Chapter 43

L A B O R MARKETS IN D E V E L O P I N G COUNTRIES

JERE R. BEHRMAN*

University of Pennsylvania

Contents

Abstract 2860
JEL codes 2860
1 Introduction 2860
2 T h e h o u s e h o l d e n t e r p r i s e m o d e l , surplus labor, d i s g u i s e d e m p l o y m e n t and
u n e m p l o y m e n t , c o m p l e t e m a r k e t s a n d separability, rural d u a l i s m 2863
2.1 Complete markets 2866
2.2 No markets 2869
2.3 Tests of separability and of market completeness 2870
2.4 Empirical studies of rural labor supplies 2871
2.5 Household formation 2875
3 L a b o r c o n t r a c t s , risks a n d i n c e n t i v e s 2876
3.1 Dominatace of household farms in agriculture 2877
3.2 Day versus longer-run labor contracts 2878
3.3 Implications of land contracts for labor 2879
3.4 Empirical studies of rural labor supplies and risk 2880
3.5 Elnph'ical studies of imperfect information and labor markets 2882
4 D e t e r m i n a n t s o f a n d r e t u r n s to h u m a n capital i n v e s t m e n t s 2884
4.1 Determinants of health and nutrition investments 2888
4.2 Productivity impact of health and nutrition 2893
4.3 Determinants of schooling 2901
4.4 Impact of schooling on economic productivity 2910
5 Urban labor markets, labor-market regulations, international trade policies
and m a n u f a c t u r i n g 2916
5.1 Urban labor market dualism 2916
5.2 The effects of labor market regulations on ~brmal-sector wages and employment 2917
5.3 The effect of trade reform and adjustment on formal-sector manufacturing labor 2918
6 Distribution and mobility 2920

* The author has benefited fi'om comments on this chapter by the editors of this volume, Orley Ashenfelter and
David Card, and from discussions of issues related to this chapter with Harold Alderman, Nancy Birdsall, Anil B.
Deolalikar, Elizabeth M. King, Victor Lavy, Robert A. Pollak, Richard Sabot, John Strauss, Duncan Thomas,
Kenneth Wolpin and, most substantially, Andrew Foster and Mark R. Rosenzweig. Behrman's time for writing
this chapter was funded in part by NIH 5-RO1-HD30907. Behrman alone is responsible for all interpretations and
any errors in this chapter.

Handbook of Labor Economics, Volume 3, Edited by O. AshenJelter and D. Card

© 1999 Elsevier Science B.V. All rights reserved.

2859
2860 J. R. Behrman

6.1 Intertemporal aspects of distribution and the Kuznets hypothesis of inverted U pattern in
inequality with development 2921
6.2 Geographical mobility 2923
6.3 Empirical investigation of urban informal-formed mobility 2926
6.4 Distributional differences among demographic groups 2927
7 Conclusions 2929
References 2930

Abstract

This chapter covers selected topics for the 80% of the world's labor force that works in the devel-
oping countries. These topics are ones that have: (1) received relatively great attention in developing
countries compared to developed economies (i.e., family enterprises, missing labor markets, geogra-
phical mobility, health/nutrition effects on productivity) because of their greater importance in
developing countries; (2) been considered more extensively for developing than developed labor
mm'kets because the nature of institutions, behaviors and available data permit more extensive
empirical examination of these topics (i.e., labor adjustments to shocks in the presence of imperfect
markets, information problems in labor markets), and (3) been considered extensively for both
developing and developed economies but with some different approaches and results for part of
the developing country literature (e.g., determinants of and labor market returns to schooling). The
discussion is organized around five broad topics: (1) The household enterprise model, surplus labor,
disguised employment and unemployment, complete markets and separability, and labor supplies;
(2) labor contracts, risks and incentives; (3) determinants of and returns to human capital investments
(including health and nutrition in addition to schooling); (4) urban labor markets, labor-market
regulations, international trade policies and manufacturing; and (5) distribution and mobility.
© 1999 Elsevier Science B.V. All rights reserved.
JEL codes: Jl; J2; J3; J4; JO

1. I n t r o d u c t i o n

This chapter is different in essence from most of the other chapters in the four volumes of
the Handbook o f Labor Economics. A l m o s t all of the other chapters are concerned with
generic and fairly narrowly focused topics in labor economics analysis, generally based on
labor markets in the United States, Canada and a few Western European and very few non-
European economies (e.g., Israel, Australia, Japan). This geographical concentration
contrasts sharply with the distribution of the w o r l d ' s labor force in recent decades, and
even more with the projected future distribution of the w o r l d ' s labor force (Table 1). The
share of the w o r l d ' s labor force in "High-income O E C D " countries (a subset of Which is
the focus of most of the chapters) was only 20% in 1965, 15% in 1995 and projected to be
10% in 2025. Therefore most of the Handbook chapters focus on a small and steadily-
becoming smaller proportion of the world' s labor force.
H o w should this narrow concentration be interpreted? One possibility is that the cover-
Ch. 43." Labor Markets in Developing Countries 2861

Table 1
The world's labor force by country income group and region a

Income group or region Millions of workers b Percentage of total

1965 1995 2025 1965 1995 2025

World 1329 2476 3656 100 100 100

Income group
High-income 272 382 395 21 15 11
Middle-income 363 658 1020 27 27 28
Low-income 694 1436 2241 52 58 61

Region
Sub-Saharan Africa 102 214 537 8 9 15
East Asia and Pacific 448 964 1201 34 39 33
South Asia 228 440 779 17 18 21
Europe and Central Asia 180 239 281 14 10 8
Middle East and N. Africa 29 80 204 2 3 6
Latin America and Caribbean 73 166 270 5 6 7
High-income OECD 269 373 384 20 15 10

Source: World Bank (1995, Table 1.1) from International Labour Organization sources.
b Ages 15-64.

age in m o s t o f these chapters is i n d e e d generic and e x a m p l e s just h a p p e n e d to be very

concentrated on the s a m e f e w e c o n o m i e s because r e l e v a n t i n f o r m a t i o n is m o r e available
for those e c o n o m i e s . But the basic analyses h o l d m o r e b r o a d l y for all e c o n o m i e s . A t a
general level, I find such a possibility intellectually attractive. I think that the basic
theoretical tools, e m p i r i c a l m e t h o d s and approaches o f labor e c o n o m i c s i n d e e d are applic-
able to all e c o n o m i e s . M o s t of the H a n d b o o k chapters, h o w e v e r , do not explicitly argue
that they are c o v e r i n g all e c o n o m i e s , but using particular e x a m p l e s f r o m a small subset of
e c o n o m i e s out of c o n v e n i e n c e . T o the contrary they often read as i f they are c o n c e r n e d
basically with the f e w e c o n o m i e s that are m e n t i o n e d explicitly. M o r e o v e r , w h i l e on a
general l e v e l it is attractive to say that the basic tools and m e t h o d s apply to all e c o n o m i e s ,
the specific application g e n e r a l l y depends on the specific institutions b e i n g e x a m i n e d ,
w h i c h w o u l d appear in m a n y cases to vary considerably across e c o n o m i e s .
This chapter c o v e r s labor e c o n o m i c s for the 80% or so o f the w o r l d ' s labor force that
lives and w o r k s in the d e v e l o p i n g countries. ~ T h e r e f o r e I a m a l m o s t forced to adopt a
different strategy than in m o s t o f the chapters in the H a n d b o o k o n L a b o r E c o n o m i c s . The
structure o f the m a j o r i t y o f the other chapters is to d e v e l o p systematically a b e h a v i o r a l
m o d e l that addresses a fairly well-defined specific issue (e.g., w o m e n ' s labor supplies, the

Svenjar's chapter covers about 5-6% of the world's labor force that also are included in Table 1 among the
low- and middle-income developing countries (according to the World Bank (1997) Albania, Armenia, Belarus,
Bulgaria, Czech Republic, Estonia, Georgia, Latvia, Lithuania, Macedonia, Poland, Romania, Russian Federa-
tion, Slovak Republic, Slovenia, and Ukraine account for 5.7% of the world's population).
2862 J.R. Behrman

impact of education on wages) that unifies the previous literature, permits a basis for
evaluating empirical studies in the literature, and perhaps is tested against some (usually
US or UK) data. That is an attractive prototype, one that in many areas is the most
promising way for advancing our knowledge of labor market issues. But it does not
seem one that I usefully can adopt in this chapter. Were I to devote the whole chapter
to, say female labor supplies in Brazil or to labor mechanisms for coping with shocks in
rural India, I could do so. But that would not seem to remedy much what I see as the
problem of the limited coverage of 80% of the world's labor force.
The compromise that I have adopted in this chapter is to discuss selected topics related
to labor markets in developing countries. By discussing several different topics I hope to
be able to cover more broadly the work in labor economics in developing countries than
would result if I were to concentrate on a topic that is as narrowly defined as are the topics
in most of the chapters in the handbook, although of course the selective topical coverage
cannot be nearly as extensive for the 80% of the world's labor force considered in this
chapter as for the 15% of the world' s labor force on which most of the H a n d b o o k chapters
concentrate. A n inevitable cost, I am afraid, of the effort to cover a number of topics is
more superficial coverage of each topic.
That leaves the question of how to select which topics to discuss in this chapter. The
decision function that I use has three criteria: (1) topics that have received relatively great
attention for labor markets in developing countries compared to developed economies
(i.e., family enterprises, missing labor markets, geographical mobility, health/nutrition
effects on productivity) because of their greater importance in the developing country
contexts; (2) topics that have been considered more extensively for developing than
developed labor markets because the nature of institutions, behaviors and available data
permit more extensive empirical examination of these topics (i.e., labor adjustments to
shocks in the presence of imperfect markets, information problems in labor markets); 2 and
(3) some topics that have been considered extensively for both developing and developed
economies but with some different approaches and results for part of the developing
country literature (e.g., determinants of and labor market returns to schooling).
From a broad aggregate perspective there are a number of systematic differences in the
distribution and composition of labor in developing versus developed economies. Because

2 Sometimesit is claimedthat data are much better for developedthan for developingcountries,whichmay be
part of the reason that studies have concentrated so much on the former. Certainlyit would appear that labor
market data often are better for developed economies because of longer-establishedsystematic data collection
procedures (although for some developingeconomics, such as India, public data collection procedures were
establishedrelativelyearly), more educated populations,and more extensiveand more regulatedmarkettransac-
tions. But there are other factors working in the opposite direction, such as lower costs of data collectionand
simplerinstitutions.The wellknownICRISATvillage-leveldata from rural India are an example.In parl because
of the low cost of labor, experimentscould be and were performed to ascertain the extent of risk aversionwith
prizes on the order of magnitudeof a months' wages (Binswanger, 1980) and enumeratorswith master degrees
resided virtuallyfull-timein the samplevillagesand collecteddetaileddata over a decade, includinginformation
of certain types that is veryhard to collectfor developedeconomies(e.g., data on exogenousproductivityshocks,
intrahouseholdfood allocation).
Ch. 43: Labor Markets in Developing Countries 2863

these differences affect the choice of topics on which analysis has focused, a brief
summary of these features based on aggregate data is useful for perspective before turning
to the selected topics. 3
Table 2 provides means for three country groups defined in World Bank (1997) by per
capita incomes in 1995 for variables related to population and GNP per capita, labor force
participation and composition, human capital, and some non-labor inputs. Some salient
features of labor markets in developing (low- and middle-income) versus developed
economies come through strongly even in such aggregate data. In the developing econo-
mies (in comparison with developed economies):
1. Agriculture and other rural labor activities are much more important even though
average labor products in agriculture are relatively much lower than those in industry;
2. Non-wage labor (largely unpaid family workers, particularly in agriculture at lower
incomes) are much more important;
3. Labor forces are growing more rapidly;
4. Labor force participation rates among 15-64 year olds are higher (particularly for low-
income countries), in part because of much lower schooling enrollment rates among
those who are in the youngest cohort in this age range;
5. Human capital investments are lower, with larger gender gaps favoring males; and
6. Non-labor production inputs per worker are much smaller.
These differences shape much of the difference in emphasis in labor economics for devel-
oping than for developed economies. The literature on developing countries, for example,
has emphasized much more household enterprises in agriculture and rural-urban mobility
than has the literature on developed country labor markets.
The remainder of this chapter is organized with reference to five broad topics, selected
as indicated above, that have received considerable emphasis in the literature on labor in
developing countries: (1) the household enterprise model, surplus labor, disguised
employment and unemployment, complete markets and separability, and labor supplies;
(2) labor contracts, risks and incentives; (3) determinants of and returns to human capital
investments (including health and nutrition in addition to schooling); (4) urban labor
markets, labor-market regulations, international trade policies and manufacturing; and
(5) distribution and mobility.

2. The household enterprise model, surplus labor, disguised employment and

unemployment, complete markets and separability, rural dualism

As illustrated in Table 2, two major features of labor markets in most developing econo-

3There are substantial limitations withaggregatedata, manyof whichare reviewedin the symposiumedited by
Srinivasan (| 994), with the considerationof labor and schoolingin Behrmanand Rosenzweig(1994) particularly
relevant for this chapter. But suchdata of necessityshapeour understandingof the broadpatterns of labormarkets
and other aspects of economies.
2864 .L R. Behrman

Table 2
Aggregate data on population and GNP per capita, labor force participation and composition, human capital
investments, and non-labor production inputs for low-, middle- and high income country groups ~

Low-income Middle-income High-income

economies economies economies

Population and GNP per capita

Population (millions) mid- 1995 3180 1591 902
Population av. ann. growth (%) 1980-1990 1.9 1.7 0.7
GNP per capita US$1995 430 2390 24930
GNP per capita average annual growth rate 3.8 -0.7 1.9
(%) 1985-1995
Population 15-64/total (%)
1980 57 58 64
1995 61 62 67
Urban population % of total
1980 21 52 63
1995 29 60 75
Urban population in cities > 1 million %
1980 7 16 31
1995 10 20 33

Labor force and composition

Labor force average annual growth rate (%) 2.0 2.0 1.1
1980-1995
Labor force part. rate (15-64) (%)
1980 86 72 70
1995 81 70 71
Female/total labor force (%)
1980 40 36 39
1995 41 38 42
Agricultural labor/total (%)
1980 73 38 9
1990 69 32 5
Industrial labor/total (%)
1980 13 28 35
1990 15 27 31
Agricultural value added per labor as % of
industrial value added per labor
1980 19 - 32
1990-1995 14 27 39
Non-wage labor as % of total labor 1980- 91 41 16
1991 b
Non-wage labor in agriculture as % of total 59 20 3
labor 1980-1991 b

Human capital
Life expectancy at birth (years) 1995 63 68 77
Adult illiteracy (%) 1995 total 34 18 <5
Ch. 43." Labor Markets in Developing Countries 2865

Table 2 (continued)

Low-income Middle-income High-income

economies economies economies

Female 45 23 <5
Male 24 14 <5
School enrollment% of age group
Primary
Female 1980 81 99 103
1993 98 101 103
Male 1980 104 106 103
1993 112 105 103
Secondary
Female 1980 26 48 -
1993 41 62 98
Male 1980 42 53
1993 - 64 97
Tertiary
1980 3 21 35
1993 20 56

Non-labor production inputs

Cropland/agriculturallaborer (km2)
1980 0.0058 0.028 0.116
1994 0.0045 0.028 O.178
Oil equivalentenergy use per capita (kg)
1980 248 1537 4644
1994 369 1475 5066

a Source: World Bank (1997, Tables 1, 4, 7, 8, 9, 12) except as indicatedin note b. Populationweightsused for
averages.
bWorld Bank (1995, Table A-2) with populationweights from World Bank (1997, Table 1)

mies are that (1) agriculture is a major (in the earlier stages of development usually the
major) sector of employment and (2) family farms/enterprises are major (often the major)
employers. Moreover, influential early two-sector aggregate development models of
Lewis (1954) and Ranis and Fei (1961) argued that in the early stage of development,
(i) workers could be shifted from traditional agriculture to m o d e m market-oriented sectors
("industry") without any reduction in agricultural output (for which reason these models
are called "surplus labor" models) and (ii) workers in traditional agriculture received their
average products (i.e., their share of total household production) so that was the private
opportunity costs of migrating to industry. Thus these models had assumptions about
agricultural households and labor markets that are central to their implications. Further,
much of the development literature on labor (and other) contracts (Section 3) and on
human resources and labor markets (Section 4) focuses on rural households.
For all of these reasons, a good starting point for considering analysis of labor markets
in developing countries - and how that analysis differs in some important respects from the
2866 Z R. Behrman

analysis of labor markets in developed economies - is to consider models of rural house-

holds.
The standard model of labor markets used for developed economies distinguishes
between labor suppliers (households) and labor demanders (firms). For a substantial
proportion of both rural and urban households in developing countries, both labor supplies
and labor demands are determined within the same institution - family farms or firms.
Models for family farms in traditional agriculture date back to Chayanov (1925). Singh
et al. (1986) and Rosenzweig (1988a) provide summaries of the literature as of the mid-
1980s. I build on the latter for my initial discussion, but I add discussion of some aspects of
more recent contributions below.

2.1. C o m p l e t e m a r k e t s

1 first consider a basic one-period model for the perfect markets case (or, more accurately,
complete except for one market case - because in most of this discussion I assume that
there is no land market). Assume there is a household with given size (M members),
demographic composition (D dependants and N workers) and land area (A). The household
welfare function depends on average consumption of the M household members (c) and
average leisure of the N workers (l = T - T w, where T is total time and Tw is the time
worked by a worker):
U = U ( c , 1). (1)

The farm production function gives the farm output (Q) as a function of land, labor used in
agricultural production (L), and other inputs used in agricultural production (F):
Q = Q(A, L, F). (2)
The household budget constraint is

PQQ - WL - PFF + WNT + I1o - P Q M C - W N l = 11 + W N T + Y o - P Q M C - WNI

= YF -- P Q M C - W N l = O,
(3)
where the P' s refer to the respective prices, W is the wage rate, Yo is other income, I I is
farm/firm profits, and YF is full income. The profit function is
/ / ( P Q , W, PF) -~ P Q Q ( A , L, F , ) - W L - P F F . (3A)

Under the assumption that the functions have the desirable properties so that there is an
interior solution, constrained maximization of the welfare function subject to the produc-
tion function and the budget constraint leads to optimal consumption, leisure, production,
sales/purchases of outputs, time worked by household members, labor used in production
and other inputs used in production.
The first-order condition for labor used in production is
Ch. 43: Labor Markets in Developing Countries 2867

PQQL = W. (4)

This also is the profit-maximizing condition for use of labor, and the profit-maximizing
condition for the other market input also is satisfied. So the constrained maximization of
household welfare yields profit maximization for the farm. At the profit-maximizing level
of labor used in farm production, the household may have positive or negative labor
supplied to the labor market ( = N T w * - L*, where * refers to the optimum levels).
Farm households with little land relative to their number of workers are positive suppliers
to the labor market and farm households with a lot of land relative to their number of
workers are negative suppliers (demanders).
In the case of complete markets for all but one input/product, the production-consump-
tion decisions can be treated separately - as if in the first step the farm household maxi-
mized full income (which is equivalent to maximizing profits because the wage rate, total
worker's time and other income are given) and in the second step, the farm household
chooses its consumption bundle as if profits were given. This means that all prices and
assets that affect the profit maximizing decision also have an impact through profits on
consumption. But any prices that affect only the consumption decision do not affect the
profit maximizing decision. 4 Note that it is full income (profits) and consumption that are
separable, and not monetized income and consumption unless labor supply is fixed. The
separability result, moreover, does not depend on the simple one-period model. If the
model is extended to include S states of the world and T time periods so that all the
variables have subscripts s and t, the problem is still recursive. Land and labor used in
agricultural production for the sth state in the tth period appear only in the farm profit
function for the sth state and the tth period so the household can maximize welfare by first
maximizing profits for each state of the world in each time period and then making
consumption decisions. The simplification is tremendous - reducing a possible risk-
adverse household's dynamic behavior in a risky environment to a simple static profit-
maximization problem.
The separability result does depend on there being no more than one missing market -
the market for land is the only missing market in the above example. Separability does not
exist if the labor market is not perfect, which has been the most -emphasized missing
second market in the literature. But other missing second markets also may cause separ-
ability not to exist (Srinivasan, 1972; Feder, 1985; Eswaran and Kotwal, 1986; Banerjee
and Newman, 1993). Consider the following simple example in Udry (1996b) in which
there are two states of nature, with a probability of ~ of state 1 and with multiplicative
production shocks 0s. With complete labor and insurance markets (but no land market and
no input F into production), the household's problem is to choose cl, c2, l, L to

maxTrU(Cl, l) + (1 - 7r)U(c2,/), (5A)

4 In the model discussed here there are not may such prices because the consumption good price is also the
production good price and the leisure price is tied directlyto the price of labor for production. But more generally
households are often modeled as consuming some goods that they do not produce (often only such goods).
2868 J. R. Behrman

s.t. PQIMCl + PQ2MC2 + W N l <-- (Pet 0j + P Q 2 0 2 ) Q ( A , N T - N1).

Separation holds and the household maximizes farm profit. In contrast, if there is no labor
market but a complete insurance market, the farm household's problem is to choose Cl, ca,
1 to
max~-U(c~, l) + (1 - ~)U(c2,1), (5B)

s.t. P Q I M Q + PQaMC2 + WN[ <~ (PQlOl + P Q 2 0 2 ) Q ( A , N T - IV[).

Separation is violated because farm output increases in the farm household's labor endow-
ment. If the labor market is complete but there is no insurance market the household's
problem is to choose Cl, c2, l, L to
maxTrU(cl, l) + (1 - 7T)U(c2, l), (5C)

s.t. PQ1MCl Jr- WN1 <- PQ10tQ(A,L) - W L + W N T and

PQ2MC 2 + W N l <-- PQ202Q(A,L) - W L + W N T .

One of the first-order conditions is

)h(PQI0~QL -- W ) + A2(PQ202Q L - W) = 0, (5D)
where A, is the marginal utility of income in state s. Separation does not hold because input
decisions depend on the ratio of the marginal utilities of income in the two states. An
increase in the household's labor endowment affects this ratio (increasing the marginal
utility of income relatively in the state in which the household has a larger production
shock if the household has diminishing absolute risk aversion) and thus changes input
decisions.
The first-order condition for the optimal leisure-consumption good combination is
UL/Uc = - WN/M. (6)
The shadow wage of leisure generally is less than the wage rate because consumption per
family member increases less than the wage if work increases by 1 h if there are non-
working household members, ff the household had no land, this condition also would hold.
Therefore it might appear that labor supply behavior would be identical between farm
households and landless households if they faced identical prices and had identical full
income.
But that is not the case. Rosenzweig (1988a) shows that the household labor supply
elasticities with respect to both wages and the consumption/production good price differ
depending, respectively, on whether the household is a positive or negative net supplier of
labor to the wage market and on whether the household is a positive or negative net seller
of the consumption/production good:
~Tw,W = -- ~I,W ~ -- W ( N T w * - L*)/YF~x,y F, (7A)
Ch. 43: Labor Markets" in Developing Countries 2869

~Tw,P ~----- 'r/l,W -- W ( Q * - MC'4:)/YFTII,YF . (7B)

The first term in each expression is the negative of the compensated price elasticity (and
must be positive in the first expression). The second term in both expressions is the
weighted full income elasticity, with the respective weights being the share of labor
market income in full income and the share of product sales in full income. The weights
in both cases can be positive or negative. Thus the overall elasticities depend, for given full
income, on the share of productive assets in total wealth.
Relation (7A) implies that the larger the share of land returns in full income for a given
full income level, the more likely households are negative suppliers of labor to the wage
market, the less are their income gains due to a wage increase, and the larger is their
elasticity of household member time worked with respect to wages. Rosenzweig (1980)
finds support for the model leading to this relation (see discussion of labor supplies below).
Relation (7B) implies that the larger the share of land returns in full income for a given
full income level, the more likely households are positive suppliers of the production/
consumption good to the market, the greater is their income gains due to a production/
consumption good price increase, and the lesser is their elasticity of household member
time worked with respect to the price of the production/consumption good. If a household
is a net supplier of these goods, leisure is a normal good, and leisure and goods are not
strong complements, the household labor supply falls with an increase in the price of
goods. If a household is a net demander of these goods, household labor supply could rise
with a price increase. 5
Even though labor supplies respond differently to price and wage changes in the
complete market model, depending on the composition of asset ownership, if there are
constant returns to scale in production, a reallocation of land among farmers does not
affect the efficiency of input use. That there is no land market, thus, does not preclude
efficiency because all other production inputs (including labor) are mobile across farms.

2.2. N o m a r k e t s

Some of the development literature (e.g., some modeling of subsistence peasant house-
holds) assumes the other extreme of no markets, or at least no labor markets. Sen (1966)
presents an autarkic model that can be viewed as a special case of relations (1)-(3), with no
purchased agricultural inputs (i.e., no F) and no labor market (i.e., no W a n d L = N T w ) . In
this case the first-order condition for the leisure-consumption tradeoff at the optimum is

UI/U ~ = QLN/M. (8)

This is identical to relation (6) above for the complete market case with the critical

5 These, or course, are partial equilibrium results. From a general equilibrium perspective, price and wage
changes may be interrelated. For instance, an upward shift in the demandfor the good due to expandingurban and
international markets might cause an upward shift in agricultural labor demands and, if labor supplies are not
completely elastic, increase rural wages.
2870 ~ R. Behrman

difference that the marginal product of labor in the household under consideration repre-
sents the relevant tradeoff rather than the market wage because the opportunity cost of
increasing leisure by a small amount is foregoing the marginal product of that labor in
household production. Parallel to relation (6), the shadow wage of leisure generally is less
than the marginal product of labor because consumption per family member increases less
than the marginal product of labor if work increases by one hour if there are non-working
household members.
This model is consistent with so-called "surplus labor" if the removal of a working
family member does not change the marginal product of labor so that total output is
unaltered. This might occur if the remaining household members each increase their
time working by N/(N - 1) so total family time working is constant. Sen calls this situa-
tion one of "disguised unemployment" because the marginal product of labor is positive
but workers can be removed from the household without a drop in output. In order for
relation (8) to hold in the new configuration of household membership, the left side of
relation (8) must adjust for the presence of one less household member and one less worker
(i.e., the constant marginal product of labor is multiplied by (N - 1)/(M - 1) which is not
equal to the N/M that was relevant before the worker's departure except in the case in
which N = M). That is, if a worker leaves, the marginal rate of substitution between
leisure and consumption tradeoff between leisure and consumption must fall because
the (in this case same) marginal product of labor has to be shared with a proportionately
larger non-working share in the household. Thus the existence of surplus labor in this
model depends critically on the characteristics of the family preference function. 6
Because each household makes its decisions in isolation, these equilibrium decisions
reflect directly the preference and production function parameters and the initial assets of
households. This is inefficient because production could be increased with redistribution of
the same assets to equalize worker/land ratios. The welfare of households in general could
be improved with the introduction of a labor or a land market.

