”11 for the twenty-first century – a widespread consensus
holds that public management is not a simple, unidimensional activity. An
implication is that, given that management is multifunctional and involves
varied behaviors, determining systematically just what differences all these
functions make in the performance of public programs is an exceedingly
difficult task. In this book we identify a number of distinct aspects and
9 Themes from the research literature

estimate their independent impacts, as well as a more complete, combined

assessment. Even here, though, we have certainly been forced to omit some
aspects of what public managers do that probably carry implications for the
outputs and outcomes of policy initiatives.
Fourth, public management in and of networks has emerged as a central
theme among many researchers in the field. Some caution that attention to
managerial actions in and on the network should not lead scholars to ignore
the important role that internal management plays in shaping performance
(Hill and Lynn 2005; O’Toole and Meier 2009). Impressively large numbers
of public management scholars, however, focus on the externally oriented,
often collaborative, efforts of managers to coproduce multiorganizational
action (a small sampling includes Provan and Milward 1995, Bardach 1998,
McGuire 2006, Agranoff 2007, Bingham and O’Leary 2008, and Rethemeyer
and Hatmaker 2008). This emerging literature, furthermore, suggests a more
significant and extensive set of cross-organizational links than those implied
in the earlier open systems perspective in organization theory. Clearly, any
serious effort to incorporate a range of managerial influences on public
program performance needs to address this growing line of work and,
possibly, growing empirical reality.12
Fifth, a review of the extant research literature reveals a somewhat cloudy
but nonetheless potentially important insight: particularly in case studies of
what public managers do and how that seems to shape outputs and out-
comes, there is often an argument or observation that managers catalyze
action. To be a little more concrete, it is often claimed that such managers do
not simply contribute directly to what happens; they also extract more
positive effort from available resources, including human resources (HR),
and they have the potential to reduce the impacts of disruptive forces. In
other words, the relationship between management and performance is
nonlinear with respect to certain other influences and should be considered
in terms of interactions. If this idea is correct, any effort to model the
influence of management on results must take into account potential
interactive effects.
There are a number of points of convergence in the research literature on
public management, therefore. Nonetheless, it is important to note that
there is no consensus among researchers regarding precisely how to theorize
about the links between management and performance. There are at least
three reasons, and noting these can help to explain why we take the approach
evident in this volume. One reason for the lack of theoretical consensus is
that, as indicated earlier, management is a complicated, multifaceted subject;
10 Performance: an evidence-based perspective

and what managers do undoubtedly has multiple influences on results.

The implication for this book is that we should explore the management–
performance connection by tapping different aspects of management and
estimating their impacts. Second, even for relatively discrete and carefully
researched aspects of management, there is no shortage of different and
competing theoretical formulations. As Hal Rainey (2009) points out,
researchers cannot agree about how to assess organizational effectiveness,
how decisions get made inside organizations, how structure influences
various aspects of organizational behavior, how motivation is shaped in
organizations and why that matters, and what leadership in organizations
actually consists of and how it generates results. There is a plethora of
theoretical ideas on these points; indeed, there is a huge surplus of them.
As a consequence, our approach – as is evident in the next chapter – is to
distill from and build upon the public management literature certain
notions that have some support there, even if that literature is often some-
what ambiguous and even though there is no consensus on precise theoret-
ical formulations.
A third reason for the lack of consensus in the literature is related to the
difficulties with assessing organizational or public program performance –
the issue explored next.

Tapping performance

It is not sufficient to say, as US Supreme Court Justice Potter Stewart did

about pornography, that we know performance when we see it. For reasons
explained more carefully in the next chapter, we think the best route for
making progress in exploring the management–performance link is through
large-N, systematic quantitative studies of many public organizations. An
initial challenge already alluded to is that the concept of performance entails
multiple aspects. We have indicated where our conceptual focus lies. In
addition, and importantly, tapping any aspect of performance requires valid
and reliable measures of performance. Unfortunately, another reason why it
has been difficult to build validated theory, and therefore consensus, is
related to this issue – a matter of method, and, in particular, a point about
measurement. To test for relationships between any aspects of management,
on the one hand, and performance, on the other, requires at least one
acceptable measure of performance.13 Preferably, several performance meas-
ures would be used, since most public programs and virtually all public
11 Tapping performance

organizations operate with multiple objectives. Consider the challenges of

doing so, however. Different public organizations are charged with highly
varied responsibilities: managing housing, determining eligibility for and
delivering welfare or social security payments, running national parks and
forests, cleaning wastewater, overseeing prisons, launching spacecraft, fight-
ing fires, defending the nation from attack. It goes without saying that no
common measure(s) can directly tap outputs or outcomes across such
diverse objectives.
Indirectly, it might seem to be possible to develop a measure at one
remove, by finding a way to calculate something akin to “degree of goal
attainment.” This approach too would be exceedingly difficult to execute,
however. The goals of many public organizations are often ambiguous, the
degree of goal ambiguity varies between agencies, and this variation also has
performance-related consequences (Chun and Rainey 2005, 2006). For fed-
eral agencies, one might be tempted to rely on some sort of common
measure, if such could be found; and at first glance there would seem to
be some candidates. Most of them, nonetheless, have significant limits. This
point can be seen by a brief overview of some candidates for analysis. The
processes initiated in the United States by the Government Performance and
Results Act of 1993 require agencies to develop strategic plans, compile
annual performance plans, and provide performance reports. The results
are not regularly audited, however, there is a great deal of leeway in agencies’
framing of their own objectives, and often – despite the efforts of scholars to
employ these data – there is no reliable way of compiling an overall measure
of performance, not to mention multiple comparable measures. Further-
more, as the US General Accounting Office (now the Government Account-
ability Office) has noted, the agencies typically under-report information on
public programs that span two or more organizations (General Accounting
Office 1999).
During the administration of President George W. Bush, in 2002, the
White House published The President’s Management Agenda, with perform-
ance as one of its themes. Following that signal, the US Office of Manage-
ment and Budget (OMB) began to issue “agency scorecards” for some, but
not all, major agencies. Each scorecard graded an agency via traffic light
signals (green, yellow, red) on how the agency was seen (by OMB and the
experts with whom it consulted) to be doing in achieving five fairly abstract
government-wide initiatives identified in the The President’s Management
Agenda – for instance, improved financial performance, or budget and
performance integration. As of 2008 these continued to be published
12 Performance: an evidence-based perspective

quarterly for a couple of dozen federal organizations. The methods and

measurement were suspect to many scholars (but, for more information, see
www.whitehouse.gov/results/agenda/Standards_for_Success_8_11_2008.pdf
(last accessed October 14, 2008)), the initiatives were not closely related to
the organizations’ programmatic outputs and outcomes, and the number of
organizations involved was rather limited, not to mention overly aggregated
(Department of Homeland Security, Department of Transportation, etc.).14
A further development along these lines during the Bush administration
was the so-called Program Assessment Rating Tool, created by OMB in 2003
and expanded since that time. PART aims to tap program purpose and
design, strategic planning, program management, program results (perform-
ance related to strategic goals), and an overall rating created from all the
other indicators. An advantage would be that PART scores were eventually
developed annually for hundreds of federal programs, not simply for agen-
cies. Unfortunately, there is no publicly available information on just how
the scores are determined by OMB budget examiners, nor is there any
assurance that different examiners are applying the same standards and
criteria to different programs. At best, only certain aspects of PART could
be plausibly associated with outputs or outcomes, and there is some evi-
dence that the scores reflect partisan concerns (Gilmour and Lewis 2006).
Efforts to measure performance have extended beyond those used by the
US national government. States and localities have struggled with these
issues. In addition, the Government Performance Project, a foundation-
supported research project referenced earlier, sought to develop consistent
performance-related information across federal agencies, across the fifty
states, and across a set of major US local governments. The effort resulted
in a great deal of valuable information regarding management and manage-
ment systems, but no common or comparable information on performance
itself was produced by the effort. The GPP, therefore, did not really test the
management–shapes–performance relationship, nor did it generate per-
formance measures that could be used for others’ analysis. Some scholars
have used the measures of management capacity that were generated by the
project as partial explanation for other interesting dependent variables, such
as state policy priorities (Coggburn and Schneider 2003), but not to explain
performance itself.
Individual-level perceptual data on public organizational performance
can be garnered from certain sources, most prominently data gathered by
the US Merit Systems Protection Board in its Merit Principles Survey, which
has been completed by thousands of federal employees in many national
13 Tapping performance

agencies. Other data sets also record similar perceptual data. Survey items
ask for responses on some performance-related matters. Drawbacks, how-
ever, include the points that (1) it is not clear to what extent federal
employees are knowledgeable and impartial with regard to their larger
organization’s overall performance, (2) several of the most relevant survey
items are focused on the individuals’ work units rather than their overall
organization’s outputs or outcomes,15 and (3) models of performance esti-
mated with independent as well as dependent variables drawn from the same
respondents may be prone to common-source bias (Brewer 2006; for a test
of common-source bias through a comparison of subjective and objective
measures of performance, see Meier and O’Toole 2010).
One locus of interesting and useful data on the performance of a substan-
tial number of general-purpose governments is the Comprehensive Perform-
ance Assessment (CPA) of the United Kingdom’s Audit Commission, and
especially the Core Service Performance (CSP) aspect.16 The CPA and CSP
have been applied to local authorities throughout the United Kingdom. The
CSP is determined for each of seven service specialties and is based largely on
archival performance indicators, supplemented by the results of the inspec-
tion and assessment of plans (Andrews et al. 2005). The archival perform-
ance indicators cover six aspects of organizational performance: quantity of
outputs, quality of outputs, efficiency, formal effectiveness, equity, and
consumer satisfaction. The inspection of services draws upon internal
improvement plans, field visits, and other documentation. Statutory plans
are assessed against the criteria of the service’s relevant central government
department. Evaluators external to the local authority conduct all assess-
ments. Each service area is given a performance score by the Audit Commis-
sion from 1 (lowest) to 4 (highest). After calculating the CSP score for each
service area, the Audit Commission derives a score for the whole organiza-
tion by weighting services to reflect their relative importance by budget, then
combining these weights with the performance score for each service area.
The CSP therefore offers a number of measurement advantages, but there
are disadvantages as well. One is that using an overall measure of the
performance of a local authority does not easily allow for exploration of
individual service areas, and thus analysis is largely limited to relatively
underspecified models. A second is that individual services are grouped into
four rough performance categories (omitting much valuable information),
and the adding of these weighted categories creates a false sense of precision
that may not exist in practice. Also, of course, additional data on manage-
ment must be gathered to try to model the management–performance link.
14 Performance: an evidence-based perspective

Some such studies have been completed (for instance, Andrews, Boyne, and
Walker 2006 and Walker and Boyne 2006).17 In any event, the British
government has now modified its approach to assessing the performance
of local authorities. The newer Comprehensive Area Assessment (CAA),
created in 2009, was not comparable to the CSP approach, and explicitly
placed significant discretion into the hands of the auditors. Furthermore, in
May 2010 the newly elected Conservative/Liberal Democrat coalition gov-
ernment scrapped the CAA as well.
Perhaps the best way of making progress is to draw from many public
organizations in one or other policy sector or service area, ideally in settings
in which there is substantial consensus that certain performance measures
are relevant, even highly salient. If such measures are collected systematic-
ally, in time series, and if additional data on control variables are likewise
available, it may be possible to supplement this information with data on
management to explore the difference that management makes. This is the
approach we have taken here (see Chapter 2). Most of our empirical work
draws from one field of public service: education. Results from education
can be supplemented by similar studies in other policy fields, and valid
general findings about public management and performance can be
developed. In fact, a number of such studies have been conducted in other
substantive fields of policy and management, both by us and by other
investigators. The results of many such studies are included in later chapters
of this book as appropriate, along with more detailed analyses of manage-
ment and performance in public education. Although the other study areas
have not produced a volume of research equal to that produced for public
education, they have generated sufficient work to indicate that the theory
and the approach can be generalized.

Our perspective: a parsimonious and formal approach

More than a decade ago we initiated a research program aimed at analyzing

the influence of management on the performance of public organizations
and public programs. We were convinced that, to be successful, such a
program would need to be theoretically based and empirically grounded,
and it would definitely need to consist of far more than a single study. After
examining the literature on public management, performance, and organ-
izational effectiveness, and after noting its many and tantalizingly interesting
assertions and not fully validated conclusions, we decided to ground our
15 Our perspective: a parsimonious and formal approach

own efforts on a parsimonious model of management and performance.18

Doing so would mean leaving aside some of the subtleties in the research
literature in favor of focusing on some especially critical aspects of
management.
On the other hand, it would also require us, to make real progress, to be
fairly specific about the theoretical ideas being explored. This approach
contrasts with an alternative that would stipulate a broad notion of govern-
ance, as has been done by Lynn, Heinrich, and Hill (2001). They indicate
that their “reduced-form model” can frame the logic of governance they
explore:
O ¼ f ðE; C; T; S; M Þ
“where
O ¼ outputs/outcomes (individual-level and/or organizational outputs/
outcomes)
E ¼ environmental factors
C ¼ client characteristics
T ¼ treatments (primary work/core processes/technology)
S ¼ structures
M ¼ managerial roles and actions” (2001: 81).
A limitation here, however, is that these researchers have not fully speci-
fied theoretical, causal linkages and therefore have not derived a set of
precise, testable hypotheses.19 Consequently, little progress has been made
through this approach in validating or further specifying an appropriate
model.
In other words, Lynn, Heinrich, and Hill explicitly “emphasize the dis-
tinction between a logic of governance and the specific theories that
researchers use to investigate questions within this logic.” Their aim is to
suggest a more abstract logic that should be able to fit any number of
theories. The approach “identifies an array of dependent and independent
concepts that investigators encounter in empirical governance research,
whether they analyze those concepts through lenses of political economy,
network analysis, systems models or institutional approaches” (81).
Our approach, by contrast, aims at theoretical specification, hypothesis
testing, and the validation of various forms of managerial influence – as
appropriate. The perspective adopted in this book fits within a theoretical
tradition developed by scholars such as James Thompson (1967), Charles
Perrow (1986), and Herbert Simon (1997). The theoretical approach
sketched in this book works at the organizational level. Another tradition,
16 Performance: an evidence-based perspective

one not chosen here, operates or is driven primarily at and from the
individual level (for instance, Argyris 1957, and subsequent publications;
Golembiewski 1962, 1990, and subsequent publications; Katz and Kahn
1978). Both traditions have their strengths. We are interested in organiza-
tional and programmatic outputs and outcomes – and the effectiveness of
management in shaping these – and judge that the former tradition offers a
more direct and appropriate means for theory building and testing for this
purpose.
In building from the tradition we have chosen, it is not imperative that we
make a choice between socialized choice-based notions of individual deci-
sion making, on the one hand, and economic rationality, on the other (see
the discussion in Lynn, Heinrich, and Hill 2001: 60–71). In tying our
investigations of public management and performance to this earlier line
of work in organization theory, furthermore, we are doing what Steven
Kelman (2008) proposes for researchers in public management.
We also chose to generate such a model on the basis of our reading of the
literature – in other words, inductively. Our approach has been to specify
and formalize into mathematical expression what we understood to be some
of the key theoretical assertions in the extant set of studies. Sometimes these
earlier studies were not precise about the nature of the relationships in the
theoretical arguments, so where we thought it necessary we added specifi-
city. One reason why many apparent theoretical insights in the field remain
neither validated nor refuted, we believe, is that they are sometimes not
formulated with sufficient specificity to be tested and given a chance to be
found wanting. Our approach, therefore, is the opposite: a formal, reductive
model that is constructed inductively but that is in some respects more
precise than the literature from which it springs. As we indicated at the time
of the first publication of the model (O’Toole and Meier 1999), we would
not be surprised to learn that some aspects of the model might need to be
reconfigured on the basis of the evidence, but we would rather be precise so
as to learn of the need for such modifications than persist in imprecision
without progress.
We concluded that the model needed to offer a prominent role for what
might be called internal management – the hands-on activities that man-
agers undertake in seeking to organize and coordinate people and resources
to get things done, as well as to reinforce and possibly enhance the routines
and standard processes through which their organizations generate results.
Equally importantly, we also concluded that the model would need to reflect
the burgeoning emphasis on networks of organizational actors – the
17 Our perspective: a parsimonious and formal approach

interdependent environment of the core organization that may be involved

in supporting, opposing, and/or co-producing outputs and outcomes.
A model reflective of much of the literature would, therefore, have to include
some recognition of the structure or institutional arrangements through
which policy-relevant action takes place, as well as the externally oriented
actions of managers as they seek to deal with their context.
These and other aspects of the model are explained further in Chapter 2.
Here we shift the focus briefly to some of the strengths of the approach we
have taken, as well as some of the limitations or sacrifices inherent in such a
research strategy.
The strengths have largely been alluded to above, and they are consider-
able. Working with a formal model allows for precision and the clear testing
of hypotheses that are deductively drawn from the inductively built specifi-
cation. Our particular approach, as will be evident in the several chapters of
the book that lay out empirical evidence, is not to test the model all at once
in its entirety; we consider that impossible. Rather, we have chosen to focus
on certain relationships and conduct studies that explore those facets in
particular. Much but not all of our empirical evidence has been drawn from
a large-N data set that is still in the process of being built but that itself has
considerable advantages – particularly given the difficulties, outlined earlier,
of tapping common performance measures across many organizations.
Chapter 2 explains the main data set and its advantages more thoroughly.
What is sacrificed by adopting such an approach? First, by using a
parsimonious model we forgo the chance to explore many of the details
and nuances in the case study literature on public management. We ignore,
for instance, the personality traits of managers, as well as organizational
culture, mission valence, the motivation of front-line workers,20 and many
other topics and variables that the literature discusses as being potentially
important. We put the behavior and priorities of public managers front and
center. Those who prize thick description will be disappointed with this
aspect of our parsimonious approach, but they will also be rewarded never-
theless with findings and insights that would be impossible to glean from a
smaller-N approach.21
Second, adopting a large-N approach means, furthermore, that aspects of
the networks/public management theme in the research literature must
largely be set aside at present. The fascinating and provocative research
findings on networks have been generated from case studies and compari-
sons across small numbers of cases. This is so because detailing the many
structural aspects of networks requires intensive investigation and the
18 Performance: an evidence-based perspective

validation of many complex interactions (for instance, Provan and Milward

1995). Intensive analysis forecloses extensive larger-N comparisons. We
explore some structural aspects of networks in this book, and we offer
carefully developed findings on the networking behavior of top managers,
but we do not explore the full details of networks. We view the tradeoff as
worth it (for a detailed methodological argument to this effect, see Meier
and O’Toole 2005), but the gains come at some cost as well. Over time, and
with sustained attention by enough researchers, sufficient data may be able
to be gathered for larger-N, more thoroughgoing network-focused studies to
be possible.
Third, we focus on what can be measured, and thus ignore or overlook the
intangibles that come into play. To be sure, we try to push the boundaries on
what can be measured, such as with our measure of managerial quality
(developed in Chapter 4). At the same time, we miss the unique behaviors
and insights that make some managers stellar. We think the tradeoff –
producing results that most organizations can attain versus identifying the
absolute best practice – is worthwhile. We seek to sketch the broad param-
eters linking management to performance rather than undertake an exhaust-
ive determination that requires excessively complex explanations.

Plan of the book

This volume is about public management and performance. It is not written

from a naive perspective that would suggest that questions of management
can be isolated from themes of politics, and, indeed, such themes are woven
into several parts of our coverage. We focus first and foremost on manage-
ment, however. Our efforts to address how some of the big questions of
politics, and especially of democratic theory, intertwine with public man-
agement and bureaucratic structures resulted in an earlier book, which can
be seen as a companion volume to this one (Meier and O’Toole 2006).
Our plan to unpack the difference that management makes begins in
Chapter 2 with our theory of public management and its formalization in
terms of mathematical modeling. The basic model is explained and distin-
guished from some other approaches that have appeared recently in the
research literature. The chapter also attends to the subject of data. Most of
our findings are drawn from a data set of about a thousand public organiza-
tions. The organizations and the nature of the data are explained in some
19 Plan of the book

detail, and we also demonstrate some of the major advantages of using such
data in a research program such as this one.
Chapters 3 through 7 provide coverage of the management–shapes–
performance evidence. Each chapter concentrates on an aspect of management,
with the major managerial functions carved from the general model for
focused attention. Systematic evidence is adduced in these chapters: detailed
findings, especially from our work in public education, plus summary
empirical evidence from other fields of policy and management and from
numerous other venues.
Chapter 3 addresses the important themes of networks and managerial
networking with other interdependent actors. Among the findings the
chapter reports are: both network structure and networking behavior are
appropriate foci of research for those interested in program performance;
networking patterns of managers exhibit some similarities even when one
compares data across national settings; networking by managers is not a
product of pressure from outside the organization but, rather, reflects
choices and actions by the managers themselves; the networking behavior
of managers contributes positively to performance; managers’ networking
behavior interacts with some resources and constraints in the environment
to boost the effects of networking in nonlinear ways; these positive outputs
and outcomes are not equally distributed among interested parties, a pattern
thus raising an issue of equity; and the direction(s) in which managers
interact outward, and the nature of the interaction partner(s), help to
explain who benefits from managerial networking.
Managers exhibit more than a mere degree of activity or “busyness” when
performing their responsibilities. Some managers are highly skilled at what
they do, while others frequently make mistakes. In Chapter 4 we develop a
measure of managerial quality and estimate the separate impact of quality
on public program outputs and outcomes. We also show that managerial
quality can limit the onset of diminishing returns in the case of networking.
The management of people and operations within the core public organ-
ization charged with delivering results is a huge topic. Indeed, it is the
primary focus of much of the literature on public management. Chapter 5
is devoted to analyses of some aspects of internal management. Among the
results developed there are: personnel and managerial stability contribute
positively to performance; and, more broadly, the management of an organ-
ization’s human capital has a substantial positive influence that can be seen
in organizations as diverse as those dealing with education and those
charged with law enforcement. Chapter 5 also shows that good internal
20 Performance: an evidence-based perspective

management can be used to deal with major events, such as budget cuts, that
can harm the organization and its outputs.
Chapter 6 turns to two often neglected aspects of public management: the
protective dimension and the nonlinearity of management’s impact. Not
only must managers be creative, help their organizations to exploit oppor-
tunities and resources in their setting, and jump-start new operations, they
also can serve a crucial performance function by buffering public programs
from perturbations and protecting them from negative shocks. Even when
shocks enter the organizational system, management capacity plays a key
role in allowing the organization to weather the disruption. In this chapter
we demonstrate how important such functions can be in dealing with such
sources of turbulence as budget cuts and natural disasters. We also explore
the form of the managerial buffering function, and we show how this general
process appears to benefit the organization’s most disadvantaged clientele.
Chapter 7 presents an initial foray into the distinction between behavior
and structural networks. This book focuses on behavioral networks, even
though much of the literature deals with the structural dimension. Our
measure of structural networks is the set of intergovernmental fiscal ties on
which local governments depend, and we explore how variation in such
funding sources between governments might shape results. We then show
how behavioral networking, as well as other managerial aspects, affect
performance within different types of structural networks.
In the final chapter we draw together the results of the research program
to date, examine how all these managerial influences aggregate, sketch
additional questions that remain to be explored, and summarize the differ-
ence that management makes.

NOTES

1. An exception is symbolic policies – those not intended to accomplish anything apart from
signaling approval or disapproval of some person, event, or institution. Symbolic policies
are obviously not intended to be implemented.
2. Here the individuals might be either people or organizations such as business firms. This
point should not imply that such policies can operate without the benefit of public
management to oversee the incentives and adjust them as priorities change or as the
incentives generate unintended consequences.
3. As is often observed, the term “bureaucracy” typically appears as one of opprobrium in
standard conversation – a shorthand for large, heavy-handed, intrusive, red-tape-encrusted,
21 Notes

inefficient government organization. Ever since Max Weber analyzed the concept
nearly a century ago (Gerth and Mills 1958), however, social science has used the
term to reference formal organization with certain distinctive features, including a
hierarchy of superior–subordinate relations, the appointment of experts on the basis
of merit criteria, fixed and limited jurisdictions, decision making on the basis of
rules, and reliance on written records. Note that, by such a designation, bureaucracy
as an organizational form applies to many for-profit and not-for-profit organizations
as well as governmental ones.
4. The budget, on the other hand, has grown considerably.
5. A much less extreme version of such an argument holds that management matters, but
what managers should do is largely driven by features of the organization’s environment
(see, for example, Woodward 1965). As becomes clear later in this book, we believe that
this approach underestimates too much the potential effect of public management.
6. As many commentators have noted, NPM is less a coherent approach than a loosely linked
cluster of ideas and reforms. Many of the specific NPM proposals have to do with bringing
business management practices into public organizations and/or tapping market forces –
for instance, by contracting functions out from public organizations. For varying versions
of and emphases within NPM, see Hood (1991), Ferlie et al. (1996), Pollitt and Bouckaert
(2000), and Barzelay (2001). NPM has shaped practices and ideas in many countries.
7. There is even a literature covering alternative ways that “publicness” might be defined
and measured. We ignore such details in the coverage here, but revisit the subject briefly
in the concluding chapter of the book.
8. It may be that private management has the same impact as public management; we do
not know, and the literature has not demonstrated findings on this point, one way or
another. We study public organizations because we are interested in public programs.
9. For an attempt to review the relevant literature and distill a set of testable propositions
regarding the determinants of effective public organizations, see Rainey and Steinbauer
1999.
10. Overlying these four, three additional aspects are interwoven: leadership influence and
emphasis, integration and alignment, and managing for results (see Ingraham, Joyce, and
Donahue 2003: 16).
11. POSDCORB was Luther Gulick’s (1937) way of summarizing the core managerial func-
tions in the relatively early days of public administration as a self-aware field. The acronym
stands for “planning, organizing, staffing, directing, co-ordinating, reporting, and
budgeting.”
12. On the one hand, many studies indicate that networks and collaboration are on the rise
in today’s world. The emerging theme of governance is reflective of this argument. On the
other hand, diachronic studies of US legislation find little evidence of change across a
nearly thirty-year period (Hall and O’Toole 2000, 2004). The actual trends over time, if
any, await further empirical research.
13. Another requirement, of course, is measures of management. We address the set of issues
related to this measurement challenge in various chapters of this book.
14. Many Cabinet departments have a wide variety of programs that are frequently not tied
closely together. These aggregations reflect a wide variety of historical and political
factors rather than any effort to group programs next to similar ones (see Seidman 1998).
22 Performance: an evidence-based perspective

15. Using work units might mean suboptimization at the work-unit level rather than optimal
performance by the entire organization.
16. The CPA includes a subjective assessment of management style, which can bias the results
if management is an independent variable in the analysis.
17. For management to play a major role, there must be discretion available to managers. In
many UK service areas, such as education, the national government restricts the range of
actions that local managers can take, thus limiting their impact.
18. More recently, one of us has suggested that the model might be used to explore
performance-related relationships across a wide range of political institutions, not merely
in bureaucratic organizations and networks of such implementing organizations (Meier
2007). This possibility is discussed in Chapter 8.
19. Lynn, Heinrich, and Hill do indicate that their reduced-form model implies an “essen-
tially hierarchically” framed perspective on governance (e.g. legislative preferences
driving the formal structures and process of public agencies, the latter then shaping the
de facto organization and management of agencies and programs, this last set of forces
influencing the “core technologies and primary work of public agencies,” etc.; 2001: 32),
although they also note that “there is likely to be endogeneity in these interactions, and
one should not assume that the causal arrow always points downward” (39, note 14). The
hierarchical formulation is based in part on the fact that many available empirical studies
test for hierarchical effects. This finding, while undoubtedly partially valid, unfortunately
does not take into account other under-researched causal paths – for instance, from
front-line bureaucrats to managers to political leaders. Where such reverse or reciprocal
causal paths have been hypothesized, evidence has validated such channels of influence;
see Meier, O’Toole, and Nicholson-Crotty (2004).
20. The exception is when the motivation of front-line workers is reflected in lower turnover –
a variable that we can measure directly.
21. Our approach is not divorced from the real work of practice. We have much experience
interacting with managers in these organizations, as will be evident from our qualitative
discussions in various places. All analysts of organizations and performance need to
understand how organizations operate and how they generate the results that they do.
2 A model of public management
and a source of evidence

How do organizations and governance systems shape performance, and how

do managers influence what happens? In this chapter, we develop a model to
answer these questions and guide our exploration of the real world of
management. We model two distinct levels: the organizational and the
network. While much of the empirical work developed later in this book
focuses on how management makes a difference on the organizational level,
many programs are implemented via complex networks that combine the
efforts of multiple organizations. This structural variance means that it is
important to discuss management in the context of the broader patterns of
governance now evident in many public programs.
Since much of the rest of the book explores these theoretical ideas
empirically, we also discuss data and data requirements. We start with an
assessment of the type of data needed to test our theories. We then discuss
the Texas schools data set, our primary data set, and its relative strengths
and weaknesses. We then note other data sets used in various work in the
public management–performance research agenda, including in some of
our own studies. Several of these other data sets have enabled researchers
to explore in other venues some of the important questions examined in
this book, and we refer to a number of such studies in later chapters.
Finally, we provide a specific discussion of performance measures for
the Texas schools data set as well as the production function used in
subsequent analyses.

Elements of a model

Our model of the relationship between public management and public

program performance begins with a set of core concepts: hierarchy, stability,
network, and management. We introduce a formal model at the organiza-
tional level and then move the modeling ideas to the network level. First,
23
24 A model of public management, and evidence

however, it is necessary to clarify some basic concepts that are both common
in the literature and also central to our approach.
Public organizations are almost always structured as hierarchies. Indeed,
bureaucracy as an organizational form is defined in part in terms of struc-
tures of superior–subordinate relations. In Chapter 1 we observed that
frequently public organizations are also linked to other organizations of
various sorts as they coproduce outputs and outcomes of public programs.
So networks, as defined in the last chapter, often contain hierarchies embed-
ded within them. At a more abstract level, we can think of hierarchy and
network as structural forms at two ends of a continuum. That continuum
would array structural forms – stable sets of relations – in terms of the
degree to which each such structural form is governed by superiors who hold
formal authority to compel action on the part of others.
Although we present hierarchies as a single type at one pole of the
continuum, in practice they vary considerably in structure. They can be
centralized or decentralized; they can vest extensive discretion in employees
or seek to limit discretion with extensive rules; they can permit the lateral
entry of employees or require that all promotions are from within the
organization. These formal arrangements themselves tell only part of the
story, since “informal organization,” which develops in every functioning
organization, is also important. For present purposes, however, we simplify
by treating hierarchy as a stabilizing arrangement. In other words, at the
extreme a “pure” hierarchy would be highly stable, even rigid, and would
resist forces that might otherwise perturb or modify it.
Formal hierarchy is not the only stabilizing force in bureaucratic
organizations. Among the other stabilizing aspects of formal organization,
aside from structural stability, are mission stability (constancy over time
in the stipulated goal(s) of the organization), procedural stability (con-
stancy in the routines, rules, and standard operating procedures of the
organization), production or technology stability (constancy in the tools
used by the organization to get things done), and personnel stability
(retention of the same individuals in the organization; see O’Toole and
Meier 2003b). Indeed, some of these stabilizing forces are likely to
reinforce each other; hierarchy, or structural stability, for example, can
contribute to procedural stability.1 These aspects of stability are discussed
further in Chapter 5.
In short, we expect hierarchies to be relatively stable systems, and we also
note that other forces also contribute to stable organizational systems. These
features of public organizations will prove to be useful in our efforts to
25 Elements of a model

model management and performance, and some of them are also examined
in the empirical parts of this book.
If “pure” hierarchy, or completely stable organization, is at one pole of the
continuum mentioned earlier, pure networks can be considered to sit at the
other end.2 There is considerable evidence demonstrating that many public
programs are not executed within or by a single hierarchical agency but are
spread across parts of two or more organizations. These can be different
organizations within a common government, but in many cases they are
parts of multiple governments, and sometimes they also include businesses
and/or nonprofit organizations. Such patterns are what we mean by net-
works. The nodes of networks can be occupied by individuals, organizations
(including hierarchies), or parts of organizations.
As with hierarchies, networked structures can vary greatly in structural
terms, from simple dyads to breathtakingly complex arrays of several dozen
organizations, and with great variation as well on a number of structural
dimensions that are typically used to characterize networked patterns: net-
work centrality, multiplexity, and so forth (O’Toole 1997). Networked action
is typical in the United States (see, for instance, Bardach 1998 and Bingham
and O’Leary 2008) as well as in other settings, from the United Kingdom
(Huxham 2000; Stoker 1999) to Sweden (Lundin 2007) and the Netherlands
(Schalk, Torenvlied, and Allen 2010), and even Thailand (Krueathep,
Riccucci, and Suwanmala 2010).
Despite the absence (or low level) of formal authority possessed by
“leaders” in a network, such patterns may over time acquire some consider-
able degree of stability. In fact, some types of networks have become known
for their relatively closed and impenetrable features – for example, the “iron
triangles” of administrative agency, legislative committee, and interest
groups sometimes seen operating for extended periods in certain US policy
fields (Freeman 1965). Our particular interest is in networks of actors
involved in implementing public programs and delivering public services.
Networks for implementation can also acquire stability over time, but those
that are not well established, but are in formation or flux due to the
establishment of a new program or significant modification in an existing
one, can be quite fluid. Indeed, in pluralist governance settings such as in the
United States, we expect networks to be structurally more open and shifting
than in implementation settings in more corporatist contexts where there is
broad agreement on processes and procedures.
Networks are particularly interesting to public management scholars
because public managers have the responsibility for trying to weave
26 A model of public management, and evidence

the interdependent parts of an implementation apparatus together into a

functioning policy delivery system. Networks such as these, quite common
for public programs, represent a considerable degree of structural fluidity
and therefore contain considerable uncertainty regarding relations, commit-
ments, understandings, power, and information (Frederickson 1999). Gov-
ernments often tap networks to deliver policy results but face a common
dilemma. The enhanced capacity of networks for action across multiple
units increases the odds of generating outputs and outcomes, but, at
the same time, adding actors in networked arrays introduces substantial
challenges of coordination and associated uncertainty.
In short, structural variation – between rigid hierarchies at one extreme
and fluid, emergent networks at the other – can be an important aspect of
the institutional setting of public programs, and it can influence the chal-
lenges that managers face as they seek to contribute to program perform-
ance. Although most programs operate somewhere between the poles of this
continuum, we want to include this range of variation in our approach to
modeling to make sure that our theory is as general as possible.3
Public management involves the coordination of people and other
resources toward the accomplishment of public purpose. The particular
activities that are encompassed by the notion of management here con-
stitute an exceedingly long list. Public managers undertake traditional
POSDCORB-like functions, as well as creative and subtle efforts. They
are involved in managing budgets, for instance, and also in devising
creative ways of gaining access to resources. Their responsibilities involve
planning, and also risk evaluation and mitigation. They manage people,
while also trying to create and motivate teams that work together across
the organization – and sometimes across different organizations. Public
management, in other words, includes myriad specific challenges and
activities.
It is impossible for us to specify and estimate the effect of all these detailed
elements of public management. So, while we build our model inductively
from the literature, we must perforce generalize and move to a somewhat
more abstract level in sketching managerial functions. One way to do so is
to distinguish managerial efforts to manage within the organization from
other activities directed externally – that is, toward the environment of the
organization. These two aspects of management are not neatly partitioned
from each other, and we should avoid any implication that there is some sort
of simple, zero-sum tradeoff between the two. For reasons that will become
apparent shortly, nonetheless, it can be helpful when formalizing the role
27 Elements of a model

of management in shaping performance to distinguish and recognize both

broad components of what managers do.
In this sense, then, our discussion of structure is related to our conceptu-
alization of management. It is common to assume that “management”
operates in and through a hierarchical structure, and it often does. Given
the prominence of networks in the practical world, however, we do not want
to assume that programs operate as hierarchies in advance of examining
them. Accordingly, we encompass as part of “management” the efforts by
actors to concert patterns of behavior across organizations and not just
within them. A part of management might involve persuasion, signaling,
and diplomacy with regard to others rather than simply issuing communi-
cations and directives along formal bureaucratic channels. This point, along
with the observation that even the management of individual organizations
operates in an open-system context, in which the core organization is
inevitably and regularly interdependent with its environment, suggests that
public management encompasses these distinguishable functions or lines
of activity.
An additional point about public management, broadly construed, needs
to be mentioned. Multiple individuals operating in the setting in which a
public program is being executed may have responsibility for, or may take
upon themselves part of the task of, management. The job is typically not
consigned to a single position or individual, and perhaps not even to actors
operating within a single organization. Managers populate multiple levels
within large organizations, and, for networks of organizations, managers
operate across the network as well. Nor is it a foregone conclusion that
the managers are all operating in coordination with the other managers.
Some managers may try to do so and fail, and other managers may also use
their managerial effort and ability to move action in directions opposed by
others – including other managers – in the setting. These possibilities make
for exceedingly complex and possibly confusing patterns of behavior.
We address these complications in two ways. The reality that there may
well be multiple managers and managerial influences, possibly pressing in
different directions, is treated by us here via a simplification: theoretically,
we model management as a vector sum of the full set of managerial efforts of
various types. In other words, we include multiple managerial functions, but
we treat different managerial vectors for each function only in terms of the
vector sum for that function.4 In most of the empirical work presented in
later chapters, we draw our data from the top manager in each public
organization. In terms of there being different managerial efforts at different
28 A model of public management, and evidence

levels of a complex institutional setting, we both simplify and complicate in

our work. With regard to the former, we largely focus our empirical work on
the efforts of top management in the core production organization and thus
mostly leave aside the additional managerial forces at lower echelons of the
organization. With regard to the latter, we undertake our modeling in two
stages, each reflecting a different level in a governance system (organizations
and networks). We thereby seek to capture the complexity of management
in networks, even if we also simplify in other ways for the purposes of
empirical analysis.

Modeling public management and performance

Our approach to modeling and exploring the relationship between public

management and program performance is to draw some basic ideas from the
extensive case study literature and to formalize some of it in terms of a
mathematical model. In this section we proceed in this fashion, step by step,
to build the model. We use the basic concepts covered in the preceding
section of the chapter, as well as some of the points of scholarly consensus
introduced in Chapter 1. Here we model management’s relationship to
performance from the perspective of a core production unit, and in the
section following we continue by modeling at the network level. Our
approach is to begin with some basic features of an organizational or
program system and then gradually introduce additional elements. Once
we have some of the basic features in place, we add aspects of management
to the model.

Inertial systems and stability

Organizations and programs, as has often been noted, are inertial systems.
What they do and what they produce today is typically very much like what
they did and produced yesterday. The pattern holds for both empirical and
normative reasons. Empirically, it is difficult to induce significant change
overnight when the routines and operations of hundreds or thousands of
individuals would need to be adjusted and coordinated anew, and perhaps in
a different direction. Normatively, one of the major advantages of bureau-
cracy as an organizational form is consistency and relative stability (Gerth
and Mills 1958); organizations, as a result, are designed to be inertial. An
29 Inertial systems and stability

inertial system means that current outputs5 can be expected to be strongly

influenced by past outputs. If one defines outputs at time t as Ot, a very
simple model of organizational or program output is

Ot ¼ b0 Ot1 þ e ð2:1Þ
where current performance is the result of past performance at time t–1,
discounted by a rate of stability, b0, and a set of shocks to the system, ε. In
this general modeling effort, we ignore the nature of the relevant outputs,
how they would be measured, and whether multiple dimensions should be
considered.6 In mathematical terms, such an inertial system is called an
autoregressive system. An autoregressive system is not the equivalent of a
static system. A static system does not change from one time period to the
next; an autoregressive system can build in change over time either internally
or in response to environmental change.7 The rate of stability of this system
is constrained to a value between zero and one. As b approaches one, the
system becomes highly stable. As the value approaches zero, the system
moves toward entropy.8
Shocks to the system, ε, can originate from a variety of forces in the
environment. Some of them are intentionally generated by other actors –
for instance, decisions from the courts that alter or constrain program
activity, actions by legislatures or executives that change priorities or alter
program funding, or antagonistic moves by those who oppose the program
or the organization that operates it. Some shocks may emanate from other
influences, such as changes in the economic or social environment. The
examples mentioned thus far are exogenous, but some may have their
source from within the system itself, such as planned organizational change,
or organization development. As will become clear shortly, we distinguish
some of the exogenous parts of ε and incorporate them into the modeling
process.
In rigid hierarchies, as indicated earlier, we expect systems to be highly
autoregressive, whereas with fluid and emerging networks there is much less
inertia. If we consider both structure and other stabilizing features of such
systems (standard operating procedures, civil service rules, and the like), we
can introduce another term, S, as an expression for the set of stabilizing
features. If we had a good measure of the set of stabilizing features, and if we
normalize it to approach 1.0 at the highest level of stability and to approach
0 at the extremely fluid and flexible end of the continuum, we could say

b0 ¼ f ðSÞ ð2:2Þ
30 A model of public management, and evidence

or that the rate of stability, b0, can be partitioned into the structural and other
related stabilizing features, S, along with other inertial elements, now b1.
The general equation would be
Ot ¼ b1 SOt1 þ e ð2:3Þ
which indicates that an increase in organizational structure and/or other
stabilizing elements results in a more inertial system.

Shocks

A major difference between fluid networks and highly stable structures such
as bureaucratic hierarchies lies in how they are affected by external shocks
from the environment. Later in this book we consider in some detail how
shocks affect organizational performance and what can be done to mitigate
any potentially negative effects. For now, we consider the issue in general
and abstract terms. Stable hierarchies generally tend to buffer or protect the
organization fairly effectively. Shocks that do penetrate a system’s protec-
tions, nonetheless, have different impacts on stable organizations from the
impacts they have on fluid networks. Although shocks are less likely to pass
through a hierarchy’s buffering apparatus, when they do they can have a very
significant impact. To see why, we return to the initial autoregressive rela-
tionship in Equation (2.1) from above for an approximation to the pattern
in a highly stable hierarchical system (where b0 ! 1):
Ot ¼ b0 Ot1 þ e ð2:4Þ
If we partition the ε into some shock Xt that penetrates the system’s buffers
with an initial impact of b2 and a random component, εt, this yields
Ot ¼ b0 Ot1 þ b2 Xt þ et ð2:5Þ
Note that, in this case, a one-unit change in Xt produces a b2 change in
Ot, all other things being equal. This effect is the impact of Xt on O for time t
only, however. Because Xt has increased the value of Ot, then in time tþ1 this
larger value of Ot will also influence the size of Otþ1. Because Ot is b2 larger
as the result of Xt, Otþ1 will be b0b2 larger as the result of the impact of Xt in
the preceding year. Such impacts continue to reverberate through the system
in future years, gradually becoming smaller (forming what is known as a
geometrically distributed lag; see Hamilton 1994) but still cumulating into a
relatively large impact.
31 Buffering

The overall impact, I, of a one-unit change in X can be determined by the

following formula, where the terms are defined as above:
I ¼ b2 =ð1  b0 Þ ð2:6Þ
A relatively small shock that gets through the system’s buffering apparatus,
consequently, can have a major, long-term influence on the system,
depending on the size of the coefficient of stability. As an illustration,
suppose the initial-year impact, b2, had a value of 1. If the coefficient of
stability is 0.99 (indicating a highly stable system), then the total impact is
100, or 100 ¼ 1/(1  0.99).9 If the coefficient of stability is only 0.7, the total
impact of X in this case falls all the way to 3.33.10 Two important points
merit reemphasis. First, relatively small changes in a system can have major,
long-run implications simply because the program structures are inertial
systems. Second, shocks that penetrate to the organization have a much
larger long-run impact in highly inertial systems than they do in less inertial
systems, because the impact of the shock continues to influence outcomes
well into the future.

Buffering

As noted above, organizations establish units or processes to buffer shocks

from the environment. Several forms of buffering can be identified (O’Toole
and Meier 2003a). Abstractly speaking, these include buffers structured as a
barricade or “wall” of some height that stops all external shocks smaller than
a given size from penetrating the organization and its operations. Alternatively,
some buffers are designed more like filters: certain issues or stakeholders –
ones more central to the organization’s goals or survival – are screened in,
while others are screened out. Another form that buffers might take is as a
dampener: external perturbations have impact, but the magnitude is
reduced by some amount (for empirical evidence on this last variant, see
Meier and O’Toole 2008). We explore this variant more carefully in
Chapter 6 of this book. Regardless of the type of buffer, in more fluid
networks buffering is more difficult to accomplish simply because the
nature of networks creates additional interdependences that cannot be isol-
ated from the technical core of the system. Because organization A is interde-
pendent with (read “linked to”) organization B, any shocks that penetrate
organization A’s buffers are likely to influence organization B regardless of the
strength of B’s buffers.
32 A model of public management, and evidence

Given the different functional forms that buffers may take, for modeling
purposes we would need to take into account the form of the buffer to
represent it mathematically. We do so here for one of the types mentioned
above: that of buffer as dampener. We can conceive of the stabilizing features
of a system (here subscripted with an “e” to indicate that the feature is
designed to interface with the environment rather than generate internal
stability – for instance, a legislative affairs or public affairs office), including
the system (such as organizational) structure, as reducing the impact of
shocks via a discounting term in the model, thus:
Ot ¼ b1 SOt1 þ b2 Xt ð1=Se Þ þ et ð2:7Þ
In this way, an increase in stabilization, Se, acts directly on the exogenous
shock to limit its impact on the system. Any shock that penetrates the
buffers of the system can still have a substantial, long-run impact on the
organization, however, if the organization is highly inertial. For a fluid
network, in contrast, buffering is relatively weak, so shocks easily reach the
system. The impact of such shocks over time is far less, however – simply
because the networks are more loosely coupled.
Equation (2.7) indicates how one might model a buffer that operates as a
dampener, reducing all environmental impacts by a given amount. Other
types of buffers would take different mathematical forms. A buffer that
operated as a barricade or a levee, for example, would be operationalized
as a more additive (subtractive in practice) process, as follows:
Ot ¼ b1 SOt1 þ b2 ðXt  Se Þ þ et ð2:8Þ
The important theoretical point about buffers is that they can be designed in
very different ways; models of organizations need to reflect the different
buffering processes with appropriate mathematical representations.

Introducing management into the picture

Hierarchies and networks are human systems for executing policy; as such,
they are not merely inertial structures and buffers but managed entities as well.
Obviously, therefore, we need to represent some of the core features of public
management in our mathematical representation – but how, exactly? We know
that all signs point to management often contributing positively to program
performance. If this contribution were to take the form of another standard
input in the production process, we could include management – “M” – as
33 Introducing management into the picture

an additive term in the model. In this fashion, if we were to add

management to Equation (2.5) from earlier, we would represent its contri-
bution as follows:
Ot ¼ b0 Ot1 þ b2 Xt þ b3 M þ et ð2:9Þ
If Xt represents a vector of all other factors that affect the system (and therefore
are viewed as shocks, whether positive or negative), such as resources, con-
straints, external demands, and so forth, then the test for whether manage-
ment matters in a program structure would be whether the coefficient for
management, b3, is significantly greater than zero.
Such a simple linear impact for management is inconsistent with our
reading of the literature, however. Much of the rich literature on public
management and performance indicates that management interacts with
other features of the system to shape results. Trying to represent this idea
in the model means considering various nonlinear mathematical forms.
Several options of this sort are possible (see O’Toole and Meier 1999 for
some possibilities), but it is also important to consider the program struc-
ture, along with other stabilizing features of the system. Indeed, one crucial
task of management is to maintain some of these stabilizing features: to
frame the goals, set the incentives, buttress the structure, and negotiate the
contributions from members and from those with whom the system inter-
acts (Barnard 1938; Simon 1997). This system maintenance function of
management, we think, can best be modeled as in the following representa-
tion, where management supplements the set of stabilizing elements in the
system in the inertial portion of the model:
Ot ¼ b1 ðS þ MÞOt1 þ b2 Xt þ et ð2:10Þ
In this equation, as stability increases, the role of management becomes less
necessary because the other stabilizing system features generate a relatively
inertial system.11 As stable structure and other such features decline, how-
ever, this system tends toward entropy unless management increases its
impact on the maintenance of steady production.
Many of the standard accounts of public management emphasize this
managerial function, and it would be possible to unpack many of the ways
that management supports the coordination and maintenance of produc-
tion efforts: managing human resources, planning and organizing and
assessing risks, allocating financial resources among production tasks so as
to support the generation of outputs and outcomes, and much more. Later
in this book we devote some attention to specific managerial subfunctions
34 A model of public management, and evidence

such as these, as well as to internal management overall. For the moment,

however, we simply note that one broad managerial function is to support
and maintain performance.
Of course, maintenance is only one function of management.12 We can
call this first managerial function M1. An equally important function of
management is to guide how the system interacts with its environment – in
modeling terms, how it deals with shocks to the system. We designate this
latter aspect of management M2. As indicated earlier in the chapter, the two
are not fully partitioned from each other, and a given manager may allocate
considerable time and effort to both (for an analysis of how shocks to the
system can generate internal managerial adjustments, see O’Toole and Meier
2010). We use different subscripts nonetheless to allow for the possibility
that these two functions can vary independently of each other yet still have
something in common that we would consider management.
M2 can be modeled, but only if the management strategy of the system is
known relative to the interdependent environment. Management could
adopt a strategy of buffering, or protecting, against environmental influ-
ences; or management could actively seek to exploit or tap the environment
for the benefit of the program’s performance. If management adopts a
buffering approach, we can model this choice as follows, with management
externally interacting with the set of stabilizing influences in the buffering
process:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 Xt ð1=Se M2 Þ þ et ð2:11Þ
In this equation, management dampens the impact of environmental shocks
and works with stabilizing features such as structure in this process.13
Management that seeks to exploit the environment rather than buffer from
it will attempt to tap or magnify some of these influences from outside –
financial resources, supportive stakeholders, and suchlike – so that they have a
performance-enhancing impact on the system. In this case, we model the
impact of management as leveraging at least some of the “X” term, as in the
following equation:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 Xt ðM2 =Se Þ þ et ð2:12Þ
To put it simply, management moves from the denominator to the numer-
ator of the second term of the equation, and in the process increases the
impact of some environmental force rather than diminishing it.
Of course, a more nuanced and realistic notion would be that manage-
ment does not simply adopt a buffering or an exploiting strategy, but, rather,
35 Introducing management into the picture

seeks to do both: buffering some influences while tapping and using

others.14 To represent both exploiting and buffering in the same model, we
combine Equations (2.11) and (2.12) and partition – and relabel – M2 as
managerial efforts to exploit, M3, and managerial efforts to buffer, M4, as
follows:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 Xt ðM3 =Se M4 Þ þ et ð2:13Þ
Rearranging the terms of this equation, we get
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 Xt =Se ðM3 =M4 Þ þ et ð2:14Þ
Equation (2.14) represents our general model of public management. The
ratio of M3 to M4 in the second, or environmental, portion of the model is a
characterization of how risk-seeking (or risk-averse) the management of the
system is. As the effort devoted to tapping environmental forces increases,
this ratio increases. As the management of the system devotes more effort to
buffering, the system becomes more risk-averse and the size of this ratio
decreases. In theory this risk ratio can be viewed as management-imposed
risk versus the normal risks associated with an uncertain environment. The
normal risks of environmental uncertainty can be tapped via an examination
of how the “X” factor, the environment, affects the degree of variation in
outcomes, O. This implies that greater risk will be associated with a larger
standard deviation in O. Within this view, management’s attempts to
increase or decrease risk by manipulating the ratio of exploiting to buffering
should also affect the standard deviation of O.
Finally, we note that we have distinguished stabilizing influences from
managerial influences in this modeling effort. Though sensible and defens-
ible, this approach omits an additional subtlety: that, over more extended
periods, stabilizing forces influence (constrain) the actions of management,
and the actions of managers can also shape the system’s structure and other
such stabilizing forces. Therefore,
St ! Mt ! Stþ1
The theory and the model, moreover, treat stabilization and some parts of
management as substitutes for each other. The interrelationships between
management and stabilization can be empirically examined with a variety of
time-series techniques, such as vector autoregression or an instrumental
variables approach within a normal time-series model. While we note these
hypothesized relationships here, however, and while we explore M and some
aspects of S in this book, we do not pursue further the interrelationship of
36 A model of public management, and evidence

S and M over extended periods. This line of research must be left to future
work, but it is a promising area in which to merge the study of public
management with the study of organization theory (Kelman 2008).
We have developed a model that includes a number of managerial func-
tions and a set of contingent relationships. Modeling management and
performance at this level has incorporated some assumptions and simplifi-
cations, but a big advantage is that we have specified precisely a set of
empirical relationships that might or might not hold in practice. Although
we begin to test many of them later in this volume, at this point the model
can be treated as a set of hypotheses. It is quite possible that this theoretical
effort will eventually be shown to be in error on certain points (one such
possibility is introduced later in the book; see Chapter 6), but we care less
about being correct in the details than about catalyzing work along these
lines. Progress can be expected only through precise and ultimately falsifi-
able predictions about managing public programs.
Our approach differs from others that have been taken in the research
literature on public management. First and most obviously, this approach
varies markedly from the most common approach in the field of public
management. Although the model has been built by relying on the extensive
case study literature, and although there is nothing about the model that
could not be explored via additional case studies, we have formalized several
of the assertions and intuitions of that earlier literature for the purposes of
more precise, large-N statistical tests. We think the advantages of this approach
in terms of reliability and external validity make the effort potentially valuable.
It should also be clear that, given the complexity of the model and the number
of hypothesized relationships, it is not a simple matter of generating one study
that can see if all the theoretical ideas set forth here are valid or invalid. The
model, in short, initiates and catalyzes a research program that needs to
encompass multiple kinds of empirical investigations. As we mentioned at
the conclusion of our initial articulation of the model, “[T]his perspective
suggests the initiation of a research agenda rather than the sketch of a one-shot
research design” (O’Toole and Meier 1999: 524, emphasis in original). Indeed,
this book synthesizes a number of the investigations that have been part of that
research agenda as it has developed over several years.
Second, the theoretical ideas sketched here are somewhat different from
those offered by certain other researchers. Even among those analysts who
frame public management in terms of multiple functions, there can be
different ways of slicing such functions aside from the way we have done
here. Mark Moore (1995), for example, carves managerial effort into three
37 Introducing management into the picture

different portions: managing upward, downward, and outward. The model

developed in this book largely does so in terms of internal management,
external exploitation, and external buffering (sometimes we combine the
latter two into a simpler measure of managing outward). We also differ from
those who have argued that managing networks is fundamentally different
from managing hierarchies (see, for instance, Provan and Milward 1995,
Mandell 2001, and Agranoff and McGuire 2003). Our model can be adapted
to the network level (see below); it also stresses managing in an organiza-
tion’s networked setting by its extensive focus on how managers deal with
manifestations arising outside the organization.
Third, our approach is different from – and, we would argue, more
specific than – other recent efforts to suggest the components of a model
of governance. The most well-known such governance model is that offered
by Lynn, Heinrich, and Hill (2001), which was discussed in Chapter 1. That
conceptualization is an interesting and potentially useful way of thinking
about governance, but it is not a theory of governance. The latter would
make relatively precise predictions about variables and their relationships.
The former can integrate extant research, identify over- and underexamined
foci of investigation (see, for instance, Hill and Lynn 2005), and thereby
provide a checklist for future work to consider. Other than clarifying a
listing of relevant variables that should be considered in theory building, it
does not generate specific research hypotheses and is largely agnostic in
choosing between theoretical perspectives.
In short, we see a logic of governance as useful but insufficient; theory
building is necessary. Our particular perspective in this regard explicitly
incorporates management into the process, provides precise predictions
about how the variables relate to each other, allows and specifies certain
relationships to be nonlinear, frames a set of research questions so that
conceptual and measurement issues can be identified and addressed, and
mimics in the abstract how organizations and their managers operate in
practice (contingent decision making that takes place in an autoregressive
system open to environmental influences). These advantages, we would
argue, are not inconsiderable; but the ultimate test is empirical.
We readily admit that we have a reductive model, one that emphasizes
parsimony. Such parsimony is always open to the charge that we have omitted
a key variable or concept. At the same time, parsimony is an advantage, because
the limited number of concepts permits us to test some highly complex
relationships (as illustrated by the nonlinear and interactive relationships).
We think the gains from parsimony exceed the costs.
38 A model of public management, and evidence

Much of the rest of this volume consists of empirical analyses, and

considerable support for several parts of the model has accumulated. It is
important to keep the overall emphasis on the model as a set of partially
tested hypotheses, though, rather than as received wisdom or fully validated
knowledge about public management in all places and for all times.
Before turning to empirical findings, however, two additional tasks ought
to be addressed. First, we should expand the modeling effort to present
some tentative ideas about managing in more complex settings – networked
settings. We do this to illustrate our belief that, although managing a
network is more difficult than managing a hierarchy, the general processes
are similar and can be managed and modeled in similar terms. Second,
we should place the ensuing empirical work in context by discussing the
data requirements for empirical studies of this model, or other models of
this sort.

Modeling management and performance in networks

Earlier in this chapter we explained that management’s institutional setting

is expected to be important; such settings range from rather stable structures
such as hierarchies to highly fluid networks of organizations (which them-
selves may contain hierarchies within them). Thus far, the model we have
developed incorporates this structural variation as part of the stabilizing
forces through the “S” term. If we try to be more precise and also complete,
nonetheless, we have to think in terms of multi-level systems, with manage-
ment operating at different levels, with different foci, and to different effects.
Although the modeling ideas that result from grappling with this further
feature of the real world generate enormous complexity, as will be seen
shortly, and although the remainder of this volume works from the more
simplified model developed in the preceding section, we think it useful here
to suggest more fully how modeling might proceed in multi-level systems
involving networks of actors. Doing so is helpful because it alerts us to some
of the issues that can arise for managers operating in complex institutional
settings, and taking this step now also suggests certain items yet to be dealt
with by researchers in the field of public management.
Moving the theoretical ideas developed here up to the network level, with
clusters of organizations and their management partially linked in pursuit of
public objectives, involves increasing the model’s complexity by an order of
magnitude (for an early development of these ideas, see Meier and O’Toole
39 Modeling management and performance in networks

2004a). Although the core concepts and the basic ideas remain the same, the
number of possible relationships and the demands that these place on data
and models increase significantly. This section merely indicates the direction
that such modeling and estimation may need to take; a full elaboration
would take substantial additional space, and the data requirements for
systematic testing of these ideas surpass significantly the kinds of data now
available. For that reason, we restrict this presentation to the outlines of how
such a modeling effort should develop.
We begin this sketch by reintroducing our basic model, but with a sub-
script, h, to indicate that the concepts are measured for a formal hierarchy –
that is, at the organizational level. This modification in symbolization, but not
conceptualization, yields the following:
Oth ¼ b1h ðSh þ M1h ÞOðt1Þh þ b2h ðXth =Seh ÞðM3h =M4h Þ þ eth ð2:15Þ

For the sake of simplicity, one can define the internal (that is, first-term)
nonoutput portion as Y and the external (environmental) term as Z, yielding
the following simple equation for a hierarchy:
Oth ¼ b1h ðYh ÞOðt1Þh þ b2h ðZh Þ þ eth ; ð2:16Þ

where Yh ¼ Sh þ M1h and Zh ¼ (Xth/Seh)(M3h/M4h).

A network established or used to implement a program would also have
similar internal and external terms. We theorize in terms of the same kind of
functional form, now subscripted with an “n” to reference the network level:
Otn ¼ b1n ðYn ÞOðt1Þn þ b2n ðZn Þ þ etn ð2:17Þ

Even networks should be expected to be somewhat inertial, though less so

than individual formal organizations. The internal term of the model
becomes much more complex, however, because it must now include both
the internal management terms for the network (Sn þ M1n), and also the
same terms for the hierarchies (Sh þ M1h) that compose the nodes of the
network:
Yn ¼ ½ðSn þ M1n Þ
ðSh þ M1h Þ ; ð2:18Þ
with two new symbols (
, ) introduced, to be explained shortly. Similarly,
the environmental term must now include both the environmental factors
for the network (Xtn/Sen)(M3n/M4n) and the environmental factors for the
hierarchies (Xth/Seh)(M3h/M4h) that comprise the nodes of the network:
Zn ¼ ½ðXtn =Sen ÞðM3n =M4n Þ
½ðXth =Seh ÞðM3h =M4h Þ ð2:19Þ
40 A model of public management, and evidence

When combined, the overall formal presentation of network management

becomes

Ot ¼ b1n ½ðSn þ M1n Þ

 ðSh þ M1h ÞOðt1Þn
ð2:20Þ
þ b2n ½ðXtn =Sen ÞðM3n =M4n Þ
½ðXth =Seh ÞðM3h =M4h Þ þ et

The two new symbols require explanation. The “” term is used to indicate
that the internal management terms of the individual organizations compris-
ing a network are aggregated in some manner, as yet undefined; the external
management terms of these units are also aggregated in some manner. The
form of aggregation, we theorize, depends on the type of interdependence
among the units comprising the network. Thompson (1967) has sketched a
simple typology of such patterns for organizations, and his notions can be
applied across units as well (see, for instance, O’Toole and Montjoy 1984).
Whether networks of organizations are pooled, sequential, or reciprocal
carries implications for how one models management.
Pooled environments around a core or focal organization and its man-
agement, whereby multiple external organizational actors contribute to
impacts on the targets of public policy but do not deal directly with each
other during their own efforts, are by definition less interdependent than
other patterns; thus resources (or constraints) from them can likely be
summed. Maintaining a supply of a particularly strategic resource from
one part of a pooled environment does not require managing relations with
the remainder of the organizational actors. Similarly, controlling the impact
of constraints imposed from a particular direction does not necessarily entail
orchestrating coalitions of actors across multiple units.
Sequentially structured environments – arrays in which an output of one unit
serves as an input for the next, and so on – suggest certain other critical
management issues: eliminating any blockages in the flows of production
between units in the environment, and taking advantage of how resources may
be aggregated. Sequential environments, or networks of organizations subject to
sequential interdependence, should be aggregated in a multiplicative manner; a
probability of failure (or success) in one relationship affects the probability of
failure (or success) of all the subsequent units in the sequential chain.
Reciprocal environments, in which the outputs of some units serve as
inputs for others, which in turn provide critical inputs for the first set,
cannot be modeled in such simple ways. These require mathematical tech-
niques that permit both positive and negative feedback in a pattern in which
the resources are not wholly exogenous to the focal organization.
41 Modeling management and performance in networks

The implications of these various archetypes of environments, or

networked patterns, for management (separate from their implications for
modeling) should be obvious. In a pooled environment, the manager merely
has to be concerned with factors that directly affect his or her own organiza-
tion. Unless other organizations are linked into its environment, the actions
that they take are of little concern to the focal organization except as
competitors. In a simple pooled relationship, the aggregate operator is likely
a simple vector summation (S). In a sequential setting, the managers have to
be concerned with the operations of all the other organizations in the
sequence. Each must either convince an errant organization to change or
adapt his or her own organization to the change in inputs. In a pure
sequential relationship, the aggregation parameter is likely a multiplicative
one (P). In networks bound by reciprocal ties, organizational management
becomes similar to network management, with a web of relationships and
concerns that have to be incorporated into any decisions. Reciprocal rela-
tionships may need to be translated into sequential relationships in two or
more directions for mathematical estimation.
Of course, any organization’s environment can contain resource
(or constraint) linkages that fit all three types of interdependence. One can
put the point in the language of networks: networks can differ from one
another in the kinds and extensiveness of interdependent relationships
between and among the various nodes.15 Mixed relationships are likely to
have some combination of different aggregation operators, perhaps includ-
ing some we have not introduced here. Aggregation questions become rather
important when one moves to the network level, because the question of
aggregation and its form applies not only to the environment of a core
organization of interest but to the network itself – for instance, regarding
how the management function is aggregated across units. In short, the new
arithmetic operator  introduced in Equation (2.20) signifies different
operations, depending on the structure of interdependence within the
network.
The second term (
) is included to show that the internal network
management needs to be related to the aggregated internal management of
the hierarchies, and that the environment of the network needs to be related
to the aggregated environments of the hierarchies. Exactly how these elem-
ents are combined (addition, multiplication, and so forth) remains to be
discovered.
The network-level model outlined sketchily here suggests why deciphering
management in and of networks is more complex and demanding than the
42 A model of public management, and evidence

management of simple hierarchies. In a two-node network, the demands

might not be insurmountable; as the number of nodes increases, however,
the ability of either analysts or managers to consider all factors simultan-
eously soon exceeds the bounds of rational capacity. Managers, we think,
use a variety of coping techniques to allow themselves to manage the
network, and they may be able to make use of some heuristics from bodies
of work such as game theory to sort through certain kinds of circumstances
(see, for instance, O’Toole 1996). Coping techniques might include satisfi-
cing, rational shielding from nodes, ignoring some interdependences,
decoupling or “negative coordination” (Scharpf 1993) from nodes, adding
structure to the network environment, and so forth. The exact strategies can
be determined only via empirical analysis of how managers operate in these
networked situations.
For reasons that should be obvious at this point, data are not available
that would allow us to fully specify and test the network-level model in
enough cases to glean patterns of findings. Accordingly, the ideas in this
section should be considered initial, untested steps toward a fully developed
theory of public management and performance in networks. We also saw in
Chapter 1 that challenges often make it difficult to test performance-related
theoretical ideas of any sort across large numbers of cases. The next section
describes the main data set that we employ to explore the relationship
between management and performance in much of the remainder of
this book.

The Texas school district data set

An important challenge to the development of evidence-based public man-

agement is the availability of adequate data sets for systematic investigation
of these theoretical notions to see if they are actually valid. Public manage-
ment in particular has been slow to develop general data sets that can be
used to answer multiple questions important to the field (this coverage is
drawn from Meier and O’Toole 2009b). There is no public management
equivalent of the American National Election Study or the Panel Study of
Income Dynamics, two widely used data sets important in other realms
of social science. In our work we have tapped data on the management of
various other public services, but the bulk of our empirical work has relied
on another data set with some important advantages for this purpose. We
have started the construction of a data set on the school districts in the state
43 The Texas school district data set

of Texas (we say “started” because the process is ongoing). Rather than
beginning from scratch, we have opted to build on an existing data set that
had a wide range of performance indicators for more than 1,000 public
organizations over an established period of time – the Texas school district
data set. To that data set we have added managerial measures with a series of
surveys of top managers undertaken in 2000, 2002, 2005, 2007, and 2009,16
plus an additional survey of top managers concerning how they responded
to the devastating effects of Hurricanes Katrina and Rita, which unexpect-
edly ravaged that region in 2005.

The Texas context

Because we rely heavily on school districts in one state, some description of
the Texas policy context is in order. In response to a nationwide study
questioning the performance of public education (A Nation at Risk;
see National Commission on Excellence in Education 1983), Governor
Mark White launched a radical reform of Texas schools in the mid-1980s.
Standards for both students and teachers were raised. Students were
required to take more courses and more rigorous courses. State aid to local
school districts was increased in order to address the problems generated by
local variation in property values. The reforms were so fundamental that
academics took precedence over the cultural lynchpin of Texas schools:
football.17
Along with the reforms a state-level accountability system was established.
Students were required to take a series of standardized tests, with the
aggregate results published widely. The release of these test results is front-
page news throughout the state. Many top managers have performance
clauses related to these tests in their contracts. In addition to test scores,
the Texas Education Agency (TEA) also collects a wide array of data on
system finances and the characteristics of the student population. These
additional data permit the estimation of statistical models with elaborate
controls.
Education has remained on the statewide agenda continuously since the
early reforms. Subsequent governors have also stressed issues of perform-
ance. Testing systems have been refined and some early problems in regard
to validity were identified and addressed. The level of financial commitment
has not necessarily corresponded with the rhetorical efforts, and there are
continued concerns with equity issues.
44 A model of public management, and evidence

The structure of Texas districts

Texas districts tend to be very similar on some structural dimensions but vary
dramatically on others. All districts but one are independent school districts,
which means that they are governed by a locally elected school board that has
the power to levy taxes in support of education. The school boards hire a
professional administrator, the superintendent, to be the chief operating
officer of the schools. The superintendent has a great deal of discretion; he
or she sets the agenda for school board meetings, proposes the district budget,
establishes the schools’ curriculum, and oversees all personnel processes.
The superintendent has the formal authority to hire and fire managers
(principals, assistant superintendents, etc.) and general authority to move
personnel to different locations or positions. Teachers’ unions are relatively
weak in the state; and, even in the large districts, managers have substantial
control over who teaches for them. These formal powers are limited some-
what by a significant teacher shortage in the state as well as informal norms
and traditions.
These structural commonalities contrast dramatically with the vast other
differences – differences that are to be expected in a highly diverse state that
contains 8 percent of all US school districts. The districts range from wealthy
to poor. Even with substantial state aid, per student instructional spending
ranged from a low of $3,069 in 2007 to a high of $21,206. Correspondingly,
the percentage of students who qualify for free or reduced school lunches
(a poverty measure) ranged from 0.0 percent to 99.9 percent. Racially, the
state is highly diverse; the student population is 14 percent African American,
47 percent Latino, 35 percent Anglo, 3.4 percent Asian, and 0.3 percent
Native American. Individual districts vary greatly on these dimensions.
To illustrate, the percentage of black students in a district ranges from
0.0 percent to 86.9 percent while the Latino percentage runs the full range
from zero to 100.
In sum, these units of analysis are all school districts and share some charac-
teristics, but they are exceptionally varied on many other dimensions and thus
constitute a valuable source of information about public management. It is worth
emphasizing that more public employees work in the field of education in the
United States than in any other policy sector, and this Texas sample represents
more than 1 percent of all governments of any type in the country.
We did not seek to become the foremost experts on the management of
Texas school districts and to purposely irritate readers, reviewers, and
45 The structure of Texas districts

editors with a narrowly focused set of studies. As the research has unfolded,
however, each survey, each development of a new measure, has made this
data set more and more valuable and more and more superior to alternative,
available data sets – in five ways. First, the data set contains multiple
measures of management and multiple measures of performance in addition
to a wide variety of control variables. The control variables are chosen to fit
with the rather well-developed production function research literature on
public education. In this book we explain the details of the management
measures and control measures, as well as the different ways of tapping
performance. The multiple measures of performance allow investigation of
the fact that public organizations have multiple goals and may need to
emphasize one goal at the expense of another. The multiple measures of
management and the extensive controls mean that we can rule out alterna-
tive explanations of our findings and thus provide evidence that the results
are not spurious due to underspecification. The multiple measures of man-
agement also reflect the inherent complexity of the process by which public
managers influence performance.
Second, by having data on the same organizations over time, we can
address questions of causality (see O’Toole and Meier 2004b) and can
replicate studies for different time periods to determine if findings remain
valid (O’Toole and Meier 2004a, 2006). Both processes augment the
existing general advantages of a large-N approach. Third, the large size of
the data set – as many as 1,000 cases over a ten-year period – means that
complex relationships that include multiple interactions can be tested
without being limited by collinearity. Too frequently complex theories of
management are based on only a small number of cases (for example, Miles
and Snow 1978). Fourth, school districts have some valuable characteris-
tics. They are the most common public organization in the United States,
and similar organizations exist in virtually all countries. They are highly
professionalized organizations that are generally decentralized and vest
substantial discretion in street-level bureaucrats. To be sure, many public
organizations have different characteristics and thus limit generalizations,
but a large number of public organizations share these characteristics.
Fifth, the data set is accessible; we provide the data to all scholars who
request it, and we have invited other scholars to suggest additional items to
include. These factors have made the returns to investment increasingly
positive.18
We should also note, however, that the evidence-based research agenda
has also used a wide variety of other data bases, including local law
46 A model of public management, and evidence

enforcement agencies (Nicholson-Crotty and O’Toole 2004), Columbian

local governments (Avellañeda 2009b), UK local authorities (Walker,
O’Toole, and Meier 2007), the federal government and its use of PART
scores (Petrovsky 2006), institutions of higher education (Hicklin 2006),
unemployment insurance agencies (Wenger, O’Toole, and Meier 2008),
and the US presidency (Vaughn and Villalobos 2009). While we rely
primarily on the Texas school district data set in the coverage that follows,
we also tap some of our studies from other fields and data sets in our
examination of public management and performance. Further, where
appropriate we review findings from a number of others’ empirical set-
tings and relevant data sets, when these bear on our efforts to understand
the effects of management on performance. Accordingly, several such
analyses are summarized in relevant portions of the empirical chapters
that constitute the bulk of this book.

