Political Restraint of the Market

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
and Levels of Criminal Homicide:
A Cross-National Application of
Institutional-Anomie Theory *

STEVEN F. MESSNER, University at Albany, SUNY

RICHARD ROSENFELD, University of Missouri — St. Louis

Abstract

This article examines the effects on national homicide rates of political efforts to
insulate personal well-being from market forces. Drawing upon recent work by Esping-
Andersen and the institutional-anomie theory of crime, we hypothesize that levels of
homicide will vary inversely with the "decommodification of labor." We develop a
measure of decommodification based on levels and patterns of welfare expenditures
and include this measure in a multivariate, cross-national analysis of homicide rates.
The results support our hypothesis and lend credibility to the institutional-anomie
perspective. The degree of decommodification is negatively related to homicide rates,
net of controls for other characteristics of nations.

Interest in explaining differences among nations in rates of crime and violence

is as old as the sociology of crime itself. The quantitative measurement of these
differences by the nineteenth-century moral statisticians Quetelet and Guerry
marks the beginning of scientific criminological inquiry (Beirne 1993). Marx
also refers to national crime data in the course of developing his critique of
the inherent flaws of capitalism. "There must be something rotten in the very
core of a social system," Marx (1859) writes, "which increases its wealth
without diminishing its misery, and increases in crimes even more rapidly
than in numbers" (New York Daily Tribune).
In recent decades, there has been a resurgence of interest in cross-national
criminological inquiry as reflected in a growing body of literature on the
structural determinants of homicide rates.l Although there are some

`Presented at the 46th annual meeting of the American Society of Criminology, 9-12
November, 1994, Miami, FL. We are grateful to the anonymous referees for helpful
comments on an earlier draft of this manuscript. Direct correspondence to Steven F. Messner,
Department of Sociology, SUNY-Albany, Albany, NY 12222.

© The University of North Carolina Press Social Forces, June 1997, 75(4):1393-1416
1394 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
discrepancies in this literature, the research is supportive of Marx's general
suspicion that fundamental features of the economic system affect societal
levels of crime. A finding that has emerged with remarkable consistency is
that high rates of homicide tend to accompany high levels of inequality in the
distribution of income (Krahn, Hartnagel & Gartrell 1986; Neuman & Berger
1988). The cross-national research also indicates that there are additional
features of a society's political economy, beyond that of income dispersion,
that are systematically related to homicide rates. For example, evidence
suggests that levels of homicide are associated with measures of the degree of
economic discrimination against social groups (Messner 1989) and measures
of the generosity of social welfare policies (Fiala & LaFree 1988; Gartner 1990,
1991).
The present article explores further the relationship between basic features
of the economic and political systems of societies and levels of criminal
homicide. Our specific focus is on the role of the market as a mechanism for
distributing the material resources for personal well-being. Markets play a vital
role in all capitalist societies, but, in some of these, physical survival and social
position are not as dependent on market considerations as in others. Esping-
Andersen (1990) has recently used the concept of the "decommodification of
labor" to refer to policies that promote reliance on, or insulation from, pure
market forces, and he has developed techniques for measuring this concept for
a small sample of advanced capitalist nations. In this research, we build upon
Esping-Andersen's work and propose a proxy measure of the
decommodification of labor that can be used in multivariate analyses for a
reasonably large sample of nations. We link the decommodification of labor
specifically with crime by drawing upon a recently proposed macrosocial
perspective in criminology: institutional-anomie theory. Our basic hypothesis
is that homicide rates and decommodification vary inversely: the higher the
level of political protection from the vicissitudes of the market, the lower the
national homicide rate.

Decommodification and Crime in Market Society

Esping-Andersen's work on decommodification is part of a long standing

intellectual and political tradition that emphasizes the importance of the
welfare state in stabilizing market societies (e.g., Bellah et al. 1992; Marshall
1950; Polanyi 1944; Tawney 1920). Decommodification refers in the most
general sense to the empowerment of the citizenry against the forces of the
market. Decommodified social policies permit actions and choices by citizens
— to get married, have children, seek higher education, engage in political
activity — that are, in principle, unconstrained by market considerations.
Decommodification frees people from the market.
There is some irony in this conception of the interrelation between the
market and the state. In the classical writings of Enlightenment thinkers such
 Institutional-Anomie Theory / 1395

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
as Adam Smith and David Hume, markets are depicted as social arrangements
that liberate individuals from the restraints imposed by traditional institutions:
the "free market" is an arena for the unfettered pursuit of self-interest (see
Hirschman 1992). That the market itself impedes the exercise of free choice is
a key intellectual claim of Marxist and social-democratic critics of modern
capitalism. That citizens possess social rights and entitlements that transcend
market considerations is the principal institutional claim made on the modern
capitalist economy by the welfare state (Marshall 1950).
The basic issue of accommodating the market to the functioning of other
social institutions is also relevant to the concerns of modernization theorists.
In Parsons's (1966) influential formulation, modernization entails the
increasing differentiation and interdependence of institutional subsystems.
The decommodification policies of the welfare state can be viewed from this
perspective as an equilibrating mechanism in highly differentiated societies?
In general, the concept of decommodification has been highly useful in
attempts to understand the institutional functioning of modern market
societies.
In Esping-Andersen's usage, decommodification refers to the granting of
services and resources to citizens as a matter of right, thereby reducing their
reliance on the market for sustenance and support (1990). It entails
"emancipation" of citizens from the market in the most fundamental sense:
"citizens can freely, and without potential loss of job, income, or general
welfare, opt out of work when they themselves consider it necessary" (23).
Decommodification involves considerably more than a society's level of
expenditure on social welfare policies and programs. It reflects the quality as
well as the quantity of social rights and entitlements. Three essential
dimensions of entitlements are encompassed by decommodification: ease of
access to them, their income-replacement value, and the range of social statuses
and conditions they cover (Esping-Andersen 1990).
It is useful to think of a continuum of decommodified social policies along
which societies may be arrayed. Near one end would be societies with highly
decommodified policies, defined by nearly universal and nonconditional
entitlements, with benefit levels close to average market incomes 3 , covering
most or all of the relevant causes and conditions for assistance (e.g., sickness,
old age, unemployment, parenthood). Societies located near the other end
would display correspondingly weaker decommodification, reflected in strict
eligibility criteria for assistance, benefit levels well below prevailing market
incomes, and a narrow range of statuses and conditions meriting assistance.
At the extremes, fully decommodified policies would pay everyone a "social
wage" guaranteeing a socially acceptable level of earnings regardless of
market participation, and fully "commodified" policies would require strict
and complete dependence on the market for the resources necessary for
survival. Although no existing society can be found at either of the ideal-typical
extremes of the continuum, market societies are enormously variable with
1396 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
respect to the level and types of social assistance available to their populations
and the associated degree of decommodification.
Esping-Andersen does not relate the notion of decommodification directly
to the phenomenon of crime. However, Messner and Rosenfeld's institutional-
anomie perspective provides a plausible theoretical basis for predicting a
relationship between the levels of serious crime in market society and the extent
to which labor has been decommodified.
Institutional-anomie theory builds upon the classical anomie tradition,
attributing high levels of crime to interrelated cultural and structural dynamics
(Messner & Rosenfeld 1997; Rosenfeld & Messner 1994; cf. Chamlin & Cochran
1995). With respect to culture, a basic premise of the theory is that market
mechanisms and arrangements are conducive to anomic pressures. Markets
presuppose a materialistic goal-orientation among actors, and they promote
a calculating, utilitarian orientation towards social relationships (Hirschman
1992:139). When these orientations develop to an extreme degree, anomie in
the Mertonian sense is likely to ensue (Merton 1968). Goals — especially, but
not exclusively, materialistic ones — receive strong cultural support, whereas
the normative means regulating conduct begin to lose "their savor and their
force" (Merton 1964:226). In such an anomic environment, actors are
preoccupied with "outcomes" (Merton 1968:211), and the efficiency rather than
the legitimacy of the means governs behavior. The resulting attenuation of
normative controls is likely to lead to high levels of deviant behavior, including
crime 4
.

