Public Disclosure Authorized

=
Public Disclosure Authorized

~~~~~25038
August 2001
Rural Development Strategy
Background Paper #2

The Role of Agriculture in

Public Disclosure Authorized

Economic Development and

Poverty Reduction

An Empirical and Conceptual

Foundation
Public Disclosure Authorized

Alexander Sarris

The World Bank

Rural Development Family
 Rural Development Strategy
Background Paper #2

The Role of Agriculture in

Economic Development and
Poverty Reduction

An Empirical and Conceptual

Foundation

Alexander Sarris

Work in Progress The World Bank

Rural Development Family
This report is one in the series of background studies prepared for the 2001 update of the World
Bank's Rural Development Strategy. This series was created to disseminate findings of work in
progress and to encourage the exchange of ideas among Bank staff and all others interested in
development issues. This paper carries the name of the author and should be used and cited
accordingly. The findings, interpretations, and conclusions are the author's own and should not
attributed to the World Bank, its Board of Directors, its management, or any member countries.

This paper has been reviewed for publication by the Rural Development Strategy Background
Paper Series Editorial Committee: Robert Thompson (Chair), Jock Anderson, Shawki Barghouti,
Csaba Csaki, Cees de Haan, Gershon Feder, Sushma Ganguly, and Kees Van Der Meer.
Contents
Acronyms ...................................................................... iv

Acknowledgement ..................................................................... iv

Foreword ....................................................................... v

Executive Summary ..................................................................... vii

1. Introduction ...................................................................... 1

2. How Does Agriculture Grow? ...................................................................... 3

3. Agriculture Development and Overall Growth ...................................................................... 7

4. Agriculture, Growth, and Poverty ..................................................................... 17

5. The Channels Via Which Agriculture Development Reduces Poverty ....................................... 21

6. Emerging Issues of Agricultural Development and Poverty Reduction ...................................... 31

6.1 Globalization ..................................................................... 31
6.2 Biotechnology ..................................................................... 32
6.3 Changing Roles of Public and Private Sectors ................................................................ 33

7. Modeling Public Policy Towards Agriculture in the Context of Growth and Poverty
Reduction ...................................................................... 35

8. Conclusions and Implications for the World Bank ..................................................................... 37

References ...................................................................... 43

Appendix: A Two Sector Model of Public Policy Towards Agriculture in the Context of Growth
and Poverty Reduction ..................................................................... 51
Theoretical Structure ..................................................................... 51
Empirical Simulations ..................................................................... 62
Appendix Tables ..................................................................... 65
Acronyms
ADLI Agriculture Demand Led Industrialization
CGIAR Consultative Group on International Agricultural Research
DES Dietary Energy Supply
GDP Gross Domestic Product
GMO Genetically Modified Organisms
IFAD International Fund for Agricultural Development
IPR Intellectual Property Rights
PFI Prevalence of Food Inadequacy
OECD Organization for Economic Cooperation and Development
TFP Total Factor Productivity
WDR World Development Report

Acknowledgements
The author is grateful for comments received during a presentation of this paper at World Bank
Headquarters on February 14, 2001. Additional thanks go to Harold Alderman, Jock Anderson,
Csaba Csaki, Gershon Feder, W. Martin, and Roland Schurmann for helpful observations and
comments to an earlier draft. Alan Zuschlag provided editorial support.

The preparation of this study has been supported by the Hellenic Consultant Trust Fund.

iv
Foreword
Poverty reduction is the overarching objective of the World Bank, and with 75 percent of the
world's poor living in rural areas, rural development is a key element in achieving progress in
this objective. At President Wolfensohn's request, the rural family has prepared a revised rural
development strategy, Reaching the Rural Poor. This has been done in close cooperation with
the regions and the other sectoral units active in the rural space. The objectives of the new
strategy are to revitalize the World Bank's activities in the rural areas by: (a) adjusting the
strategic framework; and (b) formulating a program of concrete and implementable actions.

The new rural development strategy addresses a rural situation which is different from the past,
and a rural population which confronts many new problems, especially the challenges and
opportunities facing the poor with regard to globalization. The new vision and articulation of a
development strategy builds upon the strengths of past efforts as well as incorporates new ideas
from other models. In this context, our priorities are geared to fulfill World Bank poverty
reduction objectives in the rural sector. We are convinced that the following critical components
of a rural development strategy will contribute most to accelerated growth in rural economies
and, co'nsequently, to measurable poverty reduction: crafting efficient and pro-poor policies and
institutions; facilitating broad-based rural economic growth; improving access to, and
management of natural, physical, and human assets; and reducing risk and vulnerability for the
rural poor.

A number of studies on both global and regional issues, as well as a broad portfolio analysis
were commissioned to support the development of the new strategy. These studies provided a
rich foundation for both the regional action plans and the corporate strategy. This study is one of
the selected number of background papers which have been published in the Rural Development
Strategy Background Paper Series to provide Bank staff and others with a more in-depth look at
some of the issues surrounding rural development, beyond what is covered by the strategy
document itself. This paper, and others in the series are available on line at:
www.worldbank.org/ruralstrategy. Additional information on obtaining other papers from this
series can also be found at the end of this report.

Robert L. Thompson
Director of Rural Development
The World Bank
vi
Executive Summary

The purpose of this paper is to provide a review of the issues related to the question of the role of
agriculture in promoting overall growth and poverty reduction. The major questions investigated
are the following. Under what conditions does additional or disproportional support for
agriculture will be both overall growth enhancing as well as poverty reducing? Is there a trade-
off between faster growth and poverty reduction in the context of agricultural development?
Given that overall support for agriculture as an efficient growth enhancing and poverty reducing
strategy can be justified, what are the most appropriate ways to support agriculture so as to
maximise the effectiveness of such support in generating growth and reducing poverty?
Sustained growth in an economy requires the continuous improvement in total factor
productivity (TFP), and this requires public expenditures for infrastructure and human capital
developments. At early stages of development, the economy needs an engine of growth. Such an
initial engine of growth can come from a variety of sources, such as the development of domestic
or export agriculture, industry, tourism, etc., and the key issue for development is what is done
with the increased incomes and savings that come about from the initial engine of growth.
While in the now developed countries it appears that the major stimulus to early growth was
industrial innovations, accompanied but not led by agricultural innovations, there have been
some developing countries in the recent past, notably India, China and Taiwan, where growth
was led by agricultural broad based productivity changes. In these countries it seems that the
major source of the demand for the increased product of the agricultural sector was domestic, as
there were substantial levels of initial poverty. Hence improved agricultural incomes directly led
to increases in the domestic demand for the larger quantities of food produced domestically.
Consequently the domestic terms of trade for the increased agricultural product did not decline
so as to negate the improvements in TFP. A substantial part of the cost of their agricultural
productivity improvement was shouldered by external donors, including the major contributions
of the international agricultural research centers, that are members of the CGIAR. Finally, it
appears that given the substantial rural population densities in these countries, the cost per
agricultural producer of improving agricultural productivity was relatively low, as there was no
need for experimentation in too many climatically different locations.
The situation may well be different in many of the late developing countries, such as those in
many parts of Africa, in the sense that the cost per beneficiary of agricultural productivity
improvement may be high because of low farm population densities. At the same time that
overall donor support is declining, necessitating a larger contribution to domestic productivity
improvement by the state. Can agriculture play a leading role in the settings of the late
developing countries, and should cash strapped governments devote a substantial share of their
meager resources to agricultural TFP improvements?
The paper first examines the way in which agriculture grows. The basic ingredients that make up
for faster agricultural TFP growth are known, and include agricultural R&D, extension, rural
infrastructure, and human capital such as education, and health. However, it was pointed out that
while we know the variables affecting agricultural TFP, the profession is much less sure about
the magnitudes of the elasticities of TFP with respect to the various factors above, the as well as
the ways in which these elasticities are affected by other conditions. Both structural parameters
ii The Role ofAgriculture in Economic Development and Poverty Reduction:

such as distributional variables, as well as institutional factors, such as the degree of market
imperfections may impinge on these elasticities. While considerable partial knowledge on these
issues exists at the micro level for several developing countries, their contribution to agricultural
TFP growth or to the magnitude of the elasticities of TFP growth with respect to other variables
is much less well analyzed. This set of issues then is an area of considerable lack of knowledge,
where more research would have a high pay-off.
The review revealed that there is by no means an unequivocal theoretical argument for
agricultural development as an engine of growth. If any, recent theoretical contributions tend to
rely on different structural properties between agriculture and non-agriculture concerning
external effects, notably learning by doing, to argue against agricultural productivity
developments as engines of growth in an open economy. It was noted in the review, however,
that such treatments rely on asymmetric assumptions about the economic characteristics of
agriculture and non-agriculture. Hence they cannot be utilized as guides to policy, unless one has
more information at the sectoral level about the nature of external effects and learning by doing
properties of the various sectors.
An important contribution of recent theoretical research has been to point out that the way in
which domestic incomes, are affected by the improved agricultural productivity, or any other
stimulus by another leading sector, and the consequent consumption and saving-investment
patterns are crucial for determining the pattern of growth. If improved agricultural productivity
leads to higher incomes of the poor who spend it on domestic non-tradables, then agricultural
growth will induce non-agricultural rural growth and via employment multipliers to decreases in
poverty. If, on the other hand, the fruits of agricultural development or other leading sector
economic development lead to increases in the incomes of the rich, then the important
determinant for growth will be where the additional savings are spent. If they are spent for
domestic labor intensive investments, then there may still be growth, and the poor may well
benefit from employment creation. If, however, they are spent on imported luxuries, or invested
abroad, there will be little growth stimulus. Thus the pattern of the distribution of increased
incomes from the initial stimulus is the most important determinant of subsequent growth.
Agricultural development can contribute to both growth and poverty reduction if the fruits of the
initial productivity stimulus is concentrated on those who respend it domestically (through
consumption or investment) on labor intensive products with low import dependence.
There may be both direct and indirect contributions to poverty reduction from agricultural
development, depending on the structure of the incomes of the poor. It was noted that while
recent empirical literature has highlighted the important role of agriculture for poverty reduction
in labor abundant agrarian countries like India and China, the relationship is not universal. In fact
a major contribution of the recent empirical growth literature has been to point out that the
elasticity of poverty with respect to agricultural productivity improvements depends on initial
distributional variables. This is consistent with the literature on the role of agriculture for growth,
which highlights precisely the distributional dependence of the contribution of agricultural TFP
growth to overall growth. However, it seems that the elasticity of poverty reduction should not
depend only on distributional variables. A variety of institutional and other features, which may
be of importance, cannot be captured precisely in cross-country growth analysis, and need to be
studied through specific country case studies.
The review revealed that there seem to be a set of conditions that make agricultural development
both growth enhancing as well as poverty reducing. These conditions involve the following:
 An Empiricaland Conceptual Framework iii

* Agriculture must account for a large share of aggregate employment.

* Initial distribution of land must be equitable and property rights must be well specified.
* The technological improvements must not be risk increasing, nor should they require
substantive private capital to be implemented.
* The marginal expenditure shares (both from consumption as well as from investment) of
the direct beneficiaries of agricultural growth for labor intensive local nontradables must
be large.
* There must be an excess supply of underutilized local labor resources.
• There must be complementary improvements in the provision of human capital assets at
the local level (education and health), as well as improvements in marketing
infrastructure (e.g. roads).
* There must an income and price elastic source of demand for the increased product of
agricultural, wither domestically (in the case of food crops), or internationally (in the case
of exports).
What if the above conditions that make agricultural productivity growth have a beneficial impact
on overall growth and poverty reduction do not hold? The literature is moot on this point. The
following points were, nevertheless, raised in this context. First, if agriculture does not account
for a large share of aggregate employment, then it cannot easily be a leading sector for growth or
poverty reduction. Investments in agricultural productivity growth must then be judged by
comparison with investments in productivity growth of other sectors. Perhaps the best policies
under such circumstances maybe investments in human capital, so as to allow agricultural
producers to adapt more efficiently to technological changes. Second, if it is difficult to change
the initial conditions in the short run, then, while working toward changing them in the medium
to long run, one may want to avoid the types of productivity growth that may favor adversely
factors that are unequally distributed. For instance, if the land distribution is skewed, then public
research in agricultural development may be better targeted towards labor intensive techniques
rather than land augmenting ones. Third, if there is no large local excess supply of labor, that
could be mobilized when the demand for rural non-tradables increases, then perhaps it is prudent
to concentrate on improving the marketing infrastructure, and hence lower the marketing cost
between rural and urban areas. This would prevent the prices of local non-tradables from
increasing too much from any agricultural stimulus, by essentially making such products more
tradable. The above points are not easily generalized, and the presence of only a subset of the
above conditions implies that agricultural development policy in such circumstances must be
judged on an individual country basis.
The paper highlighted some important emerging issues concerning agricultural development.
Globalization can create opportunities, for instance through improved agricultural terms of trade,
or improved market access for a country's agricultural exports. However, it may also pose
additional risks, such as increased domestic price instability. The point made is that globalization
must be accompanied by domestic market enhancing policies in order to produce positive overall
results from agricultural growth and for poverty reduction.
Biotechnology was also seen to pose risks and opportunities. Participation by developing
countries in the still to be shaped international rules of the game concerning technology transfer
iv The Role ofAgriculture in Economic Development andPoverty Reduction:

and IPRs seems the best avenue for developing countries in order not to be left out of these
developments.
The increasing role of the private sector in both developing countries was seen to pose new
challenges for the public sector in the context of agricultural development. Joint private-public
financing and responsibility for infrastructure, and research may open opportunities that were not
previously available. The role of the public sector in the provision of some type of price and/or
income insurance seems to be particularly important in this context.
The elaboration of a two-sector theoretical model of agricultural growth, based on external
effects of public infrastructure investments, and the ensuing empirical simulations, suggested
some tentative conclusions concerning the role of agriculture in growth enhancement and
poverty reduction. First it was shown that maximum growth involves the provision of adequate
public infrastructure, financed by higher levels of taxation. However, it was shown that there is a
negative relationship between agricultural and non-agricultural taxation for optimum growth. It
was seen that higher elasticities of agricultural TFP with respect to public expenditures imply
faster real income increases for the poor, while lower overall rates of growth. This trade-off is
rather surprising and needs further study and elaboration. It was also seen that the optimum share
of public investment expenditures devoted to agricultural TFP improvement depends of the
degree of trade dependence of the economy and the non-agricultural sector, increasing with
higher degrees of trade dependence. This is contrary to earlier theoretical results. Finally it was
shown that the labor intensities of the agriculture, as well as the non-agricultural sectors, and the
aggregate saving rates are important in determining the maximum possible growth under a
strategy of agricultural development.
The implications of the above review and analysis for the rural development strategy of the
World Bank, are the following. First, it should be clear that the relative importance of the
agricultural sector as a leading sector for growth and poverty alleviation depends on country
specific geographical and economic structure variables. There can be no single agricultural
growth and poverty reduction strategy to fit all developing countries. It rather seems that the
emphasis on agriculture as a leading sector should depend on a set of criteria of the type
mentioned above. Country specific strategy and policy work could then concentrate on the
presence of the conditions that may be conducive to making agriculture a growth and poverty
alleviation pole. Some of these conditions were elaborated above, but it seems that there needs to
be more organized analysis of these criteria and conditions, combined with specific indicators, to
provide specific empirical and country relevant guidance.
Second, there is no doubt, given the considerable externalities involved in agricultural TFP
growth, that external assistance should concentrate on creating an environment that facilitates the
productivity of any attendant domestic investments. For instance, if the government of a
developing country is keen on rural infrastructure, perhaps, external resources could support the
development of the other necessary ingredients for growth and poverty reduction, such as rural
education and health, with emphasis, of course, on sustainability of investments. This can be
justified under the notion that there are complementarities between various types of agriculture
related public interventions. It must be noted in this context that there seems to be considerable
lack of knowledge concerning the dependence of elasticities of both growth, as well as poverty
reduction to different types of agricultural development spending, on institutional, as well as
other structural factors. This seems an area where more country specific, as well as cross-country
research is needed.
 An Empirical and ConceptualFramework v

Another point has to do with the type of agricultural research that is supported. In many
developing country agrarian settings, the reservation unskilled wage for the economy is close to
the average product of labor in agriculture. This for instance seems to be the case in many land
abundant but labor constrained economies in Africa. This implies that in such settings,
agricultural research and productivity enhancement, in order to be both growth enhancing, as
well as poverty reducing, must aim at increasing the average productivity of labor in agricultural
production, without making production more labor intensive. This is different from settings with
labor abundance, and land scarcity, where the reservation unskilled wage may be close to the
marginal product of labor in agriculture. In such settings, agricultural research should try to
increase the productivity of land, so as to increase the marginal product of agricultural labor.
A major strategic issue has to do with the efficacy of agricultural productivity growth as an
engine of growth, as a function of rural population density. In settings of low rural population
densities, the cost of agricultural productivity enhancement may be very large relative to the
benefit per affected household. In such settings the elasticity of TFP growth, as well as the
elasticity of poverty reduction with respect to various types of public expenditures may be low.
The empirical simulations of the model suggested that low values of the agriculture TFP
elasticities imply low shares of public capital devoted to agriculture TFP enhancement. This
underscores the fact that the factors that determine the size of the elasticities of agricultural TFP
with respect to public spending must be clearly understood, before recommendations about a
rural development strategy are made.
There are several areas where additional empirical cross-country work is needed by institutions
such as the World Bank. First, it is not clear how many contemporary developing countries fulfill
the conditions that were identified as necessary for a win-win (growth promoting and poverty
reducing) agricultural development strategy. While for some conditions the verification is
relatively easy, examples being the share of agriculture in total employment, the initial land
distribution, and the existence of under-utilized rural labor resources, for others the verification
is by no mean straightforward. This may well be a worthwhile empirical project for the World
Bank Rural Development and/or Research Departments.
Second, even if all conditions can be empirically checked, it will most likely be the case that not
all of them are satisfied in any one developing country. Does this mean that agricultural
development policies should take second priority? Or is it the case that there is a core subset of
the conditions that if satisfied can justify a vigorous agricultural development strategy? This
poses the problem of agricultural development policy under a second best world. For instance,
should development effort concentrate on setting the initial conditions for agricultural
development right as a prerequisite for public investments in agriculture, or should the two
proceed simultaneously? Unfortunately, very little is known about this.
Thirdly, are all conditions identified of equal importance? For instance is the provision of
adequate roads more important than the provision of human capital? Should rural education be
enhanced before investments in agricultural research and extension are made? The answers to
these and other related questions are not easy, and are not universal, as they are intimately tied
with the historical and institutional context of any given country, as well as on an appraisal of the
speeds with which any one policy can be implemented and have an impact. Little, however, is
known about the types of conditions that should be considered as part of the core needed for
agricultural development to be effective.
vi The Role ofAgriculture in Economic Development and Poverty Reduction:

Given this rather inconclusive state of affairs regarding the role of agriculture in growth and
poverty reduction, one way to proceed for the Rural Development Department of the World
Bank to augment its practical and relevant knowledge, would be to conduct a cross-country
comparative review of agricultural sector strategies and performances over the past twenty or
thirty years. Such reviews have been done at the initiative of the World Bank in the past, but with
very different focus, and have produced wide ranging policy relevant results. The new focus
should be the relationship between agricultural productivity growth and overall economic growth
and poverty reduction performance.
 1. Introduction
Two of the major themes in the development literature as well as thinking, that have received
additional emphasis in the 1990s, have been growth and poverty reduction. The new endogenous
growth theory has highlighted the importance of several factors conducive to faster economic
growth, such as human capital, infrastructure, sound monetary and fiscal policies, democracy and
political stability, trade openness, corruption, and others, while considerable effort has been
given to exploring relationships between growth and inequality as well as poverty. This
essentially macroeconomic approach to growth, has placed much less emphasis on sectoral
aspects of growth and poverty reduction. This, lack of sectoral emphasis, however, gives little
practical guidance to policy makers who have to make decisions about the allocation of public
resources, as well as sources of funds to finance public expenditures. Similarly the latest World
Bank Development Report for 2000/2001 titled "Attacking Poverty," that emphasizes three
themes, opportunity, empowerment, and security, is notable for the relatively limited discussion
of sectoral priorities in reducing poverty and enhancing growth.
It is well known that the majority of the world's poor live in rural areas. Of the about 1.2 billion
people in the world that are estimated to live on less than one dollar a day, about three quarters
work and live in rural areas, and depend to a large extent on agriculture. This would seem to be
good reason for support of rural poverty reduction strategies, and labor intensive agricultural
growth. Yet, since the mid-1980s, aid in support of agriculture has fallen sharply in both absolute
as well as relative terms, inducing slower growth in staple food yields and lower elasticity of
poverty to overall growth. The recent IFAD Rural Poverty Report 2001 (IFAD 2001) mentions
that real net aid disbursements to developing countries have fallen from 2.7 percent of their GDP
in 1992 (or 0.33 percent of OECD GDP) to 1.4 percent of their GDP in 1998 (or 0.24 percent of
OECD GDP). Over the same period, of this smaller aid disbursements, the proportion of
sectorally allocated aid going to agriculture, forestry and fisheries has declined from 20.2 percent
to 12.5 percent.
It is not clear why there has been such a decline in support for agriculture. Lipton (2000)
suggests that this could be justified under the following arguments:
* if public action were more cost-effective in reducing urban poverty;
* if the role of agriculture and the rural sector in supporting and advancing poor people in low
income countries has declined;
* if rural people gained more from urban poverty reduction than vice versa;
* if rural anti-poverty spending deterred successful urbanization;
* if rural anti-poverty spending induced less economic growth than urban poverty reduction; or
* if labor-intensive methods for small farmers and orientation of support for staple food
production has disadvantages in the context of more globalized markets.
Lipton suggests that none of these arguments holds true and some of his and other arguments
why this is so will be considered later.
Nevertheless, in the context of public resource allocation, the major questions that policy makers
may ask concerning support for agriculture are the following. Under what conditions does
additional or disproportional support for agriculture will be both overall growth enhancing as
well as poverty reducing? Is there a trade-off between faster growth and poverty reduction in the
2 The Role ofAgriculture in Economic Development and Poverty Reduction:

context of agricultural development? Given that overall support for agriculture as an efficient
growth enhancing and poverty reducing strategy can be justified, what are the most appropriate
ways to support agriculture so as to maximize the effectiveness of such support in generating
growth and reducing poverty? These questions will form the basis of the review and discussion
of this paper.
The discussion will start in chapter 2 with a brief review of the factors that contribute to
agricultural growth. Then the role of agriculture in enhancing overall growth is considered in
chapter 3. In chapter 4 the relationship between agricultural growth and poverty reduction is
reviewed. Chapter 5 reviews the channels through which agriculture contributes to poverty
reduction. Chapter 6 discusses some emerging issues in the context of agricultural development
Chapter 7 presents a simple two-sector theoretical model of agricultural growth and poverty, in
order to highlight some of the relationships discussed in the review. Chapter 8 summarizes the
conclusions. The effort throughout is be to identify the conditions that make agriculture a leading
sector for development and growth.
2. How Does Agriculture Grow?
Before we discuss the role of agriculture in poverty reduction it seems appropriate to review the
process through which agriculture grows. Concerning agricultural growth and its components,
early research (Binswanger et. al. 1987) showed that the major determinants of agricultural
supply are physical capital, infrastructure, human capital, research, extension, and rural
population density. Prices were found to be weak determinants of agricultural supply, consistent
with earlier studies of aggregate agricultural supply response to price (Bond, 1983). Similarly
Antle (1983) showed that the major determinants of total factor productivity (TFP) in agriculture
in cross-country regressions is education, research and infrastructure. More recent research
(Mundlak, Larson and Butzer, 1997, Mundlak, 1999) has confirned these results and has
specified that technological change in agriculture is incorporated into increased agricultural
production through the increases in physical capital stock. In cross country regressions that
incorporate both country specific and time effects, the result is that constant returns to scale
cannot be rejected, and that the shares of capital, land, labor and fertilizer are respectively 0.37,
0.47, 0.08 and 0.08. These are different when time effects are included, which is the way most
cross-country production functions have been estimated. In such regressions the elasticity of
capital is lower (around 0.34), that of land is practically nil, the elasticity of labor is 0.26, and
that of fertilizer is 0.43. In other recent analyses without time effects (Craig et. al (1997) the
production elasticity of land was found to be around 0.35, that of labor was 0.25, and that of
fertilizer 0.04. Capital elasticities were quite low in this study, but quite significant were the
contributions of infrastructure, human capital and research variables.