Measures of performance

Although virtually all programs have multiple goals and are therefore subject
to multiple performance indicators, some objectives are defined by the
political environment as being more important than others (O’Toole and
Meier 2004a). This study incorporates eleven different performance indica-
tors in an effort to determine how public management affects a variety of
organizational outcomes.
Although each performance indicator is salient to some portion of the
educational environment, the most noticeable by far is the overall student
pass rate on the statewide examination; called the Texas Assessment of
Academic Skills (TAAS) until 2002, it was then replaced by the Texas
Assessment of Knowledge and Skills (TAKS). The exams are standardized,
criterion-based tests that all students in various grades have to take. Initially
the exam was given in grades 3, 5, and 7, and as an exit exam. Currently
grades 3 to 8 must take the exam as well as the exit exam, which at different
times has been given in grades 10 or 11. The current system is also develop-
ing a series of end-of-course exams, such as 9th grade algebra, to assess the
learning of specific course material. The exit exam is a high-stakes test, and
students are required to pass it to receive a regular diploma from the state of
Texas. TAAS/TAKS scores are used to rank districts, and the examination
results are without question the most visible indicator of performance used
47 Measures of performance

to assess the quality of schools. Our measure is the percentage of students in

a district who passed all (reading, writing, and mathematics) sections of the
TAAS/TAKS.
The TAAS/TAKS exam has not been without controversy and challenges.
Standardized tests clearly do not measure all the relevant aspects of an
education system. There have also been unsuccessful court challenges argu-
ing that the tests are discriminatory on the basis of race and ethnicity. Much
concern has been expressed about cheating, given the high-stakes nature of
the test. Cheating is made more difficult because Texas keeps control of the
tests until administration and also is responsible for grading the tests.
Elaborate statistical procedures scan the tests for evidence of cheating
(erasures of wrong answers changed to right answers, etc.). In the few cases
in which cheating has been found, teachers and administrators have been
fired, and schools have had their state-assigned performance score reduced.
The institutionalized methods of cheating on the exams have been more
interesting. Bohte and Meier (2000) provide an extensive study of efforts to
manipulate exam scores by exempting students from the test, particularly
exempting students as a result of limited English skills or because they are
assigned to special education. They find not just sizable incentives to exempt
students from the exam but also patterns that correlate with theoretical
reasons to cheat (lack of resources, smaller districts, etc.). In 1997 the Texas
state legislature attempted to restrict this process by requiring that all
students be tested.
Despite the criticisms of the TAAS/TAKS, it has become generally
accepted as a measure of performance for evaluating schools. Many districts
use these scores to evaluate their superintendents, and many districts also
use them to assess the performance of principals. The Houston Independent
School District (HISD) and some of the other larger districts use the test
scores to evaluate and reward teachers. Over time these evaluation tools have
become far more sophisticated, moving from simply looking at raw test
scores to the use of elaborate econometric models that seek to isolate the
value added by the teacher or the school.
TAAS/TAKS scores have some useful statistical and practical advantages.
The scores are normally distributed except when the scores become too
high and a ceiling effect (districts cannot score above 100 percent) limits the
upper end of the distribution. This occurs because districts improve over
time, either because their education has improved or because students
become more used to the test form. The Texas Education Agency, as a result,
periodically adjusts the tests to make them more difficult (the transformation
48 A model of public management, and evidence

from TAAS to TAKS involved a significant increase in difficulty). These

year-to-year movements and adjustments mean that all statistical models need
to account for these annual fluctuations with a set of fixed-effects controls.
One of the contributions of the Texas school reform movement was a
focus on racial and economic equity. The reforms required that data be
gathered and reported on the basis of both race and income. The formal
state accountability system, in fact, requires a given level of performance on
all racial subgroups. Four other TAAS/TAKS measures are also useful as
performance indicators: pass rates for Anglo, black, Latino, and low-income
students.19 Low-income students are defined as those eligible for free or
reduced-price school lunches; this is an income criterion, established by the
federal government, that is linked to the official poverty level.
Many parents and policy makers are also concerned with the performance
of school districts regarding college-bound students. Four measures of
college-bound student performance were used: the percentage of students
who took either of the college board exams, the average ACT (American
College Testing) score, the average SAT (Scholastic Aptitude Test) score, and
the percentage of students who score above 1,110 on the SAT (or its ACT
equivalent). Texas is one of the few states in which both the ACT and the SAT
are taken by sufficient numbers to provide reliable indicators of both; as
with samples drawn from other states, there is no correlation between these
scores and the number of students taking them if the proportion of tested
students is more than 30 percent of the total eligible to be tested (Smith
2003). Texas scores on the ACT and SAT are generally uncorrelated with the
percentage of students taking the exams. Because most colleges and univer-
sities require either the ACT or the SAT, students who do not take one of the
exams are unlikely to go on to attend college. The 1,110 measure, the
equivalent of the top 20 percent nationally, is defined by the state of Texas
as an indicator of college readiness.
The college-related scores, or higher-end performance scores, are clearly
distinct from the TAAS/TAKS scores. The twenty intercorrelations between
the TAAS/TAKS and the college scores average 0.27, or about 7 percent
shared variance (the highest correlation is between SAT scores and the
overall TAAS/TAKS pass rate, 0.48). Interestingly, the college indicators are
not highly intercorrelated, except for the correlation between the 1,110þ
measure and the average SAT and ACT measures (0.75 and 0.76, respect-
ively). The correlation between ACT scores and SAT scores is only 0.58,
which is surprising, since both are intended to measure the potential for
students to succeed in college.
49 The production function

The final two measures of performance might be termed bottom-end

indicators: attendance rates and dropout rates. High attendance rates are
valued for two reasons. Students are unlikely to learn if they are not in class,
and state aid is allocated to the school district based, in part, on average daily
attendance. Attendance, as a result, is a good indicator of low-end perform-
ance by these organizations; the measure is simply the average percentage of
students who are not absent. The attendance measure is distinct from the
other measures of performance. Its highest correlations are 0.35 with the
overall TAAS/TAKS and –0.35 with the dropout rate.
Dropout rates are plagued by serious problems of measurement. Schools
have no incentive to determine if a student who does not return to school has
dropped out or is attending school elsewhere. Reported dropout rates are
widely conceded to be an underestimate. In addition, there are questions about
whether or not to count a person pursuing a General Educational Development
(GED) test as a dropout. The state of Texas has also changed its measure of
dropouts during the period of this study – going from a six-year dropout rate
(the average dropout rate for grades 7 to 12) to a four-year dropout rate (grades
9 to 12). Alternative measures of dropouts based on the size of cohort that
graduates versus the size of that cohort in earlier years are greatly affected by the
high rates of mobility, particularly minority student mobility, of Texas students.
Given all these problems, one needs to be skeptical about the results of analysis
on dropout rates. For most of the analysis, dropout rates are not included, but,
in some cases, dropout rates are one of the better – albeit flawed – indicators of
a district’s performance dimension: how well the school serves at-risk students.
The dropout measure is not highly correlated with the other performance
measures; it averages a correlation of only –0.17.
With eleven different performance indicators, we do not intend to subject the
reader to a tedious discussion of each performance indicator in every empirical
analysis. Rather, for the purposes of validating our major managerial concepts,
we use most of the indicators. When our analysis becomes more specialized, we
select performance indicators with given characteristics (e.g. high task difficulty,
low-end performance, etc.) that provide the best theoretical test of the model.

The production function

Any assessment of public program performance must control for both task
difficulty and program resources. For school districts, neither of these types
of elements is under the substantial control of the districts themselves, and
50 A model of public management, and evidence

therefore they can be considered key parts of the vector of environmental

forces. Fortunately, a well-developed literature on educational production
functions (Hanushek 1996; Hedges and Greenwald 1996) can be used for
guidance. Eight variables, all commonly used, are included in our analysis:
three measures of task difficulty and five measures of resources.
Schools and school districts clearly vary in how difficult it is to educate
their students. Some districts have homogeneous student populations from
upper middle-class backgrounds. Students such as these are quite likely to
do well in school regardless of what the school does (see Burtless 1996).
Other districts with a large number of low-income students and a highly
diverse student body will find it more difficult to attain high levels of
performance, because the schools will have to make up for a less supportive
home environment and deal with more complex and more varied learning
problems (Jencks and Phillips 1998). Poor and minority students often
lack the in-home learning tools (computers, books, etc.) that are common
in middle-class homes. Texas also has a large immigrant population and
thus needs to provide programs for students with a native language other than
English. Our three measures of task difficulty are the percentages of students
who are black, Latino, and low-income. The last-mentioned variable is
measured by the percentage who are eligible for free or reduced-price school
lunches. All three measures should be negatively related to performance.
While the linkage between resources and performance in schools has been
controversial (see Hanushek 1996 and Hedges and Greenwald 1996), a
growing literature of well-designed longitudinal studies confirms that,
like other organizations, schools with more resources generally fare better
(Wenglinsky 1997). Five measures of resources are included. The average
teacher salary, percentage of a district’s expenditure funded by state aid, and
class size (see Molnar et al. 1999, Graue et al. 2007, and Dee and West 2008)
are directly tied to monetary resources. The average years of teaching experi-
ence and the percentage of teachers who are not certified (Laczko-Kerr and
Berliner 2002) are related to the human resources of the school district. Class
size and noncertified teachers should be negatively related to student per-
formance; teacher experience, state aid, and teacher salaries should be
positively related to performance.
This set of eight production function variables is used in all analyses
conducted with this data set. Since we generally are interested in the role
that management plays in organizational performance rather than a full
specification of the determinants in education policy, we do not normally
discuss the relationships for these control variables. Only when these
51 Conclusions

variables show something unexpected or when they interact with managerial

factors to affect performance are they discussed in the text.

The presentation of findings

In the decade of research on public management we have frequently added

additional data and developed new measures. As a result, many of the
studies have been carried out over different time periods, and sometimes
with slightly different measures of management. We have tried to maintain
consistency with our published work, so, rather than rebuild an entirely new
data set and rerun all the analysis, we have usually opted here to present the
original findings. To avoid redundancy in the discussion, we often present
abridged tables that show the key relationships. In this manner we avoid
focusing on control variables that are not of substantive interest in this
project. In a few cases, such as with the analysis of managerial quality in
Chapter 4 or the budget crisis in Chapter 6, we present new analysis for two
reasons: (1) to present what we think is a better analysis; and (2) to make the
text easier to read and consistent with the analysis that preceded it.

Conclusions

There are many possible approaches to exploring the relationship between

public management and public organizational or program performance. The
approach we have adopted is to build from the inductive, primarily case
study, literature to formalize mathematically some of the general relation-
ships that seem to be suggested by the earlier work. The model we have
developed may seem simple – perhaps overly simple – in certain respects.
For instance, it is comprised in its entirety of four variable clusters: O, M, S,
and X. As mentioned in Chapter 1, many of the fine-grained details of public
organizational and managerial life do not find a home in this model. Still, as
will become clear in the chapters to follow, quite a number of issues and
relationships can be investigated through this research program.
In other ways, the model is complex – in particular, in its specification of
nonlinear and reciprocal relationships between some variables, as well as in
our efforts to begin modeling the multi-level reality of management in
networks. For this latter challenge, the work has just begun. This chapter
points to some of the issues at stake as well as some of the impediments to
52 A model of public management, and evidence

systematic investigation, but we do not yet press forward with large-N

empirical studies of management at the network level. Rather, we work on
the management–performance links at the organizational level, while also
taking explicitly into account the important fact that managers of organiza-
tions must operate externally as well as internally, and that the outputs and
outcomes of their programs can be shaped in important ways by features of
the environment with which their units are interdependent. For the former,
it is clear, no one empirical study can explore all the relationships and
interactions. The demands such an effort would place on a data set would
make it impossible to execute any single definitive study. The approach we
take, therefore, is to work from the general model, and to explore aspects of
it through several related analyses – often via some simplification of the
model – in order to focus on one or a few relationships at a time. A number
of these are presented in the remaining chapters of this book.
Because of the high and increasing value of the Texas school districts data
set, as explained in this chapter, we conduct much of our work by examining
performance-related relationships in these roughly one thousand govern-
ments over a period of several years. Where appropriate, we supplement
these core analyses with findings from some additional empirical settings.
In the next chapter, we begin the empirical exploration by focusing on the
externally oriented, networking behavior of top managers. We explain in
much more detail the governmental units and the data with which we are
working and begin to answer the question of how management shapes
results.

NOTES

1. Indeed, Max Weber defined bureaucracy in part in terms of stable decisions over
time based upon precedent, thus suggesting reinforcement across certain stabilizing forces.
See Gerth and Mills (1958).
2. Some observers might point to another sort of array: the market. The pure neoclassical
market setting is characterized by an absence of structural stability, however, aside from
some basic rules of the game – such as contract law, barriers to collusion, etc. Markets
in the classic sense, in other words, are defined largely in terms of an absence of
structure among the actors. In practice, of course, markets are often structured to
some extent. How structuring and rules affect markets is an important issue of policy
design, but we do not address it in this volume. Here we are interested in structured
relationships among relevant actors, as these vary between hierarchy and network as
the archetypical forms.
53 Notes

3. Beyond the structural variation between hierarchy and network, we also want to take
note of other stabilizing elements mentioned earlier as we consider the role of public
management. In the initial formulation of our model (O’Toole and Meier 1999), we took
structural variation between stable hierarchies and flexible networks into account; but we
omitted other stabilizing forces. The model was later expanded to include the latter as
well (O’Toole and Meier 2003b), and we work from the broader version of the model
throughout this book.
4. An interesting question in management is the degree to which treating it as a vector sum
provides a misleading picture by ignoring whether or not management is consistent
throughout the organization (Andrews et al. forthcoming (b)). The impact of managerial
consistency on performance in theory would be positive but one can envision an
organization with too much consistency – e.g. groupthink – that would lead to poor
performance. Consistency needs to be explicitly modeled; surveying multiple respond-
ents and then averaging the responses (Enticott, Boyne, and Walker 2009) provides no
more information than a single-manager survey.
5. The outputs of an organization or program are the immediate consequences of policy
and management efforts: bridges built, cases processed, environmental permits issued,
etc. Outcomes relate to the eventual impact of policy actions, along with the results of
other causal variables, on the ultimate issue or concern prompting the initial policy
intervention. An example of an outcome, for which environmental permits would be an
output, is cleaner rivers and streams.
6. All these issues can be handled through appropriate conceptualization and methods.
Indeed, we address them all in the empirical chapters of this book.
7. This interpretation affects how organizations deal with their environments. That is to say,
they can assume some degree of change and build that into the inertial aspects of the
organizations rather than treating any environmental change as something new and different.
8. If b were to exceed one, the system would generate positive feedback and eventually
explode.
9. Empirically, there is little research on what the stability coefficients are for organizations.
A value of 0.99 might well be far more rigid than anything that exists in the real world of
organizations.
10. Shocks themselves can have a variety of functional forms and both short- and long-run
impacts; with adequate data, all these impacts can be estimated.
11. We realize that sometimes internal management is aimed at changing things – for
example, due to slacking, underperformance, etc. Although this recognition suggests
that management sometimes has a destabilizing influence, that influence is only in the
short run. Once such efforts are successful at fixing organizational problems, manage-
ment will need to institutionalize the changes via stabilizing structures. We return to the
point in the concluding chapter.
12. For a somewhat different way of distinguishing and partitioning managerial functions,
see Moore’s treatment of managing upward, downward, and outward (1995).
13. It is possible for management to operate independently of the buffering structures.
Management can act in a boundary-spanning function to reach out to other organiza-
tions or monitor potential changes in the environment. Such functions might even be
institutionalized in strategic planning units or units for organizational intelligence.
54 A model of public management, and evidence

14. In this exposition, we simplify for the moment by assuming that management knows
what it is doing – that is, that management operates with considerable skill or quality, not
simply effort. As will become clear later in this book, we consider each managerial
function to contain both an effort and a quality component. We introduce and validate
a measure of managerial quality in Chapter 4, but the focus here is on effort. An
alternative way of thinking about the model at this point is that it contains an assump-
tion of some deliberativeness or choice on the part of management – as to the allocation
of managerial effort across the functions of management.
15. Networks can differ from each other in many other ways as well: the number of nodes,
the degree of centralization, and numerous other dimensions sketched by those who
employ the tools of social network analysis.
16. The 2009 survey was unique, in that the questionnaire included a variety of questions
submitted by other scholars.
17. Part of the reforms was a provision that students who did not pass their courses were not
allowed to participate in extracurricular activities such as football. The adoption of the
reforms and their extensive nature owe a great deal to the leadership of Governor White
and his designated reform advocate, Ross Perot. Although some people term these
reforms and their impact “the Texas miracle,” in reality this set of changes was a
twenty-plus-year process that involved significant effort on the part of politicians and
school officials. Overall, the state has made substantial progress in educational attain-
ment, but it still has a fair way to go.
18. It is also fair to point out that the Texas school district data set has contributed to several
literatures other than the public management research field. These include the study of
representative bureaucracy (Keiser et al. 2002), the impact of charter schools on public
school competition (Wrinkle, Stewart, and Polinard 1999; Bohte 2004), the patterns of
organizational cheating (Bohte and Meier 2000), punctuated equilibria in policy settings
(Robinson et al. 2007), and the investigation of statistical techniques (Bretschneider,
Marc-Aurele, and Wu 2005; Wagner and Gill 2005), among others.
19. The various pass rates do not correlate as highly as one might imagine. The intercorrel-
ations between the Anglo, black and Latino pass rates are all in the neighborhood of 0.67,
thus suggesting that the overlap is only about 45 percent. The individual scores
for race and class correlate more highly with the overall score, because they are sub-
components of it.
3 Public management in interdependent
settings: networks, managerial
networking, and performance

When people think of what public managers do, often the tasks and
responsibilities that come most readily to mind are those tied to the
internal functioning of a public organization: motivating staff, organizing
tasks, structuring work relationships, handling the budget and other
resources such as information technology, appraising individuals’ perform-
ance, and the like. We begin our empirical examination of public manage-
ment from another angle: the externally oriented actions of managers as
they seek to do their jobs and advance their organization’s causes. We do so
for two reasons. First, this aspect of public management is often given
short shrift in standard accounts, and yet – as explained earlier in this
volume – contemporary governance arrangements typically enmesh the
actions and objectives of specific public organizations in a web of relations
with other actors. Second, in the development of our own research pro-
gram, we began by studying the external efforts of managers and sought to
explore their performance-related implications.1 Accordingly, in this book
we proceed in like manner.

Networks and networking

As noted earlier, public programs and public organizations are often situated
in networks – arrays through which many aspects of contemporary govern-
ance are handled. Networks are structures of interdependence involving
multiple organizations or parts thereof, in which one unit is not merely
the formal subordinate of the others in some larger hierarchical arrange-
ment. Networks exhibit some structural stability but extend beyond formally
established linkages and policy-legitimated ties. The institutional glue con-
gealing networked ties may include authority bonds, exchange relations, and
coalitions based on common interest, all within a multi-unit structure
(O’Toole 1997: 45).
55
56 Public management in interdependent settings

Managing in networked settings presents a challenge, not least because

public managers cannot be expected to exercise decisive leverage by virtue
of their formal position. Influence in larger networks is more difficult to
document, predict, and model than it is in relatively simple two- or three-
party relationships. If managing in networked settings can be rather
difficult, therefore, why do policy makers situate public programs in such
arrays?
An extensive research literature developed in North America and Europe
contends that implementing programs in interorganizational networks can
offer significant advantages. The expertise and/or resources needed to
address pressing policy challenges may be spread among multiple organiza-
tions and across various sectors. Mobilizing networks of such organizations
can sometimes create the right combination of technical knowledge and
critical mass of effort. Policy problems also may touch upon several
jurisdictions simultaneously: the energy supply is an issue not only for
the Department of Energy but also for the Environmental Protection
Agency, not to mention the Treasury (balance of payments), the State
Department (relations with the Organization of the Petroleum Exporting
Countries [OPEC]), and the jurisdictions of various states (drilling in
Alaska’s North Slope, disposal of low-level nuclear waste in Nevada and
elsewhere, etc.). For such “wicked problems” (Rittel and Webber 1973)
that have no simple governmental or organizational niche for proper
treatment, a multiorganizational networked arrangement may be an appro-
priate institutional response. Often such multi-actor arrangements are
encouraged by the political dynamics of policy implementation and public
management. Adding organizational actors to the patterns of policy execu-
tion can build support for programs, and public managers often find it
useful to engage in regular patterns of interaction with actors outside their
organization so as to tap opportunities available by and from others and/
or to fend off potential disruptions and threats to the core organization’s
operations.
Inducements toward networked arrays, therefore, can be numerous.
Sometimes governments formally stipulate that programs have to involve
multiple organizational actors. Hall and O’Toole (2000) find that the great
majority of new or substantially revised public programs enacted by the US
Congress during two different time periods mandated or strongly encour-
aged the regular involvement of multiple organizations. Pressures to achieve
results during implementation often further complicate these arrays. In a
companion study to their 2000 article, Hall and O’Toole (2004) explore the
57 Networks and networking

interorganizational arrangements specified in regulations, following the

enactment of legislation. The pattern is clear: implementation brought even
more involvement of additional organizations, along with more complex
patterns of interdependence.
Beyond mandates from political leaders, networks of interdependent
actors sometimes emerge as a result of the voluntary choices of public
organizations and some of those with which they interact. “Collaboration,”
“public–private partnership,” and other like forms of mostly voluntary
linkage are now much discussed in the literature of public management
(see, for instance, O’Leary and Bingham 2009). In certain cases the net-
works that emerge consist of complex combinations of mandated (or
policy-encouraged) links along with other ties that are mostly or entirely
voluntary. These arrays may include public, private, and not-for-profit
organizations, perhaps at several different (geographical or scale) levels.
Needless to say, the public management challenges in such complex settings
can be immense, even though the prospects for significant policy-related
problem solving may offer strong encouragement for using such govern-
ance structures.
Examining the structure-related impacts on performance of a wide
variety of networks is an important long-term objective of public man-
agement and public policy research. For reasons of scope and practicality,
however, we address systematically only a portion of the relevant research
questions here. We do not investigate how network variations across
different policy problems, involving public organizations of widely
differing jurisdictions, shape performance. The literature of social net-
work analysis provides tools to characterize network variations in many
dimensions, but here we would run up against the apples-and-oranges
problem of trying to force different organizations, networks, and pro-
grams into some sort of common performance metric. For reasons
explained earlier in this book, such an approach is best avoided.
We also do not devote substantial space to a full depiction of the networks
in which the public managers subject to our empirical investigation operate.
Doing so would require detailed data gathering from many actors in each
jurisdiction, thus limiting severely the number of cases we could systematic-
ally examine. While important work on networks and public management
can be undertaken with a study of only a few cases (for instance, Provan and
Milward 1995), in such cases it is very difficult to control for other influ-
ences, and any observed regularities could be attributed to any number of
causal forces.
58 Public management in interdependent settings

In this chapter, therefore, we explore a couple of slices of the “networks”

question. In particular, we examine the behavior of top managers as they
interact with external parties in the course of trying to do their job (for a full
treatment of the advantages undergirding this approach, see Meier and
O’Toole 2005). Even for managers in similar positions in public organiza-
tions of similar types, we can expect their “networking” behavior to vary
considerably – in frequency and direction. Networking by managers does
not depict the full set of structural relationships in complex networks – for
example, regular ties between other actors within the network are omitted –
but it is difficult to conceive of how public managers can work in their
interdependent environment to shape results unless their behavior includes
interactions with others. Since our interest lies in how public managers’
actions shape performance, the networking behavior of top managers is a
logical focus to address this question.
Most of the coverage in this chapter concentrates on the networking
behavior of such managers as they work externally in the interdependent
environment of their school districts. (We include some data on networking
behavior in other types of public organizations as well to show a general
pattern, and provide a concluding summary of other work later in the
chapter.) School districts are not the most complicatedly networked of
public organizations. Indeed, the very existence of school districts as a
special form of special district is due to the pressure in an earlier era to
“insulate” education from politics, given the importance of the education
function to society.2 In today’s governance settings, however, schools and
school districts are interdependent with and relevant to the interests of a
variety of other parties in the organization’s environment (see Chubb and
Moe 1990 and Wirt and Kirst 2005). Accordingly, exploring the relevance of
managerial networking to school district performance is an appropriate task.
Indeed, if we are able to find influences on performance from managerial
networking in rather less networked settings, such a finding would suggest
that managerial networking behavior in more thoroughly networked con-
texts is likely to be even more consequential.
While our focus is mostly on the networking of public managers, we do
not ignore the structural aspects of networks. For a portion of our analysis in
Chapter 7, we introduce both managerial networking and structural dimen-
sions of the networks in which they operate; there the objective is to explore
the independent effects of each upon performance.
Although the coverage here is selective, in the senses just explained, it will
quickly become obvious that there is plenty to examine in the impacts of
59 Do public managers network with external parties?

managerial networking on program results. We proceed, therefore, from a

relatively straightforward analysis, report the empirical results, and then add
further questions and analyses to depict the patterns more completely.

Do public managers network with external parties?

The most straightforward questions have to do with whether public man-

agers engage in networking behavior – and, if so, how, and how frequently?
The answers are “Yes,” and the pattern and frequency vary by managerial
role and the nature of the other parties.
In the case of the more than 1,000 Texas school districts, we have
developed a measure of networking behavior on the part of top managers,
the superintendents of the districts. Our measure of managerial networking
is an effort to operationalize our M2 term in the model – the actions of the
manager in the networked environment of a public agency. This work
assumes that managers cannot engage in network-like behavior with other
actors in the environment without coming into contact with them. Using
the Texas school district data set, we asked top managers to rate how
frequently, from daily to never, they interact with each of a set of environ-
mental actors (five actors in the 2000 survey, eight actors in the 2002 and
2005 surveys, ten in the 2007 survey, and eleven in 2009): school board
members, other superintendents, local business leaders, the Texas Education
Agency (the state-level oversight organization), state legislators, federal
education officials (2002, 2005, 2007, and 2009 only), parent groups such
as parent–teacher associations (2002, 2005, 2007, and 2009 only), teachers
associations (2002, 2005, 2007, and 2009 only), police/fire departments
(2007 and 2009 only), other local governments (2007 and 2009 only), and
non-profit organizations (2009 only). These items have been factor-
analyzed, and consistently produce a first factor that is a general networking
measure with all positive loadings (at times one factor only is produced).
Table 3.1 displays the means and the factor loadings from the 2007
survey results. A similar pattern obtains for the networking data from
other years. It is clear that these top managers do engage in networking
behavior with a variety of external actors, though it is also clear that some
of these are more frequent interaction partners than are others. School
board members are the most frequent contact, but superintendents report
weekly contacts with other superintendents, local business leaders, the
Texas Education Agency, city/county governments, local police/fire
60 Public management in interdependent settings

Table 3.1 Factor loadings for nodes involved in managerial networking, 2007

Node Loading Mean

School board 0.5504 4.46

Teachers’ associations 0.4778 2.01
Parent groups 0.3863 2.97
Local business leaders 0.6673 3.65
Other superintendents 0.5396 3.93
Federal education officials 0.4530 1.74
State legislators 0.6042 2.42
Texas Education Agency 0.4798 3.13
City/county government 0.7105 3.06
Local police/fire department 0.6272 3.09
Eigenvalue 3.12
N 757

Table 3.2 Levels of interaction are fairly constant over time

Node 2007 2005 2002 2000

School board 4.46 4.44 4.78 4.48

Teachers’ associations 2.01 2.09 2.18 NA
Parent groups 2.97 3.02 3.02 NA
Local business leaders 3.65 3.74 3.86 3.86
Other superintendents 3.93 3.89 4.16 3.95
Federal education 1.74 1.81 1.81 NA
State legislators 2.42 2.61 2.31 2.35
Texas Education Agency 3.13 2.61 3.39 3.21
City/county government 3.06 NA NA NA
Local police/fire department 3.09 NA NA NA

departments, and parent groups. As Table 3.2 shows, however, the relative
level of interaction changes little over time. Furthermore, past the first four
or five external nodes, the factor scores are relatively insensitive to the
number of nodes that are inquired about (see the discussion below). This
point is important, since some might argue that one needs to know the
full set of interactions before it is possible to understand how managers
operate in their networked environment – a hurdle that would make it
exceedingly difficult to explore such patterns in hundreds of organizations
and with hundreds of managers. The evidence shows, rather, that most of
what one needs to know about the general networking pattern exhibited by
61 Do public managers network with external parties?

managers can be educed from information about interactions with the

most important, or most frequently contacted, nodes.
A factor score, then, which indicates the extent to which the manager
interacts with external parties, is taken as the measure of M2.3 The consistent
production of a first factor of this sort is strong presumptive evidence that
the factor represents a general networking measure.
Discussing this measure is worthwhile, since it plays such an important
role in our research. Measurement issues in regard to management more
generally are important simply because this research program is one of the
most well-developed efforts to create measures of management for use in a
large-N quantitative set of performance studies. The argument for this
measure of M2, or managerial networking, as a reliable and valid measure
of management activities rests on several grounds. First, the factor analysis
of networking items reveals that contact with environmental actors forms a
consistent pattern across nodes. Such contacts are all correlated with each
other, and always produce a generic first factor with positive loading regard-
less of how many nodes are included in the analysis.
Second, M2 is positively correlated with a manager’s time estimates of
how much of his or her effort is focused outside the organization (as
opposed to focused on internal matters; see Meier and O’Toole 2003: 698,
note 4).4 In addition, as explained shortly, the strong results in various
linkages to performance with a wide variety of indicators (O’Toole and
Meier 2003b: 54, 56) reveal a concept with a great deal of empirical import
and external validity.
Third, another way to partially validate the measure of managerial net-
working is to see if it correlates with other variables where relationships
should exist. Superintendents who are more aggressive at managerial net-
working, all other things being equal, should have a school district that has
greater community support, greater school board support, and more paren-
tal involvement. Simply stated, more aggressive networking should result in
greater support in the external environment. Our survey asked superintend-
ents to rate community and school board support on a five-point scale from
excellent to inadequate.5 The survey also asked for a similar evaluation of
parental involvement.
Table 3.3 presents three regressions showing the relationship between
managerial networking and support from the school board, the community,
and parents. To make sure that any relationships are not the result of better
past performance, district poverty, or district resources (teachers’ salaries
and revenues per student), we control for these factors. More networking is
62 Public management in interdependent settings

Table 3.3 Managerial networking improves environmental support

Dependent variables

School board Community Parental

Independent variables support support involvement

Managerial networking 0.0691 (3.94) 0.1126 (7.09) 0.0764 (4.39)

Past performance 0.0034 (1.94) 0.0073 (4.67) 0.0059 (3.40)
Low-income students 0.0033 (2.94) 0.0079 (7.64) 0.0122 (10.81)
Teacher salaries (000s) 0.0016* (0.24) 0.0193 (3.13) 0.0208 (3.08)
Revenue per student (000s) 0.0239 (1.85) 0.0204 (1.74) 0.0042* (0.33)

R2 0.02 0.09 0.10

F 9.37 50.61 55.57
Standard error 0.88 0.80 0.87
N 2,524 2,534 2,529

Notes: T-scores in parentheses. * ¼ not significant at 0.05 level, one-tailed test.

positively associated with support from the school board, support in the
general community, and the level of parental involvement. Each relationship
contributes additional evidence that we have created a reliable and valid
measure of managerial networking.
Fourth, by comparing survey results between two time points – for
instance, between 2000 and 2002 – we provide a systematic evaluation
of the concept’s reliability and validity (Meier and O’Toole 2005). The
2000/2002 assessment reveals that the exact number of nodes included in
the measure was not especially crucial; the five-node measure correlated
strongly with the eight-node measure (as did the four- and seven-node
measures). Similar patterns are found in comparing managers’ responses
between the more recent surveys. This finding does not mean that scholars
can select any set of nodes to create this measure but, rather, that they need
to select the most common nodes that occupy a manager’s time. Given
careful selection, the total number of nodes becomes less relevant; research-
ers should stress getting information on the most common nodes rather
than worrying about information on all nodes.
The 2000/2002 comparisons, along with comparisons between more
recent time periods, also reveals that M2 is very much a managerial choice.
At least in principle, measuring interactions does not reveal whether it is the
manager seeking to network with actors in the external environment, or
whether it is the other actors and/or pressure from external forces that
stimulates the networking behavior. Analysis of our data reveals that it is
63 Do public managers network with external parties?

clearly the former. Networking measures for a given organization at the two
time points were essentially uncorrelated if the organization had changed
top managers. When the same manager was in place in both years, there was
a strong positive correlation between the two measures of M2 (above 0.5 – a
notable correlation for a behavioral measure such as this one). It is clear,
therefore, that networking itself is driven largely by managers’ decisions. It is
not an epiphenomenon forced on managers by the external actors (Meier
and O’Toole 2005).
There is also evidence that managers’ networking style can help to
explain the emergence of interorganizational collaborative links, at least
during crisis periods. This evidence is also drawn from Texas school
districts and their top managers. In a natural-experiment design in the
context of Hurricanes Katrina and Rita, we sought to learn if pre-disaster
levels of managerial networking were related to the post-hurricane emer-
gence of interorganizational collaboration with particular institutions in
the settings of Texas school districts. With appropriate controls, we deter-
mined that generalized networking behavior prior to the crisis helps to
explain the development of collaborative relationships in response to the
disaster, and not simply or primarily contacts with the pre-crisis network-
ing partners (Hicklin et al. 2009).
Finally, it is noteworthy that the factor loadings in Table 3.1 and the
discussion of nodes in the current chapter include the school board as an
“external” node. One might argue that the school board is the primary
political principal of the school district and should not be treated as a
networked, or networking, partner of the top manager. (Networks, as defined
at the outset of the chapter, reference non-hierarchical linkages.) It turns
out that the factor scores themselves do not change much, whether or not
school board interactions are included in the factor analysis. Nonetheless,
our two-year comparison and subsequent work (see O’Toole, Meier and
Nicholson-Crotty 2005: 57–8) have found that direct hierarchical linkages
might be best treated as separate interactions. While interactions with the
school board do correlate with interactions with other nodes, the portion of
the school board variance that is uncorrelated with M2 shows a much different
relationship. Unlike top managerial networking with other external actors
(see below), superintendent–school board interactions generally demonstrate
a negative impact on performance, when one controls for M2 using inter-
actions with the other nodes. Whether this negative relationship is the result
of political meddling in the administrative process or merely reflects an
endogenous fire alarm about performance cannot be determined from the
64 Public management in interdependent settings

data at hand.6 What is clear, however, is that the school board–superintendent

relationship is, in part, of a different form, with different results from the
superintendent’s normal effort to manage the interdependent environment
(see also Meier, O’Toole and Lu 2006).
In short, therefore, we have a good measure of the networking behavior of
these public managers. Superintendents vary considerably in their manager-
ial networking, and a considerable amount of such behavior is reported by
them. They interact with a variety of external actors, and their behavior
seems to be a matter of their own discretionary choice, rather than an activity
pressed upon them by their networked environment or other external forces.
Are top managers of school systems unique in this regard? The answer
would appear to be “No.” With colleagues, we have adapted the networking
measure to a completely different set of public organizations: local govern-
ments, called local “authorities,” in England. The reasons typically offered as
to why public managers deal with the network of actors in their organiza-
tions’ environments are unique neither to the United States nor to school
districts. So we should expect to see similar behavior in such settings as
English local authorities. Furthermore, top managers are not the only ones
who have opportunity and some reason to network in their environment. In
our English study, therefore, we have surveyed managers at three organiza-
tional levels in these jurisdictions. We surveyed so-called corporate officers, a
group that has a perspective on and responsibilities for the organization as a
whole; chief officers, who manage the delivery of particular services; and
service managers, who are first-line supervisors and have a sub-service view.
Our data were gathered in 20037 via an e-mail survey of a sample of
102 English local authorities. The authorities were selected to be representa-
tive on a number of background variables. In each authority, questionnaires
were sent to three corporate informants, the chief officer, and three man-
agers in each of seven service areas. The total number of potential inform-
ants was 2,299, and the number of respondents was 1,026, thus reaching a
response rate of 44.5 percent.
We asked the local managers about their interactions with eight different
types of external actors: elected members of the authority, user group
representatives, trade unions, local business leaders, voluntary sector actors,
Members of Parliament, managers in other local councils, and central
government officials. This group of actors is roughly equivalent in terms
of functions to the nodes about which we have surveyed in the environments
of school districts in Texas. Two sets of findings in particular are worth
noting in connection with the present discussion. First, as with top managers
65 Do public managers network with external parties?

Table 3.4 Mean scores for interactions between different groups of officers in English local
authorities

Corporate officers Chief officers Service managers

Mean Mean Mean

Elected members 5.07 5.42 4.40*

User group representatives 2.68 3.24 3.03*
Trade unions 2.74 3.10 2.64*
Local business leaders 2.87 2.83 2.27*
Voluntary sector actors 2.87 3.20 3.01*
Central government officials 3.01 3.12 2.68*
MPs 2.16 3.31 2.42*
Managers in other councils 3.61 3.59 3.51

Note: * ¼ significantly different from others at p < 0.05.

of school districts, managers in local authorities report interaction patterns

that, upon factor analysis, produce a clear first factor on which all the pairs
of interactions load positively. Once again, therefore, a general managerial
networking measure emerges from the data. Second, the specifics of the
interactions vary by managerial level or responsibility. Table 3.4 displays the
mean scores for interactions between different groups of local government
officers. The most important group of local actors that all three types of
managers are likely to interact most frequently with were elected members,
with interaction more than once a week for all managerial types. This
pattern is to be expected, given the elected members’ special relationship
with local government and its managerial cadre. By contrast, there was no
one group of networking partners with which officers were least likely to
interact with. Chief officers were the most likely to interact across all the
nodes. Corporate officers, by contrast, are the least likely to interact with
others. Two reasons may explain the pattern. Senior managers may spend
time managing downward and internally offering buffers within the organ-
ization. They are also likely to be focusing their time on non-management
issues such as policy development (for more detail regarding this analysis of
networking by managers in English local authorities, see Walker, O’Toole,
and Meier 2007).
In short, while there are differences in the networking behavior of man-
agers between US school districts and English local authorities, the broad
patterns are similar. Clearly, a part of the public manager’s function, wher-
ever he or she is located, is managing in and with the interdependent
66 Public management in interdependent settings

environment. It is also clear that managers with different functions and

located at different organizational levels report some differences in the
details of their networking activities. Such differences are also to be
expected, since different managers face different challenges and responsi-
bilities, as well as different relationships of interdependence with other
actors and organizations. The kinds of findings we develop regularly from
Texas school districts, then, would seem to be part of a broader pattern for
public managers more generally.
We turn now to a particularly interesting question: given that managers
interact with some frequency with other actors and organizations in their
environment, does this networking carry impacts on the performance of
their organizations? Some of the case study literature indicates that the
answer may be “Yes,” and our model clearly specifies a relationship – indeed,
a set of relationships. We explore this important subject next.

Managerial networking and performance

When public managers operate in networked settings, they face many

options for action and many strategic choices. Arguably the most important
of these is the decision about how much time and energy to work in the
networked environment, and in which directions – that is, with which
external actors. After all, managers must also consider that efforts might
have to be devoted to buffering program activities from the potentially
turbulent and uncertain impacts from the interdependent environment.
Managers also must devote attention to the internal management of their
units. Many other strategic choices confront managers who function in
networks, of course, including which issues to raise with others in the
network, what positions to take, what style to exhibit, how to balance
short-term versus long-term needs, and so forth. The foundational choices,
however, are how much and with whom to undertake networking action
itself. We concentrate on this topic in the next portion of this chapter.
As noted earlier, school districts are by no means the most thoroughly
networked settings in which public managers operate, but there are reasons
to expect patterns of interdependence that should carry performance impli-
cations.8 The technical and political demands placed upon school district
superintendents encourage them to develop, solidify, and use ties with other
important actors in their environments. The most important of these are
typically their own school board9 (the elected body responsible for sketching
67 Managerial networking and performance

broad policy for the district), the relevant state-level educational department
(a source of primarily formula-based funding that varies in importance from
state to state, as well as a unit that issues some regulations that apply to the
local districts), state-level legislators (who frame general educational policy),
local business leaders (who can play crucial roles in supporting the locally
enacted taxing decisions that drive much of school district revenue), and
other superintendents (professional colleagues and sources of experience
and innovation in the turbulent world of public education). Other parties
may also be tied to the operations of school districts, and, as explained
earlier, we have surveyed top managers about broader patterns of potential
interaction; but the interaction partners just mentioned can be expected to
be the most important, and most regularly contacted, in the typical case.
In contemporary American public education, in which funding issues are
critical and many ostensibly separate policy problems – e.g. drug abuse,
broken families – intrude in highly visible ways in the educational process,
schools have become battlegrounds for a range of policy disputes (Chubb
and Moe 1990; Meier and Stewart 1991). Efforts to reform schools and
influence educational policy are frequently debated and adopted in realms
where the school district is only one voice among many. Accordingly,
superintendents may have reason to devote managerial energy and effort
to understanding and leveraging their networked environment. This net-
work orientation is more extensive in the US context than in most other
countries. An Organisation for Economic Co-operation and Development
(OECD) (1995: 52) study of fourteen national education systems found
that more education decisions had to be made in consultation with others
(44 percent) in the United States than in any other Western democracy. US
systems were also rated the lowest in terms of local – that is, school district –
autonomy.
Superintendents manage their districts – a headquarters office along with
sets of schools, which in turn are managed by school “principals” – within
this broader constellation of other actors, who may be potentially important
as sources of funds, staff, ideas, guidance, other resources, and turbulence.
The extent and kind of network to build, maintain, and use is a matter
largely under the control of the superintendent.10
Managing in the network, then, is an opportunity available to those
superintendents who recognize their interdependence and opt to try to
manage it actively. To network, in this context, is therefore a key strategic
option. Our model suggests that such managerial networking can contribute
to performance. What does the evidence show? To answer this key question,
68 Public management in interdependent settings

we simplify the model in order to focus on the relationship of most

importance here.
More specifically, for present purposes we build on an autoregressive
function, since we believe that incremental changes in performance are typical
for public programs and organizations. Adjustments to performance can be
caused by a wide array of forces, including myriad factors in the environment –
and also the efforts of managers. Although the general model is
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 ðXt =Se ÞðM3 =M4 Þ þ et
as explained earlier in the book, we are particularly interested here in
probing the externally oriented portion of the model. In particular, we want
to isolate the impact of M2 (where M2 ¼ M3/M4) – that is, public manage-
ment aimed at tapping and leveraging the opportunities presented by the
actors in the environment of the core unit, while also protecting the program
from hostile or disruptive forces. This means that the basic research issue to
be investigated has to do with the second or environmental term in the
general model, and in particular with the M2 specified in that term.
Rearranging terms in the equation yields the following:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 Xt M2 ð1=Se Þ þ et
The first term in an autoregressive model such as this would surely be the
dominant one in empirical settings: current performance can be expected
to be heavily driven by past performance. Our primary interest now,
nonetheless, is in the second term. We want to know whether exerting
managerial skill and effort in the interdependent environment matters for
program outputs and outcomes. If it does, the impact of M2 over more
extended time periods can be expected to be considerable, as it feeds into
output through each cycle (via the lagged dependent variable) and thus
amplifies its impact.
Two options for exploring the second term can be indicated, each simpli-
fying the general model in reasonable ways. One way would be to retain the
autoregressive feature as well as the key elements representing managerial
networking (M2) and environmental forces (X), while screening out the
other managerial functions as well as the structural stabilizing variable (S).
There is no prima facie reason to expect the other aspects of public manage-
ment within a given case to covary with the managerial networking function,
so the term M1 can be dropped from a simplified model.11 The S terms, as
well, can be omitted for present purposes. While the structural setting
matters, the empirical context we are examining consists of a set of
69 Managerial networking and performance

managerial cases – Texas school districts – highly similar in most basic

structural features. Not only are they all in the same policy sector, they are
also the same particular type of program setting, institutional design, and
managerial level: system superintendents. By selecting a set of cases in which
public managers confront structurally similar settings, structural variation is
minimized; and we can initially examine their strategic choices about how to
interact in the network without confounding impacts arising from widely
differing structural contexts. In later analysis we add structure back into the
discussion.
Accordingly, then, a simplified version of the model can be considered:
Ot ¼ b1 Ot1 þ b2 Xt M2 þ et
This equation is simply the one preceding it, after the removal of S and the
other managerial form, M1, as variables. The model in this form is clearly
underspecified (it leaves out some important determinants of performance),
but the simplification does allow for the testing of important components of
the general model.12 Using this version, we can test the proposition that
managerial networking matters for performance. We can also explore
whether and how networking managers deal with shocks or perturbations
from the environment (X) in fashions different from managers who do not
network. These are substantial advantages.
One disadvantage that this version carries, nonetheless, is that, given the
degree of dominance that the autoregressive term exerts in the model,
detecting the impact of M2, which appears in the much smaller second term,
may be difficult. For this reason, an empirical test involving this form of the
simplified model can be considered as a rather stringent one.
Another adjustment could also be useful to examine. This version drops
the autoregressive term altogether, in the interest of focusing on the environ-
mental impact itself. Thus:
Ot ¼ b2 Xt M2 þ et
Estimating this form means sacrificing some further explanatory power for
the purpose of conducting an empirical test more sensitive to the operation
of M2 and its interaction with a matrix of environmental forces.
Both versions are included in the analysis that follows. Despite the fact
that these forms of the model omit some influences on program perform-
ance, they are not themselves so simple. Each includes a particular kind of
public management contribution, and each specifies a nonlinear relation-
ship between management and the forces in the networked environment.
70 Public management in interdependent settings

Each, in other words, represents a more complex model than the linear,
additive versions more typical in multivariate analysis. A linear version of
the simplified model depicted in the last equation is
Ot ¼ b2 Xt þ b3 M2 þ et
Here managerial networking would make a difference, but not by interacting
with the set of environmental forces. Similarly, a linear version of the auto-
regressive simplification introduced above can be depicted in like fashion:
Ot ¼ b1 Ot1 þ b2 Xt þ b3 M2 þ et
Estimating these last two equations empirically would constitute a test,
generally speaking, of whether managerial networking matters; but the
simplified forms of our model – the two equations preceding these two –
must also be explored to check for the nonlinear relationships that, we have
argued, seem to be called for by the extant case study depictions.
The following hypotheses are the focus of the current investigation.
H1: managerial networking matters – in a positive direction – for program
performance. School system output/outcome is higher if superintendents
exert management effort in the networked environment surrounding them.
H2: managerial networking matters for how management relates to both
educational system inputs and environmental perturbations. In practical
terms, managers in networked settings deal with environmental shocks in
different ways from public managers who do not manage in networks. We
expect managerial networking to interact in a nonlinear fashion with the
vector of environmental forces to which school districts are exposed.
H3: the way that networking managers tap their surroundings is to exploit
opportunities and buffer impediments to program performance. That is, the
form of the nonlinear function can be expected to show managers
tapping resources in their networked environments to enhance program
performance. To the extent that environmental shocks challenge or
threaten program performance, managerial networking – by competent
managers – can be expected to protect the core performance bureaucracy
from these forces.
We conducted this empirical test of managerial networking and perform-
ance on a subset of Texas school districts. The 2000 superintendents man-
agement survey provided information about management styles, goals, and
time allocations (return rate of 55 percent). Of these, 507 responses were
usable in our analysis.13 We pooled five years of data (1995 to 1999) on
performance and control variables to produce a total of 2,535 cases for
71 Managerial networking and performance

Table 3.5 Factor loadings for managerial networking, 2000 survey data

Indicator Loading

Frequency of contact with school board members 0.60

Business leaders 0.73
Other superintendents 0.67
State legislators 0.68
Texas Education Agency 0.51

Note: Eigenvalue ¼ 2.07.

analysis. All data other than the survey were taken from the data sets of the
Texas Education Agency.
Along with the set of production function controls that were outlined in
Chapter 2, we must also include a measure of managerial networking in the
empirical analyses. In this case, our measure of managerial networking is
developed as described above – and with the original five nodes included in
the interaction pattern. We asked each superintendent to note how often
they were in contact with each of these others, on a six-point scale ranging
from daily to never. Superintendents inclined toward networking with the
key actors in the district’s environment should interact more frequently with
all five other sets of actors than should superintendents with a traditional
hierarchical (internally focused) management style. A composite networking
scale was created via factor analysis (see Table 3.5). All five items positively
loaded on the first factor and produced an eigenvalue of 2.07; no other
factors were significant.14 Factor scores from this analysis were then used as
a measure of managerial networking or M2, with higher scores indicating a
greater networking orientation.
Clearly, this measure is simplified. It ignores all aspects of networking
aside from frequency and direction – for instance, skill, reputation, and a
number of strategic considerations.15 Further, it taps a particular kind
of networking activity: interactions of managers in clusters of dyadic inter-
actions. Networks can range considerably in the extent to which they are
integrated and the degree to which all actors are directly linked to the full
range of others. Even so, the measure taps the effort managers choose to put
into managing externally, in the networked environment.16 Furthermore,
the factor-analytic results suggest that the notion of managerial networking
as a strategic choice is a coherent concept that makes empirical sense.17
Our measure of program output18 or performance (O) in the present
analysis is the percentage of students in each school district who pass
72 Public management in interdependent settings

state-required, standardized reading, writing, and mathematics tests each

year. For the period in question, the examination was the Texas Assessment
of Academic Skills. More details on this examination and measure were
provided in Chapter 2.
Our strategy of analysis here is to begin with relatively simple models and
build up to more complex variants that provide stronger tests of our theory.
We start with tests to determine whether management (managerial network-
ing) matters at all in the performance of these school systems (thus testing
H1), then move on to tests of whether the relationship of management to
performance is nonlinear, and, if so, how (H2, H3). A second, more strin-
gent, set of tests then takes place within an autoregressive model of program
performance.
We begin with the linear, additive relationship depicted earlier in one of
the equations:
Ot ¼ b2 Xt þ b3 M2 þ et
In this model the question is whether management matters when one
controls for the constraints and resources facing the school district (the
vector of X variables). The test of this model, handled via standard multiple
regression, appears in Table 3.6. The column designated “Base model”
contains all the X variables, and the next column (“Networkþ”) adds the
management measure to this equation. The X variables generally predict as
expected, with negative relationships for all constraints (percentage of black,
Latino, and low-income students), noncertified teachers and class size, and
positive relationships for the resource variables (with the exception of state
aid). When the managerial networking variable is added to this equation, it
produces a strong positive coefficient. Programs characterized by greater
managerial networking are programs that generate somewhat higher outputs.
Because this management variable is measured as a factor score (mean ¼ 0,
standard deviation ¼ 1), virtually the entire range of management falls
between þ3 and 3. This range suggests that networking by top managers
may contribute as much as four percentage points to a district’s pass rate, all
other things being equal.19 Although this variable is by no means the most
important factor in performance, changes of this magnitude are substantively
significant and well worth pursuing. By this test, H1 is supported.
Nonlinear relationships can be tested in a variety of ways. Our theory
suggests that management interacts with the resources and constraints in the
environment – that it exploits resources and mitigates constraints. One form
of that relationship was shown in this equation introduced earlier:
73 Managerial networking and performance

Table 3.6 Management and organizational performance: additive linear estimation

Dependent variable ¼ student exam pass rates

Independent variables Base model Networkþ

Managerial networking – 0.7035 (4.60)

Resources
Teacher salaries (000s) 0.4875 (4.49) 0.4665 (4.31)
Class size 0.3199 (4.83) 0.3117 (4.72)
Teacher experience 0.2048 (2.10) 0.1943 (1.90)
Noncertified teachers 0.1874 (5.28) 0.1873 (5.30)
Percentage state aid 0.0127ns (1.53) 0.0173 (2.09)
Constraints
Percentage of black students 0.2153 (13.35) 0.2167 (13.49)
Percentage of Latino students 0.1099 (10.43) 0.1091 (10.39)
Percentage of low-income students 0.1671 (11.12) 0.1670 (11.16)

R2 0.58 0.59
Standard error 7.65 7.62
F 294.96 276.07
N 2,534 2,534

Notes: T-scores in parentheses. Dummy variables for individual years not reported.
ns ¼ not significant.

Ot ¼ b2 ðXt M2 Þ þ et
The classic way to test this relationship is to compare the interactive form to
the linear form in the following equation:
Ot ¼ b2 ðXt M2 Þ þ b3 ðXt Þ þ b4 M2 þ et
The key test is whether the vector of coefficients b2 is statistically significant –
that is, whether it adds additional explanatory power to a linear model.
The problem with this model is that the interaction terms frequently
generate so much collinearity20 that individual coefficients cannot be
precisely estimated. The actual coefficients are important, because we have
specific hypotheses about how managerial action affects the environmental
variables – that is, it should increase the impact of resources and reduce
the impact of constraints.
To test these specific estimates, an alternative approach is necessary. We
divide the sample into two parts: districts with high reported networking by
top managers (those with scores above 0) and districts with low managerial
networking (scores below 0):
74 Public management in interdependent settings

Table 3.7 Management and organizational performance: nonlinear impacts

Dependent variable ¼ student exam pass rates

Level of management networking

Independent variables High Low

Resources
Teacher salaries (000s) 0.7727 (4.55) 0.2835 (1.98)
Class size 0.6620 (5.90) 0.1211ns (1.47)
Teacher experience 0.1256ns (0.85) 0.4556 (3.51)
Noncertified teachers 0.1100 (2.20) 0.2638 (5.30)
Percentage state aid 0.0189ns (1.39) 0.0073ns (0.69)
Constraints
Percentage of black students 0.1846 (7.20) 0.2291 (13.49)
Percentage of Latino students 0.1003 (6.18) 0.1147 (10.39)
Percentage of low-income students 0.1966 (8.53) 0.1537 (11.16)

R2 0.57 0.61
Standard error 7.60 7.61
F 124.23 176.83
N 1,154 1,380

Notes: T-scores in parentheses. Dummy variables for individual years not reported.
ns ¼ not significant.

Ot ¼ b2 ðXt Þ þ et M2 > 0
Ot ¼ b2 ðXt Þ þ et M2 < 0
The results of these models produced by splitting the sample are shown in
Table 3.7. The constraints can be examined first. Our theory suggests that, for
high levels of managerial networking, the size, or impact, of each of these
should drop in absolute value. Although the coefficients do not change
dramatically, in two cases (blacks and Latinos) the hypothesis is confirmed.
In one case (low income) it is not. In terms of the four cases of resources (state
aid is not significant and can be ignored), three are as predicted. Districts with
more networking on the part of superintendents get more out of teacher
salaries and reductions in class size and are less affected by noncertified
teachers. Teacher experience shows an interesting pattern of significance for
the low-networking districts and insignificance for the high-networking
districts. While this result ostensibly contradicts our hypothesis, it means that
high-networking districts are not affected by having less experienced teachers.
Five of the seven relationships found in Table 3.7 are consistent with our
nonlinear, interactive theory of management. Although this pattern might
75 Autoregressive models

not seem like strong support for the theory, examining the individual
coefficients provides additional corroboration. Most of the relationships
differ from each other in only marginal ways. The differences between four
sets of the relationships are substantial, however. Districts characterized by a
high managerial networking style get 2.7 times the impact from higher
teacher salaries, receive 5.5 times the impact from smaller classes, get only
42 percent of the negative impact of noncertified teachers, and are
not affected at all by inexperienced teachers.21 Even with the relatively
crude test presented here (a simple measure of managing in the networked
environment – a deliberately underspecified model), therefore, management
does matter; and it matters by interacting with program resources and
constraints in predicted directions. H2 (nonlinearity) and H3 (direction of
relationships) are supported by these tests.
The pattern of relationships merits some additional comment. Manage-
ment is about choice and decision making. Quite clearly, managers allocate
more time and effort to some constraints and resources than to others. As a
result, expecting all resources to become more valuable and all constraints to
become less negative may not only be expecting too much, it might also
conflict with what the manager is trying to do. In other words, the results
may be evidence that managers make strategic networking choices beyond
the fundamental ones of “How often?” and “With whom?” An effective
manager might well focus on a small number of strategic factors that can
be manipulated to get better results, while at the same time accepting some
modest negative tradeoffs on less important variables. The relationships in
Table 3.7 are consistent with such an interpretation. The negative findings
are relatively small, as are a few of the positive findings. Three of the impacts
are substantial – those regarding teacher salaries, class size, and noncertified
teachers. Getting large positive results on these three variables more than
compensates for the modest negative changes on other factors.

Autoregressive models

Because organizations stress standard operating procedures, specialization,

and consistency, they tend to be relatively predictable and stable from year to
year. They are, as we have explained, autoregressive systems. A stronger and
more difficult test of our theory of management in the networked environ-
ment involves moving to an autoregressive model in which current perform-
ance is determined in part by past performance. Again, our strategy of
76 Public management in interdependent settings

Table 3.8 Management and organizational performance: autoregressive and nonlinear models

Dependent variable ¼ student exam pass rates

Independent variables Base model Networkþ SWAT

Managerial networking – 0.1719 (1.65) –

Performance (t – 1) 0.7172 (63.65) 0.7162 (63.48) 0.7042 (63.49)
Resources
Teacher salaries (000s) 0.3679 (6.75) 0.3704 (6.80) 0.4520 (8.63)
Class size 0.0752 (1.69) 0.0750 (1.69) 0.1913 (4.08)
Teacher experience 0.1448 (2.38) 0.1526 (2.50) 0.1285 (2.10)
Noncertified teachers 0.0947 (3.95) 0.0947 (3.95) 0.1159 (5.30)
Percentage state aid 0.0074ns (1.39) 0.0064ns (1.21) 0.0049 (0.89)
Constraints
Percentage of black students 0.0586 (5.19) 0.0593 (5.25) 0.0521 (4.48)
Percentage of Latino students 0.0412 (5.76) 0.0413 (5.79) 0.0475 (6.53)
Percentage of low-income students 0.0165ns (1.62) 0.0165 (1.69) 0.0135 (1.27)

R2 0.81 0.81 0.82

Standard error 5.18 5.18 2.11
F 1187.17 1069.46 1296.65
N 2,534 2,534

Notes: T-scores in parentheses. ns ¼ not significant. Classical statistical significance does not apply to
SWAT models.

analysis will be the same. First, we examine whether management matters at

all; and then we examine whether or not the impact is nonlinear. The
autoregressive model analogous to the linear model just estimated – one
treating management as simply another additive input – is
Ot ¼ b1 Ot1 þ b2 Xt þ b3 M2 þ et
The basic linear model without management is shown in the first column of
Table 3.8. The lagged dependent variable dominates the equation, thus
strongly supporting our notion that such organizations are indeed inertial
systems. All the same, the parameter estimate (just above 0.7) remains at a
distance from 1.0, thus indicating that the past does not rigidly determine
current performance. The constraints remain negative, though with dimin-
ished impact; similarly, the resources remain positive but with smaller
impacts. The lagged dependent variable essentially limits the influence of
these variables to their impacts on changes from year to year, and this short-
term impact by definition must be smaller than a one-shot estimate of
impact on a cross-section. Despite the stringent nature of this test,
77 Autoregressive models

management continues to have a positive impact on performance. For the

full range of this variable, management could make a difference of as much
as one percentage point per year on the pass rate. While this may not appear
to be substantial, with the autoregressive model these impacts continue to
affect future performance for several years into the future (see Pindyck and
Rubinfeld 1991 and O’Toole and Meier 1999).22 By this stringent test,
therefore, H1 is again supported.
Assessing the nonlinear impacts in an autoregressive model is somewhat
more difficult, given that the autoregressive term so dominates the equation.
The result is that the remaining coefficients are often less stable, so that a
standard interaction often has too much collinearity, and dividing the sample
could leave too little variance to provide efficient estimates. We do explore the
issue via sample splitting (to follow shortly), but we also utilize an alternative
approach: the substantively weighted analytical technique (SWAT), a form
of exploratory data analysis that allows one to focus on interesting subsets of
data (Meier and Gill 2000). We first employ SWAT to examine the question
of nonlinearity, and then we explore the question with sample splitting.
Applied to the current case, the logic undergirding SWAT is as follows. If
management has a nonlinear/interactive relationship with forces and inputs
from the environment, then those units characterized by high levels of
managerial networking should operate with a different set of relationships
from the average organization. Hypothetically, this suggests that, if one
replicated the analysis reported in Table 3.8 with a sample of units but only
had those with high levels of networking by superintendents, the regression
coefficients would change.
To get at this hypothetical situation, SWAT asks what would happen if the
population of units were to contain many more organizations with high
levels of managerial networking and many fewer with low levels of such
behavior by top managers. SWAT creates such an artificial universe by
reweighting cases in the existing sample. Comparing the regression from
this sample from a hypothetical population to the sample from the existing
population should provide some leverage on whether externally oriented
management matters more or in different ways in these sets of situations. In
the specific case, we designated those school districts with networking scores
above 1.36 (about 10 percent of the total) as having a high level of such
activity by management. We used a higher threshold in this situation rather
than just the top half, because the autoregressive specification is likely to
wash out relatively small differences in management activities. School dis-
tricts with a managerial networking score below 1.36 were weighted at 0.1
78 Public management in interdependent settings

compared to weights of 1.0 for those above this threshold. This process
artificially creates a sample that has only one-tenth as many districts in the
low category and ten times as many (relatively) in the high managerial
networking category. The results of this weighted regression are shown in
column 3 of Table 3.8.
The coefficients in column 3 are not parameter estimates (since they deal
with a hypothetical universe), but they are informative when contrasted with
the ordinary least squares (OLS) regression coefficients because they show
how a set of units with a higher degree of networking by managers might use
resources differently. The previous results in Table 3.7 suggest that teacher
salaries, noncertified teachers, and class size are likely to be the key variables.
Most of the SWAT/OLS differences are relatively small, but in two cases –
teacher salaries and class size – the differences in coefficients are substantial.
In both cases the implication is that programs with more managerial
networking get more out of their resources. The SWAT equation stressing
high levels of management has a teacher salary coefficient approximately
24 percent larger, and the class size coefficient is 254 percent larger than that
for the same variable in the base OLS regression.
The consistency of these two relationships with the same relationships in
the nonautoregressive model in Table 3.7 is reasonable evidence that one
aspect of management is likely to interact with program inputs from the
environment to produce outputs above what would be expected in a strictly
linear relationship. Again, H2 and H3 are supported even in the estimations
of an autoregressive model.
Finally, for even more evidence on this question, we undertake another
analysis by splitting the sample in a couple of fashions. Table 3.9 divides the
school districts into five quintiles by level of performance on the TAAS. The
top quintile, for example, has a mean student pass rate of 79.2 compared to
73.9 for all districts and 64.8 in the lowest quintile. Care must be exercised in
partitioning a sample, particularly when partitioning on the dependent
variable, because each subset is designed to be unrepresentative of the entire
sample. The prediction levels in Table 3.9 increase dramatically compared to
the simple linear, autoregressive estimation in Table 3.8; in the middle three
quintiles less than 1 percent of the variance is left unexplained.23
Quite clearly, we might have some intuition as to why networking by
managers might matter more or less as organizations perform better. The
regressions in Table 3.9 show that networking’s impact on performance is
relatively stable in the middle three quintiles; these estimates are also statis-
tically more reliable than those in Table 3.8. For both the highest- and
79 Autoregressive models

Table 3.9 Managerial networking at different levels of educational performance

Quintiles of performance: 5 ¼ best

Independent variable 5 4 3 2 1

Past performance 0.6645 (50.70) 0.7189 (187.09) 0.7150 (201.04) 0.7201 (174.85) 0.7330 (42.78)
Managerial networking 0.2792 (2.05) 0.1579 (4.70) 0.1626 (5.99) 0.1766 (5.26) 0.3727 (2.18)

R2 0.91 0.99 0.99 0.99 0.92

Standard error 2.93 0.75 0.60 0.76 3.71
F 485.13 9,253.65 13,210.42 10,394.95 485.13
N 518 531 484 522 519
Mean dependent
variable 79.2 77.5 75.8 71.5 64.8

Notes: T-scores in parentheses. All equations control for the eight control variables used in Table 3.7.

lowest-performing organizations, the management coefficient is much

larger.24 Why might this be the case?
In a smoothly running organization that is attaining adequate perform-
ance, the demand for creative management and the opportunities to use
that management might be relatively few. An organization interested in
optimizing rather than satisficing (or, alternatively, one seeking to change
its level of performance dramatically) is more likely to seek out opportun-
ities to exploit inside the organization or in its environment. An aggressive
superintendent in this regard might seek a larger bond issue for capital
expansion, try new programs for parental involvement, or use traditional
resources in nontraditional ways. Seeking higher levels of performance
relative to environmental constraints (as these models are set up) requires
taking more risk, and management efforts (networking) and skills should
come more into play.
For those units at the low end of the performance scale, the function of
managerial networking is probably somewhat different but also equally
important. These organizations are performing poorly, and that perform-
ance is likely recognized by both the district and various actors in the
district’s environment. In a poorly running unit, perhaps, almost any
improvement will get some returns. Good external management in such a
situation is likely to matter more, because it compensates for inadequate
processes and decisions in other parts of the core organization. Such leader-
ship could also have a salutary impact on internal morale as members see
actions being taken that could improve the organization.
80 Public management in interdependent settings

Table 3.10 Managerial networking interactions with resources and constraints

Level of managerial networking

Independent
variable All 1.0 1.25 1.5 1.75 2.0

Past performance 0.7172 0.6726 0.6431 0.6302 0.5942 0.4732

(63.65) (24.08) (17.00) (13.92) (11.34) (5.65)

R2 0.81 0.80 0.84 0.82 0.80 0.77

Standard error 5.18 5.17 4.37 4.51 4.45 4.54
F 1,187.17 199.09 148.33 101.87 70.94 31.82
N 2,534 450 260 205 165 95
Mean dependent 73.9 74.6 75.5 76.6 76.4 77.3
variable
Mean managerial 0.0 1.59 1.91 2.04 2.14 2.35
networking

Notes: T-scores in parentheses. All equations control for the eight control variables used in Table 3.7.

The autoregressive coefficient representing past performance also merits

comment. As organizational performance increases, the size of the autore-
gressive parameter decreases; the parameter in the top quintile is statistically
smaller than the estimate for all organizations. This finding suggests that
high-performing organizations are less constrained by past performance than
are organizations with weaker performance. Because a networking-style
management interacts with the environment, as our model expects, this
pattern is consistent with tapping opportunities externally and, in the process,
reducing organizational rigidity.
This relationship reveals a paradox of organizational management. At the
highest levels of performance, stability is a good thing. As performance in an
organization declines, stability has less value, simply because the organiza-
tion is reproducing poor performance. The results of Table 3.9 suggest that
stability is greatest exactly when stability is of the least value to the
organization.25
Table 3.10 looks at the same pattern but in a different way; it presents the
results for the districts run by superintendents who rate highly on the
networking variable. Since we are interested in probing what happens when
managers undertake frequent and extensive networking, we focus on the
high-networking cases for special attention here. Subsets of the sample that
include larger networking values can be compared with the full set of cases
81 Interpretation and implications

(column 1). The first subset (column 2 in the table) includes only superin-
tendents with networking scores above 1 (or one standard deviation above
average) – about 18 percent of all managers. Subsequent regressions, listed in
successive columns of the table, raise this standard by 0.25 standard devi-
ations in a series of steps until only the top 4 percent remain (those scoring
above 2.0). This incremental process of examination illustrates how the
relationships evolve at different levels of networking activity. Because we
are selecting progressively less representative organizations, our interpret-
ation should be cautious and avoid assessing patterns where the relation-
ships are not strong. These findings contain all the previous controls, but we
show only the relevant coefficients.
As networking by top managers increases, the autoregressive term
declines slowly, until the management variable is 1.5 standard deviations
above the mean, and then precipitously. This pattern suggests that net-
working performs its desired function; rather than being trapped by past
routines and behaviors, well-networked managers generate more flexibil-
ity for their organizations to change. This link should be viewed as the
first step in a two-step process of managing the organization overall: first,
exploiting the environment to create change in the unit; and, second,
then structuring the changes to produce higher performance.26 Changing
the size of the autoregressive component in the model dramatically
changes the long-run impact of other variables, because the current
values of the independent variables will continue to affect performance
in the future by feeding back through the autoregressive term. The
finding suggests, therefore, that the influence of managerial networking
ramifies forward into the future and can enhance performance substan-
tially in the longer term.

Interpretation and implications

Thus far this analysis has shown that managerial networking is related to
performance, but we have not fully demonstrated the process by which
managing in the networked environment generates better results. We have
several hypotheses. First, management’s greater attention to the environment
might create buffers from external shocks and thus permit lower-level per-
sonnel (teachers, principals) to be more effective. Second, a networking style
might encourage a more decentralized internal management approach – an
approach advocated by much current education reform literature. Third, the
82 Public management in interdependent settings

networking might expose the superintendent to innovative programs oper-

ated by other districts. Fourth, managerial networking might convince exter-
nal stakeholders to grant more autonomy to the school district and thus allow
the district to exploit the expertise it has. In all four cases, organizations might
be able to use resources more effectively.27 These and other hypotheses for the
linkage between networking by managers and performance are all plausible;
future research can be designed to evaluate these hypotheses. Indeed, some of
them are partially explored later in this book.
The evidence here indicates that public managers need to consider
networking outward an important tool of administrative success, not merely
a luxury in which to engage if there is extra time. Networkers in our sample
spent less time running internal operations than did others, but the tradeoff
paid off in results. Nevertheless, managerial networking may not be an
unmitigated good. Its contributions to performance may see diminishing
returns at higher levels of networking. We explore this issue later in the
book. In addition, networking may assist certain goals, and certain stake-
holders, more than others. We tackle this subject next.

Distributional consequences

So far, we have seen that managerial networking is a phenomenon that can

be discerned in various sorts of jurisdictions, that it helps to shape perform-
ance, and that it interacts with selected constraints and resources – the “X”
vector – in the interdependent environment in so doing. Consistent with
these findings, as implied by our model and as framed in the research
literature in the field (and from which the model is developed), the theme
of networks, networking, and public management has had a rather positive
aura. Networks can assist program and policy delivery, networking can assist
performance, and more of both – it is often implied – is a good thing. The
bulk of this literature frames the emergence of networks in terms of a
tendency or necessity to use multiple linked social actors, often multiple
organizational actors, to achieve collective purposes. Corollary attention,
unsurprisingly, has been directed at logically related issues, such as how to
manage networked arrays, how to measure and improve performance in
networked settings, and how to understand network operations through
empirical theory. With this attention to networks and management has
come an implicit notion among researchers that network development,
use, and performance are topics that carry little direct political import –
83 Network management: the functionalist perspective

aside from the obvious point that the performance of networks might itself
be of interest to a broader public.
The inadvertently depoliticized analysis of networks in recent research has
neglected issues that should be part of the research agenda, however. In this
section we outline ways that networks and network management point
toward significant political issues.28 We then focus on one political dimen-
sion of networks and their performance: the likelihood that, rather than
being neutral producers of collective goods while enmeshed in a broader
environment, networking managers respond to the stronger and more
politically powerful elements of their surroundings, thus magnifying the
tendency toward inequality already present in the social setting. This
dynamic – what we call the “dark side” of managing networks – has been
largely unexplored by network researchers. Such patterns should not be
unexpected, however. The reasons are explicit in longstanding streams of
research that have been ignored in the work done thus far on networks. We
report some empirical results that give considerable credence to the dark
side hypothesis. In so doing, we argue that there is a need for systematic
study of the political aspects of networks and their management.

Networks and network management: the functionalist perspective

The standard portrayal of networks attributes the multiactor features of

program implementation and management to the demands placed on pro-
grams and their administrators. Among the causal factors frequently men-
tioned as drivers of networked program execution are the increasingly
“wicked” character of public problems (Rittel and Webber 1973), the real-
ities of increasingly dense program environments, the expertise-reliant char-
acter of modern governance, the requisites of program design in multilevel
systems, and the demands placed on program managers in complex
settings. Although each of these arguments has a political dimension, the
production-focused and partnership-framed perspective obscures political
themes with their distributional aspects, instead emphasizing the managerial
requisites generated in and for such arrays.
The theoretical claims, parallels, and distinctions among these strands of
causal logic represent a complex and somewhat confusing pastiche. The
point to be emphasized, however, is that the political interpretation of
networks, in terms of their likely causes and consequences, seems largely
lost in the analytical picture. Researchers seem to buy into a production
84 Public management in interdependent settings

logic of one sort or another regarding network formation and operations,

and one result is a blindness toward the distributional consequences of
network actions.
The point can be put another way. The bulk of research on networks and
public management effectively reenacts a network version of the venerable
politics/administration dichotomy. This statement holds in two respects.
First, instrumental logic is used to explain network patterns, typically with
an emphasis on program or clientele needs. This theme gives little attention
to certain political drivers of network formation and use that have little to
do with program needs and more to do with incentives that can operate on
political leadership. Second, researchers typically ignore important political
issues about what networks do, how they perform, and how they can be
directed toward goal achievement. The modal study of networks and public
management recognizes that program results matter for stakeholders. But
these results – the dependent variables tapping performance – are treated
in a rather sterile fashion, as products of a production system, without
attention to distributional aspects or contest between stakeholders. Instead,
such studies emphasize management, facilitation, coordination, and related
themes. The politics of network performance, in several relevant respects, is
virtually ignored.
From other literatures, however, we can sketch three political themes
regarding network-associated impacts. One in particular, the last-mentioned
below, admits to systematic exploration.
First, while some recognize that additional actors are often needed during
implementation to build support for program operations (Pressman and
Wildavsky 1984), researchers have not considered the possibility that the use
of networks can also be a way of distancing state actors from controversial
policy efforts. The choice of networks can be a function not of increasing
problem-solving capacity but of authoritative actors dodging difficult or
costly responsibilities. Networks can be a symbolic-political choice when
there is pressure for state action yet disincentives for the state to address
policy problems definitively. This side of the network issue has been absent
from systematic investigation, but it is likely involved in the design of
institutional arrangements for addressing such policy issues as HIV/AIDS,
family planning services, and some aspects of social welfare policy.
Second, networks can have another political effect that has typically been
ignored in the research literature: the incorporation of additional perspec-
tives or constraints that shift the policy emphasis during implementation.
One way this result can develop is through the dynamics of coproduction.
85 Network management: the functionalist perspective

While it is generally recognized that adding actors increases constraints as

well as opportunities, network research has not systematically explored the
ways that coproduction can shift the goals and preferences of public pro-
grams. Instead, the challenge stemming from the addition of network actors
has largely been framed in terms of rendering the pattern less easily man-
aged. The emphasis has been on the complexity of coproduced effort – a
coordination problem – rather than the potential shift in the core of what
public programs managed through networks actually do. “Adequate man-
agement” (Kickert, Klijn, and Koppenjan 1997: 9) is seen as the challenge –
one that is best met by more energetic and more talented managerial efforts.
Once again, the emphasis is on an instrumental rather than a political point.
Adding actors does more than complicate, however; it tilts the balance of
power. Determining the scope of involvement shapes the definition of issues
and goes a long way toward determining who wins and who loses on policy
questions (Schattschneider 1960).
The third way that the addition of network actors can carry political
import is through straightforward political pressure. Here the table is tilted
again. In this variant, even if production occurs primarily through a core
organization, other network parties influence the pattern toward a skewed
distribution of program results. In short, a bias in performance can derive
both from coproduction as well as from the dynamics of managerial
response to pressure from network actors as a core organization responds
to its networked environment.
The facts of life regarding public management in a political environ-
ment are hardly new to analysts of the twenty-first century. Decades of
research have validated the point that agencies and their management
must develop support in their setting, and that doing so can mean
sacrificing the primary agenda of policy, particularly if it involves social
change, in the interests of survival. For instance, Philip Selznick’s classic
study, TVA and the Grass Roots (1949), defines and illustrates the notion
of cooptation with vivid exactitude. Cooptation and the difficult tradeoffs
it implies have been staples of the analysis of public management and
bureaucratic politics for a considerable period. Curiously, however, these
basic facts of life seem to have been largely forgotten by enthusiasts of the
network perspective. The more public programs are designed to alter the
existing order, the greater the threat of the program to those who benefit
most from the status quo. A result is heightened emphasis on capturing
benefits of the program during execution by those who are best pos-
itioned to shape the details of program implementation. By design,
86 Public management in interdependent settings

moreover, networks are leaner and weaker in the face of larger institutions
and significant individual actors in a policy system.
One way of explicating the point has to do with networks, network nodes,
and the pattern of exchanges that can be so important in facilitating network
action. Virtually all assessments of public management patterns recognize
that networks are built around exchanges between the nodes in the network,
often with managers framing and brokering the exchanges. An exchange
implies that node A provides something to node B, and vice versa, in such a
way that the overall aggregation is better off. This positive-sum view of
networks and networking overlooks the fact that each node enters the
network with a distinct set of goals. Only a portion of these goal sets overlap.
Despite the extensive literature on cooptation, the ability of network nodes
to shape the direction of public programs has not been carefully investi-
gated. For public organizations that seek multiple goals – that is, all public
organizations – the risk is that network interactions will emphasize some
goals to the detriment of others. The literatures on interest groups and on
citizen participation indicate that network nodes seek greater benefits for
goals that are favored by more entrenched interests and downplay efforts
that favor disadvantaged clientele.

An empirical test

This last-mentioned aspect of the network politics of program management

is amenable to systematic analysis. Do the benefits of managerial network-
ing, documented earlier in this chapter, accrue disproportionately to those
who already have more than others? Public school systems are an ideal
setting to test this notion, because they display a wide variety of goals and
can sit within networked settings. Because schools seek goals that benefit
different races and social classes differently, and because networks are more
likely to be populated by actors and organizations that already possess
political resources, particularly at the critical loci of such networks, our
working hypothesis is that managers who expend greater effort in working
the network will improve educational performance more for goals that
benefit their relatively advantaged clientele than for goals that benefit their
disadvantaged clientele.
We used data for five years (1995 to 1999) on the performance of Texas
school districts and the set of control variables introduced in Chapter 2. We
also used the survey responses we have collected from top managers. The
87 An empirical test

total possible number of cases for this analysis is 2,535. To perform the
analyses, we used a simplified version of our model, which asserts a positive
relationship between managerial networking and performance, controlling
for a set of resources and constraints operating on the organizations. We
already knew that managerial networking improves performance; but we
also wanted to explore how such positive impacts vary across performance
measures that refer to, or are salient for, different school system constituencies
that provide part of the networked environment for the core educational
organizations. To investigate these questions, then, we used our measure of
managerial networking (we used the original five-node factor score for this
purpose), as well as the sets of suitable performance measures and appropriate
control variables introduced earlier.
The basic hypothesis is that networking contact will contain biases that
have distributional consequences for the performance of public organiza-
tions. In this case, because we know that participation and interest group
action is positively correlated with socioeconomic status, superintendents
who network are more likely to be exposed to portions of their networked
setting that will seek benefits for the better off or higher-status students
rather than for disadvantaged students. We would expect the networking
measure, therefore, to be positively correlated with test scores for Anglo
students, with ACT test scores, SAT test scores, and the percentage of
students who exceed the college criterion on these tests (1,110 on the SAT
or its ACT equivalent). We would not expect significant positive relation-
ships for those indicators that reference the performance of disadvantaged
students: TAAS pass rates for black, Latino, and low-income students,
attendance rates, and dropout rates. These hypotheses are supported by
the interest group/participation literature (Zeigler and Peak 1972; Verba
and Nie 1972; Salisbury 1984; Scholzman 1984), the urban services literature
(Lineberry 1977; Mladenka 1980; Jones 1985), and a substantial literature in
education policy (see Tyack 1974, Bowles and Gintis 1976, Kozol 1991, and
Meier and Stewart 1991).
Regression estimations were developed for each of the ten performance
indicators outlined above. The specification includes all control variables
plus the measure of managerial networking. Dummy variables for each year
were also included. These were usually jointly significant, reflecting an
upward trend in the performance data during this period.29
The last column of Table 3.6 displays the results for the overall TAAS pass
rate performance. As explained earlier, the adjusted R-squared is approxi-
mately 0.59, indicating a reasonable amount of explained variance.
88 Public management in interdependent settings

Table 3.11 The impact of network interaction on disadvantaged student indicators

Performance measure Networking slope T-score R2 N

Latino pass rate 0.4081 1.56 0.36 2,310

Black pass rate 0.2437 0.64 0.37 1,568
Low-income pass rate 0.1168 0.61 0.51 2,518
Dropout rate 0.0424 2.04* 0.16 2,514
Class attendance 0.0028 0.18 0.24 2,534

Notes: All equations control for teacher salaries, percentage of state aid, class size, teacher
experience, percentage of teachers not certified, percentage of black, Latino, and low-income
students, and yearly dummy variables. * ¼ significant at p < 0.05, two-tailed test.