Institutional-anomie theory also assigns a critical role to structural

dynamics and, more specifically, to the balance among major social institutions
(e.g., the economy, the family, the polity). In all societies of any complexity, the
integration of social institutions is inherently problematic. This is because the
claims of the social roles associated with the respective institutions are
potentially contradictory and competing. For example, the demands and value-
orientations associated with economic roles are at times incompatible with
those of familial roles. The resolution of these conflicting claims in the course
of ongoing social interaction yields a distinctive pattern of institutional
relationships for the society at large — a distinctive "institutional balance of
power" (Messner & Rosenfeld 1997:68-79).
According to institutional-anomie theory, the form of institutional structure
that is particularly conducive to high levels of crime is one in which the
economy dominates the institutional balance of power. 5 Economic dominance
occurs when: (1) economic goals are assigned high priority in comparison
with noneconomic goals; (2) the claims of economic roles are typically honored
at the expense of those of noneconomic roles when conflicts occur; (3) social
standing tends to be more highly dependent on the performance of economic
roles than of noneconomic roles; and (4) the calculating, utilitarian logic of
the marketplace penetrates other institutional realms 6 .

Economic dominance leads, in turn, to high rates of crime via two

complementary processes. First, this type of institutional imbalance provides
 Institutional-Anomie Theory / 1397
fertile soil for the growth of the anomic cultural pressures associated with

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
market arrangements. This is because the noneconomic institutions that bear
primary responsibility for cultivating respect for social norms, such as families
and schools, are less capable of fulfilling their distinctive socialization
functions. Second, economic dominance weakens the external social controls
associated with institutional attachments. When the economy dominates the
institutional balance of power, noneconomic roles become relatively
unattractive. The result is relatively tenuous institutional engagement, weak
social control, and high rates of crime.
This concept of economic dominance in the institutional balance of power,
we propose, can be joined with Esping-Andersen's notion of
decommodification to derive a hypothesis about societal levels of crime. As
noted above, decommodification signals that the balance of institutional
power in market society has shifted from the economy toward the polity; it
implies that purely economic values and criteria are accommodated to
collective, political considerations. The market is not permitted to operate
according to its inherent logic alone but rather is subjected to political
restraints. In other words, the decommodification of labor can serve as an
indicator of one important dimension of the institutional balance of power —
the balance between the economy and the polity. A greater degree of
decommodification indicates a lower level of economic dominance in this
particular institutional interrelationship. Given the general logic of
institutional-anomie theory, then, the decommodification of labor should vary
inversely with societal levels of crime, including the most serious of crimes —
homicide.
We are aware of no previous efforts to join institutional-anomie theory with
the concept of decommodification in the analysis of cross-national variation
in homicide rates. Nevertheless, there is evidence consistent with our basic
hypothesis. Fiala & LaFree (1988) find that measures of welfare expenditures
are inversely related to child homicide rates in a cross-sectional analysis of
39 developed countries. Research by Gartner (1990) indicates that these
beneficial effects of welfare policies apply to homicide victimization more
generally. In a pooled, cross-sectional time-series analysis of 18 capitalist
societies observed at five-year intervals between 1950-80, Gartner discovers
significant negative effects of indicators of welfare spending on homicide rates
for all age-sex-specific groups. Finally, Pampel and Gartner (1995) have
examined the effects of a scale of "collectivism" on homicide rates in a cross-
national analysis of the same 18 advanced capitalist societies studied by
Gartner. The collectivism scale combines Esping-Andersen's
decommodification index with indicators of corporatism, consensus
government, Leftist political rule, and "governability" (the absence of violent
political conflict). The collectivism scale has negative main effects on homicide
rates, and it reduces the positive effect of the relative size of the youthful
population on annual changes in homicide rates.
1398 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
These studies lend plausibility to our general hypothesis, but they are
limited in important respects. As noted, the research by Fiala & LaFree (1988)
is restricted to child homicide (see also Gartner 1991), while the results of the
research by Gartner (1990) and Pampel & Gartner (1995) pertain only to the
experiences of the 18 most advanced capitalist nations (albeit with
observations for multiple time periods). The present study goes beyond these
earlier efforts by developing a theoretically grounded measure of
decommodification that can be employed in a multivariate analysis of overall
homicide rates for a reasonably large sample of contemporary nations.
Before describing our measure of decommodification, it is important to
confront a key conceptual issue. Esping-Anderson (1990) explicitly focuses
his analyses on "advanced" nations and, more specifically, on the "advanced
capitalist democracies" (1-2). He does so because these are the nations with
the economic and political capacity to achieve a high degree of
decommodification. The advanced capitalist democracies have sufficiently
large economic surpluses to enable appreciable segments of the population
to withdraw from the market, and they have political structures that are
conducive to the emergence of class coalitions supportive of
decommodification.
A legitimate question to raise, therefore, is whether the very concept of
decommodification can be applied to a heterogeneous sample of nations at very
different levels of development. We base our analysis on the assumption that
decommodification is a meaningful property with which to describe industrial
and industrializing nations in general because the provision of basic social
security is a concern in virtually all such societies. This assumption is consistent
with the underlying rationale for the comparative data sets on social transfers
published by the International Labour Office (ILO), which serve as the source
for our proxy measure of decommodification. The ILO observes that social
security has become an important feature of the economy for member states in
"nearly every country" (ILO 1992:3). To assess the applicability of
decommodification to industrial nations generally, we have examined whether
the effects of our decommodification measure on homicide rates differ
significantly for the 18 nation subsample studied by Esping-Andersen and the
remaining subsample of nations. As reported below, comparable effects are
observed across these subsamples, which is consistent with our assumption
that the concept of decommodification can be usefully applied to lesser-
developed nations as well as to the advanced capitalist democracies.