The various estimates are considerably hampered by the inaccuracy of aggregate data for inputs
such as labor and capital. Similarly the interpretation of the contribution of several variables such
as those of human capital (usually proxied by variables such as adult literacy, and life
expectancy) or infrastructure (proxied for instance by variables such as road density), is
problematic, as they may be providing indirect information on the role played by conventional
inputs, such as physical and human capital.
The changes in the total factor inputs appear to account for only about half of the total growth of
agricultural output (Mundlak, 1999). The rest is accounted for by the "residual", namely what is
normally termed total factor productivity (TFP), which is basically technical change. Mundlak
(1999) suggests that the empirical evidence points to the fact that the major way technology is
incorporated into agricultural production is through physical capital. The different rates of
growth of physical capital among sectors in turn can lead to differential sectoral growth rates
along standard Rybczynski theorem logic (for an analysis of such supply side factors see Martin
and Warr, 1993). Changes in technology, however, especially those involving new discoveries in
production techniques, come irregularly, and hence cannot be planned.
There are not many studies that explore the contribution of different factors to agricultural TFP
growth. A recent monograph by Evenson, Pray and Rosengrant (1999) has estimated the
contributions of various factors to India's TFP growth in agriculture. They find that public
research and extension are the two most important factors accounting for TFP growth, with
irrigation coming next. The internal rates of return to public agricultural research in particular are
4 The Role ofAgriculture in Economic Development and Poverty Reduction:

estimated to be higher than 50 percent, which is fairly impressive. Fan, Hazell, and Thorat
(1999) similarly show, using an econometric model estimate with Indian data, that public
expenditures for research and extension have had the largest impact on agricultural productivity
growth, with rural roads, education and irrigation following with a distance. Finally, Fan, Zhang,
and Zhang (2000) found that in China the largest contribution to agricultural productivity has
come from research and development public expenditures, followed by education, rural
telephones, rural roads, and electricity. It is interesting that irrigation investments in that setting
had the lowest impact on agricultural productivity.
It thus appears from these few recent exercises that publicly financed research and extension, and
rural infrastructure in the form of rural roads, electricity, irrigation, etc., are major contributors to
agricultural TFP growth, with investments in human capital also a significant factor. This is all in
line with the conclusions of the endogenous growth theory. All these papers, however, deal only
with agricultural TFP growth. Hence they do not answer the question of whether the same funds
if invested by the public in non-agriculture could have achieved larger TFP growth there. As
Evenson and Westphal (1995) point out there are significant differences between agriculture-
related research and industrial research, with the former much more circumstantially sensitive,
namely sensitive to local conditions. Thus, to make agricultural research have a high payoff, the
large fixed cost of establishing and running technological facilities must be geared to producing
results that can possibly be adopted by a large number of producers. This explains, for instance
why returns to agricultural R&D have been so high in densely populated agrarian countries such
as those in Asia, while they are lower in sparsely populated agrarian economies, such as those of
Africa. Evenson and Westphal (1995) in their survey of many returns to agricultural R&D
studies find that in Africa of 10 reviewed studies 4 (40 percent) reported rates of return higher
than 50 percent, while among 77 reviewed studies in Asia, the number was 48 (63%).
Nevertheless, if returns to agricultural research are as high as they appear to be, the question
arises as to why they do not attract further funds devoted to such research. Perhaps the reason
may have to do with constraints on public investment budgets, or the long term nature of such
investments.
The most surprising result of recent research in total factor productivity in agriculture and
manufacturing, is that across a variety of studies it appears that the rate of growth of total factor
productivity (TFP) in agriculture has been greater than the rate of growth of TFP in industry
(Bernard and Jones, 1996, OECD, 1995, Martin and Mitra, 1999). Martin and Mitra, in
particular, in the most complete study to date, found that the average annual growth rate of TFP
in manufacturing in developing countries varied between 0.62 and 0.92 percent over the period
1967 to 1992 depending on the methodology of estimation used, while in developed countries
the range was between 1.91 and 3.29. On the other hand in agriculture the average rate of growth
of TFP in developing countries ranged between 1.76 and 2.62 percent, while for developed
countries the range was between 3.35 and 3.46 percent. For the low-income developing
countries, the average rate of TFP growth in agriculture was between 1.44 to 1.99, while in
manufacturing it was between 0.22 to 0.93 percent. Clearly the rate of growth of TFP in
agriculture seems to be higher than that of manufacturing.
The study, furthermore, found that there seems to be convergence of the growth rates of TFP in
agriculture between all countries both developed and developing ones. The same was found also
for the growth rates of TFP in manufacturing. The authors interpret their results as suggesting
that they weaken the case for policies that discriminate against agriculture in favor of the
 An Empiricaland ConceptualFramework 5

supposedly more dynamic manufacturing sector. The results suggest that the high rates of TFP
growth in agriculture reflect effective systems of developing and disseminating internationally
innovations in agriculture, and this seems to be related to the establishment in the early 1960s of
a large-scale system for international agricultural research. Thus a hypothesis is that the
"globalization" of agricultural research, has contributed to faster TFP growth in agriculture,
compared to that of manufacturing, for which a large portion of applied research is privately
funded and appropriated.
While these results are very interesting, it is not clear whether they are due to disproportionally
high public investments in agriculture, or other policies discriminating against other sectors. If,
for instance, the contribution to infrastructure or education to TFP growth is similar across
sectors, it would be no surprise if higher TFP growth in one sector is due to higher shares of
public expenditures on these factors devoted to this sector. In fact Byerlee (1996) exhibits data
that show that developing countries have invested proportionally more in agricultural research
recently than developed countries, and this would be consistent with the above results.
Also it may be the case that technological improvements in agriculture are reflected more in
increased quantities of capital in the developed countries, compared to the developing countries,
hence masking the impact of technical changes on TFP. Hence the assessment that larger TFP
growth in agriculture implies that agriculture should not be discriminated against is incomplete
without the analysis of the sectoral utilization of public growth enhancing expenditures, as well
as analysis of the elasticity of capital inputs with respect to technological improvements.
Concerning factor supply, the data in Mundlak (1999) suggest the following:
* The share of agriculture in total investment, and in total capital is smaller than its shares in
output and the labor force. This is an indication of lower capital-labor ratios in agriculture
compared to non-agriculture.
* Average labor productivity growth in agriculture has exceeded that of non-agriculture.
* The share of agriculture in total investment has been declining since 1970.
* The share of manufacturing in total investment has also been declining. This indicates that
other sectors, probably services have attracted increasing shares of investment.
* The capital-output ratio in agriculture seems to have increased over the last thirty years
(indicating capital deepening).
* In most countries the capital-labor ratio has grown over time in the economy as a whole as
well as in agriculture.
Mundlak interprets this evidence as suggesting that demand is the dominant determinant of
agricultural growth. Sluggish agricultural demand growth, due to low-income elasticity of
demand for agricultural products, implies low growth in agricultural output and investment.
Similarly the increasing capital labor ratio in agriculture reflects the out-migration of farm labor,
as well as a shift to more capital intensive techniques.
An aspect that has not been appreciated in the above literature is the contribution of the
institutional environment to agricultural growth. Of course, the reason for such neglect is that for
most of the countries studied with data before 1985-90, there had not been any institutional
change in the structure of agricultural production, to justify any attribution of growth to such
6 The Role ofAgriculture in Economic Development andPoverty Reduction:

factors. However, the institutional changes in China (Lin, 1992) as well as in Eastern and Central
Europe (Sarris, et. al., 1999), have alerted researchers to the likelihood that institutional changes
may be instrumental in accelerating agricultural growth. Apart from major institutional changes
such as those that have occurred in China and Eastern Europe, one can think of other institutional
reasons that may contribute to fast productivity growth. These include the development of
agricultural extension, and the improvement in the functioning of input and output markets.
Other factors that have not been considered in these analyses are structural ones. For instance
does the structure of land ownership matter in the efficiency with which research and extension,
or other public investment policies like infrastructure or education enhance agricultural growth?
Does the structure of production (in terms for instance of the division between crops and
livestock, or between food and non-food crops) matter?
The above studies do not consider the contribution of the policy environment for agricultural
TFP growth. Early research did tend to show that policies affected the pace of agricultural
growth (Lele, 1989), and the review of agricultural price policies in 18 developing countries by
Schiff and Valdes (1991) tended to support the view that anti-agriculture price policies are
associated with slower agricultural growth. However, it was not clear from these studies whether
it was the result of the decline in overall resources to agriculture that slowed down agricultural
growth (and this is consistent with the sources of agricultural TFP growth literature), or the
decline in the elasticity of TFP growth to specific inputs resulting from bad policies.
The above studies suggest that while the standard inputs (capital and labor) enhance agricultural
growth, it is public expenditures for agricultural research and extension, rural infrastructure, and
rural education that are important for agricultural TFP growth. They do not, however, make the
case for disproportional public expenditures on such items relative to other sectors, as a growth
enhancing strategy, albeit the exceptionally high returns to publicly funded agricultural research
seem to suggest that considerable public investment should be devoted there. The studies also do
not consider how institutional and structural factors affect the effectiveness of these types of
policies. In other words, while by now we know the factors that affect agricultural growth and
TFP growth, and in some cases we even know the elasticities of TFP with respect to these
factors, we do not know how the elasticities of agricultural TFP growth with respect to the
various variables identified above are affected by structural and institutional features of an
economy. While country specific effects in cross-country regressions have taken account of
country heterogeneity, and isolated the net contributions of the indicated variables to growth,
their inclusion has not answered the more interesting question about what influences the
elasticities of TFP with respect to the standard variables. This is a ripe topic for research.
Another issue that has arisen in recent research is the lag between technological innovations in
agriculture and overall TFP improvements. Murgai (1999) has shown that during the period of
India's Green revolution TFP growth was surprisingly low, but increased in later years after
adoption was basically complete. She attributed this pattern to three major factors. First, the
technical innovations appear to have increased the elasticity of output response to modern inputs,
and hence a large part of the increased inputs should be attributed to the Green revolution, but is
not captured in TFP figures. Second, learning by doing effects were slow in the short run, as the
major impetus of the technological developments in the short run was fast capital accumulation
to facilitate adoption. Third, it seems that there were differences in development of
accompanying infrastructure like irrigation, and these have accounted for the differences in the
rates of productivity increases.
3. Agricultural Development and Overall Growth
What is the role of agriculture in economic development? Can agriculture be a leading sector to
induce faster growth, and under what conditions? These questions are very important for
development strategy, and the choices of policy makers. What do the theoretical and empirical
literature have to say on these issues?
On the relationship between agricultural and overall growth, Stem (1994) has presented a
summary of the empirical evidence concerming correlations between agricultural and non-
agricultural or overall growth. The historical pattern supports the view that in the course of
development the share of agriculture in both output, as well as labor falls. This is the outcome of
an initial disparity between labor productivities between agriculture and the non-agricultural "the
modem" sectors, that leads resources, especially labor to move out of agriculture.
Simultaneously the capital intensity in both sectors rises.
The empirical evidence across countries points out to close positive correlations between
agricultural and non-agricultural growth rates for the period before 1980, and little or no
correlation between the same growth rates after 1980. Stern hypothesizes that after 1980 there
were considerable exogenous shocks for many countries that may have slowed down the growth
of their non-agricultural sectors, and weakened the correlation between sectoral growth rates.
The associations highlighted by Stem suggest some complementarity between agricultural and
non-agricultural growth, and this can be supported by simple theoretical models based on
demand. For instance, rising income in a closed economy would lead to rising food consumption
at a positive but slower rate than that of non-agriculture, because of the fact that the income
elasticity of demand for food is smaller than one. This thinking would then suggest a positive
association between agricultural and non-agricultural growth rates, but with the latter larger than
the former. Of course, in open economies production and consumption can differ, and it is not
clear whether such conclusions and associations can be justified; The association between
agricultural and non-agricultural growth does not, of course, say anything about any causal
relationship between the two, and similarly does not say anything about a strategy for agriculture
in the course of development.
Early development writers such as Rosenstein-Rodan (1943), Lewis (1954), Hirschman (1958),
Jorgenson (1961), Fei and Ranis (1961) regarded agriculture only as a reservoir and source of
abundant labor and transferable product and financial surplus. The role of agriculture was seen as
ancillary to the main strategy of growth, which was accelerating industrialization. Hirschman
(1958) in particular was negative on agriculture as a source of growth on the basis of its weak
forward and backward linkages needed for development. By contrast Kuznets (1968) pointed out
that in a successful development strategy, technological progress must support both
industrialization and agricultural productivity. The basis of this view is the observation that the
stylized shift of employment away from agriculture and toward industry is the consequences of
technological changes in both agriculture and industry. The revolution in agricultural
productivity, according to Kuznets, is an indispensable base of modem economic growth. A
similar view was expounded by Kalecki (1960, 1971), who based his position on the idea that
balanced growth in both wage goods and capital goods forms the basis of sustainable long run
growth. Since agriculture is the main sector producing food, the key wage good in a developing
8 The Role ofAgriculture in Economic Development and Poverty Reduction:

economy, agricultural development is essential for a successful industrialization strategy for

developing countries.
Development thinking and practice in the 1960s and 1970s tended to neglect agriculture as a
leading sector, with its emphasis on import substitution industrialization and export promotion.
This thinking was aided by the literature concerning the terms of trade of agriculture. Concerning
the terms of trade for agriculture, which can summarize the "price bias" and the rate of taxation
on agriculture, and which can be considered as a policy instrument, there is extensive literature
on its pattern in the course of development, starting with the Soviet industrialization debate of
the 1920s (Preobrazhensky, 1965). Under the assumption that agricultural production would not
suffer, given the inelasticity of aggregate agricultural supply with respect to price, and the further
assumption that industrial investment gives higher rates of return, the idea was to "force"
savings, food, and labor out of agriculture through explicit and implicit taxation in order to
finance industrial growth.
Such thinking provided the intellectual basis for policies that were applied in many countries in
sub-Saharan Africa and other regions in the 1960s and 1970s, that taxed explicitly and implicitly
agriculture. The results of such policies were disastrous for growth, leading to the adoption of
structural adjustment programs that aimed at reversing such policies (for a summary of the issues
and debate concerning agricultural taxation see Sarris, 1994). Sah and Stiglitz (1984, 1987) have
provided a framework for thinking about the agricultural terms of trade in its relation to the
overall "investible surplus" of the economy, defined as the difference between the total
production and consumption of the non-agricultural product.
The logic of their argument is fairly straightforward. Assuming that agricultural or industrial
laborers do not produce much saving for investment, and that foreign savings are constrained, the
major sources of domestic savings are private profits from non-agricultural production and
public tax revenues from exports or imports. Since the major part of non-agricultural production
cost is labor, and since wages in a developing country seem to respond to the cost of food, non-
agricultural profits can be raised by keeping the price of food and hence wages down. Also if
most of exports are agricultural, while imports are non-agricultural, the government can increase
public revenue by taxing exports and/or imports. Both of these policies imply a reduction in the
terms of trade for agriculture.
In their later work Sah and Stiglitz (1987) widened the framework and clarified further the logic
of the above argument. They showed that suppressing the terms of trade of agriculture below the
levels dictated by international prices, increases the domestic investible surplus of the economy.
In other words the agricultural sector must be taxed, or equivalently the non-agricultural sector
must be subsidized vis-a-vis world prices, in order to raise the level of aggregate domestic
investment. There is a critical level of the domestic agricultural terms of trade, which is below
the international terms of trade, and which maximizes the total level of domestic surplus.
Furthermore, the suppression of the internal terms of trade does not have to impoverish urban
workers. Similar results about the agricultural terms of trade were obtained in earlier research
dealing with growth in dual economies such as the works by Hornby (1968) and Bardhan (1970,
chapter 9), under different assumptions.
In all of the above literature the basic assumption is that the major source of domestic savings is
non-agricultural profits. This is basically a functional view of savings and income distribution.
Translated to personal income distribution this view assumes that the recipients of agricultural
 An Empiricaland ConceptualFramework 9

incomes are subsistence farmers with little savings for investment, and that recipients of non-
agricultural profits are different from the recipients of agricultural and wage incomes. This,
however, neglects the possibility that the bulk of income recipients in developing countries have
joint income from agriculture and non-agriculture, and that rural agricultural producers may
generate considerable investible savings. In any case, the source of savings and investments is an
issue that must be dealt with in the design of a development strategy.
None of the above works deal with the issue of the allocation of investment among sectors.
Furthermore, they all neglect the issue of technical change, and its relation to investments and
growth. They also all seem to neglect the welfare of agricultural producers. Finally, they all deal
with growth towards some steady state growth rate that, like the standard neoclassical Solow
model (Solow, 1956), is independent of endogenous growth generating factors.
It was only in the late 1970s and early 1980s that the role of agriculture as a leading sector was
re-emphasized in the development literature by authors such as Mellor (1976) and Adelman
(1984). These authors emphasized the importance of agricultural growth in generating demand
for locally produced non-tradable products, and thereby stimulating overall production and
growth. Such a strategy was termed Agriculture Demand Led Industrialization (ADLI) by
Adelman (1984).
Timmer (1988) has observed that research to date relating to the different views about agriculture
in the course of development, suggest three sharply different paths for appropriate policies
toward agriculture if the goal is to speed up overall growth.
The first path, grows out of a view that markets, if left alone, will function properly, and that
economic decision makers are rational and respond efficiently to economic signals. As the long
run tendency is for a decline in the proportions of agricultural output and labor, the best way to
speed up growth is to accelerate this natural tendency. Rapid technical change and declining
relative prices for agricultural products (arising out of a faster growth of supply compared to
demand) in a world of little government interference will accomplish this.
The second path associated with Mellor and Johnston (1984), is the "interrelated rural
development strategy". This strategy advocates a unimodal, namely broad based, pattern of
economic development that improves incomes, nutrition, and income distribution, while
promoting overall growth. Agricultural growth not only satisfies the criterion of growing food
for the poor smallholders to meet nutritional requirements, but also promotes a favorable
employment-oriented demand structure. Mellor and Johnston advocate considerable government
intervention to promote extension and research aimed primarily at rural smallholders.
This strategy is heavily influenced by closed economy considerations. Three key elements are
suggested as essential to meeting all objectives of agricultural development, namely massive
investment in human capital through nutrition, health and family planning services in the
countryside, creation of a complex rural organizational structure (like the ones observed in
Taiwan and Japan) for providing services to small farmers, and investments in rapid technical
change appropriate to small farmers in order to raise agricultural output and incomes
simultaneously.
The third approach to agricultural development realizes the important links of agriculture and the
macro-economy, as well as the importance of market signals and incentives, elements that are
relatively underemphasized in the second strategy. It calls for government policy intervention
10 The Role ofAgriculture in Economic Development and Poverty Reduction:

into domestic markets, but uses markets and the private sector as vehicles for these interventions.
This approach can be termed "price and marketing policy" approach, and recognizes widespread
market failures in agriculture, as well as government failures in implementation of policies. The
dilemma is how to cope with segmented and/or poorly functioning or absent rural labor, land,
and credit markets, the pervasive lack or imperfect nature of information, and the absence of
many important markets notably those for risk.
All three approaches recognize the importance of government investments in infrastructure and
agricultural research. However, the approaches differ in their emphasis among these government
interventions. The free-market approach would put greater emphasis as well as budget share on
research, the rural development strategy on human capital investments, while the price and
marketing approach on rural infrastructure to lower marketing costs. As Timmer properly
concludes, the issue is not one versus the other, as all three elements should be part of a
successful agricultural growth strategy. The real issue is one of where should scarce resources be
invested, and with what priority at different stages of development.
The above strategic approaches, however, all avoid the major issue of allocation of resources
between agriculture and non-agriculture. They deal with the issue of how a given amount or
share of investment resources should be allocated among different types of investments within
agriculture to achieve a given growth rate. They do not deal with the issue of whether agriculture
should receive a larger or smaller share of overall investment resources relative to non-
agriculture, and whether this would speed up or slow down overall growth.
The real issue from a growth perspective, however, is how to accelerate growth. The role of
agriculture must be examined in such a context if some guidelines for strategy and policy are to
be derived. Unfortunately, however, there seems to be very little research focusing on such a
problem, and then only partially. In an early paper Krishna (1982) observed using data from the
period 1960-80 that non-agricultural growth not only was correlated with that of agricultural
growth, but, furthermore, that the growth rate of agriculture was usually lower than that of non-
agriculture. He also noted based on the research of Kuznets (1961) that the incremental capital-
output ratios for agriculture are higher than those of mining and manufacturing. Since such ratios
are often used to plan investments, the implication is that to achieve a similar growth rate for
agriculture and manufacturing, a larger share of investment should be devoted to agriculture,
relative to the share of agriculture in GDP.
Another paper that also examines the allocation of investment among agriculture and non-
agriculture is the one by Das (1982). He utilizes a simple two sector (agriculture and non-
agriculture) model, with externally fixed prices, and asks the question of the optimal saving rate
and optimal investment allocation among the two sectors, so as to maximize the per capita
consumption in the steady state. Under perfect labor mobility and no sectoral wage differentials
he finds that the optimal savings ratio (the golden rule savings ratio) is equal to the share of
capital in total income, and that the optimal share of investment in each sector is equal to the
proportion of total capital earnings generated in the sector. Hence, if agriculture generates a
small share of the economy's capital income, then this rule implies that the optimal share of
agriculture in total investments is correspondingly low.
If, on the other hand, the wage in the non-agricultural (urban) sector is fixed exogenously at a
higher level than that in agriculture, and if the labor market behaves in a Harris-Todaro (HT)
fashion, namely equating the marginal product of labor in agriculture with the expected marginal
 An Empiricaland ConceptualFramework 11

product of labor in non-agriculture, then Das (1982) finds that the optimal savings ratio is higher
than the full-employment golden rule ratio. Furthermore, the optimal share of investment for a
sector is higher (lower) than its contribution to capital income if and only if the proportion of the
total labor force in the sector is greater (or less) than its share of capital income. Given that in
developing countries the labor market is likely to behave in a HT fashion, and that normally the
proportion of labor in agriculture is higher than the proportion of capital income generated in
agriculture, the above result implies that the optimal investment share for agriculture is higher
than agriculture's share in capital income.
The above results are interesting, and give some kind of yardstick that can guide total
investments in agriculture and non-agriculture. However, the conclusion concerns all
investments and not just public ones. Furthermore, the complete exogeneity of prices makes
demand not to play any role.
On the relationship between agricultural and non-agricultural growth there is very little research
as already mentioned. A major exception is the paper by Adelman (1984) that advocated an
Agriculture-Demand-Led-Industrialization (ADLI) strategy for middle income developing
countries. This strategy that resembles in some ways the "interrelated rural development"
strategy of Mellor and Johnston (1984) basically consists of building a domestic mass-
consumption market by improving the productivity of agriculture and letting farmers share in the
fruits of improved productivity. The demand linkages generated by farmers, especially the small
low income ones, are stronger with domestic industries and other non-tradables, and domestic
low capital intensity non-agricultural sectors. The strategy advocates higher shares of investment
going to agriculture, in response to higher rates of return there. Thus, investment allocations are
made functions of the relative rates of return, and the ADLI strategy is based on the observation
that investment returns are higher in agriculture than in non-agriculture at some stages of
development. This is, of course, a key observation that has also been made by Lipton (1977,
chapter 8).
The importance of agriculture for growth in the context of the ADLI strategy has been recently
demonstrated by Vogel (1994), who computed forward and backward SAM multipliers for
agriculture and non-agriculture for a variety of developing and developed countries, and plotted
them as a functions of the GDP per capita. He showed that the backward multipliers of
agriculture are much larger than the forward multipliers at all development levels, and
firthermore, that they grow across different countries, from those with low GDP per capita, until
the GDP per capita reaches middle development levels (around 2500-3000 USD). The backward
multiplier works by examining the implication of an additional unit of income for agricultural
households. It implies increased expenditures of agricultural households on non-agricultural
products, and hence increased incomes to non-agricultural households, and through a secondary
effect, further expenditures on non-agricultural goods. The forward multiplier works thorough
the demand by agriculture of inputs from non-agriculture, and attendant income effects.
The finding that the agriculture backward multiplier is quite large, supports the view that
agricultural growth contributes considerably to overall economic growth through the demand
linkage effect. The author argues that the fact of strong backward and weak forward linkages
make agriculture a candidate to be the leading sector in an economic growth strategy. This point
of course has been made repeatedly by authors working in the "growth linkages tradition," such
as Haggblade and Hazell (1989), Haggblade, Hammer and Hazell (1991), Hazell and Roell
(1983), and more recently Delgado, Hopkins, and Kelly (1998). These authors have emphasized
12 The Role ofAgriculture in Economic Development and Poverty Reduction:

the importance of non-tradable non-agricultural products in the demand of rural low-income

residents that are recipients of agricultural income. However, the logic of this approach depends
considerably on the assumption of elastic labor supply by agricultural households, or rural
residents. While this assumption may be reasonable in some labor abundant developing
countries, it may not be so in others.