2.3. Tests' of" separability and of market completeness

Tests of separability have been proposed and applied to developing countries to see if the
complete market assumptions hold approximately empirically. Of course a complete set of
competitive markets in the strict sense does not exist in rural areas of developing countries,
nor anywhere else. But a broad variety of spot markets do seem to exist and, at least in
some contexts, appeal" to operate competitively. Moreover, informal mechanisms exist that
may fill at least some of the same functions as competitive markets (e.g., insurance
functions). Furthermore, most empirical studies of the farm household model have separ-

6 Simpler models can yield surplus labor. For example, Rosenzweig (1988a) suggests that the simplest such
model is one in which there is no labor market and the household does not value leisure so the optimizing
condition is UcQL = 0 if there are enough workers in the household both before and after the departure of a
worker.
Ch. 43: Labor Markets in Developing Countries 2871

ability as a maintained assumption, so it is useful to test whether this assumption is

warranted.
If there is separability, as noted above, a farm household's dynamic behavior in a risky
environment leads to a recursive problem in which the first-step problem is static profit
maximization. Tests of separability generally have focused on the strong exclusion restric-
tions that are implied by the profit function in relation (3A); input demand and output
supply functions depend on prices and agricultural plot characteristics and on nothing else.
A basic estimation problem is that unobserved characteristics that enter into the agricul-
tural production function, such as soil quality, may be correlated with observed household
characteristics that are excluded from such relations (e.g., wealth, assets, demographic
characteristics) so that separability is wrongly rejected. A related test used by Udry
(1996b) is to compare the distribution of plot-specific random shocks on output/area
and on labor input/area among plots operated by the same individual versus the distribu-
tion across all plots in a village. The maintained hypothesis is that the distribution of inputs
among apparently identical plots operated by the same individual is efficient, so if the
distribution of random shocks among plots in a village is the same as the distribution
among plots of the same individual, the distribution across village plots is efficient.
Table 3 summarizes all of the tests of separability of which I am aware. The available
evidence, taken at face value is somewhat mixed, although with an increasing number of
studies that reject separability and thus complete markets. Most of these studies must be
qualified because of possible omitted variables that may be correlated with the included
variables that are the focus of the tests (although if a linear approximation is adequate, the
last of the studies discussed controls for such variables with fixed effects). Subject to this
qualification, the majority of these studies reject the separability assumption necessary to
consider household consumption (including leisure choices) separately from farm house-
hold production. This has important implications for modeling and estimating time alloca-
tion decisions, including those related to labor-leisure choices. This literature to date,
however, is not very satisfactory in identifying the nature of critical market incomplete-
ness, which may be important both for advancing further understanding and for consider-
ing possible policy implications. Basically it is posed in terms of the dichotomous
possibilities of complete versus non-complete markets - in terms of Fig. 1, whether the
complete-markets linear full income constraint or the no-markets production possibility
frontier is relevant. It does not distinguish, for example, to what extent the effective budget
set is kinked due to differential buying/selling prices (holding quality constant) as in the
dashed lines in this figure.

2.4. Empirical studies of rural labor supplies

Given the importance of agriculture and of own-farm/firm labor in developing economies,

not surprisingly much of the emphasis on labor supplies has been within this context. The
better empirical studies in this literature have incorporated or tested various specific
aspects of developing country contexts - distinctions between net suppliers and demanders
2872 Z R. Behrman

Table 3
Summary of tests of separability

Samples Test Result Source

Indonesian farm households Household profits affected by Accept Pitt and Rosenzweig (1986)
illness of household head
240 Zairian farm households Land cultivated per Reject Shapiro (1990)
1985-1986 household worker affected by
household size and
composition
Indonesian rice farm households Labor (family and hired) Accept Benjanain (1992)
affected by household
demographic composition
Burkino Fasian and Kenyan farm Plot agricultural supplies and Reject Udry (1996b)
households inputs affected by land in
other plots, household size,
other income sources
Niger farm households Yield affected by household Reject Gavian and Fafchamps (1996)
manpower
Pakistani farm households Labor demand relations Reject Fafchamps and Quisumbing
affected by household (1997)
demographic variables
Pakistani rural households Net harvest profits affected by Reject Behrmanet al. (1997a)
planting season calorie
consumption
Indian rural villages Land allocations conditional Reject Foster et al. (1997)
on prices affected by
population

Farm Output

M S

Complete M a r k e t ,
Full Income
C onstraint

No Market, Production
Possibility Frontier

Fig. 1. No market production possibility frontier for farm output versus leisure, complete mm-ket full income
constraint, dashed lines for lower selling than buying prices.
Ch. 43: Labor Markets in Developing Countries 2873

of labor; missing or incomplete labor, land, credit and insurance markets; and multi-period
production processes.
Rosenzweig (1980) is an early example with static labor supply estimates that attempt to
explore some of the special features of labor supplies in rural areas of developing coun-
tries. Relation (7A) above implies that the larger the share of land returns in full income for
a given full income level, the more likely households are negative suppliers of labor to the
wage market, the less are their income gains due to a wage increase, and the larger is their
elasticity of household member time worked with respect to wages. Rosenzweig finds
support for the model leading to this relation (under the added assumptions that utility
functions are homothetic and leisure is a normal good) in that landless rural Indian house-
holds have lower labor supply elasticities than do rural Indian households with land.
Skoufias (1993b) is a more recent example of a static investigation similar to household
labor supply studies for developed economies. He examines time allocation of all family
members to the labor market, domestic production, leisure and (for children) schooling in
response to market wages for different household members in rural India. Previous studies
of household time allocations in such rural contexts had been few and based on cross-
sectional data, which means that they had not been able to control for unobserved hetero-
geneity (e.g., in tastes for different time uses, in productivities) that might be correlated
with right side observed variables (e.g., education) and thus lead to biases in the estimates.
This study uses panel data that permit control for unobserved heterogeneity and controls
for possible zero censoring in the dependent variables (i.e., many sample members do not
work in the paid labor market, many children do not go to school). The estimates indicate
that increases in the market wages of one household member have substantial effects on
the time use of other household members.
Jacoby (1993) develops a methodology for estimating structural time-allocation models
for self-employed households. The opportunity costs of time, or shadow wages, of house-
hold members are estimated from an agricultural production function that is flexible with
regard to substitution among different types of family and hired workers. The household's
structural labor supply parameters are recovered from variation in these shadow wages,
using instrumental variables (e.g., fixed production inputs, household demographic char-
acteristics). Because the estimation does not rely on market wages, the implications of
utility theory and the hypothesis of efficient rural labor markets are not tested jointly and
perfect substitutability among different types of family and hired labor is not required.
Estimates are presented for peasant family labor supply behavior in the Peruvian Sierra.
These estimates suggest that these households indeed allocate their members' time as if to
maximize a household utility function in the sense that work efibrt is higher among
peasants who are more productive at the margin and thus face higher opportunity costs
of time. Skoufias (1994b) provides similar estimates, with control for fixed effects from
panel data, for rural India.
Newman and Gertler (1994) develop an estimable structural model to deal with three
aspects of farm/firm households that make empirical labor supply analysis difficult: (i)
households jointly determine the consumption and the labor supplies of household
2874 ~ R. Behrman

members each of whom may (or may not) engage in multiple activities (i.e., own-farm
production, wage labor), (ii) household members' activities are interdependent both in the
utility and the enterprise production functions, and (iii) marginal returns to working in the
household enterprise are not observed. The model consists of two types of structural
equations and an identity: marginal return functions for each activity for each household
member, the household' s marginal rate of substitution of household consumption for each
household member's leisure, and the household budget constraint. Conditional on the
structural relations assumed, the marginal returns to work in self-employment are identi-
fied without being directly observable and without estimating the enterprise production
function by using the Kuhn-Tucker conditions to infer the equilibrium values (or bounds
on such values) depending on participation decisions in wage versus sell-employment (or
both). They present limited-information estimates of this model for household consump-
tion and labor supply decisions of rural Peruvian land-holding households and then use
these estimates to simulate the impact on consumption and on welfare (through induced
leisure as well as consumption changes) of poverty alleviation programs that increase the
returns to work (for females versus males) and for direct transfers. These simulations
indicate, for example, that an increase in female labor market returns leads to larger
changes in welfare and in consumption than does an increase in male labor market returns
because of the higher estimated household evaluation of male relative to female leisure
(given that in the base case prime-age males have the least leisure).
Skoufias (1996) explores intertemporal questions of substitution of labor supplies
among household members in a study that basically applies the intertemporal modeling
first developed for developed economies to the rural developing country context. He
explicitly presents the model and the related estimation issues (e.g., unobserved hetero-
geneities in marginal utilities of wealth, missing wage variables) and explores carefully
the sensitivity of the results to alternative assumptions and gives thoughtful discussion of
why the estimates are of interest for understanding behavior and for informing policy
decisions. He finds that the female intertemporal elasticity of substitution is significant but
small (as in estimates for the United States), with significant differences related to land
ownership and the production stage, which suggest that credit constraints limit intertem-
poral substitution across periods. The estimates for males are negative or zero. The low or
negative elasticities suggest that there are not strong labor-leisure tradeoffs, so seasonally
targeted programs such as public work programs during slack periods are likely to be
effective in increasing household welfare. The sensitivity of the estimates to land owner-
ship and production stage suggest that credit constraints are more serious for landless and
small farm households in preharvest periods, so better developed credit markets would
benefit relatively these relatively poor households.
Section 3 also summarizes several other recent studies that adopt a dynamic approach to
labor supplies in rural areas of developing countries.
Ch. 43: Labor Markets in Developing Countries 2875

2.5. Household formation

Most of the empirical analysis of labor issues related to households in both developing and
developed economies takes as given the households that are observed in the data. But
households are not immutable permanent institutions. Instead they change over time due to
behavioral decisions related to marriage and separations of individuals and of families in
extended households and mortality.
Jacoby (1995) investigates polygyny, an institution that has been widespread in many
parts of the world at different times and subject to substantial study, but not to prior
modeling and empirical testing within explicit economic models that incorporate women' s
productivity and the effects of incomplete markets and gender divisions of labor and
ownership. While there have been speculations by Becker, Boserup and others about
the economic determinants of polygyny and how they would change with development,
such speculations previously were not systematically modeled and tested empirically.
Jacoby develops a structural model of the demand for wives from a lifecycle model of
marriage and agricultural production decisions - given incomplete labor and land markets,
female specialization in non-cash crops and male control of land - that permits the
identification of wealth versus price (substitution) effects in a framework in which
wives are explicitly recognized as an alternative to other forms of productive capital (so
the latter is endogenous, not given). He then estimates this model with micro panel data
from C6te d'Ivoire that permit control for measurement errors and unobserved hetero-
geneities across farms and in preferences by using the panel features of the data and
sample-cluster-level crop shares (assumed to be independent of individual farmer prefer-
ence heterogeneities) and lagged profit function residuals (with measurement errors that
are independent of those in current profits) as instruments. He first estimates agricultural
technology from profit functions (with control for fixed effects) and then uses the esti-
mated farm-specific technology to estimate the demand for wives (with control for hetero-
genous preferences for wives, endogenous expenditures, and measurement errors in profit
heterogeneity). He is sensitive in these estimates to the assumptions that he is making and
how they relate to his underlying modeling, and explores how robust are the estimates to
some alternative assumptions. His estimates suggest that geographical variation in crop-
ping patterns leads to variation in female productivity, which induces demands for differ-
ent numbers of wives. That is, such demands are greater where women are more
productive and therefore cheaper conditional on wealth in the presence of incomplete
land and labor markets. But with the process of development through expansion of
"male" export crops, although the demand for wives has increased due to greater wealth,
the substitution effect of higher "prices" for wives (due to their lesser productivity with
the changed crop composition and maintenance of gender specialization across crops) has
lessened the extent of rural polygyny.
Other recent studies explore how better earnings endowments attract more schooled-
wives in rural India (Behrman et al., 1995, 1997b), how assortative mating on preferences
regarding schooling causes biases in the usual estimates of the impact of parental school-
2876 J. R. Behrman

ing on child education (Foster, 1996) in Bangladesh, and how exogenous technological
change can affect own-farm labor and human capital returns and therefore induce breakups
of extended households with impact on measured income inequalities in rural India (Foster
and Rosenzweig, 1999). Some of these studies are discussed in more detail below in
Section 4.

3. Labor contracts, risks and incentives

Much of the development literature, as discussed in Section 2, has assumed extreme

possibilities regarding labor markets. One extreme is that there are rigid, institutional
determined, exogenous wages with surplus labor in rural areas. But this assumption
does not seem relevant even for densely-populated South Asia, the area which inspired
much of the literature on surplus labor. Rosenzweig (1984), for example, uses district level
and household level data to test various hypotheses about the functioning of rural labor
markets in India. Assuming that the land market is imperfect and land ownership is
exogenously fixed, and that geographical mobility across districts is unimportant, he
develops a competitive model of the Indian agricultural labor market with two types of
labor - male and female, and three types of households - landless households, households
with small plots, and households with large plots. Usin~ district level data, Rosenzweig
shows that wage rates vary systematically with variation in the factor availability, contrary
to the prediction of the exogenous wage hypothesis and consistent with the competitive
model.
But the other extreme assumption of complete markets also does not seem warranted,
either, a priori or on the basis of empirical tests of complete markets some of which are
summarized in Section 2 or on the basis of other evidence. For example, Ryan and
Ghodake (1984) report that in daily agricultural markets in villages in semi-arid tropical
India, male laborers were not able to obtain work in about a seventh of the days that they
were available. More fundamentally, the complete markets model cannot account for why
the family farm is the dominant organization in rural areas of developing countries, nor for
the existence of contractual arrangements such as sharecropping and the co-existence of
spot labor markets and longer-run implicit or explicit contracts.
In the 1970s and 1980s there was a rapid expansion in theoretical literature concerned
with how incomplete markets together with some of the special features of agricultural
production shape labor and land arrangements in rural economies. There has followed an
expansion of empirical tests of various aspects of these models.
These models generally emphasize one or the other of two principle themes, both of
which related to the basic multistage agricultural production technology.
(1) Risk: agriculture is risky, particular in poor environments, because of the importance
of fluctuations in weather (or other aspects of the state of nature, including disease and pest
virulence), a critical input, within the multistage agricultural production process. To
illustrate, the production process can be considered to have two stages, planting and
Ch. 43: Labor Markets in Developing Countries 2877

hm'vesting, 7 so that the production function in relation (2) becomes

Qp = QP(A,Lp, Fp, Wp), (9A)

Qh = Qh(A, Qp,Lh, Fh, Wh), (9B)

where the subscript p refers to the planting stage and the subscript h refers to the harvesting
stage, W is weather, and Qp is an intermediate output. Assume that the weather realizations
occur at the start of each stage. In the planting stage, labor, land and other inputs have to be
committed without knowledge of the weather that will be experienced in the harvest stage
(although possibly with knowledge of the distribution of harvest weather). In the absence
of insurance markets and capital investments to mitigate the impact of weather (e.g.,
irrigation systems, greenhouses), risk-adverse farmers might seek contractual labor
arrangements that substitute in part for absent insurance markets.
(2) Labor effort and incentive problems: labor consists of both time and effort, so the
production functions in (9A) and (9B) further become
Qp = QP(A,LpEp, Fp, Wp) (9C)

Qh ----Qh(A, Qp, LhEh, Fh, Wh), (9D)

where E is the average effort of agricultural workers so LE is the labor in efficiency units.
Both time and effort affect negatively the welfare of their suppliers, so the utility function
in (1) for the ith period (production stage) becomes

U i = U ( c i , li, el) , with U e < 0, (1A)

where e is average effort of household workers. Effort in some important agricultural tasks
(e.g., weeding and application of fertilizer and pesticides as opposed to harvesting) cannot
be costlessly or cheaply monitored because of the combination of production lags, imper-
fect observability by farmers of the intermediate product, the spatial dispersion and
heterogeneity in production conditions; there is not a distinct market for effort separate
from the market for labor time; and the time-wage alone insufficiently rewards effort.
Therefore contractual labor arrangements might be developed to create incentives for
laborers to provide effort. 8

3.1. Dominance of household farms in agriculture

Binswanger and Rosenzweig (1986) claim that the dominance of the household farm in
developing country agriculture is due, at least in part, to household enterprises being able

7The productionprocess may have other stages between planting and harvesting that may be important for
some purposes, but the basic points for the present discussion are illustrated by collapsing all of the pre-harvest
stages into the planting stage.
8Work effort and incentive problems also are claimed by some to be important in non-agriculturalsectors of
developing countries. See Section 5.1 below.
2878 ~ R. Beh~nan

to deal relatively well with incentives for efforts in difficult-to-monitor tasks (because
household members are the residual claimants on net revenues and have a long-run rela-
tion with the farm) and with risks (because of the relative effectiveness of family risk-
sharing and consumption-smoothing arrangements in the market context of developing
countries). 9 As the household size increases, however, monitoring becomes more difficult
and the incentives for effort decline because the residual (profit) is shared among more
household members. Therefore, there tends to be a limit to the size of effective agricultural
households. Risk-sharing options tend to increase with numbers, but all family members
do not have to be co-resident to exploit risk-sharing possibilities. In fact, to the extent that
there are locally correlated shocks, having family members dispersed through migration
and marriage is likely to increase the risk-sharing possibilities. Large landowners can limit
their hiring of wage workers (with the accompanying incentive problems) by renting their
land out to other households because tenancy arrangements can make other households
residual claimants on profits with the accompanying incentive effects. For a few crops,
however, (i) there are large scale economies and coordination problems in harvesting and
processing and/or (ii) there is need for sustained care across crops cycles. These techno-
logical features may lead to a plantation system with large numbers of hired workers.

3.2. Day versus longer-run labor contracts

Day (" spot" ) and longer-run ( "annual," "crop-cycle," "permanent," "attached servant" )
labor contracts co-exist in many rural areas of developing countries. Explanations have
been proposed for this phenomenon based on both risk and monitoring effort.
Because of the uncertainty of weather in the harvest stage (even if effort is monitorable
costlessly), both net buyers and sellers of labor face risks with regard to the harvest wage
in the two-stage production process described above. Both net sellers and net buyers of
labor, even if risk neutral, find it optimal to reduce exposure to risk by hedging with both
types of contracts. The more risk adverse are households, the more attractive are the
longer-run contracts. Therefore risk-adverse net labor selling households are willing to
accept crop-cycle contract wage rates below the expected value of wages from the spot day
market, and risk-adverse net labor buying households are willing to offer crop-cycle
contract wage rates above the expected value of wages from the spot day market. Because
poorer households tend to be more risk adverse (e.g., Binswanger, 1980) and net labor
selling households tend to be poorer (i.e., due to less land ownership), crop-cycle wage rate
contracts are likely to be below expected spot wage rates.
If planting period work effort cannot be monitored until the harvests are realized (even if
there is no risk), employers have incentives to hire crop-cycle workers only for the plant-
ing period and additional spot workers if necessary for the harvest (Eswaran and Kotwal,
1985a). Crop-cycle contract workers are induced to provide the fight level of effort in the

9Ben-Porath(1980) and Pollak (1985) also discuss advantagesof householdsand families and how they relate
to the completenessof various markets.
Ch. 43: Labor Markets in Developing Countries 2879

planting period because the expected worker welfare (inclusive of work effort) of the crop-
cycle contract exceeds the expected worker welfare of a series of spot contracts over the
crop cycle and because only if a worker is revealed to have devoted the fight level of effort
at harvest time will she/he be offered a crop-cycle contract for the next crop cycle. In this
model, in contrast to the case discussed in the previous paragraph, crop-cycle contract
workers are better off than spot contract workers and tend to continue across crop cycles in
their privileged labor market positions.
Annual or crop-cycle contracts are important in some rural areas (e.g., West Bengal, see
Bardhan, 1983), but not in others (Bell and Srinivasan, 1989). This may reflect differen-
tials in risks or in labor incentives across areas, differentials in alternative mechanisms for
sharing risks or for inducing efforts across areas, or differentials across areas in the
correlations between harvest-stage wage rates and gross harvest incomes because, if
this correlation is high, the net income risk of net labor buyers is not tied tightly to
wage rate risk.

3.3. Implications o f land contracts f o r labor

Sharecropping and other tenancy arrangements are common in many developing coun-
tries. There is a considerable literature on land tenancy arrangements in developing coun-
tries (see Bell, 1988; Binswanger et al., 1995 and the references therein), some of which
has implications for the allocation of and the returns to labor. If a tenant household
sharecrops land (with a marginal share s of gross output from the sharecropped land
Q~), works part-time in own production (Q°), and works part-time in the labor market at
wage W, the equilibrium condition for maximizing the household welfare in (1) subject to
the production relation in (2) and the household resource constraint is to allocate labor
among production on own land (L°), production on sharecropped land (LS), and the labor
market so that

PQOQ°/OL ° = SPQ3Q"/cgL" = W. (10)

If the marginal share crop rate s is less than one and there are no other contractual
stipulations, marginal returns to labor (and to all other variable inputs) are higher on
the sharecropped land than in own production or wage work. Theoretical modeling of
sharecropping to explain its existence in light this apparent inefficiency as compared with
fixed rent contracts is concerned, again, with dealing with risk and with incentives for
efforts.
If there are no insurance markets and no missing markets for effort, landlords and risk-
adverse sharecropping tenants share production risks, and can share them optimally if
optimal allocation of inputs on sharecropped land is enforceable. However, as Newbery
(1975) noted, risk reduction to the same degree can be obtained by the tenant household by
dividing household workers' time between a risky activity (own production) and a riskless
alternative (crop-cycle labor contract work). Therefore the risk sharing explanation of
sharecropping depends on the nature of alternative risk-reducing options.
2880 Z R. Behrman

If there is no risk but there is the double coincidence of no markets for managerial skills
of landlords and for work effort of tenants, sharecropping can provide incentives for
landlords to provide managerial skills and for tenants to provide work effort (Eswaran
and Kotwal, 1985b). In this case, a fixed rental contract is inferior because it does not elicit
managerial efforts of landlords. If tenants gain experience or new technologies appear for
which the landlords do not have managerial skills, the advantage of sharecropping over
fixed rents is likely to decline. If labor tasks become routinized, wage labor is likely to
become more attractive relative to sharecropping.
Empirical studies have largely focused on the question of whether there is inefficiency,
not what determines contract choices. Studies that compare input intensities for the same
farmer on different plots with different land contracts find that observed input intensities
are lower on sharecropped than on fixed rent or own plots (Bell, 1977; Shaban, 1987).
Subject to the qualification that these results are not due to unobserved factors such as
unobserved aspects of soil quality and water availability, they suggest that sharecropping
does lead to less inputs than the two alternatives as in relation (10). Bell and Sussangkarn
(1985) report that tenants with greater risk-sharing activities (e.g., receiving more transfer
payments, with more non-agricultural household workers, with greater landholdings) are
more likely to engage in riskier tenant contracts. But this association does not demonstrate
causality because it may just reflect endogenous choices so that in riskier environments
tenants choose a portfolio of means to cope with the risk. Bell and Srinivasan (1989) report
that in ten villages in the Indian Punjab owners and tenants are more likely to share in
allocation decisions under sharecropping than fixed-rent contracts, as in the Eswaran-
Kotwal model. But there is little information on longitudinal developments in tenancy
arrangements, how they are affected by experience of tenants and by changes in markets
and in technology, and what are the implications for labor allocations and labor returns.

3,4. Empirical studies of rural labor supplies and risk

Fafchamps (1993) considers sequential labor decisions under uncertainty for small farmers
in a developing country in order to attempt to reconcile expressed concern for possible
manpower shortages with low average labor inputs. His approach is related on a general
level to that used by Rust (1987) and Wolpin (1984, 1987), but explicit differences (e.g.,
finite horizon, non-stationarity, continuous decisions, the state space not discretized and
data available only on the final, not the intermediate, state of nature). He posits a simple
structural model with: three agricultural production stages (planting, weeding and harvest-
ing with the labor demands for the third of these proportional to product); nested constant
elasticity of substitution utility and production functions (with a priori equal utility effects
of leisure for the planting and weeding stages and moderate risk aversion in the former and
constant returns to scale in the latter) that are identical across households within a region;
households differing only with regard to land assets; no labor, land or intertemporal
markets; production shocks that are independent across the three stages; and rational
expectations concerning production shocks. Euler equations for this stochastic control
Ch. 43: Labor Markets in Developing Countries 2881

problem are invertible. The two decision variables are for planting labor and for weeding
labor. He presents FIML estimates for a 3-year panel of small farmers in Burkino Faso.
Vuong's non-nested model specification test indicates that the stochastic control estimates
dominate those that result from considering a deterministic control problem. The estimates
indicate that farmers are willing to supply considerable labor hours if there are expected
returns to doing so (i.e., the intertemporal elasticity of substitution of leisure and the
elasticity of substitution between consumption and leisure both are high), but that the
low level of labor effort commonly observed in this area is the result of both low labor
productivity in rain-fed agriculture in this environment and of farmers' awareness that, in
the absence of a labor market, overly ambitious initial production plans lead to seasonal
labor constraints on production. Therefore there may be considerable gains from technol-
ogies that permit farmers to have greater control (particularly over water) and greater
flexibility to respond to states of nature as they develop, as well as from development of
the labor market.
Rose (1995) explores the impact of risk on labor supplied to wage markets, also within a
seasonal, multi-stage framework with a planting stage and a harvesting stage in the latter
of which weather shocks are revealed. She distinguishes between ex ante and ex post
market labor supply responses of agricultural cultivating households to deal with risky
production on their own farms. The ex ante response is to the riskiness of weather distri-
butions in a particular location and the ex post response is to the weather realization. She
develops a two-stage model that incorporates both ex ante and ex post responses to wage
as well as crop income risk within a stochastic dynamic programming framework. She
then uses a 3-year national stratified random panel dataset for landed households in rural
India that is merged with another panel dataset containing characteristics of the districts in
which households live, including most importantly a 21-year series of rainfall data that is
used to compute a measure of the exogenous rainfall risk faced by farmers as well as other
indicators of bad weather. She specifies and estimates labor market earnings (related to
supply) and profit functions. The profit function is estimated with fixed and random effects
techniques, and Honore's procedure for estimating a fixed effects Tobit model is used to
estimate the labor earnings equation. The results indicate that: (1) profits increase and
labor supply falls in periods of good weather and high rainfall; (2) households facing
greater weather risk supply more labor to the market and receive lower profits than those in
less risky environments; (3) the effects of weather risk on production is reduced by
irrigation and by accessibility to banks and moneylenders; and (4) the availability of
non-agricultural employment increases labor income but also reduces profits by with-
drawing resources from production and exacerbates the responses of profits and labor
supplies to shocks and risk.
In another study Rose (1999b) examines another aspect of labor supplies related to risk,
in this case the sex of a baby. She investigates the impact of a "gender shock" (i.e., birth of
a girl) on time allocations in rural Indian households because of income (due to the need to
pay dowries in the marriage market) and substitution effects (due to the higher returns
from investing in sons than in daughters). She presents an intertemporal model that
2882 £ R. Beh~nan

generates predictions for these effects, conditional on whether or not the household is
constrained in the credit market. She presents careful empirical estimates that control for
unobserved fixed effects (e.g., preferences) and endogenous sex-related infant mortality.
Her empirical results indicate that the gender shock results in a decline in male leisure for
poorer households, but an increase for less poor households. For all households, women
work less following the birth of a son than following the birth of a daughter. These results
are consistent with the model predictions if poorer, but not less poor, households are
credit-constrained.