Data and Methods

MEASURING DECOMMODIFTCATION

Esping-Andersen's measure of decommodification encompasses three

primary dimensions of the underlying concept: the ease of access to welfare
 Institutional-Anomie Theory / 1399

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
benefits, their income-replacement value, and the expansiveness of coverage
across different statuses and circumstances. A complex scoring system is used
to operationalize each of these dimensions of decommodification for the three
most important social welfare programs: pensions, sickness benefits, and
unemployment compensation (1990). This scoring system reflects the
"prohibitiveness" of conditions for eligibility, the disincentives for and
duration of entitlements, and the degree to which benefits replace normal
levels of earnings (1990). The indices for these three types of social welfare
programs are then aggregated into a combined index reflecting the overall
decommodification characteristic of a given nation's social welfare system.
Esping-Andersen is able to operationalize decommodification in this
unique way by using highly detailed information on social policies from an
original data source — the Svensk Socialpolitik i International Bleysning (the
SSIB data files). The data were collected at the Swedish Institute for Social
Research over an eight-year period, beginning in 1981, through contacts with
numerous officials in government departments and statistical offices in
different nations (1990). Although Esping-Andersen's approach to measuring
decommodification is highly appealing from a theoretical standpoint, his
measure has been constructed for only 18 capitalist nations. The explanatory
scope using this measure will therefore apply to only the most highly
developed market societies, and the small size of the resulting sample will
seriously limit the possibilities for including decommodification in
multivariate statistical analyses. Moreover, the procedures employed by
Esping-Andersen to construct his index require data on social policies that
are not available in published sources.
To overcome these limitations, we have developed a proxy measure of
decommodification for an appreciably larger sample of nations (maximum
N = 45). The proxy measure is based on data compiled by the International
Labor Office (ILO) on the financial operations of national social security
systems. These data include information on absolute and relative levels of
expenditures for social security programs, on funding sources for these
programs, and on the distribution of the expenditures across different program
types (e.g., unemployment benefits, family allowances, work-related injuries).
Our approach is based on the assumption that general expenditure patterns
reflect the underlying logic of social welfare systems. Consequently, indicators
of these general patterns are likely to be correlated with the more refined and
theoretically informed measure of decommodification developed by Esping-
Andersen.
We have examined the relationships between Esping-Andersen's
decommodification index and a variety of indicators of social security
expenditures in the 1980s for the 18 advanced capitalist nations included in
the SSIB data files. The indicators encompass four important features of the
social security systems: (1) the priority given to social welfare spending, as
reflected in expenditures as a percent of total gross domestic product (ILO
1992, Table 3); (2) the generosity of social welfare spending, as reflected in
1400 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
average annual expenditures per head of population in U.S. dollars (Table 5);
(3) the financing of social security systems, as reflected in the percentage of
total receipts according to origin (Table 8); and (4) the range of entitlements,
as reflected in the percentage distribution of benefit expenditures across
different program types (Table 10)?
Esping-Andersen's decommodification index is strongly associated with
three of the indicators of expenditure patterns. Expenditure levels as a percent
of GDP and average annual expenditures per capita exhibit large positive
correlations with the decommodification index (.75 and .81, respectively). An
indicator of the distribution of expenditures across program types — the
percent of total benefit expenditures allocated to employment injuries — also
yields a sizeable correlation with the decommodification index: r = -.67. The
negative sign of this coefficient is theoretically meaningful because it implies
that a large share of welfare benefits is not available to all citizens as a basic
entitlement but, rather, is contingent on participation in the labor market. Only
employed workers can receive benefits for employment injuries. Welfare
systems that impose this type of restriction on access are therefore less
decommodifying than those covering a wider range of circumstances
independent of market participation (e.g., programs such as family allowances
and maternity benefits).
Analogous results are obtained in a principal components factor analysis
of the decommodification index and the full range of expenditure indicators.
The decommodification index, average annual benefits per household,
expenditures as a percent of GDP, and the percent of benefit expenditures
allocated to employment injuries all load highly on the same factor. These
four measures thus exhibit a high level of shared variance, suggesting that
they converge on a common, underlying dimension 8 .

Given these results, we have computed a proxy decommodification index

by summing the z-scores for the three indicators of expenditure patterns that
are highly intercorrelated with Esping-Andersen's index and that cluster along
the same dimension. 9 The resulting composite index is highly correlated with
Esping-Andersen's original decommodification measure: r = . 84. Our proxy
measure thus exhibits strong predictive validity for the 18 nations with data
from the Swedish data source, the SSIB.
It is possible to compute the decommodification proxy measure for a fairly
large sample of nations (a maximum of 55) using ILO (1992) data for the mid-
1980s. The presence of missing data on other variables (explained below) limits
the sample size for the analysis to 45 cases. The proxy measure of
decommodification yields a respectable level of reliability for this sample: alpha
coefficient = .702.

DEPENDENT VARIABLE

The dependent variable for the analysis is the homicide rate per 100,000
population as reported in the World Health Organization's (WHO) World
 Institutional-Anomie Theory / 1401

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
Health Statistics Annual (various years). WHO defines homicide as death by
injury purposely inflicted by others. One limitation of the WHO data on
homicide is that underdeveloped nations, especially those in Africa and Asia,
are not well represented in this data source (Krahn, Hartnagel & Gartrell 1986).
Therefore caution should be exercised in generalizing our findings to the larger
population of nations. In addition, the WHO data on homicide may be biased
because they necessarily exclude deaths with undetermined cause, some of
which may be homicides. Nevertheless, Kalish (1988) argues that the WHO
data serve as the best source of information on homicide for international
comparisons because they are "based on an actual count of deceased persons"
and therefore are not susceptible to biases resulting from intercountry
differences in the treatment of "attempted homicides. 10 "

To minimize the effect of random yearly fluctuations in homicide levels,

we follow the conventional practice of computing multi-year averages (Kick
& LaFree 1985; Krahn, Hartnagel & Gartrell 1986). The averages refer to the
1980-90 period, or in cases with missing data, to the subset of years within
that period for which data are available. We employ an extended time period
for measuring homicide (a maximum of 11 years) because decommodification
is conceptualized as a basic structural feature of societies that is not likely to
vary in a meaningful way over the short run.
Examination of the univariate distribution for homicide rates reveals
considerable skewness. The value for the highest nation (Colombia, 41.2) is
approximately 6.5 times the mean value for the sample at large (6.27). We
accordingly convert homicide rates to natural logarithms to reduce skewness
and induce homogeneity in error variance." Although the log transformation
successfully reduces the overall degree of skewness in the homicide distribution,
a possible outlier remains at the lower tail of the distribution. The logged value
for Syria (-3.00; untransformed value = .05) is considerably smaller than the
value for the nation (Egypt) with the next smallest value (-.38; untransformed
value = .69). The possibility that this case is an atypical one that distorts the
regression estimates is addressed in the statistical analysis.