Recently, in relation to the revival of discussion about growth rates, in the context of the
"endogenous growth literature", there has been a small number of papers dealing with
agricultural growth, the terms of trade, and overall economic growth. An early paper by
Thirlwall (1986), provided a simple two sector framework, that related the internal terms of trade
for agriculture to the equilibrium growth rates of the two sectors (agriculture and industry),
which in the steady state are assumed equal, and hence equal to the overall growth rate of the
economy.
He showed that for a closed economy, the equilibrium growth rate will be larger, the higher is
the productivity of investment in industry and agriculture (namely the incremental output capital
ratio), the higher is the agriculture savings rate, and the lower are industrial wage costs per unit
of output. In such a steady state, and under the assumption of unitary income elasticities of
demand, the equilibrium terms of trade are constant, and, furthermore, the ratio of outputs in the
two sectors is constant and dependent only on the agricultural sector savings rate, and the wage
bill per unit of output in industry (namely the industrial labor share).
Thirlwall showed with his model that the basic constraint on growth is technical change in
agriculture, and that the steady state rate of agricultural growth is independent of the terms of
trade. Technical progress in agriculture (assuming no discovery of new land, which affects the
growth rate in the same way as technical progress) will relax the constraint on industrial growth,
and it is only this that will do so. In other words technical progress in industry only affects the
internal terms of trade, but not the long run equilibrium growth rate of the economy.
International trade in Thirlwall's model acts so as to provide, through exports, an additional
source of demand for industrial output, in addition to domestic demand that comes from
agricultural growth. Thus the rate of growth of demand for industrial output becomes a weighted
average of the rates of growth of domestic and export demand. As export demand comes to
dominate domestic demand for industrial output, the rate of growth of industry becomes
externally constrained at a rate that is independent of the rate of growth of demand coming from
the agricultural sector, and this is a turning point in the country's history. Thus the model that
Thirlwall analyzes points out that "...in the course of development we expect a healthy
agricultural sector to be the driving force behind industrial growth in the early stages, superseded
by export growth in the later stages. In this sense the model reinforces the belated recognition of
agriculture's importance in the early stages of development, and lends support to export led
growth theory in the later stages."
While Thirlwall's model emphasizes the role of capital productivity and the agricultural savings
rate in growth, it does not discuss how these can be raised, or affected by intersectoral labor
movements or investment allocations or technical change. In his model the gross productivity of
investment in the two sectors is a constant, and this seems unrealistic, as reallocations of labor in
the course of development significantly change the sectoral marginal products of capital. Also it
views investment in each sector as determined by the savings of each sector, a very restrictive
assumption that omits the possibility of inter-sectoral reallocation of savings. The domestic
An Empirical and ConceptualFramework 13

agricultural terms of trade in the model are endogenous and constant in the steady state. This is
the result of the assumption of unitary income elasticities for the products of the two sectors.
Finally, if one assumes neoclassical production functions for the products of the two sectors, and
if one assumes that in the steady state there is no intersectoral labor movement, then Thirlwall's
model implies in the steady state zero growth rate of per capita incomes in both sectors, just like
the neoclassical Solow model.
Another model by Canning (1988) is more in the spirit of the endogenous growth literature.
Canning considers a closed economy that produces three products, namely agricultural goods
destined for consumption, manufactured goods also destined for consumption, and manufactured
goods, destined for investment. He builds a general equilibrium model of the economy, in which
demand for agricultural goods is subject to Engel's law, namely an income elasticity of demand
smaller than one. Consumption in his model comes from labor and land ownership, and this
implies a farm structure, where farmers (who may be visualized as owning their land) do not
save. Savings are derived only from industrial profits, and are equal to the sum of investments in
all sectors. Non-agricultural production is subject to production with increasing retums, and it is
this assumption that drives the results of the model.
Under constant values for population and land, Canning shows that his model implies a constant
and stable steady state aggregate capital stock. However, the per capita capital stock is an
increasing non-linear function of the population, and this implies that under a growing
population, there is non-zero per capita capital stock growth, and hence positive per capita
income growth. Remarkably, this growth does not require technological change, but rather the
continuous expansion of manufacturing. The latter, because of increasing returns technology, can
produce ever-increasing amounts of output at diminishing cost.
Canning's model is compatible with a diminishing share of labor force in agriculture, as well as
with declining internal agricultural terms of trade. The reason that such developments do not
come into conflict with an assumption of no technological improvement in agriculture, is that the
lower prices of manufactured goods allow agriculture to adopt techniques that are increasingly
capital intensive, thus alleviating the constancy of land.
Canning's model does not consider foreign trade, and no policy variables. Furthermore, his
"short run" is characterized by complete depreciation of all capital, something that would be
more appropriate for overlapping generations models, and also by full adaptation of industrial
structure (namely number of firrns) to eliminate profits due to scale economies. However, both of
these assumptions are more appropriate for the "long run." In this sense his model, albeit
offering interesting insights is not appropriate as a guide for thinking about policy towards
agriculture and growth.
A more recent model fully in the endogenous growth model tradition is that of Matsuyama
(1992). Matsuyama considers a two-sector model of agriculture and non-agriculture (industry).
Agriculture is characterized by traditional diminishing returns production function of labor, but
industry is characterized by the accumulation of knowledge as a byproduct of aggregate
production. These external economies that arise from learning by doing are external to the firms
and increase industrial productivity, while there is no such knowledge accumulation in
agriculture. Labor is assumed to move freely between the two sectors so as to equalize short run
marginal products. There is no capital in his model. He also assumes a demand system that
implies an income elasticity for demand of the agricultural products smaller than one.
14 The Role ofAgriculture in Economic Development and Poverty Reduction:

His model in the short run is a simplified Ricardo-Viner (two sector three factor) model, and is
used to show that, when the economy is closed, the share of employment in manufacturing (or
agriculture, as the model assumes full employment) is constant over time and positively related
to the level of exogenous agricultural productivity. The domestic terms of trade is a function of
labor allocation and is passive in the model, namely it does not determine the rate of growth.
Growth is determined by labor allocation. As the only sector that can grow is industry, because
of learning by doing, the more labor that is allocated to industry the larger is the rate of growth.
The share of labor that is employed in industry, in turn in the closed model is related positively
to the level of agricultural productivity. Thus in the closed economy, agricultural productivity
increases are crucial in inducing growth.
However, this result seems to depend crucially on the closed economy assumption. When the
world is composed of two similar economies of the above type, then Matsuyama shows that there
is a negative link between agricultural productivity and growth. The reason is that a more
productive agriculture that employs more people, slows down the rate of labor transfer to
industry, and since industry exhibits external learning effects, slows down growth.
The rather counterintuitive results of Matsuyama do not imply that an economy with more
productive agriculture is necessarily worse off than one with less productive agriculture. They
just imply that the rate of growth is smaller. The major contribution of Matsuyama's paper is to
demonstrate that the relationship between agricultural productivity and growth depends on the
openness of the economy. His model depends on the assumption of no international knowledge
spillovers. It is also rather special in the sense that it does not include a non-tradable sector, and
in that it assumes that both agriculture and manufacturing production depend only on labor and
not on capital. Finally it depends on the assumption that learning by doing exists only for the
non-agricultural sector. There is no a-priori reason why this should be so. It is more likely that
learning by doing exists in sectors using technologies with economies of scale, and applying
modern techniques. There are many traditional industrial sectors that are not subject to
economies of scale, just as there are agricultural production technologies that are subject to
modern techniques and economies of scale. A better distinction in that model would perhaps be
between modern and traditional sectors.
Another theoretical work that examines the role of agriculture in the course of development is
that of Taylor (1991). Taylor utilizes a structuralist model of an economy, namely one that
incorporates "stylized facts" about economic behavior, rather than functions that derive from
optimizing behavior of agents. His model includes two major sectors, one that is largely quantity
adjusting under non-full employment in the short run and another that is price adjusting (a fix
price, flex price specifications).
Taylor explores the implications of an "agriculture-first" strategy in the sense that agricultural
investment is increased autonomously. Taylor attributes considerable significance to the internal
terms of trade, and it is via the different changes in the internal terms of trade that the results
obtain. He shows that the way in which an agriculture first strategy affects long run growth
depends on whether the movement in the internal terms of trade make agriculture's income fall
or rise. In the latter case growth in enhanced, while in the former it is not. Thus, the value of the
various elasticities that determine the domestic terms of trade, as well as the various
distributional parameters are crucial in determining the contribution of agriculture to growth.
This is especially important as most of the studies of technical change in agriculture have found
that the major gainers from new agricultural technology are the consumers of agricultural
 An Empirical and ConceptualFramework 15

products. This tends to benefit urban poor consumers as well as small farmers and rural landless.
It is thus the domestic elasticities of demand for food that matter for the contribution of
agricultural technological change to poverty reduction, and it is clear that these are influenced by
the degree of openness of an economy (Evenson and Westphal, 1995).
Another recent model that analyzes a two sector (agriculture-industry) economy, is that of Skott
and Larudee (1998). Their model assumes increasing returns to industry and non-increasing
returns to agriculture. Under autarchy they show that successful industrialization requires
adequately high agricultural labour productivity growth, otherwise the economy may fall into a
de-industrialization trap. This is compatible to what Matsuyama (1992) showed for a closed
economy, and this is no surprise as both assume increasing returns to industry. However, Skott
and Larudee diverge from Matsuyama in showing that under free trade economic growth and
industrialization are possible if the initial conditions are such that the economy is not in a de-
industrialization trap. Such a trap, in turn is possible if the initial level of industrialization is not
too high, and the wage share in industry is high. These conditions may resemble the initial
conditions of some countries in Sub-Saharan Africa in the post-adjustment period and Eastern
Europe in the post-transition period, and hence such a model may be helpful in understanding the
stagnation in these regions during the last ten years.
The above models and theories point out that the degree of openness, especially in the presence
of economies of scale, is a key factor in understanding the role of agricultural productivity
growth in speeding up overall growth. They also point out that since that demand factors are
crucial in determining whether agricultural productivity growth is helpful for overall growth, the
distribution of income and gains from growth is a key factor in this issue. They finally point out
that the composition of demand among tradables and non-tradables seems to be an important
element of the agriculture-first theories. To-date, however, there has been no theory or
framework integrating all the above elements. The models also do not consider the issue of how
agricultural productivity growth is to be achieved and how it is to be financed.
Do policies matter in the pace of agricultural growth? The empirical work of Mundlak, Cavallo
and Domenech (1989) and Coeymans and Mundlak (1993) is the most serious attempt to date to
relate macroeconomic and other policies to the internal terms of trade and growth. The empirical
model that they employ is a non-full employment small economy model that assumes investment
functions related to sectoral profitabilities. The overall savings-investment balance in their
models comes from the external sector, which is assumed to be able to provide enough "savings"
to finance the investments that are desired domestically. In this sense their models are
"structuralist" according to Taylor's (1991) terminology, as the two most important
macroeconomic "closure rules", namely the one governing the labour market and the one that
concerns the savings-investment balance are clearly non-neoclassical. A major innovation of
their models is the endogeneity of the technological adaptation in agriculture. This is modeled by
varying parameters of a Cobb-Douglas production function, where the parameters are made
functions of various other exogenous and policy variables.
The authors find that agricultural taxation and macroeconomic policies influenced considerably
the pace of agricultural and overall economic growth in Argentina and Chile. They also show
that the degree to which macroeconomic policies affect agricultural and overall growth depends
on the extent of "tradability" of the sectors, namely the shares of products produced by various
sectors that are tradable. Taxation of agriculture, both direct and indirect has slowed down the
adoption of better technologies in production, and hence slowed down growth. They also find
16 The Role ofAgriculture in Economic Development and Poverty Reduction:

that government investment influences positively private investment, as one would expect, but
the financing of investment makes a difference, as deficit financing via domestic borrowing
crowds out private investments.
4. Agriculture, Growth, and Poverty Reduction
Is growth in agriculture conducive to reducing overall poverty? If yes, is it more likely to do so
than growth in non-agriculture? What are the conditions that make agricultural growth a major
contributor to poverty reduction? These are the major questions that policy makers may be
interested in when thinking about development strategies that are pro-poor. What does the
literature say on these issues, and are there clear answers?
The ways in which agriculture can affect the overall poverty level in a country can be direct and
indirect. The direct way implies that agricultural growth lowers directly the degree of poverty in
rural areas and the whole economy. The indirect way implies that the way agricultural growth
contributes to overall poverty reduction is through the contribution of agriculture to overall
growth, and through the latter's contribution to poverty reduction.
Consider the contribution of overall growth to poverty reduction. This issue can be investigated
by examining the relationship between growth and income distribution, or directly via the
relationship between growth and poverty reduction. On the former, there is considerable recent
literature that re-examines the Kuznets inverse U hypothesis, namely the proposition that starting
at a low level of income, the income distribution worsens at early stages of growth, before it
improves. This hypothesis received support in the 1970s (e.g. by Ahluwalia, 1976), but recent
research has weakened the earlier results and has shown that much of the earlier results were due
to omitted country-level effects (Anand and Kanbur, 1993, Bruno, Ravallion and Squire, 1996).
In fact the review of Bruno, Ravallion and Squire (1996) showed that the income distributions in
most countries tend to be stable over time, while they vary considerably across countries.
Similarly Ravallion and Chen (1997) find that aggregate income growth in 43 countries did not
associate itself with reduced inequality.
This suggests that the major influence on poverty reduction is broad-based overall growth. This
seems to be supported by a variety of other recent studies, such as the ones by Fields (1989),
Squire (1993), Lipton and Ravallion (1995), and Deininger and Squire (1996). The last one of
these studies, shows that in periods of aggregate growth, the income of the poor increased in 88
percent of the cases. Birdsall, and Londono (1997) similarly find that the elasticity of income
growth of the poor with respect to aggregate income growth is 1.3.
The fact that country specific effects seem to be important in explaining the differences among
income distributions in different countries has given rise to some work trying to explain these
cross country differences in more detail. A recent paper by Bourguignon and Morrisson (1998)
suggested that a major explanatory factor in cross-country differences in income distribution is
the degree of dualism in economies, measured by the relative labor productivity between
agriculture and non-agriculture. Other important explanatory factors in cross country differences
in income distributions were also agriculture related, namely the amount of land per capita and
the share of land cultivated by small and medium farmers. Bourguignon and Morrisson interpret
their findings as suggesting that "...in many countries increasing the level of productivity in
traditional agriculture may have become the most effective way of reducing inequality and
poverty."
The fact that recent research points out that general growth improves all incomes and hence
under a constant distribution reduces absolute poverty, does not mean that the pattern and
18 The Role ofAgriculture in Economic Development and Poverty Reduction:

structure of growth do not matter. A paper by Ravallion and Datt (1996), that used time series of
income distribution data from several Indian states, showed that while urban income growth
contributed to urban poverty reduction, it did not contribute to rural poverty reduction, or overall
national poverty reduction. On the other hand rural income growth contributed to both urban and
rural poverty reduction, as well as national poverty reduction in India. These results, that are
compatible with the Bourguignon Morrisson findings, provide the strongest evidence to date that
rural growth is more pro-poor than urban growth. The Ravallion Datt analysis showed that 85
percent of the large reduction in poverty in India during the period of analysis was due to
agricultural growth, and this is very strong evidence that agricultural growth is pro-poor. The
finding of Ravallion and Datt is contrary to the earlier finding of Quizon and Binswanger (1986,
1989), who showed, using a partial equilibrium multimarket model for India, that the agricultural
growth effects of the Green revolution did not benefit the rural poor. The analysis of Quizon and
Binswanger, however, only considered agricultural incomes and did not consider spillover
effects to non-agricultural incomes. The authors in fact showed that the main way to help the
poor, were rises in non-agricultural incomes (of both rural and urban residents). The issue then
is, whether initial rises in agricultural incomes help increase the non-agricultural incomes that
eventually help the poor. These authors did not go into this subject.
Another paper by Ravallion and Datt (1999) showed, using the same state level Indian data as in
Ravallion and Datt (1996), that the factors that were instrumental at reducing poverty were
higher average farm yields, higher state development spending, higher non-farm output (both
rural and urban), and lower inflation. The elasticities of poverty reduction with respect to the
various variables above were the same across states, except those with respect to non-farm
output. Thus, while differences in agricultural output growth had the same impact on poverty
irrespective of state, the degree of non-farm output growth had different impact on the poor
depending on the state. Yet another paper by Datt and Ravallion (1998) found that rural absolute
poverty is negatively related to the real rural wage and the average farm yield, and positively
related to the relative price of food. Thus lower food prices were in that case poverty reducing,
largely because a large number of the Indian rural poor are landless, and hence net food buyers.
They estimated both short run and long run elasticities of various poverty measures to average
agricultural yields, and found that while the short run total (direct and indirect) elasticities were
negative and between -0.18 and -0.41, the long run elasticities were substantially larger (from -
0.88 to -1.93), implying that increases in farm agricultural productivity have substantial effects in
the long run, namely after the rural agricultural and non-agricultural labor markets have had time
to adjust.
While the number of studies investigating the structure of growth and its impact on poverty
reduction is small and concentrated on India, they all seem to point out that agricultural growth is
poverty reducing, and more so than other types of growth. Is this a result that can be generalized
across countries? The evidence suggests that this is not so. For instance Timmer (1997) finds that
the impact of agricultural growth on poverty reduction depends on income distribution. Timmer
utilizes the concept of the "elasticity of connection" between growth in the overall economy and
growth in the per capita incomes of the poor, specified as either the bottom twenty or forty
percent of the income distribution. If growth is uniform for all incomes, then this elasticity
should be equal to one for all income groups. He finds by simple regression of the log of per
capita income of a given income quintile on the log of average income per capita (including
fixed effects), that these elasticities for the poorest quintiles are, smaller than one, while those for
An Empiricaland ConceptualFramework 19

the top quintiles are larger than one. He then tries to account for these differences by permitting
the elasticities to vary as functions of distributional variables.
He finds that the overall income inequality, measured by a variable intended to capture the
relative income gap, between rich and poor', does not affect much the estimated aggregate
elasticities of connection. He then goes on to investigate whether differences in sectoral labor
productivities matter. He finds that income inequality affects considerably the elasticity of
poverty reduction with respect to different types of sectoral growth. For instance in countries
where the relative income inequality is large, the elasticity of connection between agricultural
labor productivity, and the per capita income of the bottom quintile is very small and not
statistically different from zero. The same holds for the elasticity of connection with respect to
non-agricultural labor productivity. The elasticities for the top quintile, by contrast are larger
than one. For countries with small relative inequality, the elasticity of connection is close to one
for both types of income, and slightly higher for agriculture. This basically says that the
contribution of agricultural productivity to poverty reduction is a function of the inequality in the
country, with unequal countries having a low elasticity of connection.
Timmer uses his results to estimates the impact on the poor and non-poor from a uniform
aggregate economic growth of 5 percent in per capita terms. Assuming that the composition of
growth is such that agriculture grows by 3 percent, while non-agriculture grows by enough to
keep the overall 5 percent growth rate, he finds that over a 25 year period, in low inequality
countries the per capita incomes of those in the bottom quintile grow by 241 percent, while the
incomes of those in the top quintile grow by 211 percent, thus narrowing the income distribution.
By contrast with the same assumptions the per capita incomes of the poor in a high inequality
country would grow by only 73 percent, while those of the top quintile would increase by 273
percent, worsening substantially the income distribution.
DeJanvry and Sadoulet (2000) find results similar to those of Timmer (1997), as far as the
impact of distributional variables on the elasticity of poverty reduction to overall growth. Using
data on rural and urban poverty from Latin American countries, they find that the elasticity of
urban poverty to overall income growth is -0.95, and they also find that agriculture related
variables do not affect urban poverty. On the other hand they find that the degree of income
inequality does affects this elasticity, with low inequality increasing considerably the absolute
value of this negative elasticity, while high inequality almost erases the ability of income growth
to reduce urban inequality. Concerning sectoral aspects, they find that growth in the service
sector is important in reducing urban poverty, but growth in agriculture or manufacturing is not.
Concerning rural poverty, they find that the elasticity of rural poverty reduction to overall growth
is smaller at -0.75. They also find that high levels of initial rural poverty make the absolute value
of this negative elasticity very small, while low levels of initial rural poverty increase its absolute
value considerably. Hence the elasticity of poverty reduction with respect to overall growth
depends on distributional variables. In contrast to Ravallion and Datt's (1996) work, however,
they do not find significant impact on either rural or urban poverty reduction from agricultural
growth. They find instead that it is only service sector growth that reduces rural poverty. It thus
appears that the role of overall income growth in poverty reduction depends on the distribution

' The variable Timmer utilizes is a dummy, which is equal to one if the difference between the average per capita
income of the top quintile and the bottom quintile is larger than twice the average per capita income of the economy.
20 The Role ofAgriculture in Economic Development and Poverty Reduction:

of income in the economy, and the role of agriculture in reducing poverty depends on the country
context.
That asset inequality can affect the poverty impact of growth is relatively easy to visualize. If, for
instance agricultural land distribution is highly skewed and agricultural productivity growth
favors the products produced by the large landowners, then it is not hard to see that agricultural
growth would not be poverty reducing. Such a conclusion is supported by data analyzed by
Adams and He (1995), who find that increases in agricultural crop income tend to be inequality
increasing in Pakistan because of the skewed distribution of land, while increases in livestock
income are equalizing. It thus appears that unequivocal agricultural growth, may not always be
poverty reducing, and that initial asset distribution, in particular with respect to land, may matter,
coupled with the particular way agricultural growth is stimulated.
The relationship between initial income or asset distribution and subsequent growth has been the
subject of considerable recent research (for a thorough review see Aghion, Caroli and Garcia-
Penalosa, 1999). The main links that the literature has identified, through which inequality can
affect the rate of economic growth are credit market imperfections and political economy
considerations that may affect the voting behavior of the "median voter." Empirically it seems
that there is strong association between more equitable income or asset distribution and
subsequent growth (for recent empirical analyses see Deininger and Squire, 1998, and Ravallion,
1997). The aspect of this literature that is relevant for our discussion here is that initial
inequitable land distribution makes the relationship between agricultural growth and overall
growth much weaker. The other question that it raises is whether agricultural growth under
inequitable land distribution can worsen the income distribution, and hence make subsequent
growth slower. In such a case agricultural growth could slow down overall growth. There do not
appear, however, to be studies to deal adequately with these issues.
It thus appears that the evidence reviewed suggests that agricultural development is associated
with overall and rural poverty reduction, but the relationship maybe conditioned by initial asset
and income distribution, as well as country characteristics.
5. The Channels Via Which Agricultural
Development Reduces Poverty
There are basically three ways through which the poor (or anyone else for that matter) can
improve their real incomes. First is through increases in the productive assets they own. This can
be done either through their own investments, out of their own savings or borrowing, or through
increases in publicly provided but privately appropriated assets, such as health and education.
The second mechanism is by improved employment and returns to the assets the poor already
own. Such improved returns could obtain, for instance, through increased utilization of unused
land, profits from increases in prices for the products the poor produce and sell, or increases in
employment and wages. The final channel is through increased productivity of the assets the
poor own. This could involve, for instance, increased land or labor productivity, namely
increased output per unit of land or labor at unchanged prices. How does agricultural
development contribute to these three channels?
The answers to the above question depend on the structure of assets of the poor, on the structure
of their income sources, on the structure of various institutions that mediate between the poor
and the rest of the economy (such as markets, family networks, etc.), and on the dynamic
economic andi social processes that create and maintain poverty. In other words they depend on
the static and; dynamic profile of poverty. Concerning the sources of income of the different
classes of the poor in a country, it is useful to classify them as income from agriculture (normally
divided by income from crops and livestock, or as income from food and non-food, or income
from tradable and nontradable products depending on the data and context), income from farm
and non-farm labor employment, profit income from own enterprise activity, income from land
rentals, and income from various other sources such as transfers, remittances, dividends etc.
The profiles of the poor differ considerably in different countries and regions. For instance many
of the poor in South-East Asia are rural smallholders, with substantial portions of their income
coming from agriculture, but also many others are rural landless, relying primarily on farm and
non-farm labor income. In much of Sub-Saharan Africa, the poor are mainly rural with the bulk
of their incomes from agriculture. In Latin America a large part of the poor are urban based,
relying for income on informal enterprise activity and non-farm labor.