3.5. Empirical studies of imperfect information and labor markets"

In addition to imperfect insurance and effort markets, there are other types of incomplete-
ness related to rural labor markets that may affect the type of labor contracts and indeed
may be illuminated if the same individuals participate in different labor contracts. Imper-
fect information in labor markets with heterogeneous labor is widely conjectured to be a
common feature of such markets. Only recently has there been systematic empirical
research investigating its importance in the developing country context in a series of
studies by Foster and Rosenzweig. These studies investigate some phenomena about
which there have been considerable conjectures in markets for developed country labor
markets as well as for developing country labor markets, but for which the nature of
institutions, behaviors and data have made possible more satisfactory empirical investiga-
tions in the developing than in more developed country contexts.
Two little researched but important issues in the study of labor markets with hetero-
geneous workers is how employers select workers and how worker contributions are
rewarded if employers have imperfect information about work effort. Foster and Rosenz-
weig (1994a) present evidence that employers have imperfect information with regard to
the productivity of heterogenous workers by obtaining direct measures of the complete-
ness of employer information. Therefore they are able to consider the implications of such
information asymmetries and evaluate the extent to which casual rural labor markets in
developing countries exhibit these attributes using econometric tests from three large
micro datasets from rural areas of Asia. They find: (1) there is considerable variance in
productivity that is not associated with workers' characteristics observed by employers -
from one-fifth to two-thirds of the productivity variance; (2) there is adverse selection of
less productive workers into the time-wage sector of the labor market, with a 10% increase
in the unobserved component of a worker's productivity increasing the share of labor
market work time that the worker spends in piece-work by 6.6%; (3) employers discrimi-
nate statistically by paying time-wages that are 25-60% higher for men because the
distribution of productivity is higher for men than for women, but do not have taste
preferences regarding the gender of their employees (i.e., they pay the same for perceived
productivity independent of gender); (4) employers exhibit learning over time by obser-
ving workers, which exacerbates wage inequalities between men and women because the
latter have less labor market experience; and (5) nutrition affects productivity substantially
Ch. 43: Labor Markets in Developing Countries 2883

but is not rewarded in the time-labor market presumably because of problems in monitor-
ing productivity. This is the first paper to my knowledge to address critical labor market
questions regarding how employers with imperfect information about workers' character-
istics and productivities select which heterogenous workers to employ and how workers
are rewarded.
The inability of employers to fully observe worker effort has a central role in many
contractual models of the labor and land markets that are summarized above. While there
is fairly good evidence regarding the disincentive effects of easily observed material input
use associated with sharecropping (e.g., Bell, 1977; Shaban, 1987), there is no evidence
regarding the impact on (much-harder-to-observe) work effort under such incentive
systems prior to Foster and Rosenzweig (1994b). In this paper they develop a simple
multi-stage model in which worker health is affected by effort and by calorie intakes
through the energy balance condition and that permits the use of time-series information
on worker health and the inputs to worker health (i.e., calories) to measure the effort effects
of different labor payment schemes that award workers differentially. The Euler conditions
from this model are tested empirically, with careful attention to estimation and specifica-
tion issues (e.g., controlling for simultaneity and for individual worker effects using
within-round data on payments received under different payment schemes). The estimates
indicate that time-wage payment schemes as well as share-tenancy are associated with
substantial moral hazard. Workers supply about a third more effort when working on a
piece-rate scheme or on their own land than when working for time wages or under share
tenancy. Thus this paper provides the first systematic empirical evidence about the impor-
tance of moral hazard in labor markets and finds that it is substantial in the particular
empirical context considered.
How workers are matched with "jobs" is a fundamental issue in labor economics. A
number of matching mechanisms have been posited. Foster and Rosenzweig (1996a)
develop and estimate a Roy model of the allocation of heterogeneous workers to alter-
native productive tasks. This paper demonstrates that with data on piece rates and time
rates for the same workers it is possible to distinguish among three determinants of
worker-task allocations: (a) differences in the productivity of workers at different tasks;
(b) preferences of workers for different tasks; and (c) preferences of employers for differ-
ent types of workers. In the context of rural agriculture to which the methodology is
applied, the empirical application attempts to explain the extent to which these three
factors explain the over-representation of women in weeding activities relative to men.
Using data from the Philippines, it is found that the greater proportion of women allocated
to weeding is due their lower skill level, that women do not have a preference for weeding,
and that employers do not prefer women to perform weeding tasks. The results further
indicate that employers engage in statistical discrimination in the time-wage sector in that
they assign women to weeding because women are on average of lower skill than are men.
This paper thus provides an explanation for the ubiquitous gender specialization in tasks in
agricultural societies. It shows more generally that inferences about the relative impor-
tance of worker and employer preferences as determinants of the allocation of workers to
2884 J.R. Behrman

tasks cannot be made without a careful assessment of the importance of comparative

advantage and information asymmetries.

4. D e t e r m i n a n t s of and returns to h u m a n capital investments

Human resource investments are hypothesized to play a major role in labor market
outcomes in developing countries, as in developed economies. However, there are at
least two major differences. First, while there has been considerable emphasis on school-
ing as for developed economies, investments in human resources in health and particularly
in nutrition have received relatively much more emphasis for developing economies
because such investments are thought to have relatively high productivity effects in
very poor contexts. As is illustrated in Table 4, moreover, initial 1965 investments in
health and nutrition (as represented in this table by life expectancies at birth relative to
those predicted by per capita income in a cross-section for that year) have more predictive
power for economic growth over the next quarter century than do initial schooling invest-
ments (again, in 1965 and relative to those predicted by per capita income). Second, at
least some of the available studies of both the determinants and the effects of human
resources in developing countries place considerable emphasis on ways in which incom-
plete markets, such as are discussed in Sections 2 and 3, shape such investments.
On a general level the framework for considering the determinants of human capital
investments in developing countries is the same as that used for developed economies.
Differences relate to the differences such as those in the completeness of markets and in
the roles of households. Many of the essential features are summarized in B e c k e r ' s (1967)
W o y t i n s k y Lecture.
Private maximizing behavior leads to human resource investments at the level at which
the private present discounted marginal benefit of the investment equals the private present
discounted marginal costs of the investment. Fig. 2 A provides an illustration for one

Table 4
Estimates of the associations of initial 1965 human resources relative to the levels predicted by cross-country
regressions with subsequent economic growth for the 1965-1990 quarter century~

Dependent variable Initial schooling Initial life Constant R2 F N

real per capita GDP expectancy
annual growth 1965-1990

Row l 0.39 (4.0) 1.8 (9.1) 0.15 16.3 85

Row 2 0.14 (4.7) 1.7 (8.2) 0.18 23.3 96
Row 3 0.15 (1.1) 0.10 (2.6) 1.7 (8.9) 0.21 12.1 85

"t statistics are in parentheses to right of point estimates. The initial human resource positions are the actual
values minus the values predicted by a cross-country regression on a polynomial in per capita income for 1965.
Schooling is the expected schooling for a synthetic cohort. For more details see Behrman (1994b).
Ch. 43." Labor Markets in Developing Countries 2885

MarginalBenefits, (A) MarginalBenefits, (B)

MarginalCosts MarginalCosts

Benestfiaarginal
~l~~M
(Solid)~

R*
~ ' - ~ kMarginal Benefits
Marginal I Marginal ~ '~1 "dDashed)
Costs I Costs I ~
I I i~ ".
I
I i; 'i, \ \ \ \.
I
I
I Human Resource l I HumanReseurce
I Investments I I Investments
I
H* H* H**
MarginalBenefits,
MarginalCosts
(c)

Marginal
Costs (Solid)

R*

I
R*' i....l..nn.mmm~ MarginalCosts
(Dashed)
|
|
|
|
Human Resource
B Investments
H*

Fig. 2. (A) Private marginal benefits and private marginal costs of human resource investments. (B) Private
marginal benefits and private marginal costs of human resource investments, with higher (dashed) and lower
(solid) marginal benefits. (C) Private marginal benefits and private marginal costs of human resource investments,
with higher (solid) and lower (dashed) marginal costs.

individual. The m a r g i n a l private benefit curve depends on the expected private gains (e.g.,
in wages/salaries i n labor markets) due to h u m a n capital investments. The m a r g i n a l
private benefit curve is d o w n w a r d - s l o p i n g because of d i m i n i s h i n g returns to h u m a n capital
2886 ,1. R. B e h r m a n

investments. ~0 The marginal private cost may increase with human capital investments
because of the increasing opportunity costs of more time devoted to such investments
(especially for schooling and training) and because of the increasing marginal private costs
of borrowing on financial markets. For a human resource investment at level H*, the
private returns net of costs are maximized.
If the marginal private benefit curve is higher for every level of human capital invest-
ment as for the dashed line in Fig. 2B, all else equal, the equilibrium human capital
investment (H**) and the equilibrium marginal private benefit (r**) both are greater.
The marginal private benefit curve m a y be higher for one of two otherwise identical
individuals except for the difference noted below because one individual (or whomever
is investing in that individual, such as the parents of young children): 11 (1) has greater
endowments (e,g., more ability and drive) that are rewarded in schooling and in post-
schooling labor markets; 12 (2) has lower discount rates so that the future benefits of human
capital investments have greater value at the time of the decision whether or not to invest;
(3) has human capital investments options of higher quality (e.g., access to higher quality
public schools or public health services) so that the marginal private benefits for a given
level of investment are higher, and the equilibrium investments greater; 13 (4) has better
health and a longer expected life due to complementary investments, so that the post-
investment period in which that individual reaps the returns to the investment is greater
and therefore the expected returns greater; (5) has greater marginal private benefits to a
given level o f such investments because of more extensive labor markets or labor market
discrimination that favors that individual due to gender, race, language, family, village, or
ethnic group; (6) has returns to human resources investments that are obtained more by the
investor or the relevant decision maker (e.g., if traditional gender roles dictate that children
of one sex, but not the other, provide old-age support for their parents, parental incentives
may be greater to invest in children who are likely to provide such support unless there is
an exactly compensating adjustment elsewhere such as in marriage markets); (7) has lower
discount rates, given risk aversion, because of better means for coping with risks through
insurance markets, public safety nets or whatever; (8) has greater marginal private benefits

J0Diminishing marginal retm'ns might be expected (at least at sufficientlyhigh investments levels) because of
fixed genetic endowments and because human capital investments take time (such as schooling and training) and
greater investments imply greater lags in obtaining the returns and a shorter post-investment period in which to
reap the returns from the investment.
lJ For some of these comparisons (e.g., the last three) the otherwise identical individuals would have to live in
different economies.
~2This means that to obtain an estimate of the impact of human capital investments on some outcome, one
cannot just consider the associationbetween the human capital investment and the outcome (i.e., the association
between years of schooling and wage rates), but one must control for the endowments underlying the different
human capital investments.
13if the investor (or the investor's family) must pay for greater quality, investment does not necessarily
increase with a higher quality option. What happens to the equilibrium investment depends upon where tile
marginal private cost curve for the higher quality option is in addition to the location of the marginal private
benefit curve.
Ch. 43: Labor Markets in Developing Countries 2887

to a given level of investment because of being in a more dynamic economy in which the
returns to such investments are greater; (9) has greater marginal private benefits to a given
level of such investments because of greater externalities from the human capital invest-
ments of others in the same labor market; or (10) lives in a more stable economy so that the
discount rate for future returns is lower and thus the marginal private benefit of future
returns greater. ~4
If the marginal private cost is lower for every level of human capital investment as for
the dashed line in Fig. 2C, ceteris paribus, the equilibrium human capital investment
(H***) is greater, with the marginal private benefit (r***) at the higher investment
level lower. The marginal private cost might be lower for a number of possible reasons.
Compare two otherwise identical individuals except that one individual: (1) has lower
private cost access to educational and health services related to such investments because
of closer proximity to such services or lesser user charges; (2) has less opportunity costs
for time used for such investments (e.g., due to gender specialization in household and
farm tasks performed by children); (3) faces lower utility costs of such investments
because of cultural norms that favor some activities associated with such investments
more for some individuals than for others (e.g., in some societies, it is not thought desir-
able that girls past puberty intermingle with males outside of the family in transit to school
or in school so that the preference costs of schooling are lower for boys than for girls); 15 or
(4) is from a household with greater access to credit because of greater wealth or status or
better connections or better capital and insurance markets.
This simple framework systematizes four critical general points for identifying the
determinants of human capital investments and what is the impact of human capital
investments on outcomes of interest - in the present context labor productivity. 16 First,
the determinants and the expected outcomes are interrelated, as in any investment deci-
sion. Therefore the determinants depend not only on the conditions at the time of the
investment, but on expectations regarding the context in which the returns from the
investment will be reaped. There also m a y be interactions between the various human
resource investments, for example with better health and nutrition increasing the expected
returns to schooling. Further, to identify the impact of human capital investments on a
particular outcome, it is important to control for individual, family, and community
characteristics that reflect the conditions under which the investments were made. Other-
wise the estimated effect includes not only the impact of the human capital investment, but
also the effects of individual, family, and community characteristics that directly affect the
outcome of interest and are correlated with the human capital investment because they

14Some of these possibilities tie directly into the new economic growth models that have received a lot of
attention in the past decade (e.g., the first is consistent with Stokey's, (1991) emphasis on the heterogeneity of
individuals, the seventh with a product composition more conducive to learning-by-doing as in Lucas (1988) and
Stokey (1991), and the eighth with the externalities broadly emphasized in this literature).
/5 For this case the marginal utilities of marginal private benefits and costs are equated.
16And for understanding under what conditions there may be efficiencyreasons for governments or for private
firms to subsidize human resource investments (Section 6.1).
2888 J. R. Behrman

partly determine that investments. Second, empirically observed returns to human capital
investments are for a given macro economic, market, policy, and regulatory environment.
The actual returns may change substantially with changes in that environment, such as
those associated with changing from administrated to market prices, opening up an econ-
omy more to international markets, establishing greater macro balance, eliminating regu-
lations on migration, or lessening discrimination in labor markets. Third, the marginal
private benefits of human capital investments in a particular individual may differ depend-
ing upon the point of view from which they are evaluated: (i) there may be externalities
such as those emphasized in the "new neoclassical growth models" or capital/insurance
market imperfections so that the social returns differ from the private returns; (ii) there
may be a difference between who makes the investment decision (e.g., parents) and in
whom the investment is made (e.g., children) which may result in gender (or birth-order)
differentials in incentives for investments in children given traditional gender (birth-order)
roles in old-age care for parent; and (iii) some forms of human capital investment may
have returns broadly throughout the economy and others may have returns only in specific
activities or productive units. Fourth, if the marginal private benefits equal the marginal
social benefits and if the marginal private costs equal the marginal social costs, optimizing
investments in human capital by private investors are socially efficient.

4.1. Determinants of health and nutrition investments

There is a substantial literature on the determinants of health and nutrition by behaviors of

households and other entities in developing countries (see Behrman and Deolalikar, 1988;
Jimenez, 1995; Strauss and Thomas, 1995; World Bank, 1995). I limit attention here to a
small subset of those studies, those that are most related to labor markets through focusing
on the role of income and on expectations regarding labor market outcomes.

4.1.1. Household income

If all relevant markets were complete and the only difference between two individuals
were that they came from households with different incomes, there would be no differ-
ences in human capital investments in the two individuals. However, it is widely believed
that there are associations between human capital investments in individuals in developing
(as well as developed) countries and income. This may reflect that such investments have
some consumption components and/or that income is associated with some of the deter-
minants of human capital investments discussed with respect to Fig. 2, such as ability,
discount rates, access to capital and insurance or other markets, and access to public
services.
Empirical studies of household behavior determining health and nutrition investments
in developing countries generally have included income indicators among the right-side
variables, usually with reference to credit market imperfections if any explicit rationale is
given. In the past decade the greatest emphasis related to income in this literature has been
Ch. 43: Labor Markets in Developing Countries 2889

on (1) the magnitude of income-nutrition associations and (2) whether there is complete
income pooling.
The magnitude of income-nutrition associations are of interest because on the order of
magnitude of a billion people in the developing world are thought to be malnourished
(which is widely viewed as undesirable in itself in addition to any productivity effects) and
some influential observers have argued that the most effective way to eliminate malnour-
ishment is to increase income (e.g., World Bank, 1981). Engel curves for food purchases
for poor people typically indicate income elasticities of the magnitude of 0.6-0.8, from
which many observers concluded that nutrients consumed by members of poor households
would increase by about 6-8% for every 10% increase in poor households' income. (Some
have argued that for very poor households the nutrient elasticities with respect to income
would be higher, greater than one, e.g., see Lipton, 1983.) Inferences from such empirical
estimates underlay widespread optimism about reductions in malnutrition with income
increases. A revisionist position emerged in the past decade, however, that questioned
whether nutrient (in a particular, calorie) income elasticities were nearly this large based
on claims that previous estimates had overstated calorie-income associations because of
ignoring (a) the distinction between household food purchased/produced and food
consumed by household members that may be strongly associated with income due to
provision of food to household employees, mendicants and animals and wastage, (b)
measurement error that biased the estimated associations upwards (e.g., regressing food
expenditures on total expenditures), and (c) intra-food group substitution towards more
expensive nutrients associated with income (see Behrman and Deolalikar, 1987; Bouis and
Haddad, 1992; Alderman, 1993; Bouis, 1994; Subramanian and Deaton, 1996).
At this point the prevalent view seems to be the calorie-income associations are some-
where between those implied by the previous conventional wisdom and the revisionists,
suggesting a moderate role for income increases in lessening malnutrition. But it is striking
that for the most part this fairly extensive literature does not place the investigation of the
determinants (and impact) of nutrients within the context of the particular market config-
urations faced in developing countries, nor is there much attention to timing issues regard-
ing income receipts and food expenditures. A recent at least partial exception to this
statement is Behrman et al. (1997a), who investigate calorie demands and the impact
calorie consumption on farm profits in rural Pakistan. They positive a two-stage produc-
tion process as in relations (9C) and (9D) above in which planting season labor markets do
not reward greater efforts due to better-nourished workers because of monitoring problems
and credit markets do not permit poor households to borrow for nutrition investments in
the planting stage the returns from which occur at harvest time. They find that for poor
agricultural households (defined by small landholdings), the income elasticity for plant-
ing-stage calories is one, in contrast to a value of about zero in the harvest stage when
nutrients are much cheaper and harvest piece worker rates directly reward better current
nutrition. Thus placing the nutrient investment within the particular context of incomplete
markets and multi-stage agricultural production leads to different insights regarding the
nature of this human resource demand and its relation to incomplete markets.
2890 J. R. Behrman

The income pooling question pertains to whether households effectively pool individual
incomes or whether it matters which members of the households control income. This
question originally arose in regard to bargaining models for intrahousehold allocations in
developed economies (Manser and Brown, 1980; McElroy and Horney, 1981). But, as
emphasized by McElroy (1990), the majority of the efforts to provide empirical tests of
whether incomes are pooled by household members have been for developing countries,
often with emphasis on investments in health and nutrition (e.g., Schultz, 1990b; Thomas,
1990, 1993, 1994; Strauss and Thomas, 1995; Haddad et al., 1996 and the references
therein). Many commentators summarize these studies to imply that (a) income is not
pooled in developing countries and (b) resources under control of women has much greater
impact on human capital investments in health and nutrition than do resources under
control of men. While conventional wisdom has been shaped considerably by these
studies, I find them less persuasive than do many because they do not control for unob-
served abilities and preferences that arguably are correlated with the indicators of indivi-
dual control over resources that are used (see Behrman, 1997a for further discussion).
These studies like those related to income and health/nutrition investments, moreover,
generally do not place the analysis very well into the specific market and institutional
contexts of developing economies.

4.1.2. Expected labor market returns

One important implication of standard models of human capital investments, as empha-
sized above with respect to Fig. 2, is that such investments are predicated in part on their
expected returns. Most of the empirical literature on such investments in developing
countries (and that on developed economies) does not directly incorporate this possibility
because of the problems in representing such expectations. I now consider briefly two
studies on health and nutrition in developing countries that do attempt to represent such
expectations.
Rosenzweig and Schultz (1982) argue that differential child mortality rates - w i t h
mortality being the equivalent to very poor health and nourishment - by sex across
India reflect differential expected labor market returns to investing in the human resources
of boys versus girls. They develop a simple model consistent with this argument, and then
present estimates of the model, using current adult labor force experience to represent the
experience expected for current children (arguing that such a representation is good for the
period that they consider because labor markets were relatively stable in that era). Their
estimates suggest that differential boy-girl mortality patterns in different parts of India are
consistent with the hypothesis that households invest in the children in whom the returns
are greatest (i.e., reinforce endowment differentials in light of market opportunities).
Pitt et al. (1990) develop a model that incorporates linkages among nutrition, labor-
market productivity, health heterogeneity, and the intrahousehold distribution of food and
work activities in a subsistence economy. A household is assumed to have individuals in m
classes (defined by age and sex so that within a class the health and wage production
functions are the same for all members of the household). The household maximizes its
Ch. 43: Labor Markets in Developing Countries 2891

consensus preference function that is defined over the health, food consumption, and work
effort of each individual (with positive effects of health and food consumption and nega-
tive effects of work effort) subject to (i) a budget constraint that posits that income from
labor and other sources must be greater than or equal to expenditures on food and other
consumption and (ii) production functions for health and wages for each class:

tI)k = hk(Nj, Ej, Gj), (11)

wkJ = wk(Hj, Ej), (12)

where H!J is the health of the jth individual in the kth class, Nj is the nutrient or food
consumption of the jth individual, E~ is the work effort of the jth individual, Gi is the health
endowment of the jth individual that is observed by household members but not by social
scientists, and Wjk is the wage rate for the jth individual in the kth class. 17 Nutrients and
endowments are posited to have a positive effect on health, health a positive effect on wage
rates, and effort a negative effect on health and a positive effect on wage rates. Health is
assumed to increase the marginal product of effort in producing wages, with all the
endowments effects working through health.
The first-order conditions indicate that the marginal cost of allocating nutrients at the
margin to an individual is lower the greater the extent to which that person's health
improves with more nutrition and that person's wage increases with better health. If
different classes of individuals participate in different work, as appears widely to be the
case in developing countries with respect to gender, and the wage effects of health vary
across types of work, the marginal costs of food allocated to different classes of individuals
may vary substantially. Within a class the distributions of food and work effort across
individuals depend on the distribution of endowments among those individuals. Compen-
sation or reinforcement can be examined by investigating the first derivative of health with
respect to endowments, which includes the partial effects on health through both work
effort and nutrient intakes. In the case in which endowments enter additively in the health
production relation, there is compensation (reinforcement) if the sum of these two partial
effects is negative (positive). The cross effect o f j ' s endowment on i's nutrient consump-
tion is more negative if the household preference function is non-linear with the consump-
tion of i and j as substitutes the stronger is the relation between health and effort
productivity for j.
To explore empirically whether there is compensation or reinforcement, estimates of the
endowments first must be obtained. To do so, the health production function is estimated
directly and, based on the parameter estimates and the actual nutrients consumed and work
effort expended by each individual, individual-specific endowments are calculated. There
are two problems that must be dealt with in this "residual" endowment method. First,
because endowments are not observed by social scientists and they influence household

17 W o r k time is a s s u m e d to be the same for all individuals b e c a u s e there are not data on time allocations a n d
b e c a u s e casual observations suggest that there is very little leisure in the s a m p l e area.
2892 J. R. Behrman

allocations, OLS estimates of the health production technology are not consistent. They
therefore use as instruments "food prices, labor-market variables reflecting labor demand,
and exogenous components of income" under the assumption that such variables "deter-
mine resource allocations but do not directly affect health status, given food and activity
levels." (Pitt et al., 1990, p. 1145) Second, the residually-derived endowments are likely to
be measured with systematic error because of random measurement error in the observed
inputs into the health production function such as individual nutrients, which carry over to
cause errors in the estimated endowments that in turn causes biases in the estimated impact
of the endowments on allocated variables. These biases tend to make households appear
more compensatory than they really are. t8 To obtain consistent estimates Pitt et al. (1990)
use instrumental variables in the form of estimated health endowments for weight-for-
height, m i d - a r m circumference, and skinfold thickness from other survey rounds than the
one for which the allocation estimate is being m a d e under the assumption that the period-
specific measurement errors are not correlated across time periods.
The data requirements for this study are considerable: individual specific observations
on nutrient intakes, health outcomes, and work effort; sufficient cross-sectional variation in
exogenous instruments needed for consistent estimation of the health production function;
and repeated observations on individuals to purge estimated endowments of measurement
errors. They use data from the 1981-1982 Bangladesh Rural Nutrition Survey of 385
households in 15 villages and F o o d and Agricultural Organization/World Health Organi-
zation ( F A O / W H O ) classifications of the 14 occupations provided in the data as "very
active" and "exceptionally active" to characterize higher than normal work effort and
control for whether women were lactating or pregnant in the sample period to control for
non-work nutrient use. Estimates of the health production function for weight-for-height
suggest that the impact of calories is understated and the signs of the coefficients of the
work effort variables wrong if OLS is used instead of simultaneous estimators. Then the
residual endowments obtained from the consistently-estimated health production technol-
ogy were used for the households with longitudinal data to obtain consistent estimates of
the impact o f individual endowments on individual nutrients. These estimates suggest
reinforcement in the sense that individuals with better endowments receive more nutrients
once there is control for the measurement error problem noted above (which, if not
controlled, leads to estimates that are opposite in sign, suggesting compensation); these
effects are about ten times larger for males than for females, which is consistent with their
model, given that their data indicate that women do not participate in energy-intensive
activities. Within-household estimates by gender with age-specific endowment effects
suggest that reinforcement is significant for males 12 years of age or older and for both
males and females in the 6 - 1 2 year age range, but that compensation m a y occur for those

~sPitt et al. (1990) show that, if the true impact of such endowments on nutrients is positive, the estimated
impact will be downward biased. But if the true impact is negative, the classical measurement error bias is
towards zero (and therefore positive) while the bias due to the correlation of the estimated endowment with the
measurement error in nutrients is negative, so the overall effect is indeterminate.
Ch. 43: Labor Markets in Developing Countries 2893

under 6 years of age of both sexes (although the standard errors are large); for females 12
years of age or older the sign of the coefficient is positive but the magnitude is very small
and the standard error very large. Next, they explore what the impact of (instrumented)
endowments is on household income and on participating in an exceptionally active
occupation (in the absence of data on individual wage rates or earnings). Their estimates
suggest that there is a pecuniary return to health and effort, that adult males with higher
endowments are more likely to undertake exceptionally energy-intensive work, and that
adult female health endowments are relatively unimportant (in comparison with those for
adult males) in determining activity choices or household income. Finally, the net effect of
a change in own endowments on own health are calculated from the estimated health
production functions and the estimated endowment effects on the nutrient and work effort
variables in those production functions; the elasticities of own health with respect to own
endowments are 0.88 for adult males and 0.97 for adult females. Thus, on net Bangladeshi
households exhibit compensatory behavior with respect to adult health endowments so
that these elasticities are less than one, with adult males being " t a x e d " to the benefit of
other household members more than females. Therefore, by incorporating the expected
impact of nutrient investments in an integrated manner with the estimates of that impact,
not only do they c o m e to fuller understanding of how nutrient investments work but also to
a different understanding of the nature of intrahousehold allocations of nutrients, a ques-
tion on which there has been considerable debate. 19

4.2. Productivity impact o f health and nutrition

In poor countries many people, including workers and students, have poor health or
nutrition. There have been many conjectures that such poor health and nutrition has
negative effects on productivities in the labor force and in forming human capital to be
used subsequently in the labor force (e.g., in schooling success).