CONTROL VARIABLES

Data have also been collected on additional characteristics of nations to serve

as controls. Previous comparative research on homicide typically includes
some combination of indicators of the general economic well-being of national
populations and of demographic structure (see LaFree & Kick 1986, and
Neuman & Berger 1988, for comprehensive reviews). Consistent with this
research, we have collected information on the following socioeconomic and
demographic characteristics of nations: 12 gross national product per capita
in U.S. dollars; infant mortality rate (under age 1); life expectancy at birth;
percent of the population over 64 years of age; average annual population
growth 1980-85; percent of the population urban; males per 100 females.
1402 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
GNP per capita and the sex ratio are logged transformed to reduce the
effect of cases with extreme values. With the exception of the measure of
population growth, the time of measurement for these socioeconomic and
demographic characteristics is 1985, the midpoint of the 1980-90 period, or
the closest year with available data. The sources for the sex ratio and percent
over 64 years of age are the Population Reference Bureau's (1987) Population
Data Sheet and United Nations' (various years) Demographic Yearbook. The other
measures are taken from the World Bank's (1987) World Development Report.
In the comparative homicide literature, age structure is typically measured
by an indicator of the relative size of the young population. This approach is
based on the assumption that the young population is at a relatively high
risk of offending (see Krahn, Hartnagel & Gartrell 1986). We employ a measure
of the relative size of the elderly population because this variable has been
identified as a key determinant of welfare expenditures (Wilensky 1975; see
also Pampel, Williamson & Stryker 1990), and because the elderly are likely
to have low homicide offending rates (thereby creating the possibility of
spuriousness in the bivariate relationship between homicide rates and social
welfare measures). Not surprisingly, measures of the youthful population and
the elderly population are strongly correlated. The correlation between percent
less than 15 years of age and the percent over 64 for the sample of 45 nations
is -.90. Thus, the results of our regression analyses are highly similar if the
former measure of age structure is substituted for the latter.
Several of the control variables are strongly intercorrelated. To simplify
the regressor space and lessen the problem of multicollinearity (Land, McCall
& Cohen 1990), a principal components analysis has been performed on these
socioeconomic and demographic variables. The results reveal that all the
measures except the sex ratio cluster along a single dimension (the eigenvalue
for the principal component is 4.2, and the variance explained is 71%). The
positive pole of this dimension reflects socioeconomic development, as
indicated by high life expectancy, high GNP/capita, low infant mortality,
relatively large elderly populations, slow population growth, and high levels
of urban development. These measures have been combined into a
"development index" using the loadings from the principal components
analysis as weights.
Two additional variables are also relevant to the analysis on both empirical
and theoretical grounds. As noted earlier, previous cross-national research
indicates that economic inequality is one of the more important structural
correlates of homicide rates. It seems likely that decommodification and
economic inequality are inversely related to one another. Decommodification
should reduce the dispersion in incomes as well as lessen reliance on the
market for economic well-being. However, to the degree that decommodified
social welfare practices reflect the broader balance of power between the polity
and the economy, as suggested by institutional-anomie theory,
decommodification is expected to have an effect on the level of crime
independent of its relationship with inequality.
 Institutional-Anomie Theory / 1403
FIGURE 1: Scatterplot of Homicide Rates and Decommodification Scores for Esping

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
Andersen's 18 Nation Sample

2.5

9
= 1.5

u 1 • Finland

• Canada
U • Italy
• 0s Cealand •Belgium
0 •Auslda
S
• Switzerland • Sweden
Franceja
p m Denmark * 0oi '°' a Y
0
• OK kelhedands
Ireland • •Japan

-0.5

10 15 20 25 30 35 40
De-Commodification Score

We include two measures of economic inequality. One is the commonly

used Gini coefficient of household income distribution. The primary data
source for this measure is Hoover (1989), supplemented in a few cases with
data from Krahn, Hartnagel & Gartrell (1986). Unfortunately, the available
data for the Gini coefficient of income inequality refer to the period "circa 1969"
(Hoover 1989), which is earlier than our time period of interest (1980-90).
Although income distribution is_ a reasonably stable feature of societies (Muller
1988), the measurement error associated with this time lag could attenuate
the effects of income inequality in our analyses.
The second measure reflects an ascriptive form of economic inequality:
economic discrimination against social groups. The specific measure is an
ordinal rating scale based on expert judgments about the extent to which
groups experience objective economic disadvantages that are attributable to
deliberate discrimination. A nation's score represents the most extreme level
of economic discrimination experienced by any minority group in that nation.
The economic discrimination index is based on information contained in the
"Minorities at Risk" data file compiled by Gurr (Gurr & Scarritt 1989).
Data for all variables except the Gini coefficient and economic
discrimination are available for the maximum sample of 45 nations (see
Appendix A). In the analysis of this sample, mean substitution is applied to
cases with missing data on the two measures of economic inequality. We also
conduct parallel analyses on the subsample of nations with complete data
across all variables to assess the sensitivity of the results to mean substitution.
1404 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
TABLE 1: OLS Regressions of the Average Homicide Rate, 1980-90, on
the Proxy Decommodification Index and Controls

Equation

1 2 3 4
Independent variable

Development index -.104* -.017 -.097* -.097*

[-.342] [-.057] [-.355] [-.3631

Gini coefficient of 3.802 3.274 1.474 1.432

income inequality [.2461 [.212] [.1081 [.108]

Economic discrimination .213* .172 .151* .146

index [ .2261 [ .1821 [ .1801 [ .1761

Sex ratio (in) -5.709* -5.754* -5.509* -5.551

[-.248] [-.250] [-.268] [-.157]

Decommodification index - -.209* - .161* -.161*

[-.386] [-.334] [-.3371

RZ .357 .402 .547 .548

Adjusted R 2 .293 .326 .487 .480

N 45 45 44 39
a Standardized regression coefficients are reported in brackets. Homicide rates are log (In)
transformed.
* Unstandardized regression coefficient is at least 1.5 times its standard error.