Another differentiating aspect across countries is the existence of different farming systems in
different agro-ecological zones and parts of the world. The recent FAO farming systems study,
done for the World Bank (FAO, 2000) exhibits the heterogeneity in farming systems across the
world, but also highlights the fact that even within the same agro-ecological zone there may be
several farning systems that coexist.
Along with the static description of poverty, of significant importance are the dynamic poverty
processes, namely institutional features that create and, more importantly, maintain poverty. An
early description of a variety of such mechanisms, as they apply to the rural sector, is given by
Jazairy, et. al (1992), based on the experiences of IFAD in dealing with rural poverty related
projects. They include dualism, population pressures, resource management and environmental
degradation in fragile settings, natural production cycles inducing production risk, social
22 The Role ofAgriculture in Economic Development andPoverty Reduction:

marginalization of women, cultural and ethnic factors, and exploitative intermediation

mechanisms. In that volume an attempt was made to indicate the importance of these various
mechanisms in different countries. Overall they managed to characterize the rural poor as falling
largely into the following functional classes.
* Smallholder farmers
* Landless rural residents
* Nomadic pastoralists
* Ethnic indigenous groups
* Artisanal fishermen
* Displaced or refugee populations
* Households headed by women
It should be clear from this characterization that agricultural growth has different poverty
reducing and growth implications under different settings, and for different poor groups.
Consider increases in private productive assets. One mechanism through which such assets can
be augmented, and especially benefiting the rural poor, is land distribution, land reform, or
general enhancement of property rights to land. There are considerable issues concerning land
relations in agricultural development, and they have taken new dimensions in the context of the
transition of many countries in Central and Eastern Europe (for useful surveys see Binswanger
et. al, 1995, Swinnen, 1997, de Janvry et. al., 2001). They will not concern us here, as they form
a major subject on their own. For this survey it will be assumed that the landed poor own or have
access to given amounts of land, through some form of tenure. The landless poor, of course, do
not have agricultural land.
However, a major issue, which is related to agricultural development and its role in reducing
poverty, is security of land ownership or tenure rights. These rights are far from secure in many
developing countries, and this may be a major impediment to land augmenting technical change,
that will enhance the value of land. This could be, for instance, a major problem in most African
countries, where the land tenure systems are such that they provide very weak private ownership
rights. Access to land has many advantages for poverty reduction, and often for achieving
efficiency gains. For instance access to farm land can give value to many factors that are
underutilized otherwise by poor (e.g. family labor), can lower the cost of using household factors
of production (e.g. family labor through the lower transactions and supervision costs), can
provide food security and insurance when food prices rise, etc. (for a complete discussion see
deJanvry et. al., 2001).
Thus the poverty and growth implications of any land augmenting technical change in
agriculture will depend considerably on the existing land tenure system, because it is the
appropriation of the benefits of technical change that is at issue. As Adams and He (1995)
showed, agricultural development concentrated on technological change in crop production
tended to worsen income distribution in rural Pakistan, as most of the poor were landless, and as
increased crop income tended to favor the owners of land. Hayami (2000) illustrates vividly the
different growth paths of agricultural development since the nineteenth century in the
Philippines, Indonesia and Thailand, and attributes the different agricultural growth trajectories
to the evolution of agrarian structures in these countries. In the Philippines, bimodal and dual
agrarian structures, while initially efficient due to the early substantial expansion of the land
frontier, eventually turned into a disadvantage because of the inefficiencies of large-scale
 An Empiricaland ConceptualFramework 23

agriculture with its monitoring needs for hired labor. On the contrary, in Thailand and Indonesia,
despite similar early vent-for-surplus agrarian development, the agrarian structure that was
maintained was largely unimodal and smallholder based, that facilitated later agricultural growth
and development. More research on this issue is needed.
Consider increases in private productive assets through investment. It is well known that most
poor people face credit constraints, hence most of their investments are made using own funds
out of personal savings. To have savings, of course, implies that households can meet their basic
food and other needs first, out of whatever income they have. The evidence from household
surveys suggests that the poor do have savings, often of the order of 20-30 percent of their gross
incomes. If there are variations in the incomes of the poor, and of a magnitude that can reduce
basic needs satisfaction below some minimum acceptable levels, then there is vulnerability. That
there is considerable vulnerability among the poor around the world is well documented in
WDR2000, in!chapter 8.
Under vulnerability the poor may devote a considerable portion of whatever savings they have
into liquid forms of non-productive assets, as insurance. Such assets can take the form of grain
stocks or animals in rural areas, gold and jewelry in non-farm households, etc. This has been
documented in several analyses of microeconomic behavior (see references in page 143 of the
WDR2000). The poor, in response to external risks, may devote a disproportional portion of their
savings to such unproductive self-insurance and hence reduce their investments in more
productive activities. Thus, the need for precautionary savings may reduce the growth
opportunities of the poor, and may create poverty traps. For instance, Rosenzweig and Wolpin
(1993) found ihat in rural semi-arid India, poor farmers are less likely to invest in irrigation
equipment than in bullocks, despite the fact that the return to the former is higher than the return
on the latter, because bullocks can be sold in times of need, while pumps cannot. Similarly
Fafchamps and Pender (1997) showed using similar panel data from ICRISAT that the
indivisibility of profitable investments, such as wells, coupled with the need to have cash on
hand for insur'ance purposes, made it very difficult for poor households to undertake such
investments. lir the same vein, in many parts of the world, the need to maintain some income
when adverse shocks occur, induces parents to pull children away from school (an acknowledged
profitable investment) and send them to work. This clearly prevents human capital accumulation
and leads to persistent poverty across generations. It is clear that under such conditions, what is
needed is some institutional mechanism to provide in a reliable and credible way cheaper
insurance to the poor, in order to let them utilize in a more productive way their own savings.
The second major way in which poor can expand their own assets is through acquisition of
human capital such as education and better health. The role of the government in provision of
such assets is crucial and has been reviewed extensively in WDR2000 (chapter 5). It will not be
discussed further here, as it does not pertain directly to action to improve agricultural growth.
However, it must be mentioned that human capital assets by households such as education can
make for more efficient use of other productive services. An example is the use of irrigation
equipment, which was found in Vietnam to be more efficient by better educated farmers (Van de
Walle, 2000). Thus it appears that there are complementarities between human capital variables
and the productivity of physical capital. This implies that agricultural productivity enhancing
measures, such 'as provision of infrastructure and new technologies, will produce higher returns
when implemenited by more educated producers, or when accompanied by action to strengthen
the education of those affected. Another aspect of public sector provision of human capital
24 The Role ofAgriculture in Economic Development and Poverty Reduction:

services is that it appears that the poor do better with some of all, rather than with a lot of one
type of service and little or none of the others (Lipton, 2000).
Agricultural development involves productivity increases, and this can occur through either new
techniques of production, or through productivity enhancing infrastructure and human capital
investments. These are the main mechanisms identified earlier in section 2 that create
agricultural growth, and it is these mechanisms that must be considered in their possibility to
alleviate rural and urban poverty. There are direct as well as indirect ways in which agricultural
development can contribute to poverty alleviation. The direct way involves direct improvement
in the incomes of the rural poor, through adoption of improved techniques, or increases in the
productivity of their agricultural assets such as land. Such increases in productivity can come
about through agriculture-related research, and extension, as well as agriculture related
infrastructure investments, such as irrigation, and rural electrification. The extent to which such
agricultural productivity improvements lead directly to income increases of the poor depends on
the extent to which the poor produce the products for which improved techniques become
available, as well as the degree of adoption of the new techniques by the rural poor.
Consider new techniques of agricultural production. These normally involve the possibility of
higher crop or animal yields. For crops this can involve improved yields for food or non-food
crops. While both can lead to improved incomes, increased yield of staple foods has the
advantage that a portion can be consumed directly by poor producing households since the
income elasticity of demand for staples is normally larger than zero. This implies that the
increase in marketed surplus out of increased production of foods by poor rural producers will be
smaller than the increase in production, and this avoids large price declines of staples when the
products are not perfectly traded, the markets are imperfect, or the price elasticity of demand in
the rest of the economy is small. That such imperfections are prevalent in the rural areas of
developing countries is by now well accepted, and a substantial part of development economics
research over the past twenty years has been devoted to examinations of the implications of such
imperfections (for useful surveys see Bardhan and Udry, 1999, Bell, 1989, Stiglitz, 1989, Besley,
1995).
Concerning adoption, it is not at all assured that the poor agricultural smallholders will adopt the
improved techniques so as to benefit directly. The major reasons involve uncertainty and risk
about the new technology (for a survey of agricultural technology adoption see Feder, Just and
Zilberman, 1985), plus issues involving the availability of the minimum initial capital that may
be needed to implement the new techniques. Under conditions where adoption is perceived as
risky, and in addition requires capital outlays, it is quite likely that the early adopters are the
better off farmers. This may create initially adverse consequences for the poorer farmers if the
increased production of the progressive farmers depresses domestic prices. This may either
marginalize the poorer farmers or may accelerate their tendency for adoption. In any case
historically the Green Revolution seems to have had negative initial effects on the smaller
farmers, but the later impacts were positive (Byerlee, 1996).
Binswanger and Von Braun (1993) have reviewed the issues of agricultural technological change
and poverty alleviation. They note that country experience suggests that agricultural yield
increasing technological change, coupled with better infrastructure to improve commercialization
of the increased production has been crucial to expanding agricultural growth, food supply and
employment, all of which are crucial to the poor.
An Empiricalahd ConceptualFramework 25

While there are cases that may be cited where technological change and commercialization has
been blamed 'for the decline of welfare of the poor, they show that in all such cases there are
other factors that have been responsible for these adverse effects. They note that uneven regional
agricultural development, while it may lead initially to adverse consequences for the residents of
the regions that have not adapted, due to the treadmill effect, will probably benefit the poor
through consumption improvements. They also showed that late adoption by the poor, something
that may be expected in view of the higher relative risks faced by the poor, does not necessarily
put them in a disadvantage. They also cite examples from Ethiopia and Sudan, where agricultural
technological change led to substantial increases in the returns to land, and led to eviction of
tenants, thus marginalizing the poor. However, the impoverishment in such cases was not due to
the development of agricultural technology, but rather to other accompanying anti-poor policies.
Binswanger and Von Braun also review the issue of targeting technological change to the poor,
and they find that the scope of targeting technological change to the poor is limited.
The major way, however, through which the poor may benefit from agricultural technological
change is indirect. Mellor (1999) makes the point that while recent empirical evidence, such as in
the papers reviewed above, "...make a powerful case that it is agricultural growth and essentially
only agricultural growth that brings about poverty decline in low income countries with a
substantial agricultural sector", the explanation for this relationship is older and associated with
the work of Johnston and Mellor (1961), Mellor (1961, 1976, 1995), Mellor and Johnston
(1984), and Mellor and Lele (1973). Mellor (1999) has reiterated this explanation, pointing out
that the main channels through which agricultural productivity increases impact on poverty
reduction are non-agricultural employment generation, increases in staple food output through
yield increases so as not to increase unduly domestic prices for the foods that are the major wage
goods, and shifts towards more high valued labor intensive agricultural commodities, that
stimulate demand for agricultural labor.
Concerning employment generation of agricultural productivity increases, Mellor (1999) makes
the point that algricultural employment is not likely to be very much stimulated by improvements
in land or labor saving technology of production of staple foods, because the elasticities are
rather low, normally much smaller than one. He suggests that a much more likely contributor to
agricultural employment generation is the stimulation of production of high value labor intensive
commodities such as fruits and vegetables. However, such a stimulus must come from increases
in demand for these products, that are in turn stimulated by higher incomes. Thus, one needs
higher incomes to generate such rural employment growth. He then goes on to suggest that the
major stimulus to rural employment is not from agriculture, but from rural based non-agricultural
activities. He suggests that employment elasticities from rural non-farm activities are close to
one.
The major way that has been identified by the literature, in which agricultural growth contributes
to overall growth and simultaneous poverty reduction is the stimulation of demand for non-
tradable labor intensive non-agricultural activities, through the demand linkage effect mentioned
earlier. Supply of such activities is normally assumed to be very elastic under the hypothesis of
underutilized labor resources in rural areas of developing countries. Hence, the increase in
demand is assumed to lead to an almost one-for one increase in supply, and this is what accounts
for the large multipliers. The estimated multipliers from increased agricultural output to overall
output are in the vicinity of 1.4-1.8 in most studies (e.g. Hazell and Roell, 1983, Haggblade,
Hazell, and Brown, 1989, Delgado, et. al., 1998, ), and can reach values as high as 3. However,
26 The Role ofAgriculture in Economic Development and Poverty Reduction:

in cases where the price elasticity of supply of labor is not infinite, then these multipliers are
smaller (Haggblade, Hammer, and Hazell, 1991).
Mellor (1999) also makes the point that development of urban-based formal sector
manufacturing in the absence of agricultural growth is not likely to reduce poverty. The reason is
that formal sector manufacturing growth through borrowed techniques from abroad, is most
likely to be capital intensive. This implies that while the wages of some lucky formal sector
employees may be high, the reservation wage of those who supply the pool of potential
employees, namely the average product of labor in agriculture will not rise. The consequence is
that more rural people may migrate to the cities in search of high paying formal sector
employment, with the result of larger urban unemployment, lower urban wages, and higher urban
poverty. This is a pattern that seems to have been followed in many Sub-Saharan Africa
countries.
Of course, the demand stimulus for higher valued agricultural products, and for rural based non-
agricultural activities does not have to come strictly from the agricultural sector. Broad-based
increases in urban incomes can also lead to a stimulus for rural incomes, especially if the
marketing margin from rural to urban areas is small. This indicates the two conditions that must
be fulfilled so that urban based growth can stimulate poverty reducing rural income growth,
namely the broad based nature of urban growth, and the reduction of the cost of rural-urban
marketing.
The channels through which agricultural growth and public investments in agriculture contribute
to overall income growth and poverty reduction have been made the object of empirical analysis
by two recent studies by Fan, Hazell and Thorat (1999) for India, and Fan, Zhang, and Zhang
(2000) for China. In both studies an econometric multiple equation model is estimated that
attempts to explain rural poverty, total factor productivity in agriculture or simply agricultural
production, rural wages, rural non-farm employment, the agricultural terms of trade and several
infrastructure variables. It is found in both studies that rural poverty is negatively and
significantly associated with agricultural total factor productivity (or growth in agricultural
production), rural wages, and non-farm rural employment. Agricultural total factor productivity
is explained by a sequence of infrastructure related variables, such as irrigation, rural roads,
literacy, electrification, as well as by expenditures on agricultural R&D. Wages and non-farm
employment in turn are determined by both agricultural and non-agricultural production
variables, as well as infrastructure spending variables. The most important variables in these
empirical analyses in terms of both TFP improvements, as well as rural poverty reduction are
investments in agricultural R&D, and investments in rural roads, followed by education. The
elasticities differ by country, which attests to the fact that local conditions and institutional
specificities matter. The results are, nevertheless, compatible with the view that it is not only
agricultural productivity growth that is important for poverty reduction, but also complementary
human and nonhuman infrastructure investments. The results also tend to corroborate the view
that it is through the stimulation of non-farm rural non-tradable activities that agricultural growth
helps to reduce poverty and at the same time enhance overall growth. It is clear that more studies
of individual countries are needed to generalize these findings.
All the above raise the question of the conditions that are conducive for agricultural growth to
have beneficial impact on overall growth. Delgado, et. al (1998) have outlined these conditions.
The first condition is that agriculture must account for a large share of aggregate employment.
The second is that agricultural growth must be equitable and evenly distributed. In other words it
 An Empiricaland ConceptualFramework 27

must allow a, large number of rural people to increase their incomes and hence demand. This
condition will be fulfilled when agricultural growth is targeted to products that are produced with
labor intensive technology and by a-broad range of rural producers. Initial asset distribution,
especially for land matters. The third condition is that the consumption patterns of the direct
beneficiaries of agricultural growth must be such that large shares of the increments to income
are spent on labor-intensive local nontradable goods and services. In other words the growth
multipliers are likely to be larger the less open the rural economy is, in the sense that the bulk of
the local economy consists of production and consumption of nontradables. The final condition
is that there mnust be a supply of underutilized local resources to make the supply of local
nontradables elastic, so as not to choke the increased demand for local nontradables by undue
increases in pnces.
Note that one 'of the key conditions is that agricultural growth must be broad based, in the sense
of benefiting a large share of the rural people. It is in this sense that agricultural growth is
poverty reducing. In order for such a condition to hold the productivity changes induced by
whatever interventions are made must be such as to touch a large share of the rural producers.
Nevertheless, Mellor (1999) points out that an initial skewing of the benefits of agricultural
growth towards the higher income rural producers is not in conflict with poverty reduction. He
points out that if these producers spend a large portion of their extra incomes on local
nontradables, then eventually the poor will benefit through local employment creation. This,
however, may take some time. Another aspect of the relationship between agricultural
development and poverty reduction is the relationship between agricultural development and
nutritional improvements. At low levels of income, rural households in developing countries are
not very well nourished, they spend a large share of their income on own produced staples, and
the income elasticity of demand for such staples is high. Thus agricultural productivity growth,
especially for staples, apart from improving incomes, can improve the nutritional status of many
of these households. Better nutritional status can in turn improve overall labor productivity, as
suggested by a variety of micro studies (see for instance Strauss, 1986, Behrman and Deolalikar,
1988). Better labor productivity can have growth implications. In a recent empirical analysis
Arcand (2000) 'finds a significant relationship between the initial level of the prevalence of food
inadequacy (PFI) or the dietary energy supply per capita (DES), and subsequent overall growth.
Given that these two nutritional variables are normally strongly associated with the overall levels
of poverty in a country, Arcand's results suggest that the initial degree of undernutrition and
poverty is detrimental to growth. Arcand subjects his results to a battery of robustness tests,
which preserve a significant relationship between the level of undernutrition and subsequent
growth.
Arcand further'finds that there are nutritional growth traps, in the sense that at low levels of
undernutrition, growth rates are sensitive to improvements in nutrition, while at higher levels of
undernutrition the relationship is much weaker. This suggests that at low levels of overall
nutrition it is efficient for a country to adopt policies that improve the overall nutrition level of
the population. Given the normally large shares of undernourished rural populations in most of
low-income developing countries2 , such policies almost invariably include agricultural
productivity growth for the rural poor. It is not clear whether Arcand's nutritional variables are

2
Arcand's tests identify 54 countries where nutritional improvement is likely to be helpful in improving overall
growth. These countries are mostly concentrated in Africa (35 out of 54), but there are also large Asian countries
included such as Afghanistan, Bangladesh, India, Philippines, and Thailand.
28 The Role ofAgriculture in Economic Development and Poverty Reduction:

proxies for the overall degree of poverty in a country. Even if they are, however, they still imply
that poverty alleviation in the form of decreasing the number of undernourished is efficient, in
the sense of improving growth. The results of Arcand are consistent with historical work
accounting for growth in now developed countries. For instance Fogel (1994) claims that
improved nutrition may have accounted for 30 to 50 percent of the growth in per capita income
in Great Britain between 1790 and 1980.
It thus appears from the above review of the literature that the conditions that can make
agricultural productivity growth to be both overall growth enhancing as well as pro-poor are the
following.
* Agriculture must account for a large share of aggregate employment.
* Initial distribution of land must be equitable and property rights must be well specified.
* The technological improvements must not be risk increasing, nor should they require
substantive private capital to be implemented.
* The marginal budget shares of the direct beneficiaries of agricultural growth for labor
intensive local nontradables must be large.
* There must be an excess supply of underutilized local labor resources.
* There must be complementary improvements in the provision of human capital assets at the
local level (education and health), as well as improvements in marketing infrastructure (e.g.
roads).
The consequences of agricultural development for the poor can be direct, through improved
agricultural incomes, or indirect, through the impacts on employment, wages, prices of products,
and productivity of non-farm assets. A major contribution of the research on agricultural growth
and poverty over the past decades has been to point out that the indirect impacts can be as large
or even larger than the direct ones, but may take some time to be realized.
DeJanvry et. al (2000) have shown that the shares of direct and indirect effects on poverty
reduction from agricultural TFP growth are vastly different in different institutional and
economic settings. They note that in an Asian context the indirect effects are likely to be much
larger than direct effects, and this implies that most of the benefits from agricultural TFP growth
on the poor arise from increased employment and unskilled wage increases, as verified by the
various studies of agriculture and poverty reduction in India. In Africa, the direct effects are
much more important, and this suggests that targeting technological change on poor farmers is
essential for poverty reduction. In Latin America by contrast the indirect effects are much larger
than the direct effects, but there the benefits to the poor from technological improvements are
likely to come through the declines in food prices.
What if the above conditions that make agricultural productivity growth have a beneficial impact
on overall growth and poverty reduction do not hold? Should one abandon agriculture and public
investments that favor it? Should one adopt policies to set the initial conditions right before
investing in agricultural development? The answers to the above questions are certainly not
clear-cut or generalizable, as they imply a real world where only second best policies can be
applied, but certain points may be made. First, if agriculture does not account for a large share of
aggregate employment, then it cannot easily be a leading sector for growth or poverty reduction.
Investments in agricultural productivity growth must then be judged by comparison with
investments in productivity growth of other sectors. Perhaps the best policies under such
circumstances maybe investments in human capital, so as to allow agricultural producers to adapt
 An Empiricaland ConceptualFramework 29

more efficiently to technological changes. Second, if it is difficult to change the initial conditions
in the short run, then, while working toward changing them in the medium to long run, one may
want to avoid the types of productivity growth that may favor adversely factors that are
unequally distributed. For instance, if the land distribution is skewed, then public research in
agricultural development may be better targeted towards labor intensive techniques rather than
land augmenting ones. Third, if there is no large local excess supply of labor, that could be
mobilized when the demand for rural non-tradables increases, then perhaps it is prudent to
concentrate on improving the marketing infrastructure, and hence lower the marketing cost
between rural and urban areas. This would prevent the prices of local non-tradables from
increasing too much from any agricultural stimulus, by essentially making such products more
tradable. The, above points are not easily generalized, and the presence of only a subset of the
above conditions implies that agricultural development policy in such circumstances must be
judged on an individual country basis.6. Historical lessons from country experience 3
Agricultural development played different roles in growth and poverty reduction in the now
developed countries. The development paths of developed countries like the United Kingdom
and the United States, relied primarily on capital intensive industrial development in the early
phases that induced slow growth in labor demand. As at the same time the labor supply was
increasing, the real wages of unskilled workers, the primary factor determining incomes of the
poor, grew slowly and hence there was little impact on poverty. During the early phases of
industrialization of these countries, namely from the late eighteenth to the mid-nineteenth
centuries, agricultural productivity increased only a little. In the United Kingdom, the early
agricultural technological improvement of the eighteenth century, consisted of the development
of an integrated crop-livestock system of husbandry, which was more labor intensive than the
traditional one, but increased the productivity of land, while the increases in net returns to labor
were small. However, the total employment increased, and this gave work to many
underemployed workers. The English agricultural revolution did not supply surplus labor for
industry, but rather food for the rapidly rising population from which both agricultural and
industrial labor was recruited. Hence real wages did not rise. It is only since 1840 that the
incidence of poverty in the United Kingdom started declining, and this because of a decline in
the overall labo,r force growth that led to real wage increases.
The expansion of the land frontier in the United States induced labor saving technological
change, and increased overall production. The increase in agricultural productivity and
consequent expansion of production, led to slow secular declines in the prices of food, and it is
this, along with, the slowdown of the growth rate of labor supply that induced real wage increases
and consequent declines in poverty. It is also only since the mid-nineteenth century that in the
US agricultural technology and labor productivity in agriculture started improving, thus inducing
a more balanced pattern of growth, and favoring more labor intensive techniques.
By contrast in countries like Japan development in the nineteenth century occurred through a
broad based strategy to improve subsistence-oriented agriculture, a strategy that led to both
reduced poverty as well as improved growth. The Japanese model was transplanted to Taiwan,
which under the Japanese colonial administration, by the 1920s had a well established
agricultural research system, irrigation infrastructure, availability of chemical fertilizers, and a

3 Thisvery brief description is taken mainly from Hayami and Ruttan (1971) and Mellor (2001), and is not meant to
be comprehensive.