4.2.1. Nutrition-based efficiency model

The nutrition-based efficiency model of Leibenstein (1957), Mazumdar (1959), Mirrlees
(1975) and Stiglitz (1976) systematizes the possible impact of nutrition on productivity as

~9Limitations in the data mean that some qualifications are appropriate. The use of the instruments for the
health production function to obtain consistent estimates of the production function coefficients and of the
residual endowments depends upon the assumption that there are no allocated inputs into the production of
health that are not observed, a strong assumption. If women's time in household production (not observed), for
example, has an effect on health, instrumented nutrients and work effort may be representing in part the impact of
women's time allocations since such allocations presumably respond to the same set of exogenous instruments.
The assumption that measurement errors in nutrient intakes are not correlated across periods may be strong if the
intrahousehold allocation of food was altered to favor certain groups identified by age or sex because of the
presence of outside dietary investigators. The measure of work effort based on 14 occupational categories,finally,
is quite crude, ignores what probably are substantially variations within such categories, and may impart a gender
bias since some have claimed that the FAO/WHO estimates understate energy used in various household
activities performed primarily by females.
2894 J. R. Behrman

jJ
j/

Fig. 3. Effortnutritionlocus with optimal work effort at M.

a possible explanation of downward rigidity of rural wages in poor labor markets that may
be associated with surplus labor. This, thus, is an alternative theoretical explanation for
surplus labor to the household models discussed in Section 2.
Central to the nutrition-based efficiency model is a modified agricultural production
function that is identical to that in relation (2) except the labor time is multiplied by
efficiency per unit time (E) as in relations (9C,D) which in turn depends on nutrients
consumed and therefore consumption (c):

Q = Q(A, L E ( c ) , F). (13)

The efficiency per unit time as a function of c is zero until some m i n i m u m consumption
level and then increasing in c at a declining rate over the relevant range as in Fig. 3. 2o
Under the assumptions that there is an infinitely elastic supply of workers at the wage W,
that employers can appropriate all the additional product that workers with better nutrition
produce, that c depends only on W so that E ( c ) = E ( W ) and that the farm's land area is
fixed, profit maximization implies that employers select what wage they pay and how
much labor time they hire to maximize P Q Q ( A , L E ( W ) , F ) - W L - PFF. The first-order
conditions for this maximization imply

W*/E = 1/Ew, (14)

where * indicates the maximizing choice. Thus the efficiency wage, which minimizes the

20There is some difference in the literature regarding what this functionalform is at very low levels of c.
Mirrlees (1975) and Stiglitz(1976) posit a non-convexitywith initiallyincreasingmarginaleffects of c (i.e., with
Ecc > 0) and then declining marginal effects (with E~c < 0). Bliss and Stern (1978) and Dasgupta and Ray
(1986a,b) posit a discontinuityat some minimumconsumptionlevel from no effect to decreasingmarginaleffects
(with Ecc < 0). The non-convexityin the former case leads to some peculiarities, such as unequal distribution
among household members may be optimal even if the family welfare functionis additive in individualfamily
member's utilities. With both forms the convex region is what is relevant for the basic possible explanationof
downwardly-rigidwages and unemployment.
Ch, 43: Labor Markets in Developing Countries 2895

cost per level of effort, is chosen so that the average cost per unit of effort just equals the
marginal cost per unit effort.
In the simplest form of this model in which workers have no alternative income sources,
there would be no savings, no dependants, and no unemployed workers. If workers have
alternative sources of consumption (i.e., full income in relation 3 includes not only full
earnings but also positive net profits from own farming or positive other income) and if
employers are informed about workers' other income sources and their family composi-
tion, time wages vary depending on workers' alternative income sources and family
composition. Bliss and Stern (1978) show that if the employer is a monopolist, the
employer pays out time wages so that consumption of workers is equalized, which implies
lower time wages for landed than for landless workers (so that the former are hired before
the latter). Dasgupta and Ray (1986, 1987) show that if there is perfect competition, each
worker receives the same payment per unit of work effort, so those with higher levels of
alternative consumption sources who supply more effort per time unit receive higher
wages per time unit. Thus workers with more land (or other alternative consumption
sources) and fewer dependants receive higher time wages and workers with limited
enough alternative consumption sources and numerous enough dependants so that the
time wage is at or below the efficiency wage may be unemployed (in which case redis-
tribution of assets towards these workers may increase output). Thus, predictions about
patterns of time wages and employment among potential workers with differing alternative
income sources and number of dependants depends on the labor market structure.

4.2.2. Empirical estimates of health~nutrition effects on labor productivity

Rosenzweig (1988a) reviews the evidence then available regarding the predictions of (a)
coexistence of high unemployment rates and downwardly-rigid wages, (b) wage diversity
and (c) direct productivity effects of nutrition in rural areas of developing countries. He
concludes that there is no support for (a) and (b) and that, prior to Strauss's (1986) study
(see below) there was no persuasive evidence on (c) because it was not clear from such
studies whether higher nutrient consumption caused higher productivity or higher income
associated with higher productivity caused greater food and nutrient consumption. More
recent studies are consistent with his summary regarding (a) and (b) (e.g., Richards, 1994
on Egypt).
Strauss (1986) is the first study of which I am aware that investigates the impact of
nutrition on productivity with control for their possible simultaneous determination. He
estimates the effect of a family' s average intake of calories per adult consumer-equivalent
on the productivity of on-farm family labor in Sierra Leonean agriculture. One of his
production function inputs is "effective family labor," which is a non-linear function of
actual on-farm family labor hours and the average availability of calories per consumer-
equivalent in the household. Effective family labor has a statistically significant coefficient
estimate in the agricultural production function, and effective family labor increases
significantly, at a diminishing rate, with available calories, calculated on a per consu-
mer-equivalent basis. Strauss estimates the output elasticity of available calories on a per
2896 J. R. Behrman

consumer-equivalent basis to be 0.33 at the sample mean, 0.49 at 1500 calories a day, and
0.12 at 4500 calories a day. His estimates imply that an increase in caloric intake results in
a substantial increase in the efficiency of an hour of labor (e.g., a laborer who consumes
4500 calories a day is 20% more productive than one who consumes 3000 calories a day).
These findings are robust to several alternative specifications and to changes in the instru-
ments used for his first-stage calorie estimates. Strauss also notes that labor market wages
are not significantly related to caloric consumption, which is consistent with markets being
incomplete because nutrient-related productivity is difficult to observe.
Subsequent to Strauss's study, there have been a mamber of other empirical investiga-
tions that attempt to investigate the impact of nutrition and health on output/income/profits
or wages primarily in rural areas of developing countries 21 with some effort to control for
the endogenous choices that led to the observed nutrition and health states (Table 5). These
studies use some combination of nutrient and health indicators that refer to different time
periods: (i) calories, which refer to recent food consumption and energy availability, (ii)
weight for height or BMI (body mass index, weight/height2), each of which is a common
measure of short-run nutrition and health status, and (iii) height, which is a common
measure of long-run nutrition and health status. All of these studies attempt to control
for the possible endogeneity or omitted variable bias for the first two indicators through
using instrumental variable and/or fixed effects estimators. Most of the studies assume that
height can be treated as independent in such estimates - i.e., that height is independent of
any unobserved characteristics in the disturbance term that affect labor productivity -
although a few studies control for such possibilities. 22 The studies vary in their coverage
of the three groups of nutrition and health indicators. If the true specification includes all
three of these indicators, if all three depend on some common characteristics (e.g., genetic
health endowments) or if one is an input into the production of another 23 or if the same
anthropometric indicator is used to construct more than one of these three groups of
indicators (i.e., height is used for both b and c), and if all three are not included (or
controlled for) in the specification, the estimates of the included indicators may be biased
(despite the use of instruments) because they are representing in part the incorrectly
excluded nutrition/health indicators.
Though the estimates are somewhat mixed, for most part they suggest significant effects
of nutrition/health on agricultural production, net profits or wages and some variations by
gender with effects if anything more likely to be significant or larger for males than for

22 Thomasand Strauss (1997) is the only studyin this table with estimateson urban areas. Satinand Alderman
(1988) report that estimates that they made for urban areas were not very robust and therefore they do not give
these estimates, but only those for rural areas.
22 Aldermanet al. (1996b) use instrumentedheight with instrumentsfrom parentalhousehold characteristics.
Deolalikm" (1988) presents household and individualfixed effects estimates that control for height, but do not
permit estimation of the effect of height. Some of Haddad and Bouis' (1991) alternative estimates also use
individualfixed effects.
23For example, BMI productionfunctionswith calories included amongthe inputs are presented in Pitt et al.
(1990) and in Foster and Rosenzweig (1994b), both of which are discussed above.
Ch. 43: Labor Markets in Developing Countries 2897

v "~ v

~ qo
~ o .--; c , i
r I I

0 ~ C~ 0
I I I

I I I I

I I

©
J
©

~,--&
o

~ ~
~'~
~ ~~ X V A ~
2898 J. R. B e h r m a n

2~
z
~o
. ~ =~
aa

~7 .=.

~,0 o=
~~ ~ eo

o o ~ e,i o ~ e.i c~ c5
I I
~.~ 2~
on ".~

"a
~0.~ .~ ©
e,t) 0 0 ~E

I I

~.~ ~
Ns ~ o ©
© 0 0 ~-
~.~
r..) I

N~2
o
o

d a

I
g~. i
o

o
o
>~
?, © m "0 0~ ,a= ,s=
Ch. 43." Labor Markets in Developing Countries 2899

Cz
"~ "~
~ ~ oA
.~ o . ~

r~
•} ~o ~

~ ~.~ ~ ~o "~

•~ ,.~

0 0 0

~ ~ ~'~

. e~ °o ~~ ~ 0

~
~ ~ -~ ~ .~
o e~

~ rm

~.~ ~ ~ •
0
e~

~ ~~ o

8
U
~.~.- . ~ ~ ~ =.~ ~.g

e~x
2900 J. R. Behrman

females presumably because of gender divisions of labor with males more concentrated in
tasks in which strength and bursts of energy are important.
These estimates also shed some light on possible intormation questions relating to
nutrition and health that tie in to the basic multistage agricultural production technology
that is discussed in Section 3. If the production process can be considered to have two
stages, planting and harvesting, 24 the production functions are (9C) and (9D). Assume that
the weather realizations occur at the start of each stage. In the planting stage labor, land
and other inputs have to be committed without knowledge of the weather that will be
experienced in the harvest stage (although probably with knowledge of the distribution of
harvest weather). As discussed in Section 3, effort in some important agricultural tasks
(e.g., weeding and application of fertilizer and pesticides as opposed to harvesting) cannot
be costlessly or cheaply monitored because of the combination of production lags, imper-
fect observability by farmers of the intermediate product, and the spatial dispersion and
heterogeneity in production conditions even on one farm; there is not a distinct market for
effort separate from the market for labor time; and the time-wage alone insufficiently
rewards effort.
The studies that are summarized in Table 5 are generally consistent with nutrition
having short term effects on labor effort and productivity that are not easily observed
during the agricultural planting season. 25 For five of the six agricultural production func-
tion/net profit estimates and explicit piece-rate wage estimates, there are significant posi-
tive effects of calories. Foster and Rosenzweig (1994b) further suggest that in one of the
two other cases in which there is a significant positive effect - that for peak season rural
Indian labor in Behrman and Deolalikar (1989b) - harvest piece rate wages dominate so,
even though the piece rate wages are not separated out explicitly, this estimate reflects the
same phenomenon. For only one of the other seven estimates of rural market wage effects,
in contrast, are the estimates of calories significantly positive.
In some of the studies in the top part of Table 5, moreover, the multistage nature of
production is central to the analysis. For example the piece rate wages are for the harvest
stage during which monitoring of harvesting productivity is relatively costless and the
estimated significant impact of calories on net profits in Pakistan is for planting stage
calories on harvest profits, with no significant impact of harvest calories on harvest profits,

24The productionprocess may have other important stages between planting and hm:vesting,but the basic
points are illustratedby collapsingall of the pre-harvest stages into the planting stage.
25Thomas and Strauss (1997, pp. 177-180) find significanteffects of calories for wage work but not for sell:
employment in urban Brazil and claim, in explicit contrast to the interpretationof Foster and Rosenzweig
(1994b), that "employers can observe the outputs (of better nutrition),such as better general health, improved
pallor and higher levelsof energy and effort, and those indicatorsmay be usedin settingwages of their workers."
But, as noted in Table 5, for most the sample that Thomas and Strauss use, the significantestimates that they
report imply negativeeffects of calories on wage rates received, which seems puzzlingif employerscan observe
indicators of greater calories consumedand such calories affect labor productivity.Therefore, althoughthey may
be correct that their results suggeststrongpositivewage effects of calories for those who consumevery low levels
of calorieseven thoughthe self-employmentselectedby such workersdoes not have suchreturns,the estimatesof
the negative wage effects of calories for most of the sampleraise questionsabout their interpretation.
 Ch. 43: Labor Markets' in Developing Countries 2901

which is consistent with the monitoring problem being particularly severe for planting
stage activities. The two studies of which I am aware that explicitly compare nutrient
demand elasticities for small farmers between the planting and harvest seasons, moreover,
report much larger elasticities in the planting season than in the harvest season in India and
Pakistan (Behrman and Deolalikar, 1989a; Behrman et al., 1997a). 26 This pattern of
elasticities is consistent with limited capacities for transferring resources across produc-
tion stages for small farmers and relatively high returns from using any extra resources to
increase own-farm productivity through consuming more calories in the planting stage
because of the absence of labor market rewards for such productivity in that stage due to
the monitoring problem, but the relative absence of such observability problems in the
harvest stage.
Thus these estimates, although based on a few samples, suggest that information
problems on work effort are significant for low-income agriculture in developing countries
- particularly in planting and other pre-harvest production stages, ff poor households had
better means of transferring resources over times or if there were better means of monitor-
ing work efforts, such problems would be lessened. 27 The extent of efficiency gains that
could be obtained, however, are not well-quantified.
Swamy (1997), finally, also provides a simple test of the nutrition-efficiency wage
model. He uses the estimates of calorie effects on productivity in rural India from Behrman
and Deolalikar (1989b) that are most favorable to this model (i.e., indicating the largest
response), and shows that, contrary to the claims of these models, a wage cut would lower
the cost per efficiency unit of labor.

4.3. Determinants o f schooling

Conventional wisdom is that, while the returns to health/nutrition investments may be

relatively high in poor and stagnant societies, schooling is the human resource with the
highest return in labor markets and elsewhere at somewhat higher levels of income and in
dynamic economies. Most of the empirical studies of the determinants of schooling in
developing countries have focused on household income and parental income, with a few

26There are many studies of the responsivenessof nutritionto income and some controversyover the magni-
tude of these responses that is discussed in Section 4.1, but most studies do not consider the possibility of
differential responses dependingon the stage of productionin agriculturaleconomies.
27It is useful to note that the rural employmentschemes that often are advocated to address rural seasonal
incomeproblemsdo not fully address this problemcausedby the nnobservabilityof effort becauseany household
member who is workingon such a scheme cannot be working simultaneouslyon own-farmproductionin which
the greater effort is rewarded. Such schemes may improvethe capacity of poor households to exploitthe greater
productivitywith better nutritionin own-farmproductionif they help provide additionalplanting-stageincome
through the use of some of the household'slabor on such employmentschemes so that the rest of the household's
labor can be better nourishedand therefore more productivein own-farmwork. But the increasedproductivityof
household own-farmlabor must offset the reductionin suchlabor in order for there to be incentivesto participate
in the employmentscheme.
2902 .I.R. Behrman

studies on other determinants such as opportunity costs, health and nutrition, risks and
expected returns.

4.3.1. Household income

As for health and nutrition investments, the implications of the standard human capital
investment model are that household income in itself should not affect schooling invest-
ment unless there are incomplete markets, Most empirical studies of schooling include
household income, apparently usually because of the perception (sometimes explicit) that
household income facilitates schooling investments because of imperfect capital and
insurance markets.
A recent survey in the Appendix of Behrman and Knowles (1999) of associations
between schooling investments with household income for 42 studies for 21 (mostly
developing) countries reports that estimates for about three-fifths of the schooling indica-
tors used in these studies yield significant associations between household income and
schooling. Among the cases in which income elasticities can be estimated from the
information provided in the studies, the median is 0.07, with the estimated income elasti-
cities tending to be a little higher for poorer samples and with small inverse associations of
the estimates with income reported in a number of studies. 2s Such low values for most of
these elasticities present a puzzle for those who perceive that there are high intergenera-
tional correlations in income and that the income returns from income-associated school-
ing investments are a major mechanism through which intergenerational income
correlations are generated. This survey suggests that one reason that the estimated income
elasticities in many studies are low is the use of income indicators that may be contami-
nated by relatively large measurement errors as a representation of the true longer-run
income constraint and possibly endogeneity. To illustrate, explorations with Vietnamese
data suggest that using predicted income/expenditures yields estimates on the order of
magnitude of 50-60% higher than using current annual income measures. This survey also
suggests that another reason that most studies might underestimate income-schooling
investment associations is that indicators of schooling investments generally are limited
to schooling attainment or enrollments. But cognitive achievement (or other school
outcomes) may differ significantly with income for a given level of schooling attainment
and the age of completing a given schooling level also may be inversely associated with
income (leaving more post-schooling time to reap the returns from schooling). Illustrative
estimates for Vietnam in Behrman and Knowles (1999) suggest that such considerations
add significantly to the income-schooling investment associations, as does selectivity
regarding who continues in school (i.e., only high-ability children from poor families,
but almost all children from better-off families). Thus, this survey suggests that most of the

28 The largest estimates - those over 0.20 - are for low-income countries, areas or time periods: Crte d'Ivoire,
Ghana, Nepal, Taiwan for the 1940-1949 birth cohort, Northeastern Brazil, and rural Pakistan. But these are the
only cases in which the estimates surveyed exceed 0.20. In several cases beyond these six the specifications used
allow non-linear income associations and find diminishing marginal income relations, although the changes in the
elasticities implied by these inverse associations are small.
Ch. 43: Labor Markets in Developing Countries 2903

empirical literature on schooling investments probably underestimates the importance of

household income, in part because of measurement issues and in part because systematic
conceptual frameworks are not used for the investigations.
One recent study of schooling in developing countries that lays out much more system-
atically than most of the literature how income and schooling investments might be related
in the presence of incomplete financial markets is Jacoby and Skoufias (1997). They note
that there had been considerable prior emphasis on both financial markets and human
capital as major factors in development, but not on their interaction. They also note that
there have been a number of recent studies to test the implications of incomplete financial
markets in both developing and developed economies, but that most of these studies shed
little light on the mechanisms by which consumption smoothing is attained. They inves-
tigate how child school attendance responds to seasonal income fluctuations in agrarian
Indian households using panel data. They study responses to aggregate and household
idiosyncratic and anticipated and unanticipated income shocks. They posit a dynamic
model of school attendance with different degrees of financial market completeness and
note that with incomplete markets consumption and schooling investment decisions are
not separable. Their estimation strategy is to relax successively restrictions on the relation-
ships between school attendance and income shocks implied by successively more incom-
plete financial markets, all with control for unobserved household heterogeneity. Their
results indicate that seasonal variations in school attendance are a form of self-insurance
that significantly reduces the schooling of children in households that are vulnerable to
risk, which is likely to be a costly form of insurance, particularly for poorer households.
The results have a number of potentially important policy implications, such that expand-
ing schools without understanding the nature of financial risks and market constraints may
have more limited effects on education than expected, effective compulsory schooling
laws or restrictions against child labor may have substantial negative effects on household
welfare, and improved shortterm credit and insurance markets may have important long-
term benefits in the form of greater human capital investments.

4.3.2. Parental schooling

Conventional wisdom is that: (1) mother's schooling has widespread positive substantial
effects on child education; (2) these effects tend to be much larger than those of father's
schooling; and (3) therefore, ceteris paribus, there is a stronger efficiency case (given
education externalities) for subsidies for female than for male schooling. Behrman
(1997b) first discusses a general framework for thinking about the impact of mother's
schooling on child education and then surveys what is known on the basis of all 237
estimates on 22 (mostly developing) countries that were located. Examination of available
estimates in light of this general framework suggests that knowledge on the impact of
women's schooling on child education generally could be improved with more clarity
about what model is estimated, roles of possibly important unobserved variables such as
preferences and abilities, distinctions between particular and more-general total effects,
and use of broader indicators of both mother's and child's education that capture outcomes
2904 J. R. Behrman

rather than primarily time-in-school inputs. Taken at their face value the central tendency
of current estimates is consistent with the "widespread" and "positive" part of point 1 of
the conventional wisdom, but not with the "substantial" part of point 1, or for the claim
that the effects of mother's schooling tend to be much greater than those of father's
schooling - and therefore not with a efficiency argument for large subsidies for female
schooling, or for larger subsidies for female than for male schooling.
Most studies, however, include among right-side variables some that possibly are
determined partially by mother's schooling. On the basis of a priori considerations, a
few studies that explore the effects of such procedures, and new estimates that characterize
all estimates that have been located, the usual specifications lead to a substantial under-
estimate of the total effect of mother's schooling and a smaller upward bias in the esti-
mated relative impact of mother's versus father's schooling, with control for income and
less so school characteristics biasing the estimated effects towards mother's schooling and
control for number of children and community characteristics biasing the estimates some-
what less towards father's schooling.
Most existing studies do not control for possible biases in the estimated effects of
mother's schooling due to unobserved (by analysts) abilities and preferences that directly
affect child education and that are correlated with mother's schooling. A few studies
suggest that unobserved preference and ability endowments may affect importantly the
estimated impact of mother's schooling on child education, with estimates generally
(although not always) biased upwards by the failure to control for these endowments.
They also suggest that marriage market considerations may be critical for analyzing the
impact of mother's schooling on child education, and that such considerations at least in
some contexts increase the estimated impact of mother's relative to father's schooling. But
these studies also point to the sensitivity of the results to how such endowments are
controlled, including the limitations of partial controls through observed indicators. There-
fore it is critical for interpretation that the underlying model be spelled out explicitly and
used directly as a guide to the estimation method because estimates using behavioral data
are necessarily conditional on particular assumptions about the underlying model and
explicit modeling makes it clear on what the interpretation is based.
I now review two recent studies of the role of parental schooling in child education in
developing countries that deal with some of the problems noted in this survey.
Behrman et al. (1999) examine the role of parents' schooling in child education in the
Green Revolution period in rural India. While the Green Revolution increased the rates of
return to men's schooling, given the gender division of labor there is not evidence of an
impact on the direct economic returns to women's schooling. Yet men in areas that
benefited from the Green Revolution married more-schooled women. This is somewhat
of a puzzle because, within household bargaining models, such women obtain a larger
share of the economic pie without contributing directly to the size of the pie. Among the
possible explanations are that men have pure consumption demands for more-schooled
wives and that more-schooled women contribute indirectly to the household by raising
more-educated children, a public good within the household.
Ch. 43: Labor Markets in Developing Countries 2905

This study examines the latter possibility. It presents household fixed effect estimates,
controlling for unobserved characteristics of the father's household, 29 for Indian farm
household children's daily school and study hours with and without instl-umenting
mother's schooling (literacy). In this case the instruments are local technological shocks
when the father was of marriage age that the authors argue are independent of the distur-
bance term in the within-household estimates for time that children spend studying or in
school. 3° The instrumented estimates indicate an impact of m o t h e r ' s literacy that is more
than double the uninstrumented estimates. Also of interest is the impact of the control for
the father' s family endowments by using within-household estimates in a context in which
extended households make possible such estimation. OLS estimates of the determination
of children' s school and study hours yield significant effects of mother being literate and of
father having primary schooling. But within-household estimates, while still yielding
estimates that imply that mother being literate has a significantly positive effect of
about the same magnitude (with the exact magnitude depending on the instrumenting
discussed above), yield estimates of the effect of father's primary schooling that are less
than a fifth of the OLS estimate and that are very imprecisely estimated (and would not be
judged non-zero even at the 50% level of significance). That is, in this case, the apparent
direct effect of father's schooling of more-or-less the same magnitude as of mother's
schooling in standard O L S estimates evaporates in within-household estimates while
the estimated effect of mother' s schooling is robust to the estimation alternatives consid-
ered. Thus in the OLS estimates the estimated direct impact o f father's schooling on child
educational time use is strongly contaminated and biased upwards by proxying for house-
hold preferences regarding time use and possibly household resources. To the extent that
the within-household estimates of the effect of father' s schooling differ from the OLS ones
because of the control for household resources, of course, father' s schooling still may have
an important indirect effect. However, the authors downplay this possibility because, if
there were such an effect, it also would seem to be reflected in subhousehold allocations of
household resources so that father' s schooling would still seem to be important even in the
within-household estimates.