Results

Before turning to the multivariate analysis of homicide rates for the full sample
of nations, it is instructive to examine the bivariate relationship between levels
of homicide and Esping-Andersen's original decommodification index for
the 18 advanced capitalist nations with available data. As expected, the
Pearson correlation coefficient is inverse and statistically significant at the
.05 level despite the small sample: r = -.48. 13 Nations with greater
decommodification scores thus tend to have lower homicide rates.
Figure 1 presents the scatterplot for these two variables. A striking feature
of the scatterplot is the distinctiveness of the U.S. Even with homicide rates
expressed in natural logarithms, the rate for the U.S. is unusually high. The
U.S. also has a very low decommodification score (the second from the bottom),
suggesting that this case plays a major role in producing the observed inverse
association. 14 It is thus important to determine whether decommodification
exhibits the predicted association with homicide rates in a larger sample of
nations not as sensitive to the influence of any single case.
 Institutional-Anomie Theory / 1405

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
The bivariate relationship between the decommodification proxy measure
and homicide rates for the sample of 45 nations is highly similar to that
observed with the original measure in the smaller 18-nation sample: r = -.52
(see Appendix B). Of course, the nations in this larger sample are quite
heterogeneous, raising the possibility that at least some of the simple
association between decommodification and homicide is confounded with
the effects of other structural characteristics of nations.
To assess this possibility, we turn to the multivariate analyses. Table 1
reports the estimates from four multiple regressions. The first column of the
table provides a baseline model that includes only the control variables.
Consistent with past research (Krahn, Hartnagel & Gartrell 1986; Messner
1989), the two indicators of economic inequality — the Gini coefficient of
income dispersion and the index of economic discrimination against social
groups — yield moderate positive effects on homicide rates, although the
coefficient for the Gini coefficient does not quite reach statistical significance.
Both the development index and the sex ratio are negatively related to homicide
rates. The negative coefficient for development is consistent with the
"modernization thesis" on crime, which predicts a decline in rates of violent
crime with greater urbanization and industrialization (Gurr 1989; Shelley
1981). This finding is similarly compatible with arguments that development
is associated with reduced opportunities for the kinds of interpersonal contacts
that lead to homicide (LaFree & Kick 1986). In addition, the negative effect of
development probably reflects demographic factors captured by the composite
index (an elderly population and low population- growth rates).
The negative association observed for the sex ratio is counter-intuitive. It
indicates that low homicide rates tend to be found in nations with large
numbers of males relative to females. This association is contrary to individual-
level research on criminal violence, which shows higher levels of victimization
and offending for males, but it is compatible with arguments by Messner &
Sampson (1991). They suggest that low sex ratios may promote family
arrangements that are conducive to crime and that counterbalance the crime-
reducing compositional effect at the macro-level of relatively small male
populations.
In equation 2, the decommodification proxy index is added to the baseline
model. The results are consistent with theoretical predictions.
Decommodification exhibits a significant, negative relationship with homicide
rates net of the control variables. The standardized coefficient (B = -.386) is
moderately strong and is the largest for any of the predictors in this model. 15
Comparing across equations 1 and 2, including the decommodification
measure in the model reduces slightly the effect of the Gini coefficient and the
economic discrimination index, while the coefficient for the sex ratio remains
virtually unchanged. The most dramatic change is observed for the
development index, the effects of which become trivial in equation 2.
Equation 3 estimates the same model as equation 2 for the subsample of
44 nations without Syria, an outlier on homicide. Excluding this case raises
1406 / Social Forces 75:4, June 1997
TABLE 2: Effects of the Proxy Decommodification Index on the Average

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
Homicide Rate, 1980-90, across Alternative Specifications

Samples
A B C
Model
1. Excluding the -.227* -.258* -.232*
development index [-.419] [-.535] [-.487]
2. Excluding the -.225* -.166* -.159*
Gini coefficient [-.414] [-.345] [-.333]
3. Excluding the economic -.240* -.187* -.187*
discrimination index [-.443] [-.387] [-.392]
4. Excluding the -.208* -.160* -.175*
sex ratio (in) [-.383] [-.331] [-.367]
N 45 44 39

Standardized regression coefficients are reported in brackets. Homicide rates are log (In)
transformed.
b Sample A is the maximum sample yielded with mean substitution for income inequality
and economic discrimination; sample B excludes Syria; sample C is the sample yielded
with listwise deletion of cases with missing values.
* Unstandardized regression coefficient is at least 1.5 times its standard error.

the explanatory power of the model appreciably (compare the adjusted Res)
and increases the observed effect of the development index in comparison with
the previous model. The effect for the decommodification index is lessened
slightly in equation 3 in comparison with equation 2, but it remains significant
and moderately strong.
In the final equation (equation 4), the analysis is repeated without mean
substitution, which is required in the analysis of the larger samples for cases
with missing values on the Gini coefficient and the economic discrimination
index. The results for the theoretically strategic measure prove to be very similar
to those in the previous equations. Nations with high scores on the
decommodification index tend to have low homicide rates, net of the effects
of the other variables in the models. 16
To assess further the sensitivity of the effects of the decommodification
proxy to alternative specifications, we re-estimate the regressions deleting each
of the control variables, one at a time. These analyses are performed on the
full sample of nations, the subsample without Syria, and the subsample
without mean substitution. Table 2 reports the regression coefficients for the
decommodification proxy across these alternative specifications for the
respective samples. The results reveal a highly robust pattern. Consistent with
 Institutional-Anomie Theory / 1407

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
our theoretical argument, the coefficients for decommodification are negative,
significant, and moderately strong in all specifications.
Our findings for other predictors of homicide are less stable, as shown in
Table 1, although the overall patterns of relationships are theoretically
meaningful. The instability in the estimates for the development index may
reflect problems of multicollinearity. Despite our efforts to simplify the
covariance structure of predictors through principal components indexing, a
troublesome degree of multicollinearity remains for this variable. 17 We also
suspect that the rather unimpressive effects of income inequality are
attributable, at least in part, to measurement error (noted earlier) resulting from
the time lag between the measurement of inequality and homicide rates.
Finally, we consider the possibility that the observed effect of
decommodification applies only to the nations originally studied by Esping-
Andersen and not to other nations in the sample. We do so by creating a
dummy variable coded 1 for nations in Esping-Andersen's sample and 0 for
the other nations, and by constructing a product term for this dummy variable
and the decommodification proxy. The product term is then added to the
regression models for each of the three samples reported in Table 1, along with
the constituent terms. In all cases, the coefficients for the product term fail to
attain statistical significance. This finding suggests that the net effect of
decommodification on homicide rates for the nations originally studied by
Esping-Andersen is comparable to the effect observed for the other nations in
the analysis.