30 The Role ofAgriculture in Economic Development and Poverty Reduction:

well developed transportation and marketing system, so as to facilitate the provision of marketed
surplus to Japan. The introduction of new rice varieties in the 1920s induced large increases in
fertilizer use, and increased yields considerably. In contrast to the English agricultural
revolution, in Taiwan, labor productivity and the labor earnings of individual cultivators rose
concurrently with the rise in land productivity, which resulted from the introduction of modem
biological technology. The increased incomes of farmers permitted the transfer of resources to
non-agriculture, primarily small rural based industries, as well as to government through
taxation. However, the government helped the agricultural transformation by improving rural
infrastructure, and heavily investing in rural education and rural public health. The consequence
was not only fast agricultural and non-agricultural growth, but also a decline in poverty.
The success of the green revolution in India and other Asian countries has been well documented
and needs no further discussion. In India, for instance, in the twenty year period since the
introduction of the new varieties, poverty levels dropped nearly to half. This occurred partly due
to increased employment of the poor in agriculture, but mostly because higher farm incomes led
to increments of expenditures in local employment intensive goods and services, which in turn
spurred the creation of many rural small scale enterprises to provide these goods with labor
intensive techniques, in a setting with abundant and underemployed labor.
The story of China has already been mentioned. The Chinese success story relied heavily on
agricultural development, spurred by both broad based agricultural technological improvements,
as well as local infrastructure investments and rural education. The increase in rural incomes led
to increases in demand for other products that were also provided largely by rural-based small-
scale labor intensive industries. The number of absolutely poor as a consequence declined
substantially from 250 million in 1978 to 125 million in 1985 and to less than 100 million in the
mid-1 990s.
The experience of Bangladesh is similar. Rapid broad based agricultural technological change, in
conjunction with improvements in rural infrastructure, and coupled with the provision of a food
security system for the poor, has made Bangladesh much less vulnerable to floods, and has
reduced poverty (Mellor, 2001).
Mellor (2001) cites the example of Malaysia, a country with physical agricultural resources
similar to those of Western Africa, as a success story of agricultural development and poverty
reduction. The agricultural growth of Malaysia came from exportable cash crops like rubber and
palm oil, rather than food crops. However, Malaysia adopted a policy of support for research and
rural infrastructure that led to considerable yield increases that induced large income increases
from the many small producers of these products. The intensification of rice production also led
to increases in rural incomes that were translated to increased non-farm employment. The
multipliers from such agricultural development strategy have been documented by Hazell and
Roell (1983) and they were large.
Mellor provides other examples of successful agriculture led growth and poverty reduction
stories. These include Kenya, Costa Rica, and Indonesia. In all cases the major impetus to
success has been agricultural broad-based growth, with spillovers to non-agricultural sectors.
Also in all cases the agricultural transformation has been aided by infrastructure investments and
rural education. These are examples that reinforce the points made earlier about the conditions
that are conducive to agriculture led growth and poverty reduction.
6. Emerging Issues of Agricultural Development
and Poverty Reduction
In this chapter we discuss some issues that are bound to become more important in the future, as
they pertain to agricultural development and poverty reduction.

6.1 Globalization
Globalization implies both fewer trade restrictions, as well as that the various trading countries
may be less insular in the future, allowing for lower protection, better transmission of
international prices and general market signals to domestic markets, but also allowing domestic
shocks to be absorbed through international transactions. There are several issues that arise in
this context. First, the general tendency towards trade liberalization is likely to lead to increased
average international prices for most products (Goldin and van der Mensbrugghe, 1995). While
agricultural trade liberalization has been the subject of discussion mostly among developed
countries, there are implications for developing countries. These involve significant average rises
in prices for cereals like wheat, rice, feedgrains, and sugar. Many of these products are of import
or export significance for developing countries. Sadoulet and deJanvry (1992) found that
increases of world prices for cereals and animal products are likely to have detrimental welfare
impacts on the landless and small farmers, as well as the urban poor in African and Asian
archetype economies, representative of countries with high levels of poverty.
Second, there may be implications about the instability of international prices. While theory
suggests that mnore trade liberalization implies more stable prices, there were concerns expressed
during the 1995-96 cereal price rise that international price instability might have increased as a
consequence of agricultural trade liberalization. This, however, does not appear to be the case
(Sarris, 2000).
Third, general liberalization of marketing in many developing countries since the onset of
structural adjustment programs, has led to the abolition or restructuring of most marketing
parastatals, abolition of pan-territorial prices, abolition of other marketing controls, etc. Such
measures, while removing the price biases against several products, have also removed the
previous stability of prices. They also may have led to more realistic incorporation of regional
marketing margins into producer prices, a process that may have led to secular price declines and
greater price instability for producers located in regions that are not well integrated with the main
domestic and international markets.
Thus the consequences of trade liberalization and globalization for the poor agricultural
producers depends on the types of products they produce, and on their location, and it is likely
that there are gainers and losers among the poor within any given country. Nevertheless, if better
response to the signals generated by increased globalization is to be sought, then it appears that
more emphasis must be placed on rural infrastructure so as to lessen the cost of marketing, and
make remote regions more integrated with the rest of the economies.
32 The Role ofAgriculture in Economic Development and Poverty Reduction:

6.2 Biotechnology
The developments in biotech research suggest that both the structure of research as well the
ownership of research will be different in the future. Biotechnology is based on the
understanding of how biological organisms function at the molecular level, and manipulation of
the DNA molecules from which genes are made to achieve desirable outcomes. The introduction
of intellectual property rights (IPRs) has led to substantial new research with extensive
applications to agriculture.
There are considerable potential benefits of agro-biotechnology for poverty reduction (de Janvry
et. al. 2000). These include:
* Yield increase in staple food crops produced in tropical and semi-tropical environments,
and in peasant farming systems.
* Area expansion towards less favored lands.
* Multiple cropping and shorter maturation periods.
* Cost reductions via resource savings.
* Risk reduction via lower susceptibility to pests and viruses, as well as to frosts, droughts,
etc.
* Improved storability via insect resistance, delayed maturation, etc.
. Nutritional improvements as food and feed.
* Environmental benefits, like reduction of chemical pesticides.
There are also risks for poverty reduction such as:
* Staple food crops produced in tropical and semi-tropical environments are bypassed by
research.
* Labor displacement by diffusion of herbicide tolerant plant varieties.
* Production in developed and middle level developing countries of substitutes for crops
previously produced by developing countries.
* Tailoring of research towards products not fit for poor consumers.
* Terminators genes to enforce IPR may raise costs by preventing reproduction of seeds.
De Janvry et. al. (2000) suggest that new biotech research results in traits that can be usable over
a wide range of local conditions, and this implies that if the institutions and information are
available developing countries do not need to engage in fundamental research. However, the
access to new technology may be impaired by the increasing tendency towards patenting of new
gene technologies by biotechnology firms, and the wider enforcement of IPRs. This may result in
improved seeds that are more expensive, incorporating the cost of innovation, as well as patent
protection. If the adoption of such seeds requires more initial working capital, the adoption of
genetically modified seeds (GMOs) may become more difficult for the poor. Other studies
surveyed by Perrin (1999) also suggest that developing countries may not benefit from the
strengthening of IPRs. Hence, the way in which biotech is applied to developing country
agriculture may be very important for the poor. De Janvry et. al. suggests a variety of
institutional measures to make the benefits from biotech research filter more easily towards the
poor.
 An Empiricaland ConceptualFramework 33

6.3 Changing Roles of the Public and Private Sectors

There are three important ways in which the role of the public sector has been changing in the
agriculture of developing countries. The 'first concerns the role of the public sector in providing
price stability and market insurance in agriculture. It has already been mentioned that structural
adjustment has reduced the role of monopolistic public marketing agencies, and has increased
price and possibly income instability for all agricultural producers, and most likely also for the
poor. While there is no doubt that the better functioning of markets, through wider participation
of market agents, for products, inputs, labor, and capital are bound to benefit all, and especially
the poor, the increased market uncertainty may make it even more difficult for the poor to escape
poverty. The reason may have to do with the level of precautionary savings that need to be held
by poor households to buffer undesirable income variations in the presence of incomplete
insurance andlcredit markets. If this amount depends on the degree of risk faced by households
(Paxson, 1992), then the increased market uncertainty implied by the disengagement of the state
from marketing may induce more precautionary savings by the poor, and hence limit the savings
available for more profitable productive investments. This suggests that the role of the public
sector in the context of more open markets must be redefined more as having an insurance
function, rather than a direct marketing function.
The second aspect where the role of the public sector is changing involves the provision of
infrastructure and community services such as health, education, sanitation, etc. While in the past
the provision of such services was envisioned almost exclusively as the domain of the public
sector, the new thinking sees a more mixed approach where both public and private provision are'
involved. There are considerable issues involved in both the financing of infrastructure and
service projects, as well as the pricing of the services (for a review see Jimenez, 1995). Given the
proven contribution of infrastructure investments on productivity of rural residents, and the
generally redistributive role of most infrastructure investments, it is important that the new
mixed approach does not end up making the poor shouldering a larger relative burden for
infrastructure provision.
The third role involves research. It was already seen above that the private sector is much more
involved in new biotech research compared to earlier periods. It has also been observed that there
is increasing concentration of patents in these new fields (de Janvry et. al. 2000). The current
patenting system on life forms allows private appropriation of knowledge on genomics, the basic
material of biotech research (Wright, 1998), and this may prevent access of relevant materials for
research in developing countries. However, the patenting system in this field is still evolving.
There may also 'be undue concentration of knowledge in few firms through the concentration in
the ownership of the many patents involved in the production of a final product. De Janvry et. al.
2000 note that in 1999 75% of the patents obtained by the largest 6 agro-biotech firms were
obtained through acquisitions of smaller biotech and seed companies. This may prevent open
rights to the use of some biotech products by institutions of developing countries. Finally,
enforcement of IPR legislation leads to more incentives for innovative activity, but at the same
time may make ,the cost of new products higher for the poor. Thus there may be a trade-off
between stronger enforcement of IPR laws and poverty reduction.
34 The Role ofAgriculture in Economic Development and Poverty Reduction:
7. Modeling Public Policy Towards Agriculture in
the Context of Growth and Poverty Reduction
In the previous chapters it was seen from the literature review that while there are some
theoretical attempts to model agriculture in the contest of growth and poverty alleviation, they
have not incorporated the issue of how productivity growth is to be implemented and financed.
To explore these issues, a simple two-sector model of agriculture and growth was built, that
incorporates explicitly both TFP growth as well as financing of TFP improvements in agriculture
and non-agriculture. The idea is to obtain general results about the tradeoff between taxation of
the two sectors, as well as allocation of public capital to the two sectors. As far as the author is
aware this is the first such model that has been analyzed in the literature. The details of the
model structure and the theoretical and empirical results are exhibited in the appendix. Here we
only discuss the structure and the highlights of the results.
The model considers a small open economy with two sectors, agriculture and non-agriculture.
The economy is assumed to export the agricultural product, while the non-agricultural product is
non-traded. However, a non-domestically produced intermediate product is imported and used in
the production of the non-agricultural sector. The non-agricultural sector is the only one
producing investment goods. Such an economy, resembles many of the poor economies in
Africa, as well as in some other parts of the world.
The model assumes that infrastructure or other development spending (e.g. for research, human
capital, etc.) improve the total factor productivities of the two sectors, with different elasticities.
However, the expenditures for these productivity enhancements must be financed, and this is
done by taxation of both the agricultural as well as the non-agricultural sectors, at different rates.
It is assumed that the government devotes a fraction of its total public resources for investment in
the agricultural sector. Full employment of labor is assumed, and this makes the growth linkages
of the model smaller than what they would be under a non-full employment structure. The
dynamics of the model assume that capital allocation is more sluggish in the short run, compared
to labor. Capital accumulation occurs through the saving of private agents, while the government
is assumed to have a balanced budget.
The dynamic accumulation equations of the model imply a growth rate that is a function of the
share of public capital devoted to agriculture, the labor productivities of the two sectors, and the
degree of trade openness of the economy. However, the resulting expression for the growth rate
is too complicated for general results, and some simulation exercises are performed and tabulated
in order to obtain an idea of the contribution of different factors to economic growth. Under
some assumptions about the structure of poverty, it is also possible to assess the contribution of
agricultural TFP growth on poverty reduction.
The first conclusion that obtains is that for optimal, namely maximum under the various
constraints, growth there is a negative relation between the two sectoral tax rates. A high
agricultural tax, rate implies a low non-agricultural tax rate and vice-versa. Also a zero
agricultural tax' rate necessitates a positive non-agricultural tax rate. This is necessary, as by
assumption the existence of public capital is essential for production in both sectors, and hence at
36 The Role ofAgriculture in Economic Development and Poverty Reduction:

least one of the sectors must be taxed to generate revenues to maintain public capital. In fact for
every agricultural tax rate one can compute the optimum non-agricultural tax rate that maximizes
the growth rate of the economy. Hence, one only needs to consider the agricultural tax rate as a
policy instrument.
A second empirical result is that at any given degree of openness, higher values of the
agricultural TFP elasticities, namely greater ease with which agricultural public sector
investments enhance agricultural production, imply lower rates of overall growth. However, they
imply higher rates of growth of the real incomes of the poor. Hence it appears that variations of
agricultural TFP elasticity imply a tradeoff between higher overall growth rates and higher rates
of real income growth of the poor.
A third conclusion is that in order to maximize overall growth, the rate of agricultural taxation is
a decreasing function of the degree of trade dependence of the non-agricultural sector. Hence, an
economy with high trade dependence in the production of its non-tradable good, does not need
high degrees of agricultural taxation, in order to maximize growth. The opposite is true for
countries with low degree of trade dependence. The optimum share of public capital going to
agriculture is smaller than the share of agriculture in GDP, or the share of labor in agriculture.
However, the optimal share of public capital going to agriculture is a positive function of the
degree of trade openness.
Finally, it is appears that low labor intensities in both sectors imply low rates of growth, but high
shares of public capital in agriculture. On the contrary, high labor intensities in both sectors
imply high growth rates in steady state, and low shares of public sector capital in agriculture.
Interestingly, going from low to high degrees of labor intensity of agriculture, the implied steady
state rate of growth rate of the economy increases, but the real income of the poor first declines
and then increases.
The conclusion from the simulations is that the parameters that matter a lot in determining the
usefulness of agricultural development as an engine of growth as well as poverty reduction are
the degrees of labor intensities of the two sectors, the degree of trade openness, the elasticity of
agricultural TFP with respect to public capital, and the rate of domestic savings.
Concerning policy, the conclusions of the simulations are that some agricultural taxation
certainly helps the overall rate of growth in small open economies, with an agricultural sector
that produces the bulk of exportables. The appropriate non-agricultural tax rate is inversely
proportional to the agricultural tax rate. The share of public sector capital that must be devoted to
agriculture is not related to the share of agriculture in GDP, or the share of labor employed in
agriculture, but is rather a function of the labor shares of the two sectors, the import intensity of
non-agriculture, and the elasticity of agricultural TFP with respect to public capital.
 8. Conclusions and Implications for the World Bank
Sustained growth in an economy requires the continuous improvement in TFP, and this for a
developing country requires public expenditures for infrastructure and human capital
developments Such improvements allow private agents to obtain higher levels of income, and
this makes possible higher levels of consumption and investments. However, it appears that at
early stages of development, in order for an economy to move from a situation where it is
"...saving and investing 4 or 5 percent of its national income or less, to a an economy where
voluntary saving is running at about 12 to 15 percent" (Lewis, 1954), which Lewis regards as the
central problem of economic development, the economy needs an engine of growth. Such an
initial engine of growth can come from a variety of sources, such as the development of export
agriculture, industry, tourism, etc., and the key issue for development is what is done with the
increased incomes and savings that come about from the initial engine of growth.
While in the now developed countries it appears that the major stimulus to early growth was
industrial innovations, accompanied but not led by agricultural innovations, there have been
some developing countries in the recent past, notably India, China and Taiwan, where growth
was led by agricultural broad based productivity changes. In these countries it seems that the
major source of the demand for the increased product of the agricultural sector was domestic, as
there were substantial levels of initial poverty, and hence improved agricultural incomes directly
led to increases in the domestic demand for the larger quantities of food produced domestically.
Hence the terms of trade for the increased agricultural product did not decline so as to negate the
improvements in TFP. Another important historical characteristic of these economies is that a
substantial part of the cost of their agricultural productivity improvement was shouldered by
external donors, including the substantial contributions of the international agricultural research
centers, members of the CGIAR. Finally, it appears that given the substantial rural population
densities in these countries, the cost per agricultural producer of improving agricultural
productivity was relatively low, as there was no need for experimentation in too many
climatically different locations.
The situation may well be different in many of the late developing countries, such as those in
many parts of Africa, in the sense that the cost per beneficiary of agricultural productivity
improvement may be high because of low farm population densities, at the same time that overall
donor support is declining, necessitating a larger contribution to domestic productivity
improvement by the state. Can agriculture play a leading role in the settings of the late
developing countries, and should cash strapped governments devote a substantial share of their
meager resources to agricultural TFP improvements?
The review made in this paper first examined the way in which agriculture grows. It was seen
that the basic ingredients that make up for faster agricultural TFP growth are known, such
as agricultural R&D, extension, rural infrastructure, education, etc. However, it was pointed
out that while we know the variables affecting agricultural TFP, the profession is much less
sure about the magnitudes of the relevant elasticities, as well as the ways in which these
elasticities are affected by other factors. For instance it may well be that both structural
parameters such as distributional variables, as well as institutional factors, such as the degree of
38 The Role ofAgriculture in Economic Development and Poverty Reduction:

market imperfections may impinge on these elasticities. While considerable partial knowledge on
these issues exists at the micro level for several developing countries, their contribution to
agricultural TFP growth or to the magnitude of the elasticities of TFP growth with respect to
other variables is much less well analyzed. For instance how does the contribution of the
provision of rural roads to agricultural TFP is affected by other market imperfections? This set of
issues then is an area of considerable lack of knowledge, where more research would have a high
pay-off.
The review also revealed that there is by no means an unequivocal theoretical argument for
agricultural development as an engine of growth. If any, recent theoretical contributions tend to
rely on different structural properties between agriculture and non-agriculture concerning
external effects, notably learning by doing, to argue against agricultural productivity
developments as engines of growth in an open economy. It was noted in the review, however,
that such treatments rely on asymmetric assumptions about the economic characteristics of
agriculture and non-agriculture, and hence cannot be utilized as guides to policy unless one has
more information at the sectoral level about the nature of external effects and learning by doing
properties of the various sectors.
One important contribution of recent theoretical research has been to point out that the way in
which domestic incomes, are affected by the improved agricultural productivity, or any
other stimulus by another leading sector, and the consequent consumption and saving-
investment patterns are crucial for determining the pattern of growth. If, for instance,
improved agricultural productivity leads to higher incomes of the poor who spend it on domestic
non-tradables, then agricultural growth will induce non-agricultural rural growth and via
employment multipliers to decreases in poverty. If, on the other hand, the fruits of agricultural
development or other leading sector economic development lead to increases in the incomes of
the rich, then the important determinant for growth will be where the additional savings are
spent. If they are spent for domestic labor intensive investments, then there may still be growth,
and the poor may well benefit from employment creation. If, however, they are spent on
imported luxuries, or invested abroad, there will be little growth stimulus. Thus the pattern of
the distribution of increased incomes from the initial stimulus is the most important
determinant of subsequent growth. Agricultural development can thus contribute to both
growth and poverty reduction if the fruits of the initial productivity stimulus is
concentrated on those who respend it domestically (through consumption or investment) on
labor intensive products with low import dependence.
Chapter 5 took up the issue of agriculture's contribution to poverty reduction. It was noted that
there may be both direct and indirect contributions to poverty reduction from agricultural
development, depending on the structure of the incomes of the poor. It was noted that while
recent empirical literature has highlighted the important role of agriculture for poverty reduction
in labor abundant agrarian countries like India and China, the relationship is not universal. In fact
a major contribution of the recent empirical growth literature has been to point out that the
elasticity of poverty with respect to agricultural productivity improvements depends on
initial distributional variables. This is consistent with the literature on the role of agriculture
for growth, which highlights precisely the distributional dependence of the contribution of
agricultural TFP growth to overall growth. However, it seems that the elasticity of poverty
reduction should not depend only on distributional variables. A variety of institutional and
 An Empiricaland Conceptual Framework 39

other features, which may be of importance, cannot be captured precisely in cross-country

growth analysis, and need to be studied through specific country case studies.
The review revealed that there seem to be a set of conditions that make agricultural
development both growth enhancing as well as poverty reducing. These conditions involve
the following:
* Agriculture must account for a large share of aggregate employment.
* Initial distribution of land must be equitable and property rights must be well
specified.
* The technological improvements must not be risk increasing, nor should they
require substantive private capital to be implemented.
* The marginal expenditure shares (both from consumption as well as from
investment) of the direct beneficiaries of agricultural growth for labor intensive
local non-tradables must be large.
* There must be an excess supply of underutilized local labor resources.
* There must be complementary improvements in the provision of human capital
assets at the local level (education and health), as well as improvements in
marketing infrastructure (e.g. roads).
* There must an income and price elastic source of demand for the increased
product of agricultural, wither domestically (in the case of food crops), or
internationally (in the case of exports).
The review highlighted some important emerging issues concerning agricultural development.
Globalization can create opportunities, for instance through improved agricultural terms of trade,
or improved ml arket access for a country's agricultural exports. However, it may also pose
additional risks, such as increased domestic price instability. The point made is that globalization
must be accompanied by domestic market enhancing policies in order to produce positive overall
results from agricultural growth and for poverty reduction.
Biotechnology was also seen to pose risks and opportunities. Participation by developing
countries in the still to be shaped international rules of the game concerning technology transfer
and IPRs seems the best avenue for developing countries in order not to be left out of these
developments.
The increasing role of the private sector in both developing countries was seen to pose new
challenges for the public sector in the context of agricultural development. Joint private-public
financing and responsibility for infrastructure, and research may open opportunities that were not
previously available. Nevertheless, the role of the public sector in the provision of insurance
seems to be enhanced in the new context.
The elaboration of a two-sector theoretical model of agricultural growth, based on external
effects of public infrastructure investments, and the ensuing empirical simulations suggested
some tentative conclusions concerning the role of agriculture in growth enhancement and
poverty reduction. First it was shown that maximum growth involves the provision of
adequate public infrastructure, financed by higher levels of taxation. However, it was shown
that there is a negative relationship between agricultural and non-agricultural taxation for
optimum growth. It was seen that higher elasticities of agricultural TFP with respect to
public expenditures imply faster real income increases for the poor, while lower overall
40 The Role ofAgriculture in Economic Development and Pbverty Reduction:

rates of growth. This trade-off is rather surprising and needs further study and elaboration. It
was also seen that the optimum share of public investment expenditures devoted to
agricultural TFP improvement depends of the degree of trade dependence of the non-
agricultural sector, increasing with higher degrees of trade dependence. This is contrary to
earlier theoretical results. Finally it was shown that the labor intensities of the agriculture, as
well as the non-agricultural sectors, and the aggregate saving rates are important in
determining the maximum possible growth under a strategy of agricultural development.
The implications of the above analysis for international financial institutions involved in
development, such as the World Bank, and other development banks, are the following. First, it
should be clear that the relative importance of the agricultural sector as a leading sector for
growth and poverty alleviation depends on country specific geographical and economic
structure variables. There can be no single agricultural growth and poverty reduction strategy
to fit all developing countries. It rather seems that the emphasis on agriculture as a leading sector
should depend on a set of criteria of the type mentioned above. Country specific strategy and
policy work could then concentrate on the presence of the conditions that may be conducive
to making agriculture a growth and poverty alleviation pole. Some of these conditions were
elaborated above, but it seems that there needs to be more organized analysis of these criteria and
conditions, combined with specific indicators, to provide specific empirical and country relevant
guidance.
Second, there is no doubt, given the considerable externalities involved in agricultural TFP
growth, that external assistance should concentrate on- creating an environment that
facilitates the productivity of any attendant domestic investments. For instance, if the
government of a developing country is keen on rural infrastructure, perhaps, external resources
could support the development of the other necessary ingredients for growth and poverty
reduction, such as rural education and health, with emphasis, of course, on sustainability of
investments. This can be justified under the notion that there are complementarities between
various types of agriculture related public interventions. It must be noted in this context that
there seems to be considerable lack of knowledge concerning the dependence of elasticities
of both growth, as well as poverty reduction to agricultural development spending on
institutional, as well as other structural factors. This seems an area, where more country
specific, as well as cross-country research is needed.