Table 3.12 The impact of network interaction on advantaged student indicators

Performance measure Networking slope T-score R2 N

White pass rate 0.8097 5.31* 0.42 2,506

Average ACT score 0.0670 2.50* 0.38 2,220
Average SAT score 5.0762 3.49* 0.50 1,836
Percentage above criterion 0.5512 2.80* 0.30 2,416

Notes: All equations control for teacher salaries, percentage of state aid, class size, teacher
experience, percentage of teachers not certified, percentage of black, Latino, and low-income
students, and yearly dummy variables. * ¼ significant at p < 0.05, two-tailed test.

Relationships are all in the expected directions and also significant. Of

particular interest is the impact of managerial networking, which – as we
have seen – shows a positive relationship to TAAS scores. The maximum effect
size for this variable is more than four points on districts’ overall pass rate.
Such an impact – particularly from the top position in the system, one far
removed from the core of the educational process – can be seen as substantial,
and the impact of management can show impressive results over time.
For the most salient performance indicator, managerial networking con-
tributes to positive results. This result fits with the expectations developed
from the research tradition of Selznick and others. How does networking
effort play out across the range of performance measures? We ran nine
additional regression analyses to determine the answer; the results for those
targeting performance for the relatively disadvantaged parts of the educa-
tional constituency appear in Table 3.11, while those measures of interest to
more powerful parts of the public school networked environment are
reported in Table 3.12. Both tables provide summary reports: they include
89 An empirical test

the results for managerial networking but omit the portions of the estima-
tions pertaining to the controls.
Table 3.11 summarizes five direct tests of the hypothesis by showing
managerial networking’s impact on five performance indicia that matter
most to minority constituents, the poor, and/or low performers. The pattern
is striking. For each of Latino students, black students, and low-income
students,30 managerial networking does not add to performance with any
statistically significant impacts. The same can be said for attendance. All
these performance measures are of more interest to marginalized constitu-
encies of the school system network. Only for dropout rates does managerial
networking seem to matter. This anomaly may result from the poor quality
of the dropout data, however. Data on dropout performance are the least
reliable of those analyzed in this study.
Table 3.12 shows the contribution of managerial networking to perform-
ance for four indicators relevant to advantaged (that is, top-end and/or
Anglo) students. For all, the impact of managerial networking is clearly
positive and significant. This is what one would expect if managers engaged
in the network are influenced by and attentive to what those with power
would prefer. For the Anglo pass rate, average SAT, average ACT, and
percentage of SAT above 1,110 (or its ACT equivalent), managerial network-
ing adds to performance. These are all indicators that are of considerable
interest to relatively influential or privileged constituencies.
The results overall are clear: the estimations retrieve Selznick’s insight
with detailed findings. Those parts of a networked constituency that are
influential and care about the performance results have managerial network-
ing assisting what they do; those parts dealing with more marginal or less
salient issues are less – or not – influenced by managerial networking.
Selznick’s argument is strongly supported by the findings. It is worth noting
that here, as well as for the results to be discussed next, the findings represent
distinct impacts from networking, not simply distributional inequities gen-
erated in school districts. In other words, the results indicate a set of
systematic relationships between networking and distributional impacts.
Networking actions generate greater inequalities than the school system
would have without network activity.
Further insight as to what is likely occurring in these settings can be gained
by taking a more thorough glimpse inside the managerial networking activity
reported by the district superintendents. To do so, we replace the overall
networking factor scores with the reported degree of networking, respectively,
with each node. We enter each node or networking partner into separate
90 Public management in interdependent settings

Table 3.13 The impact of network interaction on advantaged student indicators

Interactions with TAAS Anglo tests ACT SAT Criterion

School board members 0.589 (3.39)* 0.572 (3.33)* 0.011 (0.37) 0.87 (0.52) 0.200 (0.89)
Local business leaders 0.268 (1.72)# 0.450 (2.92)* 0.104 (3.71)* 8.18 (5.48)* 1.093 (5.49)*
Other superintendents 1.011 (5.77)* 0.974 (5.62)* 0.013 (0.42) 1.20 (0.71) 0.037 (0.16)
State legislators 1.504 (4.21)* 1.170 (4.73)* 0.056 (1.29) 6.33 (2.70)* 0.398 (1.24)
Texas Education Agency 0.631 (3.15)* 0.569 (2.88)* 0.061 (1.72)# 1.97 (1.05) 0.256 (0.98)

Notes: T-scores in parentheses. All equations control for teacher salaries, percentage of state aid, class size,
teacher experience, percentage of teachers not certified, percentage of black, Latino, and low-income
students, and yearly dummy variables. * ¼ significant at p < 0.05. # ¼ significant at p < 0.10. N ¼ 1,110.

Table 3.14 The impact of individual network nodes on disadvantaged student indicators

Performance measure

TAAS tests for

Interactions with Blacks Latinos Low-income Attend Dropout

School board members 0.184 (0.44) 0.059 (0.20) 0.746 (3.43)* 0.117 (6.97)* 0.049 (2.15)*
Local business leaders 0.768 (1.95)# 0.500 (1.91)# 0.535 (2.75)* 0.025 (1.66)# 0.016 (0.80)
Other superintendents 1.180 (2.82)* 0.850 (2.91)* 0.837 (3.80)* 0.078 (4.51)* 0.057 (2.49)*
State legislators 0.975 (1.59) 0.419 (1.01) 0.307 (0.98) 0.009 (0.38) 0.060 (1.81)#
Texas Education 0.556 (1.14) 1.040 (3.11)* 0.733 (2.94)* 0.038 (1.93)# 0.027 (1.03)
Agency

Notes: T-scores in parentheses. All equations control for teacher salaries, percentage of state aid, class size,
teacher experience, percentage of teachers not certified, percentage of black, Latino, and low-income
students, and yearly dummy variables. * ¼ significant at p < 0.05, two-tailed test. # ¼ significant at
p < 0.10, two-tailed test.

regression analyses that are otherwise specified identically to those performed

for the overall networking measure. We explore the impact of interaction with
each node on each of the ten performance measures already reported. These
additional analyses amount to fifty estimations – ten each for each of the
nodes (school board members, local business leaders, and so forth). The
results of these analyses are reported in Tables 3.13 and 3.14. The tables omit
the findings for the controls in favor of reporting only on the impacts of each
of the networking contacts of the school district top managers. Table 3.13
summarizes the findings for the all-pass rate (the most highly salient per-
formance measure) and the advantaged student indicators, while Table 3.14
provides a parallel set of results for the disadvantaged student indicators.
Although we do not have measures of the goals of each of the nodes, in a
few cases clear expectations can be inferred. Local business leaders are likely
91 An empirical test

to push for improvements at the elite end of the educational spectrum since
their own children are likely to be relatively advantaged in the education
system. The Texas Education Agency is most associated with its exam, the
TAAS, and it sets standards for students by race and ethnicity. Other
superintendents are likely to reflect professional interests, and professional
educators in the United States are likely to push education benefits or have
ideas for new programs that affect both haves and have-nots. The exact
preferences of the political actors – that is, school boards and state legislators –
will depend on the composition of their constituencies, and systematic data
on the “electoral” constituencies – that is, who voted for the office holder – are
not available.
The analyses provide some hints as to what may be going on as top
managers of the school districts interact with their environment. While
not definitive, the results suggest possible causal links and, thereby, plausible
production processes. For business leaders the pattern is especially clear:
such contacts help on every measure of advantaged student performance
tested and hurt on four out of five measures of disadvantaged student
performance.31 In this case, cooptation is a likely explanation. More contact
with business leaders probably exposes top school-district managers to the
complaints, concerns, and preoccupations of the local business elite, from
whom some support (for instance, for the district’s revenue-raising agenda)
may be crucial. To the extent that superintendents use their discretion to
direct or redirect attention to these matters, some sacrifice to the more
marginalized clientele may follow.
Superintendents’ interactions with their counterparts in other districts
contribute to performance on seven of the ten measures, including on all five
measures tapping disadvantaged students’ results. What is likely happening
here is information sharing and professional assistance to colleagues, thus
suggesting that collegial professional interaction can be a route for diffusion
of innovations and relatively equitable performance boosts across organiza-
tions and governments. Interaction with the Texas Education Agency is also
helpful to several measures of performance: three each for the advantaged
and disadvantaged groups. Most of these have to do with performance on
the TAAS, a subject of obvious concern at the state level. This pattern is
consistent with what one might expect from interaction with a regulatory
agency – which is, in effect, how the TEA operates.
The more intriguing results are those for the other two external links for
the superintendents. Contacts with school board members do not help
performance, and for half the measures more networking with school board
members impedes performance. The negative impacts are spread across both
92 Public management in interdependent settings

advantaged and disadvantaged students. Whatever is going on in these con-

tacts, the results do not seem to be aspects of cooptation in the usual sense. For
instance, TAAS results for both Anglo and low-income students are negatively
associated with more contact with the school board. Somewhat surprisingly,
contacts with state legislators show some impacts; these are all positive with
regard to performance, and three of the four significant impacts show up on
measures tapping advantaged student or generally salient measures.32
These findings do not fully demonstrate what is happening as managers
engage in networking activity with an array of external parties. They do show,
however, that the benefits of this activity are unevenly distributed, and also
that these consequences might be traced to contacts with particular actors.
Networking with the external world can offer perils as well as prospects, and
understanding the political and distributional dimensions of such settings
can help to explain what is likely to be produced via networked public action.

Implications

Network researchers have appropriately emphasized the complex and inter-

dependent nature of many of today’s public programs and pointed to the
challenges faced by public managers who are responsible for concerting
policy-relevant action. In implicitly (or otherwise) suggesting that the issues
are those of coordination and management alone, however, much of the
recent exploration of networks and policy implementation ignores poten-
tially crucial political dimensions of network creation, coproduction, and
cooptation. This chapter indicates that these omissions are important and
that systematic research on the political aspects of networks and their
performance impacts is needed.
The last portion of the analysis in particular makes a strong case that
networks and their management are not likely to produce leveraged perform-
ance without distributional implications. In a sense, these results validate a
venerable theme. As Selznick argued decades ago, administrative units situated
in an interdependent political environment must find ways to build support –
particularly among those elements of their setting that have the clout and
resources to matter to the agency’s future prospects. This political dynamic
does not disappear when agencies operate in networked contexts; it is probably
exacerbated. As we have indicated, however, the point has not been a promin-
ent part of the recent and extensive research treatment of networks and public
management. In this analysis, it is not only sketched, it is supported with
93 Generalizing about managing in the network

systematic evidence covering hundreds of organizations – and managers – over a

several-year period. Treating managerial networking as a cognitive or technical
challenge misses the mark, for it obscures the likely tilting of the policy table
toward well-established and influential interests. Managerial networking does
not eliminate this bias; if anything, it can accentuate it.
In school districts in Texas, at a minimum, managerial networking does
boost educational performance, but most improvements accrue to the more
privileged portions of the constituency, not to the marginalized ones. Net-
work activity and management matter, but these elements are not ways of
overcoming inequities in service delivery. Exposing managers to the pres-
sures of their surroundings, particularly to influential actors with a distri-
butionally related agenda, appears to push them to respond to the most
influential portions of the network. Networking in other directions or with
other types of actors may produce benefits – or even costs – without
catalyzing further inequities as a result, however. Positive, mixed, negative,
and zero-sum games are all plausible. The details matter. Managerial net-
working is not a substitute for politics, nor is it a more sanitized and thereby
acceptable form of political activity. It produces the kinds of patterns and
dilemmas that social scientists have been documenting for years.
Although the empirical findings presented here are limited to Texas school
districts, two reasons suggest that similar patterns would be found in other
managerial settings where some networks and networking regularly operate.
First, school districts are public organizations with relatively common prob-
lems involving the incorporation and management of networks. These
findings are most likely to apply to organizations that share the characteris-
tics of school districts: highly professionalized and decentralized organiza-
tions with a great deal of managerial discretion. Second, the story told by the
data fits longstanding theories about organizations and their environments;
in effect, the moral is that we need to think of public management networks
in the broader context of organization theory.

Generalizing about managing in the network

Managerial networking, or measures of M2, have been successfully applied in

a number of areas other than Texas school districts. All the studies use
regression models similar to those presented in this chapter; all control for
a wide variety of other factors that could also influence program outcomes.
To retain the focus on the linkage between managerial networking and
94 Public management in interdependent settings

performance, in our review of such research we do not discuss the control

variables but, rather, focus on the relationship between managerial network-
ing and performance.
Donahue et al. (2004) use data from the Government Performance Project
to examine US state government agencies charged with overseeing human
resources activities, and agencies assigned the management of state debt.
Managerial networking is measured via contact and the reciprocity of
contact with various external stakeholders, such as governors and their staff,
legislators and legislative staff, other agencies, etc. For the human resources
agencies, they relate managerial networking to subjective perceptions of the
quality of personnel hired by the state, and also the level of employee
turnover. They find that the quality of the contacts rather than just the
volume of the networking positively affects the quality of hires, and that this
measure of networking is also associated with low employee turnover rates
(Donahue et al. 2004: 140).
For the debt administration agencies, managerial networking is measured
by the establishment of contacts with financial networks, the use of outside
financial advisors, the reliance on negotiated sales (and thus interaction with
clientele) rather than the more arm’s-length process of auctioning debt, the
degree of underwriting training for the staff (which might equally be
considered a measure of internal management), and the degree of overall
control of the issuance of debt in the state. For outcome variables, they use
the percentage of debt issued via competitive sales and the percentage of
debt that was sold via requests for proposals (rather than given to a single
underwriter). With several measures of managerial networking and two
different outcomes, Donahue et al. consistently find positive relationships
between managerial networking and outcomes, with a majority of the
relationships attaining statistical significance.
Meier, O’Toole, and Hicklin (2009) examine the networking behavior of
266 college and university presidents in the United States (for both public
and private universities). They create a measure of managerial networking
via a factor score that is very similar to the measure used for superintend-
ents. Nine external nodes are included that encompass political actors, other
bureaucratic agencies, business and community leaders, and other college
presidents. A single factor solution is generated. Greater levels of managerial
networking are positively associated with the university’s six-year graduate
rate, a measure of efficiency often used by state higher education governance
bodies. They also investigate efforts to create a more diverse faculty in terms
of the percentage of new faculty hires that are African American. In this case
95 Generalizing about managing in the network

M2 appears to work for public universities only; the impact of managerial

networking is negative for private universities, but the positive impact for
public universities cancels this out.
Nicholson-Crotty and O’Toole (2004) apply the model presented in
Chapter 2 to 570 municipal police departments in the United States while
using the percentage of index crimes cleared as a measure of performance.
Index crimes are the crimes that the Federal Bureau of Investigation (FBI)
uses to create serious crime indexes for both the nation and for individual
jurisdictions. Nicholson-Crotty and O’Toole create a measure of police
networking activities, especially related to their community policing activ-
ities; in a similar manner to the superintendent’s measure, theirs is based on
contacts with stakeholders. They find not only that managerial networking
correlates positively with the percentage of index crimes cleared but also that
it interacts with past performance to generate an even larger impact on
performance.
Jacobson, Palus, and Bowling (2010) investigate managerial networking
by state government agencies using the 1994 and 1998 American State
Administrators Project. Although their primary purpose was to examine
management behaviors and how they vary by gender, their findings are
directly relevant to the present study. As a dependent variable the authors
use whether the agency has adopted a series of government reforms that
come under the rubric of the “Reinventing government” initiatives. Basic-
ally, these initiatives focus on the use of incentives and the elimination of
restrictions on public managers. The authors create four measures of man-
agerial networking: contacts with political principals, contacts with citizens
and clientele, contacts with peers, and perception of the contacts initiated
with political principals. All these managerial networking variables were
positively correlated with “Reinventing government” outcomes, with the
exception of contact with other agencies – which was negative. Given that
the reinvention movement is generally a politically imposed reform, all these
relationships were in the direction hypothesized by the authors.
Andrews et al. (2010c) make a direct effort to replicate the managerial
networking measure in this chapter, using 69 English local governments
for 2002 and 2003. English local governments are multifunctional and
have authority over the implementation of most public programs other
than health care. These researchers create a networking measure using the
same frequency of contact scales as in the Texas studies and inquire about
eight nodes, including local and national political actors, other bureau-
crats, local stakeholders, trade unions, etc. As reported earlier in this
96 Public management in interdependent settings

chapter, this study also finds a single, general factor that could be termed
managerial networking. Andrews et al. find a different and differential
impact of networking on overall local government performance (the
outcome measure is established by the national government to explicitly
assess the performance of local governments across a variety of policy
areas), however. They find that poor performance brings contacts from
elected officials, and that such contacts serve as wakeup calls to adminis-
trators in regard to their level of performance. Local administrators then
increase their networking with user groups – an activity that, in turn, has
a positive impact on future performance. This cycle of poor performance,
wakeup calls from elected officials, networking with user groups and
subsequent increases in performance is very similar to that found by
Hawes (2006) with Texas education data.
Cohen, Vaughn, and Villalobos (2010) provide additional support for the
generality of the networking to performance relationship in their study of
the management of the US Office of the President. Explicitly using the
management theory in Chapter 2, they analyze more than 300 surveys from
individuals who interacted with the president’s chief of staff. The dependent
variable is the respondents’ rating of the effectiveness of the chief of staff.
They have an excellent measure of buffering (M4), with the chief of staff
taking the role of guardian of the president’s time; they have a separate
measure of being accessible that can be an M2 measure. Both external
management variables had a positive impact on how effective the chiefs of
staff were rated.

Conclusions

It seems clear that managerial networking contributes to public organiza-

tional performance, and not merely for top managers of Texas school
districts. The sets of relationships between this facet of management and
performance have been explicated with some care here. We have covered
considerable ground in this chapter in terms of the externally oriented
networking activities of public managers and a set of performance-related
impacts of such behavior. Despite this fact, we have not yet concluded our
analysis of the M2 function. We return to some additional aspects of
networking later in this book, but first we introduce some additional aspects
of management – both because these are interesting in themselves and
because they lay the groundwork for further investigations.
97 Notes

NOTES

1. The field of public management initially attempted to distinguish itself from the field of
public administration by arguing that it focused on actions at the very top of the
organization and, by definition, paid greater attention to the organization’s environment.
Although we do not see these two fields as separate, this study is central to the concerns
of scholars who identify with either of the fields.
2. Needless to say, in reality school districts are anything but insulated from politics.
3. This measure taps managers’ efforts to interact with their interdependent environment. It
does not reach to, or at least distinguish, efforts to manage at the network level – that is,
to manage or orchestrate the full set of interdependent actors en masse. Estimating this
aspect of public management is complicated (O’Toole 2000b). As indicated earlier in
this volume, we have made efforts to model management at the network level as a part of
this research program but do not delve deeply into this set of issues in the current book.
4. Findings in a study of English local government managers, designed around a similarly
constructed measure, mirror those obtained from the Texas data set; see Walker, O’Toole,
and Meier (2007).
5. The findings presented on this point rely on data from the 2000 survey responses. A more
complete exposition is presented by Meier and O’Toole (2003).
6. Work by Hawes (2006) indicates that the fire alarm pattern holds. Low performance at
time 1 is associated with greater school board contact at time 2, which is in turn associated
with higher performance at time 3. Hawes describes a cycle of political intervention
designed to improve performance. We have analyzed a somewhat similar pattern in data
on the performance of English local authorities (see Andrews et al. 2010c).
7. A similar survey was conducted several years later, in 2008, and the results were generally
quite similar.
8. The findings and discussion in this part of the chapter rely on Meier and O’Toole (2001).
9. For reasons explained earlier, for certain purposes the school board can be considered a
political principal rather than an external node.
10. Some of the networked relationships are mandatory and imposed from the environment.
An instance is the authority given to state auditors to check on fiscal matters. In most
cases, however, the superintendent can develop new relationships or seek to alter
mandated relationships in such a way as to benefit the district.
11. The actual correlation between managerial networking and interaction with one’s
school principals (a hierarchical form of management, thus part of M1), in the data
analyzed below, is 0.18, suggesting that, while the two are distinct, they are not
contradictory.
12. Underspecification is clearly less of a problem in the autoregressive form of the model,
since any omitted variables are likely to affect current outputs via past outputs. Only
those variables that affect current outputs but not last year’s outputs, therefore, are likely
to be a problem. The greatest concern should be whether there are omitted factors that
covary with elements included in the model.
13. Districts that responded to the survey were no different from nonrespondents in terms of
enrollment, enrollment growth, students’ race, ethnicity and poverty, or test scores. There
98 Public management in interdependent settings

were slight differences in a few other factors. Respondents had 0.48 more students per
class, paid their teachers $200 more per year, but had annual operating budgets of about
$100 per student less.
14. In this measure of M2, we include school boards as an element of the networked
environment of the superintendent. We treat school boards as part of the environ-
ment in this part of the analysis, since we think that on balance this notion is more
appropriate, but we have also performed the entire set of analyses again with M2
measured only on the other four types of interactions. The results are very similar to
those reported here. Omitting boards from the study strengthens the impact of M2 on
performance modestly in the linear versions of the simplified model and weakens
slightly the evidence on nonlinearity. The correlation between the two measures of
managerial networking is 0.96. The networking factor correlates at –0.27 with time
spent managing the district (in contrast to time spent in contacts outside the
organization).
15. Later in this book we introduce other measures of management, including managerial
quality.
16. Clearly, both “sides” can initiate interactions. Anecdotal evidence suggests that skillful
superintendents generally do not wait passively to be contacted, and our later surveys
have been designed to distinguish how much of the networking behavior reported by
managers was initiated by them. We do so by asking who initiated the last interaction
with each of the several nodes. Two findings of note can be summarized here. First,
managers do indeed report that they usually initiated the last interactions. Second, as
Goerdel (2006) has shown, a measure of managerial networking focused on the man-
agerially initiated interactions shows an even stronger relationship with performance
than does the measure used here. We treat this point later.
17. Of course, networking can occur at other levels of the organization, and this
measure will underestimate total networking by the organization. Some of the net-
work links are also clearly more important than others, and equal weighting might
obscure this. These and other measurement problems are likely to attenuate any
relationships found.
18. In policy analysis terms, test scores are an outcome rather than an output.
19. Exactly how managerial networking can influence performance is discussed below.
20. Collinearity refers to a situation in which two or more predictor variables in a multiple
regression analysis are highly correlated with each other. In such a situation, the coeffi-
cients generated for those variables are typically unstable. When collinearity is high,
standard errors are inflated but the estimation is unbiased. Collinearity impedes our
ability to develop precise estimates of the coefficients involved.
21. The resource measures should be thought of as general measures of resources from the
environment rather than specifically teacher salaries and class size. Access to resources
correlates with both teacher salaries and class size, as well as a variety of other factors.
22. The long-run performance of the management variable has a value of approximately 0.6,
which means the maximum total impact is approximately 3.6 percentage points – an
estimate very similar to the estimate for nonautoregressive models.
23. Only the within-quintile variance is being explained in these analyses. The middle
quintiles eliminate a great deal of the between-district variance.
99 Notes

24. As one would expect, the standard errors increase at the extremes, thus suggesting some
caution in interpreting the results. We view these findings as suggestive until confirmed
by other empirical studies.
25. This finding has implications for management theory and how to manage organizations –
that is, in the degree of hierarchical structure that managers should create (Drucker 1967).
26. Management in this situation is both a decision to act and then a match of the strategy
with the situation (see Lynn 1984). The decision to act in no way guarantees that the
strategy then selected will pay off.
27. Visits to various schools indicate that the process might be that networking behavior is
also associated with other behaviors that improve the levels of organizational cohesion.
Good and bad schools both frequently have the same programs; the difference is often in
the commitment of teachers and administrators to making programs work. Later in this
volume, we explore how managerial quality and also the quality of human capital in the
system help shape outputs and outcomes.
28. For a complete treatment of this topic, see O’Toole and Meier (2004b). Additional
analyses along the same lines have been carried out by Meier, O’Toole, and Lu (2006);
and O’Toole and Meier (2006).
29. A few exceptions can be noted. None of the year dummies was significant for attendance.
In three other cases – average SAT score, average ACT score, and percentage scoring above
1,110 in the SAT – the dummy for 1996 was not significant, but the succeeding years were
consistent with the upward trend.
30. The performance measure is the TAAS pass rate for these subgroups.
31. This summary includes relationships significant at p < 0.10.
32. The fourth impact is on dropouts.
4 Managerial quality and performance

Our basic model hypothesizes managerial influences on public organizational

and program performance, when managers exert effort on external manage-
ment as well as when they perform the standard internal functions that
comprise managers’ responsibilities. Chapter 3 has demonstrated that man-
agers do operate externally – presumably to buffer against negative shocks,
and also to exploit resources and opportunities in the organization’s environ-
ment on behalf of the agency and its programs. Indeed, that chapter illus-
trated the nonlinear interaction of managerial networking with key resources
for school districts. The chapter also showed that managerial efforts outward
generate performance dividends, although these are not neutrally distributed
to stakeholders; networking can have inequitable distributional conse-
quences. Before we address the subject of internal management (Chapter 5),
we need to revisit both managerial functions and introduce an aspect of
internal and external management that is implied in the initial model but
thus far not incorporated into the empirical analyses: the actual quality of
management. We proceed to show that quality not only affects performance
but links to managerial networking in interesting, nonlinear ways.
The “M” terms in the model obviously refer to managerial functions that
have both a quantity, or degree of activity, aspect as well as a quality
component. Our measure of managerial networking, introduced in the
preceding chapter, obviously has advantages – including validity and reli-
ability; but it lacks a “quality” component. Managers who network can do so
without much talent or perspective; alternatively, they can correctly assess
their organization’s environment, allocate their networking skills wisely, and
tap and/or protect from the most salient forces in the agency’s setting. We
can expect a considerable range of quality to be present across the individ-
uals managing public organizations. This chapter explores this aspect of
public management.
Doing so is important because a basic tenet of public administration is
that the quality of public management can make the difference between
100
101 The Gordian concept of public management quality

success and failure in the delivery of public policy results (Lynn 1984).
Despite this widespread belief, the notion has rarely been carefully tested.
In this part of the book, we develop a measure of managerial quality suitable
for certain kinds of empirical settings and then test whether high-quality
management contributes positively to public program performance.1 Public
education, an important policy field, once again provides the context for the
investigation (Raffel 2007).
This relatively straightforward test of the management quality hypothesis
confronts a number of challenges. The notion of managerial quality itself,
although often used in teaching, research, and practice, is seldom clarified in
a way that facilitates systematic investigation. Difficulties of measurement on
this score have also impeded research. In addition, many other influences
shape what happens via public programs, so the research needs to take into
account these realities. The next sections of this chapter treat these chal-
lenges systematically; then our research on the management quality hypoth-
esis is presented and discussed.

The Gordian concept of public management quality

As indicated earlier, the proposition that public management contributes to

the performance of government is at the core of a great deal of scholarship;
but few systematic efforts have tested for the relationship empirically. The
case study and qualitative literature, on the other hand, indicates that good
management can be a particularly critical contributor to program success
(see, for instance, Doig and Hargrove 1987, Hargrove and Glidewell 1990,
Behn 1991, Thompson and Jones 1994, Ban 1995, Riccucci 1995, Cohen and
Eimicke 1995, and Holzer and Callahan 1998). Indeed, this body of work
suggests multiple and complex channels of managerial influence.
All the same, the conceptual issues are immense. A consideration of
management’s hypothesized impact on program performance, for instance,
must incorporate some attention to the notion of leadership – a theme of
substantial importance among researchers. The literature on leadership is
huge and complex, however (see Rainey 2009). Rainey and Steinbauer’s
(1999: 18–19) succinct characterization serves as a daunting reminder of
the difficulties of capturing this key notion in a satisfactory and easily
measurable form: “[T]he topic of leadership is vast, richly elaborated, and
inconclusive. . . Enough listings of desirable leadership skills and qualities
102 Managerial quality and performance

could be gathered to build another great pyramid. They vary widely, and
none of them can claim conclusive validation.”
The growing emphasis on quality – and quality management – in recent
years (see Beam 2001) overlaps the attention to leadership in public pro-
grams. Interestingly, an examination of this theme also reveals an unresolved
tension as to what kinds of broad managerial efforts are likely to be most
critical for delivering performance. Much of the attention to quality or
excellence in recent years, in the United States and elsewhere, has focused
on the value of “entrepreneurial” management for achieving results. The
popularity of Osborne and Gaebler’s (1992) volume illustrates this point,
and the National Performance Review of the Clinton years – a reform effort
with direct intellectual ties to the same perspective – reflected a similar
emphasis (Gore 1993; see Rainey 2003: 408–11). The new public manage-
ment, more broadly, emphasizes these themes. Some analysts have seen in
these approaches a diminished view of management, however (Lynn 2001),
or one, they argue, likely to limit what public agencies can deliver (see
Goodsell 1993 and Moe 1994). Terry (2002) in particular contends that
administrators perform a key function by executing “conservatorship”:
preserving established institutional forms and activities that have developed
over time and would be difficult to reestablish (for a more elaborate discus-
sion of this point, see O’Toole and Meier 2007).
Indeed, while risk-taking, entrepreneurial activities can sometimes bring
benefits, protective, conserving efforts can be especially valuable under other
circumstances (see Meier et al. 2007). As we have argued earlier, the multiple
managerial functions, which likely work through different causal pathways,
should all be considered by those who desire to probe the connection
between management and performance. Although this general point may
be valid, any systematic effort to explore the link between management
quality and performance across a large number of cases must confront a
nearly intractable measurement challenge. If high-quality public manage-
ment embraces a multitude of difficult-to-define dimensions and if different
strategic approaches and managerial orientations might be appropriate
under different difficult-to-specify conditions, how can one test the propos-
ition that good management contributes to good performance across the
spectra of cases and circumstances?
The conceptual complexity thus fuels a serious measurement challenge. In the
broader literature beyond the public sector, efforts have been made to measure
the quality of management (Bloom and van Reenen 2007), but so far the criteria
used have not been applied in empirical studies of public management.
103 The Gordian concept of public management quality

With respect to the public sector, for some years now the Government
Performance Project has developed comprehensive measures of government
management systems via a criterion-based approach. Most of this research
effort has been devoted to measures of management itself, and management
capacity, rather than managerial impacts, although some relationships
between these measures and managerial (intermediate) outcomes have been
demonstrated (Donahue, Selden, and Ingraham 2000).
A few additional notes of progress have been sounded in the effort to
probe with systematic work the link between elements of public manage-
ment and ultimate program performance. Wolf (1993) examines subjective
assessments of agency leadership, and finds that these are correlated with
agency effectiveness. Hennessey (1998) suggests a relationship between
public organizational performance and leadership, defined in terms of
Bennis’s (1993) four competences, on the basis of data from nine offices
in two federal agencies. His core argument is that leaders shape organiza-
tional culture and, thereby, performance. Attention is directed primarily
to reinvention efforts, however, and only secondarily to performance
itself. Further, the small number of cases, subjective measurement of
leadership features, and lack of controls attenuate the conclusiveness of
the work.
Rainey and Steinbauer (1999) have proposed a “theory of effective gov-
ernment organizations” incorporating a number of features that might
explain effectiveness. Several characteristics they analyze are part of, or, at
minimum, closely related to, public management – including the develop-
ment of human resources (see Chapter 5), various elements of task design,
and, in particular, leadership characterized by certain attributes. Rainey and
Steinbauer craft their argument on the basis of a review of existing literature
on the likely determinants of effectiveness. While they do no testing, they do
“posit” that leadership is likely to “emerge as” among the most important
drivers of effectiveness in governmental organizations (28).
Brewer and Selden (2000) report a systematic empirical project based on
Rainey and Steinbauer’s theoretical argument. They explain a large portion
of the variance in federal employee perceptions of organizational perform-
ance, as interpreted in rather broad terms, across twenty-three agencies. The
model they develop and test includes a leadership and supervision measure,
which is positively related to perceptions of performance, although its
predictive power is relatively slight. The measure is limited to employee
perceptions of how their immediate supervisors rate; and, as Brewer and
Selden note, “leadership and supervision may contribute to organizational
104 Managerial quality and performance

performance indirectly” (704, emphasis in original). Indeed, several other

variables they analyze that contribute more to explaining the variance in
performance are likely influenced by management as well.
These findings and arguments are provocative, but they are limited in a
number of ways and clearly not definitive. Most of the empirical work is
cross-sectional, and it is important to test for the impact of public manage-
ment by incorporating a longitudinal dimension as well. Most of the meas-
ures of performance are perceptual and/or intermediate, and thus may be
biased, given that the respondents are evaluating their own performance (see
Meier and O’Toole 2010). In addition, the measurements developed thus far
capture only a limited part of the concept of quality management as it has
been understood by scholars.
If characterizing and measuring managerial quality is challenging, even
more demanding is the task of doing so for individuals in specific manager-
ial positions. The general task of individual performance appraisal in the
public sector has been notoriously difficult to conduct (Kellough 2006; see
also Murphy and Cleveland 1995). The approach adopted in the present
investigation does not resolve the host of issues under dispute, but it does
rely on decision making by knowledgeable political principals in contact
with the particular managers whose impact is being analyzed here. To be
precise, the method relies on assessments revealed in salary determinations.
This approach might seem ironic, since when individual performance
appraisals are used in public agencies to determine pay – so-called “pay
for performance” systems – researchers have consistently noted serious flaws
(Ingraham 1993; Rainey 2009). Under certain conditions and with certain
caveats, we argue below, decisions about pay can provide a defensible
indirect measure of management quality, particularly given the conceptual
and measurement difficulties associated with developing a more direct yet
still feasible alternative (for a similar approach to measuring quality at the
middle manager level, see Johansen 2008).
In the next section, we sketch our general approach to measuring man-
agerial quality in the kinds of settings that will be analyzed later: public
education via the Texas school districts data set. Our focus is once again on
the top managers there: school superintendents. We propose an aspect of
managerial salary as a reasonable proxy measure for testing the management
quality hypothesis. We then develop the specifics of our empirical measure
and finally model our tests of school system performance. We then show
how managerial quality can interact with managerial networking to avoid
problems of diminishing returns.
105 Measuring superintendent quality

Measuring superintendent quality

As outlined above, what is needed from public managers seems to vary by

program, agency, time, and context. Any simple measure of the quality of
management, therefore, is likely to be biased.2 Those in the best position to
know and evaluate what and how managers are doing are knowledgeable
observers in the local setting at the time, particularly those with access to
information about managerial behavior, organizational morale, environ-
mental demands, and performance results. The strategy in this study is to
tap into the judgments of just such a set of individuals who observe the
managers – school system superintendents – on a day-to-day basis: members
of the school board.
Rather than seeking attitudinal judgments by the school board on man-
agement quality (thus merely moving the problems of definition from
researchers to practitioners), we assume that actions reveal evaluations. Each
school board makes an annual assessment of the superintendent’s perform-
ance and then sets his or her salary for the following year. In that determin-
ation, we think that management quality plays a role – not an exclusive role,
but a role nonetheless. Similarly, deciding the compensation to offer a new
superintendent contains an inherent quality assessment.
Quite clearly, political principals face limitations in judging managerial
quality – in particular, limitations in access to relevant information. For
governmental jurisdictions that perform only one policy function, these
limits are less severe. To the extent that political leaders in such situations
are interested in attending to the quality of management in their jurisdic-
tions, they know where to look and are undistracted by competing or
overlapping responsibilities. School districts are among the governmental
jurisdictions fitting this stipulation.
Furthermore, isolating on the managerial quality aspect of a superintend-
ent’s salary is facilitated by several characteristics of the market for superin-
tendents. That market can be characterized as competitive with substantial
information. School district managerial talent is mobile within the state
(and somewhat mobile across states). While some superintendents remain
for extended periods in one locale, most individuals typically move through
several districts as they pursue their careers. With few exceptions, positions
are filled after open searches that are often conducted with the assistance of a
search firm. Superintendents seeking to move (the average tenure in Texas is
approximately 5.3 years) will know the salary paid to the previous
106 Managerial quality and performance

superintendent and can access an extensive state database on the district and
its characteristics. Similarly, the hiring district will have extensive infor-
mation about how the candidate’s current district (or school, if it decides
to hire a principal) has performed, and assessments of an individual super-
intendent candidate’s management ability are relatively easy to obtain via the
established network of school board members. In short, a manager with a
good track record is likely to have several options, so that a school district
seeking to hire such a manager will need to offer a premium, all other things
being equal. There are no regulatory floors or ceilings regarding compen-
sation. The sheer range of salaries in the study ($35,000 to $205,228 in 1999
[mean ¼ $74,400; standard deviation ¼ $24,087]) supports the notion that
market dynamics are at work.3
Salary premiums operate within a salary structure that recognizes basic
understandings about the job, however (see Ehrenberg, Chaykowski, and Ehren-
berg 1988a, 1988b). First, the most significant determinant of salaries, both
normatively and empirically, is the size of the district; as the size of the job
expands, salaries increase proportionately. Second, human capital factors such
as education, experience, and training will result in additional adjustments to
salaries. Third, personal characteristics of the individual are likely to affect
salaries. Particularly relevant are such factors as race, ethnicity, and gender.
Although discrimination might play a role here, some districts, such as large
inner city districts, will prefer a minority superintendent for political reasons.
Fourth, because the relationship between salaries and performance can be
expected to be reciprocal – that is, superintendents could also be rewarded for
performance in the past – a control for prior school district outputs is needed.
Our strategy of analysis is to take variables measuring each factor that
should influence the manager’s salary and use them to predict the manager’s
actual salary. The residual from this equation – that is, the variance in salary
not accounted for by job size, human capital, personal characteristics, and
past performance – will contain the assessment of managerial quality (for a
similar residuals-based measure in a different context, see Palmer and
Whitten 1999: 629). This measure is quite clearly a messy one, since the
residual contains all those factors not included in the model – such as the
ability to sell oneself, experience and renown as a football coach, physical
characteristics and other irrelevant factors, as well as the assessment of
quality. The impact of this measurement error will attenuate any relation-
ships between the quality measure and organizational outputs, however
(Carmines and Zeller 1979; Bollen 1989: 159–67). The measurement error,
as a result, creates a bias in favor of null findings.
107 Measuring managerial quality

Measuring managerial quality

How might a superintendent, a single manager at the top of the hierarchy,

actually affect student performance? In the last chapter we discussed how
external efforts on the part of top managers might shape organizational
outputs and outcomes, and in the next chapter we examine some of the
details of how internally directed management can contribute. Here we note
several causal paths, some internal to the organization and others partially
external. Our interactions with superintendents, administrators, and
teachers suggest several ways that top managers can boost performance.
First, superintendents can recruit, train, and reward talented mid-level
administrators (school principals) and teachers. Of particular importance
is recruiting individuals who share organizational goals in regard to student
standards and approaches to education. Second, superintendents, like all
organizational leaders, can motivate employees to invest greater effort in the
organization. Superintendents do so by providing and communicating a
vision for the organization. They can also generate greater commitment by
handling the inevitable problems that arise in the environment; in particular,
they can provide political cover for teachers and administrators. Third,
superintendents can affect the student learning environment by mandating
the adoption of specific educational reforms. The list of possible reforms is
endless, and reforms need to be matched to the specific needs of the students
and the skills of teaching faculty. Part of this matching process relies on the
management ability of the superintendent. Fourth, superintendents can
contribute to the predictability and reliability of the system for those who
operate within it. They can provide stable processes and avoid the disrup-
tions of policy churn (Hess 1999) and other activities that interfere with the
process of educating children.4 Finally, superintendents can acquire more
resources for the organization. While most of these resources will then
appear in the district’s budget, some may be intangible – for instance, the
goodwill of local business leaders or the support of parent groups.
To generate the residuals-based measure of managerial quality, we use a
relatively common salary model from the literature (see Ehrenberg, Chay-
kowski, and Ehrenberg 1988b). The dependent variable is the logged annual
compensation for the superintendent. The log transformation is used to ease
the problem of skewed data resulting from the large salaries associated with
Texas’s megadistricts. The log transformation also permits the relationships
to be interpreted as elasticities. This salary figure includes only the official
108 Managerial quality and performance

base salary; it omits the perks some districts offer, such as club memberships,
cellphones, and transportation benefits that are not reported to the state
of Texas.
Three district characteristics are included as independent variables: the
district’s total budget, tax rate, and average revenue per student; all three
variables are logged. Total district budget is our measure of district size,
which should be the strongest predictor in the model. The tax rate is
included because some earlier work contends that superintendents are
rewarded for keeping taxes low (Ehrenberg, Chaykowski, and Ehrenberg
1988b). Revenue per pupil is a measure of wealth; certain districts will pay
higher salaries simply because they can afford to do so. For some districts
this decision is a matter of civic pride.
Four human capital characteristics are included: experience as a superin-
tendent, tenure in the current job, age, and the possession of a doctorate.
The first three variables are measured in years; salaries should increase both
with total experience as a superintendent (most of this experience will have
been in other districts) and time in the current job. Age is commonly
included in models such as these, even though it is considered a surrogate
for experience, which is already in the model.5 In terms of education,
virtually all superintendents have a master’s degree (98 percent), so the most
salient distinction is the possession of a doctorate, which should be posi-
tively related to salary.
Three personal characteristics are included: whether the superintendent is
female, black, or Latino. The predicted signs for these variables are ambigu-
ous, depending on whether a district might see it as an advantage to hire a
superintendent with a given demographic. Data on salaries, district charac-
teristics, human capital, and personal characteristics were provided by the
Texas Education Agency from their administrative database.
Finally, we include the previous year’s test scores in the model. Because we
think perceived managerial quality is affected by prior performance, and
because quality then affects future performance, over time there is reciprocal
correlation. We cannot control for prior test scores without adjusting for this
endogeneity, or the quality measure’s impact will be biased downward. The
appropriate method is to purge the reciprocal causation via an instrumental
variables technique. We do this using six student characteristics and district
resources (percentage of black, Latino, and low-income students, teacher
salaries, class size, and instructional funding) as instruments; the purged
measure of prior performance is then included in the model. Five years
(1995 to 1999) of data are used in the model, and dummy variables for
109 Measuring managerial quality

Table 4.1 Determinants of superintendent salaries

Dependent variable ¼ logged annual compensation

Independent variable Slope Error T-score

District characteristics
Logged budget 0.1641 0.0017 95.07
Logged tax rate 0.0272 0.0161 1.69
Logged revenue/pupil 0.0683 0.0092 7.45
Human capital
Past experience 0.0022 0.0003 7.94
Current job tenure 0.0009 0.0002 3.63
Doctorate 0.0532 0.0045 11.79
Age 0.0004 0.0002 1.95
Personal characteristics
Female 0.0025 0.0009 2.85
Black 0.0941 0.0183 5.16
Latino 0.0165 0.0081 2.03
Past performance 0.0009 0.0003 3.16

R2 0.78
Standard error 0.1251
F 1193.92
N 5,127

Note: Coefficients for individual years not reported.

individual years are included to account for the general increase in salaries
over this time period.
The results of the salary model appear in Table 4.1. The predictive ability
of the model (78 percent) compares favorably to other models in the
literature; and, with one exception, all the variables are in the predicted
direction.6 That exception is the tax rate, which has a slight positive associ-
ation with salary rather than a negative relationship, thus indicating that
superintendents are not systematically rewarded for keeping taxes low.
Although the relationships in the model are interesting in terms of both
personnel management and educational policy, discussion of them is beyond
the scope of the present investigation. The objective of this part of the
analysis is merely to remove as many “non-quality” factors from the super-
intendent’s salary as possible. The regression residuals are then standardized
(converted to a mean of zero and a standard deviation of one) for use in the
subsequent analysis. Because salaries are set before the school year begins,
the quality measure has a natural one-year lag in its relationship to current
organizational performance. As a consequence, any relationships that are
110 Managerial quality and performance

found can be attributed to quality influencing performance rather than

district performance influencing boards’ decisions about superintendent
compensation.

Modeling performance

Our measure of management quality can now be related to educational

performance. Given the preliminary nature of such a measure in public
management, this effort might be viewed as an attempt to determine if the
measure has external validity – since managerial quality should affect organ-
izational performance when one controls for the resources and constraints
on the core organization. The section first identifies the control variables in
the model and then discusses the measures of organizational performance.

Control variables
As discussed in earlier chapters, any assessment of organizational perform-
ance must control for both the difficulty of the job faced by the organization
and the resources in its possession. We use the well-developed literature on
educational production functions for guidance and include the same set of
control variables used in Chapter 3. These eight variables are three measures
of task difficulty and five measures of resources.7 These are used strictly as
controls, to make sure that any findings we have relative to management
quality are robust to the inclusion of factors normally linked to educational
performance.

Performance measures
This chapter incorporates eleven different performance indicators in an
effort to determine if management quality affects a variety of organizational
outputs. The most salient is the student pass rate on the Texas Assessment of
Academic Skills.8 Our measure is the percentage of students who pass all
(reading, writing, and mathematics) sections of the TAAS.
Four other TAAS measures are also useful as performance indicators.
TAAS scores for Anglo, black, Latino, and low-income students are included
as measures of performance indicators. TAAS scores are linked most directly
to basic skills and performance levels for all students. Many parents and
policy makers are also concerned with the performance of school districts
111 Modeling performance

regarding college-bound students. Four measures of college-bound student

performance are used: the average ACT score, the average SAT score, the
percentage of students who score above 1,110 on the SAT (or its ACT
equivalent), and the percentage of students who take either test. As men-
tioned in Chapter 2, Texas is one of the few states where both the ACT
and the SAT are taken by sufficient numbers to provide reliable indicators
of both.
The final two measures of performance might be termed bottom-end
indicators: attendance rates and dropout rates. Dropout rates, while it is
conceded that they contain a great deal of error, are frequently also used to
evaluate the performance of school districts. The official state measure of
dropouts is the annual percentage of students who leave school from eighth
grade onward.

Findings
The first school district performance measure assessed is the overall TAAS
score; these results are presented in the first two columns of Table 4.2. The
proposed measure of managerial quality is positively and significantly
related to school district performance. Since the measure is standardized,
and thus ranges between approximately 3 and þ3, these equations suggest
that the maximum impact of quality management is approximately 5.3
points on the TAAS. Although management quality is clearly not the most
important factor in determining test scores, in substantive terms 5.3 points
is a meaningful amount of change (the standard deviation of TAAS scores is
approximately 12.5). To check for omitted variables bias, we ran regressions
with forty-one additional variables, without affecting the findings here.
These variables included additional student characteristics, budget expend-
itures in various categories, teacher assignments, and additional measures of
district wealth.
To explore a bit more how management quality might work through
other factors known to influence performance, a second regression in
Table 4.2 adds three variables: parental involvement, community support,
and student attendance. Parental involvement and community support were
assessed via a superintendents’ survey; because they reflect the impressions
of the superintendents, these measures might contain some bias.9 All three
new measures are positively associated with organizational performance; in
the case of student attendance, the relationship is a strong one. Including
these factors in the model reduces the size of the management coefficient.
112 Managerial quality and performance

Table 4.2 The impact of management on performance: standardized tests

Dependent variable ¼ TAAS pass rate

Independent variable Slope T-score Slope T-score Slope T-score

Management quality 0.8866 7.76 0.4888 3.23 0.8334 5.13

Managerial – – 0.6418 4.27
networking
Parental support – 0.3984 2.01 –
Community support – 0.9572 4.51 –
Student attendance – 3.7705 20.90 –

R2 0.59 0.67 0.60

Standard error 8.00 6.78 7.51
F 574.29 298.36 264.25
N 5,126 2,498 2,502

Notes: Coefficients for annual dummy variables are omitted. Equations also control for
teacher salaries, state aid, class size, teacher experience, noncertified teachers, and the
percentages of black, Latino, and low-income students.  ¼ significant at p < 0.05.

These relationships suggest that some of the impact of quality management

operates through increasing community support and parental involve-
ment.10 Even with the addition of the attendance and support scores,
however, management quality as defined in this study has a significant and
positive effect on the overall performance of the organization.11
This final regression in Table 4.2 includes the managerial networking
measure from Chapter 3 (four nodes). Networking is uncorrelated with
the managerial quality measure (R2 ¼ –0.01), and thus both positively
influence organizational performance. The size of the coefficient changes
only because the networking measure is survey-based, and thus the third
equation reflects the missing data from survey nonresponse. Controlling for
networking has little impact on the strong positive relationship between
quality and performance. Because both measures have a mean of zero and a
standard deviation of one, they can be directly compared to each other. The
impact of managerial quality is about 30 percent larger than the impact of
managerial networking on the TAAS scores.
A measure of managerial quality should be general; it should be related to
a wide variety of organizational outputs. The relationship should, of course,
vary across different measures of outputs, because some problems are likely
to be more sensitive to the quality of management in the organization. As
problems become more intractable, for example, one would expect that
113 Modeling performance

Table 4.3 Management quality and other measures of performance

Performance measure Slope T-score R2 N

Latino pass percentage 0.4832 2.53 0.38 4,243

Black pass percentage 0.7014 2.68 0.38 2,965
Anglo pass percentage 0.8700 7.60 0.41 5,053
Low-income pass percentage 0.8998 6.17 0.50 5,093
Average ACT score 0.0817 3.94 0.36 4,248
Average SAT score 3.1534 2.85 0.50 3,516
Percentage above 1,110 0.6535 4.23 0.29 4,682
Percentage tested 0.0113 0.05 0.12 4,601
Dropout percentage 0.1241 8.21 0.16 5,026
Class attendance 0.0866 7.49 0.24 5,126

Notes: All equations control for teacher salaries, instructional expenditures per student,
class size, teacher experience, percentage of teachers not certified, percentage of black,
Latino, and low-income students, and yearly dummy variables.  ¼ significant at p < 0.05.

management would matter less, simply because what the organization could
do to solve such problems is more limited.
Table 4.3 presents the regression coefficients for management quality and
the ten additional performance indicators. Each equation also controls for
all the variables included in the first regression in the table.12 The perform-
ance of the managerial quality variable can be appropriately characterized as
stunning. For nine of the ten additional performance indicators, manage-
ment quality is significantly related to performance in the predicted direc-
tion (the exception is the percentage of students who take college boards).
This pattern of relationships, along with those in Table 4.2, amounts to
strong evidence that the residual-based measure of managerial quality is
tapping at least in part some aspects of how well superintendents manage
their districts.13
We also replicated the results of Table 4.3 and included the managerial
networking variable. For all the dependent variables but black pass rates and
the percentage of student who took either the SATor the ACT, the relationships
were statistically significant and in the correct direction. As one final robust-
ness check, we reestimated the equations with a lagged dependent variable.
Such a test is stringent, since it requires managerial quality to have an impact
over and above last year’s scores – in short, to continue the improvement
that generated the higher salary in the first place. Table 4.4 reports an
abridged set of results, which show that managerial quality maintains a
statistically significant impact for five of the outcome indicators – Anglo
114 Managerial quality and performance

Table 4.4 Management quality in an autoregressive specification

Performance measure Slope T-score R2 N

TAAS pass rate 0.1049 1.39 0.84 5,125

Latino pass percentage 0.0782 0.50 0.59 4,398
Black pass percentage 0.3811 1.82 0.56 2,850
Anglo pass percentage 0.1742 2.16 0.68 5,037
Low-income pass percentage 0.1293 1.20 0.73 5,083
Average ACT score 0.0435 2.24 0.47 4,018
Average SAT score 1.4379 1.59 0.68 3,216
Percentage above 1,110 0.3853 2.68 0.38 4,632
Percentage tested 0.0441 0.21 0.33 4,485
Dropout percentage 0.0843 6.19 0.32 5,024
Class attendance 0.0150 2.17 0.73 5,125

Notes: All equations control for teacher salaries, state aid, class size, teacher experience,
percentage of teachers not certified, percentage of black, Latino, and low-income students,
and yearly dummy variables.  ¼ significant at p < 0.05.

pass rates, ACT scores, college boards above 1,110, dropouts, and attendance.
If the criterion is relaxed to a 0.10 with a one-tailed test, given the direction
specified, the results for the overall TAAS rate, the black pass rate, and SAT
scores are also significant in the correct direction.

Managerial quality and managerial networking: diminishing returns?

At the outset of this chapter we noted that our measure of managerial

networking focused only on the quantity of the networking, not on the
quality of the networking. The creation of a management quality measure
provides us with the opportunity to investigate further the relationship
between managerial networking and organizational performance. What,
for example, is the functional form of the link between management and
performance? Our model specifies some hypothesized relationships, and we
have seen that there is evidence in support of some of its clearest notions.
What about the details, though? One possibility, of course, is that the two are
linked in a straightforward linear fashion over the full range of the manage-
ment variables: additional increments from management, or various types of
management, may add some regular and fairly constant amount of outputs
or outcomes. Alternative possibilities, however, are myriad. The relationship
might be curvilinear; or there may be diminishing returns; or there may be
115 Managerial quality, networking: diminishing returns?

some critical managerial contribution, past which point performance accel-

erates more rapidly. Alternatively, the relationship might be exceedingly
complex, in which case plotting a management-versus-performance curve
might encompass twists and turns of confusing sorts. Sketching the possi-
bilities de novo would produce graphic patterns reminiscent of so many
serpents lying in the sand – some stretched out lazily and signaling simple
forms, others arching and wiggling across the sand (for a complete expos-
ition of this analysis, see Hicklin, O’Toole, and Meier 2008).
In the initial formulation of the model, we explicitly suggested some
nonlinearities – in particular on managerial networking externally into the
interdependent environment, and specified nonlinearities with respect to
resources and constraints in the organization’s setting. A careful further
examination of the model and its features suggests that other nonlinearities
may also be expected, however. These have to do with, for instance, oppor-
tunity costs and potential tradeoffs across different managerial tasks; the
potential for diminishing returns from management; and the role that even
internal resources, particularly those related to personnel, can play in pro-
viding a partial support or substitute for explicit top managerial effort.
This section focuses on one particular managerial responsibility – managerial
networking, or managers’ interactions externally in support of the public
organization and its tasks – in order to explore the nature of the functional
form between this aspect of management and organizational performance. The
empirical work presented thus far has demonstrated overwhelming support for
the positive effect of networking (M2) on performance. More networking is
correlated with an increase in performance, when controlling for resources and
constraints in the environment as well as selected other managerial influences.
Networking seems to be the gift that just keeps on giving. With the extant
evidence, it could be concluded – rather implausibly, granted – that managers
who want to increase performance should devote as much time as possible to
externally oriented networking.
Does this inference make sense, though? Are there times and circum-
stances when it is probably better to network less, perhaps in the interest of
fulfilling another managerial requisite? Is it possible for managers to devote
too much time to managing the external environmental – to the detriment
of the organization’s performance? Could the benefits of networking be
contingent on the characteristics or talents of individual managers? Could
organizational characteristics affect this relationship? First we explore the
reasons to anticipate a possible nonlinear relationship, and then we outline a
number of contingencies that might affect this relationship.
116 Managerial quality and performance

Resources, time, and the managerial balancing act

There are reasons to expect that very high levels of networking may have
diminishing, or perhaps even negative, effects on organizational perform-
ance. Most of the literature on the benefits of networking focuses on how
managers form relationships with other organizations and stakeholders to
secure benefits, fend off disruptions, and identify opportunities. Managers
often network with others in an effort to attract and acquire more resources
for the organization. The environment and the relevant network actors do
not have infinite resources, however. While managerial networking should
result in considerable payoffs much of the time, there could be a limit to
these payoffs – meaning that at some point there is nothing, or, at least, less,
to gain from more external interactions.
A related point has to do with opportunity costs. Managers must perform
functions internally within their institution, not just outside. At some point,
the time spent on trying to extract the last bit from resources or buffer the
agency from all potential disturbances could have been better spent else-
where. The formal model sketched in Chapter 2 implies some sort of
balancing between managerial responsibilities without specifying how and
when the tradeoffs actually appear. More precisely, the model does not
specify the relationship between M1 (core internal functions) and M2 (exter-
nal, or networking, management). Although there is no reason to believe
that these two functions are necessarily a zero-sum effort – with a full
tradeoff at the margin between the two – managers who spend most of their
time on one component to some neglect of the other may generate perform-
ance setbacks, or, at least, suboptimal results.
Regardless of whether the limits on the benefits of networking come from
increasing difficulty in tapping opportunities/fending off perturbations
from the environment, or from tradeoffs among the multiple functions of
management,14 the consequence should be the same: at higher levels of
managerial networking, still more networking could be expected to produce
little in the way of positive performance payoffs for the organization. This
relationship can be tested with an hypothesis: the relationship between
networking and performance is nonlinear, with diminishing returns at the
higher levels.
We might also consider contingencies. Although we expect that, in the
aggregate, higher levels of networking will have diminishing effects, some
managers and certain public organizations may be able to ensure that time
117 Resources, time, and the managerial balancing act

spent networking continues to result in steady performance dividends, even

at higher levels of networking. Drawing on the two explanations as to why
the relationship may be nonlinear, we can speculate on what types of
managers and organizations may be able to avoid diminishing returns.
Indeed, we have already noted two dimensions of such managerial activ-
ity as networking: a “quality” or skill component, along with the “quantity”
or effort aspect. The original model does not specify the relationship
between these two elements, but it seems reasonable to expect that one
influences the functional form of the other. The most straightforward way
of interpreting the relationship would be that management quality mitigates
the otherwise expected diminishing returns hypothesized above.15 Second,
the model clearly suggests that aspects of “management” and aspects of
personnel/staffing perform some overlapping functions. Each, accordingly,
should be considered for generating nonlinear relationships for managerial
networking.
If the explanations for the general, hypothesized nonlinear relationship
sketched above are correct, a highly skilled manager might be able to avoid
negative returns to networking by being just active enough in the organiza-
tion’s environment to gather the crucial benefits feasibly available. This
expectation is grounded in a basic rational choice framework, which sug-
gests that managers are likely to try to assess when the “costs” of networking
(in terms of time spent and opportunities forgone) can no longer be
expected to be outweighed by the benefits that they expect to reap. The best
managers should be more adept at weighing the costs and benefits, and they
will therefore be the most efficient with their time, thus avoiding spending
too much time in the networked setting, presumably at the expense of other
managerial possibilities that also contribute to performance.
This chapter has demonstrated that managers vary considerably in qual-
ity, and the abilities of individual managers can have a strong impact on
their behavior and their organization’s performance. For the best managers,
therefore, we may find that this nonlinear effect disappears – a relationship
we test with a second hypothesis: the relationship between networking and
performance is positive and linear, when controlling for managerial quality.
The second explanation as to why increased networking may not always
be positively related to performance centers on the tradeoffs made by top
managers when choosing to apportion time and effort between internal and
external management. Organizations obviously need a certain amount of the
former to function properly – what Woodrow Wilson (1887) referred to, in a
household analogy, as the “management of the fires and the ovens.”
118 Managerial quality and performance

Neglecting these managerial requirements can be expected to result in poor

performance. Managers must find the right balance in splitting their time so
that internal needs are not neglected and external opportunities are suffi-
ciently exploited. Organizations may differ, however, in the extent to which
these tradeoffs bind tightly. Some managers may be able to hire and use
administrative staff to take care of straightforward but important internal
managerial functions, thus freeing up the top manager to spend more time
externally. These staff become managerial capacity – the ability to take
action if opportunities arise. If a manager has an executive staff that can
coordinate some of these internal management functions (and so an aspect
of M1 – a subject explored more systematically later in this book), significant
tradeoffs may be obviated (see Chapter 6). This line of reasoning leads to
additional hypotheses: the relationship between networking and performance
is nonlinear for those organizations with fewer administrative staff; and the
relationship between networking and performance is positive and linear for
those organizations with more administrative staff.
Our argument about the diminishing marginal returns of networking can
be tested by simply adding a squared value of networking to our basic equa-
tion, with the expectation that the slope for the linear term will be positive and
the slope for the squared term negative. For the empirical analysis, we use once
again the Texas school districts data set. Our analysis includes our survey data
from superintendents as conducted in 2000. To these survey responses, we
added eight years of data (1995 to 2002) from the Texas Education Agency.16
Because this analysis involves a pooled time-series approach, we included
dummy variables for the individual years (fixed effects) to deal with serial
correlation. We then assessed the degree of heteroskedasticity with pooled
diagnostics, and found the levels well within acceptable limits.
Our measure of performance is the percentage of students in a district
who passed all (reading, writing, and mathematics) sections of the TAAS.
For managerial networking we use the four-node factor score derived from
the survey responses. We also include the same measure of managerial
quality that was introduced in the preceding empirical analysis. Two sets
of hypotheses specify that the nonlinear relationship between networking
and performance is conditioned by another variable: administrative cap-
acity. Administrative or management capacity can be measured in a straight-
forward manner; it is simply the percentage of employees who are classified
as central office staff (see Chapter 6). This measure includes assistant
superintendents and staff positions such as the business manager, the human
resources function, and others, but it excludes staff assigned at the school
119 Resources, time, and the managerial balancing act

Table 4.5 The nonlinear relationship between networking and performance

Dependent variable ¼ student pass rates on TAAS

Linear Nonlinear

Independent variables Slope T-score Slope T-score

Managerial networking 0.883 7.54 1.970 4.21

Managerial networking squared – 0.232 2.40

R2 0.611 0.612
F 271.93 245.59
N 4,182 4,182

Notes: All equations control for teacher salaries, state aid, class size, teacher experience,
percentage of teachers not certified, percentage of black, Latino, and low-income students,
and yearly dummy variables. F-test for added variable yields F-statistic of 5.176 and
probability of < 0.0165.

level (principals, counselors, etc.). Our eight usual controls, resources and
constraints from the environment, are also included in all analyses.
Table 4.5 provides the abridged regression results for the linear and non-
linear estimation of managerial networking’s impact on performance (con-
trol variables not shown). Because managerial networking is a factor score, it
has a standard deviation of one; therefore, a one standard deviation change
in managerial networking is associated with a 0.883 percentage point
increase in TAAS exam pass rates. Over the full range of this variable, this
effect size translates into a maximum possible impact of approximately 4.7
percentage points on a district’s pass rate for this time period. The impact is
both statistically and substantively significant. This result is similar to
analyses presented earlier for different time periods.
Table 4.5 displays the results for the nonlinear estimation of managerial
networking – that is, it includes a value for networking squared. As pre-
dicted, the linear term is positive and significant and the squared term is
negative and significant. This pattern represents a classic case of diminishing
returns. The slope for managerial networking can be calculated for any value
by simply taking the first derivative of the equation and substituting in
values. Since the variable’s range was rescaled to be positive and runs from
0 to þ5.3, the slope is steepest at low levels of networking. At a networking
value of 0 (the least amount of contact with other nodes), a one-unit change
in networking is associated with a 1.97 point increase in the TAAS pass rate.
The impact gradually declines until a networking value of 4.25, where the
slope is zero – that is, additional networking above this value does not
120 Managerial quality and performance

80

Performance 79

78

Linear estimation
77
Nonlinear estimation

76

75
0 2 4 6
Networking

Figure 4.1 Linear and nonlinear relationship

contribute anything further to performance. In the aggregate, therefore, and

as the first hypothesis states, the relationship between managerial network-
ing and performance is clearly nonlinear. Figure 4.1 depicts the patterns for
the linear and also the more accurate nonlinear estimations graphically. For
the latter, the “serpentine” relationship arches gracefully.
We then test the second hypothesis – that, when one controls for quality,
the nonlinear relationship is expected to disappear. Table 4.6 presents the
findings for the test of this hypothesis. Here it is apparent that the squared
term is not significant, although it remains negative. The size of the net-
working coefficient itself is similar to that in the previous model, meaning
that, even when we control for quality, networking is still a significant
predictor of performance; managerial networking and managerial quality
contribute distinguishable and positive impacts on performance. We no
longer see the diminishing returns, however.
Why, exactly, does the nonlinear relationship disappear when we control
for quality, as Figure 4.2 shows? Are better managers able to network more
effectively (so that they always see returns), or are they able to network
more efficiently (they avoid networking to the point where they see dimin-
ishing returns)? A simple look at the data can help to answer this question.
Whereas the networking variable for the entire sample ranges from 0 to 5.3,
the range of networking for the best managers (those managers in the top
5 percent on the quality measure) reaches a maximum value of only 4.32.
The highest-quality managers limit their network activities before they
generate negative returns. This fact lends some support to the idea that
121 Organizational differences

Table 4.6 Nonlinear effects of networking, controlling for quality

Dependent variable ¼ student pass rates on TAAS

Independent variables Slope T-score

Managerial networking 1.519 3.27

Networking squared 0.144 1.50
Managerial quality 0.691 5.60

R2 0.624
F 235.58
N 4,114

Notes: All equations control for teacher salaries, state aid, class size, teacher
experience, percentage of teachers not certified, percentage of black, Latino, and
low-income students, and yearly dummy variables. F-test for added variable
yields F-statistic of 31.35 and probability of F < 0.001.

84

82 Controlling for quality

Basic nonlinear model
Performance

80

78

76

0 2 4 6
Networking

Figure 4.2 Controlling for quality

better managers are more efficient, or at least more perceptive, with the
ability to gauge when additional time spent in the networked environment
may not garner substantial payoffs.

Organizational differences

To evaluate the third and fourth hypotheses, on the impact of central office
staff size on the management–performance relationship, we split the sample
into two groups to test for how the presence of administrative human
122 Managerial quality and performance

Table 4.7 Nonlinear effects of networking on performance: impacts of the relative size of central
administrative staff

Dependent variable ¼ student pass rates on TAAS

Independent variables Smaller central staff Larger central staff

Managerial networking 1.382 (2.38) 2.978 (2.03)

Networking squared 0.199 (1.70) 0.412 (1.28)

R2 0.69 0.43
F 203.27 18.12
N 2,256 829

Notes: All equations control for teacher salaries, state aid, class size, teacher experience,
percentage of teachers not certified, percentage of black, Latino, and low-income students,
and yearly dummy variables. T-scores are reported in parentheses.

resources centrally can affect the relationship between networking and

performance. We are interested in whether managers who have a greater
management capacity as indicated by a larger central administrative staff
are able to spend more time networking without seeing the diminishing
returns that are present for the sample as a whole. Again, the basic logic is
that, when top managers can take advantage of internal help to share in
some of the responsibilities, the superintendent is thereby freed up to
spend more time on external management that also can reap performance
dividends.
Table 4.7 presents the findings for these hypotheses. The two groups into
which the sample is split are: those organizations with a leaner adminis-
trative staff (operationalized as the percentage of administrators in the
district who work in the central district office) and those with a more bulky
central administrative staff. The data for this variable has a mean of 1.88
percent. Because the mean is near to two, we split the sample so that the
districts with 1 percent or fewer are in the “low central staff ” category, and
those districts with 3 percent or more are included in the “high central staff ”
group. We analyzed each subset via multiple regression to see whether the
nonlinearities related to central staff and the management–performance
question are as hypothesized.
Our results mostly support these two hypotheses. The relationship
between networking and performance is nonlinear17 for superintendents of
those districts with fewer administrators in the central office (and thus
capable of a lower level of “substitute” or supportive M1 to assist the top
manager), but for districts with more central administrators (and thus a
123 Generalizing the management quality measure

84

Performance 82

High central staff

80
Low central staff

78

76
0 2 4 6
Networking

Figure 4.3 Differences by size of central staff

higher M1 capacity) the relationship between networking and performance

is positive, linear, and substantively larger. Figure 4.3 offers a picture of
the differences in these relationships, with the former “serpent in the
sand” displaying a rather marked curvature, while the latter slithers straight
ahead.
This set of findings offers some real evidence that the effect of network-
ing on performance may be contingent on the administrative capacity
inside the organization, not merely the effort and skill of the networking
top manager. The impacts of public management, in turn, are functions of
the activities and skills of multiple managers; and the behavior of one
person (or, in this case, the effectiveness of the behavior of one person)
may be affected by the presence and activity of other managers in the
organizations – in part through processes of adaptation, delegation, and
backstopping.

Generalizing the management quality measure

Managerial quality measures have been successfully applied in a number of

areas other than Texas school districts. All the studies use regression models
similar to those presented in this chapter; all control for a wide variety of
other factors that could also influence program outcomes. To retain the
focus on the linkage between management and performance, we do not
discuss the control variables but, rather, focus on this key relationship. The
managerial quality studies cover the broadest range of organizations, includ-
ing Columbian local governments, federal government agencies, state
124 Managerial quality and performance

agencies, the executive office of the US president, and even major league
baseball teams (although this last-mentioned group is obviously not a set of
public organizations). Given this wide range of cases, what “quality” means
will vary a great deal. In a developing country, for example, one would not
expect the market to generate an intrinsic measure of managerial quality;
more rudimentary measures of quality will be used.
Avellañeda (2009a, 2009b) uses the basic theoretical model and applies it
to municipal governments in Columbia. Given the location in a developing
country and given that local government in Columbia is not highly profes-
sionalized, her works suggests that the concept of quality is highly portable
to different public sector contexts. Her studies cover forty cities over a
several-year period of time. The measure of mayoral quality – the top
manager is the elected mayor of the city – includes both level of education
and prior service at the local government level. While these would not be
impressive indicators of quality in a developed country, there is substantial
variation on these features in Columbian municipalities. Avellañeda (2009b)
has as her dependent variable the percentage of eligible children who actually
attend school; education is a municipal function in Columbia. Despite an
extensive set of control variables to deal with structure, politics and elec-
tions, socio-demographics, and even citizen displacement by armed groups,
she finds that mayoral education and experience are both positively correl-
ated with educational outputs – although the impact is significantly
dampened in areas with the presence of armed guerrilla groups. Avellañeda
(2009a) relates mayoral quality to the ability to collect greater levels of local
property taxes and per capital social spending. Again, the presence of local
armed groups has a dramatic negative effect on the positive relationships she
finds between managerial quality and program outcomes.
Jacobson, Palus, and Bowling (2010) examine state-level administrators
and the adoption of new public management reforms (essentially the use of
incentives and the limitation of restrictions on managers). Their purpose is
to estimate a model that incorporates both internal management and man-
agerial networking rather than to create a measure of managerial quality;
they also find that the manager’s level of education is positively related to the
adoption of these reforms, however.
Petrovsky (2006) examines federal agencies using the Program Assessment
Rating Tool. PART scores, which were discussed briefly in Chapter 1, are an
effort initiated by the George W. Bush administration to evaluate federal
agencies and programs by using employees of the Office of Management and
Budget (OMB) to do the evaluations. Petrovsky’s strategy is to pull out the
125 Conclusions

management element of the PART scores and use this as a measure of

managerial quality for the agency. There are twenty “Yes” and “No” man-
agement questions scored by OMB personnel that focus on various quality
aspects of the agency’s management. He then uses as a dependent variable a
separate performance score created in a similar way that focuses on goal
attainment, efficiency, and a series of other factors. Petrovsky finds a con-
sistently strong positive correlation between management quality and
agency performance even after correcting for a partisan bias variable.
In a somewhat different type of study using the basic management model,
Hill (2009) attempts to determine the impact of managerial turnover on
performance using a pooled time-series analysis of North American major
league baseball teams. As part of the analysis, Hill creates a reputation-based
measure of managerial quality. His analysis with an extensive set of controls
shows that managerial quality matters for two measures of performance:
whether the team scores more runs than expected and whether the team
wins more games than predicted (Hill 2009: 565).
Cohen, Vaughn, and Villalobos (2010), using the formal theory specified
in Chapter 2, examine the management practices of the chief of staff in the
US Office of the President. Their dependent variable is how effective the
chief of staff is as rated by a survey of experts in the administration.
Although they present two variables – the chief of staff ’s experience and
the chief of staff ’s working relationship with the president – as stability
measures, both are very similar to the quality measures in the Columbian
mayors study. Both measures of experience/quality are positively related to
the perceived effectiveness of the chief of staff.

Conclusions

The analyses in this chapter offer three principal contributions. First, we

develop and apply an uncommon measure of public management quality,
thus fleshing out an aspect of managerial influence that is implicit in our
model but that was not tapped directly with our examination of managerial
networking. The application of the quality measure relies on avoiding an
underspecified model for explaining salary variations, as well as on the
notion that the mobility, information, and compensation for managers in
the empirical setting approximate the labor market assumptions of neoclas-
sical economics. We have argued that both conditions hold here. To the
extent that these conditions do not hold, in fact, we would expect null
126 Managerial quality and performance

results. This chapter, therefore, offers an innovative, albeit indirect, overall

measure of public management quality. The most important limitation here
has to do with the specialized nature of the measure, or, at least, its restricted
applicability. Most settings of interest do not approximate the required
conditions, although investigations of some other situations – certain add-
itional educational systems (see Johansen 2008), some public authorities, or
quasigovernmental entities, for instance – might be able to use and perhaps
improve on the approach taken here.18 Tapping public management quality
in many other circumstances, however, will require tackling more directly
some of the tough issues about what quality means, how it is related to
leadership, and from what sources the requisite quality judgments can be
derived.
Second, this research offers the fullest rigorous test to date of the propos-
ition that public management quality contributes positively to performance.
The results are clear and convincing. If the assumption is made that the
measure of quality is valid, then the almost completely consistent results
across eleven measures of performance are firm evidence indeed.19 That
these results obtain despite any likely measurement error for management
quality creating a bias toward null findings is particularly striking. With all
the appropriate controls for the educational setting, the quality of superin-
tendents’ management makes a difference. Whether the focus is on pass
rates, dropout rates, or the performance of specialized groups of students,
such as those from low-income families or those aiming to attend college,
management matters considerably. This set of results is even more striking
given that the focus here is on only one managerial position – that of
superintendent – at the top of the district’s organizational structure. Since
almost all school systems include additional managerial layers – at a min-
imum, school principals – the overall impact of management is probably
even higher (see Johansen 2008).
The relationship of management quality to performance is also likely to
be complex. The impact of management quality appears to be partially
channeled through the mobilization of parental and community support.
In addition to contributing directly to operations, then, and to dealing with
political principals and external regulatory authorities, managers appear to
contribute to performance by mobilizing the efforts of others who have
allied interests in delivering results at the local level.
Third, managerial networking contributes positively to performance, but
the returns diminish at higher levels of networking. This finding is consist-
ent with our notions that managers apportion time and effort across
127 Conclusions

managerial functions, presumably because multiple contributions are

needed, and the distribution of managerial time and effort is likely driven
in part by managerial assessments of marginal returns to performance. It is
also consistent with a conceptualization of public management as goal-
directed activity aimed at coordinating people and resources to produce
public value. Almost any model of the manager as an intendedly rational
actor would imply a consideration of just such tradeoffs.
Even more interesting is the additional set of findings concerning the func-
tional form of this relationship. How does managerial quality mediate the
relationship? As the results displayed in Table 4.6 and Figure 4.2 attest, skillful
managers are able to avoid diminishing returns; and additional evidence
suggests that they do so, at least in part, by economizing on their investment
in external interactions – thereby avoiding too much of a good thing. High-
quality management thus contributes directly – and also via its influence on the
shaping of at least one important, externally directed management activity.
These findings are also consistent with the argument supporting the
model, but it is helpful to note that they focus only on the behavior and
influence of the top manager. The justification for the model, plus virtually
the entire literature of the field, indicates that contributions of additional
actors supporting the management function should also matter. While we
do not test this proposition directly, we do so obliquely by examining the
extent to which central staff capacity supportive of top management can
reduce the internal/external tradeoff for school district superintendents. The
findings (Table 4.7 and Figure 4.3) support the argument that central staff
can buttress performance by supporting and contributing to internal man-
agement, thereby eliminating – or, at least, substantially reducing – the need
for top managers to balance their internal and external responsibilities. By
contrast, low central staff capacity can mean not only diminishing returns
but, past a certain point, negative returns on performance from additions to
external managerial effort. Bulking up with central staff should not be
considered mere administrative padding, however; staff do entail fixed costs,
and these need to be considered in managerial decisions. When we incorpor-
ate a full set of controls regarding resources and constraints, though, our
analyses show that significant staff capacity is supportive of performance,
and some of this contribution is apparent in the way that others, particularly
top managers, are able to add performance value. Substantial central staff
capacity alters the pattern of the externally conjured “serpent in the sand.”
All these findings are consistent with themes developed by astute obser-
vers and analysts of public management. That public management quality
128 Managerial quality and performance

matters, of course, is hardly news to specialists in public management and

public policy. The field of public administration has developed a rich
literature arguing for this notion; but to find management quality influ-
encing performance directly, and consistently, in a data set spanning hun-
dreds of governments over a several-year period is particularly persuasive
evidence. While this chapter reports in detail on only one set of adminis-
trative units – and in one policy field, at one level of government, and, in
fact, at only one level of management within that set – it indicates clear
support for theoretical arguments that have been articulated for years by
scholars and practitioners in the field. Any who doubt the importance of
management and managerial quality for what can be delivered by public
education in the United States should note the implications of this analysis
for identifying a critically important point of leverage: in Texas, at a min-
imum, public management quality itself, not simply influences such as
district spending or students’ home circumstances, makes a difference.
Furthermore, the consistent support for the importance of managerial
quality in a number of other recent studies far removed from education in
Texas suggests the general nature of the relationship.
In many respects, the results of this set of investigations are suggestive.
Considerable additional work is warranted. The links between public
management in its various additional guises and the results of interest to
multiple stakeholders need to be explored more thoroughly, as is done in
other chapters of this book. Even more work needs to focus on settings
beyond those of public education. Considering the issues in the broadest
possible context will require additional conceptual, theoretical, and meas-
urement advances. More complex models need to be considered, as a part of
this agenda. In addition, the implications for public managers themselves
need to be unpacked more fully. In short, even if it can now be argued
with persuasive evidence that the quality of public management shapes
policy outputs, most of the important challenges remain to be addressed.

