Intrahousehold Access to Land and Sources of Inefficiency:

Theory and Concepts

Marcel Fafchamps, †
Stanford University

May 1998

Abstract1

This chapter examines the allocation of land and other productive resources within rural

households of poor countries. Building upon the existing literature, it provides a consistent

framework from which to study productive efficiency and intrahousehold equity. The topics dis-

cussed include: returns to scale and household centralization; specialization and gender casting;

separate spheres and commitment failure; and labor market cartelization and discrimination.

Implications for intrahousehold access to land are discussed in detail. We show that intrahouse-

hold productive inefficiency should not arise unless household members are prevented from

entering into enforceable contracts with each other. Our analysis predicts that intrahousehold

inefficiency increases with factors that exacerbate commitment failure such as short time horizon,

low assets, unequal stakes in the household, and poor external enforcement. Granting plots of

land to dependents appears to be a response to commitment problems. Patrimonial laws and cus-

toms regarding inheritance and divorce can similarly be understood as efforts to mitigate commit-

ment failure within the household although, possibly, at the expense of women and young adults.

________________
† Department of Economics, Stanford University, Stanford, CA 94305-6072. Tel.: (650) 723-3251.
[email protected].
1 I benefitted from comments on earlier drafts from Agnes Quisumbing, Lawrence Haddad, Chris Udry, Elisabeth
Sadoulet, and participants to WIDER conferences on access to land held in May 1997 in Helsinki and April 1998 in
Santiago de Chile.
 Intrahousehold Access to Land and Sources of Inefficiency:

Theory and Concepts

The literature on women and development has documented women’s lack of access to land,

credit, and farm inputs in many developing countries (e.g., Boserup (1970), Gladwin and McMil-

lan (1989), von Braun and Webb (1989)). It also have provided evidence of gender casting and

other restrictions on female labor (e.g., Brown and Haddad (1995), Fafchamps and Quisumbing

(1998)). This situation has been identified as a source of inequality between men and women, and

as encouraging abuse (e.g., Jones (1983, 1986)). In addition, it has been claimed to be inefficient

(e.g., Jiggins (1989), Udry (1996b)). In response to these claims, aid agencies have set up target-

ted programs to deliver credit and inputs to women and, to a lesser extent, to open their access to

land. The purpose of this paper is to propose a framework that integrates efficiency and equity

issues and to identify conditions under which intrahousehold allocation of resources can be

inefficient.

Implicit in much of the lay literature is the assumption that women’s unequal access to land

is a source not only of inequality but also of inefficiency. For instance, it is often claimed that

poor countries cannot develop unless women are granted equal access to productive resources

(e.g., Boserup (1970)).2 The rationale behind these claims is not entirely clear, however.3 The

main objective of this paper is to clarify the debate by identifying conditions under which ine-

quality is indeed a source of inefficiency. To do so, we must distinguish between access to land

within households and across households. Although the two are often discussed interchangeably,
________________
2 Calls for better female access to land have been strongest in Sub-Saharan Africa (Gladwin and McMillan
(1989)). This is somewhat surprising because, relative to other regions of the world such as South Asia or Latin
America, women in Sub-Saharan Africa have a comparatively easier -- though by no means equal -- access to land.
3 Some claim that women produce ’all the food’ and thus should be given more land (e.g., Gladwin and McMillan
(1989), Jiggins (1989)). Jiggins (1989), for instance, claims that in Sub-Saharan Africa ’[women] produce
approximatively 60% of the food that is consumed by rural households’. It is unclear how women could produce this
much food if they do not have access to land. Others argue that women should be given better access to credit so that
they can hire enough manpower to cultivate their fields. Egyir (1998), for instance, points out that Ghanaian women
who farm independently ’[hire] male labor to take on strenuous activities’. But the fact that these women must hire
outsiders to farm suggests that they have more land than they can cultivate -- i.e., that they were given too much land
for an efficient allocation of factors of production, not too little.
 2

conceptually they are quite separate. Access to land for male and female dependents is deter-

mined within the household; access to land for female headed and female operated households is

determined outside the household.4 The processes for allocating resources vary between the two.