29Another of the studies that is included surveyed in the survey summarized above also controls for childhood
family effects, in this case for the mothers, by using data on adult sisters and half-sisters in Nicaragua (Behnrlan
and Wolfe, 1984). For completed schooling for females the within-estimates of mother's schooling are 30% of
OLS estimates and the within estimates of father's schooling are 40% of the OLS estimates. For household
income the within estimates of mother's schooling are significantlynegative in contrast to insignificantnegative
estimates for OLS, while the within estimates of father's schooling are 70% greater than the OLS estimates (and
significantly positive). These results are suggestivethat controlling for mother's endowments also may affect the
estimates importantly, and in some cases as much or more so for mother's as for father's schoolingeffects. But
generalizing from these estimates is somewhat risky because of their dependence on half-sisters to obtain within
effects. Also they do not control for measurement error, the effects of which, as is well-known, are exacerbated
with within estimates, although the result that the within estimates are larger in absolute magnitude in several
cases could not come from the classical measurement error model.
3oAs the authors note, if mothers' preferences related to child schooling are heterogenous and known at the
time of"marriages, then the instruments used may not be independent of the disturbance term in the child's time
use relation.
2906 J. R. Behrman

In another recent paper Foster (1996) argues that estimates of parental schooling on child
education can be seriously biased if marriage partners self-select on the basis of unobserved
characteristics. To deal with this issue, he develops a model of the marriage market in
which potential mates care about the human capital of their offspring (a public good within
marriage) as well as their own private consumption. Under the assumption of transferable
utility, child investment is shown to depend on the income and tastes for offspring school-
ing o f each of the marital partners. The problem in estimating the decision rule is that, with
selective marriages, the unobserved traits of existing marital partners are not orthogonal.
The paper develops a simulation method for correcting for the selection bias that involves
explicitly solving approximately for the marriage market equilibrium. Using data from
rural Bangladesh, the estimates indicate that marital selection is quantitatively important,
significantly diminishing the effect of husband's traits by 35-55% and augmenting the
effect of w i f e ' s traits by 13-16% on the desired schooling of children. 31 This effect is
separate from biases due to mother's schooling being a proxy in part for her own unob-
served tastes and productivity in child education, which are not considered in this study.

4.3.3. O p p o r t u n i t y costs
If markets were complete they would incorporate all costs of schooling, including most
importantly the cost of time. But labor markets for children, among others, in many contexts
are quite limited or non-existent. So private schooling investments decisions generally tend
to value time of children differentially and inefficiently. A few studies of developing
countries address directly the nature of opportunity costs of children in attending school.
Rosenzweig and Evenson (1977), for example, find that the combination of incomplete
labor and land markets in rural India result in significant positive effects of land ownership
on child on-farm labor and thus significant negative effects of land ownership on child
school attendance. Thus, despite the generally positive relation between household
resources and child schooling noted above probably due to incomplete capital and credit
markets, certain forms of household resources - in particular land that is a complement
with child labor for which market substitutes are not readily available - in some market
contexts cause a reduction of child schooling.
In most societies there is gender specialization in the provision of home health care,
with females providing most such care. Pitt and Rosenzweig (t990) develop and imple-
ment a method for estimating the effects of infant morbidity on the differential allocation
of time of family members within the context o f a household model in which health is
determined simultaneously. Identification of the effects of the health of person k on the

3LFoster shows that these directions of bias can occur when the unobservable component of assortative mating
is large relative to the observable component for women and there is no unobservable component for men (say,
because they primarily are income earners based on observed characteristics). The intuition is that, in this case,
the husband's schooling is positively correlated with the wife's unobservable so that, in estimates that do not
control for marriage selection, the estimated effect of the husband's schooling is overstated. This effectively
means that there is in the disturbance term an expression equal to the true minus the estimated effect of husband's
schooling times the wife's unobservable, which is negative so that the wife's schooling effect is underestimated.
Ch. 43: Labor Markets in Developing Countries 2907

behavior of personj when the behavior of personj may affect the health of person k (e.g.,
through child care) is not easy in part because it is difficult to find instruments that directly
affect i's health but not directly that o f j (net of any indirect effects through i's health).
They assume that households have a consensus preference function defined over the home
time and health of each household member and a composite jointly-consumed consump-
tion commodity with heterogeneity in such preferences across households. This preference
function is maximized subject to a budget constraint (which includes the wage for each
household member type as well as non-labor earnings) and a health production function
(which includes the home time of each household member, the health of every other
household member to allow for intrafamily health externalities inclusive of contagion
and/or health efficiency effects on home time, and private health-related goods and
selTices). They posit that the linearized demand relations for home time of household
members i andj conditional on the health of household member k in which the coefficients
on price of private health-related goods is the same for i and for j (e.g., if the health
production function is the same for i and j), which is a critical (and perhaps strong)
identifying assumption. The differenced version of these relations then gives the difference
in the home time of i andj as a function of the difference in their wage rates, any difference
in the impact of the price of the jointly-consumed composite commodity price on their
home time use, and any difference in the impact of the health of k on their time use.
Conditional on the assumptions underlying this relation, a consistent estimate of the
impact of the health of k on the difference in home time use between i andj is obtained by
using the prices of health-related goods as instruments. The data requirement for estimat-
ing such relations are severe: information on child health, the activities of all household
members, and the prices of health-related goods, as well as a large enough sample so that
there are enough families with the family types of interest with whom the within estimates
can be made (i.e., mothers, teenage daughters and sons, and infants). The 1980 Indonesian
Socioeconomic Survey linked with other information on prices and health programs has
such data for 5831 households. However, for both health and time allocations what is
available in this dataset are discrete indicators (dichotomous for health, trichotomous for
activities - labor force, school, home time), so Pitt and Rosenzweig adopt a fixed effects or
within-family logit procedure that is parsimonious in terms of parameters to be estimated,
permits identification, and controls for possible selectivity of households into this subsam-
ple. The estimates obtained indicate that teenage daughters were significantly more likely
to increase their participation in household care activities and to reduce their participation
in market activities and at school in comparison with teenage sons in response to increased
morbidity of infant siblings. Moreover, such estimates differed markedly from the esti-
mates obtained if there was not control for the simultaneity of child health determination
and time uses of household members, although the conclusions need to be qualified
because the critical identifying assumption is strong.

4.3.4. Impact of nutrition~health on schooling

As noted above, one implication of the standard human capital investment model is that
2908 J. R. Behrman

human resource investments may interact. In the presence of incomplete markets for
capital and insurance, such effects are likely to be inefficient. There have been consider-
able claims lbr both developing and developed economies that better child nutrition and
health, for example, cause better schooling success, with long-run benefits in terms of
economic productivity in the labor market that may not be realized efficiently due to
incomplete markets. But evidence is quite limited because numerous studies based on
socioeconomic surveys fail to model the process clearly and, perhaps for that reason, fail
to consider the endogeneity of child health, measurement error, and the impact of unob-
served fixed and choice inputs. Until recently, the available studies using behavioral data
that were used to justify the claim of positive child health/nutrition on child success did not
permit clear interpretation because the choice element of child nutrition and health was not
controlled. Three recent studies for developing countries attempt to deal with these esti-
mation problems.
Glewwe and Jacoby (1995) explore one dimension of the relation between child health/
nutrition and child school performance. They are sensitive to the treatment of child health/
nutrition as a choice rather than predetermined as in the previous literature and explore
how robust their estimates are to alternative methods of controlling for choices affecting
child health/nutrition within an explicit model of economic behavior (although their need
to depend on noisy recall data for their a priori most persuasive estimates limits their
success in their empirical application that otherwise uses basically cross-sectional beha-
vior data). Previous studies in this literature, moreover, had focused on fairly static analy-
sis of relations between indicators of child health/nutrition and outcomes such as test
performance and grade completed controlling for age. Glewwe and Jacoby instead
consider the dynamic sequence of age of initial enrollment (that they first demonstrate
can have a substantial impact on lifetime wealth), progress through school, and age of
school completion and entry into the post-school workforce. Their results indicate that
delays in enrollment are responsive to early child malnutrition, although the estimated
effect is reduced substantially (by almost two fifths) if there is control for unobserved
family and community variables, suggesting that indicators of child health/nutrition in part
proxy for such unobserved factors in previous estimates. Their empirical results therefore
suggest that: (a) estimates of the impact of child health/nutrition on child schooling
success may be quite sensitive to the underlying behavioral assumptions and the nature
of unobserved variables, (b) if there is not control for behavioral choices in the presence of
unobserved household and community factors the estimated impact of child health/nutri-
tion on child schooling success is overestimated substantially so most of the previous
studies in this literature may be fundamentally misleading regarding the magnitude of the
impact of child health/nutrition on child schooling success, and (c) an important channel
through which child health/nutrition may affect earnings through schooling pertains to the
age when children start school, a channel that had been largely ignored in the previous
literature on child health/nutrition and schooling success.
Bebrman and Lavy (1997) show that a priori the biases resulting from ignoring house-
hold decisions affecting child health/nutrition in the presence of unobservable in estimates
Ch. 43: Labor Markets in Developing Countries 2909

of the impact of child health/nutrition on child schooling success may be positive or

negative depending on which of a number of household allocation behaviors dominate
and what is the nature of any unobserved choice inputs in educational production. Then
illustrative empirical analysis, using rich data from Ghana, is presented, with the following
results: (1) IV estimates based on observed family and community characteristics similar
to those used in other studies suggest a downward bias in OLS. (2) Family and community
fixed effects estimates suggest that the direction of the bias in standard estimates is upward
and that the true effects of the range of observed child health on school success is not
significant despite the strong association that leads to the appearance of an effect in
standard OLS or IV estimates using family and community variables. (3) The usual
assumption that there are no unobserved choice inputs in educational production probably
leads to an upward bias in the estimated impact of child health on schooling even if there is
good control for the endogeneity of child health and measurement error. (4) Child health
also does not significantly affect child cognitive achievement through schooling attain-
ment; consideration of the relations that usually have been used to investigate such a
possibility, moreover, suggests that the coefficients that are usually estimated are not
coefficients that represent the impact of child health on child schooling. (5) The preferred
estimates control for unobserved family and community fixed effects and are robust to
other estimation problems, so the standard estimates overstate the impact of child health in
the observed range on child schooling success.
Alderman et al. (1997) employ longitudinal data to investigate the impact of child
health/nutrition on school enrollments in rural Pakistan using an explicit dynamic
model for their preferred estimates. These estimates use price shocks when children
were of preschool age to control for behavior determining the child health/nutrition
stock measure. They indicate that child health/nutrition is three times as important for
enrollment than suggested by "naive estimates" that assume that child health/nutrition is
predetermined rather than determined by household choices in the presence of unobserved
factors such as preferences and health endowments. These results, therefore, reinforce
strongly the importance of using estimation methods that are consistent with the economic
theory of households to explore the impact of some choice variables on others using
socioeconomic behavioral data.

4.3.5. Impact of expected returns on schooling

Human capital investments are made under imperfect information with learning by poten-
tial investors about both individual ability and the returns to schooling. Ability varies
across individuals (e.g., due to genetic variation) and returns to schooling vary across local
conditions for adoption of new knowledge (e.g., due to variations in the suitability of new
agricultural technologies across space because of soil and weather differentials for the
Green Revolution).
Yamauchi-Kawana (1997) considers, within a particular developing country context,
the problem that households have in assessing whether to invest in schooling when the
returns to schooling may have changed. First he models the schooling investments within a
2910 J.R. Behrman

two-period framework. Household members must decipher from uncertain production

processes the contribution of schooling to output under a new technological regime.
The schooling investment decisions must be made in the first period before these uncer-
tainties are resolved in the second period. He shows that adjustments in perceptions of the
school return are faster if aggregate income volatility is less, if population density is
greater, and if there is the optimal number of highly-schooled versus low-schooled adults
in the community from which to infer the returns to schooling. Because each household
learns from others, but no one takes into account that others are learning from themselves,
there is an externality from the social learning.
He then uses this framework to guide analysis of a national stratified random rural panel
dataset from India at the start of the Green Revolution (in which new crop varieties created
at international agricultural research institutions in the Philippines and Mexico became
available in India, but the suitability of which varied considerably across locales because
of varying soil and weather conditions so substantial learning was involved regarding their
local suitability in each community). First, he estimates farm profit functions with panel
data to control for unobserved farm productivity factors in order to infer farmers' abilities
and village-specific schooling return differentials. Next, he investigates the school enroll-
ment response to these estimated signals for ability and for the returns to school, with
learning weights that differ depending on the assets and volatility in each particular
context. These estimates imply an estimate of the optimal village proportion of educated
population for learning - and suggest that on the average the actual proportion of educated
households in the sample villages was significantly less than the optimal level for the
purpose of learning due to the positive externality provided to others when particular
farmers, some with and some without education, explore the new technologies.

4.4. Impact of schooling on economic productivity

There are literally hundreds of micro studies that purport to investigate the impact of
schooling on economic (e.g., wages, agricultural productivity) productivity in developing
countries within a static framework (see the surveys in Schultz, 1988; Behrman, 1990a,b,
1997b; King and Hill, 1993; Psacharopoulos, 1994; Strauss and Thomas, 1995). Table 6
reproduces a well-known summary of many of the studies on the wage outcomes. A few
studies tie together micro estimates of the impact of human resources with the distribution
of income or of earnings (e.g., Blau et al., 1988; Lam and Levison, 1991; Psacharopoulos
et al., 1992; Lam and Schoeni, 1993, 1994).
An effective way to summarize many of these results has been through the calculation of
the real rates of return to the costs incurred in schooling. This has been effective because
rates of returns permit comparisons among a wide range of investments, both within the
schoolillg sector and elsewhere in the economy. Typically these rates of return have been
calculated by comparing the direct economic outcomes for individuals with different
amounts or types of schooling (or for different types of individuals) and calculating the
rate of return to the private costs (primarily the time costs but perhaps also tuition, books
Ch. 43." Labor Markets in Developing Countries 2911

Table 6
Percentage returns to investments in schooling latest year, regional averages a

Region Social Private

Primary Secondary Higher Primary Secondary Higher

Sub-Saharan Africa 24.3 18.2 11.2 41.3 26.6 27.8

Asia b 19.9 13.3 11.7 39.0 18.9 19.9
Europe/Middle East/ 15.5 11.2 10.6 17.4 15.9 21.7
North Africa b
Latin America/Caribbean 17.9 12.8 12.3 26.2 16.8 19.7
OECD 14.4 10.2 8.7 21.7 12.4 12.3
World 18.4 13.1 10.9 29.1 18.1 20.3

Source: Psacharopoulos (1994, Table 1).

b Non-OECD.

and m a t e r i a l s and o t h e r p r i v a t e costs) a n d to the social c o s t s (the p r i v a t e c o s t s plus p u b l i c

s u b s i d i e s ) to obtain, r e s p e c t i v e l y , the s o - c a l l e d " p r i v a t e " a n d " s o c i a l " rates o f return to
s c h o o l i n g . 32 T h e s e e s t i m a t e s are w i d e l y i n t e r p r e t e d to i m p l y that i n d e v e l o p i n g countries:
(1) the r a t e s o f r e t u r n to s c h o o l i n g are high; 33 (2) t h e y do n o t d e c l i n e very r a p i d l y w i t h the
level o f d e v e l o p m e n t ; (3) the i m p a c t o f s c h o o l i n g , p a r t i c u l a r l y for f e m a l e s , o n n o n - m a r k e t
o u t c o m e s is c o n s i d e r a b l e and g e n e r a l l y greater t h a n that o f m a l e s ; (4) the social rates o f
r e t u r n d e c l i n e w i t h s c h o o l i n g levels 34 ( a l t h o u g h the private rates o f return do n o t n e c e s s a -
rily do so b e c a u s e o f r e l a t i v e l y h i g h p e r s t u d e n t s u b s i d i e s to h i g h e r s c h o o l i n g levels), are
h i g h e r f o r g e n e r a l as o p p o s e d to t e c h n i c a l v o c a t i o n a l s c h o o l i n g , a n d at least as h i g h o n
a v e r a g e for f e m a l e as f o r m a l e s c h o o l i n g ; (5) variability in s c h o o l i n g is a s s o c i a t e d with the
v a r i a b i l i t y in i n c o m e d i s t r i b u t i o n a n d m o r e s c h o o l i n g is a s s o c i a t e d w i t h less p r o b a b i l i t y o f
b e i n g b e l o w the p o v e r t y line; 35 and (6) t h e r e is n o t likely to b e an e q u i t y - p r o d u c t i v i t y

32 Sometimes the Mincerian (Mincer, 1974) semilog relation between wages and schooling with control for
post-schooling experience (or age) is used to calculate the private rate of return to time spent in schooling instead
of in the labor market under the Mincerian assumptions (e.g., there is equilibrium so that individuals are
indifferent among various schooling levels and characteristics such as ability and family background enter into
the wage determination function so they are not correlated with schooling). The Mincerian formulation assumes
that there are opportunity costs to schooling at all ages, an assumption that Psacharopoulos (1994) and some
others criticize.
33 Such estimates imply, in fact, that investment in schooling is such a high return investment that they are not
completely credible on these grounds alone. Investments with a real annual rate of return of 16-24% (the social
rate of retnrn to primary school in the four developing regions given in Psacharopoulos, 1994) and with reinvest-
ment of the proceeds of such investment implies that society can double the real invested assets in 2.9-4.3 years,
and the social real rate of 11-18% on secondary schooling implies the possibility of doubling real assets in 3.8-
6.3 years. These estimates, moreover, understate the true social rates of return and overstate the true time that
social assets could be doubled by marginal schooling investments if there are positive externalities to schooling as
often is claimed. If developing countries have available such investments opportunities on a fairly broad scale
(i.e., in most of its children), it would seem that much higher economic growth would be observed than ever has
been experienced for any sustained period of time.
2912 J. R. Behrman

t r a d e o f f in e x p a n d i n g s c h o o l i n g in t h e m o s t p r o d u c t i v e w a y b e c a u s e the r e t u r n s are h i g h e s t
for b a s i c ( p r i m a r y , t h e n s e c o n d a r y ) s c h o o l i n g for w h i c h f u r t h e r e x p a n s i o n is l i k e l y p r i m a r -
ily to e n r o l l m o r e c h i l d r e n f r o m v e r y p o o r f a m i l i e s a n d the t o t a l r e t u r n s are h i g h e r f o r
f e m a l e s t h a n f o r m a l e s . U n d e r the a s s u m p t i o n t h a t w a g e s are s t r o n g l y a s s o c i a t e d w i t h
p r o d u c t i v i t i e s , t h e s e c o n c l u s i o n s g e n e r a l l y are i n t e r p r e t e d to c a r r y o v e r to t h e i m p a c t o n
p r o d u c t i v i t y . 36
T h e r e are, h o w e v e r , a n u m b e r o f w e l l - k n o w n p o s s i b l e p r o b l e m s w i t h t h e m e t h o d o l o g y
s k e t c h e d out in t h e p r e v i o u s p a r a g r a p h . T h o u g h m o s t studies in this g e n r e do n o t a t t e m p t to
control for these problems, some of those that do report that such controls make consider-
able d i f f e r e n c e s in the e s t i m a t e d i m p a c t o f s c h o o l i n g . M o s t o f t h e e x i s t i n g studies do n o t
c o n t r o l w e l l f o r the b e h a v i o r a l d e c i s i o n s t h a t d e t e r m i n e w h o g o e s to w h a t t y p e o f s c h o o l
for h o w l o n g w i t h w h a t d e g r e e o f success. S i m p l e a n a l y t i c a l f r a m e w o r k s f o r s c h o o l
i n v e s t m e n t s , as well as c a s u a l o b s e r v a t i o n s , s u g g e s t t h a t i n d i v i d u a l s w i t h h i g h e r i n v e s t -
m e n t s i n s c h o o l i n g are l i k e l y to b e i n d i v i d u a l s w i t h m o r e ability a n d m o r e m o t i v a t i o n w h o
c o m e f r o m f a m i l y a n d c o m m u n i t y b a c k g r o u n d s t h a t p r o v i d e m o r e r e i n f o r c e m e n t for s u c h
i n v e s t m e n t s a n d w h o h a v e l o w e r m a r g i n a l p r i v a t e c o s t s for s u c h i n v e s t m e n t s a n d l o w e r
d i s c o u n t r a t e s f o r the r e t u r n s f r o m t h o s e i n v e s t m e n t s a n d w h o are l i k e l y to h a v e a c c e s s to
h i g h e r q u a l i t y schools. T h e r e f o r e s u c h studies i m p l i c i t l y a s s u m e t h a t s c h o o l i n g is distrib-
u t e d r a n d o m l y a m o n g s a m p l e m e m b e r s r a t h e r t h a n t h a t the d i s t u r b a n c e s in t h e r e l a t i o n s

34The social returns to schooling may be non-linear, with increases for lower and middle schooling levels and
then declines for further schooling. Barros (1992) gives an example of the relation of schooling to adjustment
capacities, which may have social implications beyond private implications because of the social costs of
unemployment. During periods of adjustment the relative gainers are those who have the less specific human
capital to lose and who can acquire new specific human capital the most cheaply (where "specific" means
"specific" to a firm or to a particular job). Those with little or no schooling are likely to have little specific
human capital to lose, but also are likely to acquire new specific human capital at great cost. Those with more
general human capital are likely to be able to acquire new specific human capital relatively cheaply, but also are
more likely to have greater specific human capital from the past the value of which may be reduced or lost due to
adjustment. The costs of adjustment are likely to be greatest for those with the greatest gap between specific
human capital and general human capital (since the cost of acquiring new specific human capital is likely to be
inversely associated with the stock of general human capital). The relation between schooling and adjustment
capacity, therefore is an empirical question on which some limited evidence for Brazil suggests important non-
linearities with maximum adjustment capacities for those with medium schooling levels.
35Psacharopoulos et al. (1992), for example, examine the relation between schooling and income inequality
and poverty in the Latin American and Caribbean region. A decomposition of the inequality in the distribution of
workers' income (including only individuals over 15 years of age in the labor force with positive income)
indicates that variations in schooling attainment are associated with about a quarter of the income inequality.
Also low schooling attainment is the characteristic most associated with being in the bottom 20% of the
distribution of workers' income; on average those with no schooling have a 56% probability of being in the
bottom 20% of the workers' income distribution, while those with primary schooling have 27% probability, those
with secondary schooling 9% probability, and those with university schooling 4% probability. These results are
characterized by Psacharopoulos et al. (1992, pp. 40, 48) to indicate "the overwhehning preeminence of educa-
tion" and that "clearly... education is the variable with the strongest impact on income inequality."
36The third conclusion and a small subset of the studies underlying the other conclusions use direct measures of
productivity, not wages, as the dependent variables.
Ch. 43: Labor Markets in Developing Countries 2913

used to explore the impact of schooling on various outcomes are correlated with schooling
due to the failure to control for such factors so that the estimates in such studies probably
suffer from omitted variable biases. The association of schooling with labor market
outcomes such as wage rates and agricultural productivity (as well as with household
outcomes such as fertility and child health) does not necessarily represent causality
because in most estimates years of schooling is representing not only time in school,
but also factors that are correlated with years of school such as abilities, discount rates,
family backgrounds, and schooling qualities. To obtain insight into the impact of years of
school on such outcomes, one needs to control for these other factors, as do to a certain
extent some - but not many - of the existing studies.
A number of "revisionist" studies for developing countries, parallel to a similar litera-
ture for developed economies, have explored the impact o f some of these estimation
problems on estimated schooling returns with data or specification modifications of the
standard earnings function framework by controlling for: school quality (Behrman and
Birdsall, 1983), unobserved shared family background of adult siblings and of members of
the same household (Behrman and Wolfe, 1984; Behrman and Deolalikar, 1993), usually
unobserved abilities through new tests (Boissiere et al., 1985; Knight and Sabot, 1990;
Glewwe, 1996), selectivity (Schultz, 1988), dropout and repetition rates (Behrman and
Deolalikar, 1991), 37 measurement error, school quality and behavioral choices regarding
school attendance (Alderman et al., 1996b). In earlier surveys I have claimed that those
studies that do incorporate such controls for developing countries tend to find that the
"standard estimates" (i.e., those without such controls) may overstate the impact of
schooling attainment by as much as 40-100%, probably more so for primary schooling
and underestimate the relative importance of school quality improvements (Behrman,
1990a,b). 38 The recent ferment in studies of such questions for the United States (see
the chapter by Card in this H a n d b o o k ) has re-emphasized the point that random measure-
ment error and other estimation problems may mean that some of these studies may not
overestimate schooling attainment effects as much as I earlier suggested, although there is
not yet a clear consensus regarding the rates of return to schooling in the United States,
there also has been increasing emphasis on relatively high returns to school quality in that
economy, and the issues addressed in the recent literature raise questions about some but
not all o f the estimates in the "revisionist" literature on rates of return to schooling in
developing countries. At this point I perceive that the "standard" estimates for developing