NOTES

1. Much of the material covered in this chapter has been presented by Meier and O’Toole
(2002). We have redone the analysis to include managerial networking in the equations
and have also included analysis by Hicklin, O’Toole, and Meier (2008).
2. Alternatively, it might be tautological: good management is whatever seems retrospectively
to have produced good results – a special problem when perceptual measures of both
management and performance are used (see Wolf 1993).
129 Notes

3. Since 1999 superintendent salaries in the large districts have approached $400,000.
4. In the next chapter we analyze another stability-relevant theme: personnel and manager-
ial stability, and how these influence performance.
5. The two are moderately correlated, but the relationship is not strong enough to pose a
collinearity problem.
6. In analyses in subsequent chapters in which managerial quality is included as an
hypothesized influence on performance, we use a measure calculated in the same way.
Recalculating the residuals-based measure with other, more recent, time series data
produces an even higher predictive ability in the salaries’ equation.
7. The number of candidates for inclusion in education production functions is virtually
limitless. Because many of the variables measure the same thing or relatively similar
things, collinearity in these models is a problem. As a result, some variables may have an
inappropriate sign. Because our concern is with having sufficient controls in the model
rather than estimating the precise impact of each control variable, we are less concerned
with collinearity.
8. In a 2000 survey of superintendents by the authors, 45.5 percent of superintendents rated
TAAS scores as their top priority; no other goal was endorsed by more than 13 percent of
superintendents. An additional 46.8 percent of superintendents rated TAAS scores as
“very important.” Surveys in later years show an even higher priority for TAAS or its
replacement.
9. The specific question asked the superintendent to rate parental involvement and com-
munity support on five-point scales that ranged from excellent to inadequate. This 2000
survey had a 57 percent response rate, thus reducing the total number of cases for
analysis.
10. A path analysis of the results of this second analysis shows that 72 percent of the impact
of management quality is direct, with the other 28 percent indirect through community
and school board support.
11. Note that we are limiting our analyses to linear specifications in this portion of the
chapter. We recognize that our measure of managerial quality is likely to be controversial.
To provide focus on the management quality measure, therefore, we have opted for
relatively simple models of management. As explained early in this book, we believe that
management operates in a contingent and nonlinear manner conditioned by structural
context. Some of the chapters include analyses involving more complex models of
management.
12. When community and school board support are included in the equations, the results are
similar, except that the impact of management on black TAAS scores is no longer
significant.
13. The n-size varies for these equations because the state reports results only when five or
more students per district meet the category. Some districts, for example, do not have
sufficient minority students to generate results.
14. The most likely answer of all is: from both.
15. The relationship between managerial quality and performance is not subject to these
diminishing marginal returns and has a strictly linear relationship.
16. The difference in years reflects when the original analysis was done. Replicating all the
analysis with 1995–2002 data produced similar results.
130 Managerial quality and performance

17. The squared term is significant only at the 0.10 level.

18. We believe that this approach will work for other management positions subject to
competitive markets for which measures of program performance are available. Munici-
pal agency heads such as fire chiefs, police chiefs, and public works heads may fit these
conditions in some jurisdictions.
19. Alternatively, if one views this empirical study as a check on external validity, as
explained earlier, the results are highly encouraging.
5 Internal management and performance:
stability, human resources, and
decision making

We have now seen evidence that managerial networking helps to shape

organizational performance and also that the quality of management makes
a difference for outputs and outcomes. We have explored these matters first
in this book because we tackled them early in our research program. Much
of what public managers do involves activity and decisions within the
organization, however. This set of possible managerial efforts is represented
by the M1 term in our model introduced in Chapter 2:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 Xt =Se ðM3 =M4 Þ þ et
In the model, the term seems straightforward enough, but that appearance
can be deceptive. The reason is that most of what has conventionally been
treated as public management consists of various aspects of internal man-
agement – management within and of the organization – rather than actions
directed externally. The forms of internal management are myriad. They
include the various aspects of managing people, or human resources man-
agement – such as hiring, orienting new personnel, classifying positions,
defining jobs, retaining and promoting people, disciplining and even firing
employees, counseling staff, crafting training and development programs,
handling grievances and other complaints, resolving interpersonal disputes,
dealing with issues of diversity, and much more. Internal management also
involves many other functions, including managing and allocating budgets,
operating financial management systems, structuring and restructuring
work units and tasks, motivating personnel, shaping the culture of the
organization, and – in some organizations – using performance data to help
manage various of these other managerial operations.
Internal management, in short, is not a simple and undifferentiated
function but, instead, encompasses the bulk of the managerial functions
mentioned in Chapter 1. It would probably be a job of a lifetime, or even
more than that, to try to estimate the independent and combined effects of
all these varied and complex aspects of internal management on public
131
132 Internal management and performance

organizational performance. No researchers have taken on this Herculean

task; and, even if some do, chances are that they would have modest success
at best. What we have done is to tackle three limited and yet substantively
significant parts of internal management to explore its performance results:
maintaining organizational stability, managing human resources, and
making effective decisions.
We explore these aspects of internal management because of their obvious
importance in virtually all organizations. Given the kinds of organizations
involved in public education and the nature of our core data set, we can
expect the internal management of school districts to be a critical element in
their overall performance. Three aspects of internal management can be
distinguished. One is the longevity, or stability, of staff and management in
an organization. For various reasons, sheer stability in terms of the human
capital in an organization might be expected to boost performance (but see
Meier and Hicklin 2008). A second is the quality of human capital in the
organization and its effective management. Quality and the management of
it should contribute to results, for obvious reasons. After all, it is not only
the quality of top management overall that can be important, but also the
quality of the workforce – and the efforts to develop it. A third aspect is the
ability to make good decisions, particularly when faced with major prob-
lems. To examine this situation, we analyze what school districts do when
faced with a budget crisis. What actions do they take to limit the impact of
budget cuts on organizational performance?
These matters do not completely cover the full range of internal manage-
ment issues and challenges, of course, and they certainly do not reach many
other aspects of internal management that can be important in public
organizations.1 Exploring them should nonetheless provide some insights
regarding the role of internal management in helping to shape results.

Stability, change, and the performance of public organizations

Ideas swing wildly in and out of fashion; take, for instance, the notion of
stability in administrative organizations. For decades conventional wisdom
simply assumed that stability contributes to public management perform-
ance. Such core bureaucratic features as standard operating procedures,
regular structure, incremental decision making, and fixed rules are emblem-
atic of the persisting features of such organizations. At least since the early
years of the last century, scholars linked stable patterns to efficient
133 Stability and public management: out of fashion

functioning. As explained in Chapter 2, stability has often been seen as

virtually the sine qua non of bureaucracy (Gerth and Mills 1958: 228).2
The literature on public management in recent decades has heavily
emphasized the contrary themes of organizational change, adaptability,
entrepreneurship, and reform, however. The term “bureaucracy” has
become equated with stodgy, hidebound, and inefficient operations. Much
emphasis among recent proponents of good government has been on ways
to escape from or banish bureaucracy (Osborne and Plastrik 1997) – and
move toward alternative forms and processes.
In this section of the chapter, we begin a systematic empirical exploration
of the link between an aspect of stability and public organizational perform-
ance. We use the Texas school district data set to do so. We establish a basis
for this inquiry by reviewing features of the relevant literatures, clarifying the
core concept, and narrowing the empirical focus to one aspect of stability:
constancy in personnel.

Stability and public management: an out-of-fashion statement

Our analysis of personnel stability in public organizations can be put into a

broader context: the general role of stability in administrative systems. Few
ideas these days seem as retrograde as the quaint notion that stability can be
helpful in the world of public management. One need only look as far as the
high-visibility “Innovations in government” project supported over many
years by the Ford Foundation and centered at the Kennedy School of Govern-
ment at Harvard. Similarly, the National Performance Review during the
Clinton–Gore years emphasized the advantages of freeing agencies from the
heavy hand of extant structures and processes. Successive waves of adminis-
trative reforms have stressed different and sometimes conflicting “solutions”
to a broad set of managerial challenges, with the common premise that a
disruption of existing patterns was beneficial (Light 1997). The impetus for
innovation is just as lively within particular policy fields. Educational policy,
for example, has emphasized curricular changes, the introduction of new
technologies, the recruitment of new teachers, and new kinds of teachers,
along with an assortment of organizational and incentive-based experiments
ranging from high-stakes testing to merit pay to charter and magnet schools
(Elmore 1997). Nothing seems hotter than novelty.
Scholarly experts in public management have similarly pushed themes
that critique stability. Organizational change, organizational development
134 Internal management and performance

and renewal, planned change, and, of course, the range of efforts to spark a
new public management in many countries have been of intense interest to
researchers, particularly those desirous of “breaking through bureaucracy”
(Barzelay and Armajani 1992). Research has followed the manifold govern-
mental efforts to innovate and reform, with particular attention to deter-
minants of innovation and how to develop and institutionalize change (for a
review of these themes and some of this literature, see Rainey 2003, espe-
cially 355–89; for a comparative cross-national consideration of NPM, see
Barzelay 2001). While innovation has produced considerably less than its
strongest proponents claim (Light 1998), the emphasis remains clearly
positive (see Altshuler and Behn 1997). Some work has been influential
among practitioner and research audiences alike (Osborne and Gaebler
1992), and additional dimensions of the innovations theme continue to be
advocated (Behn 2001). Stability, in contrast, rusts at the bottom of the
public manager’s toolbox.
Despite the attention given to change, reform, and entrepreneurship,
some dissents can be heard. Terry (2002) has critiqued the perspectives
sketched above and argues strongly, instead, for the importance of “adminis-
trative conservatorship,” whereby administrative leadership cultivates and
protects the core competences, values, and institutional elements of agency
life that are accumulated over extended periods. Administrative executives,
Terry indicates, are “conservators because they are entrusted with the respon-
sibility of preserving the integrity of public bureaucracies and, in turn,
the values and traditions of the American constitutional regime” (Terry
2002: xv, emphasis in original; see also Spicer and Terry 1993).
Distinct but similar arguments have been offered by a disparate set of
scholars. The Blacksburg group has emphasized the importance of a long-
term “agency perspective” that may serve as a guarantor of the public interest
(Wamsley et al. 1990). Kaufman, particularly in his classic The Forest Ranger
(1960), shows administrative routines and ingrained patterns of oversight to
be important “centripetal forces” that lend coherence to an otherwise chaotic
policy setting rife with opportunities for atomistic decision making.
Indeed, motivating the study reported here is the notion that stability is
not necessarily the bane of those committed to high performance, but can
offer opportunities for enhanced program achievement. Although this prop-
osition had been a truism of standard organization theory (see Perrow
1986), it seems to have become lost in the rush to embrace entrepreneurial
notions of public management, the enactment of innovations of all sorts,
and various forms of reinvention and change.3
135 Stability in administrative systems

Stability in administrative systems

As we have observed several times, administrative systems are fundamentally

inertial: once put into operation, they tend toward stability. Earlier we noted
that stability means, quite simply, constancy in the design, functioning, and
direction of an administrative system over time. Administrative stability can
be seen along a number of related but distinguishable dimensions.4
Structural stability: the preservation of organizational features over time.
Structural stability itself is multidimensional and includes such elem-
ents as size, formalization, differentiation, span of control, and so
forth.
Mission stability: the consistency over time of the goals of an adminis-
trative unit. When bureaus are asked to change course with frequency,
they may experience disruptions. One of the distinctive features of
public agencies, furthermore, is that their mission is for the most part
externally determined (Wilson 1989): policy changes, as established by
political executives, legislatures, and/or judicial determinations, exert
profound impacts on the missions of agencies and therefore on the
stability these units experience.
Production or technology stability: Lynn, Heinrich, and Hill (2000) con-
tend that governance systems are characterized by a mode of produc-
tion or type of technology, and that altering the form of production
essentially shifts governance arrangements.5 Analysts of public admin-
istration have long been aware of the importance of agency technology,
particularly “core technology” (Thompson 1967).
Procedural stability: Related to production but distinct from it is the set of
rules, regulations, and standard operating procedures used in a public
agency. Units pursuing the same missions with similar technologies
sometimes develop quite different procedures for getting the job done.
Welfare-to-work programs illustrate this variation across the United
States and even across offices within a given state (Sandfort 1999).
Stable procedures create opportunities for coordinating action across
large numbers of individuals without overwhelming their capacity
(Allison 1971).
Personnel stability: The types of stability mentioned above all deal with
features of the administrative system. Bureaucracy, according to Weber,
is characterized by career employees, so an additional element of stable
administration is the people who occupy positions within the
136 Internal management and performance

organization. If the positions and/or their relationships shift over time,

a system experiences instability. Even if the structural and procedural
aspects remain constant and the goal of a public agency persists,
however, changes in the personnel themselves can represent an import-
ant variety of instability.

Personnel stability

Personnel stability is the focus of empirical analysis in this section of the

chapter. Selection criteria, motivation, and agency incentive systems have
often received scholarly attention (Ban and Riccucci 2002). The stability of
the personnel over time has been much less frequently explored, though.
One reason may be the classic view that individuals within the system are
“career” bureaucrats.6
Leadership stability has been a concern, however; a point of persistent
tension between US politicos and career appointees is the relative imper-
manence of the former, who constitute, in the famous phrase of Heclo
(1977), a “government of strangers.” Turnover among politicos has exacer-
bated the difficulties involved in building competence, mutual trust, and
long-term commitment (see Dunn 1997).
Other aspects of personnel instability have received some attention
recently. A few jurisdictions have abandoned commitments to lifetime merit
appointments in favor of flexibility and responsiveness; the state of Georgia,
for instance, no longer offers job protection to new employees. At the
national level, analysts have noted that careers in public service have become
more varied. Those seeking such careers, particularly individuals with
advanced degrees in public affairs, now work in the private and nonprofit
sectors as well as in government; they are much more likely to change
agencies, organizations, and even sectors several times over the course of a
career (Light 1999). While these flows of human resources can bring
fresh perspectives to public organizations, they may also engender
complications.
Personnel experts sometimes express concern about burnout-generated
turnover (Golembiewski 1990). In public education itself, personnel shortages
in key fields such as mathematics and science have made headlines in recent
years, and teacher burnout has been the object of policy change in a number of
jurisdictions. Educational system administrators have been increasingly diffi-
cult to recruit for extended tours of duty (Hess 1999). Ironically, then, even as
137 Personnel stability

public management and public education press for change, concerns have been
raised about the performance consequences of personnel instability.
In this section we explore this issue by estimating models of how personnel
stability influences school district performance in Texas. We are interested
especially in two forms of personnel stability: the durability of top-level
public managers and the retention of front-line teaching professionals.
Why should either type of stability matter? Top managers navigate in a
complex environment; they need time to learn the basic demands of the job.
Assessing the surroundings, both inside and outside the administrative
system, can take time. Even the most skillful managers can be expected to
improve efficacy by learning their institutional, political, resource, personal,
and administrative contexts. Time also permits other stakeholders to ascer-
tain top management’s intent and style. Over an extended period, and
particularly among managers who do high-quality work, this familiarity
can breed trust. Top managers who have developed reputations can use
longevity to exercise power.7
Stability alone, of course, is not management’s sole contribution to
program performance. The quality of management exhibited by a given
manager can be expected to have an impact. Moreover, the sheer degree of
managerial activity devoted to monitoring and negotiating the external
environment can also pay off, as we have seen earlier. Accordingly, the
analysis reported below considers these additional dimensions of manage-
ment as well as the issue of stability.
Front-line professional workers, particularly in so-called “street-level bur-
eaucracies” (Lipsky 1980) that deal with unpredictable needs and demands
from clients, can also benefit from longer periods on the job. Extended time
in high-stress front-line positions can lead to burnout and departure, of
course; but, for those who endure, the multifaceted skills acquired in the
“trenches” can make a significant difference in performance. Classroom
teaching surely fits this pattern. Veteran teachers learn how to juggle the
many tasks involved in delivering high-quality instruction. They gradually
see how to translate pedagogical theories into workable practices in their
own particular setting. They also can learn over time how to sort through
the distractions that can absorb energy and attention during a school
day. They will have developed experience with difficult cases and multicul-
tural nuances. Many of these craft-like skills, developed through years of
experience, are only partially transferable to other districts with different
mixes of students and different curricula. Sheer time in position in a local
setting can help.
138 Internal management and performance

These characterizations offer a rationale for the hypothesis that personnel

stability on the part of both top-level managers and front-line workers can
assist in delivering program outputs. In the next subsection we place this
stability hypothesis into our more general model of public management and
public program performance.

Modeling the impact of stability

In considering the hypothesized impact of personnel stability – and stability

more generally – upon the performance of an administrative system, we
begin with our general model:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 ðXt =Se ÞðM3 =M4 Þ þ et
In the model, the Ss are composites of the various kinds of stability outlined
earlier: structural stability, mission stability, and so forth. More stability
means that current operations in an agency have more impact on future
performance – that is, the larger the impact of the autoregressive term and
the smaller the impact of the second, or “environmental,” term.
Although we expect many kinds of stability to matter for public agencies,
the focus here is on personnel stability, or SP. Other aspects of stability
within public organizations, such as school systems, are also worthy of
investigation, when appropriate measures for them are available.8 “Person-
nel stability” itself is something of an abstraction, in that different kinds of
personnel can exhibit different degrees of stability. This analysis focuses on
personnel stability for two kinds of system employees: top system managers
(school district superintendents) and front-line workers (teachers).
Although the model treats stability as something separate from public
management, it is clear that management itself influences personnel stability.
In certain important respects, therefore, personnel stability is another aspect
of M, or management – primarily internal management, or M1. Since the terms
S (including SP) and M1 are additive and thus substitutable in our formaliza-
tion of the theory, it makes no difference for estimation purposes as to whether
we regard personnel stability as stability or as management. Clearly, it is both.
In analyzing the impact of personnel stability on performance, we also
include our measures of managerial quality and managerial networking; we
already have seen considerable evidence that they shape outputs and out-
comes. Since probing for multiple nonlinear impacts of personnel stability
via a single research design is impractical, we simplify the model further for
139 Modeling the impact of stability

present purposes by retaining stability only in the first term of the model.
The upshot is a deliberately underspecified model designed to explore some
of the issues raised in the general formulation. With these adjustments, the
model reduces to
Ot ¼ b1 ðSP þ MQ ÞOt1 þ b2 ðM2 Xt Þ þ et
The particular interest in this investigation is personnel stability – both of
top managers and of front-line workers. Since the autoregressive form
means that a large part of any variance is likely to be explained by the lagged
dependent variable Ot1, it can be difficult to pick up the influence of other
variables. Accordingly, we test both this model and a further simplified form
that excludes the lagged dependent variable:
Ot ¼ b1 SP þ b2 MQ þ b3 ðM2 Xt Þ þ et
Both equations are used in the analysis here. Our usual set of eight appro-
priate controls, represented by Xt, a vector of environmental forces, is
included in the analysis.
Once again we use data drawn from a set of Texas school districts. We
incorporate superintendent responses from our initial survey and pooled
five years (1995 to 1999) of data on performance and control variables to
produce a total of 2,535 cases for analysis.
Personnel stability can be a recurring issue in such districts. School
district managerial talent is mobile within the state (and somewhat mobile
across states). While some superintendents remain for extended periods in
one locale, most individuals move among several districts as they pursue
their careers. Further, districts themselves seek to replace their superintend-
ents for various reasons. This movement inevitably means instability at top
managerial levels; those in the system must adjust to a new top manager and
that person’s influence on a range of district decisions. Indeed, shifts in the
top managers often trigger other personnel changes near the top, among
deputies, assistant superintendents, principals, and so forth. The average
tenure of superintendents in Texas is 5.3 years in their positions; their mean
tenure within the district in any capacity is 8.7 years.
Stability in the teacher corps should matter as well. Inexperienced teachers
are likely to be less effective as they engage in trial-and-error searches to
determine which of their academic skills actually matter in the classroom.
The teacher shortage in particular specialties compounds the difficulty. To
recruit new teachers on a regular basis, at the very least, school systems must
devote significant budgetary resources to human resources management.
140 Internal management and performance

Some systems find it necessary to take particularly sizable and costly efforts;
the Houston Independent School District, for instance, employs a recruiter
in Moscow, Russia. The average period of teacher experience in Texas school
districts is 11.6 years.

Measures of personnel stability

We examine two measures of personnel stability: one for school district
superintendents (managerial stability) and a second for school district
teachers (teacher stability). Managerial stability is simply the number of
years the superintendent has been employed by the district in any capacity.9
Teacher stability is measured as the percentage of teachers employed by the
district during the preceding year who continue to work for the district this
year. In other words, teacher stability is measured as 100 minus the year’s
turnover rate. The mean for this variable is 85.1 percent with a standard
deviation of 8.1.10 For both measures, higher scores mean more stability.
Interestingly, the two forms of stability are unrelated to each other empiric-
ally (the correlation is 0.09), thus suggesting that different forces shape
personnel patterns in these two loci.

Other measures
Two measures of public management are included as potential explanatory
variables in this analysis: managerial quality and managerial networking.
Both measures have been introduced in earlier chapters. Eight control
variables, all introduced earlier, are included; three are measures of task
difficulty and five measures of resources.
We use ten performance measures in this portion of our analysis: the
overall student pass rate on the Texas Assessment of Academic Skills, TAAS
scores for Anglo, black, Latino and low-income students, three measures of
college-bound student performance (average ACT score, average SAT score,
and the percentage of students who score above 1,110 on the SAT (or its
ACT equivalent)), along with attendance rates and dropout rates.

Findings

The first line of analysis here focuses on explaining the overall TAAS pass
rate. Table 5.1 displays abridged results of regression analyses for two
141 Findings

Table 5.1 The impact of management on performance: standardized tests II

Dependent variable ¼ TAAS pass rate

Independent variable Slope T-score Slope T-score


Managerial networking 0.6846 4.58 0.1977 1.96
Management quality 0.9182 5.57 0.1732 1.56
Stability – teachers 0.1374 5.53 0.0511 3.05
Stability – managers 0.0739 4.51 0.0251 2.27
Lagged pass rate – 0.7083 55.04

R2 0.61 0.82
Standard error 7.43 4.99
F 239.58 678.43
N 2,503 2,503

Notes: Coefficients for annual dummy variables and control variables (teacher salaries,
state aid, class size, teacher experience, noncertified teachers, and the percentages of black,
Latino, and low-income students) are not shown.  ¼ significant at p < 0.05, one tailed test.
Time period ¼ 1995–9.

models, the first omitting the autoregressive term and the second including
it.11 The tables omit the coefficients for the control variables.
In each equation, both measures of personnel stability are positively and
significantly related to school district performance. The impact of teacher
stability is slightly more than that for managerial stability in both cases.
While clearly not the most important determinant of districts’ standardized
test performance, both kinds of stability contribute to the explanation.
Since the measure of teacher stability in the sample ranges between 44.4
(55.6 percent of a district’s corps of teachers departed in one year)12 and 100
(zero turnover), the maximum impact of teacher stability is considerable:
more than 7.6 percentage points on a district’s pass rate, if the equation
without the lagged dependent variable is used, and almost three percentage
points even in the much more stringent autoregressive specification.13
Managerial stability also contributes to district performance. Here the
measure is in years of experience in the district; since the range in the data
set is forty-one years, the maximum impact of this feature is less but still
worth noting: approximately three percentage points or one percentage
point on the pass rate, respectively, depending on whether the equation
excludes or includes the lagged pass rate. Since the two elements of stability
are uncorrelated, the combined maximum impact of stability could amount
to almost eleven percentage points.14 Considering the short shrift given such
142 Internal management and performance

unglamorous organizational features as stability in recent years, these posi-

tive performance impacts are quite remarkable. Clearly, some enhancement
in outcome is due to the leverage gained by retaining those who know
the system and applying their talents to the educational and managerial
tasks at hand.
In addition, management itself is positively and significantly related to
district performance on pass rates. In the first equation reported in Table 5.1,
both measures of management boost performance, beyond the impact
attributable to personnel stability. Since both management measures are
standardized and thus range between approximately 3 and þ3, these
equations suggest that high-quality management’s maximum impact is
approximately 5.5 points on the TAAS, and managerial networking’s max-
imum impact is approximately 4.1 points. Even while taking into account
the stability impacts, therefore, management quality and networking influ-
ence district pass rates. For the equation including the lagged dependent
variable, managerial networking continues to have a positive and statistically
significant impact while management quality has the correct sign but just
misses the 0.05 threshold for statistical significance.
Taken as a whole, the analyses reported in Table 5.1 constitute evidence
that management and stability affect school district performance. The fact
that even the autoregressive form of the production function continues to
show the importance of personnel stability as well as management is a
particularly impressive demonstration.
The overall pass rate on the TAAS exam is an important and salient measure
of school district performance, but not the only one. Table 5.2 presents the
regression coefficients for the four stability and management variables for the
nine additional performance indicators. Table 5.3 repeats the same analyses but
with the inclusion in each case of the lagged dependent variable.
The results in Table 5.2 support our theoretical arguments regarding
stability and confirm the importance of management as well. Of the
thirty-six coefficients reported, all but three are properly signed; approxi-
mately two-thirds are statistically significant.15 In every analysis, one or
more of the stability and/or management variables are statistically signifi-
cant. In seven of the nine additional analyses, one or both measures of
personnel stability are statistically significant; in eight of the nine analyses,
the same can be said for one or both of the management measures. Both
stability and management contribute positively to performance in statistic-
ally significant ways in all but three estimations. This pattern is strong
evidence that personnel stability contributes to educational performance,
143 Findings

Table 5.2 Management, stability, and other measures of performance

Management Stability

Performance measure Networking Quality Teacher Manager R2 N

Latino pass percentage 0.5394 (2.07) 0.7722 (2.74) 0.0634 (1.37) 0.0659 (2.36) 0.36 2,283

Black pass percentage 0.4920 (1.29) 0.5691 (1.47) 0.3417 (4.54) 0.1093 (2.93) 0.37 1,548
  
Anglo pass percentage 0.7728 (5.20) 0.7986 (4.88) 0.1347 (5.44) 0.0781 (4.79) 0.45 2,475
Low-income pass 0.2693 (1.42) 0.9401 (4.48) 0.1345 (4.20) 0.1265 (6.07) 0.52 2,492
percentage
Average ACT score 0.0445 (1.64) 0.0932 (3.14) 0.0072 (1.41) 0.0018 (0.64) 0.36 2,196

Average SAT score 3.7900 (2.59) 3.6348 (2.28) 0.8699 (2.96) 0.1592 (1.05) 0.50 1,814
Percentage above 1,110 0.3743 (1.89) 0.7481 (3.40) 0.0636 (1.81) 0.0185 (0.86) 0.30 2,387
Dropout percentage 0.0368 (1.76) 0.1014 (4.40) 0.0006 (0.17) 0.0016 (0.71) 0.16 2,483
Class attendance 0.0015 (0.10) 0.0655 (3.89) 0.0120 (4.71) 0.0000 (0.02) 0.24 2,503

Notes: T-scores in parentheses. All equations control for teacher salaries, instructional expenditures per
student, class size, teacher experience, percentage of teachers not certified, percentage of black, Latino, and
low-income students, and yearly dummy variables.  ¼ significant at p < 0.05, one tailed test.

Table 5.3 Management, stability, and performance: autoregressive estimation

Management Stability

Performance measure Networking Quality Teacher Manager R2 N

Latino pass percentage 0.2787 (1.32) 0.3523 (1.56) 0.0148 (0.39) 0.0377 (1.68) 0.58 2,216

Black pass percentage 0.2640 (0.87) 0.3989 (1.29) 0.1254 (2.02) 0.0519 (1.73) 0.59 1,490
 
Anglo pass percentage 0.3110 (2.85) 0.1909 (1.59) 0.0576 (3.16) 0.0336 (2.82) 0.71 2,467

Low-income pass 0.2028 (1.47) 0.1580 (1.03) 0.0700 (2.99) 0.0543 (3.56) 0.75 2,491
percentage
Average ACT score 0.0346 (1.35) 0.0631 (2.27) 0.0055 (1.11) 0.0001 (0.05) 0.45 2,091
Average SAT score 1.0705 (0.87) 1.3360 (1.01) 0.4979 (1.94) 0.0893 (0.72) 0.67 1,655
Percentage above 1,110 0.2458 (1.33) 0.4301 (2.09) 0.0498 (1.51) 0.0134 (0.67) 0.39 2,366
Dropout percentage 0.0134 (0.72) 0.0669 (3.27) 0.0014 (0.45) 0.0006 (0.31) 0.34 2,483
Class attendance 0.0013 (0.14) 0.0126 (1.25) 0.0066 (4.36) 0.0004 (0.44) 0.71 2,503

Notes: T-scores in parentheses. All equations control for teacher salaries, instructional expenditures per
student, class size, teacher experience, percentage of teachers not certified, percentage of black, Latino, and
low-income students, a lagged dependent variable, and yearly dummy variables.  ¼ significant at p < 0.05,
one tailed test.

and also clearly supports the notion that management itself matters, aside
from the personnel impacts captured by stability.
Table 5.3 shows the coefficients for the nine autoregressive analyses.
Picking up the impacts of independent variables is quite difficult in such
144 Internal management and performance

cases; the results further support the importance of personnel stability,

however. Here thirty-four of the thirty-six coefficients are correctly signed,
and all equations show one or more of the independent variables as statis-
tically significant. Half the personnel stability coefficients are statistically
significant, and two-thirds of the nine equations show personnel stability in
some form as statistically significant. The management measures do not
perform as well, particularly managerial networking (see Chapter 3 on the
distributive aspects of M2). Nonetheless, management influences perform-
ance in positive and statistically significant fashions for one-half of the ten
performance measures. Indeed, if the criterion for statistical significance is
relaxed to p < 0.10, twenty-five of the forty coefficients in the autoregressive
estimations are significant. Given the stringency of the tests used to produce
the results displayed in Table 5.3, and given the range of performance
indicators considered, the results are noteworthy – especially regarding
personnel stability.
Some of the details clearly deserve more careful attention and further
analysis; one example can suffice. The analysis for pass rates among black
students indicates that a maximum of almost seven points in a district’s pass
rate can be explained by teacher stability, even controlling for the lagged
dependent variable. This level of impact for a regular cadre of teachers on
such students is phenomenal. It may be that students from particular family
backgrounds or certain circumstances are especially sensitive to stability
among the role models and mentors in their midst at school,16 or perhaps
experience matters in adjusting pedagogy to the needs of individual stu-
dents. If so, policy makers and educational administrators would do well to
attend to some of these differential impacts.
In short, the results of this study support the notion that personnel stability
can be an important determinant of public organizational performance, at
least for education; and that the impact of stability can be particularly
strong for certain measures and/or clients of public organizations. Given the
presumed importance of teachers themselves in the educational process,
that stability among teachers would be more important than that at the level
of top management in school districts makes sense; and, indeed, the strength
of the relationship between stability and performance for teachers exceeds
that for district superintendents in most but not all of the ten estimations.
Nevertheless, stability at the top seems often to matter as well. This
finding is noteworthy for at least two reasons. First, superintendents are
insulated from the classroom, so to be able to demonstrate such impacts on
performance from this level in the organization constitutes a particularly
145 Nonlinear relationships

telling result. Second, other relevant aspects of management are already

controlled for in the analysis. In particular, managers’ activity in the network
of external parties is already taken into account, as is management quality. It
would appear, then, that managerial stability per se can be helpful.
This finding should be treated with appropriate caution. Sometimes
organizational change can be a good thing, especially in underperforming
organizations. The results here pertain only to personnel stability; they
cannot necessarily be generalized to all forms of stability, though they
certainly render the broader stability hypothesis more intriguing. The evi-
dence, furthermore, does not support a sweeping castigation of all personnel
changes. The stability-related findings hold after controlling for managerial
quality. They surely do not buttress any argument that bad managers (or, for
that matter, poor teachers) should be retained merely to keep things con-
stant. Further, the value of personnel stability might actually lie in part in the
abilities of experienced, knowledgeable, and widely respected people – both
teachers and superintendents in the case of school districts – to initiate and
implement some of the right kinds of changes at the right time.17 All in all,
then, these findings support carefully framed arguments for personnel
stability but do not promote an uncritical conservatism.

Nonlinear relationships

All findings reported thus far involve linear estimations for the full set of
school districts. We are interested in exploring nonlinear elements as well,
however, since both the theory and some earlier analyses indicate that these
can be expected when dealing with public managerial and related relation-
ships. Nonlinear impacts can be assessed either via interaction terms or by
examining relationships with different subsets of the sample. Here we
explore nonlinear relationships among the independent variables via phys-
ical controls – that is, by partitioning the data set.
We divide the school districts into quartiles four different times – by
values of each of the key independent variables successively (that is, man-
agement quality, managerial networking, teacher stability, and managerial
stability) – and examine changes in the impacts of the other independent
variables on performance.18 For each quartile of partitioning and each of the
independent variables, we calculate the regression coefficients for the other
variables. The coefficients can then be graphed, as in Figures 5.1 to 5.4. Each
figure shows the full set of coefficients for each quartile of the partitioned
146 Internal management and performance

1.400

1.200

1.000

0.800

0.600

0.400

0.200

0.000
Q1 Q2 Q3 Q4
Regression slopes for
Superintendent stability Teacher stability Networking

Figure 5.1 The interaction of management quality with networking and stability: quartile regression
coefficients

0.160

0.140

0.120

0.100

0.080

0.060

0.040

0.020

0.000
Q1 Q2 Q3 Q4
Regression slopes for
Teacher stability Networking Management quality

Figure 5.2 The interaction of superintendent stability with teacher stability, management quality, and
networking
147 Nonlinear relationships

2.000

1.500

1.000

0.500

0.000
Q1 Q2 Q3 Q4

–0.500

–0.100

Regression slopes for

Superintendent stability Teacher stability Management quality

Figure 5.3 The interaction of networking with management quality and stability

0.300

0.250

0.200

0.150

0.100

0.050

0.000
Q1 Q2 Q3 Q4

Regression slopes for

Superintendent stability Networking Management quality

Figure 5.4 The interaction of teacher stability with superintendent stability, management quality,
and networking
148 Internal management and performance

variable: twelve coefficients in each figure. To illustrate, in Figure 5.1 the

solid line shows the size of the networking coefficient for each quartile of the
management quality variable. For the lowest quartile (those near the bottom
in managerial quality), the coefficient is 0.83; for Q2 the figure drops to 0.52;
for Q3 it rises to 0.675; and then it climbs to 0.921 for Q4. This generally
U-shaped pattern means that networking matters more for those highest
and lowest in managerial quality.
The figures, taken as a set, confirm the expectation of nonlinearity. Fully
linear relationships would show (roughly) identical and constant coefficients
on each graph – four horizontal and nearly equivalent lines at some height
reflecting the impact of that variable. For all four management and stability
variables, nonlinearity is clearly evident. Even in the case reflecting the most
consistency of impact (Figure 5.1), the size of the superintendent stability
coefficient varies by 332 percent from lowest to highest. In the cases
of quartiles of managerial networking (Figure 5.3) and teacher stability
(Figure 5.4), the shifts in strength are substantially more dramatic.
Second, some of the relationships are particularly interesting. Explicit
comparisons of these results with the model do not constitute definitive
tests, since the tests involve some simplification of the original formulation.
All the same, one expectation sketched earlier – that more stability would
mean a greater impact from the first term of the model and lessened impact
from the second – is supported in the case of teacher stability (Figure 5.4):
more stability is accompanied by a dropoff in the impact of networking.
When managerial quality is high, the impact of networking and teacher
stability is high, but management stability is less important (see Figure 5.1).
This pattern fits with the notion that the best managers may be able to
leverage more from their own actions and the other variables they can
influence, directly or indirectly. Management quality, in turn, seems to
matter more when superintendents engage in a great deal of networking in
their environments (Figure 5.3; see also Chapter 4). The interaction of
teacher stability with management quality is also interesting: the former
has its greatest impact at very low and high levels of managerial quality
(Figure 5.1). Conversely, teacher stability is far less important when net-
working is high (Figure 5.3). These illustrations are only a few of the ways
that management and stability interact. Probing these interactions and the
practical and theoretical reasons for them is fertile ground for future man-
agement scholarship. The findings here also suggest that practical strategies
focused on any single aspect of management are likely to be contingent on
the entire matrix of activities under examination here.
149 Conclusions about stability

Conclusions about stability

Stability has received little attention in recent times from management gurus.
The vaunted virtues of entrepreneurship and reengineering, change and
reinvention, are the coins of today’s reformist realm. The best thinking about
public management over the decades offers justification for considering this
issue afresh, however. Administrative arrangements are autoregressive systems,
and change certainly entails costs as well as possible benefits. The model we have
developed suggests that stability in public programs is a dimension worthy of
systematic investigation and that it could shape performance for the better.
Stability takes on many forms in administrative systems. We have
explored just one of these, personnel stability, and its impact on educational
performance. We have also incorporated an explicit consideration of other
aspects of public management. Our analyses reconfirm the importance of
the other aspects of management while offering substantial support for the
notion that personnel stability at both managerial and front-line levels
contributes positively to performance. The findings regarding stability are
persuasive in a number of respects. They are unambiguous on the most
important and salient performance indicator, persist in analyses of many
other measures of performance, and can even be documented in most
autoregressive estimations. Further, the examination of interactions among
the independent variables of management and stability indicates that not
only do the variables matter, the relationships between them are nonlinear
and complex. The model contends that management actions are likely to be
contingent on stability. This empirical evidence suggests that the impact of
management actions on performance is likely to be contingent on the full
range of management decisions and how the various features interact.
Although we have referred to “management” and “stability” variables, in
an important sense all four have to do with public management. The latter
two reference an aspect of what is usually referred to as personnel manage-
ment. Unlike items such as jurisdictional wealth or student characteristics,
personnel stability derives from the administrative system itself and those
who comprise it. While not totally in the control of school district leaders,
these variables are susceptible to influence by the individuals who make
decisions about how such organizations are run. In a real sense, therefore, all
four variables tap aspects of public management.
This point is worth an explicit mention because, taken as a whole, the set
of four independent variables accounts for an impressively large slice of
150 Internal management and performance

educational performance in Texas school districts. For those who would say
that public management constitutes, at best, a tiny part of the explanation as
to why programs work as they do, we would say: look to education in Texas
for a rather dramatic rebuttal. We expect that, as additional scholars develop
similar, theoretically informed indicators of management in other public
organizations, the results will be similar.
This investigation might seem to amount to a fairly comprehensive
exploration of the issues in question. This section of the chapter reports
on the role of four public management variables in shaping educational
performance, uses ten performance measures, includes numerous controls,
tests for relationships with and without the autoregressive form, and
unpacks interactive effects. In these respects, the study offers convincing
evidence. It is nonetheless important to emphasize that the results are but
partial findings regarding a broader and more complex set of issues.
Indeed, as emphasized at the outset of the chapter, personnel stability is
merely one theme even within the realm of internal management. We next
turn to a second set of HR issues: the quality of human capital in an
organization and its effective management.

The human side of public organizations

Attracting and developing skilled and motivated people at all levels is a core
function in the management of public organizations’ human resources.
McGregor admonished students of management to take note of the “human
side of enterprise” (1960; see also 2006).19
Of all the internal managerial functions, the management of public
organizations’ human capital is surely one of the most important and often
discussed (see Light 2008). Indeed, today’s frequent references to the federal
government’s “human capital crisis,” the attention devoted to HR strategy by
the US Comptroller General (see, for instance, Walker 2001), and the
arguments of a number of scholars that human resources management is
critical all amplify the theme (Ingraham, Selden, and Moynihan 2000;
Bilmes and Neal 2003; Breul and Gardner 2004; Kellough and Nigro 2006).
Within the United States, the great majority of public employees pursue their
careers at the local level, and far and away the vast majority of these – more than
6.7 million – work in the field of public education, especially elementary and
secondary education (Nigro, Nigro, and Kellough 2007: 4–5, 8). Here we
examine the impact over a several-year period of human capital and its
151 Human capital management and the model

management on the performance of Texas school districts, while controlling for

a range of additional influences – other resources, constraints, and managerial
variables – that also might shape outputs and outcomes. This study appears to
be the first systematic empirical exploration of this sort in the literature on
public management in attempting to isolate the link between human capital
and performance in a set of several hundred public organizations.
We proceed by placing this investigation in the context of our model of how
public management helps to shape program results. We then sketch the theor-
etical background that serves to support our focus on human capital and its
management as a key HRM function, and of internal management generally.
We then describe our measures, provide some evidence for the reliability of
our measure of public human resources management, and report the results for
the estimations of human capital on organizational performance.

Human capital management and the model

Our full public management model is complex. We proceed here as we have

been doing thus far: testing portions of the model, often through partial and
simplified specifications. Because all the public organizations in this study
are school districts, we are going to assume that the major structural/
stability factors (other than the management-related personnel stability
measures) are essentially constant and can be dropped from the model, thus
resulting in the following:
Ot ¼ b1 ðM1 ÞOt1 þ b2 ðXt ÞðM3 =M4 Þ þ et
We then initially drop the autoregressive term in the model, to avoid missing
significant long-term relationships simply because the lagged dependent vari-
able can overwhelm small but significant influences. We estimate the model
both with and without the lagged dependent variable, however. In addition,
following earlier simplifications, we convert the (M3/M4) ratio into M2:
Ot ¼ b1 ðM1 Þ þ b2 ðXt ÞðM2 Þ þ et
Because our concern is internal management, we simplify further by moving
from an interactive model in the second term to one that is linear. This step
eliminates a portion of the relationships that are anticipated by the full
model. The simplification is justified, however, to focus attention on the
relationships of particular interest in this investigation. The nonlinearities
theorized regarding the relationship between external management, on the
152 Internal management and performance

one hand, and the opportunities and constraints in the organizational

environment, on the other, have already been partially examined. The linear
version of the last equation is
Ot ¼ b1 ðM1 Þ þ b2 ðXt Þ þ b3 ðM2 Þ þ et
The result is a relatively simple linear model that will permit us to assess the
impact of the internal management of human resources on organizational
performance while controlling for the environment and other managerial
factors. We also estimate a second set of models with a lagged dependent
variable to simulate the autoregressive nature of most organizations.

Theorizing about the content of M1

An extensive literature emphasizes the importance of the “human side” of

public organizations (for reviews and analyses, see Shafritz et al. 2001 and
Nigro, Nigro, and Kellough 2007). The recent high-visibility study of “man-
agement capacity” in the Government Performance Project theorized that
human resources management is one of four key management “subsystems”
linked via leadership and driven toward higher performance by a system of
managing for results (see Ingraham, Joyce, and Donahue 2003, Ingraham
2007, and in particular Selden and Jacobson 2007).
A recent argument based in part on formal theory concludes that the
personnel function in government is the most critical one for those concerned
about results that comport with the “core governmental values – judgment,
balance, rationality, and accountability” (Bertelli and Lynn 2006: 131; see also
103–31). A conceptual and empirical literature on business organizations
also contends that HRM generally, and especially employee development
efforts, can be important contributors to organizational performance (for
literature reviews and conceptual contributions, see Wood 1999 and Jacobs
and Washington 2003; for empirical findings supportive of this theme, see
Koch and McGrath 1996 and Li 2000). Indeed, some economic analyses have
emphasized the importance of human resources (see Ellingsen and Johannes-
son 2007, Lazear and Shaw 2007, and Pfeffer 2007), and some empirical work
has been conducted on public education with regard to teacher labor markets
and sorting (Lankford, Loeb, and Wyckoff 2002; Boyd et al. 2003), and the
impact of teacher quality on student achievement (Rivkin, Hanushek, and
Kain 2005). These studies do not explore management or the performance of
school districts as organizations, however.
153 Sample and measures

For all these reasons, it seems reasonable to expect the component of M1

that has to do with human resources management to be related to perform-
ance. Of course, even this narrowing of the subject distills a rather large set
of managerial practices and responsibilities. For the purposes of our empir-
ical analysis, we focus on what we regard as the core indicia of what should
be expected from a successful managerial program in the field of human
resources: evidence of talent at the top of the organization, mid-level
managerial quality, skilled and capable front-line people, and the quality
of programs in place to provide professional development. These are all
elements of human capital and its management. We have developed a
separate measure for top managerial quality, of course, and the detailed
discussion of that facet was provided in Chapter 4.

Sample and measures

Once again we use the Texas school district data set and our 2004/5 survey of
school district superintendents. For this version of the survey, the return rate
was 61 percent.20 Pooling four years (2004 to 2007) of data on performance
and control variables produces a total of 2,400 usable cases for analysis.
Missing data on individual items reduces this number somewhat in individ-
ual equations.21
We use our measure of managerial networking as developed earlier in this
book and used frequently throughout. One adjustment made in the analysis
in this section is that we use the factor score of interactions for the four
possible interaction nodes aside from interaction upward with the school
board. The school board interaction can be considered a principal–agent
link and is included in the equations reported here as a separate measure. We
also employ our validated measure of managerial quality, and we use both
measures of personnel stability developed in the present chapter.
We also employ the usual eight control variables, and we focus in this
section on the same ten performance measures already introduced in the
first part of the chapter.22 What remains to be explained is the measure of
human capital management.
Our measure of human capital management taps what should be present
as a part of successful human resources management: the attraction and
development of the organization’s human capital. Success in this aspect of
management should produce high-quality managers and front-line workers –
teachers in these highly professionalized organizations – and should also
154 Internal management and performance

Table 5.4 Measuring the quality of human capital: factor loadings

Indicator Loading

Quality of experienced teachers 0.74

Quality of professional development 0.65
Quality of principals’ management skills 0.75
Our people can make any program work 0.67
Recommend subordinate as superintendent 0.39
Eigenvalue 2.12

see in place an effective professional development program. Although

many other aspects of HRM could be examined, including the nature
and quality of strategic planning for HR needs, we focus here on the core:
the talent of the people and the quality of professional development
available to them.
Based on our assumption that internal management will result in the
development of the human resources of the organization, five survey items
were used to tap the quality of this aspect of internal management.23
Superintendents were asked to rate the quality of principals’ management
skills, the quality of experienced teachers, and the quality of professional
development available on a five-point scale from “Excellent” (5) to “Inad-
equate” (1). Principals are the key line managers of the organization and
generally are the chief operating officers of the individual schools. Teachers,
of course, are the primary production personnel of any school system.
Superintendents were also asked to agree or disagree with two statements,
“With the people I have in this organization, we can make virtually any
program work” and “I am quite likely to recommend a subordinate for a
superintendent position in another district,” on four-point scales from
“Strongly agree” to “Strongly disagree.” The first question taps into a
growing view in education policy that schools are successful because they
have effective human resources rather than because they adopt specific
programs (Meier et al. 2006). The second seeks to determine if the superin-
tendent is developing the skills of mid-level managers.
Our measurement strategy is to use factor analysis to extract the common
core concept of human capital from these five indicators. The results of the
factor analysis are shown in Table 5.4. The five items all load positively on
the first factor, accounting for 42.4 percent of the total variance. No other
factor meets the standard criterion of significance by producing an eigen-
value of 1.0 or greater. The factor loadings show strong correlations between
155 Sample and measures

Table 5.5 The reliability of the human capital measure

Correlation of 2004/5 and 2006/7 measure 0.51 N ¼ 467

Correlation 2005–2007 if same manager 0.60 N ¼ 327
Correlation if managers change 0.33 N ¼ 145

the indicators and the overall factor, except for the willingness to recom-
mend an employee as a superintendent elsewhere.
The measure has some limitations, as it emphasizes the quality of human
capital and only somewhat its management. We do not have many details
regarding just how top management applied HR strategies and tactics to
build its stock of human capital. The question on professional development
does tap explicitly an aspect of management’s responsibility, and the fact
that it loads reasonably well on the factor suggests that the measure reflects
aspects of human capital and also its management.
While the individual items and the factor scores show some face validity,
before using the variable in any statistical analysis we should demonstrate
whether or not the concept is reliable. We can make some assessments of this
because the 2004/5 survey questions that were used to construct the index
were repeated in a 2006/7 survey of superintendents. This second source of
data means that we can gain some leverage on the reliability of the measure
over time. To the extent that management develops human resources, we
would expect these human resources to have some relatively lasting impact
on the organization. In short, we would expect a positive correlation
between the two measures. Further, we would expect this correlation to be
even higher if the same superintendent headed the system in both years,
simply because management practices are less likely to change if the same
person remains as superintendent.
Table 5.5 presents these findings. For the 467 districts that responded to
both surveys, the correlation of the human resources measure across time
was a robust 0.51 – fairly strong for perceptual measures over a two-year
period. When the analysis is restricted to those superintendents who held the
same position in 2004/5 and 2006/7, the correlation jumps to 0.60. For those
districts that changed superintendents between these two years, the correl-
ation drops to 0.33. The pattern is consistent with the notion that the store
of human capital both reflects the efforts of the top manager (the superin-
tendent) and is also embedded in the processes and structures of the
organization, so that the gains from good management are not lost imme-
diately when the top manager leaves.
156 Internal management and performance

Table 5.6 The impact of human capital on organizational performance

Dependent variable ¼ student exam pass rates

Independent variables Slope T-score Slope T-score


Human capital 1.7332 9.42 0.4182 3.54
Managerial networking 0.3556 1.73 0.0564 0.38
School board contact 0.5490 2.72 0.1842 1.26
Management quality 0.5591 3.35 0.2823 2.34
Management experience 0.0117 0.71 0.0046 0.38
Personnel stability 0.2096 9.42 0.0386 2.34
Previous year’s pass rate – – 0.7042 46.41

R2 0.72 0.85
Standard error 7.30 5.27
F 354.98 761.51
N 2,380 2,379

Notes: Coefficients for annual dummy variables and control variables (teacher salaries,
state aid, class size, teacher experience, noncertified teachers, and the percentages of black,
Latino, and low-income students) are not shown.  ¼ significant at p < 0.05, one tailed
test. Time period ¼ 2004–7.

Findings

We first report the impact of human capital on the most salient performance
measure: the all-pass rate on the standardized, statewide examination (the
TAKS). The first two columns of Table 5.6 display the full estimation for
the model developed above, without the lagged dependent variable and with
the pass rate as the dependent variable. The other two columns show the
results when the autoregressive term is added to the first model. In the former
case, and leaving aside the key variable of interest for the moment, all the
other management measures aside from management experience are related
to performance, and in the direction expected theoretically and on the basis of
earlier work. In addition to these relationships, human capital contributes
positively to the test results, and the relationship is highly significant – with a
T score above nine. Interestingly, as well, the effect size for this variable is
substantial. Since the measure is a factor score with an effective range of 3
to þ3, the maximum effect size is more than ten percentage points on the pass
rate. This result suggests a substantively important impact.
The estimation including the lagged dependent variable has some of
the variables dropping out of statistical significance, as would be expected
157 Findings

given the high hurdle that would have to be surmounted to show per-
formance impacts controlling for the preceding year’s pass rate. Manage-
ment quality, which can be considered a measure tapping top-level
human capital, and front-line workforce stability continue to have
impacts on performance, however. Our human capital factor score
remains significant and contributes positively to districts’ pass rates.
Although the effect size is smaller in this estimation than the previous
one, it is worth bearing in mind that in the autoregressive equation the
estimated impact of the variable should be considered as building into the
base from one year to the next, and thus reverberating forward over time.
In a several-year period, this model shows, this aspect of human resources
management has a substantial effect on the statewide standardized exam
performance. This finding is striking, particularly since a considerable
portion of the human capital was already in place during the preceding
period. The fact that this variable is shown to be substantively as well as
statistically significant in an autoregressive equation and with numerous
management and other controls included speaks to the importance of
human capital and its management for delivering public policy outcomes.
The strong finding in the equation with the lagged dependent variable
suggests that it is likely that causality works in the direction we have
hypothesized, rather than the reverse.24
The overall pass rate on the statewide exam is an important perform-
ance result, but the organizations in question are expected to achieve
other educational results as well. Tables 5.7 and 5.8 show summary
results for the nine other performance measures sketched earlier in this
chapter. The tables include only coefficients and T-scores for the esti-
mated impacts of HRM, but the full model was used to develop the
results. Table 5.7 displays the results for the model without the lagged
dependent variable, while Table 5.8 shows the findings for the autore-
gressive estimations.
As the results in Table 5.7 demonstrate, human capital has a performance-
improving impact in all nine instances (a negative influence on the dropout
rate counts as an improvement). Results for subgroups of students taking
the TAKS – black, Latino, Anglo, and low-income students – all benefit from
better human resources management. Latino students benefit even more
than the others. Similar results can be seen for the other dependent variables.
For example, the maximum effect size for SAT scores is more than thirty
points. High-end performance also improves with better human resources
and their management; the percentage of students achieving college-ready
158 Internal management and performance

Table 5.7 The impact of human capital on alternative indicators of performance

Performance measure Slope T-score R2 N

Black TAKS pass rate 1.5366 4.31 0.40 1,433

Latino TAKS pass rate 2.1036 8.51 0.50 2,139
Anglo TAKS pass rate 1.5019 8.99 0.60 2,336
Low-income TAKS pass rate 1.7028 9.29 0.37 2,391
Attendance 0.0826 5.61 0.27 2,400
Dropouts 0.1203 2.32 0.28 2,317
ACT scores 0.1007 3.27 0.43 2,125
SAT scores 6.3842 4.02 0.40 1,742
College-ready percentage 0.9950 4.41 0.35 2,232

Notes: All equations control for the five management variables, teacher salaries, per student
instructional funds, class size, teacher experience, percentage of teachers not certified,
percentage of black, Latino, and low-income students and yearly dummy variables.

¼ significant at p < 0.05, one tailed test.

Table 5.8 The autoregressive impact of human capital on alternative indicators of performance

Performance measure Slope T-score R2 N


Black TAKS pass rate 0.9231 2.93 0.54 1,379
Latino TAKS pass rate 0.8263 4.05 0.67 2,117
Anglo TAKS pass rate 0.4053 3.17 0.77 2,333
Low-income TAKS pass rate 0.4778 3.55 0.67 2,387
Attendance 0.0147 1.85 0.78 2,400
Dropouts 0.0617 1.47 0.46 2,296
ACT scores 0.0709 2.52 0.54 2,044
SAT scores 2.2582 1.74 0.62 1,641
College-ready percentage 0.7436 3.59 0.46 2,196

Notes: All equations control for past performance, the five management variables, teacher
salaries, per student instructional funds, class size, teacher experience, percentage of teachers
not certified, percentage of black, Latino, and low-income students, and yearly dummy
variables.  ¼ probability p < 0.05, one tailed test.

test scores increases a maximum of nearly six percentage points of students

tested. By any measure, the overall pattern constitutes an impressive set of
results.
Table 5.8 shows the results for the nine additional performance meas-
ures when including lagged dependent variables in each equation. In eight
of these cases, or nine out of ten overall, human capital is related to
improved performance. For all four subgroups of TAKS-taking students,
159 Implications

human capital boosts results. Positive results are also estimated for
attendance as well as high-end performance. In none of the cases are
there negative performance impacts; the result for dropouts is not signifi-
cant. Given the difficulty of attaining significant results in autoregressive
estimations, and particularly since the districts’ human resources at
any time are likely to be partially tapped in the preceding period
(see Table 5.5), these results again provide support for the proposition
that good human resources and the skillful development of them can pay
important performance dividends.

Implications

Everyone “knows” that internal management contributes to the policy

outcomes produced by public organizations. Conventional wisdom, intu-
ition, and considerable quantities of theoretical argument all support this
proposition. A great deal of the case study literature also lends plausibility to
the notion. In this part of the chapter, we have developed a measure for one
important aspect of internal management – the quality or talent of the
people involved in delivering results, along with the quality of the develop-
ment program designed to enhance the contributions of these people. This
slice of internal management – the strength of an organization’s extant
human capital – omits a great deal of what public managers do to support
performance inside the organization, but it does tap a key dimension of the
internal management.
Even when controlling for several other aspects of management, including
some that have to do with external management (such as managerial
networking), we find that the strength of the human resources of the
organization provides a substantial performance payoff. The results are
found across a wide array of performance measures, and they remain for
almost all the measures when controlling for past performance. The evidence
provides especially strong support for the proposition that the “human side”
contributes to tangible, measurable results.
Moreover, the impacts of this aspect of internal management appear to
be distributed broadly across stakeholders. Human capital helps on the
most widely noted performance measure, but it also helps for important
subgroups of the population – including more disadvantaged students.
Human capital also assists in the performance of the top-end students as
well. Given the importance attached to human resources and their
160 Internal management and performance

management in the literature on management, including public manage-

ment, this pattern of findings is certainly not surprising. Given the
absence of systematic empirical research on the question, though, the
results here are important.
In recent years much attention has been devoted to an emphasis on
marketizing the context for public services and incentivizing their manage-
ment. The results from these 1,000 plus governments, however, suggest the
need for public organizations to invest in their people, at both front-line
and managerial levels, as a major priority. The findings also support the
value of significant development programs as a part of HRM. These
implications point toward treating the human capital of public organiza-
tions as a long-term asset to be built and nurtured, rather than as a set of
somewhat interchangeable individuals who must simply be surrounded by
the inducements implied by such innovations du jour as performance
contracts.
This analysis nonetheless provides only a starting point for the explor-
ation of internal management and its links to performance. Even regarding
the management of human capital, substantial additional research is war-
ranted. Longitudinal designs with multiple time points could further clarify
the questions of whether and to what extent reciprocal causation might be
present. Analyses in fields other than public education can determine
whether the performance-related impacts of human capital and its manage-
ment hold in other kinds of public organizations and in other policy
domains. Education is, after all, an especially human-capital-intensive func-
tion. Other types of highly professionalized public organizations, especially
those with substantial front-line discretion, may operate similarly; but the
performance impacts may vary with the extent of specialized education and
training, along with the degree of decentralized discretion allocated to or
near the street level. In addition, it will be important for researchers to study
the relationship between human capital and performance in very different
kinds of settings from school districts for yet another reason. For some
public functions and in some contexts, when the long-term reinforcement of
extant production patterns may not be the preferred approach, the effective
management of human capital may be more, or less, important – or, at least,
differently influential.
This human capital aspect of HRM, furthermore, is only a part of the
broader array of HR-related challenges of interest to researchers in public
management. A number of other elements can be expected to contribute to
organizational outcomes as well. These might include HR planning, the
161 Making high-quality decisions

allocation of job responsibilities across positions, the rules and culture

regarding labor relations, and numerous other facets. Moreover, of course,
the management of financial resources, the crafting of structural arrange-
ments, and many more elements of managers’ internal responsibilities still
await systematic study.
In short, how to understand fully the internal management of public
organizations and its impacts on program performance remains a daunting
but important research question. Even with regard to the narrower question
of how human capital management shapes results, we are only partially
down the road to an answer. The human side of public organizations is
clearly important, but we are far from a full and general understanding of
just how, where, and when.

Making high-quality decisions

The most fundamental part of internal management is making decisions

(Simon 1997). Examining the quality of decision making in a large-N
quantitative study is extremely difficult, since many decisions are context-
specific and can be studied only by a more in-depth analysis. What is needed
is a decision context that is relatively similar across organizations in settings
offering the ability to track how managers respond to the decision event.
This section examines just such a context – how organizations respond to a
large budget cut. Budget reductions require organizations to make decisions
about resources, personnel, and programs – and to do so while trying to
protect the core performance of the organization. How decisions are made
in response to an exogenous shock might not be similar to day-to-day
decision making, but it does provide an opportunity to observe decisions
when the decisions actually matter.25
A budget cut is one type of what scholars term an environmental shock:
any sort of disruption emanating from outside the administrative system
and buffeting the core organization. Some such shocks can be anticipated
and protected against, but, no matter how elaborate its buffering apparatus,
any public organization – as an open system – experiences shocks on
occasion. Shocks can be negative or positive; even an unexpectedly happy
budget decision approved by political leaders can constitute a perturbation
to the system and can have disruptive impacts, at least in the short term, as
managers and others devote time and attention to handling the largesse.
162 Internal management and performance

Especially important and potentially dangerous to performance, however,

can be negative environmental shocks.
We focus for present purposes on negative budgetary shocks – signifi-
cant cuts in the budgets for operations. Some likely budget cuts can be
anticipated, and some can even be avoided. On occasion, however, major
budget cuts constitute unexpected shocks to the system. For school
districts in Texas, the site of our empirical analysis, school funding is
heavily but not exclusively formula-driven. A formula would seem to be a
metric with a highly predictable result, but, even here, shocks are possible
and do occur. The state-level formula producing intergovernmental
funding decisions, and thus a part of the district budget, is heavily driven
by enrollment and by legislative willingness to provide the funds; both
can be unpredictable. The statewide budget for intergovernmental assist-
ance in education also faces a budgetary ceiling; therefore, the revenue
received by any given district is only partially a function of its own
enrollment; it is also shaped by the relative growth (decline) in enroll-
ment in the state’s other districts. Needless to say, such changes cannot be
anticipated in any detail. Additionally, locally raised revenue, on which
part of the budget is based, is mostly derived from the property tax. Here
too disruptions can enter the system. Major industrial plant closings in a
jurisdiction can cause property values, and thus property tax revenue, to
fall abruptly. Property tax reassessments can also quickly alter the revenue
picture. Sometimes such shocks are imposed or driven by political
choices, such as popular votes via referenda in some states or the decision
by Michigan in 1994 to eliminate the property tax as a source of school
funding. In short, negative budgetary shocks can be expected from time
to time; and it is an important function of management to mitigate their
impacts and help the administrative system to recover in the face of such
exigencies.
It can often be the case, in other words, that a relative lack of change in
outcomes or an absence of fluctuation constitutes a subtle signal of man-
agerial success and achievement. To the extent that managers succeed in
helping to maintain production in the face of unhappy surprises and adver-
sity, sometimes even invisibly so, the contributions of management to
performance can be essential and far from trivial. Sometimes a placid
pattern of results constitutes a major managerial achievement. Such an event
constitutes, in effect, a managerial version of Sherlock Holmes’ famous
exchange in Arthur Conan Doyle’s story “Silver Blaze” (Conan Doyle
1894: 50):
163 Making high-quality decisions

Gregory (a Scotland Yard detective): “Is there any point to which you would wish to
draw my attention?”
Holmes: “To the curious incident of the dog in the night-time.”
Gregory: “The dog did nothing in the night-time.”
Holmes: “That was the curious incident.”

Estimating the degree of managerial impact in such cases is a particularly

difficult task, since doing so involves dealing with a counterfactual: little or
nothing seems to have happened, but what would have happened had
management not intervened? Still, it is apparent that sometimes, when
performance dogs do not bark, that result constitutes a curious incident
indeed – one that may signal a key sort of managerial success.
Management’s contribution to performance in this situation functions as
a recovery operation that might (partially) return the system to regular
performance despite the environmental shock. In the remainder of this
analysis we focus on exploring whether and how management decisions
(M1) mitigate negative budgetary shocks.
Since many of the internal actions of managers are unobservable and
cannot be directly estimated via a large-N analysis such as the one undertaken
in this study, we pursue the M1 impact indirectly, albeit carefully. We first
estimate the extent to which, if at all, shocks have their amplitude reduced in
the organizational system in terms of performance results. We then examine
reported managerial goals and internal budgetary and staff allocations to see if
the evidence indicates that managerial decisions directed at such allocations
have been aimed at protecting core functions from disruption.
In the present analysis, our data are drawn almost entirely from the Texas
Education Agency. For one portion of the work we also use 2002 survey
responses of school system superintendents. The response rate was 60
percent, with sampled districts very similar to the universe of districts. All
other analyses are conducted with a data set containing eight years (1995 to
2002) of data on performance and control variables, for a maximum of 8,329
cases for analysis.

Measuring shocks
To estimate budget shocks, we follow Ratts (1999), who examines shocks to
national economies. We regress total school district revenues (logged) on
their logged values for the prior year.26 This regression essentially estimates
what school district revenues would be if past trends continued. We then
164 Internal management and performance

designate as “shocks” any year-to-year reductions in school district budgets

that exceed 10 percent of the previous budget. A total of 730 budget shocks
occurred in this eight-year period, thus affecting approximately 8.8 percent
of the cases. Although for our purposes the minimum shock is set at a 10
percent reduction, the budget shocks could be much larger, and actually
averaged 18 percent. We have also performed sensitivity analyses by
adjusting the designated minimum cut that would constitute a shock for
the purposes of analysis, and the results do not affect the basic findings of
the analysis. It is important to note that the autoregressive estimation
incorporates past budget patterns. This means that a district with a growing
budget will suffer a 10 percent plus loss from that growing trend. A district
with a flat budget would see a real drop of 10 percent. By measuring the
shock relative to past expectations, the Ratts technique focuses on the
unpredictable element in budget cuts. In all equations here, a budget shock
of 10 percent or more is converted to a dummy variable. We are confident
that this measure isolates the shock component because the models also
control for a variety of expenditures. The coefficients for the variable,
therefore, estimate the impact of the shock while controlling for the level
of resources.

Dependent variables and controls

We also employ the usual eight control variables, and we focus in this section
on the same ten performance measures already introduced in the first part of
the chapter. We deal with serial correlation with a series of year dummy
variables. Heteroskedasticity was well within acceptable ranges.

Findings

Our first question is to ask what the impact of a budget shock of 10 percent
or more is on the performance of the organization. Table 5.9 provides a
regression of overall TAAS rates on a budget shock while controlling for the
other resources and constraints of the organization. The results in the first
two columns are surprising. A budget shock has only a modest impact on
the organization’s performance; it results in only a drop of one half-point
in TAAS performance, all other things being equal, and this relationship is
statistically significant only if we use our directional hypothesis and then use
a one-tailed test of significance at the 0.10 level. The remaining control
165 Findings

Table 5.9 The impact of a 10 percent or greater budget shock on organizational performance

Dependent variable ¼ student exam pass rates

Independent variables Slope T-score Slope T-score

Budget shock 0.5006 1.62ns 0.4938 1.59ns

Lagged budget shock – – 0.9956 3.05
Control variables
Teacher salaries (000s) 0.6616 12.31 0.6596 12.28
Class size 0.3450 8.91 0.3578 9.19
Teacher experience 0.0783 1.62ns 0.0826 1.72
Noncertified teachers 0.1371 7.84 0.1386 7.92
Percentage state aid 0.0123 2.81 0.0136 3.09
Percentage of black students 0.2332 27.52 0.2329 27.50
Percentage of Latino students 0.1035 18.66 0.1031 18.59
Percentage of low-income students 0.1485 19.32 0.1472 19.13

R2 0.61 0.61
Standard error 7.66 7.66
F 820.09 773.71
N 8,321 8,321

Notes: Dummy variables for individual years not reported. ns ¼ not significant at p < 0.05.
Time period ¼ 1995–2002.

variables in the model show the predicted relationships. (The findings are
also robust to the inclusion of a set of management variables.)
Because student performance is highly autoregressive (the same students
are tested every year), it might take more than one year for a negative impact
to show up. The third and fourth columns of Table 5.9 include a one-year lag
for the budget shock. The results show a small but statistically significant
negative impact of the budget cut in the second year, approximately one
percentage point on the TAAS. Additional estimations with longer time lags
show no further impacts.
The ability of school districts to take a 10 percent or better budget cut and
produce performance results that drop only a small amount is an interesting
finding. This is clearly a case of a sizable cut, and the interesting story is the
dog that did not bark, let alone bite. The most plausible explanation has to
do with management. Public organizations are open systems, but they are
also actively managed – that is, they are not blown along at the whim of
environmental winds. They adopt procedures to dampen environmental
threats, and they can also take actions internally to ameliorate the effects
of unpleasant shocks from the environment. Budget shocks should be no
166 Internal management and performance

Table 5.10 The impact of budget shocks on other indicators of performance

Dependent variable Slope T-score R2 N

Black pass rates 0.6232 0.77 0.49 4,870

Latino pass rates 0.0424 0.08 0.46 7,477
Anglo pass rates 0.5101 1.66# 0.45 8,202
Low-income pass rates 0.1609 0.41 0.56 8,252
Attendance 0.0147 0.45 0.22 8,324
College board exams rate 1.3930 2.06 0.14 7,449
SAT scores 2.7372 0.74 0.46 5,742
ACT scores 0.0885 1.45 0.38 6,909
Percentage above 1,110, SAT 1.2834 2.84 0.30 7,525

Notes: All equations control for teacher salaries, per student instructional funds, class size,
teacher experience, percentage of teachers not certified, percentage of black, Latino, and
low-income students, and yearly dummy variables.  ¼ significant at p < 0.05, two-tailed test.
# ¼ significant at p < 0.10, two-tailed test. Significant second-year effects were found for
Anglo students and for the percentage above 1,110.

exception to this rule, as the budget process is often fraught with crises of
either economic or political origin (Rubin 2005).
Table 5.9 shows the impact of the budget shock only on the primary
performance indicator for Texas school districts. Nine additional regressions
were run on the other performance indicators to determine if budget shocks
had any other impacts, or if perhaps the shocks affected some portions of the
organization rather than others. Table 5.10 shows the abbreviated results of
these regressions. In three cases, we find statistically significant negative
impacts of a 10 percent or greater budget shock on the organization: Anglo
test score results, the percentage of students who take one or both of the SAT
and ACT exams, and the percentage of students who score above 1,110 on
the SAT or its ACT equivalent. TAAS pass rates for blacks, Latinos, and low-
income students, as well as attendance rates and ACT and SAT mean scores,
show no discernible impact (similar null results occur with a lagged shock
variable). The budget-shock impact on Anglo TAAS rates remains at the
marginal level of about one-half point and now just crosses the threshold of
significance at the 0.10 level (or the 0.05 level with a one-tailed test, not an
especially stringent test with 8,000 plus cases). A one-year lag of the shock on
this variable shows a slightly larger negative impact of 0.9 points (results not
shown). The impact on college-bound students appears greater. Approxi-
mately 1.4 percentage point fewer students take either the ACT or the SAT
(in the first year), and those students who score above 1,110 drop by about
167 Findings

Table 5.11 Superintendent priorities: what is your primary goal for improving
your district?

Goal Percentage

Student TAAS scores 70.3

College preparation 24.4
Bilingual education 3.1
Vocational education 1.7
Extracurricular 0.6
Athletics 0.3

N ¼ 650

1.3 percentage points (in the first year, and a significant second-year drop of
0.8 points [equation not shown]). These two impacts are likely cumulative;
fewer students take the exam and those who do score lower. Although a 1.4
percentage point drop from the average of 62.8 percent taking the exams is
not large, a 1.3 percentage point drop and a 0.8 percentage point drop from
the 19.0 percent who score above 1,110 is a sizable reduction.
Why might such a pattern occur whereby one set of performance
indicators is generally unaffected by a major budget reduction while others
are? One explanation might be that management seeks to protect those
organizational outputs that are more highly valued – that is, part of their
core mission (Thompson 1967). Although it might be argued a priori that
the emphasis on standardized tests in Texas will be valued more highly
than other outputs, the results from a 2002 survey of Texas school super-
intendents (see Table 5.11) provides unequivocal evidence. Superintendents
were asked to rank-order their goals among six different areas: TAAS,
college-bound students, vocational education, bilingual education, extra-
curricular activities, and athletics. Fully 70 percent of superintendents listed
TAAS as the top goal for their district. The results of Tables 5.9 and 5.10,
therefore, are consistent with the view that superintendents protected their
primary goal at the expense of some degradation in the achievement of
secondary goals.
Goals convey intentions; they do not tell us how superintendents are
actually able to keep a large budget cut from affecting their primary
performance goal. Indeed, prior research on the public management of
cutbacks offers at least three broad possibilities. One, developed by
researchers focusing on local government management, indicates that
managerial choices under fiscal stress can be fairly unpatterned and unsys-
tematic (Pammer 1990; Bartle 1996). Another is that cuts might be
168 Internal management and performance

Table 5.12 The nonbarking dog: the relationship between a budget shock and
instructional expenditures

Dependent variable ¼ logged total instructional expenditures

Independent variables Slope T-score P

Intercept 0.1849 28.15 0.0000

Budget shock 0.0518 10.83 0.0000
Lagged expenditures 0.8803 158.14 0.0000

R2 0.75
Standard error 0.12
F 12,707.20
N 8,328

absorbed in politically more acceptable ways, especially by spreading the

pain – for instance, by across-the-board cuts ostensibly affecting all parts of
the organization and stakeholders equally (Levine 1978, 1980). A third
option, suggested in the preceding paragraphs and explicated in the litera-
ture on organization theory, would be for managers to make selective and
perhaps subtle, even if difficult, choices regarding budgetary and staffing
decisions so as to minimize negative performance impacts. Unfortunately,
much of the earlier research contains little if any systematic information on
performance effects.
We have sought to rectify this omission here, as Tables 5.9 and 5.10
document. Further, we can try to sort through the different managerial possi-
bilities by tracing some of the choices made to deal with budget shocks – in
particular, by examining management’s internal budgetary reallocations in the
wake of the shock, and also certain personnel data. For instance, one obvious
option for management is to absorb most or all of the budget cut in areas that
are not directly related to the organization’s core function: the instruction
of students. As a first step, Table 5.12 examines the allocation of funds to
instruction. Because the total expenditures of each district are logged and we
control for last year’s expenditures on instruction, the shock coefficient indi-
cates that a 10 percent or greater budget cut resulted in only a 5.2 percent
reduction in instructional funds. Bearing in mind that the criterion for a shock
was a budget cut of 10 percent or more, the actual drop in revenues for the
“shocked” districts was 18.4 percent, meaning that superintendents were able
to shelter instructional expenditures from approximately 72 percent of the
impact of an across-the-board cut.
169 Findings

Table 5.13 Redistributing funds to core functions: percentage point reductions in

allocations

Function Gain/loss T-score R2 N

Instruction 1.119 6.36 0.35 8328

Central administration 0.265 3.91 0.72 8328
Campus administration 0.337 8.85 0.57 8328
Physical plant 0.484 9.22 0.61 8328
Nonoperating expenses 2.167 11.10 0.34 8328

Notes: All regression equations estimated with a lagged dependent

variable.  ¼ p < 0.05.