Allocation within the household resembles what takes place within a firm. Most households

have a hierarchical structure, with much of the control over household resources such as land and

capital concentrated into the hands of the head (e.g., Haddad, Hoddinott and Alderman (1997),

Egyir (1998)). Conflicts over the division of welfare between dependents and the head are some-

what reminiscent of labor disputes between workers and employers -- and potentially as bitter

and unpredictable. In addition to a hierarchical structure, households often institute an explicit

division of labor together with systems of delegation of authority (e.g., Cleave (1974), Egyir

(1998), Shapiro (1990), Fafchamps and Quisumbing (1998)). This is particularly true of rural

households in poor economies where self-provision of food, shelter, water, fuel, child and elderly

care, and most services is the rule -- and thus where the tasks undertaken by households are both

numerous and complex.

In contrast, the allocation of land across households typically takes either the form of gift or

of market transaction, with occasional hybrid combinations of the two (e.g., Egyir (1998), Table

2).5 Land sales, rentals, and sharecropping are market transactions: they involve two-way con-

tractual obligations that are mutually balancing. Inheritance, transfer at marriage, and other

transfers inter vivos are example of allocation mechanisms based on the gift principle. In this

chapter we focus exclusively on intrahousehold allocation; interhousehold allocation is discussed

in other chapters of this volume. As we shall argue in the conclusion, access to land by female

headed and operated households can be fully understood only after making sense of intrahouse-

hold allocation issues. This is important because, the proportion of such households appears to be
________________
4 For the purpose of this chapter, female headed households are defined as households headed by widows,
divorced, and single women; female operated households are those where the male head is absent and participates
little or not at all to household production activities.
5 E.g., rental to a friend at below market price.
 3

on the rise in certain African countries (for Ghana, see Egyir (1998)).

This chapter is organized as follows. After a brief review of the literature, we construct a

model of intrahousehold allocation of labor and land and we derive conditions for allocative

efficiency. We show that, depending on returns to scale, specialization and experience, efficiency

in production may require land and other productive inputs to be centrally managed and individu-

als to specialize in certain tasks or activities. Intrahousehold equity in welfare, however, need not

be achieved. We also propose an explanation for the observed heterogeneity in the allocation of

tasks by age and gender across regions. We then explore conditions under which commitment

failure arise, that is, when household members are unable to pre-commit their future actions and

hence find it difficult to enforce intrahousehold agreements regarding land and labor. In this case,

we show that imperfect factor markets prevent households from achieving allocative efficiency

and that access to land and other productive resources is likely to be conflictual. We argue that

the West African practice of letting female and male dependents cultivate their own plots of land

can be understood as a way of motivating dependents to work harder and of resolving commit-

ment failure between head and dependents. Implications for empirical research are summarized

at the end.

Section 1. A Brief Survey of the Intrahousehold Literature in Economics

Much of the economic literature on intrahousehold issues has focused on the distribution of

consumption among household members (e.g, McElroy and Horney (1981), Manser and Brown

(1980), Haddad and Kanbur (1990), Haddad, Hoddinott and Alderman (1997), Browning et al.

(1994)). Production decisions are often ignored and, with the exception of Becker (1981), the

emphasis is exclusively put on gender issues, with little consideration for intergenerational

issues. Such treatment may suffice for developed countries where small nuclear households dom-

inate, most people work for a wage, and children go to school and leave their parents afterwards.

Indeed, thanks to household equipment and public utilities, the time devoted to household chores
 4

has drastically been reduced. Meal preparation and child care services are increasingly pur-

chased on the market instead of being self-provided.