Another point has to do with the type of agricultural research that is supported. In many
developing country agrarian settings, the reservation unskilled wage for the economy is
close to the average product of labor in agriculture. This for instance seems to be the case in
many land abundant but labor constrained economies in Africa. This implies that in such
settings, agricultural research and productivity enhancement, in order to be both growth
enhancing, as well as poverty reducing, must aim at increasing the average productivity of
labor in agricultural production, without making production more labor intensive. This is
different from settings with labor abundance, and land scarcity, where the reservation wage
may be close to the marginal product of labor in agriculture. In such settings, agricultural
research should try to increase the productivity of land, so as to increase the marginal
product of agricultural labor.
A major strategic issue has to do with the efficacy of agricultural productivity growth as an
engine of growth as a function of rural population density. In settings of low rural population
densities, the cost of agricultural productivity enhancement may be very large relative to the
An Empiricaland Conceptual Framework 41

benefit per affected household. In such settings the elasticity of TFP growth, as well as the
elasticity of poverty- reduction with respect to various types of public expenditures may be low. It
was seen in the empirical simulations of the model (re. Table 2) that low values of the agriculture
TFP elasticities imply low shares of public capital devoted to agriculture TFP enhancement. This
underscores the fact that the factors that determine the size of the elasticities of agricultural
TFP with respect to public spending must be clearly understood, before recommendations
about a rural development strategy are made.
There are several areas where additional work is needed. First, it is not clear how many
contemporary developing countries fulfill the conditions that were identified as necessary for a
win-win (growth promoting and poverty reducing) agricultural development strategy. While for
some conditions the verification is relatively easy, examples being the share of agriculture in
total employment, the initial land distribution, and the existence of under-utilized rural labor
resources, for' others the verification is by no mean straightforward. This may well be an
interesting empirical project for the World Bank Rural Development Department.
Second, even if all conditions can be empirically checked, it will most likely be the case that not
all of them are satisfied in any one developing country. Does this mean that agricultural
development policies should take second priority? Or is it the case that there is a core subset of
the conditions that if satisfied can justify a vigorous agricultural development strategy. This
poses the problem of agricultural development policy under a second best world. For instance,
should development effort concentrate on setting the initial conditions for agricultural
development right as a prerequisite for public investments in agriculture, or should the two
proceed simultaneously? Unfortunately, very little is known about the relative importance of the
above conditions.
Third, are all conditions identified of equal importance? For instance is the provision of adequate
roads more important than the provision of human capital? Should rural education be enhanced
before investments in agricultural research and extension are made? The answers to these and
other related questions are not easy, and are not universal, as they are intimately tied with the
historical and institutional context of any given country, as well as on an appraisal of the speeds
with which any one policy can be implemented and have an impact.
Given this rather inconclusive state of affairs regarding the role of agriculture in growth and
poverty reduction, one way to proceed for the Rural Development Department of the World
Bank to obtain such knowledge, would be to conduct a cross-country comparative review of
agricultural sector strategies and performances over the past twenty or thirty years. Such reviews
have been done at the initiative of the World Bank in the past, but with very different focus, and
have produced wide ranging and policy relevant results. The new focus should be growth and
poverty reduction performance.
42 The Role ofAgriculture in Economic Development and Poverty Reduction:
References
Adams, R.H. Jr., and He, J.J. (1995). Sources of Income Inequality and Poverty in Rural
Pakistan. Washington DC, International Food Policy Research Institute, Research Report 102.
Adelman, I. (1984), "Beyond Export-Led Growth", World Development, vol. 12, No. 9:
937-49.
Aghion, P. Caroli, E., and Garcia-Penalosa, C. (1999), "Inequality and Growth: The
Perspective of New Growth Theories", Journal of Economic Literature, vol. XXXVII(4), pp.
1615-1660.
Ahluwalia, M.S. (1976), "Inequality, Poverty and Development", Journal of
Development.Economics, vol. 3, pp. 307-342.
Anand, S. and Kanbur, S.M.R. (1993), "Inequality and Development: A Critique:, Journal
of Development Economics, vol. 41, pp. 19-43.
Antle, J. (1983), "Infrastructure and Aggregate Agricultural Productivity: International
Evidence", Economic Development and Cultural Change, vol. 31, pp. 609-619.
Arcand, J.L. (2000), "Malnutrition and Growth: The Efficiency Cost of Hunger", report
prepared for the Policy Analysis Division of the Food and Agriculture Organization of the
United Nations (FAO, ESAE), Rome.
Bardchan, P.K., and C. Udry (1999). Development Microeconomics. Oxford, Oxford
University Press.
Barro, R.J. (1990), "Government Spending in a Simple Model of Endogenous Growth",
Journal of Political Economy, vol. 98, No. 5, Part 2, pp.S103-S 105.
Barro, R.J., and Sala-i-Martin, X. (1995). Economic Growth. New York, McGraw-Hill.
Behrman, J. and Deolalikar, A. (1988), "Health and Nutrition", in H. Chenery, and T.N.
Srinivasan (editors). Handbook of Development Economics, vol. 1, Amsterdam, North Holland.
Bell. C.~(1989), "Credit Markets and Interlinked Transactions", in H. Chenery, and T.N.
Srinivasan (editors). Handbook of Development Economics, vol. 1, Amsterdam, North
Holland.Bernard, A., and Jones, C.I., (1996) "Productivity Across Industries and Countries:
Time Series Theory and Evidence", Review of Economics and Statistics LXXVIII(l): 135-46.
Besley, T. (1995), "Savings, Credit and Insurance", in J. Behrman, and T.N. Srinivasan
(editors). Handbook of Development Economics, vol. IIIA, Amsterdam, Elsevier.
Binswanger, H.P., Deininger, K., and Feder, G. (1995), "Power, Distortions, Revolt, and
Reform in Agricultural Land Relations", in J. Behrman, and T.N. Srinivasan (editors). Handbook
of Development Economics, vol. IIIB, Amsterdam, Elsevier.
Binswariger, H., and Von Braun, J. (1993), "Technological Change and
Commercialization in Agriculture", in M. Lipton, and J. van der Gaag (editors), Including the
Poor, Washington DC, The World Bank.
Binswanger, H., Yang, M-C., Bowers, A., and Mundlak, Y. (1987), "On the
Determinants of cross-country Aggregate Agricultural Supply", Journal of Econometrics, vol.
36: 111-131.
44 The Role ofAgriculture in Economic Development and Poverty Reduction:

Birdsall, N., and Londono, J.L. (1997), "Asset Inequality Matters: An Assessment of the
World Bank's Approach to Poverty Reduction", American Economic Review, vol. 87(2), pp. 32-
37.
Bond, M.E. (1983), "Agricultural Responses to Prices in Sub-Saharan Africa",
International Monetary Fund Staff Papers, vol. 30(4), pp. 703-26.
Bourguignon, F. and Morrisson, C. (1998), "Inequality and Development: The Role of
Dualism", Journal of Development Economics, vol. 57(2), pp. 233-257.
Bruno, M. Ravallion, M., and Squire, L. (1996), "Equity and Growth in Developing
Countries: Old and New Perspectives on the Policy Issues", World Bank Policy Research
Working Paper, 1563.
Byerlee, D. (1996), "Modern Varieties, Productivity, and Sustainability", World
Development, vol. 24(4), pp. 697-718.
Canning, D. (1988), "Increasing Returns, in Industry and the Role of Agriculture in
Growth", Oxford Economic Papers, vol. 40: 463-476.
Coeymans J. E. and Mundlak, Y. (1993) Sectoral Growth in Chile: 1962-82. Washington
DC. International Food Policy Research Institute. Research Report 95.
Craig, B.J., Pardy, P.G., and Roseboom J., (1997) "International Productivity Patterns:
Accounting for Input Quality, Infrastructure and Research", American Journal of Agricultural
Economics, vol. 79(4): 1064-76.
Das, S.P., (1982), "Sector Specific Minimum Wages, Economic Growth, and some
Policy Implications ", Journal of Development Economics, vol. 10, No. 1, pp. 127-131.
Datt, G. and Ravallion, M. (1998), "Farm Productivity and Rural Poverty in India", The
Journal of Development Studies, vol. 34(4), April, pp. 62-85.
deJanvry, A., Gordillo, G., Platteau, J-P., and Sadoulet, E. (editors) (2001). Access to
Land, Rural Poverty and Public Action. Forthcoming, Oxford University Press.
DeJanvry, A., Graff, G. Sadoulet, E., and Zilberman, D. (2000), "Technological Change
in Agriculture and Poverty Reduction", background paper prepared for the World Bank WDR
2000.
deJanvry, A. and Sadoulet, E. (2000), "Growth, Poverty and Inequality in Latin America:
A Causal Analysis, 1970-94", Review of Income and Wealth, Series 46(3), pp. 267-87.
Deininger, K. and Squire, L. (1996), "A New Data Set Measuring Income Inequality",
The World Bank Economic Review, vol. 10(3), pp. 565-91.
Deininger, K. and Squire, L. (1998), "New Ways at Looking at Old Issues: Inequality and
Development", ", Journal of Development Economics, vol. 57(2), pp. 259-288.
Delgado, C.L, Hopkins, J. and Kelly, V.A. with P. Hazell, A.A. McKenna, P. Gruhn, B.
Hojjati, J. Sil, and C. Courbois (1998). Agricultural Growth Linkages in Sub-Saharan Africa.
Washington DC, International Food Policy Research Institute, Research Report 107.
Evenson, R.E., Pray, C.E., and Rosengrant, M.W. (1999). Agricultural Research and
Productivity Growth in India, Washington DC, International Food Policy Research Institute,
Research Report 109.
An Empirical and ConceptualFramework 45

Evenson, R.E., and Westphal, L. (1995), "Technological Change and Technology

Strategy", in J. Behrman, and T.N. Srinivasan (editors). Handbook of Development Economics,
vol. IIIB, Amsterdam, Elsevier.
Fafchamps, M. and Pender, J. (1997), "Precautionary Saving, Credit Constraints and
Irreversible Investment: Theory and Evidence from Semi-Arid India", Journal of Business and
Economic Statistics, vol. 15(2), pp. 180-97.
Fan, S. Hazell, P., and Thorat, S. (1999). Linkages between Government Spending,
Growth and Poverty in Rural India. Washington DC. International Food Policy Research
Institute, Research Report No. 110.
Fan, S. Zhang, L., and Zhang, X. (2000), "Growth and Poverty in Rural China: The Role
of Public Investments", Washington DC. International Food Policy Research Institute,
Environment and Production Technology Division, EPTD Discussion paper No. 66.
Feder, G., Just, R., and Zilberman, D. (1985), "Adoption of Agricultural Innovations: A
Survey", Economic Development and Cultural Change, January.
Fei, J.C. and Ranis, G. (1961), "A Theory of Economic Development", American
Economic Review, vol. 51(4), pp. 533-65.
Fields, G. (1989), "Changes in Poverty and Inequality in Developing Countries", World
Bank Research Observer, vol. 4, pp. 167-186.
Fogel, R.W. (1994), "Economic Growth, Population Theory, and Physiology: The
Bearing of Long Term Processes on the Making of Economic Policy", American Economic
Review, vol. 84(3), pp. 369-395.
Food and Agriculture Organization of the United Nations (FAO), (2000), "FAO Farming
Systems Study", draft done for the World Bank Rural Development Strategy Revision, Rome.
Goldin, I. and van der Mensbrugghe (1995), "The Uruguay Round: An Assessment of
Economywide and Agricultural Reforms", paper presented at the World Bank Conference on
"The Uruguay Round and the Developing Economies", Washington DC. January.
Haggblade, S., Hammer, S.J., and Hazell, P.B.R. (1991), "Modeling Agricultural Growth
Multipliers", American Journal of Agricultural Economics", vol. 73(2), pp. 361-374
Haggblade, S., and Hazell, P.B.R. (1989), "Agricultural Technology and Farm-Non-Farm
Growth Linkages", Agricultural Economics, vol. 3(4), pp. 345-364
Haggblade, S., Hazell, P.B.R., and Brown, J. (1989), "Farm-Nonfarm Linkages in Rural
Sub-Saharan Africa", World Development, vol. 17(8), pp. 1173-1202.
Hayami, Y. (2000), "An Ecological and Historical Perspective on Agricultural
Development in Southeast Asia", World Bank, Policy Research Working Paper 2296,
Washington DC. March.
Hayami, Y, and Ruttan, V.W. (1971), Agricultural Development: An International
Perspective, Baltimore, The Johns Hopkins University Press.
Hazell, P.B.R., and Roell, A. (1983). Rural Growth Linkages: Household Expenditure
Patterns in Malaysia and Nigeria. Washington DC, International Food Policy Research Institute,
Research Report 41.
46 The Role ofAgriculture in Economic Development and Poverty Reduction:

Hirschman, A.O. (1958). The Strategy of Economic Development. Yale University

Press, New Haven, Connecticut.
International Fund for Agricultural Development (IFAD) (2001). Rural Poverty Report
2001. Oxford, Oxford University Press.
Jazairy, I., Alamgir, M., and Panuccio, T. (1992). The State of World Rural Poverty: An
Inquiry into its Causes and Consequences. International Fund for Agricultural Development
(IFAD), Rome.
Jimenez, E. (1995), "Human and Physical Infrastructure: Public Investment and Pricing
Policies in Developing Countries", in J. Behrman, and T.N. Srinivasan (editors). Handbook of
Development Economics, vol. IIIB, Amsterdam, Elsevier.
Johnston, B.F. and Mellor, J.W. (1961), "The Role of Agriculture in Economic
Development", American Economic Review, vol. 51(4), pp. 566-93.
Jorgenson, D.G. (1961), "The Development of a Dual Economy", Economic Journal, vol.
71, pp. 309-334.
Kalecki, M. (1960), "Unemployment in Underdeveloped Countries", Indian Journal of
Labor Economics, vol. 3(2), pp. 59-61; reprinted in M. Kalecki, (1976), Essays on Developing
Countries, Sussex, UK, Harvester Press.
Kalecki, M. (1971). Selected Essays on the Dynamics of the Capitalist Economy 1933-
1970, London, Cambridge University Press.
Krishna, R. (1982), "Some Aspects of Agricultural Growth, Price Policy and Equity in
Developing Countries", Food Research Institute Studies, vol. XVIII, No.3: 219-260.
Kuznets, S. (1961), "Long-Term Trends in Capital Formation Proportions", Economic
Development and Cultural Change, July.
Kuznets, S. (1968). Toward a Theory of Economic Growth with Reflections on the
Economic Growth of Nations. New York, Norton.
Lele, U. (1989), "Sources of Growth in East African Agriculture", World Bank Economic
Review, vol. 3(1), pp. 119-144.
Lewis, W. A. (1954), "Economic Development with Unlimited Supplies of Labour",
Manchester School of Economic and Social Studies", vol. 22(2), pp. 139-181.
Lin, J. Y. (1992), "Rural Reforms and Agricultural Growth in China" American
Economic Review, vol. 82(1): 34-51.
Lipton, M. (2000), "Rural poverty reduction: the neglected priority" background paper
for the World Bank World Development Report 2000, Washington DC, processed.
Lipton, M. (1977). Why Poor People Stay Poor: A Study of Urban Bias in World
Development. London, Temple Smith.
Lipton, M. and Ravallion, M. (1995), "Poverty and Policy", chapter 41, in J. Behrman,
and T.N. Srinivasan (eds.), Handbook of Development Economics, vol. IIIB, Amsterdam, North
Holland.
 An Empirical and ConceptualFramework 47

Martin, W., and Mitra D. (1999), "Productivity Growth and Convergence in Agriculture
and Manufacturing", Washington DC, World Bank, Policy Research Working Paper 2171,
August. Forthcoming in Economic Development and Cultural Change.
Martin, W., and Warr, P.G. (1993), "Explaining the Relative Decline of Agriculture: A
Supply-side Analysis for Indonesia", The World Bank Economic Review, vol. 7(3), pp. 381-401.
Mass-Collell, A., and Razin, A. (1973), "A Model of Intersectoral Migration and
Growth", Oxford Economic Papers, vol. 25 (March), pp. 72-79.
Matsuyama, K. (1992), "Agricultural Productivity, Comparative Advantage, and
Economic Growth", Journal of Economic Theory, vol. 58: 317-334.
Mellor, J.W. (2001), "Agricultural Development: So Many Successes, Such Excellent
Results", report prepared for the Food and Agriculture Organization of the United Nations,
January.
Mellor, J.W. (1999), "The Relation between Growth in Agriculture and Poverty
Reduction", paper prepared for USAID/G/EGAD, November.
Mellor, J.W. (1995). Agriculture on the road to Industrialization. Baltimore, The Johns
Hopkins University Press.
Mellor, J.W. (1976). The New Economics of Growth: A Strategy for India and the
Developing World, Ithaca, New York, Cornell University Press.
Mellor, J.W., and Johnston, B.F.(1984), "The World Food Equation: Interrelations
Among Development, Employment, and Food Consumption", Journal of Economic Literature,
vol. XXII (June), pp. 531-574.
Mellor,, J.W., and Lele, U.J. (1973), "Growth Linkages of the New Foodgrain
Technologies", Indian Journal of Agricultural Economics, vol. 28(1), pp. 35-55.
Mundlak, Y., (1999), "The Dynamics of Agriculture", Elmhirst Memorial Lecture, in
G.H. Peters, and J. Von Braun (editors). Food Security, Diversification, and Resource
Management: Refocusing the Role of Agriculture? Proceedings of the Twenty-Third
International Conference of Agricultural Economists. Aldershot, and Brookfield, Ashgate.
Mundlak, Y., Cavallo, D., and Domenech, R. (1989). Agriculture and Economic Growth
in Argentina, 1913-84. Washington DC. International Food Policy Research Institute. Research
Report 76.
Mundlak, Y., Larson, D.F., and Butzer, R. (1997) "The Determinants of Agricultural
Production: A Cross Country Analysis", Washington DC, World Bank, Policy Research
Working Paper 1827, September.
Murgai, R. (1999), "The Green Revolution and the Productivity Paradox: Evidence from
the Indian Punjab", World Bank, Policy Research Working Paper No. 2234, November.
OECD(1995), Technological Change and Structural Adjustment in OECD Agriculture,
Organization for, Economic Co-operation and Development, Paris.
Paxson, C.H., (1992), "Using Weather Variability to Estimate the Response of Savings to
Transitory Income in Thailand", American Economic Review, vol. 82(1), pp. 15-33.
48 The Role ofAgriculture in Economic Development and Poverty Reduction:

Perrin, R.K.(1999), "Intellectual Property Rights and Developing Country Agriculture",

Agricultural Economics, vol. 21 (3), pp. 221-229.
Preobrazhensky, E. (1965). The New Economics. Oxford, Clarendon Press.
Quizon, J. and Binswanger, H. (1989), "What can Agriculture Do for the Poorest Rural
Groups?", in I. Adelman and S. Lane (editors). The Balance between Agriculture and Industry in
Economic Development, vol. 4 (Social Effects), McMillan Press for the International Economic
Association.
Quizon, J. and Binswanger, H. (1986), "Modeling the Impact of Agricultural Growth and
Government Policy on Income Distribution in India", The World Bank Economic Review, vol.
1(1), September, pp. 103-148.
Ravallion, M. (1997), "Can High Inequality Developing Countries Escape Absolute
Poverty?", Economics Letters, vol. 56, pp. 51-57.
Ravallion, M., and Chen, S. (1997), "What can New Survey Data Tell us About Recent
Changes in Distribution and Poverty?", The World Bank Economic Review, vol. 11(2), pp. 357-
82.
Ravallion, M., and Datt, G. (1999), "When is Growth Pro-Poor? Evidence from the
Diverse Experience of India's States", World Bank Policy Research Working Paper 2263,
December.
Ravallion, M., and Datt, G. (1996), "How Important to India's Poor is the Sectoral
Composition of Economic Growth?", World Bank Economic Review, vol. 10(1), pp. 1-25.
Rosenstein-Rodan P.N. (1943), "Problems of Industrialization of Eastern and South-
Eastern Europe", Economic Journal, vol. 53, pp. 202-211.
Rosensweig, M.R., and Wolpin, K. (1993), "Credit Market Constraints, Consumption
Smoothing, and the Accumulation of Durable Production Assets in Low Income Countries:
Investments in Bullocks in India", Journal of Political Economy, vol. 101(2), pp. 223-244.
Sadoulet, E. and deJanvry, A. (1992), "Agricultural Trade Liberalization and Low
Income Countries: A General Equilibrium Multimarket Approach", American Journal of
Agricultural Economics, vol. 74(2), pp. 268-80.
Sah, R.K., and Stiglitz, J.E. (1984), "The Economics of Price Scissors", American
Economic Review, vol. 74, No. 1: 125-138.
Sah, R.K., and Stiglitz, J.E. (1987), "Price Scissors and the Structure of the Economy",
Quarterly Journal of Economics, vol. 112, No. 1: 109-134.
Samuelson, P.A. (1954) "The Pure Theory of Public Expenditures", Review of
Economics and Statistics, vol. 36 (November), pp. 387-89.
Sarris, A.H. (1999), "Structural Adjustment and the Rural Handicap in Developing
Countries", report prepared for the Policy Analysis Division, Food and Agriculture Organization
of the United Nations, Rome.
Sarris, A. H., (2000), "Has World Cereal Market Instability Increased?", Food Policy,
Vol. 25, pp. 337-350.
 An Empirical and ConceptualFramework 49

Sarris, A.H. (1994). Agricultural Taxation under Structural Adjustment. FAO Economic
and Social Development Paper No. 128. Rome.
Sarris, A.H., Doucha, T., and Mathijs, E. (1999) "Agricultural Restructuring in Central
and Eastern Europe", European Review of Agricultural Economics, vol. 26(3): 305-330.
Schiff, M. and Valdes, A. (1991). The Political Economy of Agricultural Price Policy.
Vol. 4. A synthesis of the Economics of Developing Countries. Baltimore. Published for the
World Bank by The Johns Hopkins University Press.
Skott,i P. and Larudee M. (1998), "Uneven Development and the Liberalization of Trade
and Capital Flows: the Case of Mexico", Cambridge Journal of Economics, vol. 22, pp. 277-295.
Solow, R.M. (1056), "A Contribution to the Theory of Economic Growth", Quarterly
Journal of Economics, vol. 70(1) (February), pp. 65-94.
Squire, L., (1993), "Fighting Poverty", American Economic Review", vol. 83(2), pp. 377-
382
Stern, N. (1994) "Growth Theories, Old and New, and the Role of Agriculture in
Economic Development", Mimeographed, report prepared for the Policy Analysis Division of
FAO.
Stiglitz, J.E. (1989), "Economic Organization, Information and Development", in H.
Chenery, and T.N. Srinivasan (editors). Handbook of Development Economics, vol. 1,
Amsterdam, North Holland.
Strauss, J. (1986), "Does Better Nutrition Raise Farm Productivity?", Journal of Political
Economy, vol.l94(2), pp. 297-320.
Swinnen, J. (editor) (1997). Political Economy of Agrarian Reform in Central and
Eastern Europe. Aldershot, Ashgate.
Taylor, L. (1991). Income Distribution, Inflation and Growth. Lectures on Structuralist
Macroeconomic Theory. Cambridge, Massachusetts. The MIT Press.
Thirlwall, A.P. (1986), "A General Model of Growth and Development along Kaldorian
Lines", Oxford Economic Papers, vol. 38: 199-219.
Timmer, C. P. (1997), "How Well do the Poor Connect to the Growth Process?", CAER
Discussion Paper No. 178, Harvard Institute for International Development, Cambridge,
Massachusetts.
Timmer, C. P. (1988), " The Agricultural Transformation", chapter 8, in H. Chenery and
T.N. Srinivasan (eds). Handbook of Development Economics, vol. I. Elsevier Science.
Van de Walle, D. (2000), "Are Returns to Investment Lower for the Poor: Human and
Physical Capital Interactions in Rural Vietnam", World Bank, Policy Research Working Paper
No. 2425, Washington DC., August.
Vogel, S.J. (1994), "Structural Changes in Agriculture: Production Linkages and
Agricultural Demand-Led Industrialization", Oxford Economic Papers, Vol. 46, pp. 136-156.
World Bank (2000). World Development Report 2000/2001: Attacking Poverty. Oxford
University Press for the World Bank.
50 The Role ofAgriculture in Economic Development and Poverty Reduction:

Wright, B. (1998), "Public Germplasm Development at a Crossroads: Biotechnology and

Intellectual Property", California Agriculture, vol. 56(2), pp. 8-13.