A second look at the budgeting actions of superintendents can be found

in Table 5.13. We took the percentage of the overall budget allocated to
five functions – instruction, central administration, school administration,
physical plant, and nonoperating expenditures – and examined the impact
of the budget shock on these percentages while controlling for the expend-
itures in the previous year. These results clearly demonstrate the privileged
position of instructional expenditures in the districts’ internal budgetary
decisions. Bearing in mind that these districts are operating with a smaller
total budget, we note that the 5.2 percent decrease in instructional expend-
itures (in Table 5.12) actually means that the percentage of all funds allo-
cated to instruction increases by 1.1 percentage points (Table 5.13).
Administration and physical plant percentages go up, but this essentially
means that they absorb less than a 10 percent cut (although more than the
5 percent in instruction). The major loser in the reallocation is nonoperating
expenses (capital expenditures, debt service, and community services), with
a reduction of 2.2 percentage points. This is a substantial cut, since non-
operating expenditures average only about nine percentage points in the
school districts’ budgets.
Squeezing out the frills and capital expenditures is unlikely to be enough
to cover the shortfall, however. Texas superintendents have extensive per-
sonnel powers that they can use. Teachers work on annual contracts, and
superintendents can opt not to offer a contract for the next year or to offer a
contract at the same salary. Table 5.14 illustrates two personnel approaches
that districts facing a major budget cut use. The first regression predicts
average teacher salaries with teachers’ salaries in the previous year and the
budget shock. The significant negative coefficient indicates that teacher
salaries in the examined districts actually drop by $278 in the year of the
170 Internal management and performance

Table 5.14 Reducing core costs: teachers’ salaries and class size

Dependent variable

Teacher salaries Class size

Intercept 3.629 0.925

Budget shock 0.278 (5.14) 0.073 (2.11)
Lagged dependent variable 0.923 (244.41) 0.917 (242.51)

R2 0.88 0.88
Standard error 1.39 0.87
F 29,908.92 30,203.83
N 8,323 8,329

budget crisis. As a point of comparison, over this eight-year time period

teacher salaries in all districts increased by approximately $1,329 per year.
A portion of the budget deficit, therefore, is made up by offering lower
average salaries to teachers. This can be done by offering continuing teachers
the same salary and replacing teachers who leave with cheaper, less experi-
enced teachers (teacher turnover averages 16 percent per year in the state).
The second equation in Table 5.14 examines class size in the same way.
Districts hit with a budget shock responded by allowing their class sizes to
increase slightly, by 0.07 students per teacher. Although this is not a large
change, it should be viewed in terms of the general trend during this time in
Texas, when student-to-teacher ratios declined from 13.4 to 12.4. In com-
bination, these two results suggest that districts replaced fewer teachers (and
some may have actually terminated teachers), thus permitting class sizes to
increase, and at the same time held the line on teachers’ salaries.
Managers obviously take multiple actions to render budgetary canines
somnolent. A related strategy is to alter the composition of the instructional
personnel. Table 5.15 examines the percentage of employees who are
teachers, teachers’ aides, and instructional support staff (audio-visual assist-
ants, etc.). Although Table 5.13 shows that the percentage of funds allocated
to teaching increased, Table 5.14 illustrates again that this is effectively a
result of instructional funds declining at a lower rate than for other types of
funding. The percentage of teachers in the budget-impacted districts does
not change from one year to the next – evidence consistent with zero growth
in total teaching positions. At the same time, the percentage of teachers’
aides declines by 0.3 percentage points and the percentage of support staff
declines by 0.13 percentage points. Although these are relatively small
171 Findings

Table 5.15 Reducing core support tasks: instruction, aides, and support staff

Dependent variable ¼ percentage of staff assigned to

Function Gain/loss T score R2 N

Teaching 0.036 0.31 0.72 8,329

Support staff 0.125 2.21 0.69 8,329
Teachers’ aides 0.312 2.52 0.62 8,329

Notes: All regression equations estimated with a lagged dependent

variable.  ¼ p < 0.05.

Table 5.16 Seeking less expensive core personnel

Dependent variable ¼ percentage of teachers in various functions

Function Gain/loss T score R2 N


Regular education 0.871 4.95 0.71 8,329
Special education 0.191 2.26 0.69 8,329
Compensatory education 0.377 3.29 0.55 8,329
Vocational education 0.041 0.60 0.68 8,329
Bilingual education 0.182 2.43 0.87 8,329
Other education 0.210 3.56 0.64 8,329

Notes: All regression equations estimated with a lagged dependent

variable.  ¼ p < 0.05.

reductions in absolute terms, the changes should be contrasted with a 20

percent growth in support staff and a 17 percent growth in teachers’ aides
over the years examined. In short, additional funds are saved by reducing the
number of “less” essential personnel.
Finally, an organization facing fiscal constraints is likely to recognize
that their core personnel vary in skill and cost. Although school districts
might like to think of all teachers as fully interchangeable, in practice
teachers are specialists in a given curriculum, and some of these specializa-
tions are in short supply. The state of Texas faces a large shortage of qualified
bilingual education instructors, and many districts pay a bonus to newly
hired bilingual teachers (or offer a higher pay scale for these professionals).
Table 5.16 provides an analysis of the percentage of teachers that fall into six
categories (regular education, special education, compensatory education,
vocational education, bilingual education, and other). Quite clearly, these
results are consistent with a strategy of responding to budgetary shocks by
seeking out or concentrating on less expensive teachers. A district facing a
172 Internal management and performance

budget shock sees a 0.87 percentage point increase in regular education

instructors but a 0.19 percentage point decline in special education teachers,
a 0.38 percentage point decline in compensatory education teachers, and a
0.18 percentage point cut in bilingual education teachers (the “other”
teacher category also declines, but the composition of this residual category
is too diverse to determine if such teachers are more or less likely to be better
paid). Because these shifts appear in the context of a constant number of
teachers (see Table 5.15), the implication is that, as bilingual or other
more expensive – i.e. special education – teachers leave, they are replaced
by less expensive regular education teachers.27
Through a variety of stratagems that can be documented with some care,
therefore, managers make a series of decisions that, in the short term,
insulate the most highly valued organizational tasks from the potentially
negative impacts of budgetary shocks. The effort was made to keep teachers
in the classroom, even if class size had to grow a little bit and even if the
district could afford only a lower-cost teacher (regular education rather than
bilingual or special education). In the short term, at least, the performance-
related results are clear: core performance is protected as much as possible,
some other goals are subtly deemphasized, and for the most part the
managerially crafted system is tweaked in order to deamplify or detour
sizable shocks from having their full effect on public education.

Implications

We have shown in a sample of 1,000 public organizations over an eight-year

period that, when faced with significant budgetary shocks, these units make
a series of internal management decisions that absorb the unpleasant event
without experiencing much if any performance decline. These findings are
valid for many but not all performance metrics, and they document short-
term impacts only. Interestingly, few if any declines are seen in delivering
results on the most salient organizational goal, and also on bottom-end
performance measures. The organizations and their managers appear to be
successful at generating desired outcomes on priority matters and for the
most disadvantaged clientele. The results constitute evidence that internal
management is fairly effective in shaping performance. In this set of cases, in
effect, the budgetary dog did virtually nothing in the night-time. That
pattern of findings constitutes not only a curious array of results but also
one that contains encouraging news for those interested in the link between
173 Implications

public management and performance. A virtual lack of impact can be taken

as strongly hinting at effective management. Considering these results, we
conclude that, despite a sizable shock delivered to school systems via the “X”
term, internal management (M1) is rather successful at propping up the
steady delivery of outcomes by making a series of key allocation decisions.
The analyses arrayed in Tables 5.12 to 5.16 considerably strengthen the
inference that managers operate systematically to make tough choices with
the objective of protecting the most valued outputs and outcomes. In these
cases, school system superintendents worked to insulate instructional
resources from unexpected cuts, redistributed resources within the dimin-
ished total budget to help instruction at the expense of ancillary tasks and
also important but less immediate capital needs, and managed the teaching
corps – and its compensation – in the interest of similar objectives. By
attrition and salary constraints, an emphasis on core teachers instead of
support staff and aides, and a focus on employing standard classroom
instructors instead of specialized teachers, they managed their organizations
toward success – as measured by a focus on performance in the core business
of the organizations.
The emphasis here, however, is on internal management. Tried-and-true
core managerial functions, such as careful budget management and the
management of human resources, are tools that public managers use to
craft policy-relevant results. These are not blunt instruments – like so many
clubs wielded to keep the budgetary dogs at bay. Rather, delicate and often
almost invisible tweaks of a set of managerial systems can be used to fashion
success in the midst of troubled times. Not only does POSDCORB live; its
injunctions receive partial validation on the basis of performance.
Lest the lessons of this study appear overly optimistic and also excessively
managerialist, however, two caveats should be entered. First, there are some
losers when budgetary shocks hit. Interestingly, the evidence suggests that
top-end and/or more privileged students fare the worst in response to
substantial budgetary shocks. While those concerned with equity might
regard this pattern as better than its obverse, it does demonstrate that
managers cannot turn budgetary mastiffs completely into sleek greyhounds.
To the extent that the public prefers, and ultimately benefits from, higher-
end achievement such as a college-ready student body, sizable cutbacks
entail costs.
Second, managers are able to handle a substantial portion of the poten-
tially deleterious cutbacks through a series of short-term choices. The
findings reported here do not speak to long-term consequences, however.
174 Internal management and performance

The analyses show that managers are able to cope rather well, but what
happens beyond short-term coping is an important and so far unanswered
question. Clearly, repeated budgetary shocks are likely to show substantially
greater negative impacts on performance. Even isolated shocks may carry
larger long-term consequences, particularly on some aspects of educational
performance – bilingual or special education, perhaps, or advanced work by
high-performing students. The protective moves visible in the school dis-
tricts studied here provide some clear benefits, but they may also function,
in effect, as initiating a process of eating the seedcorn: weakening the
educational system’s infrastructure so that it becomes progressively more
difficult over time to maintain effective performance. These matters too
deserve careful investigation.

Generalizing the impact of internal management

Internal management actions, both by themselves and via their contribu-

tions to stability, have been successfully linked to performance in a number
of settings other than Texas school districts. All the studies use regression
models similar to those presented in this chapter; all control for a wide
variety of other factors that could also influence program outcomes. To
retain the focus on the linkage between management and performance, we
do not discuss the control variables but, rather, focus on key relationships
between management or stability and performance.
In a specific application of the management model presented in Chapter 2,
Donahue et al. (2004) examine US state agencies that oversee human
resources and state agencies that manage the fiscal debt of the state and its
programs. For the HR agencies, they measure M1 in terms of the degree of
program decentralization, the extent of formal reporting on personnel
matters, and the extent of workforce planning. They find that both decentral-
ization and the extent of workforce planning positively affect the quality of
personnel hired (this is measured in terms of managerial perceptions) while
the degree of formal reporting is associated with lower levels of employee
turnover (Donahue et al. 2004: 140). For debt management agencies, Dona-
hue et al. (134) measure M1 via a survey that assesses how much control the
agency has over capital improvement plans in the state, their enforcement
powers over the use of bond-raised funds, and the agency’s control over the
repayment life of the debt. The agency outputs are the percentage of bonds
issued via competitive sales and the percentage of bonds sold via requests for
175 Generalizing the impact of internal management

proposals; both measures should be associated with lower interest rates for
debt. Although the M1 measures do not have much impact on competitive
sales, they do have a strong positive impact on the percentage of bonds
covered by requests for proposals (142).
Meier, O’Toole, and Hicklin (2009) examine the managerial practices of
266 public and private universities in the United States. They create an
internal management measure that is a factor analysis of how the university
president allocates his or her time. Using scales similar to the managerial
networking scales in terms of frequency, the president’s interactions with
thirteen different individuals or groups within the university are assessed.
A two-factor solution is found. One factor indicates a very hierarchical
managerial style, whereby actions are taken almost exclusively through direct
subordinates – that is provosts, deans, and department chairs. The other
factor is a more fluid internal management style that includes interactions
with a far larger variety of individuals and resembles more a network-like
approach than a strict hierarchical orientation to management. The study
includes two performance measures: the six-year graduation rate (the
percentage of new first-year students who graduate within six years) and
the percentage of new faculty hires who are African American. A hierarchical
management style is positively associated with six-year graduation rates but
had no impact on the effort to hire more African American faculty
members. In contrast, a more networked internal management style is
positively associated with increases in faculty diversity but is unrelated to
the six-year graduation rate. In short, one approach to internal management
appears to generate efficiency benefits while a different management style is
associated with greater equity.
Roch, Pitts, and Narvarro (2010) study a specialized aspect of internal
management in their analysis of public schools in the US state of Georgia
from 2002 to 2005. The outcome they explore is whether the individual
schools pursue school disciplinary policies that are punitive (out-of-school
suspensions, expulsions, etc.) or ameliorative (in-school suspensions,
assignments to alternative schools, etc.). Their internal management meas-
ure is based on human resources and, essentially, is a measure of how well
the demographic composition of the teaching force and the administrators
reflect the demographics of the student body, using a Euclidian distance
measure. In essence, this measure taps how successful the HR managers
have been in recruiting a diverse workforce. Roch, Pitts and Narvarro find
that this measure of internal management is associated with more ameliora-
tive disciplinary actions, especially in the case of teachers.
176 Internal management and performance

Roch and Pitts (2010) extend this work on Georgia schools by examining
the differences between public schools and charter schools. Charter schools
are schools funded by the state of Georgia but not subject to the extensive
rules and regulations that affect public schools. The idea is to free up these
schools to be innovative and let them compete for students with each other
and public schools. Roch and Pitts argue that charter schools are managed in
a distinctly different way from public schools. They are generally headed by
entrepreneurs who have a strong commitment to a specific education
philosophy and who recruit teachers and administrators who share this
philosophy. In short, such organizations rely on value congruence rather
than hierarchy to manage the organization. This strong emphasis on value
congruence in terms of educational philosophy, the authors argue, means
that some key HR variables for public schools will no longer matter in
charter schools. Again, the authors focus on race, but in this case they have
two dependent variables: ameliorative disciplinary policies and the perform-
ance of minority students on the Georgia state standardized tests. For public
schools, the authors find, as in the previous study, that the diversity of the
faculty is positively associated with ameliorative disciplinary policies and
also that it is positively associated with minority student test scores. For
charter schools, however, these relationships are insignificant – a finding
that allows the authors to conclude that the strong goal orientation of
charter schools, an internal management factor, squeezes out other values
in the implementation of educational policy.
Pitts (2009) also pursues this line of management research but focuses on
federal government agencies, using the 2006 Federal Human Capital Survey
(a large survey of federal employees with over 200,000 respondents). Pitts
creates an internal management index (M1) designed to tap the organiza-
tion’s commitment to diversity management in terms of goals and actual
implementation. He then relates this measure of internal management to a
dependent variable that asks respondents to rate how well their work group
performs, and a second dependent variable that measures job satisfaction.
Pitts finds a strong positive association between internal management (M1)
and perceived job performance. He also finds this aspect of management is
positively correlated with job satisfaction, so it might indirectly affect other
organizational processes and outcomes though increases in job satisfaction.
The studies by Pitts and colleagues fall into a research genre termed
“representative bureaucracy.” This literature documents when and under
what conditions attempts to create a workforce representative of the popu-
lation results in changes in program outcomes. The literature is not
177 Generalizing the impact of internal management

considered part of the public management field, but the basic premise – that
managerial efforts to build human resources of a specific type will generate a
predictable set of program outcomes – is a premise that is part of the public
management agenda. We do not review this extensive literature here; we
should note, however, that the basic relationship (that internal management
affects performance) has been found for child support enforcement agencies
in Missouri (Wilkins and Keiser 2006), local police forces in urban counties
(Meier and Nicholson-Crotty 2006), the federal Equal Employment Oppor-
tunity Commission (Meier, Eller, and Pennington 2005), and the Federal
Housing Administration (Selden 1997), as well as in public schools data,
both nationally in the United States and in the state of Florida (Meier and
Stewart 1991).
Nicholson-Crotty and O’Toole (2004) use the basic management model
in Chapter 2 to study 570 municipal police departments in the United States.
They create an internal management variable by using survey measures that
include both structure and the development of human resources. The
fourteen items produce a single internal management factor that they
designate as M1. As a dependent variable, the authors use the clearance rate
for indexed crimes; these are the serious crimes that the FBI uses to calculate
crime rates in the United States. The internal management measure has a
positive impact on crime clearance rates; it also interacts with past perform-
ance and generates additional impacts via this autoregressive interaction.
Jacobson, Palus, and Bowling (2010) use data from the 1994 and 1998
American State Administrators Project; this is a large survey of state admin-
istrators that asks a range of questions about these managers and how they
perform their jobs. Although their objective is to examine management
behaviors and how they vary by gender, their findings are relevant for this
study. As a dependent variable they measure how extensively “reinventing
government” initiatives were implemented. Reinventing government is a
general reform of the new public management that seeks to bring more
business management techniques and more incentives into public manage-
ment. The authors find a negative relationship between the extent of time
spent on internal management (regular day-to-day operations) and the
adoption of these reforms. The negative relationship is expected, because
these reinventing reforms are being pushed by outside political actors, and
the management measure taps into an internal rather than an external focus
for management.
Andersen and Mortensen (2010) take the O’Toole and Meier notions on
the stability aspects of internal management as the starting point and ask
178 Internal management and performance

whether stability per se is an advantage for organizations. Their organiza-

tions are Danish schools, and they include a measure of resource stability in
a large production function that predicts the performance of 140,000 school
children on standardized tests. They find that stability in resources contrib-
utes positively to student performance over and above the level of resources
or the change in resources.
Cohen, Vaughn, and Villalobos (2010) explicitly use the formal theory
specified in Chapter 2 to study the management of the US Office of the
President. They have more than 300 surveys concerning ten different
individuals who served as chief of staff to the president. The respondents
are all individuals who held positions of authority in the administration,
and they were asked to rate the effectiveness of the chief of staff. The
authors have two different measures of internal management: whether the
chief of staff adopted an administrator role (rather than a policy role)
and an assessment of the chief of staff ’s advice. Both measures of internal
management were positively associated with the respondents’ assessment
of performance.
In a variety of other investigations, therefore, and in numerous other
empirical settings, the effectiveness of various aspects of internal manage-
ment in shaping organizational outputs and outcomes has been demon-
strated. The pattern of findings is quite consistent with the more detailed
research results reported here from the Texas school districts data set.

Conclusions

This chapter has examined three elements of internal management: creating

stable personnel, managing an organization’s human resources, and making
decisions in the face of a significant budget cut. In systematic analyses we have
shown that, for public organizations, the stability of the workforce contrib-
utes to performance, as does the management of the organizations’ human
capital. In addition, basic internal management decisions can be used to limit
the harmful effects of negative shocks to the organization. In several add-
itional ways, therefore, management matters – either directly, as with human
resources management or budget decisions, or indirectly, as with the stability
of personnel. These findings do not reach to still other aspects of internal
management, but they do suggest that managerial actions directed inward as
well as those (discussed earlier in this book) that are directed externally shape
results.
179 Notes

In the next chapter we build on our notion of shocks introduced above to

examine more fully how management might shape outcomes when the
system comes under stress, primarily from unanticipated perturbations to
the organization. We ask if organizations can take actions that permit them
to weather crises. In particular, we examine the role of management capacity
and the development of buffering capacity.

NOTES

1. Sargent (2009) examines other internal management issues, such as goal setting, budget
efficiency, employee training, and technology adoption.
2. This portion of the chapter is adapted from the analysis presented by O’Toole and Meier
(2003b).
3. Overgeneralization should be avoided with regard to the impact of stability on performance.
In this chapter we argue that stability can be helpful, and test one aspect of this idea against
evidence; but we expect the overall impact of stability, as well as of certain types of stability,
to be contingent. We are exploring some of the contingencies in additional research.
4. Recently Andersen and Mortensen (2010) have considered an additional aspect of
stability beyond what is included here: the stability of resource allocation. Based on
Danish public educational data, and consistent with the argument of the present chapter,
they find that budgetary stability helps organizational performance.
5. Their primary term for this set of features is “treatments,” by which they mean “primary
work or core processes or technology” (Lynn, Heinrich, and Hill 2000: 15).
6. Despite the stereotype of bureaucracy as unchanging, substantial personnel turnover
exists, and it varies greatly across public organizations (Kellough and Osuna 1995).
7. Such stability can also reduce policy churn – the adoption of frequently changing reforms
without leaving sufficient time for implementation. Policy churn is identified by Hess
(1999) as a major problem affecting urban school system performance.
8. Structural stability is largely constant across the entire sample examined in this study. We
are pursuing some additional aspects of stability in work not reported here.
9. As a result, the measure taps both stability and an aspect of capacity – the latter in the
sense of knowledge about the organization.
10. Turnover in organizations varies widely. School districts are similar to other public
organizations in the level of turnover (see Meier and Hicklin 2008).
11. We assessed the normal problems of serial correlation and heteroskedasticity in pooled
models. We include individual dummy years to control for the changes in variables from
year to year. Diagnostics showed only marginal levels of heteroskedasticity that should
not affect the results.
12. This level of exceedingly high turnover could be a behavioral symptom that, in turn,
is driven by other causes. Several of the most plausible sources of turbulence are
included in the model via the set of controls; obviously, idiosyncratic factors are not.
Since the focus of this research is to explore the influence of the impact of
180 Internal management and performance

management and stability on performance, not to explain turnover per se, the issue is
not pursued further here.
13. We also reran the analysis for a much larger sample: all 1,000þ districts in Texas. Given
the five-year time series, this exploration amounts to a 5,000-case data set. Doing so
requires dropping two independent variables derived from the survey: managerial net-
working and managerial stability. In this estimation, teacher stability maintained its
impact and is statistically significant. The impact of management quality also continues
to appear.
14. For the autoregressive form, the comparable figure is approximately 4 percent. Although
this result may not seem overwhelming, the autoregressive term means that an increase in
performance today grows the base for future increases as well, via the lagged dependent
variable, and therefore improvements reverberate forward into the future. If the improve-
ments continued indefinitely, the total impact would be 13.3 percentage points, relatively
close to the 11 percentage points without the lag.
15. Including the analysis for overall TAAS pass rate results in totals of thirty-seven properly
signed coefficients out of forty, with more than two-thirds statistically significant.
16. The sensitivity of black student TAAS performance to personnel stability at school seems
more general. Note the enhanced impact of managerial stability as well, surely an influ-
ence at some remove from most students’ day-to-day educational experience (Tables 5.2
and 5.3).
17. Note in this regard, for example, the function of M3 in our theoretical model.
18. For these analyses, the performance indicator used is the overall TAAS pass rate. The
estimations omit the lagged dependent variable but include all other controls.
19. This section treats human resources and their management seriously but is not designed
to explore McGregor’s advocacy for so-called “theory Y” over “theory X” (or vice versa).
The analysis in this section draws on that presented by O’Toole and Meier (2009).
20. We also use the 2006/7 survey in Table 5.5; that survey had a response rate of 67 percent.
Districts responding to the surveys were no different from nonrespondents on key
variables such as enrollment, enrollment growth, students’ race, ethnicity and poverty,
or test scores.
21. In addition, performance measures (the dependent variables) are reported for school
districts only if the district had performance data on five or more students. For certain
measures, such as the black pass rate on the statewide standardized exam (white subur-
ban districts) and SAT scores (more Texas college-bound students sit for the ACT
instead), some districts have no performance data reported.
22. As of 2003 the TAAS had been replaced by the Texas Assessment of Knowledge and Skills.
The results of the two examinations correlate very highly, nonetheless.
23. These items were asked on the 2004/5 survey. They were also repeated on the 2006/7
survey. The results are presented in Table 5.4 and the actual values used are from the
2004/5 survey.
24. To investigate further the causal direction, we conducted a panel version of Granger
causality analysis. The results were ambiguous, and we could not rule out reciprocal
causation for this and the other nine indicators. Measures at two time periods, particu-
larly with different respondents for some of the time periods, did not provide enough
leverage to sort this out statistically.
181 Notes

25. The analysis here draws from that presented by Meier and O’Toole (2009a).
26. This estimate is carried out on the panel so that the individual estimate for a district is
based on its entire history in the data set. The key estimation question is whether to
estimate one set of parameters for the entire data set or use individual parameters for
each district. While in theory this distinction is important, in practice the results are
correlated at 0.96. For efficiency purposes, therefore, we used the estimates based on a
single set of parameters rather than more than 1,000 sets.
27. It might be supposed that managerial quality should have something to do with organ-
izational responses to negative budgetary shocks. Analysis of the data shows no signifi-
cant effect of quality. This finding is not especially surprising, given the salary-based
quality measure used in this data set. The top managers are likely to be rewarded for what
they do on a day-to-day basis, and budgetary crises do not occur with great frequency.
6 Nonlinearities in public management:
the roles of managerial capacity and
organizational buffering

Chapter 5 mentioned that, in the classic Sherlock Holmes tale “Silver Blaze,”
Arthur Conan Doyle’s famous detective infers an important finding lurking
behind an apparent non-event.1 Similarly, in the last section of the previous
chapter we built from something that did not happen – in this case, a very
limited negative impact on public program performance even in the face of
sizable and negative budget shocks from the environment – to highlight
ways that public managers are able to protect their organizational systems
from unanticipated and unpleasant disruptions to maintain performance
in the face of adversity. Because managers made a series of decisions that
reflected key priorities and long-term goals, the “dog that didn’t bark” in this
latter instance was a set of school systems that did not appreciably suffer – at
least in the short run.
This chapter follows the earlier analysis, at the intersection of public
management and organization theory, to explore a more general process
that bureaucracies use in the face of potentially disruptive circumstances. We
first examine the question of whether and how the presence of managerial
capacity in public organizations might provide protection or support for
public agencies facing environmental battering. Apart from the day-to-day
efforts on the part of managers to encourage efficient and effective produc-
tion, in other words, we ask if reserve capacity has positive impacts on
performance. Can capacity be activated in times of crisis to protect the
organization? This is the first core research question explored here. Chapter 5
offered a crisis response that entailed a series of decisions that are highly
specific to school districts. Our purpose is to determine if more general
principles hold that could be applied to other organizations.
The role of managerial capacity is examined as one of the nonlinear relation-
ships in our theory. Essentially, we expect managerial capacity to interact with
environmental shocks to lessen their impact. After examining two kinds of
environmental shocks – budget cuts and a natural disaster – we proceed to
probe additional nonlinear aspects of our managerial theory: the relationship
182
183 Managerial capacity and budget cuts

of managerial capacity and networking together as they affect performance and

the more general nonlinearities of organizational buffering in the context of our
theory. In all four cases we take the nonlinear aspects of our theory seriously
and seek to determine a set of interactions or different nonlinear specifications
to establish which best fits the experiences of these organizations.

Managerial capacity and budget cuts

Several recent research efforts have moved public management to the central
concern of organization theory – performance (Kelman 2008) – by offering
evidence that public management and public managers make a difference in
delivering the outputs and outcomes of public organizations (see, for instance,
Lynn, Heinrich, and Hill 2001, Meier and O’Toole 2001, 2003, O’Toole and
Meier 2003b, 2004a, 2004b, Donahue et al. 2004, Ingraham and Lynn 2004,
Andrews, Boyne, Law, and Walker 2005, Brewer 2005, Chun and Rainey 2005,
Martin and Smith 2005, Boyne et al. 2006, and Hicklin, O’Toole, and Meier
2008). Such varied managerial features as networking behavior, strategic stance,
and the stability of managerial and front-line personnel are linked to stronger
performance. What of the relative size, or capacity, of the managerial cadre,
though? The relative size of the management cadre is important theoretically
because it is a fashionable target for journalists and management consultants
who condemn bureaucracy. This chapter seeks to bring some empirical
evidence to bear on this popular nostrum.
Management capacity has attracted interest from public management
researchers and practitioners, and efforts have been made to develop data
on the relative capacity of different governmental agencies and different
units of government (Ingraham, Joyce, and Donahue 2003). These data have
not thus far been tied clearly to information about program outcomes,
however. The present chapter taps an important aspect of managerial
capacity across a large number of public organizations, and estimates the
impact of capacity in mitigating the negative performance repercussions of
downward exogenous shocks.

Capacity as managerial potential

Protecting public organizations and programs from disruption is a core man-
agerial function (for a classic depiction, see Thompson 1967; or, more
recently, O’Toole and Meier 2003a), even if the emphasis in recent literature
184 Nonlinearities in public management

has been on the proactive and entrepreneurial aspects of public management.

Protection and defense are important, even if currently underemphasized,
aspects of management; and evidence has been offered on behalf of the role
of these elements in contributing to performance. Studies of strategic man-
agement explore a “defender” approach to dealing with the organizational
environment (Miles and Snow 1978), and defenders can outperform other
strategic stances in some settings (Meier et al. 2007). Personnel stability,
including managerial stability, also contributes to outcomes (see Chapter 5).
Most significantly for present purposes, subtle managerial efforts internally
can protect the core organizational tasks from performance disruptions.
The idea here is akin to the notion of “disturbance handler,” as characterized
by Mintzberg (1973). Even if environmental (budgetary) shocks enter
the organization and threaten to wreak havoc, managers are far from
impotent; they can reallocate staff and resources toward the highest-priority
tasks within the organization and continue to deliver results with minimal
disruption (see Chapter 5), at least for a while and at least within some
limits.2
There are, accordingly, theoretical and empirical reasons to explore the
various ways that public managers either buffer or dissipate negative shocks.
This subject links the theoretical interest of public administration in organ-
izational constraints with the organization theory focus on performance
(Kelman 2008). The issue is important in practice as well, since no public
organization, no matter how well supported and how protected from its
environment, is immune from unpleasant and often unanticipated shocks.
In the present section we move beyond the aspects explored in earlier
studies to examine whether managerial capacity per se can mitigate the
impact of substantial budgetary shocks (for a recent review of the literature
on capacity, including the complex ways that the concept has been used, see
Christensen and Gazley 2008). Ingraham, Joyce, and Donahue (2003)
observe that the notion of “capacity” has been defined in varied ways (see
also Malysa 1996) but is typically considered a concept with multiple
dimensions (15). As they indicate, “By capacity, we mean government’s
intrinsic ability to marshal, develop, direct, and control its financial, human,
physical, and information resources” (15, emphasis added in the latter
instance). Ingraham, Joyce, and Donahue then go on to stipulate four
“key levers” that, they argue, drive or feed into capacity. We build from
their general definition but treat the “levers” aspect more abstractly.
In particular, these researchers reference governments’ “intrinsic ability”
to get things done, and this framing of the concept draws one’s attention not
185 Managerial capacity and budget cuts

to actual operations but to the “potential” or “reserve” available for handling

the varied tasks of management. Management capacity, therefore, is not
management effort or practice but, rather, what could be mobilized if
needed. An analogy drawn from the field of physics comes from the distinc-
tion between kinetic and potential energy, the former constituting energy in
operation or execution, the latter the possible energy available in a system.
How might the management capacity of an administrative system be
assessed? The answer is not obvious, since capacity cannot be directly and
operationally observed; it constitutes a potential for action rather than action
in practice. Capacity, moreover, probably has multiple dimensions, and there
is no real evidence regarding which aspects might be most important and
under which circumstances. Accordingly, we work from a general notion that
reaches to potential that could be mobilized in varied ways. The relative size of
the administrative corps of a system should tell us something of the capacity
of that system, but some of the administrative personnel have regular line
responsibilities and are not easily mobilized or deployed to deal in nonroutine
ways with the challenges raised by sizable budgetary shocks. Our empirical
study focuses on school districts, and so we consider the relative size of a
district’s central office staff as a rough measure of the management capacity in
the organization. Central staff have regular responsibilities, but in the typical
educational system the central office is the locus for financial planning,
human resource analysis, data gathering, and system leadership. Central
administrators would be more likely to have organization-wide views and
also be less likely to be solidly booked with running day-to-day operations.
These perspectives and related functions are those that, we could expect, are
crucial for maximizing the operational capacity of the system in times of stress
(see Yukl 2006: 364 ff.). Accordingly, we consider whether this measure of
managerial capacity of school districts can contribute to performance when
budgetary crises threaten.3
Central office staff are obviously not dead weight in such administrative
systems. Under normal circumstances, they are occupied with a variety of
tasks, including manifold analytical functions as well as efforts to diagnose
and address chronic problems facing the organizational system, whether legal,
political or production-related. One example of this last-mentioned type of
chronic challenge, for the case of school districts, would be efforts to devise
programs to improve student attendance. Doing so would probably boost
performance on other indicia over the longer term. In a sense, however,
our conceptualization of management capacity considers central staff as
representing a kind of (partial) slack in the managerial resources available
186 Nonlinearities in public management

for near-term production, and thus as a potential for action that may not be
fully realized except under relatively unusual circumstances – when tasks can
be reassigned and central staff can directly address immediate performance-
related needs. Indeed, the central office (headquarters) is also the location
where one might expect some slack in human resources, such as it is, to be
stockpiled – if there is any conscious effort to build such slack into the system.
The logic of storing slack in administrative capacity is based on the notion of
flexibility, innovation, and relative payoffs. Adding a single person to a line
production position is likely to increase production by a marginal amount,
but that person is unlikely to be usable for other functions should the need
arise. Similarly adding a person as a line administrator could well improve
day-to-day responsiveness, but these skills would not necessarily be transfer-
able in times of emergency. Storing slack within the central office provides the
greatest flexibility, however, because a central manager can be moved from
seeking grants one week to assessing the profitability of food services the next
week to an emergency fill-in for a line manager the next. Theoretically,
organizational slack of this sort is best stored at the managerial levels, where
it translates into increased management capacity.
Earlier research and theorizing on slack can help to clarify this notion.
Thompson (1967) and Galbraith (1973; see also Pfeffer and Salancik 1978)
argue that slack can serve as a buffer to help organizations absorb and survive
the effects of shocks. Slack is therefore conceptualized as resources that can, if
needed, be mobilized as inputs for the technical core during turbulent times.
Cyert and March (1963) point out that slack can also be seen as resources
available on behalf of innovation (see also Doig and Hargrove 1990: 3).
Organizations with slack are likely to be more innovative – particularly
so when the slack is managerial. Innovative organizations are more likely
to sustain their level of performance when shocks occur given their orienta-
tion of seeking new ways of dealing with problems. Organizational slack,
therefore, including managerial slack, should be positively associated with
performance in organizations experiencing shocks.

A reduced model of management and performance

To consider the role of managerial capacity in assisting administrative systems in

recovering from negative environmental shocks, we rely on a simple model. Since
we are seeking answers about how management capacity, rather than specific
managerial functions such as internal management or external networking,
187 Sample and measures

might carry implications for performance, we model the situation even more
straightforwardly than is the case in our model introduced in Chapter 2. Here we
start with a measure of managerial capacity (Mc) and a set of environmental
forces (X) that are related to organizational performance (O):
Ot ¼ b1 Mc þ b2 Xt þ et ð6:1Þ
We want to separate out from this environmental term (X) some shock to
the system; we call it X’:
0
Ot ¼ b1 Mc þ b2 Xt þ b3 Xt þ et ð6:2Þ
From this general linear model, we add a bit of complexity based on what is
suggested in some of the qualitative literature. Specifically, we would expect
management, or some forms of management such as managerial capacity, to
interact with the environmental shock and reduce the impact of the shock
on the organization. Specifically, we operationalize the following model,
which includes such an interaction:
Ot ¼ b1 Mc þ b2 Xt þ b3 X0t þ b4 Mc X0t þ et ð6:3Þ
Essentially, the argument that managerial capacity, once mobilized, can
mitigate environmental shocks suggests that b3, the coefficient for the shock,
should be negative but that b4, the coefficient for the interaction term,
should be positive and of such a magnitude as to cancel out the negative
impact of the shock.

Sample and measures

The present analysis uses data drawn entirely from the Texas Education
Agency for eight years (1995 to 2002) for a total of 8,329 cases for analysis.
Missing data on individual items reduces this number somewhat in individ-
ual equations. We are interested, then, in budgetary shocks, managerial
capacity, and their performance consequences, while we control for a set
of other influences. Each of these variables is introduced in turn.

Measuring shocks
Our budget shock measure, introduced in Chapter 5, follows Ratts (1999),
who examines economic shocks to national economies. We regress total
school district revenues (logged) on its logged values for the prior year. As
188 Nonlinearities in public management

explained in the preceding chapter, this regression essentially estimates

what school district revenues would be if past trends continued. We then
designate as “shocks” any year-to-year reductions in school district budgets
that exceed 10 percent of revenues. A total of 730 budget shocks occurred in
this eight-year period, thus affecting approximately 8.8 percent of the cases.
In all equations here, a budget shock of 10 percent or more is converted to a
dummy variable.

Managerial capacity
To deal with a shock to the organization, one would expect managers to
mobilize the available managerial capacity to analyze the nature and extent
of the shock and design strategies for mitigating the impact on the organiza-
tion. Without some surplus capacity, one would expect that the reallocating
of managerial time to deal with the shock would result in lower performance
in the short term, as managers – particularly line managers – neglect their
day-to-day duties in order to deal with the shock. Our measure of manager-
ial capacity is the percentage of school district employees engaged in central
office administration. This includes the superintendent, assistant superin-
tendents, the basic administrative support staff (budgeting, personnel, etc.),
and any centralized analytical capability. Capacity measured this way is
relatively common in the empirical studies of organizations (Dalton et al.
1980). Texas school districts are exceptionally lean in terms of administra-
tion. The average district during this time period had only 1.89 percent of
total employees allocated to central office administration, with a standard
deviation of 1.42. Nine out of ten school districts had between 0.71 percent
and 3.54 percent central administrators.

Performance indicators
A preliminary analysis incorporated ten different performance indicators
in an effort to determine how budget shocks affect a variety of organiza-
tional outcomes. The results in the analysis indicated that, for most of the
performance measures, managers were able to reallocate funding and staff
in ways that resulted in no statistically significant reductions in perform-
ance. For three of the ten measures there were negative performance
impacts in the year in question and/or in the following year (see Chapter 5).4
We focus in this section entirely on these three indicators – the Texas
Assessment of Academic Skills, the percentage of students who took either
189 Findings

the SAT or the ACT, and the percentage of students who score above 1,110 on
the SAT (or its ACT equivalent). Since these measures cover both the core task
of the organization and the more difficult high-end objectives, they should
provide a good overall view of how the organization deals with shocks.

Control variables and methods

We include the eight control variables used in earlier models. These variables
are all commonly used in education production functions (Todd and
Wolpin 2003). Assessments of the equations showed serial correlation, so a
set of dummy variables for individual years has been included in the analysis
as an adjustment. Pooled diagnostics for heteroskedasticity show only
modest evidence of heteroskedasticity; estimation with robust standard
errors generates results similar to those presented here.

Findings

Our first question is to ask what the impact of a budget shock of 10 percent
or more is on the performance of the organization. Because organizations
are highly autoregressive systems, they might be able to shrug off the impact
of a shock immediately by short-term adaptations but absorb greater losses
as a result in future years. The analysis in Chapter 5 revealed that budget
shocks affected the organization in the first and second years but had
no impacts in the third and fourth years. Accordingly, we include both
the initial shock to the organization and a shock that is lagged by one year.
Table 6.1 provides a regression of overall TAAS rates on a budget shock while
controlling for the other resources and constraints of the organization.
A budget shock has only a modest impact on the organization’s performance
in the first year; it results in only a drop of one-half of a point in TAAS
performance, all other things being equal, and this relationship is statistically
significant only if we use our directional hypothesis and a 0.1 one-tailed test
of significance. The impact of the shock in the second year is much stronger,
however, and clearly significant (a drop of about one percentage point in the
TAAS pass rate).
The impact on examination pass rates, particularly the delayed impact, is
clearly an important finding. Given how important school districts’ performance
on this metric is considered, it seems clear that, despite short-term efforts to
190 Nonlinearities in public management

Table 6.1 The impact of a 10 percent or greater budget shock on students’ state
examination performance

Dependent variable ¼ student TAAS exam pass rates

Independent variables Slope T-score

Budget shock 0.4938 1.59

Lagged budget shock 0.9956 3.05*
Control variables
Teacher salaries (000s) 0.6596 12.28*
Class size 0.3578 9.19*
Teacher experience 0.0826 1.72*
Noncertified teachers 0.1386 7.92*
Percentage state aid 0.0136 3.09*
Percentage of black students 0.2329 27.50*
Percentage of Latino students 0.1031 18.59*
Percentage of low-income students 0.1472 19.13*

R2 0.61
Standard error 7.66
F 773.71
N 8,321

Notes: Dummy variables for individual years not reported. * ¼ significant at p < 0.05,
one-tailed test. Time period ¼ 1995–2002.

protect the educational system from disruption, they are not completely success-
ful. School system superintendents overwhelmingly identify TAAS performance
as their highest priority (see Chapter 5), and budget shocks certainly impede that
objective.
Table 6.2 shows the impact of a 10 percent or greater budget cut on
the other two performance indicators – the percentage of students who take
either of the two standard national college entrance examinations and the
percentage of students who score above 1,110 on the SAT or its ACT
equivalent. For the former, the estimation shows that a budget shock reduces
the test taking rate by approximately 1.39 percentage points in the first
year – a sizable drop. This reduction in the performance criterion may be
due in part to efforts to maintain and protect other key educational activ-
ities. There is no statistically significant impact of the budget shock in the
following year, however. The second set of regression results reported in
the table show that the percentage of students who score 1,110 or above on
the SAT – a level designated by the state of Texas as “college-ready” – drops
by approximately 1.28 percentage points in the first year of the budget hit
191 Findings

Table 6.2 The impact of a 10 percent or greater budget shock on the performance of college-bound
students

Take SAT/ACT test Score 1,110þ

Independent variables Slope T-score Slope T-score

Budget shock 1.3918 2.06* 1.2819 2.84*

Lagged budget shock 0.6173 0.86 0.7977 1.68*
Control variables
Teacher salaries (000s) 0.3529 3.04* 0.4749 6.23*
Class size 1.4456 15.66* 0.1535 2.51*
Teacher experience 0.9479 8.90* 0.1444 2.05*
Noncertified teachers 0.0061 0.16 0.1118 4.30*
Percentage state aid 0.0224 2.27* 0.0608 9.36*
Percentage of black students 0.0329 1.81* 0.0024 0.20
Percentage of Latino students 0.1437 11.46* 0.0037 0.45
Percentage of low-income students 0.3617 20.11* 0.2545 21.42*

R2 0.14 0.30
Standard error 15.18 10.07
F 71.00 189.56
N 7,449 7,526

Notes: Dummy variables for individual years not reported. * ¼ significant at p < 0.05,
one-tailed test.

and then another 0.8 percentage points in the next. The total impact on
the college-ready student cohort is a matter of real concern. On average, only
19 percent of students meet this criterion; a drop of 2.08 percentage points
over two years, therefore, translates into an 11 percent drop in students
meeting this criterion.
Table 6.3 examines whether managerial capacity can mitigate the impact
on the TAAS pass rate of a 10 percent or greater budget cut. Because we are
estimating the shock in both the first year and the second year, we add the
interaction of managerial capacity with both these shocks. The estimations
control for all variables included in Table 6.1, but only the relevant coeffi-
cients are presented. The intuition about an interaction term is that the
slope of a relationship (between shocks and performance) changes contin-
gent on the value of some other variable (management capacity). To see how
this occurs, we illustrate using the impact of the shock in the first year. To do
this we need both the shock coefficient and the interaction coefficient:
O ¼ 1:24 shock þ 0:28 ðshock  capacity Þ
192 Nonlinearities in public management

Table 6.3 The impact of a 10 percent or greater budget shock on students’

state examination performance

Dependent variable ¼ student TAAS exam pass rates

Independent variables Slope T-score

Budget shock 1.2411 2.42*

Lagged budget shock 2.2501 4.28*
Management capacity 0.0089 0.11
Management capacity times
budget shock 0.2813 1.63*
Management capacity times
lagged budget shock 0.5027 2.90*

R2 0.61
Standard error 7.64
F 661.02
N 8,319

Notes: Equations also control for teacher salaries, class size, teacher
experience, noncertified teachers, percentage state aid, percentage of
Latino students, percentage of black students, percentage of low-income
students as well as dummy variables for individual years. * ¼ significant at
p < 0.05, one-tailed test.

If we rearrange the terms, we get an equation that tells of the impact of the
budget shock – that is, the slope – at any level of managerial capacity:
O ¼ ð1:24 þ 0:28 capacityÞ  shock
Various values can be substituted into this equation to calculate the effect of
a shock at a stipulated amount of managerial capacity. For example, an
organization with only 0.71 percent central office staff (the tenth percentile)
would suffer a reduction of about 1.04 in the TAAS pass rate that first year.
In contrast, a well-staffed central administration of 3.54 percent (the nineti-
eth percentile) would suffer a negative impact of only 0.26 points – a result
that is itself not statistically significant.
For the second year of the shock, a similar set of calculations can be made.
The equation for the slope is
O ¼ 2:25 lagged shock þ 0:50 ðlagged shock  capacityÞ;
or
O ¼ ð2:25 þ 0:50 capacity Þ  shock
193 Findings

To illustrate, then at the tenth percentile we would see a drop of about 1.89
percentage points, and at the ninetieth percentile the result would be –0.48
percentage points.
Two important calculations can be made with these relationships. The
first is to estimate where the slope of the line becomes zero and thus the
shock has no impact on the organization at all. This can be done by
taking the first derivative of the expression with respect to shocks and
setting the result equal to zero. For the first year of the shock, this occurs
when central office staff exceed 4.41 percent of total employment (about 5
percent of the cases); for the second year of the shock, the respective value
is 4.48 percent, or essentially the same level. One can also take the
formula for the confidence limits and calculate when the slope ceases to
be statistically distinguishable from zero (or statistically significant, in
layperson’s terms). For the first year of the shock, that occurs when
management capacity exceeds 0.764 percent (well below the mean); for
the second year of the shock, this occurs at 2.38 percent central adminis-
tration – a level exceeded by 24 percent of all school districts. In short,
greater management capacity appears to mitigate the negative impact of
budget shocks on the TAAS, and the level of capacity is well within the
range of existing organizations.
What about management capacity’s impact on what budget shocks do to
the number of students undergoing college testing? The first two columns of
Table 6.4 show the results of a regression constructed in a fashion parallel to
that for the TAAS pass rate. For this performance measure, the shock has
no impact in the first year if the management capacity measure reaches
4.01 percent of total staff – a value exceeded in 7 percent of the districts. In
the second year, districts with a central staff larger than a mere 0.34 percent
experience no negative, lagged impacts from the budget cut. Almost all the
districts – 97.9 percent – have this minimum level of managerial capacity.
The slope ceases to be statistically different from zero in the first year at a
managerial capacity value of 2.49 percent (23 percent of the districts), and
in the subsequent year at central staff size of 0.33 percent and above
(98.1 percent of the districts).
Finally, the last two columns of Table 6.4 report the results for the
percentage of students scoring above 1,110 on the SAT or its ACT equiva-
lent. The budget shock has no effect in the first year when central staff are
3.70 percent or more of the total district employment (9 percent of all
districts) and no effect in the second year when managerial capacity equals
or exceeds 1.21 percent (64.2 percent of all districts). The slope ceases to
194 Nonlinearities in public management

Table 6.4 The impact of a 10 percent or greater budget shock on the performance of
college-bound students

Take SAT/ACT Test Score 1,110þ

Independent variables Slope T-score Slope T-score

Budget shock 2.7284 2.18* 3.0814 3.51*

Lagged budget shock 2.9939 2.30* 2.7196 3.21*
Management capacity 0.7496 3.40* 0.8213 5.65*
Management capacity times 0.6801 1.32 0.7359 2.07*
budget shock
Management capacity times 1.2044 2.24* 1.1732 3.16*
lagged budget shock

R2 0.14 0.30
Standard error 15.17 10.05
F 61.14 163.66
N 7,449 7,526

Notes: Equations also control for teacher salaries, class size, teacher experience, noncertified
teachers, percentage state aid, percentage of Latino students, percentage of black
students, percentage of low-income students as well as dummy variables for individual
years. * ¼ significant at p < 0.05, one-tailed test.

have an impact distinguishable from zero at managerial capacity values of

2.63 percent (20 percent of all districts) and 1.39 percent (54.3 percent of
all districts) for the first and second years, respectively. In short, higher
levels of managerial capacity can indeed mitigate the negative perform-
ance-related impacts of sizable budget cuts for all three indicators
examined.

Implications

In the era of new public management, bureaucracy is an epithet frequently

used to criticize public organizations. Such a narrow view of bureaucracy
overlooks the need for governments to have the capacity to respond
to problems as they occur. This section has examined how organizations
respond to budget cuts and finds that a little “bureaucracy” might be a good
thing.
195 Implications

In a sample of 1,000 public organizations over an eight-year period, the

evidence indicates that, when faced with significant budgetary shocks, some
units absorb a good deal of the unpleasant event without experiencing
much, if any, performance decline. Some sacrifices to performance do,
nonetheless, occur. This section has explored the impact of managerial
capacity in mitigating the impacts of budget shocks on the most vulnerable
outcomes of public education systems.
The most obvious findings are straightforward. In these instances, the
analysis shows that managerial capacity interacts with substantial budget
shocks and reduces their impacts. At sufficiently high levels of managerial
capacity – the level depends on which year and which performance measure –
the administrative systems are protected from any performance drop. Such
impacts are defensive only – that is, bureaucracy can mitigate the negative
impact of budget cuts; as might be expected, it cannot turn them into a
positive outcome for the organization.
Earlier research in this book demonstrated that management, including
internal management, can indeed contribute to performance. In most of
those studies, the decision-making and/or behavioral moves of managers
add value. In this section, by way of contrast, the evidence supports the
notion that latent or potential managerial resources – managerial capacity –
can be mobilized to blunt the impact of negative shocks on public organiza-
tions. In terms of the model tested (Equation (6.3)), managerial capacity can
be considered an aspect of M1 feeding, when active and mobilized, into
S. Our measure of this reserve “army” of management is the relative size of
the central office staff. Although front-line workers in educational systems
are absolutely critical for delivering results, it is logical for there to be a kind
of latent performance bonus associated with central staff. The actions and
the analytical effort on the part of a centrally positioned staffer, once
mobilized to protect the system from the effects of a shock, might well be
felt at the margin in many parts of the larger organization – for instance, in
multiple classrooms and multiple schools.
Thus far, we have emphasized the contribution that managerial capacity
can make to performance. We have done so by focusing particular attention
on the impact, or non-impact, of budget shocks. Concentrating on shock
events and their consequences allows us to see in full relief the positive
aspects of carrying such capacity in administrative systems – some sort of
stabilizing mechanism, perhaps, that maintains the organizational ship on
course despite budgetary gales or even hurricanes. This interpretation is
accurate as far as it goes – but also incomplete. If only the subject of
196 Nonlinearities in public management

managerial capacity were so simple. Unfortunately, capacity can cut both

ways. It is worth looking closely in Tables 6.3 and 6.4 at the coefficients for
managerial capacity alone – that is, the impact of capacity on performance
in the absence of a budget cut. Table 6.3 shows that, with regard to pass rates
on the statewide standardized exam, managerial capacity does not constitute
a drag on the system in “normal” times, and this finding is both interesting
and substantively significant. Table 6.4 offers a different story for the high-
end performance metrics associated with college-bound students, however.
For each of these measures, managerial capacity is negatively related to
results for systems not experiencing sizable budget shocks.5
The result suggests, in other words, another key decision that should be
considered by public managers: how much to protect performance against
shock, on the one hand, versus how much to seek to maximize performance
during more typical times, on the other. A second key decision is how much
to stress basic performance versus top-end college goals. There is no gener-
ally applicable optimal point in this balancing act. The choices made by
managers in particular places and at particular times are likely to be a
function of several considerations – including the history of and expect-
ations regarding large budget cuts and other shocks, the relative salience of
various performance criteria, and the value placed on maintaining core
production in difficult times or boosting performance during the more
usual periods.
In a sense, these results and their implications might seem sensible, even
expected. They certainly offer a performance-related rationale for crafting
some degree of slack into administrative systems, and that is a far cry from
other motivations sometimes attributed to bureaucrats – such as the argu-
ment for budget (or discretionary budget) maximization sketched by some
public choice theorists.
The findings also suggest that, to the extent that slack is being built into
such organizations, it makes sense to locate it in a central office, rather than
(for instance) at the front line. The marginal degree of help in assisting
performance potentially available from a central analyst or manager, once
mobilized, is likely to be greater in terms of system results than the marginal
benefit from an additional skilled instructor – even though a cadre of the
latter is absolutely essential for educational excellence (Hanushek, Kain, and
Rivkin 1998).
This section provides some performance-tested validity to the idea that
managerial capacity generates results and raises questions about some of the
injunctions of the so-called new public management, which emphasizes lean
197 Managerial capacity and natural disasters

administrative systems and the market-oriented management of programs.

Such arrangements may carry advantages on occasion, but they are likely
to provide little assistance or protection when budgets become tight
or unstable. Rather, mobilizing managerial help to shift capacity into
actuality can provide results that the latest NPM innovations would be likely
to miss.

Managerial capacity and natural disasters

Managerial capacity appears to mitigate negative budget shocks. Is it pos-

sible to generalize about the role of managerial capacity with regard to other
shocks?6 After all, managers and organizations must sometimes cope
with sizable negative shocks that land without notice inside the organiza-
tional system and simply have to be handled. A classic case of just such an
eventuality occurred on the US Gulf Coast in 2005, when two major
hurricanes descended upon the same region within weeks of each other
and caused massive destruction and considerable loss of life. Aside from
those who were required to deal swiftly with the emergency needs of the
moment (the police, fire and rescue, disaster relief, public health, and other
such programs and agencies) many additional organizations and their man-
agers had to manage major shocks stemming from the hurricanes and their
aftermath – perturbations that had penetrated their organizations and posed
substantial managerial challenges.
Public organizations and public managers, in short, sometimes face the
“fire and rain” (to borrow singer-songwriter James Taylor’s imagery) from a
major unanticipated disruption and have to seek to mitigate its negative
impacts. In this section, we examine via a natural experimental design how
the performance of a large set of public organizations – public school
districts – in the Gulf Coast region was affected by the hurricanes of 2005.
We also explore the key question of whether and how aspects of manage-
ment capacity were able to reduce or eliminate measurable disruptive
impacts.
On August 29, 2005, Hurricane Katrina slammed into the Gulf Coast near
the Louisiana–Mississippi state line. An estimated 1,900 deaths were attrib-
uted to Hurricane Katrina and the subsequent flooding; property damage
was estimated at $81.2 billion. The flooding of New Orleans and subsequent
problems resulted in a mass evacuation of Louisiana residents. Included in
these evacuees were 46,503 students (plus their families) who were relocated
198 Nonlinearities in public management

to Texas and enrolled in Texas public schools. Of these students, a total of

35,091 remained in Texas schools until the end of the 2005/6 school year.7
While coping with the widespread devastation from Hurricane Katrina
and the relocation of thousands of students and their families, the people of
the Gulf Coast region took a second blow – this one from Hurricane Rita.
The storm made landfall near the Texas–Louisiana border on September 24,
2005. Although only seven fatalities were attributed directly to the hurricane,
the hurricane caused some $10 billion of property damage. Many damaged
facilities were schools in east Texas. The evacuation itself shut down most
schools in the Gulf Coast region; and, because schools further inland served
as evacuee centers, these schools were also closed for a period of time. A total
of 243 Texas school districts were closed for, on average, six days, with some
districts closed for five or more weeks.
The two hurricanes created two distinct natural experiments in terms of
how public organizations respond to environmental shocks. First, many
districts received an influx of students from Louisiana. Given that Louisiana
public schools are generally perceived to be inferior to the Texas schools, and
given the poor urban areas that sent evacuees to Texas, the general percep-
tion was that Texas schools received an unexpected flood of students who
were not likely to perform well in class and who, in addition, were living
with multiple challenges resulting from their evacuee status. Many struggled
with health, housing, and other difficulties. Second, many districts were then
closed for a week or more, thus creating the need to adjust curricula and
lesson plans to the shortened time period.
The important theoretical aspect of these two environmental shocks is
that they both penetrated to the technical core of the organization – that is,
the teaching of students. Many environmental shocks can be screened out
(as indicated in the second term of our model) as management seeks to
buffer or as stabilizing forces dampen the environmental shocks. For
example, a law such as the “No child left behind” act with its massive
reporting requirements might be handled by special reporting units rather
than the schools themselves. In the present case, there was no way to avoid
the arrival of new students or the closing of schools. The addition of new
students or the cancelling of class days directly affected the production
processes – that is, the schools and classrooms – of the school district. This
logic suggests that we then seek information on how school districts
mitigated the impact of the two hurricanes, and that we do so by focusing
on the first term of the model, the internal management and structural
elements.
199 Data and measurement

Data and measurement

The units of analysis are all Texas school districts with 500 or more stu-
dents.8 The smaller districts are excluded from study here because these
units often have highly fluctuating test data (our dependent variable meas-
uring performance) that are overly sensitive to the handful of students who
are examined. In such cases, the ability to control for past performance is
limited, and so the estimation of how much an intervention event affected
performance may be biased or inefficient. The 703 school districts included
in the study range widely on a variety of dimensions, including student
composition (race, ethnicity, etc.), resources, setting (urban, rural, subur-
ban), and performance.

“X”: measuring the environmental shock

Two measures of environmental shock are used in this analysis. First, the
“Katrina student influx” shock is tapped by using the percentage of the
student body in a school district that was composed of Katrina evacuees at
the end of the 2005/6 school year (thus, measured in late spring 2006). The
year-end count is used rather than the initial count, because students often
moved from temporary districts to “permanent” districts in Texas as parents
became employed. A total of 424 out of the 703 districts in the study enrolled
Katrina evacuees as students, with a range from 0 to 5.42 percent of the
district’s total enrollment; these districts contained 99.8 percent of all
Katrina evacuees enrolled in Texas public schools (see Table 6.5). Of those
districts receiving students, the average evacuee enrollment was 0.47 percent
of the overall total; but, as the standard deviation shows, the distribution is
positively skewed.
Second, the shock due to Rita is measured by the total number of days the
school district was closed because of the impact of the storm. (Rita caused a
number of Texas system closures, but Katrina, for which the brunt of the
impact occurred considerably further to the east, caused few Texas district
closures.) Table 6.5 shows that 243 districts were closed approximately one
week (5.14 days), but that the standard deviation indicates a positively
skewed distribution. Twenty-eight districts were closed more than two
weeks. Logic suggests that the relationship between days missed and student
performance is likely to be subject to a threshold effect. Missing a single day
200 Nonlinearities in public management

Table 6.5 Organizational shocks: Hurricanes Katrina and Rita

Relocated students as percentage of enrollment

Mean 0.28
Standard deviation 0.52
Low 0.00
High 5.42
Mean of impacted districts 0.47
Days of school missed due to district closure
Mean 1.36
Standard deviation 3.34
Low 0.00
High 29.00
Mean of impacted districts 5.14

of class is unlikely to cause major problems for teachers or students. To

account for such a threshold, we recalculated this variable to include days
missed only if the period of closing constituted more than one week of
school – that is, six or more days; sixty districts met this criterion.9
The two environmental shocks affected some of the same districts.
Of those districts in the study, 20 percent were hit by both shocks, while
67.1 percent were hit by at least one of the shocks.10 Because these measures
of the shocks sustained are somewhat collinear and because we attempt to
explain the response to the shocks via interactive effects, we estimate the
impact of the shocks both separately and together in the same equation. The
results are highly similar, although the impact of Katrina students is lessened
by its collinearity with the Rita/days measure.11

“O”: outcome measures

There are many ways to evaluate the success of public school systems. By one
relatively low standard, the schools were a clear success; 46,000 students were
absorbed and damaged schools reopened. A more interesting assessment is
how the shocks affected district scores on the Texas Assessment of Know-
ledge and Skills. Because the “official” TAKS pass rate, known as the
accountability subset, permits students to be excluded from the test if the
student is enrolled in special education, has limited English skills, or has not
resided in the district for a sufficient time period, we do not employ the
official rate in our analysis. Rather, we use the pass rate for all students in the
201 Data and measurement

district. This is especially important in picking up the impact of the Katrina

students, since many of them might have changed districts after the late
October deadline and thus be excluded from the accountability subset.
Because the overall pass rate we use includes all students, the pass rates are
lower than those officially used to rate and evaluate districts. For the 2005/6
school year, the average TAKS pass rate when all students are included was
66.4 percent with a standard deviation of 10.9; the all-pass rates are normally
distributed and range from thirty-one to ninety-six.
Many other performance indicators used to assess schools are not par-
ticularly valuable in the present study because the measures are not likely to
be sensitive to environmental shocks or were not collected in time. College
preparation indicators such as SAT or ACT scores, for example, are not
available for approximately one year after TAKS test results are released; in
addition, performance on such indicators reflects only the age cohort that is
taking the test in any given year. Two other possible performance indicators
are available: the “commended” pass rate and school attendance. The com-
mended pass rate is based on a much higher test score; to illustrate, in the
average district only 9.9 percent of students passed all tests at the com-
mended level of performance. Because this measure is affected by a much
smaller number of students and the overwhelming majority of students do
not meet this standard, the influx of students and the missed days is unlikely
to have much impact. Attendance results are tightly clustered, with a mean
of 95.9 percent and a standard deviation of 0.7. This lack of variation means
that finding impacts for any variables, including the hurricanes, will be
difficult. Although the analysis therefore focuses primarily on the TAKS
results, we also note in passing any impacts on these two other measures.

Control variables
Our theory specifies an autoregressive model, and that fits well the logic
undergirding the notion of environmental shocks. Our analysis, therefore,
includes the 2005 TAKS pass rate in all models that estimate 2006 perform-
ance impacts.12 The post-hurricane TAKS results are thus assessed relative to
the pre-hurricane TAKS results (a before–after research design). Although
such an estimation controls for the history of the school district by incorpor-
ating it in this lagged dependent variable, other changes in school district
resources or constraints could also affect performance for 2006. To control
for these factors, we include our five measures of resources and three
measures of constraints. All eight of these measures are change, or
202 Nonlinearities in public management

Table 6.6 Environmental shocks and student performance: the impact of students and missed
class days

Dependent variable ¼ all-students TAKS pass rate

Independent variable Slope Slope Slope

Students 0.479* (1.84) – 0.437* (1.68)

Days missed – 0.138* (2.36) 0.131* (2.23)
Lagged pass rate 0.928* (80.47) 0.928* (80.58) 0.928* (80.70)
Change in
Teacher salary 0.000 (0.98) 0.000 (0.85) 0.000 (0.91)
Instruction funds 0.136* (1.65) 0.137* (1.67) 0.134 (1.64)
Black students 0.139 (0.84) 0.020 (0.13) 0.129 (0.79)
Latino students 0.149 (1.32) 0.179 (1.59) 0.163 (1.45)
Low-income students 0.005 (0.13) 0.008 (0.22) 0.006 (0.17)
Class size 0.654* (2.98) 0.758* (3.42) 0.732* (3.31)
Teacher experience 0.069 (0.40) 0.089 (0.52) 0.064 (0.37)
Noncertified 0.034 (1.09) 0.035 (1.14) 0.032 (1.02)

R2 0.91 0.91 0.91

Standard error 3.21 3.21 3.21
F 735.05 737.55 672.53
N 703 703 703

Notes: T-scores in parentheses. * ¼ significant at p < 0.05, one-tailed test. Time period ¼ 2005/6.

differenced, measures – that is, they measure the change in the variable from
2005 to 2006. All the impact of the variables’ levels – e.g. if resources act as a
stock of capital rather than a flow – should be reflected in the lagged
dependent variable.13

Findings

Table 6.6 presents our findings for the impact of the two environmental
shocks on the Texas school districts. These shocks are estimated separately
(columns 1 and 2) as well as simultaneously in the same equation (column 3).
The third column with both shock measures included in the model shows
that a one percentage point increase in Katrina students (as a percentage of
the student body) is associated with a drop in the TAKS all-pass rate scores
of 0.437 percent (p < 0.05, one-tailed test). Although some of this drop
might have been the result of the originally enrolled students not doing well
as the result of more crowded classes and other factors, this effect size is the
203 Findings

equivalent of 43.7 percent of Katrina evacuees failing the TAKS (in contrast
to the statewide average of 33.6 percent). The maximum total impact
of Katrina students on district performance, based on the maximum of
5.42 percent evacuee students, is approximately 2.4 points on the TAKS
pass rate.14
The missed class days variable has a similar negative and statistically
significant impact on TAKS scores. Each additional day (above five total
days) that schools were closed is associated with a decline in TAKS scores of
0.131 percentage points on the pass rate. Based on the largest value of days
closed (twenty-nine), the maximum impact on performance in districts
suffering from closed schools is estimated to be approximately 3.1 points.15
The remaining factors in the equation are generally consistent with past
research. Clearly, the autoregressive term dominates the equation; it espe-
cially does so given the limitation to districts with 500 plus students. This
point is reflected in the extremely high coefficient of determination account-
ing for 91 percent of the variance in 2006 TAKS scores. Of the differenced
measures, only class size reaches the 0.05 level of statistical significance; an
increase in class size from 2005 to 2006 was associated with a (predictable)
drop in TAKS scores.
Estimating the performance results of the hurricane shocks raises the
important theoretical question of how the districts responded to the
shocks and whether there were factors that could have (or in some
districts did) mitigate(d) these negative results. An analysis of the
residuals from the equations in Table 6.6 confirms that the hurricane-
impacted districts include both positive and negative residuals; some
districts were able to take one or both shocks and still outperform
expectations. East Chambers Independent School District, for example,
was closed for twelve days and had slightly more than 1 percent of its
study body as Katrina students, yet the district scored 6.9 percentage
points above the regression line.
The parsimonious theory of public management that we use implies
that districts might mitigate shocks in one or more of three fashions:
through the stabilizing effects of structural (and other) elements, the
operations of management in supporting and reinforcing performance-
related operations, and/or the inertia that carries established practices
forward into the future (past performance). The easiest explanation to
consider is that for past performance. Prior performance is already in the
model as part of the autoregressive estimation; for past performance to
matter more than it does in Table 6.6, it would have to interact with either
204 Nonlinearities in public management

or both of the shock variables in such a manner that high-performing

districts would be less affected by the shock than low-performing districts.
We tested this idea (analysis not shown), but the impact of past perform-
ance as interacted with the two shocks was not sufficient to overcome the
negative impacts.
If high levels of prior performance do not limit the negative impact of
environmental shocks, then structure and management are the two logical
possibilities. Based on the previous study of budget cuts, managerial cap-
acity is a logical candidate in this regard. Greater central management
capacity permits a set of decisions to be made concerning how to evaluate
the incoming students,16 which schools can be assigned the evacuee stu-
dents, what resources have to be shifted to the needed schools, how one
can restructure curricula to make up for missed days, and what resources
need to be procured from outside the district. By using central management
to make these decisions and perform these tasks, the district does not
pull school-level personnel away from the day-to-day operations of the
district.
To measure central management capacity, which represents a structural
resource that might mitigate the performance impacts of negative shocks, we
use the percentage of total staff that are assigned to central office adminis-
tration. The mean for the current set of districts is 1.34 percent, with a
standard deviation of 0.63.
To determine if management capacity afforded by central structure can
mitigate the negative impact of either or both of the unexpected arrival of
students or the missed school days, we interact this variable with each of the
hurricane shocks. These results are presented in Table 6.7.
Such interactive models induce a fair amount of collinearity and funda-
mentally change the interpretation of coefficients. The standard errors
normally used to determine statistical significance need to be recalculated
as the marginal impact of the shocks is determined given the level of
managerial capacity. To illustrate, the slope for days of school closure
now depends on the value of managerial capacity, and this can be deter-
mined by taking both the slope for the days and the interaction term (from
column 3):

 0:222  days þ 0:095  ðdays  administrationÞ

Grouping the terms gives us

½0:222 þ ð0:095  administrationÞ  days

205 Findings

Table 6.7 Administrative capacity can overcome the impact of environmental shocks

Dependent variable ¼ all-students TAKS pass rate

Independent variable Slope Slope Slope

Students 0.901 (1.60) – 0.864 (1.53)

Days missed – 0.233 (1.54) 0.222 (1.47)
Central administration 0.381* (1.71) 0.454* (2.30) 0.328 (1.46)
Students  administration 0.442 (0.96) – 0.434 (0.94)
Days  administration 0.101 (0.78) 0.095 (0.73)
Lagged pass rate 0.929* (80.80) 0.928* (80.73) 0.928* (80.81)

R2 0.91 0.92 0.92

Standard error 3.21 3.20 3.20
F 617.58 619.18 531.84
N 703 703 703

Notes: T-scores in parentheses. * ¼ significant at p < 0.05, one-tailed test. Equations also
control for change in teacher salaries, per student instructional funds, class size, teacher
experience, noncertified teachers, black students, Latino students, and low-income students.

We can then use this equation to draw a line that will show the impact
of a day of school lost at various levels of administrative capacity (see
Figure 6.1). This figure, which also displays the 95 percent confidence
limits, indicates that, at low levels of administrative capacity, the impact
of missing a day of class is negative and statistically significant. As the
level of central administration increases, however, this negative impact
becomes less, and it becomes statistically indistinguishable from zero at
approximately 1.0 percent of central administrators. The slope equation
can be set equal to zero to get a point estimate of when the negative
impact ceases – that is, has a slope of zero; this occurs when central office
administration is equal to 2.33 percent of total employment. The equation
actually shows a positive slope at very high levels of central adminis-
tration, but this can be ignored for two reasons: these values are not
statistically different from zero, and only a small percentage of districts
have central administration percentages that are high (forty-eight of the
703). The logical conclusion is that, as central administrative capacity
grows, it gradually eliminates the negative impact of the environmental
shock on performance.17
Table 6.7 also shows that central management capacity has a similar
impact on the shock of Katrina evacuee students (see also Figure 6.2). The
impact of evacuee students is strongly negative and statistically significant at
206 Nonlinearities in public management

Effect on performance (change in pass rate)

1
Marginal effect of missed days
95% confidence interval

0.5

–0.5

0 1 2 3 4
Percentage central administration

Figure 6.1 The marginal impact of missed school days contingent on managerial capacity
Effect on performance (change in pass rate)

4
Marginal effect of percentage of displaced students
95% confidence interval

–2
0 1 2 3 4
Percentage central administration

Figure 6.2 The marginal impact of Katrina students contingent on managerial capacity

low levels of managerial capacity but declines in impact as central adminis-

tration increases. The point prediction of zero impact is estimated to be
approximately 1.99 percent central office staff (fairly similar to the days
impact). Again, the positive slope predictions are never statistically different
from zero, and cover only ninety-three of the 703 total districts. In short,
central management capacity appears to mitigate the negative impacts of the
hurricanes.18
207 Management capacity in non-crisis times

Implications
Natural disasters are a different type of shock from budget cuts because there
is no advance warning. As such, a natural disaster is a rigorous test of
the management capacity hypothesis. Indeed, the two hurricanes did reduce
the overall level of TAKS scores in the district (either by the influx of
students or the school closures). Management capacity interacted with these
shocks, however, and was able to reduce their impact on performance. In
short, management capacity operated as organizational slack and had an
important nonlinear impact in crisis situations.

Management capacity in non-crisis times

The previous two sections have demonstrated the crucial role of managerial
capacity in dealing with budget cuts and major shocks to the organization
such as a natural disaster. In both cases, managerial capacity had nonlinear
impacts to mitigate the negative consequences of the shock. These findings
raise the question of whether or not managerial capacity is also useful in
other management efforts and if the impact continues to be in nonlinear
interactions with other factors. One potential candidate for such a set of
relationships is managerial networking, a management activity that has
demonstrated substantial impact on performance. This section investigates
whether managerial capacity can enhance the impact of networking.19
The literature often emphasizes the necessity for – and benefits of – networking
efforts, but effective managerial action entails costs as well. Various reasons thus
suggest the plausibility of the idea that the performance-related effects of man-
agerial networking are likely to be contingent on the capacity of the managerial
cadre to perform or assist with such functions.
This point can best be seen by considering what managers typically do
when they interact with others in their networked environment. Most public
managers face an environment with myriad potential actors who might be
useful to the “core” organization and/or might seek to impede the organiza-
tion in its effort to attain its goals. These actors, in turn, are frequently
linked with yet others (Scharpf 1993, 1997).
Public managers face the prospect of sorting through a potentially large
series of interactions with other actors, individual and corporate. Some
interactions are sure to be mandatory; some are voluntary; and some
contain elements of both. Certain interactions may be one-shot, but, in
208 Nonlinearities in public management

the longer term implementation of policy, managers would need to take into
account repeat pattern formulations (see Stoker 1992). The manager must
decide which external actors to engage; how to indicate this intention and
initiate contact; and, if engagement takes place, what strategy to take. Any
such externally oriented action entails some opportunity costs; managers
cannot do everything at once. There may also be some networking inter-
actions in which both (all) actors will, on average, be better off; some zero-
sum patterns; and some negative-sum interactions in which the strategy will
be to limit losses.
In any of these, the manager has to convince the other actor(s) to engage
and to cooperate – that is, to contribute positively to the manager’s organ-
ization or program. While success might occasionally be achieved through
sheer persuasion, the manager typically must bring something to the inter-
action. Such inducements might be in the form of monetary resources,
superior information, or the capacity to take action if some consensus is
reached. By “bearing gifts” in these exchanges, the manager can offer
something of value to induce the other actor(s) to participate. In principle,
therefore, the payoff from managerial networking will be contingent in part
on the creativity a manager displays in identifying attractive options (a form
of capacity), the inducements that the manager brings to the interaction
(resources), and the ability to convert the results into concrete gains
(another aspect of capacity).
This formulation implies that managerial networking – that is, making
contacts with key actors in the environment for the purposes of identifying
and implementing mutually acceptable, even attractive, jointly determined
decisions – will depend on the resources, broadly construed, that the man-
ager possesses. Managerial capacity can be expected to be one crucial
resource needed to operate effectively in an interdependent environment.
Sorting through such complicated institutional terrain and making the
most appropriate strategic moves, therefore, calls for a considerable amount
of informational and cognitive capacity (Simon 1997). In complex public
management settings, this requirement can perhaps be met via managerial
capacity. Greater managerial capacity can contribute to the choice among
interactions in which to engage, and how. Nor is this all; engaging the right
partners in an appropriate way does not ensure successful completion.
Greater managerial capacity can be used to implement whatever agreement
occurs in the networking process. An environmental actor might even be
more willing to participate if he or she perceives that the manager has the
ability to transform the network discussions into reality.
209 Modeling the impact of managerial capacity

This conceptualization of managerial interactions suggests that manager-

ial capacity more broadly conceived can be expected not only to boost
performance on tasks amenable to managerial influence but also to enhance
the value to the core organization of managerial interactions in the interde-
pendent environment. Managerial capacity, in short, can be viewed as the
potential to analyze and evaluate relationships with environmental actors
and the potential to implement any decisions that result from those
interactions.