This fundamental transformation of family life has yet failed to reach much of the Third

World, especially rural areas. There, households are large, occasionally encompassing several

couples with their children. Household members are engaged in a multiplicity of income generat-

ing activities, both on their own account and in collaboration with others. Self-employment is the

rule -- in farming, livestock raising, crafts, trade, and services. Household chores are many and

time-consuming; they fall overwhelmingly on the shoulders of women. Children, whom there are

plenty, spend less time in school and more time helping around the house or the farm. It is cus-

tomary for young adult males to remain on their father’s farm until they are allowed to marry --

and often to stay after marriage in the case of vertically integrated households (e.g., Binswanger

and McIntire (1987), Rosenzweig and Wolpin (1985)). Third World households are thus more

complex than present day households in developed economies. A proper understanding of their

functioning requires a conceptual framework tailored to their characteristics.

Much of the descriptive literature on intrahousehold issues in developing countries -- and

much of social tension and political activism -- focuses on access to productive resources such as

land, capital, and farm inputs (e.g., Kumar (1994), Groote and Kebe (1995), Goetz and Gupta

(1996), Smith (1994)). This is hardly surprising given that the predominant avenue to financial

independence -- and thus to the freedom to choose one’s consumption and privately enjoy it -- is

by engaging in an individually managed income-earning activity such as a farm or business (e.g.,

Egyir (1998)). Third World societies have evolved detailed laws and customs governing the

transfer and use of productive resources such as land and capital, and assign specific tasks and

roles to household members according to their status (e.g., Fafchamps and Quisumbing (1998)).

These rules restrict individual choice and impose financial and work obligations to contribute to

the joint welfare of the household as a group.

 5

Financial independence is usually granted in exchange for assuming specific responsibili-

ties. In many Sahelian villages, for instance, the head of the household manages the main cereal

fields and disposes of the output, but is responsible for feeding the family. Married wives often

are granted small plots that they cultivate on their own, but they must provide the ’sauce’, that is,

the vegetables and condiments that spice up family meals. Unmarried sons may earn pocket

money by cultivating their own field, but they must first work on the household’s main fields.

Tension between household members therefore results regarding the allocation of labor among

competing income-generating activities with distinct residual claimants (e.g., Jones (1983, 1986),

Boe (1996)). Some suspect that tension over labor allocation may even be the source of

economic inefficiency (e.g., Udry (1996a), Balsvik (1995)). Gender inequality in the acquisition

of schooling, experience and skills opening the door to wage employment is also widespread

(e.g., Rosenzweig and Evenson (1977), Adams and He (1995), Alderman and Christi (1990),

Khandker (1988), Jacoby (1992), Filmer (1996)), together with occasional accusations that male

heads of household collude to exclude women or young adults from remunerative activities (see

Kevane (1996), Anderson and Francois (1996)). Domestic violence and social stigma are cited as

additional pressure certain heads of household bring to bear on recalcitrant members. Jones

(1986), for instance, reports that if a wife refuses to work on her husband’s farm, the husband is

allowed by custom to beat her (see also Lilja (1996), Lilja et al. (1996), Kevane (1996)).

While some of the themes that run through the intrahousehold literature in developing

countries are similar to those that animate the feminist discourse in developed economies, the

emphasis on access to productive resources and on the allocation of household labor to indivi-

dual income earning activities is original (e.g., Haddad, Hoddinott and Alderman (1997)). Issues

of welfare equity within Third World household are also not limited to gender distinctions (e.g.,

Haddad and Kanbur (1990, 1993), Kanbur and Haddad (1994), Quisumbing, Haddad and Peña

(1995), Dercon and Krishnan (1997)). They encompass adult dependents, mostly young males.
 6

Early formalization of intrahousehold allocation is couched in terms of a unitary model of

the household (e.g., Becker (1965)). Labor supply response and nutrition outcomes are inter-

preted as results of the maximization of a single utility function. The works of Rosenzweig

(1986), Rosenzweig and Evenson (1977), Pollack and Wachter (1975), Jacoby (1992), and

Skoufias (1993) fall within this category. If applied wisely, this approach can yield interesting

results, even if the underlying assumptions can be criticised as unrealistic. For instance, it can

explain why parents prefer to send boys to school because returns to education is higher than for

girls (e.g. Rosenzweig and Schultz (1987)); why food, when it is scarce, goes to adult males

because they are the prime income earners of the household (e.g., Pitt, Rosenzweig and Hassan

(1990)); or why women concentrate on house chores because their wage outside the household is

lower than that of men (e.g., Evenson (1978), Becker (1965), Low (1994)).