SUMMARY AND CONCLUSIONS

In this article, we have addressed the consequences of political restraints on

the market for societal levels of lethal violence. Drawing upon recent work on
the nature of welfare state regimes, the institutional-anomie theory of crime,
and previous cross-national analyses of homicide, we have derived a specific
hypothesis that overall levels of homicide will be lower in capitalist societies
that have decommodified labor by reducing dependence on the market for
personal well-being. This hypothesis has been tested in cross-national
regression analyses using a theoretically grounded measure of
decommodification, along with relevant controls.
The results support our basic hypothesis. Controlling for a wide range of
other structural characteristics of nations, the decommodification measure
exhibits a significant negative effect on homicide rates. This effect is moderately
strong and is robust across alternative specifications and varying subsamples.
Our analyses thus replicate the findings of previous cross-national studies of
homicide that have incorporated measures of social welfare spending, and
they indicate that these earlier findings are generalizable to larger samples of
nations and to more inclusive measures of homicide.
Our research also lends credibility to the theoretical perspective informing
the analysis — the institutional-anomie theory of crime — and helps to
1408 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
empirically distinguish this perspective from more conventional stratification-
based accounts of variation across societies in the level of homicide. Although
these perspectives are in many respects complementary, there are two
important distinctions between them.
First, institutional-anomie theory broadens the structural focus of
traditional economic stress or deprivation perspectives by directing attention
to aspects of the economic organization of market societies beyond the
stratification system, and to the interplay of the economy and other social
institutions. In this article, we have restricted our attention to restraints
imposed on market economies by the political system. Additional research is
needed on the role of other institutions, such as the kinship, religious, and
educational systems, in fostering or curbing crime in market societies. Such
an expanded institutional focus might help to account for nations such as
Japan and Ireland, which have exceptionally low levels of homicide among
the developed countries, and yet only moderate scores on the
decommodification index (see Figure 1). In the case of Japan, we would
attribute the low rate of homicide to the prominent role of the family and its
restraining influence on the anomic forces emanating from the market (cf. Adler
1983). It would seem promising to pursue in further research the
corresponding role of organized religion in a nation such as Ireland.
A second difference between institutional-anomie theory and more
traditional economic perspectives on societal levels of crime involves the
significance assigned by institutional-anomie theory to cultural orientations,
which ostensibly operate in tandem with features of economic stratification.
It is not possible to document any such cultural effects with the existing data
because valid and reliable measures of culture are not available for cross-
national analysis, but it is interesting to note that our development index
captures to some extent levels of economically induced deprivation via the
indicators of overall economic resources (GDP/capita) and life chances (infant
mortality and life expectancy). The effects of decommodification on homicide
rates net of the development index are thus at least suggestive of the kinds of
cultural dynamics postulated by institutional-anomie theory. 18 Nevertheless,
further research is clearly needed to clarify the precise nature of the social
mechanisms linking the welfare state, institutional balance, and levels of crime
and violence in market societies.
We close with a final comment on the practical implications of our
analysis. It is hardly an exaggeration to claim that the current era is one of
profound social change in the history of capitalism. With the fall of the Soviet
empire and with the economic reforms taking place in the People's Republic
of China, a much larger segment of the world's population is exposed to
market arrangements. Moreover, in the U.S. and other advanced capitalist
societies, there have been growing concerns about the scope, cost, and even
the very logic of the welfare state (e.g., Stevenson 1995; Whitney 1995). If the
findings reported here are sustained in subsequent research, then proposals
to substantially reduce social welfare spending and deregulate market
 Institutional-Anomie Theory / 1409

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
economies should be considered with due regard for unintended social
consequences, including possibly higher rates of criminal violence.

Notes
1. See, for example, Archer & Gartner (1984); Avison & Loring (1986); Braithwaite &
Braithwaite (1980); Conklin & Simpson (1985); Fiala & LaFree (1988); Gartner (1990);
Groves, McCleary & Newman (1985); Kick & LaFree (1985); Krahn, Hartnagel & Gartrell
(1986); Krohn (1976); LaFree & Kick (1986); Messner (1980, 1982, 1989); and Wellford
(1974).
2. From a more critical perspective, the function of the welfare state is to stabilize the
capitalist economic order and legitimate class rule (Habermas 1989; O'Connor 1973). See
Esping-Andersen (1990:12-14) for a discussion of the similarities between structural
functionalist and Marxist approaches to the modernization process.
3.Esping-Andersen (1990) defines the market replacement value of social entitlements as
the difference between benefit levels and "normal earnings or the standard of living
considered adequate and acceptable in the society" (47).
4. Merton directs attention to the contemporary U.S. in his discussion of the nature and
sources of anomie, as do Messner and Rosenfeld (1997) in their discussion. However, as
Gouldner (1970) suggests, Merton's arguments (and by extension Messner and Rosenfeld's)
can be applied more generally to societies dominated by a market economy, i.e., to
"bourgeois utilitarian societies." In Gouldner's words: "The 'almost exclusive concern with
outcomes' to which Merton refers is a distinctive characteristic of utilitarian culture; it is
not an aberration of utilitarian society but its normal cultural emphasis" (1970:68; see
also Rosenfeld & Messner 1997).
5.Messner and Rosenfeld's discussion of economic dominance in the institutional balance
of power raises themes similar to those contained in Currie's analysis of a "market society"
as distinct from a "market economy." According to Currie (1991:255), a market society
refers to "one in which the pursuit of private gain becomes the organizing principle of all
areas of social life — not simply a mechanism that we may use to accomplish certain
circumscribed economic ends."
6.John Gagnon comments on the general tendency in market societies for the logic of the
economy to permeate discourse over an ever widening range of social phenomena, both
inside and outside academic social science: "Within the social sciences there has been a
100-year struggle to extend the reach of economic metaphors and analyses to include all
aspects of mental and social life. Outside the social sciences, in practical society, a parallel
attempt to subject all forms of conduct to the discipline of commodification and pricing
has become part of the normal order" (Gagnon 1994:1078).
7.The categories for the origin of social security receipts are: contributions from insured
persons, contributions from employers, special taxes, state participation, other public
participation, income from capital, and "other" receipts. The categories for the distribution
of benefits are: sickness-maternity, employment injuries, pensions, unemployment, and
family allowances..
8. The factor loadings for the decommodification index, average annual benefits per
household, expenditures as a percent of GDP, and the percent of benefit expenditures
allocated to employment injuries are .90, .86, .82, and -.82 respectively. Although these
loadings are reasonably high, some unshared variance obviously remains. This probably
reflects the limitations of expenditure data as indicators of the "theoretical substance of
1410 / Social Forces 75:4, June 1997
the welfare state," as well as random measurement error. See Esping-Andersen (1990:18-