Modeling the impact of managerial capacity

We begin with several aspects of public management. Substantial earlier

research in this volume and elsewhere has shown positive contributions
to organizational outcomes from managerial networking in the environment
of public organizations, while interactions between political principals and
top managers were negatively related to results. The work has also shown
that managerial quality positively influences performance, as does a key
aspect of personnel management: retaining experienced employees and
thereby maintaining stability in human resources. Our opening argument
generates the expectation that management capacity should also contribute.
These several aspects of management and their expected relationships,
therefore, yield the following (with several of our usual managerial functions
relabeled here for ease of interpretation in the current analysis, and with the
use of a simplified linear additive model for the other managerial functions
already explored):
Ot ¼ b1 M2 þ b2 Mu þ b3 Mq þ b4 Mp þ b5 Mc þ et ð6:4Þ
where
Ot is some measure of organizational outcome at time t,
M2 represents external networking efforts by managers,
Mu is a measure of managerial interactions upward with political
principals,
Mq is a measure of managerial quality,
Mp is personnel stability,
Mc represents management capacity,
ε is an error term, and
b1 to b5 are estimable parameters.
210 Nonlinearities in public management

Equation (6.4) is insufficient, for two reasons. First, it lacks a vector of

control variables that represent the environmental forces that must be taken
into account in explaining outcomes. Adding such a vector, represented by
Xt to designate a set of such forces as they operate at time t, yields the
following:
Ot ¼ b1 M2 þ b2 Mu þ b3 Mq þ b4 Mp þ b5 Mc þ b6 Xt þ et ð6:5Þ
Second, our argument is not that managerial capacity adds in a linear
manner to the performance of an organization; other studies have demon-
strated that it does in some cases (Meier, O’Toole, and Hicklin 2010; O’Toole
and Meier 2010). Rather, we are interested in whether managerial capacity
can be the gift that managers bear – that is, the resources that make
interactions with environmental actors more successful. The question, there-
fore, requires estimating the specification in Equation (6.6) that includes an
interaction between managerial networking (M2) and managerial capacity
(Mc) as indicated by the new multiplicative term in the model:
Ot ¼ b1 M2 þ b2 Mu þ b3 Mq þ b4 Mp þ b5 Mc þ b6 Xt þ b7 M2 Mc þ et ð6:6Þ
If the inclusion of this new term adds additional explanation to our models,
then we can conclude that managerial networking interacts with managerial
capacity. More precisely, we are expecting the slope coefficient for the
interaction (b7) to be positive.

Data and measurement

For the empirical portion of this study, the units of analysis are all Texas
school districts. In this analysis we include eight years of performance data
(1995 to 2002) and supplement these with data from an original manage-
ment survey. The survey had a 55 percent response rate, and when combined
with the archival data produces 4,114 total cases for analysis. Missing data
on one of the performance indicators reduces the total number of cases to
3,798 in that case.20
A pooled time-series analysis such as this one needs to be concerned with
violations in the assumptions of multiple regression, particularly serial cor-
relation and heteroskedasticity. Assessments of the equations showed serial
correlation, so a set of dummy variables for individual years was included in
the analysis as an adjustment. Pooled diagnostics for heteroskedasticity
211 Data and measurement

showed only modest evidence of heteroskedasticity; estimation with robust

standard errors generated results similar to those presented here.

“M”: the management variables

Managerial networking. This variable (M2) seeks to measure the reported
behavior of school district top managers as they interact with the important
parties in the district’s environment. Our measure, introduced in Chapter 3,
is a factor score that indicates how frequently superintendents interact with
four key environmental actors.
Management capacity (Mc) is operationalized as the percentage of
employees who are located in central office administration (as opposed to
school administration). The mean percentage of central office administra-
tors for all districts in this study is only 1.89 percent, with a standard
deviation of 1.43.
Managerial quality (Mq) is a validated measure based on the residual from
a model explaining salaries of district superintendents. We replicated that
analysis for the years 2000 to 2002 and created a measure for 1995 to 2002.
Managing upward is measured as the frequency of interaction with school
board members, on a scale ranging from daily to never.
Personnel stability. Two aspects of personnel stability are included.
Managerial stability seeks to measure constancy in top leadership; it is
simply the number of years the superintendent has been employed by the
district in any capacity. Workforce stability is the percentage of teachers
employed by the district during the preceding year who continue to work
for the district.

“O”: outcome measures

This study incorporates three different performance indicators in an effort
to determine how public management affects a variety of organizational
outcomes. The performance indicators are selected to illustrate a range
of functions that the school districts perform: the overall student pass rate
on the Texas Assessment of Academic Skills, the percentage of students
who score above 1,110 on the SAT (or its ACT equivalent), and attendance
rates.
We also include the three constraint variables and the five resource
variables that are contained in previous models.
212 Nonlinearities in public management

Table 6.8 How the interaction of managerial networking and managerial capacity affects
organizational performance

Dependent variable ¼ student exam pass rates

Independent variables Slope T-score Slope T-score

Managerial networking 1.1425 9.25 0.9045 4.62

Managerial capacity 0.0075 0.08 0.0204 0.22
Networking  capacity 0.1294 1.57
School board contact 0.7894 5.76 0.7897 5.76
Management quality 0.7223 5.88 0.7185 5.85
Management experience 0.0573 4.64 0.0571 4.63
Personnel stability 0.1548 9.04 0.1551 9.06

R2 0.64 0.64
Standard error 7.13 7.12
F 341.43 326.14
N 4,114 4,114

Notes: Equations also control for teacher salaries, class size, teacher experience, noncertified
teachers, percentage state aid, percentage of Latino students, percentage of black students,
percentage of low-income students as well as dummy variables for individual years. Time
frame 1995–2002.

Findings

As noted above, the statewide standardized test, TAAS, is considered the

education system’s primary indicator, and it is used to evaluate the per-
formance of both the superintendents and the school districts in Texas.
Table 6.8 provides two regressions for this dependent variable. On the left
side is the regression without the interaction term, and on the right the
interaction between management capacity and managerial networking is
included.
First, it is apparent that, leaving aside the influences of networking
and managerial capacity, the other management measures included in
both specifications have their expected impacts; all these additional
slopes are statistically significant. We omit detailed discussion of these
in order to focus attention on the primary relationships under consider-
ation here.21
With regard to the results reported in Table 6.8, note that on the left, by
itself, managerial capacity has no impact on student TAAS performance.
213 Findings

In interpreting the interactive results on the right, care must be taken,

because the multiplicative nature of the interaction often creates relatively
artificial T-scores (they compare to a zero point that may be well outside
the range of the data). As a result, the equation has to be rearranged
to calculate the slopes if the contingent relationship is to be illustrated.
In essence this process shows that the impact of managerial networking
will vary depending on the value of managerial capacity. It is also neces-
sary to recalculate the standard errors to determine if the impact of
the interactive relationship is statistically significant at these various
levels.22
To determine the impact of managerial networking, the zero-order coeffi-
cient (0.905) and the interacted coefficient (0.129) both have to be taken, as
follows:

Networking slope ¼ 0:905 þ 0:129 ðcapacity Þ

First we can illustrate the problem with direct interpretation of the unin-
teracted slopes in an equation such as this. The formula above says that,
when capacity is zero, the slope of networking’s impact on TAAS perform-
ance is 0.905, or a one standard deviation change (M2 is a factor score) in
networking is associated with an increase of 0.905 percentage points on the
TAAS, all other things being equal. This is the equivalent of the zero-order
slope listed in the table; but, since there are no school districts with no
central office administrators, this part of the relationship is well beyond the
range of the data. The best way to interpret the interactive slope is to include
representative values of capacity and then calculate the slope of networking
at those levels of capacity. We do such a calculation using the mean value
of capacity, as well as values one standard deviation above the mean and
one standard deviation below the mean. These calculations produce the
following results.

Level of capacity Slope for networking

0.46 0.964*
1.89 1.149*
3.32 1.333*

* ¼ significantly different from zero, p < 0.05.

At low levels of managerial capacity (0.46 percent of central office staff),

therefore, a one standard deviation increase in networking is associated
214 Nonlinearities in public management

with an increase of 0.964 points on the TAAS, but at high levels of

capacity (3.32 percent of central office staff) a similar increase in net-
working is associated with an increase of 1.333 points – or about a 38
percent increase in relative impact. Although a 1.33 percentage point
increase might not seem like a great deal, this result is for a one standard
deviation change in networking. Since this is a factor measure with an
effective range from 3 to þ3, the total impact at this level of managerial
capacity could be as large as eight percentage points – a substantively
important amount.
Although our concern is not how networking influences the impact of
managerial capacity on performance, the equations show some interesting
results in this regard. A similar assessment for managerial capacity, which is
not statistically significant on the left-hand side regression, shows the
following.

Level of networking Slope for capacity

1.0 0.109
0.0 0.149
þ1.0 0.278*

* ¼ significantly different from zero, p < 0.05.

Capacity has no impact on TAAS performance except at high levels of

networking, where its impact is statistically significant but not especially
large. The networking measure ranges as high as þ3, however, so this impact
for a small number of districts will be considerably larger. The overall
conclusion from Table 6.8 is that managerial capacity enhances the positive
impact that networking has on performance; and, at high levels of network-
ing, managerial capacity also has a positive impact on TAAS performance, all
other things being equal.
Student attendance is crucial to school districts but is not easy to
influence, given that many student absences are the result of illness
or other factors outside the control of the student or the school district.
At the same time, school districts spend substantial time trying to influ-
ence the portion of attendance that is under the control of the student.
Table 6.9 uses the same format as Table 6.8 to show the interaction of
managerial capacity and networking and their impact on student attend-
ance. It is noteworthy that, on the left hand equation, both managerial
networking and managerial capacity have significant, positive impacts on
performance.
215 Findings

Table 6.9 How the interaction of managerial networking and managerial capacity affects
organizational performance: attendance

Dependent variable ¼ student attendance rates

Independent variables Slope T-score Slope T-score

Managerial networking 0.0730 5.67 0.0380 1.86

Managerial capacity 0.0975 10.12 0.0994 10.28
Networking  capacity 0.0190 2.21
School board contact 0.1281 8.93 0.1282 8.97
Management quality 0.0506 3.95 0.0500 3.90
Management experience 0.0012 0.93 0.0012 0.95
Personnel stability 0.0126 7.05 0.0126 7.07

R2 0.26 0.26
Standard error 0.74 0.74
F 69.58 66.70
N 4,114 4,114

Notes: Equations also control for teacher salaries, class size, teacher experience, noncertified
teachers, percentage state aid, percentage of Latino students, percentage of black students,
percentage of low-income students as well as dummy variables for individual years.

Calculating the slopes for managerial networking at different levels of

managerial capacity reveals the following impacts.

Level of capacity Slope for networking

0.46 0.047*
1.89 0.074*
3.32 0.101*

* ¼ significantly different from zero, p < 0.05.

The impact of managerial networking essentially doubles as the level of

management capacity increases from one standard deviation below the
mean to one standard deviation above the mean. This finding makes a great
deal of intuitive sense. While networking can create pressures or generate
ideas for how to deal with truancy problems, actually dealing with the
problems requires some capacity to act, and that capacity is almost always
lodged at the central office level.23
Performing the same set of calculations for managerial capacity at differ-
ent levels of managerial networking shows the following:
216 Nonlinearities in public management

Level of networking Slope for capacity

1.0 0.080*
0.0 0.099*
þ1.0 0.108*

* ¼ significantly different from zero, p < 0.05.

Table 6.10 How the interaction of managerial networking and managerial capacity affects
organizational performance: college readiness

Dependent variable ¼ percentage scoring above 1,110 on SAT

Independent variables Slope T-score Slope T-score

Managerial networking 0.7471 4.33 0.2196 0.73

Managerial capacity 0.7589 3.98 0.8114 4.22
Networking  capacity 0.3166 2.12
School board contact 0.2146 1.12 0.2216 1.16
Management quality 0.6733 3.88 0.6649 3.83
Management experience 0.0182 1.07 0.0184 1.08
Personnel stability 0.0783 2.89 0.0766 2.83

R2 0.32 0.32
Standard error 9.60 9.60
F 84.36 80.81
N 3,798 3,798

Notes: Equations also control for teacher salaries, class size, teacher experience, noncertified
teachers, percentage state aid, percentage of Latino students, percentage of black students,
percentage of low-income students as well as dummy variables for individual years.

Again, the interaction shows that the impact of managerial capacity on absentee-
ism increases when the superintendent engages in more effort to manage exter-
nally in the network. The relative increase from low to high (35 percent) is not as
large as it is for the networking increase contingent on capacity (118 percent), but
the impacts are strong and statistically significant at all levels.
The college board scores variable (see Table 6.10) has a different relationship
for managerial networking and performance in the noninteracted equation.
While managerial networking is strongly related to performance in a positive
direction, the relationship for managerial capacity is negative and significant.
Our experience in interviewing superintendents suggests that the negative
relationship for capacity, in part, reflects tradeoffs that superintendents have
to make. Resources committed to central office administration are resources
that cannot be committed to other programs, particularly to programs that
217 Implications

enhance the performance of the best students. Although the choice might not
always be so stark as whether to hire an advanced placement calculus teacher or
an attendance officer instead, once such hires are made they create a path
dependence in the short term, since the individuals are not fungible.
The interactive relationship shows that increases in managerial capacity
have a dramatic impact on the influence that networking has on the college
boards indicator, specifically the following.

Level of capacity Slope for networking

0.46 0.366*
1.89 0.819*
3.32 1.272*

* ¼ significantly different from zero, p < 0.05.

The impact of managerial networking at high levels of managerial capacity is

more than three times that at lower levels of capacity. The effect sizes here
are large, as the mean value of the dependent variable is 18.8 – which means
that an increase of 1.272 percentage points is a 6.8 percent increase in the
proportion of students meeting this criterion (over a range of 3 to þ3, this
impact is clearly substantial for this subset of districts).
The results for managerial capacity, which has a negative relationship with
college-bound performance, are somewhat different, but follow the same pattern.

Level of networking Slope for capacity

1.0 1.128*
0.0 0.811*
þ1.0 0.494*

* ¼ significantly different from zero, p < 0.05.

Managerial capacity’s negative impact drops by approximately 56 percent as the
level of managerial networking is increased from one standard deviation below
the mean to one standard deviation above the mean. The relationship remains
statistically significant at this level, but becomes statistically insignificant when
the level of networking approaches two standard deviations above the mean.

Implications

Our theoretical discussion argued that managerial networking and manage-

ment capacity should be expected to interact, with networking having
218 Nonlinearities in public management

enhanced positive impacts on organizational performance at greater levels of

capacity. Further, we expected to see such impacts across organizational
outcomes: for any given level of networking activity, enhanced capacity
should allow managers to leverage more organizational benefit from external
interactions. Identifying interaction partners and types of interactions,
engaging productively with promising external opportunities, negotiating
joint agreements, and implementing the results of any bargains reached
should all be more effectively accomplished with enhanced central capacity.
The findings confirm these expectations. For arguably the most important
measure of performance, pass rates on the statewide examination, the
impact of managerial networking increases as capacity increases 38 percent
across the range examined. The interactive effects for the other dependent
variables tested are even greater. The networking slope more than doubles at
higher capacity for school attendance – a result that fits with how we might
expect networking to contribute to this outcome. Interestingly, while cap-
acity itself negatively influences college-bound performance at many levels
of networking – a pattern probably reflecting the opportunity costs of
capacity for this outcome – greater capacity dramatically boosts the contri-
bution of networking to college-bound performance.

Nonlinearity and buffering: applying Occam’s razor

The previous three sections of this chapter examined nonlinearities in public

management with a focus on managerial capacity. Capacity at high enough
levels was able to mitigate the deleterious impacts of budget cuts and natural
disasters. Capacity also interacted with managerial networking in nonlinear
ways to influence performance. These findings raise the question as to
whether there might be a generic process by which capacity protects the
organization from turbulent aspects of the environment that might harm the
organization. To address this broader question, we return to our general
model of management and develop a measure of buffering. We then seek to
determine if the nonlinear predictions of our theoretical model hold in the
real world.
From the extensive empirical work documenting cases of public organiza-
tions operating in complex and turbulent environments – settings in which
the organizations are charged with carrying out policy objectives as they face
interdependence with other actors, including organizational and political
ones – it seems clear that two broad classes of forces can contribute to
219 Nonlinearity and buffering: applying Occam’s razor

protecting, insulating, or mitigating impacts on the organization from the

external environment: structural or procedural elements that help secure
an administrative system, on the one hand, and managerial contributions
to protection, on the other (see O’Toole and Meier 2003a: 112).24 As a
shorthand, we refer to any of these influences that reduce the impacts of
environmental forces on organizational or performance results as buffers,
and we refer to the dynamic of reducing such influences as buffering.
Whether structural or managerial (or both), several types of buffering
functions can be envisioned. Various particular forms of adaptive response
have been shown to be used in organizations to protect core tasks from
environmental perturbations, including the development of organizational
subunits to handle any uncertainty and instability generated externally (see,
for instance, Fennell and Alexander 1987). Indeed, even the development of
interorganizational linkages can be a means of buffering core organizational
activities (Miner, Amburgey, and Stearns 1990).
If the variety of buffering forms are framed at an abstract level, the
protective mechanism or effort could serve as a blockade insulating the
administrative system from external shocks up to, but not past, a particular
size (much like levees around New Orleans), or a selective filter allowing
some but not all external influences through the apparatus (a legislative
affairs office for contact with public officials), or a dampener reducing the
amplitude of any external influences (O’Toole and Meier 2003a: 113–14;
see also Lynn 2005: 38–9). Modeling the impacts of such different forms
of buffering would mean taking into account several rather distinct forms
of insulation, as explained in the discussion of buffering in Chapter 2,
each with its own somewhat different effect on the administrative systems
in question. Indeed, some public organizations might employ simultan-
eously several different kinds of buffering devices or effort. Investigating
all such buffers would be a useful task, but modeling and estimating
the impacts of buffering across many such organizations requires some
simplification.
We return to our base model presented in Chapter 2:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 ðXt =Se ÞðM3 =M4 Þ þ et ð6:7Þ
where
O is some measure of outcome,
S is a measure of stability, denoting structural, procedural, and other
elements that support unperturbed production,
M denotes management, which can be divided into three parts:
220 Nonlinearities in public management

M1 – management’s contribution to organizational stability through

additions to hierarchy/structure as well as regular operations,
M3 – management’s efforts to exploit the environment,
M4 – management’s effort to buffer environmental shocks,
X is a vector of environmental forces,
ε is an error term,
the other subscripts denote time periods, and
b1 and b2 are estimable parameters.
Where does buffering, as we have defined it, appear in the model? Our
conception of buffering is similar to Lynn’s (2005: 45) idea of moving buffers
out into the interface with the environment. Hence, buffering is present as
the denominator of the second, or environmental, term:
b2 ðXt =Se ÞðM3 =M4 Þ ð6:8Þ
or, after rearranging,
b2 ðXt M3 Þ=ðSe M4 Þ ð6:9Þ
This term models the impact of the set of environmental forces Xt on
outcome Ot. The impact can be leveraged by managerial effort (M3) or
buffered by the combined impacts of stabilizing forces (Se), such as structure,
as well as managerial influences aimed at protecting the production system
(M4). Note that this model simplifies by treating the buffering function in
mathematical terms solely as a dampener. This reciprocal function (1/Se M4)
essentially reduces the size of the impact that an X or environmental variable
can have (hence the division into X), by dampening the impact over time.
Eventually, other functional forms can be formally specified and tested, as
we draw once more from the extensive case study literature, but this
particular version is a useful first step. It is the “Se M4” denominator as a
whole, then, that serves as the model’s term for buffering.25
In Chapter 3 we noted that the two terms related to managing the
environment could be combined where M2 ¼ M3/M4.26 Thus M2 incorpor-
ates all efforts to manage externally in the environment, in contrast to
managing the organization, M1:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 ðXt =Se ÞðM2 Þ þ et ð6:10Þ
Although much of our preceding work has used this combined M2 term in
a series of analyses that demonstrate that managerial networking can be
effective and that its impacts can be nonlinear, as the model indicates,
they have not addressed four fundamental elements of the model. We
221 Measuring buffering

have not proposed and validated a measure of M4,27 the efforts of

managers to buffer the environment; we have not incorporated the struc-
tural elements of stability; we have not addressed whether the M2 measure
contains the M3 and M4 elements or whether there might be other
elements of M4 that have yet to be measured; and we have not addressed
the hypothetical functional form specified in their model – that is, the
reciprocal function.
This section addresses three of these questions. First, we propose a
measure of environmental buffering that encompasses both structural and
managerial activities. Second, we incorporate this measure into an explan-
ation of organizational performance that includes several other management
and stabilizing factors as well as measures of resources and constraints.
Third, we then brave Occam’s razor28 to determine if the complex nonlinear
relationship specified by theory is the most appropriate functional form, or
if a simpler estimation will provide equally good results. With regard to the
fourth question, we make the assumption that, because previous measures of
M2 did not specifically address the buffering function, prior analysis does
not explicitly include buffering actions. The model can thus be expanded to
contain both M2, composed of actions seeking environmental opportunities,
and M4, an explicit buffering function.

Measuring buffering

An organization’s effort to buffer environmental influences is likely to be

accomplished through a complex combination of structures (along with
associated procedures)29 and managerial actions (Meyer and Rowan 1977;
Miner, Amburgey, and Stearns 1990: 690; Sorenson 2003). Faced with a
turbulent environment (for a classic treatment, see Lawrence and Lorsch
1967) or one that is relatively stable but hostile, organizational leadership
has two options. First, decision makers could establish structures (Se) that
interact with the environment to absorb the environmental pulses, and in
the process shelter the organization’s core technologies (Thompson 1967).
In business firms, these structural elements would include organizational
units that deal with inventory control for inputs or post-production
marketing and distribution of outputs. In a public organization, such
structural features might include special legislative affairs or public affairs
units designed to handle requests from outside the organization, or an
222 Nonlinearities in public management

emergency response unit such as a SWAT (special weapons and tactics) team
or Delta Force. For school districts, the organizations examined here, the
response to the children of migrant workers provides an excellent illustra-
tion of buffering. The Houston Independent School District begins the
school year with about 176,000 students, with enrollment then peaking at
210,000 in November. To buffer its schools from this turbulence, HISD has
created student intake units to conduct immediate assessments of the
students (language, grade level, special needs) and assign them to appropri-
ate classes. These bilingual assessment units operate as what Lynn (2005: 41)
terms “absorbed slack,” and thus process the environmental influx and allow
schools to continue to function in regular ways.30
Second, management itself could engage in a set of buffering activities.
Management might decide that certain environmental events or influ-
ences will be ignored, while others will require the intervention of top
management, and still others should be programmed for response by
specific units – say, an accounting office primed to handle external audits
or other challenges regarding expenditures. School systems sometimes
establish their own police forces to deal with security and crime problems
rather than rely on the double environmental shock of a crime and the
arrival of an outside police force. Such a unit allows management to set
procedures for dealing with minor crimes without involving local law
enforcement. This buffering component of management (which is termed
M4) might be an ancillary function of management’s effort to interact
with the environment (M2) – that is, while seeking external opportunities,
managers can also identify negative or threatening forces. We can consider
including both aspects of management, therefore, in the model eventually
specified here.
Because the buffering function itself is a combination of both structure
and management, we opt for a unified measure of overall buffering (in lieu
of trying to separate out the two processes), stipulate operational definitions
of each, and then simultaneously test these concepts and the various ways
they could combine. This simplification then allows as a first step a valid-
ation of the measure of buffering as dampening, and an effort to probe how
it affects performance. If this effort is successful, future research will permit
an examination of exactly how the buffering process is developed and
operated.
How might we get an organizational measure of buffering? If we start with
the basic principle that organizations are autoregressive systems – that what
they do today reflects what they did yesterday – then an examination of how
223 Measuring buffering

autoregressive processes respond to novel events should be useful. Consider

the following simple autoregressive system:
Ot ¼ b1 Ot1 þ b2 Xt þ et ð6:11Þ
where O is the organization’s outcome31 and X is some type of environ-
mental shock, whether positive or negative from the standpoint of
performance.
One unit of any “X” variable affects output by b2 in the current year;
but, because this output then becomes part of the production base of the
organization (Ot1), the impact of one unit of X in the second year
becomes b1b2. Subsequent years see additional impacts, with the size of
each year’s impact declining in what is termed a geometrically distributed
lag. Although a buffering process could operate either on the X term
through b2 or on the Ot1 term via b1, in this assessment our attention
focuses on the latter.
Let us assume that there are two organizations, one with an autoregres-
sive parameter of 0.9 and another with an autoregressive parameter of
0.7. Let us further assume the occurrence of some environmental disturb-
ance that has an impact of Y on the organization. The following illustrates
the impact that this Y-level disruption has on the two organizations in
future years.

Parameter Year 0 Year 1 Year 2 Year 3 Year 4 Year 5

0.9 Y 0.9Y 0.81Y 0.73Y 0.66Y 0.59Y

0.7 Y 0.7Y 0.49Y 0.34Y 0.24Y 0.17Y

Even five years later, all other thing being equal, the environmental disturb-
ance retains nearly 60 percent of its impact for the first organization but has
fallen all the way to a 17 percent impact in the second.32 In organization-
theoretical terms, we think of the first unit as tightly coupled internally; any
disturbance, however slight, will reverberate through the organization for a
substantial period of time (see Meyer and Rowan 1977 and Powell and
DiMaggio 1991).33 The second organization is more loosely coupled internally
(probably decentralized but with absorbed slack; see Lynn 2005: 41); events
dissipate more quickly over time. This is the pattern that one would expect to
see if the organization had established structures and used managerial pro-
cesses to buffer or reduce the impact of environmental events or forces over
time.34 Loose coupling permits an organization to make a mistake and correct
that mistake without endangering the organization (Lynn 2005: 49). In this
224 Nonlinearities in public management

way it lessens the threat of environmental shocks on the entire organization.

This view is consistent with that of Meyer and Rowan (1977: 341), who suggest
that organizations “tend to buffer their formal structures from uncertainties of
technical activities by becoming loosely coupled.”
How might we get an estimate of the impact of such buffering or loosely
coupled internal structure on program performance? A simple autoregres-
sive estimation will not work, because this coefficient essentially uses a panel
of this year’s and last year’s performance for many organizations, and it is
not possible, as a result, to get an organization-specific estimate with only
two points. Because we have several years of prior data, our solution is to use
the historical data on each of the organizations to establish a baseline
measure of buffering in a subsequent period. Since our analysis here investi-
gates the years 1995 to 2002, we use data from 1986 to 1994 for each
organization to estimate this buffering process (incorporating therefore both
structural and managerial aspects). This approach allows us to have an a
priori estimate of buffering that is independent of the actual data used in the
study. The specific buffering measure correlates output (TAAS scores) at
time t with output from time tþ1 for the 1986–1994 period.35 This correl-
ation coefficient is transformed into a buffering measure by subtracting it
from 1.0 so that larger numbers indicate greater levels of buffering (looser
coupling) and lower levels indicate less buffering.
Three aspects of this measure merit discussion. First, the measure is
essentially a system component rather than an event component. It seeks
to assess the organization’s response to the environment in general rather
than in relation to any one specific event or type of external event. It is quite
likely that certain events might generate much greater effort at buffering,
depending on their salience and centrality for the organization in question.
At the same time, a systematic buffering element is likely to play a role even
in these one-time unique occasions, because systemic buffering obviously
provides some of the experience and capacity to deal with the more idiosyn-
cratic incidents.
Second, the measure is focused on outcomes. We are interested in explain-
ing the policy performance of systems rather than, for instance, their
internal operations. For that reason, the buffering measure directly
taps how an administrative system is or is not protected from having its
production – in terms of results – shaped over time by reverberations from
earlier events.
Third, the measure opts for parsimony in the composition of the meas-
urement – that is, it is a single measure generalized from past behavior and
225 Measuring buffering

does not attempt to separate out the individual influences of structure and
management. We use this strategy because we are interested in probing the
complex functional forms in our model of management.
Our strategy of analysis is to make some simplifying assumptions in the
model and proceed to test this notion of buffering as dampening, and its
various functional forms, in a step-by-step process. We start with the base
model in Equation (6.7) and regroup the elements in the second term of the
model to cluster stability and M4 together as follows:
Ot ¼ b1 ðS þ M1 ÞOt1 þ b2 ðXt M3 Þ=ðSe M4 Þ þ et ð6:12Þ
In this model, buffering generates a highly complex reciprocal relationship
that interacts with M3 and Xt (that is, Se M4 is divided into XtM3). We then
simplify this model by focusing solely on the second term of the model, thus
eliminating the autoregressive term as in Equation (6.13).36 The rationale is
that the current investigation focuses entirely on how environmental influ-
ences (the X vector) shape performance and may be mitigated by buffering.
These elements of the model all appear in the second term.
Ot ¼ b2 ðXt M3 Þ=ðSe M4 Þ þ et ð6:13Þ
Because (6.13) is a highly nonlinear form, we investigate whether the
relationships are actually this complex. To do so, we move to the simplest
possible option: a linear additive model for all variables. Equation (6.14)
displays this linear model, and also includes another term (M, operational-
ized below) to represent any other relevant management influences:
Ot ¼ b1 M þ b2 Xt þ b3 M3 þ b4 ðSe M4 Þ þ et ð6:14Þ
(6.14) becomes the basis for the first model that we test. The exact model to
be tested is based on measures developed earlier, particularly a measure of
M2 used in place of M3:
Ot ¼ b1 M þ b2 Xt þ b3 M2 þ b4 ðSe M4 Þ þ et ð6:15Þ
This model now contains a direct test of whether buffering, the last-listed
term in (6.15), contributes to performance. Then we add to this model in a
series of incremental steps to determine if more complex forms of the
relationships are warranted. First, we determine if including a reciprocal
relationship for buffering adds any additional information to the analysis, by
estimating Equation (6.16):
Ot ¼ b1 M þ b2 Xt þ b3 M2 þ b4 ðSe M4 Þ þ b5 ð1=Se M4 Þ þ et ð6:16Þ
226 Nonlinearities in public management

Because a linear relationship is simpler and more direct, our approach to

testing is to employ Occam’s razor as a selection criterion – that is, of two
competing explanations the simpler one is to be preferred unless the more
complex explanation adds significantly to our knowledge. From this per-
spective, therefore, both the linear and the nonlinear terms (b4 and b5)
should be included in the same equation. Because the theoretical model
contains not only a reciprocal function but also an interaction, we also test
Equation (6.17), which includes an interaction of buffering with managerial
networking (M2), and Equation (6.18), which interacts the reciprocal func-
tion with the environmental variables (X).
Ot ¼ b1 M þ b2 Xt þ b3 M2 þ b4 ðSe M4 Þ þ b5 ðM2 =Se M4 Þ þ et ð6:17Þ
Ot ¼ b1 M þ b2 Xt þ b3 M2 þ b4 ðSe M4 Þ þ b5 ðX=Se M4 Þ þ et ð6:18Þ

Data and measurement

We use eight years of data (1995 to 2002) from the Texas Education Agency
on organizational performance, resources, student composition, and other
relevant factors and results from our 2000 survey. Because this is a pooled
time-series analysis, we included dummy variables for the individual years to
deal with serial correlation. We then assessed the degree of heteroskedasticity
with pooled diagnostics and found the levels well within acceptable limits.
Our measure of buffering, as explicated in the preceding section, is unity
minus the correlation of school district outcomes for the period 1986 to
1994. Given that the primary statewide standardized test (the TAAS) is the
central outcome in the state’s performance appraisal system, we calculated
the buffering measure for that outcome and used it as a measure of buffering
in equations using TAAS or other outcomes as the dependent variable. The
final measure has a mean of 0.49 with a standard deviation of 0.20; it ranges
from 0.03 (a very tightly coupled system, with reverberations important over
time) to 0.999 (a school district well protected from the impacts over time of
environmental disturbances).

“O”: outcome measures

Buffering might well be an approach used to protect some goals and not
others. This study incorporates ten different performance indicators in an
effort to determine how public management affects a variety of organizational
227 Findings

outcomes: the overall student pass rate on the Texas Assessment of Academic
Skills, TAAS scores for Anglo, black, Latino and low-income students, the
percentage of students who took either of the college board exams, the average
ACTscore, the average SATscore, the percentage of students who scored above
1,110 on the SAT (or its ACT equivalent), and attendance rates.37

“M”: management variables

We take in a full set of managerial variables, including managerial network-
ing, managerial quality, managing upward, managerial stability, and work-
force stability.

“X”: environmental factors

We include the standard production function controls used in earlier
models: percentage of black, Latino, and low-income students, class size,
teacher salaries, state aid, noncertified teachers, and teacher experience.

Findings

The linear model estimates for the specification in (6.15) with the dependent
variable as the most prominent outcome measure – the overall TAAS pass
rate – are found in Table 6.11. The buffering variable is positively related to
overall performance, even controlling for a series of management variables
(M) as well as a set of variables covering a series of resources and constraints
(X). The inclusion of these controls supports the conclusion that buffering
itself matters rather than some combination of positive environmental
factors. Although the unstandardized coefficient looks large, the range of
the buffering variable is between zero and one, so this slope indicates a
maximum impact of 3.2 percentage points on the TAAS. In comparison, this
size represents approximately half the possible impact of that attributable to
the managerial networking variable. Although our concern is with the
buffering measure, the other relationships for the management variables
are consistent with earlier research: each variable is statistically significant,
and all of them except contact with the school board are positively associated
with performance.
Table 6.12 presents abbreviated information from nine additional regres-
sions, each representing the model specified in Equation (6.15) and one for
228 Nonlinearities in public management

Table 6.11 The impact of buffering on organizational performance

Dependent variable ¼ student exam pass rates

Independent variables Slope T-score p-value

Managerial buffering 3.2223 5.25 0.0001

Managerial networking 1.1810 9.58 0.0001
School board contact 0.7912 5.79 0.0001
Management quality 0.6939 5.73 0.0001
Management experience 0.0571 4.64 0.0001
Personnel stability 0.1532 8.98 0.0001

R2 0.64
Standard error 7.10
F 345.04
N 4,114

Notes: Equations also control for teacher salaries, class size, teacher experience, noncertified
teachers, percentage state aid, percentage of Latino students, percentage of black students,
percentage of low-income students as well as dummy variables for individual years.

Table 6.12 The impact of buffering on alternative indicators of performance

Performance measure Slope T-score R2 N

Black TAAS pass rate 13.0681 7.23* 0.48 2,503

Latino TAAS pass rate 2.9283 2.73* 0.45 3,745
Anglo TAAS pass rate 2.5750 4.26* 0.50 4,068
Low-income pass rate 4.8122 6.08* 0.58 4,087
Attendance 0.3847 5.94* 0.25 4,114
Percentage taking college boards 4.8069 3.38* 0.12 3,776
ACT scores 0.3265 2.60* 0.39 3,522
SAT scores 10.9686 1.54 0.48 2,902
Percentage college-ready 1.5563 1.72 0.32 3,798

Notes: All equations control for the five management variables, teacher salaries, per student
instructional funds, class size, teacher experience, percentage of teachers not certified,
percentage of black, Latino and low-income students and yearly dummy variables.
* ¼ significant at p < 0.05, two-tailed test.

each of the other dependent variables. The top four lines deal with the
impact on TAAS scores for various subsets of students. In each case, greater
buffering is associated with higher student scores. The pattern of these
coefficients is interesting. Without question, the greatest impact of buffering
is for African American students – an impact five times larger than the
229 Findings

relationship for Anglo students. Similarly, the impact for low-income stu-
dents is substantially larger than that for all students; the Latino impact,
while only marginally larger, is still greater than that for Anglos. This pattern
of results suggests that buffering is particularly valuable for the least advan-
taged of the organizations’ clientele – a finding that is especially interesting
given that managerial networking tends to be associated with gains for more
advantaged groups (see the analysis in Chapter 3).
Buffering also has additional impacts, as a glance at the other relation-
ships in Table 6.12 attests. Attendance is a basic minimum performance
indicator that is focused on disadvantaged rather than well-off students.
Although attendance does not vary a great deal and is difficult to affect,
buffering is positively related to attendance. We suspect that, if attendance
rates were analyzed by race, some larger relationships would appear for non-
Anglo groups. For the more elite measures, on the other hand, buffering
does not do as well. It is associated with more students taking college boards,
but it is also associated with lower scores on the ACT. Neither SAT scores nor
the college-ready percentage is significantly related to buffering. Again, these
relationships could be interpreted as consistent with the notion that buffer-
ing benefits the more disadvantaged clientele. Expanding the number of
students who take the college boards is a policy that benefits students who
would not otherwise go on to college.
Having established that buffering is associated with organizational perform-
ance, we turn to determining if the functional form specified by the theory is
correct. Table 6.13 adds the reciprocal of the buffering variable to the ten
equations represented in Tables 6.11 and 6.12, thus summarizing ten estima-
tions for Equation (6.16), one for each performance measure. If the reciprocal
functional form were to add explanatory power, over and above a linear
specification, we should find that the coefficients for this reciprocal variable
are statistically significant. In seven of the ten cases, the reciprocal relationship
fails to attain the 0.05 level of statistical significance. If the case of SAT scores
with the negative relationship is dismissed, there are only two cases for which
the nonlinear form appears to contribute: low-income pass rates and attend-
ance. In both cases the relationship is not strictly linear, but the increase in the
overall level of explained variation is minimal. The linear coefficient holds up
better than the nonlinear version. An overall conclusion, then, should be that –
in trimming with Occam’s razor – the relationship between buffering and
organizational performance is linear rather than nonlinear.38
Table 6.13 tests a simple reciprocal relationship, while the full theory
specifies that buffering as a reciprocal relationship interacts with managerial
230 Nonlinearities in public management

Table 6.13 Buffering: is the relationship linear or reciprocal?

Linear Reciprocal

Performance measure Slope T-score Slope T-score

Overall TAAS pass rate 2.2877 2.13* 0.1934 1.06

Black TAAS pass rate 13.3745 3.51* 0.0591 0.09
Latino TAAS pass rate 4.4454 2.00* 0.3307 0.78
Anglo TAAS pass rate 0.9695 0.92 0.3294 1.85
Low-income pass rate 7.1709 5.18* 0.4855 2.07*
Attendance 0.9084 8.03* 0.1084 5.64*
Percentage taking college boards 6.5834 2.27* 0.3805 0.70
ACT scores 0.3426 1.32 0.0034 0.07
SAT scores 52.1820 3.58* 8.1818 3.23*
Percentage college-ready 1.4133 0.76 0.0305 0.09

Notes: All equations control for the five management variables, teacher salaries, per student
instructional funds, class size, teacher experience, percentage of teachers not certified,
percentage of black, Latino and low-income students and yearly dummy variables.
* ¼ significant at p < 0.05, two-tailed test.

Table 6.14 Does the interaction with M2 add explanation to a linear model?

Performance measure Networking Buffering Ratio T-score

Overall TAAS pass rate 1.9468* 3.1950* 0.3199 2.88*

Black TAAS pass rate 0.0948 13.1852* 0.2337 0.78
Latino TAAS pass rate 1.7959* 2.9230* 0.5877 3.03*
Anglo TAAS pass rate 1.6782* 2.5435* 0.1876 1.73
Low-income pass rate 1.2212* 4.7917* 0.1714 1.20
Attendance 0.0908* 0.3841* 0.0067 0.57
Percentage taking college boards 0.3583 4.8161* 0.0971 0.39
ACT scores 0.1893* 0.3369* 0.0315 1.45
SAT scores 0.6654 10.4876 3.1806 2.69*
Percentage college-ready 0.3205 1.5781 0.1927 1.20

Notes: All equations control for the five management variables, teacher salaries, per
student instructional funds, class size, teacher experience, percentage of teachers not
certified, percentage of black, Latino and low-income students and yearly dummy variables.
* ¼ significant at p < 0.05, two-tailed test.

networking and/or environmental resources and constraints, as indicated in

Equations (6.17) and (6.18). Table 6.14 shows abridged results from ten
regressions that include an interaction term between the reciprocal and the
managerial networking variable. Because this equation also includes linear
231 Findings

Table 6.15 Does buffering interact with resources in a reciprocal manner?

Performance measure Resources Buffering X/Se M4 T-score

Overall TAAS pass rate 1.1810* 3.1165* 0.0003 0.13

Black TAAS pass rate 0.7132* 14.0069* 0.0023 0.29
Latino TAAS pass rate 0.3863 5.3821* 0.0069 1.30
Anglo TAAS pass rate 1.2284* 1.6955 0.0023 1.06
Low-income pass rate 0.8097* 7.5907* 0.0073 2.59*
Attendance 0.0749* 0.9529* 0.0015 6.46*
Percentage taking college boards 0.6006* 7.1153* 0.0063 0.95
ACT scores 0.1124* 0.4044 0.0002 0.35
SAT scores 7.2452* 50.9109* 0.1011 3.25*
Percentage college-ready 0.7871* 1.8649 0.0008 0.20

Notes: All equations control for the five management variables, teacher salaries, per student
instructional funds, class size, teacher experience, percentage of teachers not certified,
percentage of black, Latino and low-income students and yearly dummy variables.
* ¼ significant at p < 0.05, two-tailed test.

terms for both networking and buffering, these coefficients are reported in
the table as well, in case the interaction achieves statistical significance at the
expense of inducing collinearity in the other parameter estimates. In only
three cases – overall TAAS scores, the Latino TAAS rate, and SAT scores – is
the interaction term statistically significant. In all three cases, the sign of the
ratio interaction coefficient is the opposite of the two nonlinear terms (one
indicator of collinearity),39 the size of the coefficients is not large, and the
coefficients of determination increase only marginally (not shown).
A conservative interpretation of the results in Table 6.14 is that the relation-
ship between networking and buffering regarding performance is linear
rather than a more complex nonlinear interaction.
The other possible nonlinear interaction is with the “X” variables repre-
senting external resources and constraints (Equation (6.18)). Because there
are eight such X variables, and a set of eight interaction terms would
generate an excessive number of coefficients, we converted and simplified
the interaction term by first regressing performance on the eight environ-
mental factors and saving the predicted values. This step creates a vector that
contributes the full amount of explained variation to the dependent variable.
Then we ran ten regressions, one for each of the dependent variables, with
the complete set of independent and control variables and included an
interaction term between this new resources vector and the reciprocal of
the buffering measure. These results are reported in Table 6.15. Again, the
232 Nonlinearities in public management

findings are not impressive enough to conclude that the nonlinear inter-
action term is a superior specification to the linear additive specification, at
least for this set of data. Only three of the interaction terms meet the
minimum level of statistical significance, and only one of these, for attend-
ance, appears to add anything more than a minimum level of explained
variation.
By examining the functional form of the buffering relationships in com-
parison with a linear additive model, we find that the nonlinear estimations
produce some interesting individual results but are not demonstrably super-
ior to the linear additive ones. In such a circumstance, the principle of
Occam’s razor holds that the simpler linear models are to be preferred. This
finding does not mean that these relationships are linear or noninteractive in
all circumstances. For instance, in this analysis we have not probed subsets of
the sample to determine if high performers or low performers operate in a
manner different from the median organization (see Chapter 4). Nor have
we examined different or unusual combinations of the resources and con-
straints provided or imposed by the environment (turbulent or stable
environments) to see if selected configurations might produce different
results regarding the functional form among these variables. All the same,
the results here do indicate that the rather more complicated specification
entailed in the full model is not necessary in at least some important
empirical cases.

Implications
In this section, we have treated buffering as a combination of managerial
efforts and structural features that limit the impact of environmental forces
on public organizations’ performance over time. Rather than differentiating
the micro-details of managerial buffering activities, which could be quite
subtle and varied, or the structural forms that buffers might take, we have
developed a buffering measure that is designed to tap all these by encom-
passing the results of buffering over a several-year period prior to the period
under direct examination. Developing such a measure has been particularly
helpful, in that it does not rely on the details of structural form or
managerial action but, instead, gathers their accumulated influences into
a single measurable term – one that can be tapped for many organizations.
We have used the measure to estimate the impacts of buffering, and also to
explore some theoretical expectations about the functional form of its
influence.
233 Findings

The results provide not only some support for the model’s assertions
but also some Occam-guided negative findings. A key bottom line is that
buffering can help performance. In seven of the ten linear estimations, this
conclusion is supported. In only one of the ten equations does buffering have
a statistically significant impact in the direction of impairing performance.
This general support for the buffering function appears in a set of well-
specified models that contain five other management influences and eight
controls for resources and constraints. These control variables rule out many
alternative explanations for performance fluctuations, such as socioeconomic
differences among the districts or personnel and management activities. It
seems clear, therefore, that the protection of organizational production from
potentially disturbing influences can be an important function for those who
care about performance, even if “protection” sometimes comes at the price of
missed opportunities to tap new resources or acquire new support or juris-
diction. To put the point another way: public organizations can benefit both
from protecting their operations internally – for instance, by stabilizing their
personnel in front-line and managerial positions (note the full estimation
sketched in Table 6.11 as an example) – and also by insulating internal
operations from externally generated perturbations.
The results of these analyses offer more than a simple brief for the
advantages of buffering, however. For school districts in Texas, buffering
clearly helps improve performance as measured by criteria pertaining, or of
most interest, to disadvantaged students. It helps little or not at all for those
who are advantaged. This distributional dimension of the results is striking,
and highly interesting as well. In an analysis reported in Chapter 3, we
showed that managerial networking also carries distributional consequences,
but with a pattern opposite to that found here for buffering. Networking by
top managers in school districts benefits the most advantaged students and
does little for those at the other end.
The results of this study also raise questions about the functional form
through which buffering operates. The model tested here includes complex-
ity: a reciprocal for buffering, plus interactions with both managerial behav-
ior (networking) and external forces (resources and constraints). The
empirical results show only limited and sporadic support for these features
over the considerably more straightforward linear form, however. The latter
represents the notion that buffering serves as a simple input to production.
If a linear functional form can be regularly shown to be just as efficient in
explanation as one incorporating the nonlinearities and reciprocal, these
features of the model should be rejected.40
234 Nonlinearities in public management

Capacity: an English example

With one exception, neither management capacity nor buffering has been
used in other studies, so much evidence on the generality of the Texas
findings is not yet available. One study of management capacity and its
interactive effect does exist, however. Andrews et al. (2009) examine the
negative impact of unexpected immigration on the performance of local
governments in England using the national government’s evaluation score
for local governments. Immigration was a significant shock. The number of
immigrants who came to the United Kingdom with the expansion of the
European Union was seventeen times the anticipated number, and the influx
overwhelmed the services of some local governments. Andrews et al. create a
capacity measure based on central office bureaucracy similar to that for
Texas school districts, and find that local governments with greater manage-
ment capacity are able to mitigate this negative impact of immigration so
that the end result is no longer statistically significant.

Capacity, buffering, and performance: concluding thoughts

In earlier chapters we explored management’s performance-related effects as

managers work externally and internally, and we demonstrated that – apart
from sheer activity or effort – managerial quality also matters. In this
chapter we have extended the analysis of managerial and organizational
influences on policy outcomes. In doing so, we have paid sustained attention
to some so-called nonlinearities – in these cases, the possibility that man-
agerial capacity might interact with other features of the system to protect
against performance declines or enhance the value of another managerial
function. We have also explored the impact of organizational buffering on
performance.
In several analyses, we have shown that managerial capacity – reserves of
management or potential managerial effort – can provide performance
improvements. It can do so when unexpected and negative jolts hammer
the organizational system, as when budget cuts or natural disasters strike
with little warning. In such instances, capacity can partially or even
fully mitigate the otherwise expected dips in performance. Part of the value
of such capacity, therefore, is protective or defensive. Capacity can also
235 Notes

leverage additional results from already established and potentially valuable

managerial functions. In the case analyzed here, managerial networking’s
value to the organization and its clientele is enhanced as capacity is
expanded.
Managerial capacity is not costless; adding capacity means providing
funding for what sometimes may seem to be excess staff or forgoing funding
for alternative activities. Having some such capacity also produces the
performance-related benefits documented in this chapter, however. These
can be demonstrated most readily by exploring the nonlinear relationships
that have been the focus of particular attention here.
Our base model points toward an additional sort of nonlinearity, which
we have also examined in the chapter. The “SeM4” in the denominator of the
second, or environmental, term of the model represents a combination of
the structural and the managerial elements of organizational buffering. In
this chapter we have developed a measure for this aspect of the system and
analyzed its performance-related effects. Here we found, as extensive theor-
etical literature expects, that buffering aids performance. We found relatively
little evidence in our empirical data, however, that it does so in the nonlinear
way that the model stipulates. At this point, therefore, and without substan-
tial additional research, we cannot conclude that the model’s specification of
how the buffering function operates is correct.
What we can conclude, nonetheless, is that capacity and buffering provide
additional aids as public organizations and their managers seek to deliver
results. In yet further ways beyond those sketched in earlier chapters,
managerial and organizational features contribute to the production of
public services.

NOTES

1. The section draws from the presentation by O’Toole and Meier (2010).
2. Huge and continuing shocks, whether budgetary or otherwise, would take their toll,
theoretically, but there are no existing studies of this in public management literature.
3. Other measures of managerial capacity might consider education levels, skills sets of
managers, or even years of experience.
4. There were no statistically significant losses in any of the ten performance measures in the
third or fourth years.
5. In some preliminary analysis, we find that increased staffing of the central office leads to
some tradeoffs likely to benefit more disadvantaged students at the expense of students
236 Nonlinearities in public management

with greater advantages. The relative size of the central office staff is positively
associated with increased student attendance and negatively associated with drop-
out rates, and these are both desired outcomes. There is also a negative relation-
ship between central office staff and the percentage of teachers the district employs
in gifted programs. In essence, there appears to be a clear tradeoff: build manager-
ial capacity if the goal is to help the disadvantaged students or build specialized
teaching capacity if the goal is to improve the performance of high achievers.
6. This section draws from the presentation by Meier, O’Toole, and Hicklin (2010).
7. All data are from the official records of the Texas Education Agency.
8. Using the entire set of 1,043 districts, including the smaller ones, has only modest impact
on the results. In this case, the weaker of the two influences, the “Katrina” coefficient
(explained shortly), fails to attain statistical significance (the direction and final results
still hold). The results for the “Rita” coefficient remain unchanged.
9. The results using all the days were very similar but were affected more by collinearity, since
many of the Katrina evacuee districts were closed for the Rita evacuation but did not sustain
physical damage that would have prevented them from opening the following week.
10. Here we count as districts affected by one or both shocks any district with a non-zero
value for enrolling Katrina evacuee students or sustaining a period of at least six days of
closure due to Rita.
11. The hurricane-impacted districts are not appreciably different from those not affected.
Both the Katrina and the Rita districts actually had lower revenues per pupil but higher
teacher salaries (for Katrina) and larger class sizes. These differences are in part a result of
the hurricane-impacted districts being in the region of the state where wages are likely to
be higher.
12. The equations in all tables were also estimated using a three-year average of prior TAKS
scores rather than just the previous year. The three-year average tends to underestimate
the district’s record, because the TAKS was a new exam created in 2003, and there is a
positive trend in overall scores as districts and students got used to the new exam. The
results of this estimation were generally the same as those presented but showed a slightly
larger impact from the hurricane, because it underestimated the prior performance of the
district.
13. The hurricane-affected districts did not do better as the result of an influx of aid.
Instructional expenditures per student go up by $645 in districts impacted by Katrina
(versus $658 for all districts) and $533 in the Rita districts (versus $661 for all other
districts). The influx of additional state or federal funds was overwhelmed by the total
number of new students who arrived.
14. The impact of Katrina might have been larger, but the majority of evacuee students were
distributed in a ring of school districts in the Houston metropolitan area that are generally
fairly high-performing and are known for talented instruction. The districts with Katrina
students, for example, had a TAKS pass rate of 62.8 percent in 2005 compared to 61.1
percent for all other districts. They achieved this despite having lower revenues per pupil
($8,643 versus $8,945) and higher student to teacher ratios (14.3 versus 12.7).
15. The use of the TAKS exam as the most sensitive performance indicator is corroborated by
the results for the commended pass rate. The respective coefficients for the model in
column 3 for that equation were –0.184 for the Katrina students and –0.039 for the days
237 Notes

missed. Both were significant but only at the 0.10 level. The commended rate impact of
both shocks is about one-third the size of the impact on the overall TAKS pass rate. The
attendance equations show no impact for the missed class days, and a coefficient of –
0.034 for the Katrina students (significant at the 0.10 level). The attendance impacts are
tiny and thus need not be considered in detail.
16. Hurricanes aside, many of the districts regularly receive a large number of students after
the start of the school year because the families of migrant workers return to Texas after
working harvests further north. These evaluation systems are generally administered by
central management in order to avoid disrupting the schools.
17. The findings are not the result of districts with greater central office capacity simply being
wealthier districts. The correlation between central office bureaucracy and property tax
wealth per student is 0.03 (not significant). There is a positive correlation between
revenues per student (which include state and federal funds) and central office bureau-
cracy of 0.26, but for districts with 1,500 students or more this correlation is an insignifi-
cant 0.04. The measure is the percentage of employees in the central office bureaucracy,
which is likely why the measure is unrelated to district wealth.
18. We also examined how personnel stability affects the impact of the hurricanes, and found
relatively similar results (see Meier, O’Toole, and Hicklin 2010).
19. This section draws from the presentation by Meier and O’Toole (forthcoming (a)).
20. The TEA does not report data if fewer than five students are in the category. In some
smaller school districts this might mean that fewer than five students take the college
boards, and thus the percentage of students scoring above 1,110 on the SAT or its ACT
equivalent will be missing.
21. The same pattern holds generally for the estimations with attendance and high-end
performance, as reported in Tables 6.9 and 6.10. A few of the management coefficients
there are not significant.
22. Calculating the slopes is relatively simple and is illustrated in the text. Calculating
the standard errors for the marginal standard errors is more complicated and
requires using the variance–covariance matrix from the regression (see Brambor,
Clark, and Golder 2006). If the formula for the conditional slope is represented by

slope ¼ b1 þ b2Z
where Z is the interacted variable, then the standard error at any point in the regression
line can be calculated by taking the square root of [var(b1) þ Z2var(b2) þ 2Zcov(b1b2)].
23. Texas school districts generally require the schools to keep track of attendance and do the
routine follow-ups. More elaborate programs focused on truancy are generally housed in
the central office. One district, for example, works with local law enforcement agencies to
have police visit the homes of absent students to let the parents know of an unexcused
absence. This district also has a program through which, on a given day, law enforcement
officers routinely stop all school-age children visible in public places during school hours
to determine why they are not in school.
24. Of course, these two are related. Management can undertake actions that build or alter
structural features of the organizations, for instance.
25. As indicated below, we do not develop separate measures for each of the “S” and “M4”
terms of the model. The measure used here is one for the entire buffering term,
238 Nonlinearities in public management

encompassing both elements. The model’s incorporation of the buffering concept sug-
gests that its impacts on performance are likely to be similar across outcomes. It may
nonetheless be that some kinds of performance, affecting certain kinds of groups, are
more or less sensitive to the impacts of organizational buffering. This notion is not
included in the general theoretical argument being examined but is analyzed empirically.
26. This argument is theoretical rather than empirical – that is, we have not demonstrated
that the measures of these concepts actually combine in this way, nor have we demon-
strated that the measures cover the full extent of the two concepts (see below).
27. Nor have we proposed and validated a measure of M3, the efforts of managers to tap or
exploit the environment.
28. Entia non sunt multiplicanda praeter necessitatem, or “No more things should be
presumed to exist than are absolutely necessary.” The principle is attributed to William
of Ockham, a fourteenth-century philosopher and logician.
29. For the remainder of the chapter we refer to the combination simply as structure.
30. Such structures can buffer other events. When 5,000 students arrived at HISD in
September 2005 as the result of relocations from Hurricane Katrina, they were assessed
and assigned in the same manner.
31. The logic works whether the “O” variable is defined either as outputs or outcomes.
32. All things are not equal, of course, because new environmental events enter this system
every year. More recent events could swamp the later-year impacts of this initial event,
depending on their size.
33. Coupling could actually be considered on two dimensions: the extensiveness of the
interdependent links with the environment and reverberation through time. Here we
focus on the latter and assume the former as a constant. For the empirical part of the
analysis in this book, this assumption is reasonable, since the full sample of organizations
are structurally alike and similarly specialized.
34. There can be both internal and external drivers of the autoregressive parameter. We
assume that the main source of influence over variations in the parameter, among
organizations of similar structure and function, is the set of externally generated influ-
ences – the “Xs” of the model – during the preceding cycles.
35. Essentially, we estimate over 1,000 regression equations, one for each school district, to
produce these estimates.
36. Some of our management and structural variables contain an element of this internal
dimension. The choice is between leaving out management elements that are more
general or including them even though they might operate within the organization.
37. We considered using dropouts as a performance measure, but the state of Texas made one
significant change in the dropout measure during this time period so that early dropout
rates are not comparable to later ones. Accordingly, we omitted this measure from the
analysis.
38. We also experimented with a quadratic estimation, and got similar results.
39. As an illustration, Table 3.12 shows that networking is positively and significantly related
to SAT scores while Table 6.14 produces a negative and insignificant coefficient.
40. This section of the chapter has offered such tests of certain portions of that model and
has raised questions regarding functional form. We are a long way from clearly rejecting –
or accepting – such theoretical expectations in toto, however. For one thing, Chapter 3
239 Notes

has shown evidence of other nonlinearities supportive of certain complexities in

functional form – interactions between managerial networking and some external
influences, to be precise. A prudent additional step, therefore, would be to concentrate
on additional analysis of selected resources in the environment, those showing nonlinear
relationships with managerial networking, to see if these also interact with buffering.
For another, as suggested earlier, different portions of the sample may behave differently
(high or low performers, for instance), and interactions can be probed systematically in
this regard as well.
7 Public management in intergovernmental
networks: matching structural networks
and managerial networking

In this book on public management and performance, we began by

noting the importance of networked relationships among organiza-
tions as a key part of contemporary public managers’ institutional set-
tings. We also explored at length the question of how what managers
actually do shapes the results of public programs. We have analyzed
managerial networking behavior, among other aspects of management,
but thus far we have not assessed how networked structures themselves –
as distinct from the networking behavior of managers – shape policy
outputs and outcomes. To put it in terms of the elements of our model,
we have unpacked several elements of “M” but largely ignored the
structural aspects of the core organization’s environment – an aspect of
the “X” term.1
The reasons for this focus were explained in Chapter 2, and it should be
clear from the extensive reliance of our empirical exploration on the Texas
school district data set that researchers inevitably face real limitations in fully
exploring structural influences with such a sample. School districts intern-
ally look much like each other in structural terms, so the range of variation is
quite limited. Moreover, sketching the full networks in which they may
operate would require labor-intensive data gathering among many actors
in each of the more than 1,000 districts in the sample. The task is not
practical without large amounts of time and resources for extensive
fieldwork.
Nonetheless, with data now available it is possible to begin an effort to
integrate the structural aspects of such networks with the behavioral net-
working efforts of top managers. We can do so because the fiscal ties
between the districts and other levels of government are themselves struc-
tural, and data are available that allow us to establish whether these relatively
stable intergovernmental relationships – these funding-based network
links – affect performance in a manner that is distinct from the networking
behavior – the M2 – of the managers.
240
241 Public management and the intergovernmental landscape

This chapter offers such an examination. While it certainly does not cover
the full terrain of how structural networks can shape results, it does show
some of the complex causal patterns that are part of today’s managerial
setting.

Public management and the intergovernmental landscape

In the United States, intergovernmental programs have become more the

rule than the exception. Most “national” public initiatives are implemented
with active involvement by states and/or localities, and a large portion of
US-based state programs – varying by state, to be sure – are in reality state-
local efforts. The principle of federalism and the political realities of
program adoption and execution provide a heavy tilt toward collaborative
intergovernmental arrangements under most circumstances (O’Toole
2000a). Of course, the significance of intergovernmental programs is not
limited to the United States, and many other nations also confront the
challenges and opportunities of managing public purposes across govern-
mental lines. In particular, with the gradual emergence of the European
Union as an important contributor to governance in multilevel systems, the
theme of public management in intergovernmental settings is acquiring
renewed importance.2
The crucial role of public management in such programs has been
recognized by specialists in intergovernmental relations, who have empha-
sized the rise of “intergovernmental management” as the core of intergov-
ernmental relations more generally (Wright 1990). This is not to say that
management is everything in such programs. The politics of intergovern-
mental relations matter greatly, and the details of program design can be
critical. With due caution regarding the hazards of overgeneralization, the
management of intergovernmental programs should be of central interest
to scholars of governance in the twenty-first century.
Although the theme of intergovernmental management has been empha-
sized as a key part of the contemporary policy environment, theoretical work
on this topic has been slow to develop. The specialty is descriptively rich,
with many studies of the efforts and intricacies of what managers do when
they work across governmental boundaries and how important the details
can be (see, for instance, Agranoff 1986, Gage and Mandell 1990, and
Agranoff and McGuire 2003). The field has not managed to integrate these
insights into a theoretical perspective that can offer testable generalizations,
242 Public management in intergovernmental networks

however, or links to allied theoretical developments on the burgeoning

subject of governance (Lynn, Heinrich, and Hill 2001). A review of the
subject of intergovernmental relations some time ago remains valid on this
point: Beam, Conlan, and Walker (1983) point to the descriptive strengths
but theoretical weaknesses of this line of work (see also O’Toole 1990). In
this chapter, we use our work on management and performance to begin a
systematic examination of the role of public management in executing
intergovernmental public programs.

Intergovernmental ties as networked relations

As we have made clear earlier in this book, an increasing body of scholarship

argues, and in some cases demonstrates, that public management often takes
place in and on networks of actors rather than solely within the confines of
a single, hierarchical public bureaucracy framed in a dirigiste state (Hufen
and Ringeling 1990; Scharpf 1993; Bressers, O’Toole, and Richardson 1995;
Klijn 1996; Bogason and Toonen 1998; O’Toole 1998; Peterson and O’Toole
2001). In networked settings, program success requires collaboration and
coordination with other parties over whom managers exercise little formal
control. Many of these complex arrangements are required or strongly
encouraged by policy makers, others emerge through mutual agreement
among organizations or other partners who find that mutual interests are
served by working together on a regular basis.
Networked arrays may include some combination of: agencies (or parts of
agencies) of the same government; links among units of different govern-
ments; ties between public organizations and for-profit companies; public–
nonprofit connections; and more complex arrangements, including multiple
types of connections in a larger pattern. Networks can range in complexity
from simple dyads, at one pole, to bewilderingly complex arrays entailing
dozens of units, at the other (see, for instance, Provan and Milward 1991).
Beyond the issue of size or complexity, networks can vary greatly on many
other dimensions. Clearly, examining “networks” and public management in
any comprehensive sense requires investigating a great number of research
questions, network dimensions, and levels of analysis (for a preliminary
sketch of a research agenda, see O’Toole 1997).
One aspect of networks and public management worthy of close examin-
ation is the intergovernmental component. Several reasons justify such a
focus. Intergovernmental ties are frequent, and intergovernmental links are
243 Intergovernmental networks and public education

among the most common networked connections during the execution of

public programs (Hall and O’Toole 2000). Second, intergovernmental com-
ponents of networks per se have generally not been a focus of empirical
studies of networks and public management, so it is worthwhile to explore
whether this aspect of networked public management offers particular
insights. Third, given the generally atheoretical nature of much work con-
ducted on intergovernmental management, an effort to consider the topic
through the lens of recent network studies offers the prospect of some
much-needed theoretical leverage.
Intergovernmental ties involving public organizations could take one or
more of several forms. Service agreements in intergovernmental regulations,
for instance, are reasonably common. Perhaps the most frequently occur-
ring, and clearly among the most important, are regular ties triggered and
sustained by intergovernmental aid, particularly grants-in-aid. Intergovern-
mental grant programs involve one or more “donor” government(s), and
governmental agencies, in regular interaction with one or more “recipient”
government(s) and agencies (Pressman 1975). Donor units offer incentives
for recipients to undertake certain initiatives with certain emphases, and
typically attach some regulatory “strings” as a condition of aid. Recipient
units may have their effective program costs trimmed substantially by grants,
even though entering into the bargain means dealing with the preferences of
the donor and the conditions of support, usually on a regular basis. Such
programs create dependences, and also increased probabilities and scope for
public action. Over time, the intricacies of program management require
diplomatic, fiscal, and other forms of managerial skill and effort (for a
classic depiction of the subtle forms of intergovernmental influence in grant
programs over time, see Derthick 1970). In short, examining networked
patterns of intergovernmental aid provides an opportunity to begin an
exploration of the intergovernmental dimensions of public management.

Intergovernmental networks and public education

We explore intergovernmental management here in the field of public educa-

tion. Public education is not among the most highly networked public service
production and delivery sectors (Tyack 1974). Nevertheless, this policy arena
has developed into a significantly more complex and interdependent setting.
Schools are now venues for the delivery of a host of associated services
or regulatory programs, from public health (vaccination programs, the
244 Public management in intergovernmental networks

prevention of sexually transmitted diseases) to substance abuse, to the pre-

vention and control of child abuse, to the achievement of nutritional object-
ives, to the reduction of adolescent violence, to civil rights, and to the
improvement of life chances for disabled children. The “core” educational
function has been surrounded by and insinuated into a panoply of other
public objectives, and in turn a host of other organizations have become
involved in the day-to-day functioning of school district activities. Funding
and curriculum strength as well as program innovations depend in part upon
school district support from – and, in some circumstances, coproduction
with – other important stakeholders in other school districts, in the business
community, among community groups, and from elsewhere. School districts,
in short, typically operate within a network of other organizations and actors
that influence their students, resources, programs, goals, and reputation.
While this chapter concentrates on the most important intergovernmental
links in school districts’ networked environments, we do not ignore the
significant components of managers’ networking activity involving other
external parties as well. Managerial networking more generally is taken
account of in the analysis reported below, but the regular ties of school
districts with their other governmental partners receive systematic attention.
This study can be viewed, then, as an extension of earlier theoretical and
empirical work with the addition of an explicitly intergovernmental (and, as
will be explained, structural) component.
School districts can have regular ties with a variety of external parties.
Among the most important are links with other levels of government. The
key intergovernmental link for school districts in most US states is with the
state-level department of education. Typically, state education agencies do not
become directly involved in the provision of educational services. In recent
years, however, they have overcome the political support for local independ-
ence of the educational function and begun establishing and monitoring
accountability standards (Wong 1999). States – through their state education
agencies – are a critical source of funding for most school districts (again, the
extent and type of state support varies considerably from state to state), and
they also enforce some regulations, primarily about attendance and gradu-
ation requirements but also in connection with the certification of teachers.
More frequently they administer statewide standardized tests and evaluate
districts on the basis of them. The reliance of school districts on state
departments of education for significant funding and some direction on the
educational function means that state education agencies are a particularly
central actor in districts’ intergovernmental networks on a continuing basis.
245 The network theme: structure and behavior

Nationwide, state agencies provide significant funding for public educa-

tion. About one-third of all state-level expenditures went to education
nationwide in 2000 (Wulf 2002: 271), with 60 percent of this amount
provided intergovernmentally to local governments, primarily school dis-
tricts. The proportion of local educational spending deriving from state
sources varies between virtually all (as in Hawaii, where no local taxes are
used and only a tiny fraction of the overall total comes directly from the
federal government) and barely more than 10 percent of the total (New
Hampshire) (Council of State Governments [CSG] 2002: 486–7).3
One additional intergovernmental fiscal link can be significant: school
districts’ interdependence with the federal government, particularly the US
Department of Education. Unlike in some countries, the US national gov-
ernment is involved in the public education function only in relatively
peripheral ways. Even so, the Department of Education is not a minor player
in the environments of at least some school districts. Overall, approximately
12 percent of the total national budget for elementary and secondary public
education derives directly or indirectly from the federal government.4
Although this figure is not overwhelming, neither is it trivial. More import-
antly, the degree of national support varies considerably from district to
district. In some cases, particularly districts with large federal military bases,
national funding can provide an important supplement to state-level aid and
local own-source revenue.
In this chapter, we focus on these fiscal ties between school districts, on
the one hand, and the state and national educational funding agencies, on
the other. In effect, we examine the fiscal networks of school districts.
Districts raise a major portion of their own revenue, but most school
districts also depend on ties with other levels of government – especially
state education departments – for an important slice of their budget.
Management in the school districts receives primary attention. We are
interested in seeing how the degree of fiscal reliance on these other parties
in the educational network of school districts is connected to management
and its relationships to educational performance.

The network theme: structure and behavior

Sometimes, the word “network” has been used as a loose metaphor for the
interdependence characteristic of governance. Other analysts mean the term
to refer to relatively stable interorganizational structures, and they use the
246 Public management in intergovernmental networks

concepts and tools of sociology to dissect these and build empirical theory.
Yet other researchers focus primarily on networking: the actual connections
made between or among actors operating in some social space.
We avoid the metaphorical use of the term here, but we are interested in
the other two aspects of the network theme. Relatively stable arrays of
interdependent organizational actors are characteristic of the intergovern-
mental system. When grant programs continue over multi-year periods,
units develop elaborate formal understandings with each other, and the
networked sets of institutional actors approximate a structurally identifiable
cluster. (It is worth noting the use of images such as “picket fence federal-
ism” to describe US vertical links between or among donor and recipient
agencies that jointly administer public programs; see Wright 1988.) For
present purposes, in our treatment of intergovernmental relations and the
management of public education, we refer to this feature as the structural
network. Regardless of whether and how often actual interaction can be
observed between the linked entities, an ongoing fiscal link clearly defines
part of a structural relationship between the units involved.
In addition, managers may also be active in their networked environment
in efforts to build support for programs, attract partners for cooperative
effort, and fend off challenges from other actors. Some of these efforts may
in fact take place in concert with others involved in a structural network,
some may involve other dyads and other network actors – sporadically or
regularly. We refer to the set of these moves, which are in principle observ-
able, as the behavioral network, and the efforts of managers to be active in
this way as managerial networking.
Typically, researchers do not distinguish carefully between structural
features of networks and behavioral manifestations of networking. Both
signal some connection, perhaps interdependence, with other actors in a
public organization’s environment. These aspects may operate somewhat
independently of each other, however, and we view it as important to take
both into account in any comprehensive understanding of the network
theme in public management. Intergovernmental management, in particu-
lar, involves a consideration of both.
This study represents a first effort at combining structural and behavioral
networks. Accordingly, it raises a number of issues related to the substantive
area under consideration – educational policy – including what the relative
impacts of different funding sources are. Our primary focus is on manage-
ment and performance, however, and exploring these other issues is beyond
the scope of this book.
247 Intergovernmental networks: a formal treatment

Intergovernmental networks and public management:

a formal treatment

In this chapter we begin once again with our model, but concentrate on only
its second term by deliberately underspecifying the full set of relationships
we have hypothesized, as follows:

Ot ¼ b2 ðXt =Se ÞðM2 Þ þ et

The intergovernmental structural dimension can be considered part of the

environment, designated as Xi, and included as a separate term:


Ot ¼ b2 ðXi ÞX0t =Se ðM2 Þ þ et

Here X0t simply refers to the portion of the Xt vector besides the
intergovernmental structure – that is, all other environmental forces impin-
ging on the local government. These include a wide variety of resources (for
example, the extent of local wealth) and constraints (the task difficulty of
educating the children of a particular community; a school system educating
a more diverse or impoverished student population confronts a larger
challenge). We simplify further by not designating the stability relationship
as a reciprocal one:
Ot ¼ b2 Xi X0t Se M2 þ et
This leaves us with a highly complex four-way interaction between the
environment, intergovernmental structure, stability, and management. The
best test to determine if the relationships actually fit a four-way interaction
is with a set of nested hypothesis tests that contrast various interactions with
strict linear additive models. Doing so would require us to include not just
the four additive terms and the four-way interaction, but also four three-way
interactions and six two-way interactions. Such a set of tests puts a great deal
of stress on a data set by generating massive collinearity, and would likely
make pinpointing the precise relationships extremely difficult. To provide a
first step in assessing the linkage between the intergovernmental environ-
ment and the various forms of management and stability, we further sim-
plify the model to one with interactions between intergovernmental
structure and the remaining terms only.


Ot ¼ b2 Xi X0t þ Se þ M2 þ et
248 Public management in intergovernmental networks

which can be transformed to the following equation to be estimated:

Ot ¼ b2 Xi X0t þ b3 Xi Se þ b4 Xi M2 þ et
In terms of actual testing, we have two measures of intergovernmental
structures, two measures of management, and two measures of stability, in
addition to several environmental variables. This simplified model probably
gives up a fair amount of explanatory power by omitting the lagged depend-
ent variable in particular.
Although the model has been simplified in the last equation, the relation-
ships continue to incorporate a number of features we expect to find for
public management in networked settings. Behavioral networking is expli-
citly incorporated, and in the analyses that follow we add our measure of
managerial quality as well. In addition, the “S” term, as indicated above,
represents a set of stability-inducing forces. Earlier we incorporated man-
agerial and personnel stability, and we retain these elements here. The Xt
term represents a vector of environmental forces, and we include several of
the most important of these in the analyses as well. From the Xt term, which
had been relatively undifferentiated beforehand, we extract two aspects of
the intergovernmental system for closer examination and denote them
with Xi. Both variables represent financial dimensions of the intergovern-
mental environment. The first is the dependence on funds from other levels
of government, especially those from state governments, and the second is
the diversity of school district funding sources. The degree of dependence on
state and national funders renders a school system’s environment more
network-like in structural terms. We interpret school systems with high
dependence on intergovernmental aid as being situated in a more networked
setting than those that are self-funded. State aid is the primary kind of such
support, and we incorporate a measure of dependence on state aid into our
treatment of structural stability in the analyses below.5 In other words,
dependence on state aid taps what we are calling structural networks, as
distinct from the managerial networking we also examine.
In terms of the intergovernmental network itself (beyond the dependence on
state aid), another feature for school districts that should tap the degree of
structural complexity facing districts and their managers is the extent to which
a school district’s financial support is provided by a diversity of sources in the
network rather than merely one funder. Some districts might derive virtually all
their financial resources from their own-source revenues. These, presumably,
are the most independent of other networked partners, in that they exert their
own taxing authority rather than being dependent on decisions of others.
249 The units of analysis

Others might derive considerable financing from the state or the federal
government. These may be advantaged, in a sense, in having an additional
source of funds; but the management challenges in dealing with own-source
and intergovernmental funding streams, and the uncertainties connected to the
latter, render managerial networking (and perhaps other managerial functions)
more important. Those school districts with financial support stemming in
substantial measure from more than one intergovernmental funding stream
would seem to face more challenging and, potentially, more uncertain network
environments. For these reasons, the sheer diversity of funding across the
intergovernmental network can also be expected to be related to the requisites
of intergovernmental management. Dependence on state aid, in short, and
dependence on a more diverse and thus complex intergovernmental network
are features of the interdependent environment that should be related to school
district management, and also to performance.

The units of analysis

Our data are drawn from the 1,000 plus Texas school districts. District
superintendents were sent a mail questionnaire on management styles, goals,
and time allocations (return rate 55 percent with 507 usable responses). We
pooled five years (1995 to 1999) of data on performance and control
variables to produce a total of 2,535 cases for analysis. All nonsurvey data
were from the Texas Education Agency.
Our measures can be discussed in terms of parts of the model: manage-
ment (M); elements of stability (Se); the vector of environmental forces (X’);
intergovernmental structures (Xi); and program outcomes (O), or perform-
ance. These items are covered in this order.

“M”: management
Two measures of public management are included as potential explanatory
variables in this analysis: managerial quality and managerial networking. We
treat them both here as elements of the M2 term in our simplified model but
leave them distinct in our efforts to develop estimations. Managerial quality, as
reported in Chapter 4, is based on the residual from a model explaining the
salaries of district superintendents. A second measure of management is
included, as well: managerial networking. To measure the behavioral network-
ing activity of school superintendents, we use the factor score from the reported
250 Public management in intergovernmental networks

interactions with five sets of actors from the organization’s environment: school
board members, local business leaders, other school superintendents, state
legislators, and the TEA. The managerial networking measure and the man-
agerial quality measure are uncorrelated with each other (r ¼ 0.01).6

“Se”: stability
We incorporate two aspects of personnel stability in this study, as developed
in Chapter 5. Managerial stability is simply the number of years the super-
intendent has been employed by the district in any capacity. Teacher stability
is measured as the percentage of teachers employed by the district during the
preceding year who continue to work for the district. For both measures,
then, higher scores mean more stability. Data on managerial stability were
obtained from the survey respondents; data on teacher stability were pro-
vided by the Texas Education Agency. Although these measures are treated as
stability features here and in the subsequent discussion, they can also be
considered aspects of management – what is usually referred to as personnel
management. While not totally under the control of school district leaders,
these variables are susceptible to influence by the individuals who make
decisions about how such organizations are run. In a real sense, therefore, all
four variables tap aspects of public management.