Seminal contributions by Manser and Brown (1980) and McElroy and Horney (1981) chal-

lenged the unitary model of the household and proposed its replacement with a collective model

in which household members bargain over the distribution of consumption among themselves.

They show that the bargaining power of individuals depends on the resources they command, and

thus on their individual incomes. Consumption thus depends not only on the household’s total

income but also on which member earns it. Individual payoffs in case of breakdown of coopera-

tion serve as threat points in the negotiation over household expenditures, thereby helping to

shape the allocation of leisure and consumption among household members. Several possible

determinants of these threat points have been suggested in the literature, such as land ownership,

education, entitlement programs, laws and customs about the sharing of household assets in case

of divorce, and social norms regarding gender-specific tasks and parenting responsibilities (e.g.,

McElroy (1990), Lundberg and Pollack (1993)). Tests of income pooling have usually rejected

the unitary model (e.g., Manser (1993), Thomas (1993)).

Intrahousehold modeling has recently been rejuvenated by the work of Chiappori (1988,
 7

1992, 1993) and coauthors (e.g., Browning et al. (1994), Bourguignon et al. (1993), Bourguignon,

Browning and Chiappori (1995)). Chiappori’s idea is a simple but powerful one. Even though

there is no commonly accepted model of bargaining, most such models discussed in the

economic literature -- whether cooperative or not -- generate Pareto efficient outcomes (as do uni-

tary models which are a special case of collective models; see Haddad, Hoddinott and Alderman

(1997)). Efficient allocations within the household exhibit certain characteristics which can serve

as a basis for a test of allocative efficiency without assuming any particular form of bargaining.

The application of these ideas to household consumption data from developed countries has

failed to reject allocative efficiency while rejecting income pooling (e.g., Browning et al. (1994),

Bourguignon et al. (1993), Thomas (1993), Manser (1993), Fortin and Lacroix (1993)). Using

data from Taiwan, Thomas and Chen (1994) nevertheless reject Pareto efficiency for some alloca-

tions. Taken together, these results constitute evidence in favor of the collective model of the

household, leading Alderman et al. (1995) to urge analysts to think seriously about collective

model alternatives to the unitary model.

Early work by Jones (1983, 1986) and a recent paper by Udry (1996a) nevertheless suggest

that farming households in Africa fail to efficiently allocate productive resources within the

household. Rather than relying on the complicated consumption-based tests proposed by Chiap-

pori and coauthors, Udry (1996a) focuses on the production side and tests whether yields are

equalized across plots managed by different household members. He finds conclusive evidence

that plots in the hands of women and young adult males have lower yields than those operated by

male heads of household. His results and those of Jones, obtained in African farming households

where individual members are known to operate in separate spheres (e.g., Lundberg and Pollack

(1994, 1993), Katz (1994), Carter and Katz (1993)), cast doubt on the efficient collective model

of the household.

The reason for allocative inefficiency is unclear, however. Balsvik (1995) suggests that it
 8

may be due to commitment failure: because the head of the household is unwilling or unable to

credibly commit to reward the work of his wife and dependents after harvest, the latter prefer to

divert their labor to their own fields and income earning activities. Similar ideas permeate the

works of Jones (1983) and Lundberg and Pollack (1994). In a repeated long-run union, the threat

of future non-cooperation should in principle induce voluntary collaboration and solve the com-

mitment failure problem.6 If, however, certain promises are not credible and certain Pareto

improving trades are not feasible, the achievement of allocative efficiency may be prevented.