37Grade repetition is substantial in many developing countries (e.g., Latin America and the Caribbean have a
first grade repetition rate of 42%, and an overall primary school repetition rate of 29% according to recent
estimates based on a special UNESCO/OREALC survey) so the failure to control for grade repetition and school
dropouts in standard estimates may be quite important.
38Such factors are controlled generally by linking data used for the standard estimates with other information
about characteristics such as school quality, family background, and ability or by using special data on adult
siblings or family or community members to control for common unobserved characteristics (e.g., the estimate of
the difference in wage rates regressed on the difference in schooling for adult siblings controls for the additive
effect of common family and community background shared by the siblings).
2914 Z R. Behrman

countries probably overstate the true schooling returns substantially but that there remain
some open questions about this literature to which studies of developing as well as
developed economies are likely to continue to contribute.
Beyond the "standard" and related "revisionist" literature, however, recent empirical
studies of schooling in developing countries have contributed to knowledge of the impact
of schooling within a dynamic context with explicit attention to various forms of market
imperfections.
Education may enable one to deal better with uncertainty by improving one's abilities to
learn, which is likely to be particularly important in dynamic environments in which there
are technological innovations and new market opportunities (Welch, 1970; Schultz, 1975).
These notions have been formalized recently in a target-input model in which individuals
choose an allocation of resources or inputs knowing the technology of production only up
to a stochastic "target" for the level of input use (see Rosenzweig, 1995 for details and
references). With repeated production periods, in each period the priors regarding the
optimal input use are updated based on past experience. Education can affect the produc-
tion cure learning process in two ways: (1) Education can increase the precision of the
information that an individual has initially because of access to more information
sources. 39 In this case experience and education clearly are substitutes - alternative
ways of increasing the precision of one's priors. Therefore the returns to education are
high only with new technological options, and decline with more experience with any
given technology. (2) Education may enable individuals to gain more information from
each use of a technology than they would otherwise be able to gain - the more educated
may learn faster and be able to decode information acquired through experience more
effectively. If this is the only effect, at low levels of experience, education and experience
are complements rather than substitutes so that the returns to education at least initially
increase with experience with a given new technology. While this approach is stated in
terms of new technology, it should be clear that similar possibilities exist with the stochas-
tic terms and learning relating to markets as well as to technology. This role may be critical
for entrants into a new market, whether they be youth searching for good matches in the
labor market or entrepreneurs entering a new domestic or international product or input
market.
It is useful to consider in somewhat more detail two examples of such studies, Foster
and Rosenzweig (1995, 1996b). There has been renewed interest in the fundamental issue
of what causes economic growth with its multiple implications for labor, with particular
attention focused on the role of information externalities. Evidence on the existence of
such spillover effects from aggregate data has not been persuasive. Foster and Rosenzweig
(1995) empirically implement a "target-input" model of agricultural technology adoption
in which there are potential information externalities associated with adoption by neigh-
bors. The learning model yields an explicit representation for the profit function that

39For example, Thomas et al. (1991) give such an interpretation based on how the coefficientestimates of
mother's schooling declines as they include use of information sources in their conditional demand relations.
Ch. 43: Labor Markets in Developing Countries 2915

depends on own and neighbors' accumulated experience with the new technology. The
profit function is estimated as both a linear approximation and in its exact non-linear
representation using fixed-effects instrumental variables procedure that accounts for
input endogeneity. The results indicate that the experience of neighbors as well as own
experience increase farm profits; there is both learning by doing and learning from others.
The adoption decision (the amount of land devoted to high yielding variety crops) is
derived from a Markov perfect game-theoretic model. A linear approximation to that
rule is estimated using a fixed-effects procedure to control for permanent unobservables.
It is found that own and neighbors' experience provide similar information about optimal
inputs. Moreover, the finding that increasing own assets increases the level of adoption
while increasing neighbors' assets reduces adoption indicates that the learning externality
is not fully internalized by the village. The results in this paper provide what is arguably
the best evidence to date on the existence of knowledge spillovers - the extent of which are
critical for efficiency arguments for public subsidies for schooling. 4°
Foster and Rosenzweig (1996b) provide a related analysis of the relationship between
schooling and technical change. Although other papers have provided evidence on the
long standing question of whether exogenous technical change increases the productivity
returns to schooling, they have not been sensitive to the role of human capital accumula-
tion itself in the process of technical change. The analysis in this paper draws on panel data
for agricultural households in India during the period of the green revolution, which is
reasonably argued to correspond to a period of exogenous technical change that differs
across space because of differences in water and soil conditions. In addition, this paper
assesses quantitatively within a unified framework the extent to which schooling levels
respond to the increased returns and the extent to which the demand for schooling responds
to investments in schools. Geographic variation in the extent of technical change is
sufficient to identify profit function parameters that indicate the extent of technical change
in Indian districts and changes in the profitability of inputs such as schooling, irrigation,
etc. The estimation procedure takes into account that inputs evolve dynamically and that
there may be district-level unobservables that affect both profits and inputs. The results
indicate that the schooling return is on average augmented by exogenous technical change.
Moreover, the increase in the return to schooling is greater the higher is the rate of growth
in the area; having a primary education increased the impact of technical change on profits
by 70%. An approximate dynamic schooling decision rule is estimated that makes use of
the profit-function estimates of district-level technical change. In conformity with the
profit function estimates, the demand for schooling is found to increase with the level
of technical change. It is also found to increase with the availability of schools which
implies that policies that promote technical change are complementary to investments in
schools. This study thus provides perhaps the best available evidence that the returns to
school are high in the presence of new technologies.

40Besley and Case (1994) is another important study that also is concerned with the distribution of new
technologies in rural India and possible spillovers,but without the same focus on schooling.
2916 J.R. Behrman

5. Urban labor markets, labor-market regulations, international trade policies and

manufacturing

Urban labor markets differ from rural labor markets in developing economies by having:
(1) more heterogenous production and therefore more heterogenous labor and more wage
variance; (2) higher returns to education and related skills (in part in governmental occu-
pations) and therefore more concentration of more-educated and more-skilled workers; (3)
less dependence on weather and thus less seasonality and less problems due to incomplete
seasonal markets; (4) more geographical concentration of production activities so that
information is likely to be better and mobility greater with greater payoffs for job search
thus higher unemployment; and (5) more intense policy regulation and union activities in
part because of the greater concentration and greater scale economies and lesser costs of
monitoring compliance and in part because of the greater worker heterogeneity.
After a more extensive description of urban labor markets in developing countries with
emphasis on features similar to these, Rosenzweig (1988a, p. 755), concluded: "An
informed reader will see that most of the features of the low-income-country urban
environments described also characterize urban areas of high-income countries. And the
issues of the impact of governmental labor market interventions and trade unions and the
determinants and consequences of job search strategies, which appear particularly perti-
nent to such settings, form an important part of the core of modern labor economics. Few
distinct analytical models specifically targeted in any meaningful way to problems of low-
income country urban markets have emerged in the literature."
In my judgement a decade later this conclusion still holds. For that reason and because
some of the discussion above relates in part to urban labor markets (i.e., some of the
literature on the determinants of and the returns to human resource investments), I here
devote much less space to urban labor markets in developing countries than ! have devoted
above to rural labor markets in developing countries.
I first consider some aspects of urban labor market dualism, which has received more
emphasis in studies on developing countries than in studies on developed economies, and
then turn briefly to empirical studies on labor market regulations and on the impact of trade
policy changes on manufacturing - two policy areas in which some of the changes in a
number of developing countries have been larger in degree if not different in kind from
those experienced in developed economies.

5.1. Urban labor market dualism

A common description of urban labor markets in developing countries dating back at least
to Fields (1975), Mazumdar (1976) and Sabot (1977) is that they are dualistic. On one
hand there is a "formal" or "modern" sector comprised of mostly-larger, often relatively
capital-intensive, private and public relatively high-wage producers that are subject to and
more or less comply with labor market regulations. On the other hand, there is an "infor-
mal" sector comprised of small, mostly family enterprises that are relatively labor-inten-
Ch. 43: Labor Markets" in Developing Countries 2917

sive and low-wage and that are not subject to or do not comply with labor market regula-
tions.
Much of the empirical literature on urban labor markets (most of it until recently) has
focused on testing whether there are barriers to mobility between the informal and formal
sectors by comparing wage rates (e.g., Mazumdar, 1981) or by comparing estimated wage
(or earnings) relations (e.g., Heckman and Hotz, 1986; Funkhouser, 1997a,b). These
comparisons often are interpreted to mean that there is urban labor market segmentation.
But, as at least the latter studies recognize and attempt to deal with in part, such compar-
isons are difficult to interpret because workers with identical observed characteristics may
differ in unobserved characteristics (e.g., innate ability, preferences) that affect their
selection into different sectors, lifecycle wage schedules may differ with different tech-
nologies and organizations but generally only wages at a point in time are observed, the
empirical classification of sectors is arbitrary, identifying the labor payment to unpaid
family workers is difficult, and the comparisons are conditional on the correct specification
ofthnctional forms. Of more fundamental interest than such comparisons, moreover, is the
question of the extent of mobility among such urban sectors (Section 6.3).

5.2. The effects of labor market regulations on formal-sector wages and employment

Bell (1997) uses time series and panel firm data (and individual data for Mexico) to
investigate the impact of minimum wages on formal-sector wages and employment in
Colombia and Mexico. She finds virtually no effect in Mexico, which she suggests is
because the levels of minimum wages were too low to be ineffective. For Colombia, in
contrast, she finds significant negative employment elasticities that imply reductions of
formal sector employment for low-skilled, low-wage Colombian workers of 2-12% for a
10% increase in minimum wages with firm fixed-effects estimates (although the estimates
appear insignificant without control for firm fixed effects).
MacIsaac and Rama (1997) explore formal-sector labor costs in Ecuador, which are
alleged to be high because of many policy-mandated benefits that are equal to 75% of the
minimum wage. They use household survey data to describe the associations between
average hourly earnings and various observed worker and employer characteristics. They
find that the effect of the mandated benefits is mitigated by a reduction (39% on the
average) in the base earnings that is larger in the private than in the public sector but
negligible for unionized workers. As a result, total labor costs for complying employers
increase only 8% despite the substantial mandates. They also find that, despite the
mandated benefits, interindustry wage differentials are comparable to those in Bolivia,
which is alleged to have much more flexible labor markets.
Gruber (1997) estimates the incidence of a sharp change in payroll taxation in Chile
(due to the privatization of the Social Security system in 1981) on wages and employment
in reduced-form relations. He uses plant data and finds that the incidence was entirely on
wages, with no effect on employment - a result that is robust to alternative estimation
strategies to deal with possible measurement error. As he notes, an important limitation of
2918 J. R. Behrman

this approach is that the reduced-form estimates cannot disentangle the structural sources
of wage and employment changes.
Thus these thi'ee studies all conclude that most of incidence of the cost of legislated
benefits is on covered workers' wages, with relatively small impact on total labor costs and
on employment in covered sectors. If so, it is not the case that these regulations have much
impact on either the competitiveness of covered firms nor all the welfare of covered
workers. These studies are basically silent on the effects of such regulations on uncovered
workers. Nor do they investigate what determines compliance with labor regulations and
thus whether a firm effectively is in the covered or non-covered sector.

5.3. The effect of trade reform and adjustment on formal-sector manufacturing labor

The most radical trade reforms and economy-wide adjustments in recent decades have
been in developing economies and economies in transition. Yet there has been until
recently very little analysis of these experiences as compared with the much larger number
of analysis of the impact of trade reform on labor in developed economies. 4~
Revenga (1997) analyzes the impact of trade liberalization in 1985-1988 on employ-
ment and wages in the Mexican manufacturing sector. During this time period, average
tariffs on Mexican manufacturing were cut in half and the coverage of import licensing
was cut by three-quarters. She posits that firm wages are a weighted average of the union's
preferred wage outcome and the alternative industry or regional average, so that firm
wages can be decomposed into firm-specific wages from quasi rents that differ from
industry average wages, with unobserved heterogeneity in firm bargaining power. She
estimates this relation from time series firm data, using industry trade policies as instru-
ments to attempt to break the correlation with the unobserved heterogeneity in the distur-
bance term. She likewise estimates firm labor demands conditional on output, again using
industry trade policies as instruments. She first documents that many of the rents generated
by previous trade protection were obtained through a wage premium by workers in the
protected sector, which she estimates totaled 25% of workers' earnings. She estimates that
trade liberalization shifted down industry product and labor demand which in itself
reduced real wages on the average by 3-4%. But there was an additional impact of almost
the same magnitude due to the reduced rents from protection. She also finds that her
estimated sharing rule for rents from protection is associated with the share of non-produc-
tion workers among total workers, which she interprets as a measure of skills that are in
short supply so that they increase workers' bargaining power. Except for this measure,
however, she does not incorporate heterogeneity in workers into her analysis even though
it would seem that such heterogeneity in itself could account for deviations in firm wages

41There has been increasinginterest in characterizingthe changes in distribution and in labor market outcomes
that occurredwith international trade liberalization. These characterizations suggestheterogeneityin experiences,
with generally fairly quick unemployment adjustments and lowered dispersion between low- and higher-skill
wages in the earlier East Asian experience but increasing dispersion in the more recent Latin American experi-
ences. See Horton et al. (1991, 1994) and Wood (1997) and the references therein.
Ch. 43: Labor Markets in Developing Countries 2919

lrom industry levels even in the absence of protection. She also finds a significant negative
effect of reducing import quotas on employment, but no significant employment effect of
reduced tariffs. She interprets these results as being consistent with unions being able to
capture part of the rents generated by tariffs so these rents adjust rather than employment if
tariffs fall, but unions do not capture part of the rents generated by quota protection. But
this just pushes the question back a step - why can unions capture rents from tariff but not
from quota protection?
Currie and Harri son (1997) examine the labor market impact of Moroccan international
trade reform during the mid-1980s - including the virtual elimination of quantitative
restrictions on imports and a reduction of the maximum tariff from 165 to 45% over a
6-year period. They present an explicit model of an imperfectly competitive Cournot firm
in an industry in which domestic and imported goods are imperfect substitutes, there is an
industry-level quota on imports, and firms face upward-sloping labor supply curves. They
derive employment relations that depend on trade policy in part because they posit that
both the extent of market power and productivity (in the form of Hick's neutral techno-
logical change) depend on import tariffs and quotas. They note that trade liberalization can
reduce labor demand due to falling output prices but may increase labor demand due to
increased productivity and lessened market power. They also note that public-sector
enterprises may be constrained in their ability to reduce their labor force and that exporters
are more likely to reduce their labor demands the more distinct are products that they sell
in domestic markets from those that they export. They use time-series data on all Moroc-
can manufacturing firms (except those with less than 10 employees or with annual sales
less than US $11,000) for 1984-1989. They find that on the average employment was not
affected by the trade reforms, but firms in the most affected sectors and exporting firms
reduced employment significantly (3.5-6% in response to 21-24% decline in tariffs). They
explore why most private domestic-market-oriented firms did not adjust employment
significantly; they find that this does not reflect high adjustment costs due to labor market
regulation, but that these firms absorbed the loss in rents from the loss of protection
through reduced profits.
A number of questions remains unexplored in these studies - what is the impact on the
rest of the labor force including the informal manufacturing sector and all of services and
agriculture, what are the effects of entry and exit into the formal manufacturing sector,
what are the implications of workers being heterogeneous, what are the implications of the
fact that policies are choices and not clearly exogenous changes that are predetermined in a
statistical sense, what is the nature of dynamic processes, what is the impact of imperfect
information and of other aspects of incomplete markets? But these studies are examples of
a growing number of studies of major reforms in developing countries - with much more
substantial policy changes than in the much more-studied developed economies - that
should increasingly illuminate our understanding of how labor markets function.
2920 ~ R. Behrman

6. Distribution and mobility

Distribution is of interest as an objective for society that is separate from, although inter-
related with efficiency, producfion and growth - often taking the form of concern about the
poorer members of society. Distribution and mobility are intertwined. Differences in
human capital investments for otherwise identical individuals that yield differences in
their labor market (or other) returns thereby yield differences in the distribution of labor
market returns. These distributional differences, at least within the standard human capital
model that is summarized at the start of Section 4, may create incentives for human capital
investments associated with actual or potential job mobility. A central feature of devel-
opment, in fact, is the relocation of labor from less to more productive activities.
Such relocations may or may not require geographical movements. New products may
be produced or new technologies may be adopted, for example, that result in the realloca-
tion of labor to more productive activities without any geographical movement. On the
other hand, scale and conglomeration economies and limited or missing markets for other
factors may mean that labor is reallocated to more productive activities substantially by
geographical migration. There has been some, but relatively little attention in the devel-
opment literature to labor reallocations that do not involve geographical movements.
There has been substantial attention to labor reallocations that involve geographical
mobility, particularly in the form of rural-urban migrations. In most developing countries,
for example, there have been substantial migratory movements from rural to urban areas,
as well as smaller movements among rural areas and from urban to rural areas.
The human capital model of migration simply states that it pays to migrate from one
location to another particular location if the present discounted value of benefits exceeds
the present discounted value of costs and the gain is larger for that move than for any other.
The model suggests, therefore, that migration is more attractive for individuals for whom
time horizons are longer (e.g., because they are younger), who have lower discount rates
(because of taste heterogeneities or more education), who are better informed (if there is
risk aversion and insurance market imperfections), who are more adaptable (younger?
more educated?), who are better prepared to deal with up-front costs (if there are capital
market imperfections), who have less immobile capital (in some contexts, land, or loca-
tion-specific production knowledge) and who gain more from diversification (if there are
risks that are not perfectly correlated over space and the individual is linked by family or
other relations across space). Such migration tends to be equilibrating by reducing differ-
entials at the margin across areas through shifting, for example, homogenous labor from
areas in which wages are low to those in which wages are high. While this model is
presented usually with reference to migration, it clearly refers to any form of investment
in mobility.
In this section, I consider several dimensions of distribution and mobility that have
received attention in the development literature and in some cases have led to new analy-
tical approaches in labor economics.
Ch. 43: Labor Markets in Developing Countries 2921

6.1. Intertemporal aspects of distribution and the Kuznets hypothesis of inverted U pattern
in inequality with development

Because of data limitations, most empirical studies of distribution in developing countries

are based on cross-sectional annual data. Recently there have been several studies for
developing countries with intertemporal approaches to characterizing or modeling distri-
bution, some of which are now reviewed.
Kuznets (1955) hypothesized that at very low levels of income, distribution would have
to be relatively equal because of subsistence minimums, but with the process of develop-
ment distribution initially would become more unequal because those best suited by virtue
of their human capital (or lucky by virtue of their asset ownership) first would grasp new
income-earning opportunities while most members of society initially would be left
behind. But as opportunities increased with the process of development, increasing
proportions of society would have new income-earning opportunities, so eventually
inequality would decline. 4z Therefore there is an inverse U-shaped relation between
inequality and development.
This "Kuznets curve" relating to the hypothesized inverse U relation between inequal-
ity and development has received a lot of attention in the applied development literature.
Early cross-sectional studies seemed consistent with it. More recent longitudinal studies
(e.g., Anand and Kanbur, 1993) have not found support for such a relation. Fields and
Jakubson (1997) show that cross-sectional estimates that appear to support the existence of
such a relation are reversed once there is control for country fixed effects - and thus a
paradox between the previous cross-section and time-series results is resolved. They claim
that the cross-sectional estimates suggest support for the Kuznets curve even though the
time-series estimates do not because inequality is greater in Latin America than in Asia,
but the latter includes both lower and higher per capita income countries than the former. 43
While most explorations of this hypothesis have focused on aggregate data, there are
obvious implications of the hypothesis that, if human resources are not very mobile inter-
nationally, labor market returns to human capital that are in relatively scare supply as
development increases initially increase, but as more human investment is induced by
new opportunities the returns to such investments decline. Knight and Sabot (1983) explore
this possibility using cross-sectional data from Kenya and Tanzania. They find that the
expansion of schooled workers reduces the returns to schooling (the "compression effect")
and therefore reduces intraurban wage inequality more than the initial increased schooling

42 There may be other major factors that affect the income distribution as well. For example, Becker et al.
(1990) suggest that their may be multiple equilibria, with high growth options in which the cost of human capital
formation is low due to the large stock of human capital, so there are large human capital investments with
relatively low (and equal) returns.
43 Ravallion and Chen (1997) consider distributional changes in 64 spells for developing and transitional
countries based on matched household surveys in 1981-1994. They find no significant association between
growth and distribution if Eastern Europe and Central Asia are excluded, and a significantly positive relation
between growth and equality if Eastern Europe and Central Asia are included. But they do not control for the
initial level of development, so their characterization is only tangentially related to Kuznets' hypothesis.
2922 Z R. Behrman

dispersion (the "composition effect") increases wage inequality. This exploration must be
qualified, however, due to the limitation of the sample to the major urban areas in the two
countries considered and the dependence on simple earnings functions with no investiga-
tion of the implications of endogenous schooling decisions.
Deaton and Paxson (1994) build on Eden's (1980) point that standard models of inter-
temporal choice imply that for a given birth cohort earnings inequality (and therefore
consumption inequality) grows over the lifecycle because of the changing impact of the
integral of accumulated shocks. They show that this result is more general than the certainty
equivalence assumptions required to justify the random walk consumption as is implicit in
Eden's paper, and that empirical examination of whether consumption dispersion increases
with age can be used to test alternative consumption theories (since if liquidity constraints
are effective consumption dispersion will track income dispersion). They then examine the
experience in Taiwan and in two developed economies for constructed data on age-cohorts
of individuals using a succession of cross-sectional surveys over recent 11-15 year time
periods and find in all three cases within-cohort consumption becomes substantially more
dispersed over time with similar rates of dispersion - which they claim is evidence against
models of perfect insurance - and that within-cohort dispersion of earnings increases with
age. They note that inequality increases with age have a number of implications: (1) with no
links between generations, with constant inequality in earnings distributions, and with
roughly stable population distributions across birth cohorts, inequality for each birth cohort
is consistent with approximately constant inequality in society as a whole (as observed in
the US for extended periods of time); (2) but if there were strong intergenerational links
through bequests such approximately constant aggregate inequality would not be observed
(which they claim is evidence against the extreme forms of dynastic models); and (3)
countries with rapid demographic transitions and aging populations (e.g., Taiwan and a
number of other Asian developing and developed economies) would be expected to have
increasing inequality, an additional reason beyond that suggested in the well-known
Kuznets (1955) hypothesis for initially increasing and then falling inequality in the devel-
opment process. This is an interesting study that combines in a fruitful way a number of
tools from consumption theory, distribution theory, econometrics, and data analysis to
develop some new insights, although the results have to be qualified because of limited
attention to human capital investments (in their role as intergenerational links that may be
alternatives to bequests in addition to their role in affecting earning dispersions given
heterogeneity in such investments over the lifecycle), to endogenous changes in household
structures, and to the nature of the structural relations underlying the interesting descrip-
tions of intertemporal distributions presented.
Deolalikar and Gaiha (1993) is the first empirical characterization of which I am aware
of whether households in poor areas of developing countries that are beneath the poverty
line in 1 year tend to below the poverty line transitorily or permanently. They use ICRI-
SAT panel data over a decade from rural south India. They calculate alternative measures
of the percentage of households below the poverty line based on actual average poverty in
annual cross-sections (57%), expected average poverty based on estimates of income as a
Ch. 43: Labor Markets in Developing Countries 2923

function of observed characteristics (62%), "innate poverty" based on estimates with

observed characteristics and unobserved characteristics (12%), ever-below the poverty
line (88%), and always below the poverty line (21%). They first show the sensitivity of
the estimation of the percentages of households below the poverty line to be quite sensitive
to the definition and time period that are used, and that the proportion of permanent poor is
much lower than the number of transitory poor, although still quite large. This paper is
basically descriptive, but it is important because it describes an important phenomenon
about which there has been much speculation but almost no prior evidence.
Foster and Rosenzweig (1997) raise an important but essentially ignored question in the
analysis of income inequality trends, namely that of defining the observational unit. If
households are taken to be the relevant unit, as is often the case, a complete understanding
of the evolution of income inequality requires an understanding of how households are
formed and dissolved, and specifically how household formation and dissolution are
affected by economic growth. Measures of the relationship between economic growth
and income inequality that are based on repeated cross-sections cannot properly address
this issue because individuals cannot be matched across households over time. In this
paper empirical estimates of the impact of economic growth on income inequality are
obtained using longitudinal data on rural Indian households during the green revolution
period. The empirical analysis is based on the structural estimation of a behavioral model
of a farm household in which the existence of a public good and differences in agricultural
productivity among household members (heads of separate but related nuclear families)
provide a rationale for joint co-residence. In the Indian agricultural setting, the initiation of
the green revolution (exogenous technological change) altered both the income potential
of nuclear and joint households and the incentives for co-residence. The model is para-
meterized, structural parameters are estimated and counterfactual simulations are
performed. The results of these exercises show that while technical change had only a
small effect on the distribution of incomes of dynasties (of households defined by their pre-
green revolution composition), the green revolution increased income inequality of house-
holds (defined contemporaneously).

6.2. Geographical mobility

The form of mobility that is most emphasized in the development literature is geographical
migration, particularly between rural and urban areas. Much of the empirical literature
simply documents some of the basic implications of the basic human capital model of
migration, such as that wage differentials induce such migration selectively, more so for
young adults and for more schooled individuals. Behrman and Birdsall (1983), for exam-
ple, note that most of what appears to be selectivity on unobserved characteristics for
migration in Brazil largely disappears if there is control for migration selectivity not only
on schooling attainment but also on school quality. 44 Despite considerable migration,
44But Robinsonand Tomes (1983) find that the returns to migrationin Canada are significantlyoverstated if
migration selectivity on unobservedability is not controlled, althoughthey do not controlfor schoolingquality.
2924 ~ R. Behrman

however, wage differentials appear to persist for long periods of time between urban and
rural areas, although such comparisons have to be qualified because of problems in
assuring that homogenous workers are being compared and prices are being held constant.
Not all migrants into urban areas, moreover, obtain jobs in the high-paying modern or
formal sector.

6.2.1. Harris-Todaro migration model

To reconcile ongoing migratory flows with persistent urban-rural wage discrepancies
Todaro (1969) modified the basic human capital model of migration to incorporate
employment risk, and Harris and Todaro (1970) incorporated this migration relation
into an influential two-sector economy-wide model of migration, wage and employment
determination. In this model the rural labor market is assumed to function competitively,
but the urban wage is set institutionally (e.g., governmental minimum wage) above the
initial rural wage with the probability of employment in the urban sector equal to the
number of urban jobs at the institutionally-set wage relative to the number of urban job-
seekers. Therefore, with no costs for migration, migration occurs until the rural wage
equals the expected urban wage, which is the product of the employment probability
times the urban wage and the urban labor market distortion causes a misallocation of
workers across sectors and urban unemployment. If the institutionally-set urban wage is
increased and urban labor demand is inelastic, employment and output falls in both
sectors, urban unemployment increases less than proportionately to the urban wage
increase so the expected urban wage increases, and rural-urban migration occurs. If the
institutionally-set urban wage is increased and urban labor demand is elastic, employment
and output fall in the urban sector and rise in the rural sector and urban-rural migration
occurs. Whatever the urban labor demand elasticity, an urban wage subsidy with a fixed
urban wage induces rural-urban migration, reduced rural output and employment, and
increased urban unemployment.
Some aspects of the model are troublesome. First, if the urban-rural wage differential is
50-100% as has been alleged (although empirical comparisons do not seem to control well
for price and skills differentials), observed urban unemployment rates are far too low for
the equilibrium predicated by this model. For that reason, some (e.g., Fields, 1975) have
added an informal urban sector in which wages are much lower than in the modern sector
covered by the institutionally-set wage. But the difficulties in measuring the pure wage
return in the informal sector mentioned in Section 5.1 mean that empirical evidence on this
resolution is fuzzy. Second, the ad hoc exogenous minimum wage is troublesome because
governmental policies are the result of behavioral decisions that are not incorporated into
the model and because minimum wages do not appear to be binding in many cases (i.e.,
Bell, 1997 that is summarized in Section 5.2; Squire and Narueput, 1997).