Appendix. A Two Sector Model of Public Policy
Towards Agriculture in the Context of Growth and
Poverty Reduction
1. Theoretical structure
The model is a modification and adaptation to the context of this study of the model analyzed by
Sarris (1999), which in turn was based on the model analysed by Mas-Collell and Razin (1973),
using ideas from Taylor (1991, chapter 9). Consider an economy that consists of two sectors,
agriculture, denoted throughout by the subscript "a" and non-agriculture (industry for short),
denoted throughout by a subscript "n". The economy will be considered as small and open for
simplicity. Trade can be introduced in many ways. For instance we could assume that both
goods are traded, and the economy is small in international trade. Such an assumption would
impose the domestic terms of trade (barring tariffs) from abroad, and this is unrealistic, as all
economies have a substantial non-traded sector. Another way would be to assume that only one
of the two goods is traded and exported, and in return the economy imports an imperfect
substitute in consumption with the non-traded good. The approach followed is similar to this
latter one, except it makes the assumption that the imported good is incorporated into the non-
traded one before it is consumed domestically. This is made purely for analytical simplicity. The
economy as delfined is meant to emulate an archetype African economy in the sense, for instance,
illustrated by Sadoulet and deJanvry (1992).
Suppose that the economy is a small open economy that exports the agricultural product and
imports a non-competitive intermediate product, which is used in the production of the non-
agricultural sector. Suppose that the international prices of the exported agricultural product and
the imported intermediate are given and by choice of units are both equal to one. Define as
numeraire the non-agricultural price and set it equal to one. Define by e the nominal exchange
rate, which given that the non-agricultural sector is non-traded, and is assumed to be the
numeraire, is equal to the domestic relative price of the agricultural product faced by
consumers .
Let the production functions of the two sectors be given by the following Cobb-Douglas
formulas.
Ya =ALaKa (1)

Y- BLnMYK,-fl` (2)

where Yi (i=a,n) denotes the gross output of the two sectors, Li and Ki (i=a,n), denote the labor
and (specific) capital inputs into the two sectors, M denotes the imports of the non-competitive
intermediate product, a, ,Bdenote the labor shares in the agricultural and non-agricultural sectors
respectively, y is the share of the imported intermediate in the production of n, and A,B, denote
the TFPs of the two sectors. It will be assumed that the total labor force is a constant fraction of

4 An alternative assumption, and analytically identical, would be to normalise on the agricultural good, in which
case the domestic relative price of the agricultural good would be the inverse of the price of the non-traded good.
52 The Role ofAgriculture in Economic Development and Poverty Reduction:

the total population, and hence per-capita and per-worker magnitudes are simply related by a
multiplicative constant and will be treated equivalently. The per capita outputs of the two sectors
denoted by y1 (i=a,n) are given by the following formulas, which can be easily derived from the
level form of the Cobb-Douglas production functions.

Y = A(1 -,u)kk;a (3)

= (4)

where p is the share of total capital stock that is employed in the non-agricultural sector, k is the
economy wide capital-labour ratio, ki (i=a,n) are the sectoral capital labor ratios, and m is per
capita imports. In other words p = Ku/K, k=K/L, m=M/L, and ki = Ki /Li (i=a,n). The variables p
and k will be the state variables of the dynamic model. If the agricultural sector is more labor
intensive than the industrial sector, then it should hold that a> P.
Consider the incorporation of public infrastructure in the above simple model. The major issue
from an economy-wide perspective is how does infrastructure, or other public expenditures
affect the shape of the production functions. The sectoral production functions estimated, for
instance, by Antle (1983), Mundlak, Larson, and Butzer (1997), and others use relatively
standard functional forms, such as Cobb-Douglas, and various proxy measures for physical
infrastructure or human capital, such as rural road density, irrigation density, adult literacy rates,
etc., in order to estimate relevant elasticities. In models of endogenous growth, such as the ones
of Barro (1990), simple forms are also used for the influence of infrastructure on the production
functions.
There is no doubt that the total factor productivity of agriculture, as well as non-agriculture (the
"constants" A and B in the model analyzed here) are influenced by variables such as research,
extension, infrastructure, and human capital. All these variables involve an externality in the
sense that the overall economic returns from expenditures in such variables are often larger than
private returns. In recognition of this fact governments devote considerable expenditures to all
the above productivity improving activities. Studies have found that there are substantial
elasticities of TFP with respect to the above mentioned factors, and also that the same factors
also contribute to poverty alleviation (Evenson, Pray and Rosengrant 1999, Fan, Hazell and
Thorat, 1999, Fan, Zhang and Zhang, 2000). In the sequel we shall consider a simple way of
incorporating public spending in the two-sector model in order to obtain some insights.
It will be assumed that public infrastructure or other public expenditures can affect the
production of individual firms in each sector by changing the value of the TFP constant A and B.
In particular assume that the two constants take the following forms.

A= (Ga )-ga6o, (5)

La

B=(G)7-gl7 (6)

Where Gi (i=a,n) denotes the total amount of sector specific public capital, gi (i=a,n) denotes the
public capital in each sector per unit of labor employed in each sector, and the parameters s and
il are the relevant elasticities. The public capital can comprise both physical infrastructure, such
as roads, dams, electricity grids, etc, research, extension, as well as human capital such as health,
 An Empiricaland ConceptualFramework 53

literacy, etc. iIt is assumed that the production functions are affected by the per capita availability
of public capital in each sector. In other specifications, like the one in Barro and Sala-i-Martin
(1995) (chapter 4), the total public capital is considered rather than the per-capita magnitudes. It
is not clear which specification is better, but it does not matter that much. The key behavioral
assumption in the model is that the various firms and productive entities in the two sectors
consider the values of A and B as exogenous to their production decisions. Hence infrastructure
and other public expenditures are assumed to have external effects.
It is also assumed that the public good is non-rival and non-excludable. In other words each firm
within a sector makes use of all of the public capital available to the sector, and that use of the
public good by one firm does not diminish the quantity of the public good available for the other
firms within the sector. This is the standard approach to public goods, due to Samuelson (1954).
Notice, however, that we are assuming some kind of congestion of public capital across sectors.
This is represented by the fact that the A and B parameters are functions of per capita availability
of public capital to each sector. If some policy or other effect induces intersectoral labor
movement, then this, for fixed public capital available to the two sectors, induces changes in the
private production of each firm, because of congestion effects. So, for instance, a labor
reallocation towards non-agriculture, induces a decline in B and an increase in A, ceteris paribus.
Full employment of capital and labor is assumed. This is not a major assumption as the model is
designed to trace dynamic growth relations. However, it excludes "unlimited supplies of labor"
or substantial disguised unemployment. Unemployment of capital and/or labor must be specified
as due to some kind of rigidities, that are assumed away here for simplicity. For instance
unemployment could be due to fixed urban wages at levels above the rural wages, leading to
rural urban migration a-la Harris-Todaro (HT), but if such a specification is made, then some
assumption must be made about how the fixed urban wages adjust from year to year, in order to
trace the model dynamically. While a variety of specifications is possible, they do not influence
the major results of the dynamic model exhibited below. The assumption of full employment
also makes the! "growth linkage" multipliers from agricultural development small. This implies
that any growth enhancing results of agricultural development can be regarded as underestimates
of the true results in the presence of underemployed labor resources.
Full employment of factors implies the following relation between the variables defined above.
1 1 l-,u a
k kk (7)
k -k 0 'a k,

In order to finance the public expenditures, it will be assumed that agriculture is taxed at the
uniform rate ta while non-agriculture (industry) is taxed at the uniform rate tn. The tax rates can
be visualized as indirect tax rates on the final output of each sector's product. While it is doubtful
if all agriculture or all non-agriculture can be taxed at uniform rates as assumed here, the
assumption of uniform tax rates is convenient for the analysis. We do not discuss the means by
which taxation on each sector is imposed. The tax rates do not have to be positive, and negative
tax rates imply a subsidy for the relevant sector. The tax revenue of the government are assumed
to be utilized in order to provide public capital to the two sectors. A balanced budget in every
period is assumed for simplicity.
Under the assumptions concerning the international prices and the normalization mentioned
above, and given that the tax rates are indirect and uniform for all output, e represents the
54 The Role ofAgriculture in Economic Development and Poverty Reduction:

domestic relative consumer price of the agricultural good, while the relative producer price is
e( -ta)/(l-tn). The latter, may then be regarded as the real exchange rate.
The marginal products of labor and the returns to capital in the two sectors, denoted by wi and ri
(i=a,n) respectively are as follows, where we incorporate the influence of taxes.
wa = A(1-ta)eak'-a (8)

wn = B(1- tn ),Bu Yk-Yk-f6mr (9)

ra =A(1-ta)(1- a)ek1a- (10)

=B(l - t,)(l -,l8 -y)pyk-rk,-6mr (11)

Note that I13-y is the share of capital in the production of the n good.
The per-capita demand for the intermediate is given by the following expression, that derives
from the profit maximizing condition of the typical firm in the n sector.
em = y(I - tn )y, = yB(I - tn)p'Yk-Ykn-8mT (12)
If we denote by g the economy-wide ratio of public capital to total labor, and if it is assumed that
public capital in the two sectors is made up of the same good, which will be assumed to be the n-
sector good, then we can write the per capita sectoral public capital contributions as follows.
G
= pg (13)

g = L (1-p)g (14)
n (1-_2 ) (1-2l )

where p denotes the share of total public capital that is allocated to agriculture, and A is the share
of labor employed in agriculture.
As indicated above, all domestic taxes (namely all public revenues) are utilized for public
expenditures on agriculture and non-agriculture. Hence we have the following relation in level
form.
PUB = t0 eY, +t,,1Y (15)

where PUB is total domestic public spending on both sectors. If we assume that the depreciation
rate of infrastructure capital is equal to o, the change per capita in aggregate public sector capital
stock is as follows,

AG + PUB = _dg + t,eya + tnYn (16)

L -S+L 6+ 0 y+~~(6

where AG denotes the change in aggregate public sector capital stock in a year.
The per capita GDP, denoted by y, which is equal to domestic value added in domestic prices, is
given by the following expression.
 An Empiricaland ConceptualFramework 55

y=ey0 + (l-Y)Yn (17)

Private per capita income of the economy, denoted by yp is the total value added (namely GDP)
minus taxes, as follows:
yp =e(1-ta)ya +(1 -y)(1 t)y- (18)

If s denotes the average private saving (and investment) rate, and 0 denotes the share of private
expenditure spent on food, the supply demand balance for the product of the n-sector is as
follows, under the assumption that all investment is made up of n-sector products.
syp + (1- 0)( - s)yp + t,,ey, + nY =Yn (I19)
By substituting (18) in (19) we can derive an expression for e in terms of the per capita products
of the two sectors.
e = QyYan (20)
Ry0.
where the parameters Q and R are defined for notational convenience, and are given by the
following expressions:
Q =~
Q(tn) (-tn)[O(
(I - s) + Y(l -R(l -s))] (2 1a)
R = R(ta) = (1-9(1- s)) + t0 0(l- s) (21b)

In the short run it will be assumed that the sector specific capital stocks are immobile across
sectors, while!labor is freely mobile. By substituting (20) in the labor market equilibrium
condition, namely the relation that equates the marginal products of labor in the two sectors 5 , as
given in equation (8) and (9), we obtain a linear relation between the capital-labor ratios of the
two sectors, as follows.

)1kt =-(1-tn ) )
(-')k (22)

Combining (22) with the full employment condition (7), we obtain expressions for the sectoral
capital-labor ratios, in terms of the economy-wide state variables, as follows:

ka,=(I Ak (23)

kn = (lk A)(24)

where X,the share of labor in agriculture, is given by the following expression.

(1-tI-aU+,laR
A2(ta)~ (25)

5An alternative to this labour market equilibrium condition would be to equate the average product of labour in
agriculture with the marginal product of labour in non-agriculture. This would correspond to an agrarian structure of
small farmers, and given our algebraic assumptions, it would not change the subsequent analysis.
56 The Role ofAgriculture in Economic Development and Poverty Reduction:

In (25) the notationally convenient parameter U is given by the following expression

U =9(l- s) + y( - 0(1- s)) (26)

It can be readily derived from (25) by simple algebra that dk/dta<O. In other words larger
taxation of the agricultural sector induces labor reallocation towards non-agriculture, and hence a
lower share of total labor employed in agriculture, as expected. Also it can be readily shown that
dk/dO>O, and dk/dy>O. The first of these relations implies that if the share of the consumers'
expenditure that is devoted to agricultural products is larger, then so is the share of labor
employed in agriculture. The second relation implies that if the share of imported intermediates
in the production of non-agricultural non-tradables is large, then the share of labor employed in
agriculture is large, and this makes sense as in such a case the country needs to produce and
export larger quantities of the agricultural product to finance the imports.
By utilizing (12) and (19), the per-capita import of the non-competitive intermediate can be
written as a share of the agricultural product as follows:
yR
m=Uya -Y0 (27)

where the last equality just defines the parameter F. It can be easily computed that dF/dy>O,
implying that a large degree of intermediate import dependence by the non-agricultural sector
implies that a larger share of the agricultural product must be exported to finance the imports. It
can also be verified that the above expression for m corresponds to the difference between the
per capita production of the agricultural product, and the per capita quantity demanded
domestically for consumption of food. This implies that dr/dO<O, as can also be directly verified
by differentiating the expression for F in (27).
Per capita total private investment in this economy is given by syp. This can be written, by
utilizing the expressions for yp and e in equations (17) and (19) above, as a share Xof yn,, where
the share X is given by the following expression.
A)
SO = s(1-tn)(l-Ya (28)

By substituting the expressions (23) and (24) in the production functions (3) and (4), and
utilizing (27), we can write the per capita products of the two sectors in terms of the state
variables as follows:
Ya = A(1 -,),-a 2Gk-a (29)
Yn = BAYrFY2a (1- ),? PjV-4?Y (1- ,)Y(I-a) k--a (30)

There are three dynamic equations of the system. The first describes the evolution of private
capital, and is given by the following equation:

k_k== _ +5 )(31)
-(n
k k
where 6 is the rate of depreciation of private capital, which will be assumed for simplicity to be
the same as the rate of depreciation of public capital, and n is the (exogenous) rate of population
 An Empiricaland ConceptualFramework 57

(as well as labor force) growth. We could assume different rates of depreciation of public and
private capital, but this would only complicate the growth expressions without adding any new
insights.
By utilizing (30), as well as (5), (6), (13), and (14) we can write the private sector per-capita
capital stock growth rate as follows.

k = ,fYpYC (1 - p) 1 p 1 -l-aY (1- ,l)y(I a) k -aagq`6 (1-_ A),-q ,y(a-e) - (n + g )

k
(32)
The second dynamic equation describes the evolution of the share of capital employed in the n-
sector, and will be assumed to be a function of the relative rates of return to capital, as for
instance specified by Mass-Collell and Razin (1973).

,uI ra

The final dynamic equation concerns the evolution of the public sector capital stock. By the
equation for public capital (15), under the additional assumption that the difference equation can
approximate the instantaneous change of public sector capital, the growth rate of public sector
capital is given by the following equation.
-g-G-L= t°eYa+tnYn ( )(34)
g g
By utilizing (19), the above growth rate can be written as follows
g = (taQ ntR)Yn (n ) (35)
g R g
The steady state of this system will involve the same returns to private capital in the two sectors,
so that the private capital stocks in the two sectors grow at the same rates. By equating the two
sectoral rates of return to capital in (10) and (11), and utilizing (12) and (20) we can derive the
steady state value of the share of capital employed in the n-sector.
*~~~((-,l3 - y)R (6
(1-a)(l-tt)U + (I-,8-y)R (36)
Note that this value is a function of only the agricultural tax rate and not the non-agricultural tax
rate. It can also be readily seen, through straightforward algebra, that dpt*/dta >0. In other words
a larger agricultural tax rate implies a larger share of capital employed in the non-agricultural
sector in the steady state, a reasonable conclusion to expect. It can also be shown that dji*/dO<0.
In other words the larger the share of food in total expenditure, the smaller the share of capital
employed in the non-agricultural sector.
In this model there is no steady state where the per capita magnitudes are constant. This can be
seen by noting that if the right hand sides of (35) and (31) are set equal to zero, then they imply
that the ratio g/k is constant. However, by considering the right hand side of (32) it can be seen
that for this value of the ratio g/k as well as the value for p* implied by (36), the right hand side
of (32) is not equal to zero, which is a contradiction. Hence the steady state, under the
58 The Role ofAgriculture in Economic Development and Poverty Reduction:

assumption of a constant food budget share and constant savings rate, is characterized be equal
growth rates of all the relevant per-capita capital stocks, both private as well as public, and by
consequence per-capita GDP. This implies that in steady state growth, the ratios of any two
magnitudes, such as the per capita production in the two sectors, or the per capita private and
public capital stocks will be constant. By reviewing the two dynamic equations (32) and (35) it
can be readily seen that there cannot be a steady state in the above sense unless the following
condition is satisfied among the various parameters of the model.
q+ye=/3+ay or q-/=7y(a-e) (37)
This condition is a generalization of the condition that the elasticity of TFP with respect to public
capital is equal to the aggregate labor share that was proposed and utilized by Barro (1990) to
ensure steady state growth in the aggregate economy. In the sequel this condition will be
assumed, so as to assure steady state growth.
Notice that in this model the share of agriculture in GDP, which we shall denote by Sa is a
function of both the tax rates, and is independent of other state variables if 0 is constant.

Sa = eya - Q (37)
eya + Yn Q+ (1- y)R
It can be readily seen that this share is a negatively sloping function of the agricultural tax rate
and a positively sloping function of the non-agricultural tax rate as is expected. Namely dSa /dta
<0 and dSa /dtn >0. It can also be shown that dSa /dO>O, implying that a large food budget share
implies a larger agricultural sector ceteris paribus, which is to be expected.
Let us now assume that in the steady state, the aggregate per capita public and private capital
stocks both grow at the same rate c. By using (35) we can obtain that the ratio of per capita
output of the n-sector and the per capita public capital is constant and equal to the following.
yn = (n+S+4')R (38)
g taQ+tnR

Similarly by using equation (31) we can derive that the ratio of per capita output of the n-sector
and the per capita private capital is constant and equal to the following.
yn n+6+; (39)
k X
Notice that as the per capita GDP can be written, by utilizing (20), as a fixed multiple of per
capita output of the n-sector, the ratio of per capita GDP y to k is also a constant. This implies
that y and k will grow at the same rate.
By multiplying the expression in (39) with the inverse of the expression in (38) we eliminate the
variable yn , as well as the growth rate- variable 4, and we obtain an expression for the ratio of
the per capita public and private capital stocks.
g taQ+tR (40)
k XR
 An Empirical and ConceptualFramework 59

By substituting this expression in (40) in the growth rate equation for private capital (32), using
(37), and using the steady state value of pt* from (36), we obtain an expression for the steady
state growth rate of k, and hence the growth rates of g and y as well.

rpyr (1- py(P*)I-6-r (1-_

r= *)y(-a)(t,Q + tR) +ar (I _ A)'-, r(a-C)_ (n + 5)
XR
(41)
As in the steady state the share of capital employed in non-agriculture, namely the variable pt*, is
constant and a function of the agricultural tax rate, the overall growth rate 4 is a function of the
two tax rates, as well as the variable p, namely the share of public capital employed in non-
agriculture. Also note that apart from the terms that involve p directly, there are no other terms
that are functions of p. We can thus maximize the growth rate as a function of p, in order to find
the share of public capital employed in agriculture that will maximize the steady state growth
rate, irrespective of the tax rates. The maximization is straightforward and yields the following
"optimal" share of public sector capital that must be devoted to agriculture.

p*= - (42)
+ya 77+Y 7 +y

It can be readily seen from (42) that if there is no trade, in the sense that there is no agricultural
exports, and hence no intermediate good required for non-agricultural production, then the share
of public capital that needs to be devoted to agriculture, to maximize the steady state growth rate
is zero. In other words for a closed economy, to maximize the steady state growth rate, under a
fixed food budget share 0, there is no need to devote any public resources to agriculture. The
reason is basically the agricultural treadmill. Any productivity gains in agriculture are reflected
in lower relative prices of the agricultural good, as can be seen from (20). When there is trade, by
contrast, higher agricultural productivity is reflected in lower costs of intermediate imports, and
this implies that improvements in agricultural productivity will raise growth. These results are
opposite to those obtained by Matsuyama (1992).
It can be readily seen from (42) that the share of public capital that should be devoted to
agriculture for maximum growth is a positive function of y, the share of imports in the total
production of the non-agricultural non-traded good. Finally the share of public capital that should
be devoted to agriculture for maximum growth is seen to be a positive function of the elasticity
of agricultural TFP with respect to public capital, namely the parameter s, while it is a negative
function of the elasticity of nonagricultural TFP with respect to public capital. This brings to the
fore the importance of these elasticities, that have not been estimated very widely in an
aggregative form for agriculture, and even less for non-agriculture.
Notice that the non-agricultural tax rate enters the expression (41) in only two factors, namely X
and the parenthesis that involves the variables Q and R. It is easy then to maximize the growth
rate with respect to the non-agricultural tax rate while keeping the agricultural tax rate fixed (the
second order condition can be checked to hold as well). The resulting expression is the
following.

t R(/3+ay)-taU
60 The Role ofAgriculture in Economic Development and Poverty Reduction:

It can be easily seen that the tax rate in (43) is smaller than one, and can be negative, namely it
can be a subsidy. It can also be computed by straightforward algebra that the following relation
holds.

dtn < 0 (44)

dta
Thus it appears that for optimal growth there is a negative relation between the two sectoral tax
rates. A high agricultural tax rate implies a low non-agricultural tax rate and vice-versa. Note
that a zero agricultural tax rate implies a positive non-agricultural tax rate. This is necessary as
by assumption the existence of public capital is essential for production in both sectors, and
hence at least one of the sectors must be taxed to generate revenues to maintain public capital.
The relationship (43) describes the optimal trade-off.
If the expression for the optimal non-agricultural tax rate in (43), and the expression for the
optimum share of public capital going to agriculture (42) are substituted in the growth equation
(41), then we obtain an expression for the aggregate growth rate that depends only on the
agricultural tax rate. From that expression one can obtain the agricultural tax rate that maximizes
the overall growth rate. The expression is too complicated for analytical manipulation. Hence
one must resolve to numerical computations, which are not difficult.
Notice that in the model we have assumed that the parameters 0 and s, namely the economy-wide
"food" budget share, and the average saving rate are constant fractions of private incomes. The
first point here is that such constant shares do not require constant food budget and savings
shares from all the various sources of income. For instance, if it is assumed that there are three
different sources of income, namely labor income, capital income from agriculture, and capital
income from non-agriculture, then if one postulates different constant food budget shares and
saving shares for each one of these types of income, the net result is a model that is formally
equivalent to the simpler one analyzed here, except that the various expressions have more
parameters. The basic point is that one can still write an equation for the supply demand balance
for the non-agricultural product (namely equation (18)), that can in turn be solved analytically
for the domestic terms of trade as in (19). Under the production assumptions of the model this
can always be done as long as we assume that the budget and saving shares of the different types
of income sources are constant. The same results if we assumed several different income groups,
each having a certain share of labour and capital incomes in the two sectors. Hence the simpler
model was preferred.
The more serious assumption concerns exactly the one concerning constant food budget and
saving shares from different income sources. Such an assumption assumes away one of the basic
"stylized facts" about development, namely that the income elasticity of demand for food is
smaller than one, known as Engel's law. The results of Matsuyama (1992), as well as those of
Skott and Larudee (1998) and Taylor (1991) depend critically on this assumption. If we assume a
linear expenditure system instead of the unitary income elasticity demand system we have
assumed, then we could not solve explicitly for the domestic terms of trade as in (20), and we
could thus not obtain closed form expressions for the sectoral capital labor ratios and the
attendant growth rates.
We could rescue the analytical structure of the present model by making the assumption that the
value of the food budget share 0 is a decreasing function of the aggregate capital labor ratio k,
An EmpiricalandConceptualFramework 61

and correspondingly that the saving rate s is an increasing function of k, without otherwise
changing the structure of the production and other relations. Given that we have presented the
signs of the derivatives of the various key variables, such as the labor share in agriculture X, the
share of agriculture that is exported F, the share of capital employed in the n-sector [t*, and the
share of agriculture in GDP Sa , with respect to 0, we can think of the changes implied by an
assumption of not fixed 0.