One of the objectives of this paper is to examine the precise conditions under which a commit-

ment failure may arise. We leave for future work the enforcement difficulties arising from the

presence of multiple wives or nuclear households within the production unit, as in the case of

polygamous or vertically integrated households (e.g., Jacoby (1995)).7

Section 2. A Neo-Classical Model of Intrahousehold Allocation

To capture the essence of the allocative issue, consider the following stylized model of a

collective household operating in separate spheres. The household is assumed comprised of N

members, each of whom is endowed with total time Ti , land Ai 8 and labor efficiency ei --

representing schooling, skills, and experience (e.g., Becker (1981)). For the moment, individual

land endowments and the mapping of individuals into households are taken as exogenously

given. Non-productive members of the household are ignored for simplicity. Individual house-

hold members allocate their time between work on their own account, work for other members of

the household, outside wage work, and leisure. They choose how to spend their income.9
________________
6 Indeed, in many societies, laziness is often an accepted motive for divorce.
7 It is immediately apparent that many concepts developed for firms can be applied to complex household
structures, such as multiple agents models, theory of hierarchies, information circulation issues, etc. Patrimonial law --
e.g., rules of devolution of assets upon death or dissolution -- also becomes more complicated in multiple households.
8 Without loss of generality, variable A can also be thought of as a vector of land qualities and productive assets
i
including farm equipment, working capital, and livestock.
9 In other words, we rule out paternalism, that is, the possibility that a particular household member may decide
the consumption of another member against his or her will. How individual consumption is decided is irrelevant, as
long as individual preferences dictate how money is spent on someone’s consumption bundle. The production and
consumption of household public goods singularly complicates the analysis without adding to our results; it is not
considered here (see, for instance, Bergstrom (1994)).
 9

Individual consumption expenditures are financed through income from own production and

wages, net of transfers to and from other household members.

We begin by assuming that factor markets are totally absent. This assumption is lifted later

on. Formally, the decision problem facing each individual member of the household can be

denoted as follows:

Max U i (ci ) + V i (Ti − ΣLij ) (1)

ci , Lij , gij j

subject to a budget constraint

p ci = F(Ai , Σ e j Lji ) + Σ g ji − Σ gij (2)

j j i

and non-negativity constraints Lij ≥ 0 and gij ≥ 0. Variables ci stands for consumption (possibly a

vector); Lij is the labor time provided by individual i on j’s income earning activity; and gij is the

transfer from individual i to individual j. Utility is assumed non-altruistic and separable between

goods and leisure. Preferences may vary across individuals but all have access to the same pro-

duction technology F(.), which depends on total effective labor Σe j Lji and land Ai . Leisure li
j

equals total time Ti minus total labor ΣLij . When deciding their own labor and consumption,
j

individuals take transfers from others L ji and g ji as given. First order conditions yield the usual

∂U i
____ = λi p (3)
∂ci
∂V i
____ ∂F
____
= λi ei (4)
∂li ∂Li
where Li ≡ ΣL ji . Variable λi is the Lagrange multiplier associated with the budget constraint; it
j

measures the marginal utility of income. Voluntary transfers of labor and money are against

individual’s self-interest since they subtract from own consumption. In equilibrium therefore,

Lij = 0 and gij = 0 for all i ≠ j.

In this world of isolated individuals, production efficiency is not achieved. Individuals with

low land Ai and thus low returns to labor would be better off exchanging labor for consumption
 10

with better endowed individuals who would themselves benefit from hiring other people’s labor.

The absence of market implies that production decisions are a function not only of individual

assets Ai but also of tastes U i (.) and V i (.) and of individual efficiency of labor ei . This result is a

well known feature of unitary household models with missing markets (e.g., Singh, Squire and

Strauss (1986), de Janvry, Fafchamps and Sadoulet (1991)).