6.2.2. Stiglitz labor "turnover" and "efficiency wage" models

Stiglitz (1976) has proposed one resolution to the latter problem by positing that mono-
polistically competitive urban firms incur hiring and training costs with labor turnover, so
Ch. 43: Labor Markets in Developing Countries 2925

they pay a wage p r e m i u m over alternative wage rates to reduce turnover that is an inverse
function o f the urban unemployment rate. In this m o d e l urban unemployment is optimal in
the sense that the lost output due to unemployment is less than the output gain from lower
turnover costs. Stiglitz (1982) also develops other models in which the wage premium that
a firm offers determines what quality of workers that a firm can hire ("efficiency wage
worker quality m o d e l " ) and the absolute wage offered by the firm determines the work
effort of workers ("efficiency wage effort model"). In these "efficiency wage" models
unemployment persists in equilibrium. 45 As Rosenzweig (1988a) emphasizes, the optim-
ality result in the Stiglitz "turnover" model depends critically on the maintained assump-
tion that workers do not share in the cost of training and also implies that workers' wages
do not rise over their work lives (because they neither share in costs nor receive benefits
from training) - contrary to lifecycle wage patterns observed in many developing and
developed economies. Likewise the unemployment result in the Stiglitz efficiency wage
models depends on the lack of alternative contractual or sorting arrangements that mini-
mize shirking or sort workers optimally and, again, systematic empirical evidence on the
critical behavioral relations does not exist.

6.2.3. Empirical studies o f households, migration and risk

The two-sector models described so far in this section contrast with the emphasis in
Section 2 on households being central to much rural decision making and on the critical
role of incomplete markets, including those for capital and insurance, in determining
household behavior. Nor can these models explain temporary (or seasonal) migration or
remittances, both o f which are widespread in many developing countries. Since the mid-
1980s there have been several empirical papers that focused on the role of households and
risk-sharing in the absence of insurance markets as critical for understanding some impor-
tant aspects of migration in developing countries.
Lucas and Stark (1985) is the first study to investigate temporary migration and remit-
tances within a household context. They use a national survey of households in Botswana
and find evidence consistent with (1) temporary migration is in part an insurance arrange-
ment with higher remittances home if home household incomes suffer negative shocks, (2)
bequest prospects have a positive impact on who migrates and on how much is remitted,
and (3) remittances in part are returns for prior household investments in schooling.
Within the limitations of cross-sectional data, this study is suggestive of how households
and incomplete markets affect migration and remittances, although it is difficult to identify

45Others also have emphasized the importance of efficiency wage models to explain urban dualism in devel-
oping countries. Esfahani and Salehi-Isfahani(1989), for example, argue that under plausible conditions tile urban
dualism can be explained by differential observability of effort, which has the advantage over other explanations
in the literature of integrating a number of related stylized facts (e.g., difference in formal-informal sector wage
rates, unobserved productivity, unemployment, technology, factor intensity, operational size and management).
They argue that worker effort is less observablein the formal sector (because of larger enterprise size, greater task
complexity, and more extensive management structure) but under a plausible condition on the marginal disutility
of effort with respect to effort lower observability results in more effort, greater productivity, and higher wages.
2926 J. R. Behrman

some of the above possibilities from others (e.g., households that receive greater remit-
tances accumulate more wealth) without longitudinal data. Other subsequent studies have
examined further the relations among households, migration and incomplete markets -
e.g., Behrman et al. (1999), Foster (1996), Rosenzweig (1988b,c), Rosenzweig and Stark
(1989). 46 Rosenzweig (1988b), for instance, provides a formal framework for examining
some "transaction cost" issues related to the family. The model explains the geographical
pattern of marriages and the pattern of intrahousehold transfers as a response to the need to
smooth consumption in the absence of market insurance or state income maintenance
schemes. Daughters tend to "marry out" of their natal village, and the daughters in
each family are spread out geographically, as they would be if families were attempting
to diversify in the face of weather related risk, rather than concentrated geographically, as
they would be if the principal determinant of marriage destinations were the cost of
obtaining information about the marriage market.

6.3. Empirical investigation of urban informal-formal mobility

Perhaps the most emphasized aspect of developing-country urban labor markets in the
empirical literature, as noted in Section 5.1, is the possibility of dualism. But as also noted
there, most empirical studies of these markets have been limited to comparisons of wages
or wage functions, with a number of problems of interpretation.
Funkhouser (1997b) contributes to this literature by using a survey from E1 Salvador
with retrospective data that permits investigation of patterns over time in employment. A
model is set up with different wage functions by sectors, including worker unobserved
fixed effects and random effects with serial correlation. Changes in sectors occur between
periods if the second-period random error is large enough to offset any sector-specific
human capital accumulation and the serially-correlated component of the error. This
general framework then is used to consider alternative modeling/estimation strategies
regarding the number of periods and the existence of serial correlation - with insights
regarding what can be deduced (and, perhaps more importantly, what cannot be deduced)
regarding the causes of sectoral wage differentials under different assumptions. The multi-
period model, for example, permits additional insights beyond the usual single-period
framework by observing the wage changes of movers between the sectors, but if there
is serial correlation there still may be asymmetries under perfect competition of the sort
predicted by the segmented model (i.e., larger wage gains for those moving from the free
to the limited-access sector than vice versa). Examination of transition rates reveals
considerable mobility, particularly for males. Careful examination of wage change rela-
tions for various individuals reveals some interesting patterns: (1) high mobility; (2)
greater mobility for workers who are male, younger and less educated; (3) higher earnings
and higher earnings growth for workers continuously employed in the formal sector, but
highest earnings for those with some informal sector attachment for those who maintain

46The first two of these are summarizedin the discussion on parental schooling in Section 4.3 above.
Ch. 43: Labor Markets in Developing Countries 2927

formal sector jobs rather than move in or out of the informal sector; and (4) asymmetry in
the sense that movers from the informal to the formal sector have higher wage gains than
vice versa. These patterns provide some support for segmentation having an influence
(particularly 4), but this support is limited because of the considerable mobility (particu-
larly for males), the larger earnings gains for those with some informal sector attachment
for those who maintain informal sector positions, and the problem of identifying symmetry
due to segmentation from that due to serial correlation. This study contributes to the
literature by extending the perspective on modeling segmented labor markets in develop-
ing countries and lbcusing on wage changes and mobility rather than wage levels as in the
previous literature.
Pradhan and van Soest (1997) also consider informal-formal sector mobility in a sense.
They model female and male labor supplies within a static household utility maximization
framework similar to that used for a number of studies in developed economies, but with
an extension to include formal and informal sectors with both wage and non-monetary
differences (with the latter determining selection between the two sectors). Estimates for
two-adult households in urban Bolivia indicate: (i) substantial intrahousehold effects, with
elasticities in line with those reported for the United States; (ii) intersectoral wage respon-
siveness; and (iii) that non-monetary returns are greater in the formal than in the informal
sector. The identification of the sectoral work decision, however, seems to reflect an
arbitrary exclusion restriction of the non-monetary returns from the utility maximization
process that leads to the labor supply relations.

6.4. Distributional differences among demographic groups

In most societies there are some demographic groups for whom mean wages are lower,
even if there is control for observed differences in observed characteristics such as in years
of schooling and age. Examples include women in almost every society, indigenous
groups in Latin America and the Caribbean and most minority tribes and low castes in
Asia. Empirical studies largely have focused on possible differences in standard wage
relations (such as are discussed at the start of Section 4.1) as possible sources for these
wage differences.
Some suggest, for example, that as a result of these lower wages, the rates of return for
investing in the human resources of members of such groups on the average are lower than
for investing in men, members of the dominant group and of higher castes. But the leap
from low wages to low rates of return does not necessarily follow. Investments in indivi-
duals with low wages can have relatively high rates of return. This may be so for human
capital investments that take the individual's time, such as schooling and training, since
ceteris paribus low wages mean that the opportunity cost of time for such investments (a
major input into such investments) is relatively low. Also in some societies average wage
gaps conditional on schooling appear to narrow with more schooling - which may reflect a
number of factors that change with more schooling including greater labor force integra-
tion, more emphasis on intellectual rather than physical attributes, and lessening discri-
2928 Z R. Behrman

ruination. Moreover, if the proportion of females and minority groups 47 that receives
higher levels of schooling and training is relatively low, if the distributions of innate
abilities and motivation axe the same across demographic groups, and if those individuals
who receive such human capital investments tend to have relatively high abilities and
motivations, the average ability and motivational levels of women o1" of minority group
members with higher levels of schooling and training will exceed the average ability and
motivational levels of men or majority groups members with the same levels of schooling
and training. Further, a greater proportion of the returns to human resources investments in
women and some minority groups may be in informal sector, family enterprises, and
household production activities that often are not incorporated into rate of return analyses
in tbe same way that are returns in terms of labor market activities.
There have been some efforts to explore whether rates of returns to human resources,
particularly schooling, differ among demographic groups, with more-or-less standard
wage relations. On the basis of micro data, for example, Behrman and Deolalikar
(1993, 1995) report that estimated rates of return to schooling in Indonesia are higher
for females than for males for schooling above the primary level in estimates that control
for unobserved household and community effects, Psacharopoulos (1993) reports that
estimated rates of return to schooling are lower for schooling for indigenous peoples
than for those of European descent in Bolivia and Guatemala, Schultz (1993a,b) finds
little evidence of selectivity differences between males and females in Thailand, Birdsall
and Sabot (1991) present a series of studies that investigate differences in schooling
returns among groups identified by gender, ethnicity, and caste, and Horton (1996)
presents estimates of declining gender wage residual differentials in seven Asian econo-
mies. There also have been a few studies that examine possible gender differences in
productivity of health and nutrition that are reviewed in Section 4.2. As noted there,
apparently because of gender specialization in tasks, these studies tend to find greater
impact of better nutrition/health on productivities or wages of males than females.
A few studies, many of which are reviewed in some detail above, have gone beyond
estimation of standard wage relations to explore for rural areas of developing countries the
nature of gender differences in labor markets or in intrahousehold allocations that may
lead to gender differences in labor markets (and other outcomes, e.g., Sen, 1990) and the
role of marriage markets in providing insurance in risky environments with missing
insurance markets. For example, Haddad and Kanbur (t990) report that individual distri-
butions are more unequal than would be suggested by aggregation to the household level
because of gender differences, Behrman (1988a,b) and Behrman and Deolalikar (1989a,b;
1990) find that intrahousehold allocations favor males in agricultural production stages in
which food is relatively scarce and expensive and the productivity impact of nutrition/
health is relatively great for males due to gender specialization in tasks, Pitt et aL (1990)
find that calorie allocations that prima facie may appear to favor males actually tax males

47I use this term to refer to demographicgroupsthat are thought to be disadvantaged in a population even
though in some cases they may constitute a majority of the population (e.g., indigenouspeople in Bolivia).
Ch. 43: Labor Markets in Developing Countries 2929

to benefit of other household members given differential energy expenditures related to

gender task specialization, Foster and Rosenzweig (1994a,b) find that gender wage differ-
entials reflect statistical discrimination (and not taste discrimination of employers) based
on lower distributions of unobserved productivities for females than for males and imper-
fect information, Deolalikar and Rose (1998) find that the "gender shock" with the birth of
a daughter rather than a son effectively reduces household wealth and induces subsequent
increases in time devoted to labor, Jacoby (1995) finds that demands for wives in a
polygamous society depends on their agricultural productivity, Rosenzweig (1988b,c)
finds that exogamous marriages serve to provide insurance through diversifying risk,
and Udry (1996a) finds intra agricultural household inefficiencies in the distribution of
agricultural inputs across plots controlled by men and women.

7. Conclusions

Modeling of labor markets for developing countries has been distinguished from labor
economics more generally by differences in degrees of market completeness and in insti-
tutions, not by differences in kind. Earlier modeling of the development process often
made extreme assumptions about the nature of labor and other markets, particularly in
rural areas in which most people lived in developing economies. At one extreme, for
example, some influential models assume that such markets did not function, with the
result that there was "surplus labor" in agriculture that could productively be moved to
industry without a loss in agricultural output (e.g., Lewis, 1954; Ranis and Fei, 1961; Sen,
1966). At the other extreme, some influential models of reallocation of labor from agri-
culture to industry assume that rural labor markets approximately perfectly competitive
markets, although there are significant rigidities or information problems in urban markets
(e.g., Harris and Todaro, 1970; Stiglitz, 1974, 1982). In between, still other influential
models assumed that rural labor markets existed, but had rigid wages and substantial
unemployment because of nutrition efficiency wages.
The primary contribution of labor economics for developing countries in recent years
has been to develop and to test empirically tractable models that center on dynamic
behaviors of rural households within the context of some incomplete or missing markets
(e.g., insurance, information) and a range of more-or-less good substitutes for these
markets. These studies have led to a much better empirically-grounded understanding
of how such households function and what are the implications for efficiency and for
distribution of various market imperfections and what possible policy interventions
might have high payoffs. Not only have they been informative about economic behaviors
in the developing country context, but in a number of cases they have been informative
about basic labor economics issues that also are important in developed economies but
which are much more difficult to examine empirically in those economies because of data
problems and institutional differences and complexities. A few examples include the
nature and impact of moral hazard in labor contracts, the range of possibly micro adjust-
2930 J . R . Behrman

m e n t s to s h o c k s , t h e n a t u r e o f i n t r a h o u s e h o l d a l l o c a t i o n s , a n d t h e m u l t i p l i c i t y o f f u n c t i o n s
of households.

References

Alderman, H. (1993), "New research on poverty and malnutrition: what are the implications for research and
policy?" in: M. Lipton and J, Van der Gaag, eds., Including the poor (The World Bank, Washington, DC).
Alderman, H., J.R. Behrman, D. Ross and R. Sabot (1996a), "Decomposing the gender gap in cognitive skills in a
poor rural economy", Journal of Human Resources 31: 229-254.
Alderman, H., J.R. Betmnan, D. Ross and R. Sabot (1996b), "The returns to endogenous human capital in
Pakistan's rnral wage labor market", Oxford Bulletin of Economics and Statistics 58: 29-55.
Alderman, H., J.R. Behrman, V. Lavy and R. Menon (1997), "Child nutrition, child health, and school enroll-
ment: a longitudinal analysis", Policy research working paper no. 1700 (The World Bank, Washington, DC).
Anand, S. and S.M.R. Kanbur (1993), "Inequality and development: a critique", Journal of Development
Economics 41: 19-44.
Angrist, J.D. and V. Lavy (1997), "The effect of a change in language of instruction on the returns to schooling in
Morocco", Journal of Labor Economics 15: $48-$76.
Banerjee, A.V. and A.F. Newman (1993), "Occupational choice and the process of development", Journal of
Political Economy 2: 274-299.
Banerji, A. and H. Ghanem (1997), "Does the type of political regime matter for trade and labor market policies",
The World Bank Economic Review 11: 171-194.
Bardhan, K. (1984), "Work patterns and social differentiation: rural women of West Bengal", in: H.P. Binswan-
get and M.R. Rosenzweig, eds., Contractual arrangements, employment, and wages in rural labor markets in
Asia (Yale University Press, New Haven, CT) pp. 184-208.
Bardhan, K. (1993), "Women and rural poverty in some Asian cases", in: M.G. Quibria, ed., Rural poverty in
Asia: priority issues and policy options (Oxford University Press, Hong Kong) p. 316-367.
Bardhan, P.K. (1979a), "Wages and unemployment in a poor agrarian economy: a theoretical and empirical
analysis", Journal of Political Economy 87: 479-500.
Bardhan, P.K. (1979b), "Labor supply functions in a poor agrarian economy", American Economic Review 69:
73-83.
Bardhan, P.K. (1983), "Labor tying in a poor agrarian economy: a theoretical and empirical analysis", Quarterly
Journal of Economics 98:501-514.
Bardhan, P.K. (1984a), "Determinants of supply and demand for labor in a poor agrarian economy: an analysis of
household survey data from rural West Bengal", in: H.P. Binswanger and M.R. Rosenzweig, eds., Contractual
arrangement, employment, and wages in rural labor markets in Asia (Yale University Press, New Haven, CT)
pp. 242-262.
Bardhan, P.K. (1984b), Land, labor, and rural poverty (Oxford University Press, Delhi, India).
Barros, R.P. de (1992), "Human resources in the process of adjustment: Brazil", Mimeo. (IDB Research Network
on "Human Resources in the Adjustment Process" Instituto de Pesquisa Economica Aplicada Rio de Janeiro,
Brazil).
Becker, G.S. (1967), "Human capital and the personal distribution of income: an analytical approach",
Woytinsky lecture (University of Michigan, Ann Arbor, MI); (republished in G.S. Becker (1975), Human
capital, 2nd edition (NBER, New York) pp. 94-117).
Becker, G.S., K.M. Murphy and R. Tamura (1990), "Human capital, fertility, and economic growth", Journal of
Political Economy 98: S12-$37.
Behrman, J.R. (1988a), Nutrition, health, birth order and seasonality: intrahousehold allocation in rural India",
Journal of Development Economics 28: 43-63.
Ch. 43: Labor Markets in Developing Countries 2931

Behiman, J.R. (1988b), "Intrahousehold allocation of nutrients in rural India: are boys favored? Do parents
exhibit inequality aversion?" Oxford Economic Papers 40: 32-54.
Behrman, J.R. (1990a), Human resource led development? (ARTEPflLO, New Delhi).
Behrman, J.R. (1990b), The action of hmnan resources and poverty on one another: what we have yet to learn
(Population and Human Resources Department, World Bank, Washington, DC).
Behrman, J.R. (1994b), "Health and nutrition are also productive human resources", in: N. Birdsall and R. Sabot,
"Virtuous circles: human capital, growth and equity in East Asia", Mimeo. (World Bank, Washington, DC).
Behrman, J.R. (1997a), "Intrahousehold distribution and the family", in: M.R. Rosenzweig and O. Stark, eds,
Handbook of population and family economics (North-Holland, Amsterdam) pp. 107-168.
Behrman, J.R. (1997b), "Women's schooling and child education: a survey", Mimeo. (University of Pennsylva-
nia, Philadelphia, PA).
Behrman, J.R. and N. Birdsall (1983), "The quality of schooling: quantity alone is misleading", American
Economic Review 73: 928-946.
Behrman, J.R. and N. Birdsall (1988), "The reward for good timing: cohort effects and earnings functions for
Brazilian males", Review of Economics and Statistics 70: 129-135.
Behrman, J.R. and A.B. Deolalikar (1987), "Will developing country nutrition improve with income? A case
study for rural south India", Journal of Political Economy 95:108 138.
Behrman, LR. and A.B. Deolalikar (1988), "Health and nutrition", in: H.B. Chenery and T.N. Srinivasan, eds.,
Handbook on economic development, Vol. 1 (North-Holland, Amsterdam) pp. 631-711.
Behrman, J,R. and A.B. Deolalikar (1989a), "Seasonal demands for nutrient intakes and health status in rural
south India", in: D.E. Sahn, ed., Causes and implications of seasonal variability in household food secmJty
(The Johns Hopkins University Press, Baltimore, MD) pp. 66-78.
Behrman, J.R. and A.B. Deolalikar (1989b), "Wages and labor supply in rural India: the role of health, nutrition
and seasonality", in: D.E. Sahn, ed., Causes and implications of seasonal variability in household food
security (The Johns Hopkins University Press, Baltimore, MD) pp. 107-118.
Behrman, J.R. and A.B. Deolalikar (1990), "The intrahousehold demand for nutrients in rural south
India: individual estimates, fixed effects and permanent income", Journal of Human Resources 25: 665-
696.
Behrman, J.R. and A.B. Deolalikar (1991), "School repetition, dropouts and the returns to schooling: the case of
Indonesia", Oxford Bulletin of Economics and Statistics 53: 467-480.
Behrman, J.R. and A.B. Deolalikar (1993), "Unobserved household and community heterogeneity and the labor
market impact of schooling: a case study for Indonesia", Economic Development and Cultural Change 41:
461-488.
Behrman, J.R. and A.B. Deolalikar (1995), "Are there differential returns to schooling by gender? The case of
Indonesian labor markets", Oxford Bulletin of Economics and Statistics 57:97-118.
Behrman, J.R. and V. Lavy (1997), "Child health and schooling achievement: Association, cansality and house-
hold allocations", Mimeo. (University of Pennsylvania, Philadelphia, PA).
Behrman, J.R. and J.C. Knowles (1999), "Household income and child schooling in Vietnam", World Bank
Economic Review, in press.
Behrman, J.R. and M.R. Rosenzweig (1994), "Caveat emptor: cross-country data on education and the labor
force", Journal of Development Economics 44: 147-172.
Behrman, J.R. and B.L. Wolfe (1984), "The socio-economic impact of schooling in a developing country",
Review of Economics and Statistics 66: 296-303.
Behrman, J.R., N. Birdsall and A.B. Deolalikar (1995), "Marriage markets, labor markets and unobserved human
capital: an empirical exploration for southcentral India", Economic Development and Cultural Change 43:
585-602.
Behrman, J.R., A. Foster, and M.R. Rosenzweig (1997), "The dynamics of agricultural production and the
calorieincome relationship: evidence fl'mn Pakistan", Journal of Econometrics 77: 187-207.
Beln-man, J.R., A. Foster, M.R. Rosenzweig and P. Vashishtha (1999), "Women's schooling, home teaching, and
economic growth", Journal of Political Economy, in press.
2932 ~ R. Behrman

Bell, C. (1977), "Alternative theories of sharecropping: some tests using evidence from northeast India", Journal
of Development Studies 13: 317-346.
Bell, C. (1988), "Credit markets and interlinked transactions", in: H. Chenery and T.N. Srinivasan, eds., Hand-
book of development economics (North-Holland, Amsterdam) pp. 763-830.
Bell, C. and T.N. Srinivasan (1989), "Interlinked transactions in rural markets: an empirical study of Anda
produce, Bihar and Punjab", Oxford Bulletin of Economics and Statistics 51: 73-83.
Bell, C. and C. Sussangkarn (1985), "The choice of tenancy contract", Mimeo. (The World Bank, Washington,
De).
Bell, L.A. (1997), "The impact of minimum wages in Mexico and Colombia", Journal of Labor Economics 15:
S102-S135.
Ben-Porath, Y. (1980), "The F-connection: families, friends, and firms and the organization of exchange",
Population and Development Review 6:1 30.
Benjamin, D. (1992), "Household composition, labor markets, and labor demand: testing for separation in
agricultural household models", Econometrica 60: 287-322.
Besley, T. and A. Case (1994), "Diffusion as a learning process: evidence from HYV cotton", Mimeo. (Princeton
University, Princeton, NJ).
Binswanger, H.P. (1980), "Attitudes toward risk: experimental measurement in rural India", American Journal of
Agricultural Economics 62: 395-407.
Binswanger, H.P., K. Deininger and G. Feder (I 995), "Power, distortions, revolt and reform in agricultural land
relations", in: J.R. Behrman and T.N. Srinivasan, eds., Handbook of development economics, Vol. 3B (North-
Holland, Amsterdam) pp. 2659-2772.
Binswanger, H.P. and M.R. Rosenzweig (1986), "Behavioral and material determinants of production relations in
agriculture", Journal of Development Studies 22:503 539.
Birdsall, N. and R.H. Sabot eds., (1991), Unfair advantage: labor market discrimination in developing economies
(World Bank, Washington, DC).
Blau, D., J.R. Behrman and B.L. Wolfe (1988), "Schooling and earnings distributions with endogenous labor
force participation, marital status, and family size", Economica 55: 297-316.
Bliss, C. and N. Stern (1978), "Productivity, wages and nutrition; parts I and II', Journal of Development
Economics 5: 33-198.
Boissiere, M., J.B. Knight and R.H. Sabot (1985), "Earnings, schooling, ability and cognitive skills", American
Economic Review 75: 1016-1030.
Bouis, H.E. (1994), "The effect of income on demand for food in poor countries: are our databases giving us
reliable estimates?" Journal of Development Economics 44: 199-226.
Bouis, H.E. and L.J. Haddad (1992), "Are estimates of calorie-income elasticities too high? A recalibration of the
plausible range", Journal of Development Economics 39: 333-364.
Chayanov, A.V. (1925), Peasant farm organization (Moscow: Cooperative Publishing House, Moscow); trans-
lated in D. Thorner, B. Kerbaly, and R.E.F. Smith, eds. (1966), A.V. Chayanov: the theory of peasant
economy (Richard Irwin, Homewood).
Cragg, M.I. and M. Epelbaum (1996), "Why has wage dispersion grown in Mexico? Is it the incidence of reforms
or the growing demand for skills?" Journal of Development Economics 51:99-116.
Currie, J. and A. Harrison (1997), "Sharing the costs: the impact of trade reform on capital and labor in
Morocco", Journal of Labor Economics 15: $44-$71.
Dasgupta, P. and D. Ray (1986), "Inequality as a determinant of malnutrition and unemployment: theory",
Economic Journal 96:1011-1034.
Dasgupta, P. and D. Ray (1987), "Inequality as a determinant of malnutrition and unemployment: policy",
Economic Journal 97: 177-188.
Deaton, A. and C. Paxson (1994), "Intertemporal choice and inequality", Journal of Political Economy 102: 437-
467.
Deolalikar, A.B. (1988), "Nutrition and labor productivity in agriculture: estimates for rural south India", Review
of Economics and Statistics 70: 406-413.
Ch. 43: Labor Markets in Developing Countries 2933