The first point that we can make in this context is that now we cannot talk about the same steady
state growth rates for all the relevant state variables, such as k and g. However, we can still talk
about a growth path where the returns to capital are equalized across sectors. The solution to ,u*
in (36) can still then obtain, with the qualification that as 0 declines (and concurrently s
increases) the; value of p* increases, as might be expected. So in this case even in the "long run"
the share of capital employed in non-agriculture will be monotonically increasing, as long as the
food budget share keeps declining.
The second point is that the parameter indicating the share of public capital employed in
agriculture, enters the growth rate equation for private capital, as well as the growth rate equation
for public capital as can be seen from (35) after substitution of yn in a fashion that is not
dependent on 0. Hence these instantaneous growth rates can be maximized with respect to p
irrespective of the value of the food budget share. In other words the allocation rule for public
capital (42) does not depend on the food budget share.
Third, if the food budget and the savings parameters 0 and s are functions of k (the first a
decreasing and the second an increasing such function), denote by Ck and tg the instantaneous
growth rates of k and g respectively. Then the following can be derived by manipulation of the
growth equations for k and g, namely (31) and (35).

If g .t 0Q+ then -k >,4g (45)

k X

Ifgk
<taQ+R'' then k<<;g (46)
k X
The right hand' side of the inequalities in the left sides of (45) and (46) are functions of the
parameters 0 and s, which are in turn functions of k, as assumed. The above equations indicate
that the growth paths of g and k will tend towards a ratio that is represented by the fraction in the
right hand side of the two inequalities above. Hence the growth rate corresponding to this
common ratio can be considered as the "underlying tendency" at any point in time, which in turn
will be changing over time, and it is in this sense that we can interpret the steady state values of
the growth rates in the context of Engel's law. Note that the two tax rates affect this steady state
growth path.
Consider now poverty and distributional issues. Given the aggregated nature of our model it is
not easy to specify in a general sense the structure of incomes of the poor. However, we can
assume, as seems to be the case for most developing countries that the poor draw most of their
income either from agriculture, or from unskilled labor. We can also assume that in terms of
consumption the most important commodity they consume is food. This implies that we can
write the real income of a typical poor person as follows.
62 The Role ofAgriculture in Economic Development and Poverty Reduction:

Ypoor =(Q- t)y

ra)a 1 wa (47)
= #+
ypOOt (I -
e A.e
where the parameter (pdenotes the share of income derived from agricultural production, the rest
been derived from wage labor. Notice that the first term in the right hand side of (47) is
agricultural production per agricultural worker net of taxes, and not per unit of labor in the
economy, and this is why the per capita agricultural product is divided by the share of labor in
agriculture.
By using the expressions for the per capita agricultural product and the common wage rate in (3)
and (8) we can write the real income of the poor as follows.

Ypoor = A(1-ta)kI a k- Xk + (1-0)a] = A(1 -t 0 )[(l )k] + (1-)a] (48)

e k0 2 L A2I
The last equation above is the result of applying the expression for the capital labor ratio in
agriculture from (23). Under the assumptions that led to the steady state growth equations, it is
apparent that faster increases in agricultural productivity, as well as faster increases in the
economy-wide capital labor ratio are beneficial for the typical poor. It is clear that improvements
in agricultural productivity are necessary to have faster growth of the income of the poor, or
equivalently to reduce the number of the absolutely poor, as average incomes of the poor are
raised. This is the result of the inter-sectoral labor market equilibrium. In other words any
agricultural TFP improvements will increase the incomes of agricultural producers, and
consequently, through the demand effect, increase demand for n sector goods, and hence increase
the national demand for labor, and consequently real wages.
The expression in (48) can also be written, after substituting the expressions for A as

follows. Ypoor = p 6 (1-t )2a- -1(1 - ,) 1

ak'-age [( + (1- ()at] (48)
e
This shows clearly that an increased share of public resources devoted to agriculture is beneficial
for the poor, in addition to high levels of k and g. In steady state, the growth rate of the real
incomes of the poor will be a constant factor equal to (1-a+s) of the common growth rate of k
and g. If the elasticity of agricultural TFP with respect to public sector capital is larger than the
labor share in agriculture, then this factor will be larger than one, signifying that the real incomes
of the poor will be growing faster than those of the economy as a whole, and this will improve
income distribution. Otherwise the opposite will be the case.

2. Empirical Simulations

Consider numerical simulations of the above model. For all experiments a population growth rate
of 0.02 (namely two percent annually) and an annual depreciation rate of 0.05 common to both
the private and public capital are assumed. As the population growth rate and the depreciation
rate affect the steady state growth rate in an additive fashion, implications for the growth rates of
different values for these parameters can be easily derived, by just adding or subtracting the
An Empiricaland ConceptualFramework 63

respective population and depreciation rates from those assumed above, and modifying the
overall growth rate above by the total difference.
Table I presents the growth rates under different agricultural tax rates for a set of parameters that
might characterize an agrarian developing economy under alternative values of the trade
dependence parameter y. The assumed value of the aggregate saving rate is 0.2. The values for
the TFP elasticities assumed are such as to render them roughly equal. The results suggest that at
low degrees of trade openness, namely low values of 7, the agricultural tax rate should be
considerably high, and correspondingly the nonagricultural tax rate low, in order to achieve high
overall growth rates. This is consistent with the view that at low levels of trade exposure and
development,' significant domestic taxation of agriculture is optimal for growth. Notice that at
low levels of trade exposure, the optimum share of public capital devoted to agriculture is low,
considerably smaller than the share of labor in agriculture, or the share of agriculture in GDP.
The picture, however, changes considerably for high degrees of trade exposure. For large values
of the parameter y it can be seen that the optimum rate of agricultural taxation is smaller than the
non-agricultural taxation, and low, but the share of public capital going to agriculture is still
smaller than the share of agriculture in GDP or the share of labor in agriculture. Thus, the first
conclusion is that the rate of agricultural taxation in order to finance infrastructure and other TFP
enhancing public expenditures is a decreasing function of the degree of trade dependence of the
non-agricultural sector.
Table 2 exhibits the growth rates, assuming a fixed agricultural tax rate, but under different
assumptions concerning openness of the nonagricultural sector, and various assumptions
concerning the elasticities of agricultural TFP with respect to public capital. The aggregate
saving rate is kept at 0.2. The first interesting thing to note is that at any given degree of
openness, higher values of the agricultural TFP elasticity imply lower rates of overall growth.
However, they imply higher rates of growth of the real incomes of the poor. Hence it appears
that variations of agricultural TFP elasticity imply a tradeoff between higher overall growth rates
and higher rates of real income growth of the poor. This was not observed, for instance in the
figures of table 1, where both the overall income growth as well as the growth rates of the real
incomes of the poor moved together as the agricultural tax rate varied. Notice also that moving
down a column, toward economies with larger degrees of market openness, it appears that they
dictate larger shares of public capital devoted to agriculture, for a given agricultural TFP
elasticity, and imply larger rates of both overall growth, as well as real income growth of the
poor. Notice that in the bottom panel of figures the share of public capital devoted to agriculture
is not only larger than 50 percent, but is also almost as large as the (large) share of agriculture in
GDP.
Table 3 presents alternative growth rates for different values of the food budget share parameter,
as a function of the agricultural tax rates. It can be seen that while higher agricultural tax rates
imply larger growth rates and larger growth rates for the real incomes of the poor, at all levels of
the food budget shares, the growth rates are not very sensitive to the different tax rates at
different levels of the food budget shares. Looking down any column, it can also be seen that the
growth rates for given degree of agricultural taxation are not very sensitive to the different food
budget shares.
Table 4 presents the growth rates for three different values of the saving rates, under a variety of
food budget shares, and fixed values of the other parameters. While it can be seen that the
growth rates are not sensitive to the different levels of the food budget share for given saving
64 The Role ofAgriculture in Economic Development and Poverty Reduction:

rate, they are quite sensitive to different saving rates, as one might expect, with higher saving
rates implying substantially larger growth rates. It can be seen also that for lower food budget
shares, given a saving rate, the steady state growth rates are higher, and the growth rates of the
real incomes of the poor are also higher.
The final table exhibited, table 5, presents the growth rates for different values of the labor
intensity parameters in agriculture and non-agriculture, namely the parameters a and P. It can be
seen that low labor intensities in both sectors imply low rates of growth, but high shares of public
capital in agriculture. On the contrary, high labor intensities in both sectors imply high growth
rates in steady state, and low shares of public sector capital in agriculture. Interestingly, also note
that going from low to high degrees of labor intensity of agriculture, the implied steady state rate
of growth rate of the economy increases, but the real income of the poor first declines and then
increases.
The conclusion from the above simulations is then that the parameters that matter a lot in
determining the usefulness of agricultural development as an engine of growth as well as poverty
reduction are the degrees of labor intensities of the two sectors, the degree of trade openness, the
elasticity of agricultural TFP with respect to public capital, and the rate of domestic savings.
The conclusion of this section is that agricultural taxation certainly helps the overall rate of
growth in small open economies, with an agricultural sector that produces the bulk of
exportables. The appropriate non-agricultural tax rate is inversely proportional to the agricultural
tax rate, and is negative, namely turns into a subsidy, only for large values of the rate of
agricultural taxation. The share of public sector capital that must be devoted to agriculture is not
related to the share of agriculture in GDP, or the share of labor employed in agriculture, but is
rather a function of the labor shares of the two sectors, the import intensity of non-agriculture,
and the elasticity of agricultural TFP with respect to public capital.
 An Empirical"andConceptualFramework 65

Appendix Tables
Table 1. Steady state growth rates of various economies under different agricultural tax rates
Economicparameters Alpha= 0.60 Beta= 0.30 Gamma= 0.10
Theta= 0.50 Epsilon= 0.40 Eta= 0.32
Agricultural tax rates
0.1 0.2 0.3 0.4 0.5
Variables Symbols
Nonag. tax rate tn 0.310 0.260 0.208 0.156 0.102
Common growth rate % percent 5.51% 5.77% 6.03% 6.30% 6.60%
Gr. Rate of realincomes of poor percent 4.41% 4.62% 4.82% 5.04% 5.28%
Share of cap in h-sector mu 0.699 0.735 0.770 0.805 0.839
Shareofpubcapinagr. rho 0.111 0.111 0.111 0.111 0.111
Share of lab. in agr. lambda 0.564 0.520 0.472 0.421 0.365
Share of agr. In GDP Sa 0.355 0.357 0.360 0.362 0.365
Economic parameters Alpha= 0.60 Beta= 0.30 Gamma= 0.30
Theta= 0.50 Epsilon= 0.40 Eta= 0.36
Agicultural tax rates
0.1 0.2 0.3 0.4 0.5
Variables Symbols
Nonag. tax rate tn 0.428 0.373 0.314 0.252 0.184
Common growth rate % percent 4.28% 4.46% 4.61% 4.72% 4.81%
Gr. Rate of real incomes of poor percent 3.43% 3.57% 3.69% 3.78% 3.85%
Share of cap in n-sector mu 0.551 0.594 0.639 0.686 0.734
Share of pub cap'in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.620 0.577 0.530 0.478 0.420
Share of agr. In GDP Sa 0.425 0.433 0.441 0.449 0.458
Economic parameters Alpha= 0.60 Beta= 0.30 Gamma= 0.50
Theta= 0.50 Epsilon= 0.40 Eta= 0.40
Agricultural tax rates
0.1 0.2 0.3 0.4 0.5
Variables Symbols
Nonag. tax rate tn 0.551 0.496 0.435 0.367 0.289
Common growth rate % percent 5.66% 5.77% 5.81% 5.78% 5.68%
Gr. Rate of real incomes of poor percent 4.53% 4.61% 4.65% 4.62% 4.54%
Share of cap in n-sector mu 0.337 0.378 0.424 0.475 0.533
Share of pub cap in agr. rho 0.333 0.333 0.333 0.333 0.333
Share of lab. in agr. lambda 0.663 0.622 0.576 0.525 0.467
Share of agr. In GDP Sa 0.496 0.509 0.523 0.538 0.554
Economic parameters Alpha= 0.60 Beta= 0.30 Gamma= 0.60
Theta= 0.50 Epsilon= 0.40 Eta= 0.42
Agricultural tax rates
0.1 0.2 0.3 0.4 0.5
Variables Symbols
Nonag. tax rate tn 0.614 0.562 0.502 0.433 0.352
Common growth rate % percent 7.88% 7.94% 7.92% 7.80% 7.60%
Gr. Rate of real incomes of poor percent 6.30% 6.35% 6.33% 6.24% 6.08%
Share of cap in n-sector mu 0.190 0.219 0.253 0.294 0.345
Share of pub cap in agr. rho 0.364 0.364 0.364 0.364 0.364
Share of lab. in agr. lambda 0.681 0.641 0.596 0.545 0.487
Share of agr. In GDP Sa 0.534 0.550 0.568 0.586 0.606
Source. Computed
66 The Role ofAgriculture in Economic Development and Poverty Reduction:

Table 2. Steady state growth rates of different economies under various degrees of openness and agricultural TFP
elasticities

Economicparameters Alpha= 0.60 Beta= 0.30 Theta= 0.50

S= 0.20 Gamma= 0.20 Ag.x rte= 0.30
Agricultural TFP elasticities
0.10 0.20 0.30 0.40 0.50 0.60 0.70
Variables Symbols
Nonag. TFP elasticity eta 0.400 0.380 0.360 0.340 0.320 0.300 0.280
Nonag. tax rate tn 0.260 0.260 0.260 0.260 0.260 0.260 0.260
Growth rate % percent 6.99% 6.05% 5.43% 5.01% 4.73% 4.56% 4.48%
Gr. Rate of real incomes of percent 3.50% 3.63% 3.80% 4.01% 4.26% 4.56% 4.93%
poor
Share of cap in n-sector mu 0.712 0.712 0.712 0.712 0.712 0.712 0.712
Share of pub cap in agr. rho 0.048 0.095 0.143 0.190 0.238 0.286 0.333
Share of lab. in agr. lambda 0.503 0.503 0.503 0.503 0.503 0.503 0.503
Share of agr. In GDP Sa 0.401 0.401 0.401 0.401 0.401 0.401 0.401
Economic parameters Alpha= 0.60 Beta= 0.30 Theta= 0.50
S= 0.20 Gamma= 0.40 Ag.tx rte= 0.30
Agricultural TFP elasticities
0.10 0.20 0.30 0.40 0.50 0.60 0.70
Variables Symbols
Nonag. TFP elasticity eta 0.500 0.460 0.420 0.380 0.340 0.300 0.260
Nonag. tax rate tn 0.373 0.373 0.373 0.373 0.373 0.373 0.373
Growth rate % percent 7.64% 6.34% 5.46% 4.84% 4.42% 4.15% 4.02%
Gr. Rate of real incomes of percent 3.82% 3.80% 3.82% 3.87% 3.98% 4.15% 4.42%
poor
Share of cap in n-sector mu 0.547 0.547 0.547 0.547 0.547 0.547 0.547
Share of pub cap in agr. rho 0.074 0.148 0.222 0.296 0.370 0.444 0.519
Share of lab. in agr. lambda 0.554 0.554 0.554 0.554 0.554 0.554 0.554
Share of agr. In GDP Sa 0.482 0.482 0.482 0.482 0.482 0.482 0.482
Economicparameters Alpha= 0.60 Beta= 0.30 Theta= 0.50
S= 0.20 Gamma= 0. 60 Ag.tx rte= 0.30
Agricultural TFP elasticities
0.10 0.20 0.30 0.40 0.50 0.60 0.70
Variables Symbols
Nonag. TFP elasticity eta 0.600 0.540 0.480 0.420 0.360 0.300 0.240
Nonag. tax rate tn 0.502 0.502 0.502 0.502 0.502 0.502 0.502
Growth rate % percent 13.13% 10.77% 9.12% 7.92% 7.06% 6.48% 6.17%
Gr. Rate of real incomes of percent 6.56% 6.46% 6.38% 6.33% 6.35% 6.48% 6.79%
poor
Share of cap in n-sector mu 0.253 0.253 0.253 0.253 0.253 0.253 0.253
Share of pub cap in agr. rho 0.091 0.182 0.273 0.364 0.455 0.545 0.636
Share of lab. in agr. lambda 0.596 0.596 0.596 0.596 0.596 0.596 0.596
Share of agr. In GDP Sa 0.568 0.568 0.568 0.568 0.568 0.568 0.568
Source. Computed
An Empirical and ConceptualFramework 67

Table 3. Steady state growth rates of different economies under various values of food budget shares and
agricultural tax rates

Economic parameters Alpha= 0.60 Beta= 0.30 Theta= 0.30

s= 0.20 Gamma= 0.30 Epsilon= 0.40
Agricultural tax rates
0.10 0.20 0.30 0.40 0.50
Variables Symbols
Nonag. tax rate tn 0.447 0.412 0.374 0.334 0.292
Growth rate % percent 4.71% 4.74% 4.74% 4.70% 4.63%
Gr. Rate of real incomes of percent 3.77% 3.79% 3.79% 3.76% 3.70%
poor
Share of cap in n-sector mu 0.651 0.683 0.717 0.753 0.790
Share of pub cap in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.518 0.481 0.441 0.396 0.347
Share of agr. In GDP Sa 0.320 0.327 0.335 0.342 0.350
Economicparameters Alpha= 0.60 Beta= 0.30 Theta= 0.50
s= 0.20 Gamma= 0.30 Epsilon= 0.40
Agricultural tax rates
0.10 0.20 0.30 0.40 0.50
Variables Symbols
Nonag. tax rate tn 0.428 0.373 0.314 0.252 0.184
Growth rate % percent 4.28% 4.46% 4.61% 4.72% 4.81%
Gr. Rate of real incomes of percent 3.43% 3.57% 3.69% 3.78% 3.85%
poor
Share of cap in' n-sector mu 0.551 0.594 0.639 0.686 0.734
Share of pub cap in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.620 0.577 0.530 0.478 0.420
Share of agr. In GDP Sa 0.425 0.433 0.441 0.449 0.458
Economicparameters Alpha= 0.600 Beta= 0.300 Theta= 0.70
s= 0.200 Gamma= 0.300 Epsilon= 0.40
Agricultural tax rates
0.100 0.200 0.300 0.400 0.500
Variables Symbols
Nonag. tax rate tn 0.396 0.306 0.210 0.108 -0.001
Growth rate % percent 3.82% 4.18% 4.48% 4.73% 4.98%
Gr. Rate of real incomes of percent 3.06% 3.34% 3.58% 3.79% 3.99%
poor
Share of cap in n-sector mu 0.443 0.499 0.557 0.615 0.675
Share of pub cap in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.715 0.667 0.614 0.556 0.490
Share of agr. In GDP Sa 0.546 0.554 0.562 0.570 0.579
Source. Computed
68 The Role ofAgriculture in Economic Development andPoverty Reduction:

Table 4. Steady state growth rates of different economies under various degrees of saving rates and food
budget shares

Economic parameters Alpha= 0.60 Beta= 0.30 Ag.tx rte= 0.30

s= 0.10 Gamma= 0.30 Epsilon= 0.40
Food budget shares
0.30 0.40 0.50 0.60 0.70
Variables Symbols
Nonag. tax rate tn 0.365 0.332 0.288 0.227 0.137
Growth rate % percent 1.17% 1.12% 1.07% 1.02% 0.95%
Gr. Rate of real incomes of poor percent 0.94% 0.90% 0.85% 0.81% 0.76%
Share of cap in n-sector mu 0.703 0.659 0.614 0.567 0.519
Share of pub cap in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.458 0.508 0.557 0.604 0.650
Share of agr. In GDP Sa 0.354 0.413 0.477 0.546 0.620
Economic parameters Alpha= 0.60 Beta= 0.30 Ag.tx rte= 0.30
s= 0.20 Gamma= 0.30 Epsilon= 0.40
Food budget shares
0.30 0.40 0.50 0.60 0.70
Variables Symbols
Nonag. tax rate tn 0.374 0.348 0.314 0.270 0.210
Growth rate % percent 4.74% 4.68% 4.61% 4.54% 4.48%
Gr. Rate of real incomes of poor percent 3.79% 3.74% 3.69% 3.64% 3.58%
Share of cap in n-sector mu 0.717 0.679 0.639 0.599 0.557
Share of pub cap in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.441 0.486 0.530 0.573 0.614
Share of agr. In GDP Sa 0.335 0.386 0.441 0.500 0.562
Economic parameters Alpha= 0.60 Beta= 0.30 Ag.tx rte= 0.30
s= 0.30 Gamma= 0.30 Epsilon= 0.40
Food budget shares
0.30 0.40 0.50 0.60 0.70
Variables Symbols
Nonag. tax rate tn 0.383 0.362 0.336 0.304 0.264
Growth rate% percent 7.52% 7.46% 7.39% 7.31% 7.24%
Gr. Rate of real incomes of poor percent 6.02% 5.96% 5.91% 5.85% 5.79%
Share of cap in n-sector mu 0.732 0.698 0.664 0.629 0.593
Share of pub cap in agr. rho 0.250 0.250 0.250 0.250 0.250
Share of lab. in agr. lambda 0.423 0.463 0.503 0.541 0.578
Share of agr. In GDP Sa 0.316 0.360 0.406 0.455 0.507
Source. Computed
 An Empiricaland ConceptualFramework 69

Table 5. Steady state growth rates of different economies that exhibit different degrees of labor intensity
among sectors

Economic parameters Beta= 0.10 Theta= 0.50 Agtx rte= 0.30

s= 0.20 Gamma= 0.30 Epsilon= 0.40
Labor share in agriculture alpha
0.30 0.40 0.50 0.60 0.70 0.80
Variables Symbols
Nonagr. TFP elasticity eta 0.070 0.100 0.130 0.160 0.190 0.220
Nonagl tax rate tn -0.068 -0.029 0.011 0.051 0.090 0.130
Growth rate % percent 2.93% 2.91% 3.29% 4.03% 5.15% 6.73%
Gr. Rate of real: incomes of poor percent 3.22% 2.91% 2.96% 3.22% 3.60% 4.04%
Share of cap in n-sector mu 0.603 0.639 0.680 0.727 0.780 0.842
Share of pub cap in agr. rho 0.632 0.545 0.480 0.429 0.387 0.353
Share of lab. in agr. lambda 0.628 0.693 0.738 0.772 0.798 0.819
Share of agr. In GDP Sa 0.551 0.542 0.532 0.522 0.512 0.500
Economic parameters Beta= 0.30 Theta= 0.50 Ag.tx rte= 0.30
s= 0.20 Gamma= 0.30 Epsilon= 0.40
Labor share in agriculture alpha
0.30 0.40 0.50 0.60 0.70 0.80
Variables Symbols
Nonagr. TFP elasticity eta 0.270 0.300 0.330 0.360 0.390 0.420
Nonag. tax rate tn 0.196 0.235 0.275 0.314 0.354 0.393
Growthlrate % percent 3.12% 3.29% 3.79% 4.61% 5.77% 7.38%
Gr. Rate of real incomes of poor percent 3.43% 3.29% 3.41% 3.69% 4.04% 4.43%
Share of cap1 in n-sector mu 0.503 0.542 0.587 0.639 0.703 0.780
Share of pub cap in agr. rho 0.308 0.286 0.267 0.250 0.235 0.222
Share of lab. in agr. lambda 0.361 0.429 0.484 0.530 0.568 0.601
Share of agr. In GDP Sa 0.481 0.468 0.455 0.441 0.426 0.411
Economic parameters Beta= 0.50 Theta= 0.50 Ag.tx rte= 0.30
s= 0.20 Gamma= 0.30 Epsilon= 0.40
Labor share in agriculture alpha
0.30 0.40 0.50 0.60 0.70 0.80
Variables Symbols
Nonagr. TFP elasticity eta 0.470 0.500 0.530 0.560 0.590 0.620
Nonag. tax rate tn 0.459 0.499 0.538 0.578 0.618 0.657
Growth rate % percent 6.61% 7.05% 7.90% 9.14% 10.84% 13.14%
Gr. Rate of real incomes of poor percent 7.27% 7.05% 7.11% 7.31% 7.59% 7.88%
Share of cap in n-sector mu 0.336 0.372 0.415 0.470 0.542 0.639
Share of pub cap in agr. rho 0.203 0.194 0.185 0.176 0.169 0.162
Share of lab. in agr. lambda 0.253 0.311 0.361 0.404 0.441 0.474
Share of agr. In GDP Sa 0.384 0.366 0.347 0.327 0.306 0.283
Source. Computed
 - - - - -- - - -- -- -,- - . - -- - - - - - -- - -
Th.Wo d k k - .

Rural Development Department

The World Bank
1818 H Street, N.W., Room MC5-724
Washington, D.C. 20433
website: http://www.worldbank.org