Xi environment: the intergovernmental dimension

School districts differ substantially in terms of their structural network
contexts, and we focus on the intergovernmental aspect of this structure: their
dependence on state aid and their dependence on a diversity of funding.
Dependence on state aid is measured by the percentage of total school district
funds that originate from the state. The average district receives 51.5 percent
of its funds from the state, but the range is from 0 to 100 percent (allowing
for rounding errors). We dichotomize this variable with a median split, desig-
nating districts with more than 58 percent of state aid as highly dependent on
state aid. For the diversity of funding measure, we take the percentages of state,
local, and federal funds and squared them. This number is then subtracted from
10,000. A score of zero means that the district received all its funds from a single
source; higher scores indicate greater diversity, with a maximum possible score
of 6,733 (if funds came equally from each of the sources). The mean diversity
measure is 4,334 with a standard deviation of 1,217, range 198 to 6,358. We
again use a median split, with district scores above 4,735 designated as more
251 Findings

diverse. A district with a score of 4,735 might have approximately 4 percent

federal funds, 66 percent state funds, and 30 percent local funds, although many
other combinations are possible.

X0t: control variables

Any assessment of public program performance must control for both task
difficulty and program resources. For school districts, neither of these types
of elements are under the substantial control of the districts themselves, and
therefore they can be considered key parts of the vector of environmentally
influenced X’ forces represented in the model. We use the usual set of three
measures of task difficulty and five measures of resources, as employed
several times already in this volume.

“O”: performance measures

Finally, for measures for O, or performance (outcomes), we use ten different
performance indicators in an effort to determine if intergovernmental net-
work structures influence how public management and personnel stability
affect a variety of organizational processes. We continue to employ the all-
pass rate on the Texas Assessment of Academic Skills, as well as four other
TAAS measures: TAAS scores for Anglo, black, Latino and low-income
students are included as measures of performance indicators. Three meas-
ures of college-bound student performance are used: average ACT score,
average SAT score, and the percentage of students who score above 1,110 on
the SAT (or its ACT equivalent).7 The final two measures of performance
might be termed bottom-end indicators: attendance rates and dropout rates.
Dropout rates, although it is conceded that they contain a great deal of error,
are frequently also used to evaluate the performance of school districts. The
official state measure of dropouts is the annual percentage of students who
leave school from eighth grade onward.

Findings

Our strategy for testing the simplified model is to conduct regression

analyses that seek to explain the performance measures across the districts
for all ten performance measures available while also distinguishing
simple and complex intergovernmental networks for each analysis. This
latter step allows us to see how all the sets of relationships differ between
252 Public management in intergovernmental networks

more and less interdependent intergovernmental network settings (and

more and less diverse intergovernmental funding sources). This last
aspect of the approach helps, in particular, to probe for nonlinear rela-
tionships – a central aspect of the model and one often difficult to
analyze systematically. Nonlinear impacts can be assessed either via inter-
action terms or by examining relationships with different subsets of the
sample. The former, while elegant, is often plagued by severe collinearity
problems that prevent meaningful interpretation of coefficients. Accord-
ingly, we explore nonlinear relationships among several of the management
and intergovernmental-structure variables here via physical controls – that
is, by partitioning the data set. Given the large number of regression
analyses required to examine the full range of educational performance
(ten dependent variables) and the fact that we are using two different
measures to tap intergovernmental network structure, we partition the
sample in a straightforward fashion: each specification is run with esti-
mations developed separately for the top and bottom halves of the
sample, based upon intergovernmental network structure.
Forty separate regressions have been run. First, using dependence on state
aid as our measure of intergovernmental structure, we have sought to
explain outputs and outcomes on each of the ten performance measures.
Since the sample is split into high and low state aid districts for each of the
ten measures, these analyses represent twenty estimations. The state aid
measure of intergovernmental structure was then replaced with the measure
tapping diversity of funding, and all twenty analyses were rerun. These forty
sets of results are displayed in Tables 7.1 to 7.8; to provide overviews of the
patterns embedded in the accumulation of so many results, we have sum-
marized the most important tendencies in Tables 7.9 and 7.10. For the sake
of emphasis on the sets of relationships of interest, we report here only the
results for the management and stability variables; the same control variables
were used in each analysis – all eight controls explained earlier were included,
as were dummy variables for each year of the time series.
With regard to the management- and personnel-stability-related variables,
a first general point is that the hypothesized positive relationships between
management and personnel stability, on the one hand, and performance, on
the other, are clearly supported. While the relationships and their strength
vary depending on which aspect of performance is considered, the overall
pattern is unambiguous for settings reflecting both higher and lower
quantities of intergovernmental network structure. Tables 7.1 to 7.4
report on the twenty regression analyses that included state aid as
253 Findings

Table 7.1 Management and dependence on state aid

Dependent variables ¼ TAAS pass rates for

All students Black students Latino students

State aid State aid State aid

High Low High Low High Low

Networking 0.713* 0.574* 0.326 0.669* 0.702* 0.376

(0.217) (0.211) (0.647) (0.435) (0.408) (0.343)
Quality 0.678* 1.030* 0.164 0.692* 0.882* 0.774*
(0.249) (0.221) (0.719) (0.424) (0.453) (0.361)
Teacher stability 0.176* 0.089* 0.338* 0.280* 0.096* 0.005
(0.035) (0.035) (0.113) (0.099) (0.067) (0.065)
Management stability 0.097* 0.060* 0.167* 0.091* 0.113* 0.035
(0.025) (0.022) (0.068) (0.040) (0.045) (0.035)
R2 0.62 0.61 0.29 0.52 0.33 0.42
N 1,246 1,222 705 818 1,119 1,133

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard errors.
* ¼ significant at p < 0.10, one-tailed test.

Table 7.2 Management and dependence on state aid II

Dependent variables ¼ TAAS pass rates for

Anglo students Low-income students

State aid State aid

High Low High Low

Networking 0.948* (0.222) 0.457* (0.202) 0.293 (0.270) 0.254 (0.276)

Quality 0.374* (0.256) 1.017* (0.211) 0.771* (0.311) 1.116* (0.289)
Teacher stability 0.156* (0.036) 0.106* (0.034) 0.168* (0.044) 0.084* (0.047)
Management stability 0.111* (0.025) 0.060* (0.021) 0.136* (0.031) 0.120* (0.029)
R2 0.45 0.47 0.52 0.54
N 1,226 1,214 1,246 1,212

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard
errors. * ¼ significant at p < 0.10, one-tailed test.
254 Public management in intergovernmental networks

Table 7.3 Management and dependence on state aid: college aspirations indicators

Dependent variables

ACT scores SAT scores Above 1,110

State aid State aid State aid

High Low High Low High Low

Networking 0.076* 0.007 3.53* 2.74* 0.528* 0.126

(0.040) (0.037) (2.24) (1.87) (0.279) (0.281)
Quality 0.019 0.159* 3.36* 6.45* 0.098 1.283*
(0.046) (0.039) (2.51) (2.01) (0.327) (0.295)
Teacher stability 0.004 0.006 1.40* 0.08 0.053 0.099*
(0.007) (0.008) (0.42) (0.41) (0.046) (0.053)
Management 0.002 0.007* 0.05 0.32* 0.031 0.008
stability (0.004) (0.004) (0.24) (0.19) (0.032) (0.029)
R2 0.35 0.33 0.47 0.56 0.24 0.36
N 1,097 1,068 869 921 1,216 1,138

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard
errors. * ¼ significant at p < 0.10, one-tailed test.

Table 7.4 Management and state aid: low-end indicators

Dependent variables

Attendance Dropouts

State aid State aid

High Low High Low

Networking 0.007 (0.021) 0.042* (0.023) 0.080* (0.031) 0.007 (0.029)

Quality 0.064* (0.024) 0.061* (0.024) 0.108* (0.036) 0.099* (0.031)
Teacher stability 0.017* (0.003) 0.004 (0.004) 0.006 (0.005) 0.004 (0.005)
Management stability 0.000 (0.002) 0.001 (0.002) 0.002 (0.004) 0.001 (0.003)
R2 0.29 0.23 0.16 0.20
N 1,246 1,223 1,246 1,203

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard
errors. * ¼ significant at p < 0.10, one-tailed test.
255 Findings

Table 7.5 Management and funding diversity

Dependent variables ¼ TAAS pass rates for

All students Black students Latino students

Diversity Diversity Diversity

High Low High Low High Low

Networking 0.586* 0.912* 0.111 1.052* 0.438* 0.760*

(0.196) (0.229) (0.447) (0.655) (0.336) (0.343)
Quality 0.567* 1.144* 0.591 1.440* 0.113* 1.412*
(0.233) (0.231) (0.488) (0.599) (0.395) (0.402)
Teacher stability 0.159* 0.132* 0.487* 0.225* 0.143* 0.017
(0.038) (0.033) (0.105) (0.110) (0.069) (0.064)
Management 0.091* 0.038* 0.079* 0.143* 0.128* 0.025
stability (0.022) (0.025) (0.042) (0.067) (0.036) (0.044)
R2 0.63 0.58 0.44 0.30 0.39 0.34
N 1,279 1,220 865 681 1,185 1,095

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard
errors. * ¼ significant at p < 0.10, one-tailed test.

Table 7.6 Management and funding diversity II

Dependent variables ¼ TAAS pass rates for

Anglo students Low-income students

Diversity Diversity

High Low High Low

Networking 0.514* (0.200) 1.112* (0.224) 0.058 (0.246) 0.685* (0.294)

Quality 0.635* (0.238) 0.920* (0.226) 0.679* (0.293) 1.052* (0.298)
Teacher stability 0.137* (0.038) 0.142* (0.033) 0.141* (0.047) 0.135* (0.044)
Management stability 0.107* (0.022) 0.033* (0.024) 0.159* (0.027) 0.072* (0.032)
R2 0.47 0.42 0.57 0.45
N 1,226 1,199 1,278 1,209

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard
errors. * ¼ significant at p < 0.10, one-tailed test.
256 Public management in intergovernmental networks

Table 7.7 Management and funding diversity: college aspirations indicators

Dependent variables

ACT scores SAT scores Above 1,110

Diversity Diversity Diversity

High Low High Low High Low

Networking 0.009 0.121* 4.86* 1.33 0.060 0.906*

(0.035) (0.042) (1.79) (2.45) (0.233) (0.331)
Quality 0.094* 0.094* 2.66 4.26* 1.067* 0.495*
(0.042) (0.042) (2.10) (2.44) (0.281) (0.337)
Teacher stability 0.000 0.009* 0.57* 1.26* 0.038 0.096*
(0.008) (0.007) (0.42) (0.43) (0.047) (0.052)
Management 0.007* 0.006* 0.40* 0.10 0.038* 0.079
stability (0.004) (0.005) (0.18) (0.26) (0.025) (0.036)
R2 0.29 0.45 0.51 0.50 0.31 0.31
N 1,146 1,046 1,005 805 1,238 1,145

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard
errors. * ¼ significant at p < 0.10, one-tailed test.

Table 7.8 Management and funding diversity: low-end indicators

Dependent variables

Attendance Dropouts

Diversity Diversity

High Low High Low

Networking 0.026* (0.019) 0.037* (0.023) 0.062* (0.029) 0.013 (0.030)

Quality 0.033* (0.023) 0.072* (0.024) 0.100* (0.034) 0.078* (0.031)
Teacher stability 0.011* (0.004) 0.014* (0.003) 0.003 (0.006) 0.004 (0.005)
Management stability 0.001 (0.002) 0.000 (0.003) 0.002 (0.003) 0.000 (0.003)
R2 0.29 0.23 0.20 0.15
N 1,279 1,220 1,277 1,202

Notes: All equations control for teacher salaries, instructional funding, percentage of black students,
percentage of Latino students, percentage of low-income students, class size, teacher experience,
noncertified teachers, and individual year dummy variables. Numbers in parentheses are standard errors.
* ¼ significant at p < 0.10, one-tailed test.
257 Findings

Table 7.9 Summary of results: state aid

High Low Neither

Networking 8 1 1
Quality 3 7 0
Teacher stability 7 1 2
Management stability 5 2 3

Notes: Comparison of slopes for districts with a great deal of state aid with
those with less aid. Figures are which slope is larger. Neither is coded if
neither coefficient is significant or if they are equal. Median splits on all,
including federal aid.

Table 7.10 Summary of results: funding diversity

High Low Neither

Networking 2 8 0
Quality 2 7 1
Teacher stability 5 4 1
Management stability 7 1 2

Notes: Comparison of slopes for districts with a great deal of funding

diversity with those with less aid. Figures are which slope is larger. Neither is
coded if neither coefficient is significant or if they are equal. Median splits on
all including federal aid.

the measure of intergovernmental network structure. Of the eighty coeffi-

cients reported in these tables – twenty each for behavioral networking by
top managers, top managerial quality, teacher stability and management
stability – fifty are statistically significant and in the expected direction,
and only two are statistically significant in the opposite direction.8 Based
on the binomial probability distribution, the probability of a set of rela-
tionships this consistent is less than one in a million. When state funding
is replaced by our measure for funding diversity, fifty-six of the eighty
relationships are significant in the predicted direction, four in the opposite
direction. Again, the probability of this pattern of relationships if the
actual data were random is less than one in a million. Overall, manage-
ment – measured via both behavioral networking and quality – as well as
personnel stability certainly improve educational systems’ performance.
Second, although there are, clearly, general patterns, the degree of explan-
ation varies across the performance measures. This point holds whether the
sample is split by state aid or funding diversity. Our ability to explain
258 Public management in intergovernmental networks

the variance in a number of performance measures is quite good, although

the level of explanation for the so-called “low-end” performance measures is
modest. For equations explaining attendance and dropout rates, the R2
ranges between 0.15 and 0.29. The equations concerning dropouts are
particularly unimpressive (unlike for the other performance measures, this
one looks better as scores decline; accordingly, coefficients indicating con-
tributions to performance should be negatively signed), and a probable
reason is the quality of dropout data, as explained above. For both low-
end measures, additional factors are surely important in driving results.
Keeping students in school and getting them to attend on a regular basis
are particularly challenging issues when students are beset with other prob-
lems, such as substance abuse, family difficulties, and the like. That the low-
end equations leave most variance unexplained is not surprising. We include
and discuss these results below but concentrate particular attention on the
other eight equations.
Third, the relationships in different intergovernmental settings clearly
differ; in short, management interacts with intergovernmental structures
to generate nonlinear relationships just as the model predicts. In the remain-
der of this discussion, we concentrate on how managerial and personnel
stability impacts vary by intergovernmental network type.
We consider the results in Tables 7.1 to 7.4 first. An interesting finding is that
managerial networking is more important for performance in districts that
receive more state aid. The criterion for the judgment of a difference was
simply whether one coefficient was larger than the other, subject to the con-
straint that at least one of the pairs of coefficients had to be statistically
significant. In eight of the ten measures the coefficients for managerial net-
working are greater for the high-aid half of the sample; in one case – TAAS pass
rates for black students – the opposite is the case.9 In some of the cases the
difference is substantial. The effect size for networking in improving Latino
students’ state exam pass rates in high-aid districts is a maximum of more than
four points, whereas the impact for low-aid districts is not statistically signifi-
cant.10 An impact of four points might seem relatively small, but over the long
term increases of this magnitude can make a real difference. The impact of
managerial networking is even greater on pass rates for Anglo students in the
high-aid part of the sample: a maximum of about 5.7 percent; in the other half
of the sample the relationship is still positive but is less than half the size.
Networking efforts on the part of managers generally pay greater divi-
dends if another government, in this case the state, is a major source of
resources. Higher state aid means that districts are more dependent on a key
259 Findings

partner in their structural network. Major financial benefits for the district
may be contingent on managers’ treating the issue of intergovernmental
support seriously, and external managerial effort does contribute more to
results. In this case, good behavioral networks mimic structural networks.
A second pattern is discernible in Tables 7.1 to 7.4. Personnel stability also
tends to matter more for performance in those settings more dependent on
state aid. The relationship is particularly clear for teacher stability. For seven
performance measures, teacher stability contributes more in the high-aid
districts, whereas for only one – high SAT performance – is the pattern
reversed. For two performance measures, ACT scores and dropout rate,
teacher stability is significant for neither half of the sample. In some
instances, such as the all-students’ TAAS pass rate, the teacher stability
impact for the high-aid districts is twice as large.
The pattern across the performance measures is similar for management
stability, although the results are a bit less straightforward. For five cases, the
high-state-aid half of the sample gets a bigger impact on performance from
superintendents’ stability than does the low-aid half; in two cases, the
pattern is reversed. Three cases do not involve statistically significant
impacts of management stability on performance, although two of these
are the instances focused on the low-end measures.
Personnel stability helps educational performance more in the more heavily
networked settings – at least, as measured by districts’ dependence on the
primary source of intergovernmental aid. This pattern is especially interesting.
Often the characterization of networks and network management suggests that
stability and networking are somehow opposed. Either stability impedes the
entrepreneurial networking often endorsed by network enthusiasts, or net-
working promises to break through the rust of overly entrenched bureaucracy.
The findings reported here point toward a more intriguing pattern: that
stability – at least of personnel – and intergovernmental network dependence
interact positively with respect to program outcomes. Stability in at least some
forms may be a platform on which managers and others can build effective
performance in heavily networked settings.
Another way of characterizing the findings is that personnel stability
may compensate for, and be especially important in dampening, some of
the disruptiveness of structurally less stable (more networked) settings.
A second possibility is that personnel stability allows the manager to turn
network interactions into repeat games, thus allowing each side to build
trust and make credible commitments. This characterization is close to
what researchers report as typical in instances of intergovernmental
260 Public management in intergovernmental networks

cooperative agreements. If so, to treat stability in general and networked

patterns as somehow at odds would be a distortion. It may be especially
important for public managers in networked intergovernmental settings to
work to build sufficient stability into their contexts so as to be able to
operate effectively – or to find ways of balancing the compensating advan-
tages and disadvantages of various kinds of stability and instability for
delivering results. This set of findings is generally consistent with a theme
articulated by Milward and Provan (2000: 370), who have emphasized that,
in networked settings, stability in certain senses may be valuable for
effective performance.11
One other pattern is worth noting from Tables 7.1 to 7.4. Management
quality matters most in settings with less fiscal dependence on the intergov-
ernmental network. This relationship obtains in seven of the ten pairs of
regressions; in three, including the two sets of low-end performance ana-
lyses, the relationship is reversed. It is not immediately clear why manage-
ment quality should matter more in the more structurally homogeneous
districts, although here the school systems must learn to rely more on their
own fiscal resources or transactions – a kind of “fend for yourself ” aspect of
intergovernmental relations. The relationship is striking in some instances,
not least the results for standardized state test performance for Anglo
students; quality matters here for both high- and low-aid districts, but for
the latter its impact is nearly triple. (The maximum effect of quality for
districts heavily reliant on state assistance is about 2.2 percentage points on
districts’ pass rate but about 6.1 percentage points for the other half of the
sample.)
Top managers of high quality may be particularly influential in more
“standard” settings – those contexts best approximating the hierarchical and
structurally stable environments that public management scholars and prac-
titioners have been assuming for decades. If so, perhaps if high-quality
managers become trained for the more networked world of many contem-
porary public programs, they could have greater impact in more challenging
program contexts. Considering these results in conjunction with the person-
nel stability findings discussed above also suggests that, as settings become
more network-dependent, sheer managerial (and others’) stability – longevity
in the system, chances to learn the lie of the land – starts to trump
managerial quality. Such a conclusion is plausible when it is considered
that, in less structurally stable surroundings, knowing the contacts and
history of how to get things done may be a particularly prized (and scarce)
asset for managers, as well as others.
261 Findings

Another possibility is that more diverse networked program contexts are

inherently less manageable, so the impact of managerial quality is tempered
in such circumstances.12 If so, the structural context for public management
may be important in ways that need to be considered by analysts and
practitioners. The findings reported in Tables 7.1 to 7.4 are summarized in
Table 7.9. Overall, these findings support the theoretical notions contained
in our simplified specification.
Tables 7.5 to 7.8 display the results, excluding the control variables, for
the same set of analyses, with our measure of funding diversity replacing
that for the extent of reliance on state aid. Funding diversity measures how
much a school district turns to multiple sources of aid (federal, state, and
own-source) to support itself. The measure taps not the extent of reliance
on funding from the intergovernmental network or any of its components
– which would be a measure of fiscal dependence – but, rather, the extent
to which a district has to juggle a multiplicity of significant funding
streams. We are measuring something like the complexity of the intergov-
ernmental funding network. Splitting the sample between low- and high-
funding-diversity segments reflects a structural distinction between simpler
and more complex structural settings for trying to manage the education of
students.
Tables 7.5 to 7.8 contain the coefficients for the funding diversity
regressions, while Table 7.10 summarizes the findings. The more a dis-
trict’s funds are concentrated in one source – any source – the more
important managerial networking and managerial quality are to perform-
ance. For networking, eight of the ten pairs of regression results point in
this direction. For managerial quality, seven sets of regressions indicate
that greater impacts of quality are associated with less diverse funding
arrangements.13 In a number of cases, the differences are quite large. In
one, the Latino student pass rate on the statewide test, quality is actually
negatively related to performance for “complex” districts, although the
slope is small, while it has a substantial positive impact for the low-
funding-diversity half of the sample (a maximum impact of approxi-
mately 8.5 percentage points).
Funding homogeneity seems to allow top managers to focus their
networking and have a chance to produce better results. When funding
diversity is high – that is, when the funding network is complex –
managers presumably have to be effective in spreading themselves and
their efforts over more nodes. As a result, the actual impact of any given
level of networking is lessened overall. Having to deal with several
262 Public management in intergovernmental networks

important sources of funding means that managers likely have to handle

higher levels of goal conflict – a circumstance that would also complicate
the task. Whereas some literature implies that operating in a networked
world is something that adept managers should be able to master (for
example, Mandell 2001), these findings raise the question of whether
networks of a certain complexity, and perhaps a degree of goal conflict,
strain or surpass even active networkers’ abilities to perform their man-
agerial role effectively. The answer to this question carries implications
not only for those interested in networks and the performance of public
programs, but also for those who care about intergovernmental programs
and their management.
This sort of logic, combined with the findings on state aid, might lead
to the expectation that personnel stability should matter more in the
high-funding-diversity school districts. These are the ones in more com-
plex and perhaps conflictual intergovernmental networks, at least in a
fiscal sense. If the earlier discussion can be applied here, the more
complex settings should also be ones in which the stability of personnel
can contribute more to performance. The results provide some support
for this idea. Here management stability seems clearly more important in
the high-diversity settings; seven sets of equations point in this direction,
only one (for the state exam pass rate for black students) in the opposite
direction.14 In several of the cases, the differences in slopes for manage-
ment stability were considerable between the two parts of the sample.
Again, this finding makes sense, in that experienced top managers are
more likely to be effective in navigating the complex and conflictual
world of intergovernmental funding to produce effective performance.
The impacts of teacher stability on performance do not vary consistently
between high- and low-diversity settings. For five of the ten measures,
teacher stability matters more when funding is spread among several
sources. For four measures, it matters more when funding is concentrated
as to source.15 Since teachers themselves are virtually never involved in
negotiating intergovernmental aid or dealing with the complexity of a
school district’s intergovernmentally networked surroundings, the mixed
results here are not very surprising.
If we treat funding diversity, then, as a measure of network complexity,
the findings overall are rather straightforward. More concentrated
funding streams mean that top managerial networking and quality are
more important. Managerial stability becomes consistently more import-
ant with diverse funding sources. Managers matter for performance in
263 Conclusions

intergovernmentally networked settings, but they appear to have larger

impacts when the networks are not too complex and conflictual, so that
they are presumably able to concentrate their external efforts where it will
do some good. Where the setting is more complex, managerial longevity
helps more.

Conclusions

Among the most important kinds of networked contexts is the set of

situations in which the intertwining of interdependent actors is a product
of conscious design rather than evolution or chance. The most explicit and
longstanding kind of designed interdependence is the intergovernmental
governance system. In some contexts, such as the emerging forms of the
European Union, such multilevel governance is a fairly recent and
immensely important product of expanding networked action. In other
contexts, such as a range of intergovernmental programs in federal systems
such as that in the United States, intergovernmental patterns are many,
variegated, constitutionally legitimated, and of longstanding operation. All
such settings deserve systematic analytical attention, and the burgeoning
interest in networks portends increasing focus on intergovernmental gov-
ernance and management.
This trend is propitious, not only because the topic is important and will
likely grow in salience, but also because the subject of intergovernmental
management, while critically important, remains in need of theoretical
approaches. Developments on the subject of networks may be able to help
in this regard. Our effort to explore this notion has been constructed from a
general theory of networks and public management developed without
particular attention to intergovernmental program management but with
potential applicability there.
We have applied a simplified version of our model to an enduring and
distinctively American set of cases, school system management, that
nevertheless should be instructive regarding the role and importance of
public management in intergovernmentally networked settings. The
model provides help in unpacking an exceedingly complex set of rela-
tionships. It, and the findings of this chapter, can offer both hypotheses
and guidance for further research on intergovernmental management in
networked settings.
264 Public management in intergovernmental networks

The evidence is clear that managerial networking, managerial quality, and

selected stabilizing features (most systematically, personnel stability) con-
tribute positively to program performance, at least for many measures
associated with education in Texas. The evidence is also overwhelming that
nonlinear interactions between structure, management, and environmental
forces are commonplace in the world of networked public programs.
Further, discussions of networks and networking can benefit from more
conceptual clarity. In particular, the networking behavior of managers
(and others) is not the same thing as the structural interdependence that often
binds elements of networks together. Our model helps to elucidate this distinc-
tion by labeling the behavioral aspect as a component of management – M2 –
and the structural part as one aspect of the environmental vector – Xi. We have
measured these separately; in fact, we have tried to tap the latter in two ways
regarding the intergovernmental fiscal interdependence of school districts and
sought to explore how these are related. While managers may do a lot of
networking in either high- or low-dependence settings, their efforts pay off
more when the structural environment reflects more reliance on external
resources. When managers face dependences from several directions, their
networking efforts are less effective than when they can concentrate their
efforts on limited sources. Behavioral networking helps, but it helps more in
certain kinds of networks; and the intergovernmental structure is part of the
explanation.
Rather than treating stability as an enemy of networks and networking,
and the converse, this chapter shows that, in intergovernmental settings,
some kinds of stability may actually be a help or precondition to more
effective networked action. When school districts operate in more fiscally
interdependent and complex settings, managerial stability and personnel
stability seem to provide greater contributions to more effective perform-
ance. Stability as a platform for risk taking, entrepreneurial action in net-
works: this idea deserves further careful exploration, especially as
governance systems expand and multilevel intergovernmental arrays are
increasingly developed. How to deliver performance in such settings
becomes a critical issue, and the findings here suggest that certain subtleties
may be part of the answer.
The stability and networks relationship in this chapter might also be
linked to our earlier findings on managerial capacity. In that case, another
form of stability, a large managerial cadre, facilitated the organization’s
ability to perform in the face of a crisis. Crises and networks are forms of
uncertainty; stability and management capacity are both forms of
265 Notes

organizational presence, perhaps even of bureaucracy. Organizations that

run counter to less certain environments appear to be better performers in
both cases. Because these findings are generally contrary to current ideolo-
gies of management consulting, they hold important lessons for public
managers and the design of governance institutions alike.
Our findings about managerial quality also deserve mention and are
related to this theme. Quality matters, but not in the same way and to the
same extent in all settings. Finding and retaining first-rate managers clearly
pays off in performance; the analyses reported here indicate support for this
point. In the more interdependent and more complicatedly networked
intergovernmental contexts, however, the impact of good managers, ceteris
paribus, is somewhat less. A number of implications might flow from this
finding, especially if further support is found in other settings and other
policy fields.
Rather than tease these out in detail here, however, we end the chapter by
emphasizing some of the tasks yet to accomplish. The general model of
public management continues to show promise, and the role of manage-
ment, directly (in terms of networking and quality) and indirectly (in the
recruitment and retention patterns of personnel), is clearly a crucial feature
of successful performance for public education in Texas. The complex set of
managerial influences deserves analysis in other places, other policy sectors,
and other countries. Additional dimensions of intergovernmental structure,
including those beyond the fiscal, are worth attention as well. Other stabil-
izing forces should also be explored. So, while the results here are complex
and promising, they point toward yet more puzzles and should encourage
still more research.

NOTES

1. Influences emanating from the environment of a public organization are properly treated
as part of the “X” term when one is analyzing management and performance at the
organizational level of analysis. If one were examining the system at the network level of
analysis, network structural arrangements would be considered part of the vector of
stabilizing features, Sn, of the larger array (see the coverage of modeling at the network
level in Chapter 2).
2. The analyses in this chapter follow the presentation by O’Toole and Meier (2004c).
3. We cite information here from earlier years, since the data we analyze in this chapter
pertain to an earlier five-year period, but the general points made about the fiscal structure
of the system and variation nationwide continue to hold.
266 Public management in intergovernmental networks

4. The total revenue for public school systems in United States for 1998/9 was
more than $200 billion, of which approximately $24 billion derived from the national
government. The vast majority of the nationwide elementary and secondary education
dollars are raised directly by local districts, primarily from their own taxes, with some
coming from other local governments and some from charges imposed by the
districts. See CSG 2002.
5. We use state funds rather than state and federal funds because, for most districts, the
percentage of federal funds is relatively small. For a small number of districts with major
federal facilities, such as military bases, federal aid is the largest source of funds. These
“federal impact” districts are fundamentally different from districts that depend on state
sources, in that federal funds essentially substitute for local monies. Including federal
funds in the analysis results in generally similar results to those reported here, but the
federal impact districts tend to muddy the findings. The diversity measure also picks up
the federal impact districts.
6. Management quality is also uncorrelated with both measures of personnel stability (–0.02
for each), as is managerial networking (0.04 correlation with teacher stability, –0.08 with
management stability).
7. The relationship between the percentage of students taking the tests and the test scores in
Texas is actually positive but explains less than 2 percent of the variance.
8. Of the twenty-eight non-significant relationships, nineteen are in the predicted direction.
9. For one performance measure, TAAS pass rates for low-income students, neither coeffi-
cient was statistically significant.
10. Even if it were, the slope is substantially smaller.
11. Their focus is on another aspect of stability: constancy of the key network organizational
members – in contracted relations for service provision – over a several-year period.
12. Given the findings analyzed here, this may be the case not only for M1, the internal
management of operations that has been omitted altogether in the simplified model, but
also for overall managerial quality.
13. For both networking and quality, the dropouts performance equation contained anom-
alous results. Given the quality of the data, therefore, the pattern can be considered even
stronger than the summary numbers suggest.
14. The equations for the two low-end performance measures all show insignificant results
for managerial stability.
15. For the dropout rate, the teacher coefficient was insignificant for both halves of the
sample.
8 Public management and performance:
what we know, and what we need
to know

This book has presented a perspective, a model, and a large set of empirical
findings. The results speak to a broad agenda occupying many scholars and
practitioners: understanding how public managers shape agency and
program performance. In this chapter we draw the volume to a close by
undertaking two tasks. First, we review and tie together what we have
learned about public management and performance. We then sketch a new
agenda for some of what, we believe, remains to be explored.

What does the evidence show?

This research program has developed many findings about whether, how,
and how much public management influences the performance of public
organizations. Most findings explain what is going on in school districts
within one state; but, we argue strongly, this “limitation” should not dimin-
ish the record. Although this work should be replicated in other empirical
settings – and although we have initiated that ourselves, as have others
discussed in earlier chapters – the patterns analyzed here should not be
marginalized. First, the sample included in most of our empirical studies
consists of more than 1 percent of all governments in the United States.
Second, no larger sample of public organizations has ever been analyzed for
determinants of performance. Third, the analyses develop and report on
numerous types of managerial influence. While the findings sometimes
show that managers shape performance in ways that some might find to
be expected, the results do not merely theorize or speculate about such
channels of influence; they demonstrate the influence with systematic evi-
dence. This sort of contribution is unusual. The approach here, therefore, is
what can be considered to constitute evidence-based public management.
Fourth, the characteristics of these organizations suggest that many of the
results here are likely generalizable to other public organizations that are
267
268 Public management and performance

highly professionalized and operate via a relatively decentralized structure,

especially when such organizations function as street-level bureaucracies
(Lipsky 1980). Fifth, substantial practical limitations currently restrict the
options for large-N empirical studies seeking to explain performance: the
relative lack of valid and reliable performance measures; the limited sets of
such measures available in time series; the typical lack of a common per-
formance metric across the cases (the “apples and oranges” problem); the
restricted number of data sets that contain multiple performance measures,
despite the fact that public organizations are virtually never asked to do one
and only one thing; and the difficulties plaguing such available, putative
performance measures as the federal government’s Program Assessment
Rating Tool. Consequently, we certainly cannot say that the findings
developed here hold for all public organizations; but they are built on a
better-developed and sounder empirical base than has been the case for
other research programs in the field.
The model we have developed from the case study literature reduces the
innumerable complications and subtleties of the practical world to a few
clusters of variables: management; stabilizing features in and of public
organizations; external forces that work as opportunities or constraints on
programs; and, of course, performance itself. This simplified model is not
without its complexities. It considers the performance of public organiza-
tions, essentially, as inertial – and yet as subject to influences, even in the
short term, including by management. It treats management as a multifunc-
tional enterprise and indicates that different managerial functions have
different – and nonlinear – impacts on performance. It can be simplified
further by bundling managerial functions into externally oriented and
internally focused management. Moreover, as shown repeatedly throughout
this book, it offers opportunities to test parts of the model through further
simplifications, when these can be justified by the research focus of interest
and the features of the empirical setting.
Managers, the model proposes, work in the organization’s environment to
draw in resources and take advantage of opportunities for the agency and its
mandated programs – while also protecting the core organizational tasks
from disruption triggered by outside jolts or shocks. Managers also, the
model asserts, perform multiple tasks to encourage and support the internal
production of the organization. While the model does not explicitly empha-
size this point, it also implies that these managerial functions include both a
quantity element (how much effort and energy do managers invest in
various functions?) and a quality element (how good are the managers at
269 What does the evidence show?

doing what they do in seeking to shape performance?). Public organizations

also reap performance advantages from stability, the model further claims,
even though today’s popular rhetoric extols change and reinvention. The
model admits that change can boost performance, but it emphasizes the
underheralded virtues of stability. Finally, the model is also framed explicitly
in an open-systems fashion so that the world external to the public organiza-
tion offers both prospects for improvement and threats to the delivery of
policy results.
The model therefore incorporates a number of assertions and assump-
tions, some of which seem almost obvious, others of which are certainly
somewhat speculative. The important question is not how the model is built,
but what the evidence actually shows. In answer to this question, we can say:
“Many things.” In most respects the model holds up surprisingly well to
systematic empirical investigation. On a set of issues the jury is still out.
Finally, on a few points, the model may need to be revised, at least for certain
sorts of practical settings.
In summarizing here what we know about public management and
performance from our empirical studies, we begin with the second, or
environmental, term of the model, and focus especially on managerial
networking. Do public managers in fact devote substantial effort to interact
with a variety of external actors – do they network? The answer is clearly
“Yes,” at least for top managers, and pretty clearly (on the basis of our
analysis of data from English local authorities and police departments) for
other managers at subordinate levels. We also know that the variation in the
extent of networking by managers – even those in the same positions in the
same sorts of organizations – is considerable. In work not included in this
book, we have explored whether networking behavior by top managers can
be explained by features of the organization and the environment. Virtually
none of the networking behavior is related to these features of the context;
and, as Chapter 3 shows, we have strong evidence that managers’ networking
patterns are rather stable over time and related to the individual manager
and his or her characteristics.
Can we be more specific about how the managerial networking boosts
performance? Although exploring relationships in hundreds of organiza-
tions across a number of years means not being able to trace the precise
processes through which managers’ networking efforts assist production, we
expect that the interactions serve multiple functions. Managers working in
their networked environment can draw useful information into their organ-
izations (for example, with regard to programs, policies and innovations
270 Public management and performance

worth considering), can negotiate and receive technical assistance, can tap
the efforts and production potential of other organizations in patterns of
coproduction, and can strengthen the organization’s political position with
other actors and institutions. They can also help to fend off threats and
shocks from outside.
As Chapter 3 highlighted, we know that, in this set of public organiza-
tions, networking adds to performance not in a simple linear fashion but by
interacting with selective resources (opportunities) in the environment, thus
leveraging and strengthening the impact of these resources on the outcomes
of interest. In other words, the nonlinear, interactive relationship between
networking and environmental forces (the “X” term of our model) is
supported by the evidence. In a paradoxical finding, we discovered that
those organizations performing especially well are less inertial than those
seemingly more trapped in a mediocre production process.
We also know that it is possible to create a good measure of managerial
networking across many organizations without expending huge amounts of
time and expense in sketching the full networks in which managers may be
situated. Considering such networking as a general behavior and one that is
practically important makes sense: more networking by top managers means
better organizational performance. As one might expect, nevertheless, one
can overdo it. We have shown that the marginal contributions of managerial
networking by top managers decline at higher levels of networking –
although incorporating a consideration of managerial quality (an aspect to
be reviewed shortly) eliminates the diminishing marginal returns.
Although managerial networking improves performance, it does not do so
in an undifferentiated fashion. Because the externally oriented actions of
managers take place in a political environment, the external interactions
influence organizational performance to improve outcomes for the more
advantaged and powerful stakeholders in that environment. Performance is
improved on the most salient performance measure and those that are most
important to advantaged clientele, while there is little or no effect for
outcomes that matter most to the less advantaged. Managerial networking
has real payoffs, but not inevitably and not for everyone.
Although networking by top managers has been carefully investigated in
this book, we have not explored the impact of networking at other levels in
the organization. We know from the English local government data that
substantial networking takes place at different echelons in these governmental
organizations. We have not estimated any performance impacts of network-
ing at different levels within an organization, however.
271 What does the evidence show?

Managerial networking is a set of behaviors. Public organizations and

their programs often also are situated in networks – patterns of interdepend-
ence across organizations that have structural properties. We began to
explore how such networks, or network properties, might help to shape
program outcomes, in Chapter 7. That initial foray into the topic demon-
strated that structural features of intergovernmental networks and network-
ing behavior by top managers both influence educational outcomes. The
structural and behavioral elements also interact with each other, but much
remains to be learned here.
The specific investigations of managerial networking tap the amount of
networking and then explore whether it makes a difference to performance.
In recognition of the obvious fact that management entails a quality dimen-
sion as well as a quantity one, we also developed a measure of managerial
quality to assess the relationship of managerial quality to performance. We
have validated such a measure, based on the portion of top managers’
salaries that is not explained by conventional determinants of salary. We
have argued that this measure is correlated with local school boards’ assess-
ment of the quality exhibited by top management in their districts and have
shown that the measure is positively related to many indicators of perform-
ance. Chapter 4 was devoted to this work. Although we have not developed
this line of research further to additional levels of management, other recent
work shows that a similar measure of middle management quality, also
based on salary data, can be developed and is, separately, related to perform-
ance (Johansen 2008). A salary residual measure cannot be used in tapping
managerial quality in all public organizations, of course, in particular
because constrained salary ranges are often stipulated as a part of civil
service systems. Attention should therefore be devoted to alternative
approaches – including some possibilities based on surveys of decision
makers in public organizations and/or of those responsible for holding
such organizations accountable (for an interesting survey-based approach
developed for and implemented in business organizations, see Bloom and
van Reenen 2007).
With regard to the environmental term of the model, we have also
investigated externally generated shocks and how public organizations
buffer themselves from their sometimes turbulent environment. We exam-
ined two types of jolts or shocks: unexpected and sizable budget cuts
and the displacement of thousands of students due to two hurricanes –
natural disasters that were consequential for public education and many
other functions as well. We have also been exploring environmental jolts
272 Public management and performance

of another sort in a different empirical setting: unexpectedly large influxes of

eastern European immigrants into England, and the consequent strains on
the systems there for delivering public services (see, for instance, Andrews
et al. 2009, 2010b).
All these analyses, including the one focusing on English local govern-
ment, show similar patterns with respect to performance. Negative shocks,
whether budgetary or otherwise, impede performance; when the shock
penetrates the organization, though, managers are able to operate internally
to reduce the performance-related hit. Chapter 5 provided extensive
evidence in this regard, particularly concerning subtle adjustments in
budgetary and staffing patterns that focus the organization’s attention on
its core business. (We noted there, and repeat here, that such efforts over the
short term could carry negative consequences for other performance object-
ives, especially if such efforts are repeatedly called upon to mitigate shocks.
We observed the irony that, to the extent that managers are successful in
such efforts, they may weaken the political argument and coalition for
budgetary assistance going forward.) These sorts of moves mitigate the
performance-related effects of shocks and are an aspect of the internal
management of the organization. Internal management comprises a host
of possible functions. In this volume we have explored a number of these,
but substantial additional work will be required to demonstrate fully the
manifold ways internal management can shape outcomes.
Although managers can work to protect the core production processes
from unwanted disruption, they cannot always eliminate the negative per-
formance impacts of shocks from the external world. In such cases, our
studies of three kinds of shocks – unexpected budget cuts, the unexpected
arrival of needy hurricane refugees who also need to be educated, and the
arrival in England of immigrants from eastern Europe in much greater than
anticipated numbers – show an identical, and important, finding: manage-
ment capacity, measured as the relative size of the central office staff, can
mitigate the negative performance impact of such jolts to the system.
Why does management capacity help? Our earlier discussion in Chapter 6
offered a number of clues, but we cannot be completely sure without
additional research. The strongest arguments on behalf of more capacity
seem to focus on the ability to analyze nonroutine situations, such as jolts
from the environment, and to deploy human resources to address the most
pressing or fundamental challenges presented by those circumstances. We
have also been able to determine, as the evidence in the same chapter shows,
that more capacity does more than “merely” wipe out the untoward impacts
273 What does the evidence show?

of shocks. It also enhances the performance-related value of networking by

managers themselves. Here too we see the value of capacity primarily in
helping managers sort through the multiple and sometimes complex
“games” in which they must or might engage externally. Thus analytical
ability and the possible delivery of results and resources valued by external
parties enhance the simple value of networking in the environment.
Managerial capacity offers much of value to public organizations, but the
issue is two-pronged. It may be good to keep on hand substantial managerial
capacity, in case lightning (or hurricanes, or budget disaster, or huge influxes
of immigrants) strikes. The opportunity costs must also be considered,
however. A practical, and complicated, question is how to assess the relative
risks of being overwhelmed by the need for additional management help
versus the real financial (and perhaps other) costs of carrying additional
trained managerial professionals to deal with unexpected contingencies.
Different organizations, situated in different contexts, should assess the
relative risks differently. A practical issue, therefore, is how prone a public
organization is to disruptions or unpleasant shocks that are substantially out
of the ordinary. If the scenario is a rare one, perhaps it makes sense simply to
ride it out when it occurs. If, on the other hand, shocks are likely to occur
with a fair amount of frequency, building in the advantages of managerial
capacity (and thus slack) might make sense. Our advice to public managers
is direct, even if nuanced: make a determination about the amount of
managerial capacity to develop and maintain on the basis of the expectations
and needs facing your own public organization.
We have been discussing shocks and managerial capacity here as if it were
inevitable that, once turbulence buffets the organizational system, it pene-
trates and forces managerial (and other) responses – otherwise disruption
happens. We have also seen, however, that the buffering of shocks from the
environment can keep disruption away from the internal workings of public
organizations. We have treated buffering, a venerable concept in organiza-
tion theory, in terms of the denominator of the environmental term of our
model (SeM4) – a combination of stability-inducing features along with the
protective efforts of managers who interact with forces in the environment
to keep them out of the organization even before they penetrate the system
of production. Unfortunately, our research agenda has not yet progressed to
the point at which we have been able to measure – separately – the stabilizing
organizational elements and the protective moves of managers. The com-
bined buffering term was the subject of some work reported in Chapter 6,
however. Clearly, buffering helps performance. What is less clear is whether
274 Public management and performance

the relationship is nonlinear with respect to the environmental shock itself.

In our initial efforts, we concluded that the interactive effect we hypothe-
sized does not add to the explanatory power of the model, but a more
definitive conclusion awaits more extensive testing.
In short, therefore, we have explored rather extensively – even if not
definitively – the second term of our model. We have shown that an overall
measure of managerial quality also generates positive outcomes. What of the
first term – the portion of the model focused on internal management and
its purported relationship to performance? This book contains some find-
ings in this respect as well. Indeed, in this section we have already mentioned
one such finding: that managers can adjust budgets and staffing patterns to
reduce or remove disruptions emanating from outside the organizations.
The book has offered additional evidence as well, evidence centering
on human resources and their effective management. Stability in staffing
patterns – of top managers and especially of front-line professionals in
school districts – contributes positively to educational outcomes. This set
of patterns can be interpreted as partial validation of the stability-related, or
“S,” term of our model. Personnel stability is also a function of HR manage-
ment, however, and is therefore a partial manifestation of the beneficial
contribution of internal management. As additional research ensues, it
would be useful to try to estimate the impact of other elements of the
stabilizing vector. In the case of school districts, curricular stability and
stability in procedures could be of particular interest.
More generally, an important line of work that has yet to be explored has
to do with the structural stability of administrative systems. The Texas school
districts data set offers many advantages, as we have explained, but it also
has its limits. One of these is the link between structure and performance.
School districts differ in many ways, but they are structurally rather similar;
they also tend toward structural stability over time. Ideally, therefore, future
empirical work in settings offering both cross-sectional structural variation
and some structural reorganizations over time can explore this important
aspect of the “S” vector.
The management of organizations’ human capital also contributes to
performance, as still more evidence presented in Chapter 5 demonstrates –
thus validating a proposition at the heart of the specialty of public HR
management and bringing to the fore the practical importance of govern-
ments addressing what has been called the human capital “crisis” on the
horizon. As we indicated at the outset of this volume, internal management
encompasses a huge number of options and functions, often summarized by
275 What does the evidence show?

Table 8.1 Intercorrelations of management measures

“M” measures Quality Mnetworking Mstability Tstability Buffering1 Capacity

Mnetworking 0.01 x x x x x
Mstability 0.03 0.07 x x x x
Tstabililty 0.02 0.05 0.00 x x x
Buffering 0.05 0.05 0.05 0.05 x x
Capacity 0.10 0.04 0.04 0.12 0.27 x
Human resources M 0.07 0.05 0.10 0.12 0.04 0.002

Notes: 1 Technically, this term measures M4Se, the denominator of the second term in the model. It includes
a management component, and so we include it here.

the venerable term POSDCORB but sometimes extending beyond even

this set of activities. Our work has examined only a modest fraction of
these. Thus another part of the research agenda is to expand empirical
analyses to estimate the performance-relevant impacts of other internal
management efforts.
Overall, we have shown numerous measurable and practically important
influences of public management on policy outcomes. Nevertheless, we
should avoid the temptation to embrace managerialism – the assumption
that better management is a cure-all for a wide range of economic and social
problems (see Pollitt 1990: 1). Numerous other influences need to be taken
into account, and, indeed, we have already written another book about the
interaction between managers and the broader political system (Meier and
O’Toole 2006). Managers are not superheroes, even if their intention and
efforts are often noble and heroic; but neither should we underestimate the
effect that management can have. Even taking into account only those
performance results that we have thus far been able to link to one or other
aspect of management, and even omitting the obviously relevant perform-
ance impacts that middle management is almost certain to have, we estimate
roughly that we can attribute approximately 20 percent of the overall
variance in performance across organizations to the effects of top manage-
ment (see Meier and O’Toole 2009b). This is hardly a trivial quantity.
Equally interesting is the pattern of managerial influence across the several
different aspects of management. Table 8.1 shows the correlation matrix
between the six different management measures. Of the twenty-one inter-
correlations, all but one are at 0.12 or lower; and the one exception is also
exceedingly modest (0.27 between buffering and management capacity).
For practical purposes almost all the measures are virtually uncorrelated.
276 Public management and performance

Table 8.2 Practical lessons for managers

(1) Managerial networking can lead to performance gains either through the acquisition of
technical knowledge or the development of political/public support.
(2) Managerial networking can interact with some organizational resources to increase their
impact several-fold. Given this relationship, managers need to exercise choice with
regard to how to pursue environmental opportunities.
(3) Managerial networking is subject to diminishing marginal returns; the best managers
limit their networking activities before reaching this point.
(4) Managerial networking has distributive consequences. Network demands are more likely
to come from well-established and well-endowed network nodes.
(5) The impact of managerial networking depends on the structural context. Managerial
networking is more valuable in structural networks.
(6) Management needs to build excess managerial capacity to deal with shocks and other
unexpected environment problems.
(7) Management capacity enhances the impact of managerial networking.
(8) Managers need to recognize that building managerial capacity results in tradeoffs. In
particular, investments in managerial capacity can limit current levels of production.
(9) Loosely coupled organizations generally perform better; this is especially the case with
regard to disadvantaged clientele.
(10) The stability of front-line workers leads to higher productivity. This relationship is
stronger for programs that affect disadvantaged clientele.
(11) Managerial stability is positively related to performance.
(12) The development of human capital is the management activity with potentially the
largest payoff for performance.
(13) Management actions are generally independent of each other, so that improvements in
one area are not necessarily related to management abilities in other areas.

This pattern is interesting, for at least two reasons. First, it suggests that the
contributions that various management functions make to performance can
be treated basically as additive. Second, the set of (non)relationships shows
that “good management” comes in many flavors – or, more precisely, that
management is clearly a multifunctional enterprise. Managers who excel at
one aspect are no more (or less) likely to do well at another. These results are
intriguing and call for more sustained analysis both within and among these
and other managerial functions.
Although our purpose has been to contribute to the scholarly literature on
public management rather than provide advice to practicing public man-
agers, we are aware that our work has significant implications for how
organizations should be managed. Under the guise of evidence-based public
management, we have discussed this issue in some detail (see Meier and
O’Toole 2009b). Table 8.2 provides a summary of findings that we think
have practical application to the real world of public management. These
277 What remains to be explored?

should be considered no more than tentative, because we have not worked

with public organizations to apply these lessons systematically and measure
the results. At the same time, much of the list accords with what are con-
sidered to be good management practices in the literature (see Rainey 2009).
For all that we have learned, nonetheless, there is a great deal that remains
to be done. As we bring the present study to a conclusion, therefore, it is
appropriate to indicate some of what still lies before the research community.

What remains to be explored?

Even as we have summarized the research findings in this book, we have

indicated additional avenues to be explored. In particular, we note that this
research agenda needs to move to venues beyond school districts in Texas;
should encompass managerial influences below as well as including the top
echelon of public organizations; would, ideally, explore the multiple influences
of management in settings more thoroughly networked than public education;
could explore additional measures of managerial quality that could be applied
in more conventional civil service systems; might reach to include additional
aspects of internal management; should address other terms of the model that
have not been systematically analyzed thus far; and would focus as well on the
functional form of several of the relationships hypothesized in that model to
clarify and develop a better understanding of the causal patterns.
All these items on the agenda were mentioned earlier, but a few additional
comments about some of them can be helpful here. With regard to the heavy
use of the Texas school districts data set, we explained earlier its unique value
and its enhanced utility over time as additional validated measures
have been developed. We should note that we have been exploring the
management-and-performance agenda in other venues as well. Thus far,
we have developed findings from English local governments, as alluded to in
this book, and in state unemployment insurance programs. We have initi-
ated collaborations with public management researchers in two other
European countries to explore the impact of management on performance
in public education in these settings as well. The findings there should help
to test further the relationships we have found in Texas school districts and
also broaden the research agenda to include additional matters. This new
work, for example, is designed to tap managerial influences at more than one
level, and thus below the top echelon. This sort of cross-national work also
can link public management studies to their broader governance structures.
The United States situates management in the midst of pluralist governance
278 Public management and performance

arrangements, but some of the relationships might operate differently in, for
instance, more corporatist settings. Some of the emergent work, therefore,
offers the potential to explore such themes of governance, which have been
considered important but essentially unverified in recent work in the field.
In addition to these ongoing projects, we are initiating work on the man-
agement of hospitals and nursing homes. Both data sets include public,
private, and nonprofit organizations.
We have also expanded the Texas schools data set by inviting other scholars
to nominate survey items for inclusion in the superintendent surveys. Several
such questions on performance management, social capital, trust, and diver-
sity were included in the 2009 survey. We are also at the early stages of
examining data on charter schools. We have collected data on these over time
and are now working with a colleague to examine public management in these
organizations, which effectively operate as nonprofit organizations subject to
some state regulation and funding. The analyses there may speak to the
themes developed in the research literature on nonprofit organizations.
What is likely to be more challenging by far is any effort to estimate the
full range of managerial influences in more thoroughly networked settings.
As explained in Chapter 2, we expect public management in networks to
be perhaps even more consequential than in more traditional settings
(O’Toole 2000b), but practical impediments make it especially difficult to
conduct large-N empirical explorations of the question. Modeling managerial
impacts in networks requires adding more management functions and
vectors of influence, managers themselves may have strategic reasons to
disguise their own efforts to shape results in such settings, multiple managers
in very different organizations may simultaneously be attempting to move
performance along (albeit not necessarily in the same direction), and the
structural dimensions of the program setting are less transparent and may
need to be mapped with significantly more labor-intensive research efforts.
This agenda item remains important, but progress is likely to be slow.
We are more optimistic about the agenda with regard to internal manage-
ment and performance. Thus far we have triangulated around the internal
management function but not specified it as completely as would be ideal.
A number of additional managerial efforts that are focused internally can be
examined in principle so long as the data are gathered systematically from
managers or other sources. Certain additional managerial functions have
already been the focus of research effort on our part. Along with colleagues,
we have explored whether the content of managers’ (and organizations’)
strategy is related to performance in the ways hypothesized by some generic
279 Politics and management

management theorists (see Meier et al. 2007 and Andrews et al. 2010a). Strategy
can be considered an aspect of management that entails both internal and
external attention. Other management efforts that are directed primarily at
internal coordination, planning, resource allocation, communication, and so
forth are all – in principle – amenable to empirical analysis. The research agenda
going forward needs to include such studies if we are to reach a fuller under-
standing of the management–performance relationship. Some of the points just
mentioned, along with a set of additional ones, can be unpacked more fully in
this coverage of the research agenda that remains to be addressed.

Governance structures

The broader set of institutional arrangements for public managerial efforts

needs to be explored in any thoroughgoing examination (see, for instance,
Lynn, Heinrich, and Hill 2001). In this volume we have analyzed a range of
managerial influences, but we have done so while largely ignoring the
system’s larger governance structure. For the most part we have examined
independent agencies within a decentralized governance system in a frag-
mented political system. All these characteristics of governance systems vary
both within nations and between them. Similarly, the political context also
needs to be taken into account, as does the larger network in which many
public organizations conduct their efforts. Further, multiple levels in gov-
ernance arrangements can also shape policy results, and the interplay among
these is clearly worth systematic investigation (for one such effort, see Meier,
O’Toole, and Nicholson-Crotty 2004). Combining all such elements in a
complex governance structure and exploring their interactive effects across
many cases poses a formidable challenge – and especially so if the variation
in governance structures cross-nationally in quite different systems is taken
seriously. Analyzing such patterns for their performance-related effects is
necessary for a full understanding of the role management plays in imple-
menting public programs.

Politics and management

The study of governance structures brings to the forefront the relation-

ship between politics and public managerial systems – an understudied
topic. Public programs are governed by a mix of career administrators
280 Public management and performance

and political actors. Several aspects of these hybrid governance systems

should be of particular interest.
First, how public organizations and their managers interact with political
actors and institutions deserves much more careful attention. Most of the
literature on this subject is qualitative and of a case study nature. A more
quantitative literature focuses only on simple principal–agent models. These
literatures are interesting and provocative but necessarily limited in what
they can offer. Large-N studies of this topic that recognize the complexity of
the political–administrative interface are clearly in order. Second, and in a
related vein, the relative importance of political and administrative systems
for performance needs much more attention. In work that has not been
covered in this book, we have sought to determine what happens when
political bodies fail to represent the broader public (Meier and O’Toole
2004b). In the empirical settings we have examined, managerial influence
expands when political institutions do an inadequate job at their political
task. We believe that this pattern is likely to be a more general one, but the
research on it has only just begun.
Third, how public organizations and their managers marshal public
support is also important and deserving of similar careful, quantitative
examination. Finally, it may be time for public management researchers to
think big – even imperialistically. Administrative agencies are not the only
institutions in a political system that require management. The judiciary,
legislatures, and other relevant political bodies also present appropriate
venues for the systematic study of public management. These bodies pro-
duce outputs and outcomes, and they are not anarchies. There is no good
reason to restrict the systematic study of public management and performance
to bureaucratic bodies alone (Lynn 1996). One of us has already proposed that
public management as a research field should expand in hegemonic fashion to
cover the institutions traditionally included in the embrace of political science
(Meier 2007; see also Vaughn and Villalobos 2009).

Social capital

A topic that can be mentioned in conjunction with more detailed coverage

of governance structures is that of social capital. This is not the place for
anything more than a cursory mention of this subject, but it should be
clear from work done on nonprofit organizations, as well as some of the
contributions in political science, that social capital can offer a number of
281 Social capital

advantages, including the enhancement of what government organizations

themselves can accomplish (Putnam 1993, 2000). The research literature of
public management has paid little attention thus far to how social capital
helps to shape program outcomes, however.
Specific kinds of research questions present themselves. One has to do
with whether and how the general level of community organization might
leverage enhanced public program performance – for instance, can social
capital help to blunt the negative performance consequences that govern-
ments might otherwise experience when public organizations are faced
with the kinds of environmental shocks we have examined in this book?
We have begun to explore this question, and plan to address it more fully
in the future (Andrews et al. 2009). Another is whether managers of
public organizations can place themselves appropriately in their net-
worked environment so as to take advantage of key external parties and
organizations to enhance public program performance. The theory of
structural holes offers potential as a source of testable propositions
worthy of careful empirical testing in this regard (see, for instance, Burt
1992, 1997). Here too we have begun some systematic analysis (Andrews
et al. forthcoming (a)), but the field as a whole should be encouraged to
explore such questions in considerably more depth.
Yet another potentially important topic, related to the role of social capital
in public management, is how trust – sometimes considered an important
aspect of social capital more generally – operates in social systems to
facilitate exchanges between coproducing actors, enhances community sup-
port for the production of public services, reduces actors’ discount rates into
the future to achieve benefits for which the main cost must be borne in
the short term, improves flows of accurate information, and reduces coord-
ination costs among interdependent organizations and individuals (see
Edelenbos and Klijn 2007). Additional work along these lines, clearly related
to the theme of social capital, should also be revealing.
The literature on social capital distinguishes between bonding and
bridging social capital. Bonding social capital occurs in homogeneous
groups and might have negative consequences for effective governance,
since it may be correlated with hostility across groups in society. Bridging
social capital, in contrast, consists of those bonds across groups that tie
individuals to more diverse individuals in other groups. Bridging social
capital is perceived to make positive contributions to society and govern-
ance. Unexplored in the management literature is what public management
does to build both bridging and bonding social capital. Any public program
282 Public management and performance

that relies on citizen production also seeks to build social capital in the
community. The role that public management plays in fostering social
capital, particularly bridging social capital, could well be one of manage-
ment’s more important contributions to democracy.

M3 exploiting the environment

Our research program has made progress in examining several of the

relationships hypothesized in our model. We have not, however, made
progress in examining managers’ explicit efforts to tap or exploit opportun-
ities or resources in the environment. This managerial function is labeled M3
in our model. We have hypothesized that such managerial exploitation is
likely to enhance outcomes and is likely to interact with at least some of the
environmental (or “X”) forces rather than merely add in a linear fashion to
performance. At this point, however, we have not developed a defensible
measure of managerial exploitation itself (apart from its inclusion in the
broader networking measure) and are unable to document how such efforts
actually operate to shape performance. This topic too therefore remains on
the agenda and should receive attention in the future.1

Performance, writ more broadly

One theme emphasized in this book has been that public organizations are
typically charged with multiple goals and, therefore, should be assessed via
multiple performance measures. We have done so in many of the analyses
presented in earlier chapters, and we have seen that the patterns revealed
across a full array of performance measures can be quite telling. This cannot
be the end of the story, though. In all the measures used in this study,
performance is interpreted implicitly as quantity of outputs or outcomes
over some designated time period, but additional ways of conceptualizing
performance should be considered as well.
As Boyne has pointed out (2003), performance can be considered along a
number of additional dimensions. Two of the most important are efficiency
(the ratio of outputs or outcomes to financial inputs) and equity (the
“fairness of the distribution of service costs and benefits between different
groups” (Boyne 2003: 368)). In this book we have ignored the subject of
efficiency, treating financial inputs as control variables as we have sought to
283 Another M?

estimate managerial effects on service outputs and outcomes.2 Furthermore,

while we have explored an aspect of equity in our consideration of whether
networking by top managers produces benefits skewed toward the most
advantaged clientele, we have not gone into this theme in depth. For
instance, we have ignored the distribution of service costs across social
groups. Efficiency and equity have been persistently important themes in
the practical world of public management and have been recognized by
researchers as central as well. Still, as Boyne documents, their empirical
examination in larger-N studies of performance has been rare. It is time
for the studies of performance and public management to begin to address
these additional subjects as well. A range of thus far untapped dependent
variables can be the focus of research attention.
Even Boyne’s range of indicators might be perceived as too narrow from
one perspective. Many feel that education should not be narrowly defined in
terms of test scores, but should also include broader elements of producing
effective democratic citizens (Smith 2003). Similar performance goals can be
envisioned for all government agencies, as they provide fora for citizen
development (Cook 1996).
This study has relied on performance indicators established by external
political actors. It differs from some analyses on this dimension by avoiding
the use of managers’ own perceptions of how well the organization is doing
(a relatively common practice in the public management literature).
Although we find the relationship between internal perceptual measures of
performance and external objective measures of performance interesting and
plan to investigate this link further (see Meier and O’Toole 2010), the real
world of governance eschews self-assessments of performance and therefore
operates consistently with the emphasis in this book.

Another M?

In its formalization of the role of internal management in shaping perform-

ance, our model emphasizes the point that public management can
reinforce, prop up, or stabilize the ongoing production process of organiza-
tions. We have observed in this book that there are many ways of doing so,
and our empirical treatment of this function has investigated a few of these.
The model makes no mention of another potential source of managerial
influences on performance, however: efforts by management to disrupt
284 Public management and performance

internal routines and production processes in the interest not of reinforcing

current efforts but changing and improving them.
We have downplayed this potential aspect of management in our studies,
in part because we have been skeptical of the frequent claims that innovation
and entrepreneurial public management are the sole routes to high perform-
ance. Some well-established themes in the research literature as well as in
practice, however, strongly emphasize organizational change and develop-
ment. Managerial efforts to alter rather than reinforce the inertial produc-
tion systems of public organizations are certainly worthy of systematic study.
Indeed, we have noted in this volume that the more highly performing
portions of our sample of school districts are also less autoregressive than
those doing less well. Therefore, another item on the research agenda, going
forward, is to investigate managerial efforts at organizational change and the
reform of extant processes to explore performance-related effects. It might
be the case, for instance, that the results of such managerial efforts are
negative in the short term but sometimes positive over the longer haul.

Extending to other sectors

Our model was developed from the literature on public management, which
is based largely on case studies, and has been used to explore relationships in
public organizations. Nothing in the model is restricted only to traditional
government-owned organizations, however. We have already mentioned the
potential applicability to nonprofit organizations. In addition, the same
might be said with regard to private, for-profit organizations.
Here matters are not quite so simple, because “publicness” and “private-
ness” can be considered on more than one dimension, not simply ownership
(see Perry and Rainey 1988). In particular, the source of funding and the
form of social control also constitute criteria for distinguishing the sectors.
The public–private distinction might also be considered in terms of multiple
categories or via a continuum rather than a dichotomy. Bozeman (1987), for
example, has even argued that “all organizations are public” – at least in
some respects.
The research literature has not sorted out the question of whether pub-
licness, in whatever form, renders the managerial functions distinctive with
respect to determinants of organizational performance. To the extent that
management can be considered a generic function, an approach typically
taken by business management researchers, a model and set of findings valid
285 Better measures, more data

for public organizations should presumably be generalizable to organiza-

tions more broadly. Accordingly, one approach would be to test the various
hypotheses implicit in the model in empirical work on, for instance, private
and for-profit organizations. If the patterns and results are similar, such
findings would strengthen the argument for generic management. If they are
different in consistent ways, such findings would inductively build the
argument for distinctive theoretical spheres – or, at least, for treating pub-
licness as a theoretically important variable.
Another approach, however, would be to use the currently available
research literature on public and private organizations to inform a consider-
ation of how a general model – for instance, the one we have put forward in
this book – might be affected by sector with regard to how performance is
shaped. So, for example, should we expect a given quantity of the internal
management function (M1) to have a larger or smaller effect on performance
in more public or more private organizations? The current set of research
findings on publicness can be used to develop a set of testable hypotheses.
We have made an initial theoretical effort along these lines (see Meier and
O’Toole forthcoming (b)), although the argument has not been included in
this volume. The validity of such a logic, of course, awaits empirical testing.

Better measures, more data

As is clear from this book, it is possible to develop valid and reliable

measures of aspects of public management and test their performance effects
with data on hundreds of organizations. As discussion in the current chapter
makes clear, these can be improved in a number of respects. As public
management goes forward, it is important to build additional data sets that
share many of the strengths of the Texas school districts data set and perhaps
add still others – structural variation, for instance, and network characteris-
tics. To those who would criticize the school districts data set for some of its
limitations, we would say, simply, that the answer is to improve on what is
now available; to organize and collect more data on more public organiza-
tions; to do it in more policy fields; to build a longer time series; to do so
with an array of performance measures; and, above all, to gather excellent
data on management. This last kind of item – the behavior, efforts, and skill
of public managers – is the type typically neglected. Even with the current
popularity of the so-called performance movement, governments as well as
researchers have little solid information about what managers do and
286 Public management and performance

how well they operate as they tackle their responsibilities. While calling
for better measures and better data is almost a cliché, the fact is that
we will not make much more progress in understanding the relationship
between public management and performance without significant develop-
ments on both fronts.
There is much more work to be done, therefore, if we are to implement
fully the sort of research agenda that can demonstrate the entire range of
managerial impacts on the outputs and outcomes of public organizations.
It will take time, and many researchers, and undoubtedly some false starts.
At this point in the research program, we have begun the task, and a
number of other researchers have joined the effort, often by expanding
the range of questions that can be addressed with the data at hand and
sometimes by identifying new and promising sources – whether it be the
influence of diversity and representation on performance (Pitts 2005),
organizational leadership and its effects (Fernández 2005), gender and
performance (Johansen 2007), mayoral influences on results in Colombia
(Avellañeda 2009b), or managerial effects in state governments (Donahue
et al. 2004), to mention but a sample of the recent work. The questions are
large and important, the need for answers pressing, and the results thus far
promising. We welcome many more to join the challenge.

NOTES

1. What we have explored is managerial strategy, particularly the notion of prospecting,

which could be very similar to efforts to exploit the environment (see Meier et al. 2007).
Prospecting does not appear to have much impact in our analyses, thus suggesting that the
effort to exploit environmental opportunities is something different from a prospecting
strategy.
2. In some cases efficiency is clearly a secondary goal. The school districts in our study seek
to maximize outputs relative to the resources they are allocated. In this process they might
go beyond the most efficient point of production and get diminishing returns. Because
districts are evaluated on their test scores, not on whether they attained the scores in the
most efficient manner, they have an incentive to spend all their resources in order to attain
the maximum performance possible.
 Glossary

Above criterion is a US college board score of more than 1,110 on the SAT or its ACT
equivalent. This is, essentially, equal to a score that would rank a student in the top 20 percent
nationally. The Texas Education Agency defines this score as one designating college-ready
students.

ACT is a college aptitude examination, formerly known as the American College Test, that high
school students take. The exam is used primarily by Midwestern colleges and universities.
All-pass rate is the percentage of students in a district who have passed all portions of the
TAAS or the TAKS exam for that year.
Anglo pass rate is the percentage of Anglo students (white non-Latinos) in a district who have
passed all portions of the TAAS or TAKS for that year.
Attendance is a performance indicator that is the percentage of all students in a district who
attend classes on average – i.e. the average of the average daily attendance percentages.
Autoregressive systems are inertial systems. In an autoregressive system, current outputs can
be expected to be strongly influenced by past outputs. Formal organizations are typically
designed to be autoregressive. See also Outputs.

Black pass rate is the percentage of black students in a district who have passed all portions
of the TAAS or TAKS exam for that year.

Buffering (SM4) is a measure of how loosely coupled the organization is, consistent with
the theoretical notion that a loosely coupled organization is more able to take a shock to the
system and dissipate it rapidly. This measure is based on the correlation of the organization’s
outputs from 1988 to 1994 (a time period before the current study). That correlation
subtracted from 1.0 is the measure of buffering used.

Bureaucracy is a type of formal organization with certain distinctive features, including a

hierarchy of superior–subordinate relations, the appointment of experts on the basis of merit
criteria, fixed and limited jurisdictions, decision making on the basis of rules, and reliance on
written records.
Capacity; see Managerial capacity.

College boards are aptitude exams that are given to high school students to measure their
readiness for college. See ACT, SAT.

287
288 Glossary

College-ready; see Above criterion.

Dropouts are students who leave school without completing their degree program. The state
of Texas has two dropout measures: the four-year dropout measure and the six-year dropout
measure. The difference is in whether the base measure is calculated over the last four years of
school or the last six. Dropouts are only rarely used as performance indicators in this book,
because they are conceded to contain substantial measurement error.

Governance, in today’s parlance, refers to the full set of formal institutions and informal ones,
along with the associated patterns of action that produce policy results. A governance system
for a particular policy or program may involve an organization such as a public agency or a
large and complex network of organizations and actors. Governance systems may include
public, for-profit, and nonprofit units.
Human resources management; see Internal management.

Independent school district is a US governance structure for school districts. These are
essentially special-purpose units that are created for the express purpose of operating public
schools. School districts generally do not have boundaries that are coterminous with other
jurisdictions – even those that share the same name. The “independent” portion of the title
means that the school district has an independently elected governing board and has the
authority to levy taxes for the support of public education. Except for a single municipal
school district, all districts in Texas are governed as independent school districts.
Internal management (M1) is management’s efforts to manage inside the organization. We
measure internal management with five items: the superintendent’s assessment of the quality
of the principals’ management skills, experienced teachers, professional development (on a
scale of 5 ¼ excellent to 1 ¼ inadequate), and whether the superintendent agreed or disagreed
on a four-point scale with two other items: “With the people I have in this organization, we
can make virtually any program work” and “I am quite likely to recommend a subordinate
for a superintendent position in another district.” The measure is a factor analysis of these
items, resulting in a single factor.

Latino pass rate is the percentage of Latino students in a district who have passed all portions
of the TAAS or TAKS exam for that year.
Low-income pass rate is the percentage of low-income students in a district who have passed
all portions of the TAAS or TAKS for that year.
Low-income student is a student who is eligible for free or reduced price school lunches. The
exact criteria are based on federal poverty levels for a given size of household.
M1; see Internal management.

M2; see Managerial networking.

Managerial capacity is the percentage of all employees who are located in central office
administration.
Managerial networking (M2) is a measure of how frequently top managers interact with key
actors in the environment. Superintendents are asked how frequently they interact with a set
of individuals on a scale of daily (6) to never (1). The actual score is a factor analysis of the set
289 Glossary

of external actors. The initial measure included local business leaders, state legislators, other
superintendents, the Texas Education Agency, and school board members. Later measures
omitted contact with the school board and treated that as a separate measure (often termed
“managing upward”). The factor analysis consistently generates only a single significant
factor, with all indicators loading positively on that factor. An increase in the number of
external actors appears to have little impact on the measurement of this factor. Managerial
networking is also the measure of managing outward.
Managerial quality is a measure that is based on a regression equation predicting the salary of
the school superintendent. Essentially, the strategy (see Chapter 4) takes all the factors that
should predict a superintendent’s salary (district size, past performance, age, etc.) and uses
these as independent variables in a regression. The residual from this equation – that is, the
part that cannot be explained by known factors – is taken as a judgment by the school board
as to the quality of the manager.

Managerial stability (Sm) is the number of years the superintendent has been employed by the
district in any capacity.
Managing downward is a measure of how frequently the superintendent interacts with school
principals, on a scale of daily (6) to never (1).
Managing outward; see Managerial networking.
Managing upward is a measure of how frequently the superintendent interacts with members
of the school board, on a scale of daily (6) to never (1).

Networking; see Managerial networking.

Networks are structures of interdependence involving multiple organizations or parts thereof,
in which one unit is not merely the formal subordinate of the others in some larger
hierarchical arrangement.
Outcomes of an organization or program refer to the eventual impact of policy actions, along
with the results of other causal variables, on the ultimate issue or concern prompting the
initial policy intervention. See also Outputs.
Outputs of an organization or program are the immediate consequences of policy and
management efforts, such as bridges built, cases processed, environmental permits issued.
See also Outcomes.
Percentage tested is the percentage of students who take either the ACT or the SAT. This
measure is a rough indicator of students going on to college, since a student who takes neither
exam is unlikely to attend college.
Performance of an organization or public program is the achievement of such organizations
and programs in terms of the outputs and outcomes that they produce. There can be
numerous measures of performance, and the concept can be considered to have a number
of dimensions, including efficiency, effectiveness, equity, and public satisfaction.
Personnel stability (Sp) is a measure of the stability of teaching personnel. It is the percentage
of the teachers last year who were employed by the district at the start of the current year.
This calculation avoids imputing turnover to districts that are rapidly growing.
290 Glossary

Principal is the title given to the individual who oversees a school within a school district.
Principals can be thought of as middle managers who oversee the production personnel
(the teachers) of the districts.
Public management is the coordination of people and resources toward the accomplishment
of collective purpose; public management also involves tapping the interdependent organiza-
tional environment in support of such purpose and to protect the organization’s efforts from
potential disturbances.
Sm; see Managerial stability.

SM4; see Buffering.

Sp; see Personnel stability.
SAT is a college aptitude examination, formerly known as the Scholastic Aptitude Test, that is
used by most colleges and universities for admissions decisions. The test ranges from 200 to
800 for three tests – math, verbal, and analytical. The school districts generally report the sum
of math and verbal scores.
School boards are the governing body of US school districts. In Texas, school boards are
elected via a nonpartisan ballot. The board has the authority to set general education policy,
to levy taxes, and to hire a superintendent to operate the school.
Stability; see Managerial stability, Personnel stability.

Superintendent is the chief operating officer of the school district. The superintendent is
appointed by the school board, generally for a fixed-term contract.
TAAS is the Texas Assessment of Academic Skills, a standardized test administered to Texas
students until 2003. The test was administered to different grades of students in different
years (with a general expansion of the number of grades). There were also specialized tests
given to assess knowledge gained from specific courses – e.g. an end of algebra exam. This
evolved into a high-stakes test that students had to pass in order to receive a regular diploma
from the state of Texas. The data used are the percentage of students passing all TAAS exams
that they took. See TAKS.
TAKS is the Texas Assessment of Knowledge and Skills, a standardized test used in Texas from
2003 onward. It replaced the TAAS, and, although it is more difficult than the TAKS, the
scores under it are highly correlated. See TAAS.