Deolalikar, A.B. (1993), "Gender differences in the returns to schooling and school em'olhnent rates in Indone-
sia", Journal of Human Resources 28: 899-932.
Deolalikar, A.B. and R. Gaiha (1993), "Persistent, expected and innate poverty: estimates for semi-arid rural
south India, 1975-84", Cambridge Journal of Economics 17: 4 0 9 4 2 1 .
Deolalikar, A. and E. Rose (1998), "Gender, savings and production in rural India", Journal of Population
Economics, in press.
Diwan, I. and M. Walton (1997), "How international exchange, technology, and institutions affect workers: an
introduction", The World Bank Economic Review 11: 1-16.
Eden, B. (1980), "Stochastic dominance in human capital", Journal of Political Economy 88: 135-145.
Esfahani, H.S, and D. Salehi-lsfahani (1989), "Effort observability and worker productivity: towards an explana-
tion of economic dualism", Economic Journal 99: 1-19.
Eswaran, M. and A. Kotwal (1985a), "A theory of two-tier labor markets in agrarian economies", American
Economic Review 75: 162-177.
Eswaran, M. and A. Kotwal (1985b), "A theory of contractual structure in agriculture", American Economic
Review 75: 352-367.
Eswaran, M. and A. Kotwal (1986), "Access to capital and agrarian production organization", Economic Journal
96: 4 8 2 4 9 8 .
Fafchamps, M. (1993), "Sequential labor decisions under uncertainty: an estimable household model of west
African farmers", Econometrica 61:1173-1198.
Fafchamps, M. and A.R. Quisumbing (1997), "Human capital, productivity, and labor allocation in rural Paki-
stan", Mimeo. (International Food Policy Research Institute, Washington, DC).
Fafchamps, M., C. Udry and K. Czukas (1998), "Drought and saving in West Africa: are livestock a buffer
stock?" Journal of Development Economics, in press.
Fallon P.R. and R.E.B. Lucas (1993), "Job security regulations and the dynamic demand for industrial labor in
India and Zimbabwe", Journal of Development Economics 40: 241-276.
Feder, G. (1985), "The relation between farm size and farm productivity", Journal of Development Economics
18: 297-313.
Fields, G.S. (1975), "Rurahirban migration, urban unemployment and underemployment, and job search activity
in LACS", Journal of Development Economics 2: 165-187.
Fields, G.S. (1979), "Place to place migration: some new evidence", Review of Economics and Statistics 61: 21-
32.
Fields, G.S. and G.S. Jakubson (1997), "New evidence on the Kuznets curve", Mimeo. (CorneU University,
Ithaca, NY).
Foster, A.D. (1996), "Analysis of household behavior when households choose their members: MARRIAGE-
market selection and human capital allocations in rural Bangladesh", Mimeo. (University of Pennsylvania,
Philadelphia, PA).
Foster, A.D. and M.R. Rosenzweig (1993a), "Information, learning, and wage rates in low-income rural areas",
Journal of Human Resources 28: 759-790.
Foster, A.D. and M.R. Rosenzweig (1993b), "Information flows and discrimination in labor markets in low-
income countries", in: Proceedings of the World Bank Annual Conference on Development Economics
(World Bank Economic Review and World Bank Research Observer) pp. 173-203.
Foster, A.D. and M.R. Rosenzweig (1994a), "Information, learning, and wage rates in rural labor markets",
Journal of Human Resources 28: 759-790; reprinted in: T. Paul Schultz, ed., Investment in women's human
capital (University of Chicago Press, Chicago IL) pp. 138-170.
Foster, A.D. and M.R. Rosenzweig (1994b), "A test for moral hazard in the labor market: contractual arrange-
ments, effort, and health", Review of Economics and Statistics 76: 213-227.
Foster, A.D. and M.R. Roseuzweig (1995), "Learning by doing and learning from others: human capital and
technical change in agriculture", Journal of Political Economy 103:1176-1209.
Foster, A. and M.R. Rosenzweig (1996a), "Comparative advantage, information and the allocation of workers to
tasks: evidence from an agricultural labor market", Review of Economic Studies 63: 347-374.
2934 ~ R. Behrman

Foster, A.D. and M.R. Rosenzweig (1996b), "Technical change and human-capital returns and investments:
evidence from the green revolution", American Economic Review 86: 931-953.
Foster, A.D. and M.R. Rosenzweig (1997), "Household division, inequality and rural economic growth", Mimeo.
(University of Pennsylvania, Philadelphia, PA).
Foster, A.D., M.R. Rosenzweig and J.R. Behrman (1997), "Population growth, income growth and deforestation:
management of village common land in India", Mimeo. (University of Pennsylvania, Philadelphia, PA).
Funkhouser~ E. (1997a), "Demand-side and supply-side explanations for barriers to labor market mobility
in developing countries: the case of Guatemala", Economic Development and Cultural Change 45: 341-
366.
Funkhouser, E. (1997b), "Mobility and labor market segmentation: the urban labor market in El Salvador",
Economic Development and Cultural Change 46:123-154.
Gavian, S. and M. Fafchamps (1996), "Land tenure and allocative efficiency", American Journal of Agricultural
Economics 78: 4 6 0 4 7 1 .
Glewwe, P. (1996), "The relevance of standard estimates of rates of return to schooling for education policy: a
critical assessment", Journal of Development Economics 51: 267-290.
Glewwe, P. and H. Jacoby (1994), "Student achievement and schooling choice in low-income countries: evidence
from Ghana", Journal of Human Resources 29: 842-864.
Glewwe, P. and H. Jacoby (1995), "An economic analysis of delayed primary school enrollment and childhood
malnutrition in a low-income country", Review of Economics and Statistics 77: 156-169.
Gruber, J. (1997), "The incidence of payroll taxation: evidence from Chile", Journal of Labor Economics 15:
$72--S 101.
Haddad, L. and R. Kanbur (1990), "How serious is the neglect of intrahousehold inequality?" Economic Journal
100: 866-881.
Haddad, L. and H. Bouis (1991), "The impact of nutritional status on agricultural productivity: wage evidence
from the Philippines", Oxford Bulletin of Economics and Statistics 53: 45-68.
Haddad, L., J. Hoddinott and H. Alderman, eds. (1996), Intrahousehold resource allocation: methods, models, and
policy (For the International Food Policy Research Institute, The Johns Hopkins University Press, Baltimore,
MD).
Harris, J.R. and M.P. Todaro (1970), "Migration, unemployment, and development", American Economic
Review 60: 126-142.
Heckman, J.J. and V.J. Hotz (1986), "The sources of inequality for males in Panama's labor market", Journal of
Human Resources 21 : 507-542.
Horton, S., ed. (1996), Women and industrialization in Asia (Routledge, London).
Horton, S., R. Kanbur and D. Mazumdar (1991), "Labour markets in an era of adjustment: evidence from 12
developing countries", International Labour Review 130: 531-558.
Horton, S., R. Kanbur and D. Mazumdar (1994), Labor markets in an era of adjustment (World Bank, Washing-
ton, DC).
Jacoby, H. (1992), "Productivity of men and women and the sexual division of labor in peasant agriculture of the
Peruvian sierra", Journal of Development Economics 37: 265-288.
Jacoby, H. (1993), "Shadow wages and peasant family labor supply: an econometric application to the Peruvian
sierra", Review of Economic Studies 60: 903-921.
Jacoby, H. (1995), "The economics of polygyny in sub-Saharan Africa: female productivity and the demand for
wives in C6te d'Ivoire", Journal of Political Economy 103: 938-971.
Jacoby, H. and E. Skoufias (1997), "Risk, financial markets and human capital in a developing country", Review
of Economic Studies 64:311-335.
Jimenez, E. (1995), "Human and physical infrastructure: public investment and pricing policies in developing
countries", in: J.R. Behrman and T.N. Srinivasan, eds., Handbook of development economics, Vol. 3B
(North-Holland, Amsterdam) pp. 2773-2844.
King, E.M.(1990), "Does education pay in the labor market? W o m e n ' s labor force participation, occupation, and
earnings in Peru", LSMS Working paper no. 67, Mimeo. (World Bank, Washington, DC).
Ch. 43: Labor Markets in Developing Countries 2935

King, E.M. and M.A. Hill, eds. (1993), Women's education in developing countries: barriers, benefits, and
policies (For the World Bank, The Johns Hopkins University Press, Baltimore, MD).
Kingdon, G.G. (1997), "Does the labor market explain lower female schooling in India?" Paper no. 1 (Devel-
opment Economics Research Programme, London School of Economics, London).
Kingdon G.G.(1996), "Student achievement and teacher pay: a case-study of India", Paper no.74 (Development
Economics Research Programme, London School of Economics, London).
Knight, J.B. and R.M. Sabot (1981), "The returns to education: increasing with experience or decreasing with
expansion", Oxford Bulletin of Economics and Statistics 43: S1-$72.
Knight, J.B. and R.H. Sabot (1983), "Educational expansion and the Kuznets effect", American Econonfic
Review 73:1132 1136.
Knight, J.B. and R.H. Sabot (1987a), "Educational expansion, govermnent policy and wage compression",
Journal of Development Economics 26: 201-221.
Knight, J.B. and R.H. Sabot (1987b), "Educational policy and labor productivity: An output accounting exer-
cise", Economic Journal 385: 199-214.
Knight, J.B. and R.H. Sabot (1990), Educational productivity and inequality: the East African natural experiment
(Oxford University Press, New York).
Krueger, A.B. (1996), "International labor standards and trade", in: M. Bruno and B. Pleskovic, eds., Annual
World Bank conference on development economics (The World Bank, Washington, DC) pp. 281-316.
Kuznets, S. (1955), "Economic growth and income inequality", American Economic Review 45: 1-28.
Lain, D. and D. Levison (1991), "Declining inequality in schooling in Brazil and its effect on inequality in
earnings", Journal of Development Economics 37: 199-226.
Lain, D. and R.F. Schoeni (1993), "Effects of family background on earnings and returns to schooling: evidence
from Brazil", Journal of Political Economy 101: 710-740.
Lam, D. and R.F. Schoeni (1994), "Family ties and labor markets in the United States and Brazil", Journal of
Hmnan Resources 29: 1235-1258.
Lewis, W.A. (1954), "Economic development with unlimited supplies of labor", Manchester School of
Economic and Social Studies 22: 139-191.
Leibenstein, H.A. (1957), Economic backwardness and economic growth (Wiley, New York).
Lipton, M. (1983), Poverty, undernutrition and hunger (The World Bank, Washington, DC) pp. 35-49.
Lopez, R. (1986), "Structural models of the farm household that allow for interdependent utility and profit-
maximization decisions", in: I. Singh, L. Squire and J. Strauss, eds., Agricultural household models: exten-
sions, applications and policy (For the World Bank, The Johns Hopkins University Press, Baltimore, MD) pp.
306-326.
Lucas, R.E.B. (1988), "On the mechanics of economic development", Journal of Monetary Economics 21: 3-42.
Lucas, R.E.B. and O. Stark (1985), "Motivations to remit: evidence from Botswana", Journal of Political
Econmny 93: 901-918.
Macisaac, D. and M. Rama (1997), "Determinants of hourly earnings in Ecuador: the role of labor market
regulations", Journal of Labor Economics 15: S136~S165.
Manser, M. and M. Brown (1980), "Marriage and household decision-making: a bargaining analysis", Interna-
tional Economic Review 21:31-44.
Mazumdar, D. (1959), "The marginal productivity theory of wages and disguised unemployment", Review of
Economic Studies 26: 190-197.
Mazumdar, D. (1976), "The urban informal sector", World Development 4: 655-679.
Mazumdar, D. (1981), The urban labor market and income distribution: a study of Malaysia (Oxford University
Press, London).
McElroy, M.B. (1990), "The empirical content of Nash-bargained household behavior", Journal of Human
Resources 25: 559-583.
McElroy, M.B. and M.J. Homey (1981), "Nash-bargained household decisions: toward a generalization of the
theory of demand", International Economic Review 22: 333-347.
Mincer, J.B. (1974), Schooling, experience, and earnings (NBER, New York).
2936 J. R. Behrman

Mirrlees, J. (1975), "A pure theory of underdeveloped economies", in: L. Reynolds, ed., Agriculture in devel-
opment theory (Yale University Press, New Haven, CT).
Moll, P.G. (1992), "Quality of education and the rise in returns to schooling in South Africa, 1975-1985",
Economics of Education Review 11: 1-10.
Moll, P.G. (1993), "Industry wage differentials and efficiency wages: a dissenting view with South African
evidence", Journal of Development Economics 41: 213-246.
Moll, P.G. (1996), "The collapse of primary schooling returns in South Africa, 1960-1990", Oxford Bulletin of
Economics and Statistics 58: 185-209.
Newbery, D. (1975), "The choice of rental contracts in peasant agriculture", in: L. Reynolds, ed., Agriculture in
development theory (Yale University Press, New Haven, CT).
Newman, J., S. Jorgensen and M. Pradhan (1991), "How did workers benefit from Bolivia's emergency social
fund?" World Bank Economic Review 5: 367-393.
Newman, J.L. and P.J. Gertler (1994), "Family productivity, labor supply and welfare in a low income country",
Journal of Human Resources 29: 989-1026.
Otsuka, K., H. Chuma and Y. Hayami (1992), "Land and labor contracts in agrarian economies", Journal of
Economic Literature 30: 1965-2018.
Pans, E.A. and M.D. Robinson (1997), "The implications of increasing economic openness for real wages in
developing countries, 1973-90", World Development 25: 537-547.
Pissarides, C.A. (1997), "Learning by trading and the returns to human capital in developing countries", The
World Bank Economic Review 11: 17-32.
Pitt M.M. and M.R. Rosenzweig (1986), "Agricultural prices, food consumption and the health and productivity
of farmers", in: I. Singh, L. Squire and J. Strauss, eds., Agricultural household models: extensions, applica-
tions, and policy (The World Bank, Washington, DC) pp. 153-182.
Pitt, M.M. and M.R. Rosenzweig (1990), "Estimating the behavioral consequences of health in a family context:
the intrafamily incidence of infant illness in Indonesia", International Economic Review 31: 969-989.
Pitt, M.M., M.R. Rosenzweig and D.M. Gibbons (1993), "The determinants and consequences of the placement
of government programs in Indonesia", The World Bank Economic Review 7:319-348.
Pitt, M.M., M.R. Rosenzweig and M.N. Hassan (1990), "Productivity, health and inequality in the intrahousehold
distribution of food in low-income countries", American Economic Review 80:1139-1156.
Pollak, R.A. (1985), "A transaction cost approach to families and households", Journal of Economic Literature
23: 581-608.
Pradhan, M. and A. Van Soest (1997), "Household labor supply in urban areas of Bolivia", Review of Economics
and Statistics 79:300-310.
Psacharopoulos, G. (1993), Ethnicity, education and earnings: a comparative analysis of Bolivia and Guatemala
(World Bank, Washington, DC).
Psacharopoulos, G. (1994), "Returns to investment in education: a global update", World Development 22:
1325-1344.
Psacharopoulos, G., S. Morley, A. Fiszbein, H. Lee and B. Wood (1992), Poverty and income distribution in Latin
America: the story of the 1980s (World Bank, Washington, DC).
Quisumbing, A.R. (1996), "Male-female differences in agricultural productivity: methodological issues and
empirical evidence", World Development 24: 1579-1596.
Ranis, G. and J. Fei (1961), "A theory of economic development", American Economic Review 56: 533-558.
Ravallion, M. and S. Chen (1997), "What can new survey data tell us about recent changes in distribution and
poverty", World Bank Economic Review 11: 357-382.
Revenga, A. (1997), "Employment and wage effects of trade liberalization: the case of Mexican manufacturing",
Journal of Labor Economics 15: $20-$43.
Richards, A. (1994), "The Egyptian farm labor market revisited", Journal of Development Economics 43: 239-
262.
Roberts, M.J. and E. Skoufias (1997), ""The long-run demand for skilled and unskilled labor in Colombian
manufacturing plants", Review of Economics and Statistics 79: 330-334.
Ch. 43: Labor Markets in Developing Countries 2937

Robinson, C. and N. Tomes (1983), ""Self-selection and inteFprovineial migration in Canada", Canadian Journal
of Economics 15: 474-502.
Rose, E. (1995), "Ex-ante and ex-post labor supply responses to risk in a low income area", Mimeo. (University
of Washington, Seattle, WA).
Rose, E. (1999a), "Consumption smoothing and excess female mortality in rural India", Review of Economics
and Statistics, in press.
Rose, E. (1999b), "Gender bias, credit constraints and time allocation in rural India", Economic Journal, in press.
Rosenzweig, M.R. (1978), "Rural wages, labor supply, and land reform: a theoretical analysis", American
Economic Review 68: 847-861.
Rosenzweig, M.R. (1980), "Neoclassical theory and the optimizing peasant: An econometric analysis of market
labor supply in a developing country", Quarterly Journal of Economics 94:31-56.
Rosenzweig, M.R. (1984), "Determinants of wage rates and labor supply behavior in the rural sector of a
developing country", in: H.P. Biuswanger and M.R. Rosenzweig, eds., Contractual arrangement, employ-
ment, and wages in rural labor markets in Asia (Yale University Press, New Haven CT).
Rosenzweig, M.R. (1988a), "Labor markets in low-income countries", in: H. Chenery and T.N. Srinivasan, eds.,
Handbook of development economics (North-Holland, Amsterdaln) pp. 713-762.
Rosenzweig, M.R. (1988b), "Risk, implicit contracts and the family in rural areas of lowincome countries",
Economic Journal 98: 1148-1170.
Rosenzweig, M.R. (1988c), "Risk, private information, and the family", American Economic Review 78:245
250.
Rosenzweig, M.R. (1995), "Why are there returns in schooling?" American Economic Review 85: 153-158.
Rosenzweig, M.R. and R.E. Evenson (1977), "Fertility, schooling and the economic contribution of children in
rural India", Econometrica 45: 1065-1079.
Rosenzweig, M.R. and K.J. Wolpin (1985), "Specific experience, household structure, and intergenerational
transfers: farm family land and labor arrangements in developing conntries", Quarterly Journal of Economics
100: 961-987.
Rosenzweig, M.R. and K.J. Wolpin (1986), "Evaluating the effects of optimally distributed public programs",
American Economic Review 76: 470-487.
Rosenzweig, M.R. and K.J. Wolpin (1988), "Migration selectivity and the effects of public programs", Journal of
Public Economics 37: 265-289.
Rosenzweig, M.R. and T.P. Schultz (1982), "Market opportunities, genetic endowments, and intrafamily
resource distribution: child survival in rural India", American Economic Review 72: 803-815.
Rosenzweig, M.R. and O. Stark (1989), "Consumption smoothing, migration, and marriage: evidence from rural
India", Journal of Political Economy 97: 905-926.
Rust, J. (1987), "Optimal replacement of GMC bus engines: an empirical model of Harold Zurcher", Econome-
trica 55: 999-1033.
Ryan, J.G. and R.D. Ghodake (1984), "Labor market behavior in rural villages in south India: effects of season,
sex and socioeconomic status", in: H.P. Binswartger and M.R. Rosenzweig, eds., Contractual arrange-
ment, employment, and wages in rural labor markets in Asia (Yale University Press, New Haven, CT) pp.
169-183.
Sabot, R.H. (1977), "The meaning and measurement of urban surplus labor", Oxford Economic Papers 29:389
411.
Sahn, D.E. and H. Alderman (1988), "The effect of human capital on wages, and the determinants of labor supply
in a developing country", Journal of Development Economics 29: 157-184.
Sahn, D.E. and H. Alderman (1996), "The effect of food subsidies on labor supply in Sri Lanka", Economic
Development and Cultural Change 45: 125-146.
Schaffner, J.A. (1995), "Attached farm labor, limited horizons and servility', Journal of Development Economics
47: 241-270.
Schultz, T.P. (1988), "Education investments and returns", in: H. Chenery and T.N. Srinivasan, eds., Handbook
of development economics (North-Holland, Amsterdam) pp. 543 630.
2938 Z R. Behrman

Schultz, T.P. (1990a), "Women's changing participation in the labor force: a world perspective", Economic
Development and Cultural Change 38: 457-488.
Schultz, T.P. (1990b), "Testing the neoclassical model of family labor supply and fertility", Journal of Hulnan
Resources 25: 599-634.
Schultz, T.P. (1993a), "Returns to women's education", in: E.M. King and M.A. Hill, eds., Women's education
in developing countries: barriers, benefits, and policies (For the World Bank, The Johns Hopkins University
Press, Baltimore, MD) pp. 51-99.
Schultz, T.P. (1993b), "Investments in the schooling and health of women and men: quantities and returns",
Journal of Human Resources 28: 694--734.
Schultz, T.P. and A. Tansel (1998), "Wage and labor supply effects of illness in Cote D'lvoire and Ghana:
instrumental variable estimates for days disabled", Journal of Development Economics, in press.
Schultz, T.W. (1975), "The value of the ability to deal with disequilibria", Journal of Economic Literature 13:
827-846.
Sen, A. (1966), "Peasants and dualism with and without surplus labor", Journal of Political Economy 74: 425-
450.
Sen, A.K. (1990), "More than 100 million women are missing", New York Review of Books 37: 61-66.
Shaban, R. (1987), "Testing between competing models of sharecropping", Journal of Political Economy 95:
893-920.
Shapiro, D. (1990), "Farm size, household size and composition, and women's contribution to agricultural
production: evidence from Zaire', Journal of Development Studies 27: 1-21.
Singh, I., L. Squire and J. Strauss, eds. (1986), Agricultural household models: extensions, applications, and
policy (The World Bank., Washington, DC).
Skoufias, E. (1993a), "Seasonal labor utilization in agriculture: theory and evidence from agrarian households in
India", American Journal of Agricultural Economics 75: 20-32.
Skoufias, E. (1993b), "Labor market opportunities and intrafamily time allocation in rural households in south
Asia", Journal of Development Economics 40: 277-310.
Skoufias, E. (1994a), "Market wages, family composition and file time allocation of children in agricultural
households", Journal of Development Studies 30: 335-360.
Skoufias, E. (1994b), "Using shadow wages to estimate labor supply of agricultural households", American
Journal of Agricultural Economics 76: 215-227.
Skoufias, E. (1996), "Intertemporal substitution in labor supply: micro evidence from rural India", Journal of
Development Economics 51: 217-237.
Soon, L.Y. (1987), "Selfemployment vs wage employment: estimation of earnings functions in LACS", Econom-
ics of Education Review 6: 81-89.
Squire, L. and S.S. Narueput (1997), "The impact of labor market regulations", The World Bank Economic
Review 11: 119-144.
Srinivasan, T.N. (1972), "Farm size and productivity: implications of choice under uncertainty", Sankhya: The
Indian Journal of Statistics, Series B 34: 409420.
Srinivasan, T.N. (1994), "Data base for development analysis: an overview", Journal of Development Economics
44: 3-26.
Stiglitz, J.E. (1976), "The efficiency wage hypothesis, surplus labor, and the distribution of income in LDC's",
Oxford Economic Papers, New Series 28: 185-207.
Stiglitz, J.E. (1974), "Alternative theories of wage determination and unelnployment in LACS: the labor turnover
model", Quarterly Journal of Economics 88: 194-227.
Stiglitz, J.E. (1982), "Alternative theories of wage determination and unemployment: the efficiency wage
model", in: M. Gersovitz, C. Diaz-Alejandro, G. Ranis and M,R. Rosenzweig, eds., The theory and experi-
ence of economic development (George Alien and Unwin, London).
Stokey, N.L. (1991), "Human capital, product quality, and growth", Quarterly Journal of Economics 106: 587-
616.
Strauss, J. (1986), "Does better nutrition raise farm productivity?" Journal of Political Economy 94: 297-320.
Ch. 43: Labor Markets in Developing Countries 2939

Strauss, J. and D. Thomas (1995), "Human resources: empirical modeling of household and family decisions", in:
J.R. Behrman and T.N. Srinivasan, eds., Handbook of development economics, Vol. 3A (North-Holland,
Amsterdam) pp. 1883-2024.
Strauss, J. and D. Thomas (1996), "Wages, schooling and background: investments in men and women in urban
Brazil", in: N. Birdsall and R.H. Sabot, eds., Opportunity foregone: education in Brazil (For the Inter-
American Development Bank, The Johns Hopkins University Press, Baltimore, MD) pp. 147-191.
Subramanian, S. and A. Deaton (1996), "The demand for food and calories", Journal of Political Economy 104:
133-162.
Swamy, A.V. (1997), "A simple test of the nutrition-based efficiency wage model", Journal of Development
Economics 53: 85-98.
Tan, H. and G. Batra (1997), "Technology and finn size-wage differentials in Columbia, Mexico, and Taiwan
(China)", The World Bank Economic Review 11: 59-84.
Thomas, D. (1990), "Intrahousehold resource allocation: an inferential approach", Journal of Human Resources
25: 635-664.
Thomas, D. (1993), "The distribution of income and expenditure within the household", Annales de Economie et
de Statistiques 29:109-136.
Thomas, D. (1994), "Like father, like son; like mother, like daughter: parental resources and child height",
Journal of Human Resources 29: 950-989.
Thomas, D. and J. Strauss (1997), "Health and wages: evidence on men and women in urban Brazil", Journal of
Econometrics 77: 159-187.
Thomas, D., J. Strauss and M.H. Henriques (1991), "How does mother's education affect child height?" Journal
of Human Resources 26:183-211.
Todaro, M. (1969), "A model of labor migration and urban unemployment in less developed countries", Amer-
ican Economic Review 59: 138-148.
Udry, C. (1994), "Risk and insurance in a rural credit market: an empirical investigation in northern Nigeria",
Review of Economic Studies 61: 495-526.
Udry, C. (1996a), "Gender, agricultural production and the theory of the household", Journal of Political
Economy 104: 1010-1046.
Udry, C. (1996b), "Efficiency and market structure: testing for profit maximization in African agriculture",
Mimeo. (Northwestern University Evanston, IL).
van der Gaag, J. and W. Vijverberg (1988), "A switching regression model of wage determinants in the public
and private sectors of a developing country", Review of Economics and Statistics 70: 244-252.
Vijverberg, W.P.M. (1993), "Educational investments and returns for women and men in Crte D'Ivoire", Journal
of Human Resources 28: 933-974.
Welch, F. (1970), "Education in production", Journal of Political Economy 78: 35-59.
Wolpin, K.I. (1984), "An estimable dynamic stochastic model of fertility and child mortality", Journal of
Political Economy 92: 852-874.
Wolpin, K.I. (1987), "Estimating a structural search model: the transition from school to work", Econometrica
55:801 818.
Wood, A. (1997), "Openness and wage inequality in developing countries: the Latin American challenge to east
Asian conventional wisdom", The World Bank Economic Review 11: 33-58.
World Bank (1981), World development report, 1981 (World Bank, Washington, DC).
World Bank (1995), World development report 1995: workers in an integrating world (Oxford University Press,
New York).
World Bank (1997), World development report 1997: the state in a changing world (Oxford University Press,
New York).
Yamanchi-Kawana, F. (1997), "Information, neighborhood effects and investment in human capital: learning
schooling returns in a dynamic context", Mimeo. (University of Pennsylvania, Philadelphia, PA).