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
21).
9. In the construction of the index, we reverse the polarity of the item on the distribution
of benefits by using the percentage of benefits distributed to categories other than
employment injuries (i.e., 100 — "the percent distributed to employment injuries"). This
ensures that all items are scored in a consistent direction.
10.See Bennett and Lynch (1990), Huang (1993), Kalish (1988), and Messner (1992) for
discussions of the quality and comparability of international crime statistics.
11.We plotted the residuals from an OLS regression of the untransformed homicide rate
against predicted Y values. The scatterplot conforms to a classic heteroskedastic "fan"
pattern, with greater variance in residuals for higher predicted Y values (Hamilton 1992:117).
The plot for residuals with the transformed homicide rates reveals a more homoskedastic
distribution. The transformation of homicide rates implies that the modeled relationship
between homicide and other variables is curvilinear with respect to the original metrics.
For a discussion of the interpretation of regression coefficients under different
transformations; see Hamilton (1992:145-82).
12.Three additional controls were considered in preliminary analyses but were excluded
because they consistently failed to yield appreciable net associations with homicide rates:
population size, population density, and Gurr's institutionalized democracy index (ICPSR
1990).
13. Although tests of statistical significance are not technically applicable given the
nonrandom nature of the sample, we nevertheless follow the common practice of reporting
significance as a rule-of-thumb to identify nontrivial relationships. The criterion for
significance in the regression analysis is a t ratio of 1.5, which corresponds approximately
to the .05 level (one-tailed test).
14.With the U.S. removed from the sample, the correlation between the homicide rate and
decommodification score drops to r = -.25. Although still in the expected direction and
moderately strong, this correlation is not statistically significant for the remaining 17 nation
sample. The outlier status of the U.S. in the homicide distribution depicted in Figure 1 is
consistent with the hypothesis derived from institutional-anomie theory that a society
characterized by economic dominance will have unusually high levels of serious crime.
See Messner and Rosenfeld (1997). More generally, the proposition that the U.S. is distinctive
on a number of social and cultural dimensions when compared with other advanced
industrial nations is part of the thesis of "American exceptionalism." See Lipset (1996)
for a recent statement.
15.In this larger sample of nations, removing the U.S. has a minor impact on the observed
relationship between the decommodification proxy and homicide rates. The association is
still moderately inverse and statistically significant. The unstandardized regression
coefficient without the U.S. is -.200; the B coefficient is -.361.
16.We computed values of Cook's D for each of the three subsamples in Table 1 to search
for influential cases in the estimation of regression parameters. No case reaches the generally
accepted critical value of "1" on this diagnostic statistic. As a further check for a
disproportionate impact of a single case or a small number of cases on the parameter
estimates, we performed a robust regression on the full sample using the Huber iteratively
reweighted least squares technique (Hamilton 1992:183-216). The robust WLS results for
the decommodification index are virtually identical to those obtained through OLS: the
unstandardized coefficients are -.217 and -.209, respectively.
 Institutional-Anomie Theory / 1411

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
17.Appendix C reports variance inflation factors (VIF) for the independent variables in
the analyses of the respective samples. The values for the development index in all instances
exceed the conventional threshold for high multicollinearity of 4.0. This problem is
particularly severe in the analysis of the 39 nation sample without mean substitution. The
VIFs for the decommodification proxy, however, are always below the conventional criterion.
18. Similarly, in her analysis of the relationships between family structure, welfare
expenditures, and child homicide, Gartner (1991) proposes a broader interpretation of
indicators of welfare practices, suggesting that "perhaps spending on social programs
should be thought of as an indicator of a cultural orientation or social ideology inhibiting
personal violence" (238).

References
Adler, Freda. 1983. Nations Not Obsessed with Crime. Rothman.
Archer, Dane, and Rosemary Gartner. 1984. Violence and Crime in Cross-National Perspective.
Yale University Press.
Avison, William R. and Pamela L. Loring. 1986. "Population Diversity and Cross-National
Homicide: The Effects of Inequality and Heterogeneity." Criminology 24:733-49.
Beirne, Piers. 1993. Inventing Criminology. SUNY Press.
Bellah, Robert N., Richard Madsen, William M. Sullivan, Ann Swidler, and Steven M. Tipton.
1992. The Good Society. Knopf.
Bennett, Richard R. and James P. Lynch. 1990. "Does Difference Make a Difference?
Comparing Cross-National Crime Indicators." Criminology 28:153-81.
Braithwaite, John and V. Braithwaite. 1980. "The Effect of Income Inequality and Social
Democracy on Homicide." British Journal of Criminology 20:45-53.
Chamlin, Mitchell B. and John K. Cochran. 1995. "Assessing Messner and Rosenfeld's
Institutional Anomie Theory: A Partial Test." Criminology 33:411-29.
Conklin, George H. and Miles E. Simpson. 1985. "A Demographic Approach to the Cross-
National Study of Homicide." Comparative Social Research 8:171-85.
Currie, Elliot. 1991. "Crime in the Market Society." Dissent (Spring):254-59.
Esping-Andersen, Gosta. 1990. The Three Worlds of Welfare Capitalism. Princeton University
Press.
Fiala, Robert and Gary LaFree. 1988. "Cross-National Determinants of Child Homicide."
American Sociological Review 53:432-45.
Gagnon, John H. 1994. "Review Essay: The Dismal Science and Sex." American Journal of
Sociology 99:1078-82.
Gartner, Rosemary. 1990. "The Victims of Homicide: A Temporal and Cross-National
Comparison." American Sociological Review 55:92-106.
_. 1991. "Family Structure, Welfare Spending, and Child Homicide in Developed
Democracies." Journal of Marriage and the Family 53:231-40.
Gouldner, Alvin W. 1970. The Coming Crisis of Western Sociology. Basic Books.
Groves, W. Byron, Richard McCleary, and Graeme R. Newman. 1985. "Religion,
Modernization, and World Crime." Comparative Social Research 8:59-78.
Gurr, Ted Robert. 1989. "Historical Trends in Violent Crime: Europe and the United States."
Pp. 21-54 in Violence in America. Vol. 1, edited by T.R. Gurr. Sage.
Gurr, Ted Robert, and James R. Scarritt. 1989. "Minorities at Risk: A Global Study." Human
Rights Quarterly 11:375-405.
1412 / Social Forces 75:4, June 1997
Habermas, Jurgen. 1989. "The Crisis of the Welfare State and the Exhaustion of Utopian