Texas Education Agency is a state agency in Texas that oversees the state’s school districts.
The TEA is also the agency that collects a wide range of performance data on Texas schools
and school districts. The agency has regulatory authority and is the pass-through agency for
funds allocated to the school district from the states.
 References

Agranoff, Robert. 1986. Intergovernmental Management: Human Services Problem-Solving in

Six Metropolitan Areas. Albany, NY: SUNY Press.
2007. Managing within Networks: Adding Value to Public Organizations. Washington, DC:
Georgetown University Press.
Agranoff, Robert, and Michael McGuire. 2003. Collaborative Public Management: New Strategies
for Local Governments. Washington, DC: Georgetown University Press.
Allison, Graham. 1971. Essence of Decision: Explaining the Cuban Missile Crisis. Boston:
Little, Brown.
Altshuler, Alan A., and Robert D. Behn (eds.). 1997. Innovation in American Government:
Challenges, Opportunities, and Dilemmas. Washington, DC: Brookings Institution Press.
Andersen, Simon Calmar, and Peter B. Mortensen. 2010. “Policy stability and organizational
performance: is there a relationship?” Journal of Public Administration Research and
Theory 20, 1: 1–20.
Andrews, Rhys, George A. Boyne, Jennifer Law, and Richard M. Walker. 2005. “External
constraints and public sector performance: the case of Comprehensive Performance
Assessment in English local government.” Public Administration 83, 4: 639–56.
Andrews, Rhys, George A. Boyne, Kenneth J. Meier, Laurence J. O’Toole, Jr., and Richard M.
Walker. 2009. “Immigration, local government capacity and public service performance:
evidence from England.” Working paper. Cardiff University.
2010a. “Alignment and results: testing the interaction effects of strategy, structure, and
environment from Miles and Snow.” Administration and Society 42, 2: 160–92.
2010b. “The micro-politics of European immigration: local government capacity and
public service performance in England.” Working paper. Cardiff University.
2010c. “Wakeup call: strategic management, network alarms and performance.” Public
Administration Review 70, 5: 731–41.
Forthcoming (a). “Environmental and organizational determinants of external networking.”
American Review of Public Administration.
Forthcoming (b). “Vertical strategic alignment and public service performance.” Public
Administration.
Andrews, Rhys, George A. Boyne, and Richard M. Walker. 2006. “Strategy content and organiza-
tional performance: an empirical analysis.” Public Administration Review 66, 1: 52–63.
Appleby, Paul H. 1949. Policy and Administration. University, AL: University of Alabama Press.
Argyris, Chris. 1957. Personality and Organization: The Conflict between System and Individual.
New York: Harper.

291
292 References

Avellañeda, Claudia N. 2009a. “Mayoral quality and local government finance.” Public
Administration Review 69, 3: 469–86.
2009b. “Municipal performance: does mayoral quality matter?” Public Administration
Research and Theory 19, 2: 285–312.
Ban, Carolyn. 1995. How Do Public Managers Manage? Bureaucratic Constraints, Organiza-
tional Culture, and the Potential for Reform. San Francisco: Jossey-Bass.
Ban, Carolyn, and Norma N. Riccucci (eds.). 2002. Public Personnel Management: Current
Concerns, Future Challenges, 3rd edn. New York: Longman.
Bardach, Eugene. 1998. Getting Agencies to Work Together: The Practice and Theory of
Managerial Craftsmanship. Washington, DC: Brookings Institution Press.
Barnard, Chester I. 1938. The Functions of the Executive. Cambridge, MA: Harvard University
Press.
Bartle, John R. 1996. “Coping with cutbacks: city response to aid cuts in New York state.”
State and Local Government Review 28, 1: 38–48.
Barzelay, Michael. 2001. The New Public Management: Improving Research and Policy
Dialogue. Berkeley: University of California Press.
Barzelay, Michael, and Babak J. Armajani. 1992. Breaking through Bureaucracy: A New Vision
for Managing in Government. Berkeley: University of California Press.
Beam, David R., Timothy J. Conlan, and David B. Walker. 1983. “Federalism: the challenge
of conflicting theories and contemporary practice.” In Ada W. Finifter (ed.). Political
Science: The State of the Discipline: 247–82. Washington, DC: American Political Science
Association.
Beam, George. 2001. Quality Public Management: What It Is and How It Can Be Improved and
Adapted. Chicago: Burnham.
Behn, Robert D. 1991. Leadership Counts: Lessons for Public Managers. Cambridge, MA:
Harvard University Press.
2001. Rethinking Democratic Accountability. Washington, DC: Brookings Institution Press.
Bennis, Warren. 1993. An Invented Life: Reflections on Leadership and Change. Reading, MA:
Addison-Wesley.
Bertelli, Anthony M., and Laurence E. Lynn, Jr. 2006. Madison’s Managers: Public Adminis-
tration and the Constitution. Baltimore: Johns Hopkins University Press.
Bilmes, Linda, and Jeffrey Neal. 2003. “The people factor: human resources reform in
government.” In John Donahue and Joseph Nye (eds.). For the People: Can We Fix Public
Service?: 113–33. Washington, DC: Brookings Institution Press.
Bingham, Lisa Blomgren, and Rosemary O’Leary. 2008. Big Ideas in Collaborative Public
Management. Armonk, NY: M. E. Sharpe.
Bloom, Nicholas, and John van Reenen. 2007. “Measuring and explaining management
practices across firms and countries.” Quarterly Journal of Economics 122, 4: 1351–408.
Bogason, Peter, and Theo A. J. Toonen (eds.). 1998. “Comparing networks.” Symposium in
Public Administration 76, 2: 205–407.
Bohte, John. 2004. “Examining the impact of charter schools on performance in traditional
public schools.” Policy Studies Journal 32, 4: 501–20.
Bohte, John, and Kenneth J. Meier. 2000. “Goal displacement: assessing the motivation for
organizational cheating.” Public Administration Review 60, 2: 173–82.
Bollen, Kenneth. 1989. Structural Equations with Latent Variables. New York: Wiley.
293 References

Bowles, Samuel, and Herbert Gintis. 1976. Schooling in Capitalist America: Educational
Reform and the Contradictions of Economic Life. New York: Routledge & Kegan Paul.
Boyd, Donald, Hamilton Lankford, Susanna Loeb, and James H. Wyckoff. 2003. “Analyzing
the determinants of the matching of public school teachers to jobs: estimating compensating
differentials in imperfect labor markets.” Working Paper no. 9878. Cambridge, MA: National
Bureau of Economic Research [NBER] (available at http://ssrn.com/abstract=430592).
Boyne, George A. 2003. “Sources of public service improvement: a critical review and research
agenda.” Journal of Public Administration Research and Theory 13, 3: 367–94.
Boyne, George A., and Alex A. Chen. 2007. “Performance targets and public service improve-
ment.” Journal of Public Administration Research and Theory 17, 3: 455–77.
Boyne, George A., Kenneth J. Meier, Laurence J. O’Toole, Jr., and Richard M. Walker (eds.). 2006.
Public Service Performance: Perspectives on Measurement and Management. Cambridge
University Press.
Boyne, George A., and Richard M.Walker. 2006. “Strategy content and public service organi-
zations.” Journal of Public Administration Research and Theory 14, 2: 231–52.
Bozeman, Barry. 1987. All Organizations Are Public: Bridging Public and Private Organiza-
tional Theories. San Francisco: Jossey-Bass.
Brambor, Thomas, William Roberts Clark, and Matt Golder. 2006. “Understanding inter-
action models: improving empirical analyses.” Political Analysis 14, 1: 63–82.
Bressers, Hans, Laurence J. O’Toole, Jr., and Jeremy Richardson (eds.). 1995. Networks for
Water Policy: A Comparative Perspective. London: Frank Cass.
Bretschneider, Stuart, Frederick J. Marc-Aurele, and Jiannan Wu. 2005. “‘Best practices’
research: a methodological guide for the perplexed.” Journal of Public Administration
Research and Theory 15, 2: 307–23.
Breul, Jonathan D., and Nicole W. Gardner (eds.). 2004. Human Capital 2004. New York:
Rowman & Littlefield.
Brewer, Gene A. 2005. “In the eye of the storm: frontline supervisors and federal agency
performance.” Journal of Public Administration Research and Theory 15, 4: 505–27.
2006. “All measures of performance are subjective: more evidence on US federal agencies.”
In George A. Boyne, Kenneth J. Meier, Laurence J. O’Toole, Jr., and Richard M. Walker
(eds.). Public Services Performance: Perspectives on Measurement and Management:
35–54. Cambridge University Press.
Brewer, Gene A., and Sally C. Selden. 2000. “Why elephants gallop: assessing and predicting
organizational performance in federal agencies.” Journal of Public Administrative
Research and Theory 10, 4: 685–711.
Burt, Ronald S. 1992. Structural Holes: The Social Structure of Competition. Cambridge, MA:
Harvard University Press.
1997. “The contingent value of social capital.” Administrative Science Quarterly, 42, 2:
339–65.
Burtless, Gary (ed.). 1996. Does Money Matter? The Effect of School Resources on Student
Achievement and Adult Success. Washington, DC: Brookings Institution Press.
Carmines, Edward, and R. Zeller. 1979. Reliability and Validity Assessment. Newbury Park,
CA: Sage.
Christensen, Robert K., and Beth Gazley. 2008. “Capacity and public administration: analysis
of meaning and measurement.” Public Administration and Development 28, 4: 265–79.
294 References

Chubb, John, and Terry M. Moe. 1990. Politics, Markets and America’s Schools. Washington,
DC: Brookings Institution Press.
Chun, Young Han, and Hal G. Rainey. 2005. “Goal ambiguity and organizational perform-
ance in US federal agencies.” Journal of Public Administration Research and Theory 15, 4:
529–57.
2006. “Consequences of goal ambiguity for public organizations.” In George A. Boyne, Kenneth
J. Meier, Laurence J. O’Toole, Jr., and Richard M. Walker (eds.). Public Services Performance:
Perspectives on Measurement and Management: 92–112. Cambridge University Press.
Coggburn, Jerrell D., and Saundra K. Schneider. 2003. “The quality of management and
government performance: an empirical analysis of the American states.” Public Adminis-
tration Review 63, 2: 206–13.
Cohen, David B., Justin S. Vaughn, and José D. Villalobos. 2010. “Subjective performance
indicators and White House management.” Paper presented at the sixty-eighth annual
national conference of the Midwest Political Science Association. Chicago, April 23.
Cohen, Steven, and William Eimicke. 1995. The New Effective Public Manager: Achieving
Success in a Changing Government. San Francisco: Jossey-Bass.
Conan Doyle, Arthur. 1894. The Memoirs of Sherlock Holmes. London: George Newnes.
Cook, Brian J. 1996. Bureaucracy and Self-Government: Reconsidering the Role of Public
Administration in American Politics. Baltimore: Johns Hopkins University Press.
Council of State Governments. 2002. The Book of the States 2002. Lexington, KY: CSG.
Cyert, Richard M., and James G. March. 1963. A Behavioral Theory of the Firm. Englewood
Cliffs, NJ: Prentice-Hall.
Dalton, Dan R., William D. Todor, Michael J. Spendolini, Gordon J. Fielding, and Lyman W.
Porter. 1980. “Organizational structure and performance: a critical review.” Academy of
Management Review 5, 1: 49–64.
Dee, Thomas, and Martin West. 2008. “The non-cognitive returns to class size.” Working
Paper no. 13994. Cambridge, MA: NBER (available at www.nber.org/papers/w13994).
Derthick, Martha. 1970. The Influence of Federal Grants: Public Assistance in Massachusetts.
Cambridge, MA: Harvard University Press.
Doig, Jameson W., and Erwin C. Hargrove. 1987. Leadership and Innovation: A Biographical
Perspective on Entrepreneurs in Government. Baltimore: Johns Hopkins University Press.
1990. Leadership and Innovation: A Biographical Perspective on Entrepreneurs in Govern-
ment, abridged edn. Baltimore: Johns Hopkins University Press.
Donahue, Amy K., Willow S. Jacobson, Mark D. Robbins, Ellen V. Rubin, and Sally C. Selden.
2004. “Management and performance outcomes in state government.” In Patricia W.
Ingraham and Laurence E. Lynn, Jr. (eds.). The Art of Governance: Analyzing Manage-
ment and Administration: 123–51. Washington, DC: Georgetown University Press.
Donahue, Amy K., Sally C. Selden, and Patricia W. Ingraham. 2000. “Measuring government
management capacity: a comparative analysis of city human resources management
systems.” Journal of Public Administration Research and Theory 10, 2: 381–411.
Drucker, Peter. 1967. The Effective Executive. New York: Harper & Row.
Dull, Matthew. 2006. “Why PART? The institutional politics of presidential budget reform.”
Journal of Public Administration Research and Theory 16, 2: 187–215.
Dunn, Delmer D. 1997. Politics and Administration at the Top: Lessons from Down Under.
University of Pittsburgh Press.
295 References

Edelenbos, Jurian, and Erik-Hans Klijn. 2007. “Trust in complex decision-making networks: a
theoretical and empirical exploration.” Administration and Society 39, 1: 25–50.
Ehrenberg, Ronald G., Richard P. Chaykowski, and Randy A. Ehrenberg. 1988a. “Determin-
ants of the compensation and mobility of school superintendents.” Industrial and Labor
Relations Review 41, 3: 386–401.
1988b. “Are school superintendents rewarded for performance?” In D. H. Monk (ed.).
Micro-Level School Finance: Issues and Implications for Policy: 337–64. Cambridge, MA:
Ballinger.
Ellingsen, Tore, and Magnus Johannesson. 2007. “Paying respect.” Journal of Economic
Perspectives 21, 4: 135–49.
Elmore, Richard F. 1997. “The paradox of innovation in education: cycles of reform and the
resilience of teaching.” In Alan A. Altshuler and Robert D. Behn (eds.). Innovation in
American Government: Challenges, Opportunities, and Dilemmas: 246–73. Washington,
DC: Brookings Institution Press.
Enticott, Gareth, George A. Boyne, and Richard M. Walker. 2009. “The use of multiple
informants in administration research: data aggregation using organizational echelons.”
Journal of Public Administration Research and Theory 19, 2: 229–53.
Fennell, Mary L., and Jeffrey A. Alexander. 1987. “Organizational boundary spanning in
institutionalized environments.” Academy of Management Journal 30, 3: 456–76.
Ferlie, Ewan, Lynn Ashburner, Louise Fitzgerald, and Andrew Pettigrew. 1996. The New Public
Management in Action. Oxford University Press.
Ferlie, Ewan, Laurence E. Lynn, Jr., and Christopher Pollitt (eds.). 2005. The Oxford Hand-
book of Public Management. Oxford University Press.
Fernández, Sergio. 2005. “Developing and testing an integrative framework of public sector
leadership: evidence from the public education arena.” Journal of Public Administration
Research and Theory 15, 2: 197–217.
Forbes, Melissa, and Laurence E. Lynn, Jr. 2005. “How does public management affect
government performance? Findings from international research.” Journal of Public
Administration Research and Theory 15, 4: 559–84.
Frederickson, H. George. 1999. “The repositioning of American public administration.” PS:
Political Science and Politics 32, 4: 701–11.
Freeman, J. Leiper. 1965. The Political Process: Executive Bureau–Legislative Committee
Relations. New York: Random House.
Gage, Robert W., and Myrna P. Mandell (eds.). 1990. Strategies for Managing Intergovern-
mental Policies and Networks. New York: Praeger.
Galbraith, Jay R. 1973. Designing Complex Organizations. Reading, MA: Addison-Wesley.
General Accounting Office. 1999. Managing for Results: Opportunities for Continued Improve-
ments in Agencies’ Performance Plans. Washington, DC: Government Printing Office.
Gerth, H. H., and C. Wright Mills (eds.). 1958. From Max Weber: Essays in Sociology. Oxford
University Press.
Gilmour, John D., and David E. Lewis. 2006. “Assessing performance budgeting at OMB.”
Journal of Public Administration Research and Theory 16, 2: 169–86.
Goerdel, Holly T. 2006. “Taking initiative: proactive management and organizational per-
formance in networked environments.” Journal of Public Administration Research and
Theory 16, 3: 351–67.
296 References

Golembiewski, Robert T. 1962. The Small Group: An Analysis of Research Concepts and
Operations. University of Chicago Press.
1990. “Differences in burnout, by sector: public vs. business estimates using phases.”
International Journal of Public Administration 13, 4: 545–59.
Goodsell, Charles. 1993. “Reinventing government or rediscovering it?” Public Administration
Review 53, 1: 85–6.
Gore, Al. 1993. From Red Tape to Results: Creating a Government that Works Better and Costs
Less. Washington, DC: Government Printing Office.
Graue, Elizabeth, Kelly Hatch, Kalpana Rao, and Denise Oen. 2007. “The wisdom of class size
reduction.” American Educational Research Journal 44, 3: 670–700.
Gulick, Luther. 1937. “Notes on the theory of organization.” In Luther Gulick and Lyndall
Urwick (eds.). Papers on the Science of Administration: 3–35. New York: Institute of
Public Administration.
Guyatt, G., J. Cairns, D. Churchill, et al. 1992. “Evidence-based medicine: a new approach
to teaching the practice of medicine.” Journal of the American Medical Association 268,
17: 2420–5.
Hall, Thad E., and Laurence J. O’Toole, Jr. 2000. “Structures for policy implementation: an
analysis of national legislation, 1965–1966 and 1993–1994.” Administration and Society
31, 6: 667–86.
2004. “Shaping formal networks through the regulatory process.” Administration and
Society 36, 2: 186–207.
Hamilton, James D. 1994. Time Series Analysis. Princeton University Press.
Hanushek, Eric. 1996. “School resources and student performance.” In Gary Burtless (ed.).
Does Money Matter? The Effect of School Resources on Student Achievement and Adult
Success: 43–73. Washington, DC: Brookings Institution Press.
Hanushek, Erik A., John F. Kain, Steven G. Rivkin. 1998. “Teachers, schools, and academic
achievement.” Working Paper no. 6691. Cambridge, MA: NBER (available at http://ssrn.
com/abstract=122569).
Hargrove, Erwin C., and J. C. Glidewell (eds.). 1990. Impossible Jobs in Public Management.
Lawrence, KS: University of Kansas Press.
Hawes, Daniel P. 2006. “Haven’t we been here before? Political institutions, public manage-
ment, and bureaucratic performance.” Paper presented at the conference “Empirical
studies of organizations and public management.” College Station, TX, May 6.
Heclo, Hugh. 1977. A Government of Strangers: Executive Politics in Washington. Washington,
DC: Brookings Institution Press.
Hedges, Larry V., and Rob Greenwald. 1996. “Have times changed? The relation between school
resources and student performance.” In Gary Burtless (ed.). Does Money Matter? The Effect
of School Resources on Student Achievement and Adult Success: 74–92. Washington, DC:
Brookings Institution Press.
Heinrich, Carolyn J. 2007. “Evidence-based policy and performance management.” American
Review of Public Administration 37, 3: 255–77.
Hennessey, J. Thomas. 1998. “‘Reinventing’ government: does leadership make the differ-
ence?” Public Administration Review 58, 5: 522–32.
Hess, Frederick M. 1999. Spinning Wheels: The Politics of Urban School Reform. Washington,
DC: Brookings Institution Press.
297 References

Hicklin, Alisa K. 2006. “The quest for diversity: increasing minority student representation at
American public universities.” Unpublished Ph.D. dissertation. Department of Political
Science, Texas A&M University, College Station.
Hicklin, Alisa, Laurence J. O’Toole, Jr., and Kenneth J. Meier. 2008. “Serpents in the sand:
managerial networking and nonlinear influences on organizational performance.” Journal
of Public Administration Research and Theory 18, 2: 253–73.
Hicklin, Alisa K., Laurence J. O’Toole, Jr., Kenneth J. Meier, and Scott E. Robinson. 2009.
“Calming the storms: collaborative public management, Hurricanes Katrina and Rita,
and disaster response.” In Rosemary O’Leary and Lisa Bingham (eds.). The Collaborative
Manager: New Ideas for the Twenty-First Century: 95–114. Washington, DC: Georgetown
University Press.
Hill, Carolyn J., and Laurence E. Lynn, Jr. 2005. “Is hierarchical governance in decline?
Evidence from empirical research.” Journal of Public Administration Research and Theory
15, 2: 173–95.
Hill, Gregory C. 2009. “The effect of frequent managerial turnover on organizational per-
formance.” Social Science Journal 46, 3: 557–70.
Holzer, Marc, and Kathe Callahan. 1998. Government at Work: Best Practices and Model
Programs. Thousand Oaks, CA: Sage.
Hood, Christopher. 1991. “A public management for all seasons.” Public Administration 69,
1: 3–19.
Hufen, Hans, and Arthur Ringeling (eds.). 1990. Beleidsnetwerken: Overheids-, semi-
overheids-, en particuliere organisaties in wisselwerking. The Hague: VUGA.
Huxham, Colin. 2000. “The challenge of collaborative governance.” Public Management
Review 2, 3: 337–52.
Ingraham, Patricia W. 1993. “Of pigs in pokes and policy diffusion: another look at pay for
performance.” Public Administration Review 23, 4: 348–56.
(ed.). 2007. In Pursuit of Performance: Management Systems in State and Local Government.
Baltimore: Johns Hopkins University Press.
Ingraham, Patricia W., Philip G. Joyce, and Amy Kneedler Donahue. 2003. Government
Performance: Why Management Matters. Baltimore: Johns Hopkins University Press.
Ingraham, Patricia W., and Laurence E. Lynn, Jr. 2004. The Art of Governance: Analyzing
Management and Administration. Washington, DC: Georgetown University Press.
Ingraham, Patricia W., Sally C. Selden, and Donald Moynihan. 2000. “People and perform-
ance: challenges for the future of the public service: the report from the Wye River
conference.” Public Administration Review 60, 1: 54–60.
Jacobs, Ronald L., and Christopher Washington. 2003. “Employee development and organi-
zational performance: a review of literature and directions for future research.” Human
Resource Development International 6, 3: 343–54.
Jacobson, Willow S., Christine Kelleher Palus, and Cynthia J. Bowling. 2010. “A woman’s
touch? Gendered management and performance in state administration.” Journal of
Public Administration Research and Theory 20, 2: 477–504.
Jencks, Christopher, and Meredith Phillips (eds.). 1998. The Black–White Test Score Gap.
Washington, DC: Brookings Institution Press.
Johansen, Morgen S. 2007. “The effect of female strategic managers on organizational
performance.” Public Organization Review 7, 3: 269–79.
298 References

2008. “Measuring middle manager quality and its effect on organizational performance.”
Paper presented at the third conference on “Empirical studies of organizations and
public management.” College Station, TX, May 2.
Jones, Bryan D. 1985. Governing Buildings and Building Government: A New Perspective on the
Old Party. University: University of Alabama Press.
Katz, Daniel, and Robert L. Kahn. 1978. The Social Psychology of Organizations, 2nd edn. New
York: Wiley.
Kaufman, Herbert. 1960. The Forest Ranger: A Study in Administrative Behavior. Baltimore:
Johns Hopkins University Press.
1991. Time, Chance, and Organizations: Natural Selection in a Perilous Environment, 2nd
edn. Chatham, NJ: Chatham House.
Keiser, Lael R., Vicky M. Wilkins, Kenneth J. Meier, and Catherine Holland. 2002. “Lipstick
or logarithms: gender, identity, institutions and representative bureaucracy.” American
Political Science Review 96, 3: 553–64.
Kellough, J. Edward. 2006. “Employee performance appraisal in the public sector: uses
and limitations.” In Norma M. Riccucci (ed.). Public Personnel Management: Current
Concerns, Future Challenges, 4th edn.: 177–89. New York: Longman.
Kellough, J. Edward, and Lloyd G. Nigro. 2006. “Personnel policy and public management:
the critical link.” In J. Edward Kellough and Lloyd G. Nigro (eds.). Civil Service Reform in
the States: Personnel Policy and Politics at the Subnational Level: 1–10. Albany, NY: SUNY
Press.
Kellough, J. Edward, and Will Osuna. 1995. “Cross-agency comparisons of quit rates in the
federal service: another look at the evidence.” Review of Public Personnel Administration
15, 4: 58–68.
Kelman, Steven. 2008. “Public administration and organization studies.” In James P. Walsh
and Arthur P. Brief (eds.). The Academy of Management Annals, vol. I: 225–67. New
York: Lawrence Erlbaum Associates.
Kickert, Walter J. M., Erik-Hans Klijn, and Joop F. M. Koppenjan (eds.). 1997. Managing
Complex Networks: Strategies for the Public Sector. London: Sage.
Kim, Sangmook. 2005. “Individual-level factors and organizational performance in govern-
ment organizations.” Journal of Public Administration Research and Theory 15, 2: 245–61.
Klijn, Erik-Hans. 1996. Regels en sturing in netwerken: De invloed van networkregels op de
herstructurering van naoorlogsewijken. Delft: Eburon.
Koch, Marianne J., and Rita G. McGrath. 1996. “Improving labor productivity: human
resource management policies do matter.” Strategic Management Journal 17, 5: 335–54.
Kooiman, Jan. 2003. Governing as Governance. London: Sage.
Kozol, Jonathan. 1991. Savage Inequalities: Children in America’s Schools. New York: Crown
Publishers.
Krause, George A., and James W. Douglas. 2005. “Institutional design versus reputational effects
on bureaucratic performance: evidence from US government macroeconomic and
fiscal projections.” Journal of Public Administration Research and Theory 15, 2: 281–306.
Krueathep, Weerasak, Norma M. Riccucci, and Charas Suwanmala. 2010. “Why do agencies
work together? The determinants of network formation at the subnational level of
government in Thailand.” Journal of Public Administration Research and Theory 20, 1:
157–85.
299 References

Laczko-Kerr, Ildiko, and David C. Berliner. 2002. “The effectiveness of ‘Teach for America’
and other under-certified teachers on student academic achievement: a case of harmful
public policy.” Education Policy Analysis Archives 10, 37, http://epaa.asu.edu/epaa/
v10n37.
Lankford, Hamilton, Susanna Loeb, and James Wyckoff. 2002. “Teacher sorting and the plight
of urban schools: a descriptive analysis.” Educational Evaluation and Policy Analysis 24, 1:
37–62.
Lawrence, Paul R., and Jay W. Lorsch. 1967. Organization and Environment: Managing
Differentiation and Integration. Cambridge, MA: Harvard University Press.
Lazear, Edward P., and Kathryn L. Shaw. 2007. “Personnel economics: the economist’s view of
human resources.” Journal of Economic Perspectives 21, 4: 91–114.
Levine, Charles H. 1978. “Organizational decline and cutback management.” Public Adminis-
tration Review 38, 4: 316–25.
(ed). 1980. Managing Fiscal Crisis: The Crisis in the Public Sector. Chatham, NJ: Chatham
House.
Li, Ling X. 2000. “An analysis of sources of competitiveness and performance of Chinese
manufacturers.” International Journal of Operations and Production Management 28, 3:
375–92.
Light, Paul C. 1997. The Tides of Reform: Making Government Work, 1945–1995. New Haven,
CT: Yale University Press.
1998. Sustaining Innovation: Creating Nonprofit and Government Organizations that Innov-
ate Naturally. San Francisco: Jossey-Bass.
1999. The True Size of Government. Washington, DC: Brookings Institution Press.
2008. “A government ill executed: the depletion of the federal service.” Public Adminis-
tration Review 68, 3: 413–19.
Lineberry, Robert. 1977. Equality and Urban Policy: The Distribution of Municipal Public
Services. Beverley Hills, CA: Sage.
Lipsky, Michael. 1980. Street-Level Bureaucracy: Dilemmas of the Individual in Public Services.
New York: Russell Sage Foundation.
Lundin, Martin. 2007. “Explaining cooperation: how resource interdependence, goal congru-
ence, and trust affect joint actions in policy implementation.” Journal of Public Adminis-
tration Research and Theory 17, 4: 651–72.
Lynn, Laurence E., Jr. 1984. “The Reagan administration and the renitent bureaucracy.” In
Lester M. Salamon (ed.). The Reagan Presidency and the Governing of America: 339–70.
Washington, DC: Urban Institute.
1996. Public Management as Art, Science, and Profession. Chatham, NJ: Chatham House.
2001. “The myth of the bureaucratic paradigm: what traditional public administration
really stood for.” Public Administration Review 61, 2: 144–60.
2006. Public Management: Old and New. New York: Routledge.
Lynn, Laurence E., Jr., Carolyn J. Heinrich, and Carolyn J. Hill. 2000. “Studying governance
and public management: Why? How?” In Carolyn J. Heinrich and Laurence E. Lynn, Jr.
(eds.). Governance and Performance: New Perspectives: 1–33. Washington, DC: Georgetown
University Press.
2001. Improving Governance: A New Logic for Empirical Research. Washington, DC: George-
town University Press.
300 References

Lynn, Monty L. 2005. “Organizational buffering: managing boundaries and cores.” Organiza-
tional Studies 26, 1: 37–61.
McGregor, Douglas. 1960. The Human Side of Enterprise. New York: McGraw-Hill.
2006. The Human Side of Enterprise, annotated edn., updated and with commentary by
J. Cutcher-Gershenfeld. New York: McGraw-Hill.
McGuire, Michael. 2006. “Collaborative public management: assessing what we know and
how we know it.” Public Administration Review 66, 6 (Supplement): 33–43.
Malysa, Lani Lee. 1996. “A comparative assessment of state planning and management
capacity: tidal wetlands protection in Virginia and Maryland.” State and Local Govern-
ment Review 28, 3: 205–18.
Mandell, Myrna P. (ed.). 2001. Getting Results through Collaboration: Networks and Network
Structures for Public Policy and Management. Westport, CT: Quorum Books.
Martin, Stephen, and Peter C. Smith. 2005. “Multiple public service indicators: towards
an integrated statistical approach.” Journal of Public Administration Research and Theory
15, 4: 599–613.
Meier, Kenneth J. 2007. “The public administration of politics, or what political science could
learn from public administration.” PS: Political Science and Politics 40, 1: 3–9.
Meier, Kenneth J., Carl Doerfler, Daniel Hawes, Alisa K. Hicklin, and Rene R. Rocha. 2006.
“The role of management and representation in improving performance of disadvan-
taged students: an application of Bum Phillips’s ‘Don Shula rule.’ ” Review of Policy
Research 23, 5: 1095–110.
Meier, Kenneth J., Warren Eller, and Michael Pennington. 2005. “Race, sex, and Clarence
Thomas: representation change in the EEOC.” Public Administration Review 65, 2: 171–9.
Meier, Kenneth J., and Jeff Gill. 2000. What Works: A New Approach to Program and Policy
Analysis. Boulder, CO: Westview Press.
Meier, Kenneth J., and Alisa K. Hicklin. 2008. “Employee turnover and organizational
performance: testing a hypothesis from classical public administration.” Journal of Public
Administration Research and Theory 18, 4: 573–90.
Meier, Kenneth J., and Jill Nicholson-Crotty. 2006. “Gender, representative bureaucracy,
and law enforcement: the case of sexual assault.” Public Administration Review 66, 6:
850–60.
Meier, Kenneth J., and Laurence J. O’Toole, Jr. 2001. “Managerial strategies and behavior in
networks: a model with evidence from US public education.” Journal of Public Adminis-
tration Research and Theory 11, 3: 271–95.
2002. “Public management and organizational performance: the impact of managerial
quality.” Journal of Policy Analysis and Management 21, 4: 629–43.
2003. “Public management and educational performance: the impact of managerial
networking.” Public Administration Review 63, 6: 675–85.
2004a. “Conceptual issues in modeling and measuring management and its impacts on
performance.” In Patricia W. Ingraham and Laurence E. Lynn (eds.). The Art of Govern-
ance: Analyzing Management and Administration: 195–223. Washington, DC: Georgetown
University Press.
2004b. “Unsung impossible jobs: the politics of public management.” Unpublished paper.
2005. “Managerial networking: issues of measurement and research design.” Administration
and Society 37, 5: 523–41.
301 References

2006. Bureaucracy in a Democratic State: A Governance Perspective. Baltimore: Johns

Hopkins University Press.
2008. “Management theory and Occam’s razor: how public organizations buffer the
environment.” Administration and Society 39, 8: 931–58.
2009a. “The dog that didn’t bark: how public managers handle environmental shocks.”
Public Administration 87, 3: 485–502.
2009b. “The proverbs of new public management: lessons from an evidence-based research
agenda.” American Review of Public Administration 39, 1: 4–22.
2010. “I think (I am doing well), therefore I am: assessing the validity of administrators’
self-assessments of performance.” Paper presented at the sixty-eighth annual national
conference of the Midwest Political Science Association. Chicago, April 24.
Forthcoming (a). “Beware of managers not bearing gifts: how management capacity
augments the impact of managerial networking.” Public Administration.
Forthcoming (b). “Comparing public and private management: theoretical expectations.”
Journal of Public Administration Research and Theory.
Meier, Kenneth J., Laurence J. O’Toole, Jr., George A. Boyne, and Richard M. Walker. 2007.
“Strategic management and the performance of public organizations: testing venerable
ideas against recent theories.” Journal of Public Administration Research and Theory 17, 3:
357–77.
Meier, Kenneth J., Laurence J. O’Toole, Jr., and Alisa K. Hicklin. 2009. “Comparing the
impact of public and private sector management: a preliminary analysis using colleges
and universities.” Paper presented at the annual meeting of the American Political
Science Association. Toronto, September 3.
2010. “I’ve seen fire and I’ve seen rain: public management and performance after a natural
disaster.” Administration and Society 41, 8: 979–1003.
Meier, Kenneth J., Laurence J. O’Toole, Jr., and Yi Lu. 2006. “All that glitters is not
gold: disaggregating networks and the impact on performance.” In George Boyne,
Kenneth Meier, Laurence J. O’Toole, Jr., and Richard Walker (eds.). Public Service
Performance: Perspectives on Measurement and Management: 152–70. Cambridge
University Press.
Meier, Kenneth J., Laurence J. O’Toole, Jr., and Sean Nicholson-Crotty. 2004. “Multilevel
governance and organizational performance: investigating the political-bureaucratic
labyrinth.” Journal of Policy Analysis and Management 23, 1: 31–47.
Meier, Kenneth J., and Joseph Stewart. 1991. The Politics of Hispanic Education: Un Paso
Pa’lante y dos Pa’tras. Albany, NY: SUNY Press.
Meyer, John W., and Brian Rowan. 1977. “Institutionalized organizations: formal structure as
myth and ceremony.” American Journal of Sociology 83, 2: 340–63.
Miles, Raymond E., and Charles C. Snow. 1978. Organizational Strategy, Structure, and
Process. New York: McGraw-Hill.
Milward, H. Brinton, and Keith G. Provan. 2000. “Governing the hollow state.” Journal of
Public Administration Research and Theory 20, 2: 359–79.
Miner, Anne S., Terry L. Amburgey, and Timothy M. Stearns. 1990. “Interorganizational
linkages and population dynamics: buffering and transformational shields.” Adminis-
trative Science Quarterly 35, 4: 689–713.
Mintzberg, Henry. 1973. The Nature of Managerial Work. New York: Harper & Row.
302 References

Mladenka, Kenneth R. 1980. “The urban bureaucracy and the Chicago political machine: who
gets what and the limits to political control.” American Political Science Review 74, 4:
991–8.
Moe, Ronald C. 1994. “The ‘Reinventing government’ exercise: misinterpreting the problem,
misjudging the consequences.” Public Administration Review 54, 2: 111–22.
Molnar, Alex, Philip Smith, John Zahorik, Amanda Planer, and Anke Halbach. 1999. “Evalu-
ating the SAGE program: a pilot program in targeted pupil–teacher reduction in
Wisconsin.” Educational Evaluation and Policy Analysis 21, 2: 165–77.
Moore, Mark H. 1995. Creating Public Value: Strategic Management in Government. Cam-
bridge, MA: Harvard University Press.
Moynihan, Donald. 2006. “What do we talk about when we talk about performance?
Dialogue theory and performance budgeting.” Journal of Public Administration Research
and Theory 16, 2: 151–68.
2008. The Dynamics of Performance Management: Constructing Information and Reform.
Washington, DC: Georgetown University Press.
Murphy, Kevin R., and Jeanette Cleveland. 1995. Understanding Performance Appraisal: Social,
Organizational, and Goal-Based Perspectives. Thousand Oaks, CA: Sage.
National Commission on Excellence in Education. 1983. A Nation at Risk: The Imperative for
Education Reform. Washington, DC: Government Printing Office.
Nicholson-Crotty, Sean, and Laurence J. O’Toole, Jr. 2004. “Public management and organ-
izational performance: the case of law enforcement agencies.” Journal of Public Adminis-
tration Research and Theory 14, 1: 1–18.
Nigro, Lloyd, Felix Nigro, and J. Edward Kellough. 2007. The New Public Personnel Adminis-
tration, 6th edn. Belmont, CA: Thomson Wadsworth.
O’Leary, Rosemary, and Lisa Blomgren Bingham (eds.). 2009. The Collaborative Public
Manager: New Ideas for the Twenty-First Century. Washington, DC: Georgetown Univer-
sity Press.
O’Toole, Laurence J., Jr. 1990. “Theoretical developments in public administration: implica-
tions for the study of federalism.” Governance 3, 4: 394–415.
1996. “Rational choice and the public management of interorganizational networks.” In
Donald F. Kettl and H. Brinton Milward (eds.). The State of Public Management: 241–63.
Baltimore: Johns Hopkins University Press.
1997. “Treating networks seriously: practical and research-based agendas in public admini-
stration.” Public Administration Review 57, 1: 45–52.
1998. Institutions, Policy, and Outputs for Acidification: The Case of Hungary. Aldershot:
Ashgate.
(ed.). 2000a. American Intergovernmental Relations: Foundations, Perspectives, and Issues,
3rd edn. Washington, DC: CQ Press.
2000b. “Different public managements? Implications of structural context in hierarchies
and networks.” In Jeffrey Brudney, Laurence J. O’Toole, and Hal G. Rainey (eds.).
Advancing Public Management: New Developments in Theory, Methods, and Practice:
19–32. Washington, DC: Georgetown University Press.
O’Toole, Laurence J., Jr., and Kenneth J. Meier. 1999. “Modeling the impact of public
management: implications of structural context.” Journal of Public Administration
Research and Theory 9, 4: 505–26.
303 References

2003a. “Bureaucracy and uncertainty.” In Barry C. Burden (ed.). Uncertainty in American

Politics: 98–117. Cambridge University Press.
2003b. “Plus ça change: public management, personnel stability, and organizational per-
formance.” Journal of Public Administration Research and Theory 13, 1: 43–64.
2004a. “Desperately seeking Selznick: cooptation and the dark side of public management
in networks.” Public Administration Review 64, 6: 681–93.
2004b. “Parkinson’s law and the new public management? Contracting determinants and service
quality consequences in public education.” Public Administration Review 64, 3: 342–52.
2004c. “Public management in intergovernmental networks: matching structural networks
with managerial networking.” Journal of Public Administration Research and Theory 14,
4: 469–94.
2006. “Networking in the penumbra: public management, cooptative links, and distribu-
tional consequences.” International Public Management Journal 9, 3: 271–94.
2007. “Public management and the administrative conservator: empirical support for
Larry Terry’s prescriptions.” Administrative Theory and Praxis 29, 1: 148–56.
2009. “The human side of public organizations: contributions to organizational perform-
ance.” American Review of Public Administration 39, 5: 499–518.
2010. “In defense of bureaucracy: public managerial capacity, slack, and the dampening of
environmental shocks.” Public Management Review 12, 3: 341–61.
O’Toole, Laurence J., Jr., Kenneth J. Meier, and Sean Nicholson-Crotty. 2005. “Managing
upward, downward, and outward: networks, hierarchical relationships and perform-
ance.” Public Management Review 7, 1: 45–68.
O’Toole, Laurence J., Jr., and Robert S. Montjoy. 1984. “Interorganizational policy imple-
mentation: a theoretical perspective.” Public Administration Review 44, 6: 491–503.
OECD. 1995. Decision-Making in 14 OECD Education Systems. Paris: OECD.
Osborne, David, and Ted Gaebler. 1992. Reinventing Government: How the Entrepreneurial
Spirit Is Transforming the Public Sector from Schoolhouse to Statehouse, City Hall to the
Pentagon. Reading, MA: Addison-Wesley.
Osborne, David, and Peter Plastrik. 1997. Banishing Bureaucracy: The Five Strategies for
Reinventing Government. Reading, MA: Addison-Wesley.
Palmer, Harvey D., and Guy D. Whitten. 1999. “The electoral impact of unexpected inflation
and economic growth.” British Journal of Political Science 29, 4: 623–39.
Pammer, William J. 1990. Managing Fiscal Strain in Major American Cities: Understanding
Retrenchment in the Public Sector. Westport, CT: Greenwood Press.
Perrow, Charles. 1986. Complex Organizations: A Critical Essay, 3rd edn. New York: Random
House.
Perry, James L., and Hal G. Rainey. 1988. “The public–private distinction in organi-
zation theory: a critique and research strategy.” Academy of Management Review 13, 2:
182–201.
Peters, B. Guy, and Donald J. Savoie (eds.). 2000. Governance in the Twenty-First Century:
Revitalizing the Public Service. Montreal: McGill-Queen’s University Press.
Peterson, John, and Laurence J. O’Toole, Jr. 2001. “Federal governance in the United States
and the European Union: a policy network perspective.” In Kalypso Nicolaidis and
Robert Howse (eds.). The Federal Vision: Legitimacy and Levels of Governance in the
United States and the European Union: 300–34. Oxford University Press.
304 References

Petrovsky, Nicolai. 2006. “Public management theory and federal programs: a test using
PART scores.” Paper presented at the conference “Empirical studies of organizations and
public management.” College Station, TX, May 5.
Pfeffer, James. 2007. “Human resources from an organizational behavior perspective: some
paradoxes.” Journal of Economic Perspectives 21, 4: 115–34.
Pfeffer, Jeffrey, and Gerald R. Salancik. 1978. The External Control of Organizations:
A Resource Dependence Perspective. New York: Harper & Row.
Pindyck, Robert S., and Daniel L. Rubinfeld. 1991. Econometric Models and Economic
Forecasts. New York: McGraw-Hill.
Pitts, David W. 2005. “Diversity, representation, and performance: evidence about race and
ethnicity in public organizations.” Journal of Public Administration Research and Theory
15, 4: 615–31.
2009. “Diversity management, job satisfaction, and performance: evidence from US federal
agencies.” Public Administration Review 69, 2: 328–38.
Pollitt, Christopher. 1990. Managerialism and the Public Services: The Anglo-American Experi-
ence. Oxford: Blackwell.
Pollitt, Christopher, and Geert Bouckaert. 2000. Public Management Reform: A Comparative
Analysis. Oxford University Press.
Powell, Walter W., and Paul J. DiMaggio (eds.). 1991. The New Institutionalism in Organiza-
tional Analysis. University of Chicago Press.
Pressman, Jeffrey. 1975. Federal Programs and City Politics: The Dynamics of the Aid Process in
Oakland. Berkeley: University of California Press.
Pressman, Jeffrey, and Aaron Wildavsky. 1984. Implementation: How Great Expectations in
Washington Are Dashed in Oakland; Or, Why It’s Amazing that Federal Programs Work at
all, This Being a Saga of the Economic Development Administration as Told by Two
Sympathetic Observers Who Seek to Build Morals on a Foundation of Ruined Hopes,
3rd edn. Berkeley: University of California Press.
Provan, Keith G., and H. Brinton Milward. 1991. “Institutional-level norms and organiza-
tional involvement in a service-implementation network.” Journal of Public Adminis-
tration Research and Theory 1, 4: 391–417.
1995. “A preliminary theory of interorganizational network effectiveness: a comparative
study of four community mental health systems.” Administrative Science Quarterly 40, 1:
1–33.
Putnam, Robert. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton
University Press.
2000. Bowling Alone: The Collapse and Revival of American Community. New York: Simon &
Schuster.
Radin, Beryl A. 2006. Challenging the Performance Movement: Accountability, Complexity, and
Democratic Values. Washington, DC: Georgetown University Press.
Raffel, Jeffrey A. 2007. “Why has public administration ignored public education, and does
it matter?” Public Administration Review 67, 1: 135–51.
Rainey, Hal G. 2003. Understanding and Managing Public Organizations, 3rd edn. San
Francisco: Jossey-Bass.
2009. Understanding and Managing Public Organizations, 4th edn. San Francisco: Jossey-
Bass.
305 References

Rainey, Hal G., and Paula Steinbauer. 1999. “Galloping elephants: developing elements of
a theory of government organizations.” Journal of Public Administration Research and
Theory 9, 1: 1–32.
Rattsø, Jørn. 1999. “Aggregate local public sector investment and shocks: Norway 1946–
1990.” Applied Economics 31, 4: 577–84.
Rethemeyer, R. Karl, and Deneen M. Hatmaker. 2008. “Network management recon-
sidered: an inquiry into management of network structures in public sector
service provision.” Journal of Public Administration Research and Theory 18, 4:
617–46.
Rhodes, R. A. W. 1997. Understanding Governance: Policy Networks, Governance, Reflexivity
and Accountability. Maidenhead: Open University Press.
Riccucci, Norma M. 1995. Unsung Heroes: Federal Executives Making a Difference. Washing-
ton, DC: Georgetown University Press.
2005. How Management Matters: Street-Level Bureaucrats and Welfare Reform. Washington,
DC: Georgetown University Press.
Rittel, Horst, and Melvin Webber. 1973. “Dilemmas in a general theory of planning.” Policy
Sciences 4, 2: 155–69.
Rivkin, Steven, G., Erik A. Hanushek, and John F. Kain. 2005. “Teachers, schools, and
academic achievement.” Econometrica 73, 2: 417–58.
Robinson, Scott, Kenneth J. Meier, Floun’say Caver, and Laurence J. O’Toole, Jr. 2007.
“Explaining policy punctuations: bureaucratic centralization, organizational size, and
the punctuated equilibrium theory of public agency budgets.” American Journal of
Political Science 51, 1: 140–50.
Roch, Christine H., and David W. Pitts. 2010. “Differing effects of representative bureaucracy
in charter schools and traditional public schools.” Unpublished paper. Department of
Public Management and Policy, Georgia State University, Atlanta.
Roch, Christine H., David W. Pitts, and Ignacio Navarro. 2010. “Representative bureaucracy
and policy tools: ethnicity, student discipline and representation in public schools.”
Administration and Society 42, 1: 38–65.
Rose, Richard. 1993. Lesson-Drawing in Public Policy: A Guide to Learning across Time and
Space. Chatham, NJ: Chatham House.
Rubin, Irene S. 2005. The Politics of Public Budgeting: Getting and Spending, Borrowing and
Balancing, 5th edn. Washington, DC: CQ Press.
Salisbury, Robert. H. 1984. “Interest representation: the domination of institutions.” Ameri-
can Political Science Review 78, 1: 64–76.
Sandfort, Jodi R. 1999. “The structural impediments to human service collaboration: the case
of welfare reform.” Social Service Review 73, 3: 314–39.
Sargent, Stephen. 2009. “The impact of internal management on organizational perform-
ance.” Unpublished Ph.D. dissertation. Deptartment of Political Science, Texas A&M
University, College Station.
Schalk, Jelmer, René Torenvlied, and Jim Allen. 2010. “Network embeddedness and public
agency performance: the strength of strong ties in Dutch higher education.” Journal of
Public Administration Research and Theory 20, 3: 629–53.
Scharpf, Fritz W. (ed.). 1993. Games in Hierarchies and Networks: Analytical and Empirical
Approaches to the Study of Governance Institutions. Frankfurt: Campus.
306 References

1997. Games Real Actors Play: Actor-Centered Institutionalism in Policy Research. Boulder,
CO: Westview Press.
Schattschneider, E. E. 1960. The Semi-Sovereign People: A Realist’s View of Democracy in
America. New York: Holt, Rinehart and Winston.
Scholzman, Kay Lehman. 1984. “What accent the heavenly chorus? Political equality and the
American pressure system.” Journal of Politics 46, 4: 1006–32.
Seidman, Harold. 1998. Politics, Position, and Power: The Dynamics of Federal Organization,
5th edn. New York: Oxford University Press.
Selden, Sally C. 1997. The Promise of Representative Bureaucracy: Diversity and Responsiveness
in a Government Agency. New York: M. E. Sharpe.
Selden, Sally C., and Willow Jacobson. 2007. “Government’s largest investment: human
resource management in states, counties, and cities.” In Patricia W. Ingraham (ed.). In
Pursuit of Performance: Management Systems in State and Local Government: 82–116.
Baltimore: Johns Hopkins University Press.
Selznick, Philip. 1949. TVA and the Grass Roots: A Study in the Sociology of Formal Organiza-
tion. Berkeley: University of California Press.
Shafritz, Jay M., David H. Rosenbloom, Norma M. Riccucci, Katherine C. Naff, and Al C.
Hyde. 2001. Personnel Management in Government: Politics and Process, 5th edn.
New York: Marcel Dekker.
Simon, Herbert A. 1946. “The proverbs of administration.” Public Administration Review 6,
1: 53–67.
1997. Administrative Behavior: A Study of Decision-Making Processes in Administrative
Organizations, 4th edn. New York: Free Press.
Smith, Kevin B. 2003. The Ideology of Education: The Commonwealth, the Market and
America’s Schools. Albany, NY: SUNY Press.
Sorenson, Olav. 2003. “Interdependence and adaptability: organizational learning and the
long-term effect of integration.” Management Science 49, 4: 446–63.
Spicer, Michael W., and Larry D. Terry. 1993. “Legitimacy, history and logic: public adminis-
tration and the constitution.” Public Administration Review 53, 3: 239–45.
Stoker, Gerry (ed.). 1999. The New Management of British Local Governance. Basingstoke:
Macmillan.
Stoker, Robert. 1992. Reluctant Partners: Implementing Federal Policy. University of Pittsburgh
Press.
Terry, Larry D. 2002. Leadership of Public Bureaucracies: The Administrator as Conservator,
2nd edn. Armonk, NY: M. E. Sharpe.
Thompson, Fred, and Lawrence R. Jones. 1994. Reinventing the Pentagon: How the New Public
Management Can Bring Institutional Renewal. San Francisco: Jossey-Bass.
Thompson, James D. 1967. Organizations in Action: Social Science Bases of Administrative
Theory. New York: McGraw-Hill.
Todd, Petra E., and Kenneth I. Wolpin. 2003. “On the specification and estimation of the
production function for cognitive achievement.” Economic Journal 113, 1: F3–F33.
Tyack, David. 1974. The One Best System: A History of American Urban Education. Cambridge,
MA: Harvard University Press.
US Census Bureau. 2008. 2007 Census of Governments. Washington, DC: Government
Printing Office.
307 References

Vaughn, Justin, and José Villalobos. 2009. “The managing of the presidency: applying theory-
driven empirical models to the study of White House bureaucratic performance.”
Political Research Quarterly 62, 1: 158–63.
Verba, Sidney, and Norman Nie. 1972. Participation in America: Political Democracy and
Social Equality. New York: Harper & Row.
Wagner, Kevin, and Jeff Gill. 2005. “Bayesian inference in public administration research.”
International Journal of Public Administration 28, 1: 5–35.
Walker, David M. 2001. Human Capital: Taking Steps to Meet Current and Emerging Human
Capital Challenges. Washington, DC: General Accounting Office.
Walker, Richard M., and George A. Boyne. 2006. “Public management reform and organiza-
tional performance: an empirical assessment of the UK Labour government’s
public service improvement strategy.” Journal of Policy Analysis and Management 25, 2:
371–93.
Walker, Richard M., Laurence J. O’Toole, Jr., and Kenneth J. Meier. 2007. “It’s where you are
that matters: an empirical analysis of the networking behaviour of English local govern-
ment officers.” Public Administration 85, 3: 739–56.
Wamsley, Gary L., Robert N. Bacher, Charles T. Goodsell, Philip S. Kronenberg, John A. Rohr,
Camilla M. Stivers, Orion F. White, and James F. Wolf. 1990. Refounding Public Adminis-
tration. Newbury Park, CA: Sage.
Wenger, Jeffrey B, Laurence J. O’Toole, Jr., and Kenneth J. Meier. 2008. “Trading speed for
accuracy? Managing goal conflict and accommodation in the US Unemployment Insur-
ance Program.” Policy Studies Journal 36, 2: 175–98.
Wenglinsky, Harold. 1997. How Educational Expenditures Improve Student Performance and
How They Don’t. Princeton, NJ: Educational Testing Service.
Wilkins, Vicky M., and Lael R. Keiser. 2006. “Linking passive and active representation
by gender: the case of child support agencies.” Journal of Public Administration Research
and Theory 16, 1: 87–102.
Wilson, James Q. 1989. Bureaucracy: What Government Agencies Do and Why They Do It.
New York: Basic Books.
Wilson, Woodrow. 1887. “The study of administration.” Political Science Quarterly 2, 2: 197–222.
Wirt, Frederick, and Michael Kirst. 2005. The Political Dynamics of American Education.
Berkeley, CA: McCutchan Publishing.
Wolf, Patrick J. 1993. “A case survey of bureaucratic effectiveness in US Cabinet
agencies: preliminary results.” Journal of Public Administration Research and Theory 3,
2: 161–81.
Wong, Kenneth K. 1999. Funding Public Schools: Politics and Policies. Lawrence: University
Press of Kansas.
Wood, Stephen. 1999. “Human resource management and performance.” International Jour-
nal of Management Reviews 1, 4: 367–414.
Woodward, Joan. 1965. Industrial Organization: Theory and Practice. Oxford University Press.
Wright, Deil S. 1988. Understanding Intergovernmental Relations, 3rd edn. Pacific Grove, CA:
Brooks/Cole.
1990. “Federalism, intergovernmental relations, and intergovernmental management:
historical reflections and conceptual comparisons.” Public Administration Review 50, 2:
168–78.
308 References

Wrinkle, Robert D., Joseph Stewart, and Jerry L. Polinard. 1999. “Public school quality,
private schools and race.” American Journal of Political Science 43, 4: 1248–53.
Wulf, Henry S. 2002. “Trends in state government finances.” In Council of State Govern-
ments. The Book of the States 2002: 269–74. Lexington, KY: CSG.
Yukl, Gary. 2006. Leadership in Organizations, 6th edn. Upper Saddle River, NJ: Prentice Hall.
Zeigler, L. Harmon, and G. Wayne Peak. 1972. Interest Groups in American Society. Englewood
Cliffs, NJ: Prentice-Hall.
 Index

Note: page numbers in italic indicate Glossary items.

above criterion 287 distributional consequences of buffering 233
ACT scores 48, 287 English local government example 234
administrative function environmental factors (X) 227
conservatorship role 134 findings 227–32
role in managerial networking 115, 117–18, forms of adaptive response 219
121–3 implications for performance 232–3
stability 133–4, 135–8 management variables (M) 227
all-pass rate 287 managerial buffering elements 218–19, 221, 222
Anglo pass rate 287 measurement 226–7
Appleby, Paul 2 modeling a measure of buffering 219–21,
attendance rates 49, 287 222–6
autoregressive models 75–81 Occam’s razor principle 226, 232, 233
autoregressive function 68, 69, 70 outcome measures (O) 226–7
impact of networking on future selective filters 219
performance 75–7 structural buffering elements 218–19, 221–2
nonlinear future impacts of networking types of buffering functions 219
77–81 bureaucracy 132–3, 134, 182, 183, 194–7, 287
autoregressive systems 287 burnout-generated turnover 136, 137
modeling 28–30 Bush, George W. 11–12, 124–5
public organizations as xii
capacity see managerial capacity
black pass rate 287 change see organizational change
budget shocks, summary of findings 271–4 collaboration through networks 57
see also decision-making study; managerial see also managerial networking
capacity and budget shocks college boards 287
buffering (SM4) 287 college-bound student performance
effects of managerial slack 185–6 measures 48
impacts on performance 234–5 college-ready student 288
impediments to performance 68–75 common-source bias 13
modeling 31–2 Comprehensive Performance Assessment
summary of findings 273–4 (Audit Commission) 13–14
buffering and managerial capacity 185–6, 194–7, Conan Doyle, Arthur 162–3, 182
202–6, 207, 218–35 conserving activities 102, 134
approach to measuring buffering 221–6 Core Service Performance (Audit Commission)
blockade-type buffers 219 13–14
buffering mechanisms 219
dampener-type buffers 219 dark side hypothesis, managerial networking
data 226–7 82–3, 83–6, 86–93
definition of buffering 218–19 data sets xiii
definition of buffers 218–19 developing additional data sets 285–6

309
310 Index

data sets (cont.) governance 288

exploration of other data sets 277–8 future research 279
extending to other sectors 284–5 model 15, 37
variety of data used 45–6 patterns 3
see also Texas school district data set government, policy implementation
decision-making study 161–74 challenges 1–4
budgetary shock context 161–3 Government Performance and Results Act
controls 164 (1973) 11
dependent variables 164 Government Performance Project (GPP) 12,
environmental shocks 161–2 103, 152
findings 164–72
implications 172–4 hierarchy 24–5
long-term costs 173–4 human resources management 274
maintaining performance during budget contribution to performance 152–3, 159–61
cuts 162–3, 172–4 findings 156–9
measuring budget shocks 163–4 human capital as a long-term asset 160
mitigating negative budgetary shocks implications 159–61
162–3, 172–4 link with performance 150
short-term costs 173 measures 153–5
distributional consequences modeling the impact of 151–2
of buffering 233 need for investment in people 160
of networking 82–3, 83–6, 86–93, 270 sample 153–5
dropout rates 49, 288 theorizing about the content of M1 152–3
Hurricane Katrina 197–8
economic equity measure 48 Hurricane Rita 198
effectiveness of public management 2, 7–8, 14
efficiency and performance 2 independent school district 288
English local government example, managerial individual performance appraisal 104
capacity 234 inequality, consequences of networking 82–3,
entrepreneurial management 102, 133–4 83–6, 86–93, 270
environment inertial systems
exploiting opportunities in (M3) xiii, modeling 28–30
68–75, 282 public organizations as xii
organizational relationships with xiii see also autoregressive models; autoregressive
threats from xiii systems
environmental factors (X) 227 innovation in public management 133–4
environmental shocks 161–2 institutional design and reputation 6
modeling 30–1 intergovernmental aid 243
nature of response to 68–75 intergovernmental networks 240–6
summary of findings 271–4 behavioral aspect of networking 245–6
term (X’) 186–7 intergovernmental aid 243
see also decision-making study; managerial intergovernmental landscape 241–2
capacity and budget shocks; managerial intergovernmental management 241–2
capacity and natural disasters intergovernmental ties as networked
environmental typology 40–1 relations 242–3
pooled environments 40, 41 network terminology 245–6
reciprocal environments 40, 41 networked environment of public education
sequentially structured environments 40, 41 243–5
equity and performance 2 structural networks 240–1, 245–6
European Union 263 intergovernmental networks and public
evidence-based approach to public management 247–65
management 4–5 behavioral aspect of networking 249–50, 264
conclusions 263–5
Ford Foundation 133 control variables (X0t) 247, 251
311 Index

findings 251–63 findings 189–94

influences on performance 263–5 identifying slack in managerial resources 185–6
intergovernmental context 250–1 implications 194–7
management measures (M2) 249–50, 264 location of slack in the system 196
managerial networking 249 managerial capacity measurement 188
managerial quality 249 managerial reserve available 183–6
managerial stability 250 measures 187–9
measures 249 measuring shocks 187–8
modeling 247–9 methods 189
outcome measures (O) 251 organization theory approach 183
performance measures (O) 251 performance indicators 188–9
stability measures (S) 250 performance priorities 195–6
structural dimension (Xi) 247, 250–1, 264 positive aspects of bureaucracy 194–7
teacher stability 250 potential to mitigate budgetary shocks 183–6
units of analysis 249 protection and defense functions 183–6
internal management (M1) 288 reduced model 186–7
functions 131–2 sample 187–9
theorizing about the content of M1 152–3 views on bureaucracy 183
internal management and performance 131 managerial capacity and buffering 185–6, 194–7,
human resources management 150 202–6, 207, 218–35
making quality decisions 161–74 approach to measuring buffering 221–6
organizational stability 132–50 blockade-type buffers 219
range of functions and challenges 131–2 buffering mechanisms 219
review of studies in other settings 174–8 dampener-type buffers 219
summary of findings 272, 274–5 data 226–7
definition of buffering 218–19
Kennedy School of Government 133 definition of buffers 218–19
distributional consequences of buffering 233
Latino pass rate 288 English local government example 234
leadership 101–2, 103–4 environmental factors (X) 227
leadership stability 136 findings 227–32
local governments, policy implementation 1–2 forms of adaptive response 219
low-income pass rate 288 impacts on performance 234–5
low-income student 288 implications for performance 232–3
management variables (M) 227
M terms, quality component 100 managerial buffering elements 218–19, 221, 222
M1 see internal management measurement 226–7
M2 see managerial networking modeling a measure of buffering 219–21, 222–6
M3 see environment Occam’s razor principle 226, 232, 233
management quality hypothesis 100–1 outcome measures (O) 226–7
managerial capacity 288 selective filters 219
as managerial potential 183–6 structural buffering elements 218–19, 221–2
in non-crisis times 207–18 types of buffering functions 219
summary of findings 272–3 managerial capacity and managerial networking
managerial capacity and budget shocks 183–97 207–18
amount of protection to build in 195–6 control variables 211
approach to measuring managerial capacity data and measurement 210–11
185–6 findings 212–17
buffering effect of managerial capacity 194–7 impacts on performance 217–18
buffering effect of managerial slack 185–6 implications 217–18
capacity as managerial potential 183–6 links between 207–9
control variables 189 management variables (M) 211
costs of surplus capacity 195–6 managerial capacity (Mc) 211
environmental shock term (X’) 186–7 managerial networking (M2) 211
312 Index

managerial capacity and managerial networking measuring impact on performance 68–75

(cont.) measuring M2 61–4
managerial quality (Mq) 211 mechanisms of performance enhancement 81–2
managerial stability 211 nature of response environmental shocks 68–75
managing upward 211 nature of the link with performance 114–15,
modeling the impact of managerial capacity 116–21
209–10 network-related activities 8, 9
networking strategy and resources 207–9 nonlinear impacts on future performance 77–81
outcome measures (O) 211 nonlinear impacts on performance 68–75,
personnel stability variables 211 114–15, 116–21
potential impacts on performance 207–9 opportunity costs 115, 116
role of managerial capacity in networking 207–9 organization theory context 93
workforce stability 211 organizational influences on 121–3
managerial capacity and natural disasters 197–207 political issues 82–3, 83–6, 86–93
autoregressive model 201–2 politics of program management 85–6, 86–93
buffering effects 202–6, 207 potential for diminishing returns 115, 116–21
control variables 201–2 public–private partnerships 57
data 199–202 quality (skill) aspect 100, 117
environmental shock term (X’) 199–200 quantity (effort) aspect 117
findings 202–6 role of administrative support staff 115,
Hurricanes Katrina and Rita 197–8 117–18, 121–3
implications 207 strategic choices by managers 66–8, 75
measuring the environmental shock (X’) studies of performance effects 93–6
199–200 summary of findings 269–71
nonlinear effects 202–6, 207 tendency toward inequality 82–3, 83–6, 86–93
outcome measures (O) 200–1 tradeoffs 115, 116
managerial functions, modeling 26–7, 278–9 use of resources and time 116–21
managerial influence voluntary linkages 57
extent of 275 managerial networking and managerial
modeling 32–6 capacity 207–18
patterns of 275–6 control variables 211
managerial networking (M2) 55–96, 288 data and measurement 210–11
autoregressive models 75–81 findings 212–17
buffering impediments to performance 68–75 impacts on performance 217–18
building support for programs 56 implications 217–18
collaboration 57 links between 207–9
comparing different public organizations 64–6 management variables (M) 211
dark side hypothesis 82–3, 83–6, 86–93 managerial capacity (Mc) 211
diminishing returns 115, 116–21 managerial networking (M2) 211
distributional consequences 82–3, 83–6, managerial quality (Mq) 211
86–93, 270 managerial stability 211
exploitation of environmental opportunities managing upward 211
68–75 modeling the impact of managerial capacity
fending off threats 56 209–10
functionalist perspective 83–6, 86–93 networking strategy and resources 207–9
government requirement for 56–7 outcome measures (O) 211
impact on future performance 75–7 personnel stability variables 211
impact on organizational performance 66–75 potential impacts on performance 207–9
implications 81–2 role of managerial capacity in networking 207–9
influence of managerial quality 120–1 workforce stability 211
interactions with external parties 59–66 managerial networking and managerial quality
managerial balancing act 116–21 see managerial quality and managerial
measuring extent and frequency 59–66 networking
measuring impact of M2 68–75 managerial quality 289
313 Index

managerial quality and managerial networking elements of a model 23–8

114–23 environmental shocks 30–1
managerial balancing act 116–21 governance model 15
nature of the link with performance 114–15, hierarchy 24–5
116–21 inductive approach 16
nonlinear relationships 114–15, 116–21 inertial (autoregressive) systems 28–30
opportunity costs 115, 116 influence of multiple managers 27–8
organizational differences 121–3 limitations of the approach 17–18
potential for diminishing returns 115, 116–21 managerial functions (M) 26–7
quality (skill) aspect of networking 117 managerial influences 32–6
quantity (effort) aspect of networking 117 network-level modeling 38–42
role of administrative support staff 115, networks 16–17, 25–6
117–18, 121–3 operation at multiple levels 27–8
tradeoffs 115, 116 parsimonious and formal approach 14–18
use of resources and time 116–21 role of internal management 16
managerial quality and performance 100–28 shocks 30–1
approach to measurement 104, 105–6 stabilizing forces 24, 35–6
causal paths 107 strengths of the approach 17
characterizing managerial quality 101–4 testing hypotheses 36, 38
entrepreneurial management 102 theoretical influences 15–16
individual performance appraisal 104 underlying principles xiii
leadership 101–2, 103–4 modeling
management quality concept 101–4 managerial functions 26–7, 278–9
management quality hypothesis 100–1 managerial influence 32–6
managerial quality studies 123–5 modeling performance 110–14
measurement challenge 102 control variables 110
measuring managerial quality 107–10 findings 111–14
measuring superintendent quality 105–6 performance measures 110–11
modeling performance 110–14 multiorganizational action 3
pay for performance systems 104
performance appraisal 104 National Performance Review 133
protective and conserving activities 102 natural disasters see managerial capacity and
quality component of the M terms 100 natural disasters
quality management perspective 102 network-level modeling 38–42
review of related studies 103–4 pooled environments 40, 41
risk-taking activities 102 reciprocal environments 40, 41
salary-based measures 104, 105–6 sequentially structured environments 40, 41
salary model 107–10 typology of environments 40–1
summary of research findings 125–8, 271 networking
managerial stability (Sm) 289 as behavior 245–6
managerialism xii, 4, 5, 275 uses of the term 245–6
managing downward 289 see also managerial networking
managing outward see managerial networking networks 25–6, 289
managing upward 211, 289 advantages of 56–7
Merit Principles Survey 12–13 approach to investigation 57–9
mission stability 24, 135 characteristics of 55
model of public management and performance 23, coordination challenges 25–6
28–38, 38–42 dealing with “wicked problems” 56
aims of the approach 15 fluidity of 25–6
alternative approach 15 inducements to use 56–7
approach to modeling 36–8 managerial challenges 56
autoregressive (inertial) systems 28–30 means of avoiding difficult issues 84
buffering 31–2 nodes 25
data sets used for testing xiii policy-related problem solving 56
314 Index

networks (cont.) structural buffering elements 218–19, 221–2

political pressure within 85–6, 86–93 types of buffering functions 219
shifts in policy emphasis 84–5
stability in 25 Occam’s razor principle 226, 232, 233
structural and behavioral elements 271 opportunity costs, managerial networking
structural meaning 240–1, 245–6 115, 116
uses of the term 245–6 organization theory 182, 183
see also intergovernmental networks context for networking 93
new public management xii, 5, 7, 102, 134, organizational change 133–4
177, 194, 197 managerial efforts to bring about 283–4
nonlinear relationships between variables organizational goal ambiguity 6
xiii, 182–235 organizing through structures and procedures xiii
buffering 218–35 outcomes of an organization or program 289
future impacts of networking 77–81 outputs of an organization or program 289
managerial capacity and budget shocks 183–97
managerial capacity and managerial PART (Program Assessment Rating Tool) 6, 12,
networking 207–18 124–5, 268
managerial capacity and natural disasters pay for performance systems 6, 104
197–207 percentage tested 289
managerial capacity and performance 234–5 performance 289
managerial capacity in non-crisis times 207–18 and effectiveness 2
managerial networking and managerial and efficiency 2
capacity 207–18 and equity 2
managerial networking and managerial quality and governance patterns 3
114–15, 116–21 and organizations 3
managerial networking and performance 68–75 and public management 3–4
managerial quality and managerial networking and public satisfaction 2
114–15, 116–21 broader measures of 282–3
public management and performance 9 definitions 2
stability in public management 145–8 link with public management 2
summary of findings 270 nonlinear relationship with management 9
nonlinearity and buffering 218–35 research findings 6–7, 9–10
approach to measuring buffering 221–6 performance appraisal 104
blockade-type buffers 219 performance measurement 10–14
buffering mechanisms 219 challenges 10–11
dampener-type buffers 219 developing better measures 285–6
data 226–7 link to management effectiveness 14
definition of buffering 218–19 measurement process examples 11–14
definition of buffers 218–19 Texas school district data set 46–9
distributional consequences of buffering 233 performance targets 6
environmental factors (X) 227 personnel stability (Sp) 24, 135–6, 274, 290
findings 227–32 personnel stability analysis 136–8
forms of adaptive response 219 burnout-generated turnover 136, 137
impacts on performance 234–5 conclusions about stability 149–50
implications for performance 232–3 consequences of personnel instability 136–8
management variables (M) 227 findings 140–5
managerial buffering elements 218–19, 221, 222 influence on performance 140–5
managerial capacity in English local interaction between variables 145–8
government 234 leadership stability 136
measurement 226–7 measures of personnel stability 140
modeling a measure of buffering 219–21, 222–6 measures used 140
Occam’s razor principle 226, 232, 233 modeling personnel stability (Sp) 138–40
outcome measures (O) 226–7 modeling the impact of stability (S) 138–40
selective filters 219 nonlinear relationships 145–8
315 Index

policy implementation public management research, findings

avoiding difficult issues 84 approach to analysis 267–8
challenges faced by governments 1–4 budget shocks 271–4
definitions of performance 2 buffering of shocks 273–4
multiorganizational action 3 distributional consequences of networking 270
networks of organizations 3 environmental shocks 271–4
patterns of governance 3 extent of managerial influence 275
performance 2 features of the model 268–9
role of public management 2 generalization from findings 267–8
role of subnational governments 1–2 human resources management 274
scale and scope of public organizations 3 internal management 272, 274–5
shifts in policy emphasis 84–5 managerial capacity 272–3
policy-related problem solving, use of networks 56 managerial networking 269–71
political issues 270 managerial quality 271
in networking 82–3, 83–6, 86–93 nature of the data set 267–8
future research 279–80 nonlinear interactions 270
political pressure within networks 85–6, 86–93 patterns of managerial influence 275–6
political science perspective 4–5 personnel stability 274
politics of program management 85–6, 86–93 political issues 270
population ecology perspective xii, 4–5 practical applications 276–7
principal (of a school) 290 relevance of the findings 267–8
procedural stability 24, 135 stabilizing forces 274
production function variables 49–51 structural and behavioral elements
production stability 24, 135 of networks 271
program management, political issues structural stability 274
85–6, 86–93 what the evidence shows 267–77
protective activities, management function 102 public management research, future agenda
public education, networked environment 243–5 additional data sets 285–6
public management 290 broader assessment of performance 282–3
as impediment to service delivery xii developing better measures 285–6
definition 2 expansion of the Texas schools data set 278
managerialist view xii exploiting the environment (M3) 282
new public management view xii exploration of other data sets 277–8
population ecology view xii extending to other sectors 284–5
range of theoretical approaches xi–xii governance structures 279
role in policy implementation 2 managerial efforts to bring about
views on effectiveness xi–xii change 283–4
public management and performance 2, 3–4 modeling the full range of managerial
evidence-based perspective 4–5 functions 278–9
managerialist perspective 4, 5 politics and management 279–80
multiple influences 5 social capital 280–2
new public management perspective 5 what remains to be explored 277–86
political science perspective 4–5 public management studies, review
population ecology perspective 4–5 competing theories 10
research findings 6–7, 9–10 difficulty of assessing performance 10–14
public management principles xii–xiii distinctiveness of public management 7
autoregressive (inertial) systems xii effectiveness of public management 7–8
dealing with threats in the environment xiii importance of public management 7–8
exploiting opportunities in the environment xiii interactive effects 9
managing within the organization xiii key themes and findings 7–9
mathematical model xiii lack of consensus about performance 9–10
nonlinear relationships between variables xiii managers catalyze action 9
organizing through structures and procedures xiii multifaceted nature of management 8–9, 9–10
relationships with the environment xiii network-related activities 8, 9
316 Index

public management studies, review (cont.) procedural stability 135

nonlinear relationship with performance 9 production stability 135
public sector performance 6–7, 9–10 relationship to performance 132–3
range of approaches 5–7 stabilizing forces 24, 25, 35–6, 274
public organizations structural stability 24, 25, 135, 274
multiorganizational action 3 support for 134
networks of organizations 3 technology stability 135
scale and scope of 3 unpopularity of , 133–4
public–private partnerships 57 see also managerial stability; personnel stability
public satisfaction, and performance 2 stabilizing forces 24, 25, 274
modeling 35–6
quality management perspective 102 strategic choices, managerial networking 66–8, 75
quality of public management see managerial structural and behavioral elements of
quality networks 271
structural fluidity 25–6
racial equity measure 48 structural networks 240–1, 245–6, 247, 250–1, 264
representative bureaucracy 176 structural stability 24, 25, 135, 274
research see public management research subnational governments, policy implementation
resources, measures of 50 challenges 1–2
risk taking by managers 102 substantively weighted analytical technique
(SWAT) 77–8
S see stability superintendent 290
salary-based measures of quality 104, 105–6 quality measurement 105–6
salary model 107–10
SAT scores 48, 290 TAAS (Texas Assessment of Academic Skills)
school boards 290 scores 46–9, 290
Selznick, Philip 85, 89, 92 TAKS (Texas Assessment of Knowledge and Skills)
shocks see environmental shocks scores 46–9, 290
Sm see managerial stability tapping performance 10–14
SM4 see buffering link to management effectiveness 14
social capital 280–2 measurement process examples 11–14
Sp see personnel stability performance measurement challenges 10–11
stability (S) in public management 132–50 task difficulty, measures of 50
administrative conservatorship 134 technology stability 24, 135
administrative stability 133–4, 135–8 Texas Education Agency (TEA) 290
bureaucracy 1, 132–3, 134, 182, 183, 194–7 Texas school district data set xiii, 42–52, 267–8
burnout-generated turnover 136, 137 ACT scores 48, 287
changing views on 132–3 attendance rates 49, 287
conclusions about stability 149–50 college-bound student performance
consequences of personnel instability 136–8 measures 48
criticism of 133–4 dropout rates 49, 288
drive for innovation 133–4 economic equity measure 48
emphasis on organizational change 133–4 future expansion 278
entrepreneurial management 133–4 measures of resources 50
influence on performance 140–5 measures of task difficulty 50
interaction between variables 145–8 performance measures 46–9
leadership stability 136 presentation of findings 51
measures of personnel stability 140 production function variables 49–51
mission stability 135 racial equity measure 48
modeling personnel stability (Sp) 138–40 research value of the data set 44–6
modeling the impact of stability (S) 138–40 SAT scores 48, 290
nonlinear relationships 145–8 structure of the Texas districts 44
personnel stability 135–6 TAAS/TAKS exam scores 46–9, 290
personnel stability analysis 136–8 Texas context 43
317 Index

United Kingdom Merit Systems Protection Board 12–13

Audit Commission performance measures 13–14 Office of Management and Budget 11–12
English local government example 234 The President’s Management Agenda (2002)
United States 11–12
General Accounting Office 11
Government Accountability Office 11 “wicked problems,” use of networks 56
Government Performance and Results Act
(1973) 11 X see environmental factors