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
Energies." Pp. 284-99 in Jurgen Habermas On Society and Politics: A Reader, edited by
Steven Seidman. Beacon Press.
Hamilton, Lawrence C. 1992. Regression with Graphics: A Second Course in Statistics. Duxbury
Press.
Hirschman, Albert O. 1992. Rival Views of Market Society and Other Recent Essays. Harvard
University Press.
Hoover, Gary A. 1989. "International Inequality: A Cross-National Data Set." Social Forces
67:1008-26.
Huang, W.S. Wilson. 1993. "Are International Murder Data Valid and Reliable? Some
Evidence to Support the Use of Interpol Data." International Journal of Comparative
and Applied Criminal Justice 17:77-89.
ICPSR. 1990. "Polity II: Political Structure and Regime Change, 1800-1986: Summary."
Inter-University Consortium for Political and Social Research, Machine-Readable Datafile
(Study #9263; Principal Investigator: Ted Robert Gurr).
ILO (International Labour Office). 1992. The Cost of Social Security. International Labour
Office.
Kalish, Carol B. 1988. International Crime Rates. U.S. Department of Justice.
Kick, Edward L., and Gary D. LaFree. 1985. "Development and the Social Context of
Murder." Comparative Social Research 8:37-58.
Krahn, Harvey, Timothy F. Hartnagel, and John W. Gartrell. 1986. "Income Inequality and
Homicide Rates: Cross-National Data and Criminological Theories." Criminology
24:269-95.
Krohn, Marvin D. 1976. "Inequality, Unemployment, and Crime: A Cross-National
Analysis." Sociological Quarterly 17:303-313.
LaFree, Gary D., and Edward L. Kick. 1986. "Cross-National Effects of Developmental,
Distributional, and Demographic Variables on Crime: A Review and Analysis."
International Annals of Criminology 24:213-35.
Land, Kenneth C., Patricia L. McCall, and Lawrence E. Cohen. 1990. "Structural Covariates
of Homicide Rates: Are There Any Invariances Across Time and Social Space?"
American Journal of Sociology 95:922-63.
Lipset, Seymour Martin. 1996. American Exceptionalism: A Double-Edged Sword. Norton.
Marshall, T.H. 1950. Citizenship and Social Class. Cambridge University Press.
Marx, Karl. 1859. "Europe: Population, Crime, and Pauperism." New York Daily Tribune
(16 Sept.).
Merton, Robert K. 1964. "Anomie, Anomia, and Social Interation." Pp. 213-42 in Anomie
and Deviant Behavior, edited by Marshall Clinard. Free Press.
. 1968. Social Theory and Social Structure. Free Press.
Messner, Steven F. 1980. "Income Inequality and Murder Rates: Some Cross-National
Findings." Comparative Social Research 3:185-98.
. 1982. "Societal Development, Social Equality, and Homicide: A Cross-National
Test of a Durkheimian Model." Social Forces 61:225-40.
. 1989. "Economic Discrimination and Societal Homicide Rates: Further Evidence
on the Cost of Inequality." American Sociological Review 54:597-611.
. 1992. "Exploring the Consequences of Erratic Data Reporting for Cross-National
Research on Homicide." Journal of Quantitative Criminology 8:155-73.
 Institutional-Anomie Theory / 1413

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
Messner, Steven F., and Robert J. Sampson. 1991. "The Sex Ratio, Family Disruption, and
Rates of Violent Crime: The Paradox of Demographic Structure." Social Forces 69:693-
713.
Messner, Steven F., and Richard Rosenfeld. 1997. Crime and the American Dream. 2d ed.
Wadsworth.
Muller, Edward N. 1988. "Democracy, Economic Development, and Income Inequality."
American Sociological Review 53:50-68.
Neuman, W. Lawrence, and Ronald J. Berger. 1988. "Competing Perspectives on Cross-
National Crime: An Evaluation of Theory and Evidence." Sociological Quarterly 29:281-
313.
O'Connor, James. 1973. The Fiscal Crisis of the State. St. Martin's Press.
Pampel, Fred, and Rosemary Gartner. 1995. "Age Structure, Socio-political Institutions,
and National Homicide Rates." European Sociological Review 11:243-60.
Pampel, Fred, John B. Williamson, and Robin Stryker. 1990. "Class Context and Pension
Response to Demographic Structure in Advanced Industrial Democracies." Social
Problems 37:535-50.
Parsons, Talcott. 1966. Societies: Evolutionary and Comparative Perspectives. Prentice-Hall.
Polanyi, Karl. 1944. The Great Transformation. Rinehart.
Population Reference Bureau. Various years. World Population Data Sheet. Population
Reference Bureau.
Rosenfeld, Richard, and Steven R Messner. 1994. "Crime and the American Dream: An
Institutional Analysis." Advances in Criminological Theory 6:81-103.
1997. "Markets, Morality, and an Institutional-Anomie Theory of Crime." In The
Future of Anomie Theory, edited by Nikkos Passas and Robert Agnew. Northeastern
University Press.
Shelley, Louise I. 1981. Crime and Modernization. Southern Illinois University Press.
Stevenson, Richard W. 1995. "A Deficit Reins in Sweden's Welfare State." New York Times
(2 Feb.), Al, A5.
Tawney, R.H. 1920. The Acquisitive Society. Harcourt Brace.
United Nations. Various years. Demographic Yearbook. United Nations Publishing Service.
Wellford, Charles F 1974. "Crime and the Dimensions of Nations." International Journal of
Criminology and Penology 2:1-10.
Whitney, Craig R. 1995. "In Europe, Touches of Leanness and Meanness." New York Times
(1 January), p. E5.
Wilensky, Harold L. 1975. The Welfare State and Equality: Structural and Ideological Roots of
Public Expenditures. University of California Press.
World Bank. 1987. World Development Report. Oxford University Press.
World Health Organization. 1980-90. World Health Statistics Annual. World Health
Organization.
1414 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
APPENDIX A: Sample of Nations

Argentina Mauritius
Australia Mexico
Austria Netherlands
Belgium New Zealand
Brazil Norway
Canada Panama
Chile Peru
Colombia Portugal
Costa Rica Singapore
Denmark Spain
Dominican Republic Sri Lanka
Ecuador Sweden
Egypt Switzerland
El Salvador Syria
Finland Thailand
France Trinidad
Germany, Federal Republic United Kingdom
Greece United States
Guatemala Uruguay
Ireland Venezuela
Israel
Italy
Jamaica
Japan
Kuwait
 Institutional-Anomie Theory / 1415
APPENDIX B: Correlations and Univariate Statistics

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
I. Correlation Matrix
Y X1 X2 X3 X4 X5
(Y) Average homicide rate (in) 1.00 -.52 -.44 .51 .34 -.02
(XI) Decommodification index 1.00 .82 -.63 -.28 -.38
(X2)Development index 1.00 -.68 -.19 -.46
(X3)Gini coefficient 1.00 .34 .17
(X4)Economic discrimination index 1.00 .13
(X5)Sex ratio (In) 1.00
(N = 45)
II. Univariate Statistics
Mean Standard
Deviation
Average homicide rate (in) .97 1.29
Decommodification index .00 2.38
Development index .00 4.24
Gini coefficient .40 .08
Economic discrimination index 1.63 1.37
Sex ratio (In) 4.59 .06
(N=45)
1416 / Social Forces 75:4, June 1997

Downloaded from https://academic.oup.com/sf/article/75/4/1393/2233716 by University of Iowa Libraries/Serials Acquisitions user on 02 February 2021
APPENDIX C: Variance Inflation Factors for Fully Specified Models across
Different Samples of Nations

Sample

Maximum Excluding No Mean

Syria Substitution
Predictors
Development index 4.2 4.3 7.6
Gini coefficient 2.2 2.3 2.9
Economic discrimination index 1.2 1.2 1.3
Sex ratio (in) 1.4 1.3 2.6
Decommodification index 3.3 3.3 3.4

N 45 44 39