Suppose now that a complete competitive market exists in which individuals can freely

exchange labor at the same wage w per unit of effective labor. The production and consumption

choices of individuals operating in their separated spheres becomes:

Max U i (ci ) + V i (Ti − ΣLij ) (5)

ci , Lij j

subject to a budget constraint

p ci = F(Ai , Σ e j Lji ) + Σ w ei Lij − Σ w e j Lji (6)

j j i

and the usual non-negativity constraints Lij ≥ 0. First order conditions yield the usual

∂F(Ai , Li )
_________
w= (7)
∂Li
Labor use and thus output no longer depend on individual preferences and labor efficiency, but

they still are a function of individual land endowments Ai . The wage rate w clears the market for

labor. Labor flows from resource-poor to resource-rich individuals. Everybody is at least as well

off and many are strictly better off, although better endowed individuals continue to enjoy a

higher level of welfare. How gains from trade are divided across buyers and sellers of labor

depends on labor supply and demand elasticities.

The existence of a competitive labor market implies that labor resources are allocated

efficiently within the household as well: since all household members face the same opportunity

cost for effective labor, workers are efficiently allocated to individual production units. There is

no need for intrahousehold bargaining. Intrahousehold labor transactions need not, however, take

the form of an explicit wage contract. Household members may be compensated for the work
 11

they provided to others in an implicit manner, e.g., through gifts or a larger share of household

consumption expenditures. Formally, this situation can easily be represented in our model by let-

ting the size of the gifts from other household members to i be a function of the labor provided to

them, i.e.:

g ji = wei Lij (8)

As long as the ’brownie points’ that individuals gain by helping out other household members are

commensurate to the wage they would earn outside the household, their labor is allocated

efficiently within the household. In equilibrium, if households are sufficiently similar, labor tran-

sactions between households may be infrequent without necessarily implying that labor is not put

to the best use.

The existence of a labor market does not, by itself, ensure productive efficiency, however,

unless Ai is not a vector but a single land input and the production function F (A, L) exhibits con-

stant returns to scale (CRS). In this case, the existence of a competitive labor market ensures that

expected yields are the same for all individuals on all plots -- maintaining our assumption of

common technology. Evidence that yields differ then constitutes an indication that labor markets

are inefficient (e.g., Gavian and Fafchamps (1996), Udry (1996b)). When Ai is a vector of inputs,

however, markets for all factors of production except one must exist for production efficiency to

be achieved. Short of that, output will continue to depend on individual resource endowments.

If production is not CRS, efficiency in production requires that markets exist for all factors

of production.10 Exchange can take a variety of forms (e.g., rentals, sales, or partnerships) which,

for our purpose, are largely irrelevant so long as efficiency is achieved. Assume that it does. What

will efficient production look like? The answer depends on returns to scale. If they are decreas-

ing, the atomisation of production is optimal: each individual should be an independent producer.
________________
10 There are many reasons why production techniques may not be perfectly replicated from field to field and thus
why returns to scale in agriculture may not be constant. All ultimately have to do with the presence of imperfectly
traded inputs, such as credit, management, experience, and rental services of indivisible equipment and infrastructure
(e.g., Feder (1985), Eswaran and Kotwal (1986)).
 12

If they are increasing, all production should be undertaken by a single individual who will buy

and rent all factors from the market. If returns to scale are first increasing then decreasing, the

optimal size of the production unit may require that several, though not all individuals pool their

productive resources together; individual production will be sub-optimal.

These results, although hardly original, seem to have been overlooked in much of the lay

discourse on intrahousehold allocation. Efforts by heads of household to centralize farm manage-

ment and gather control over land and other productive resources are customarily interpreted as a

devious ploy to deprive women of their rights. This may very well be the case, as we shall argue

when we discuss commitment failure. But one should also be open to the possibility that centrali-

zation is in fact efficient, especially, as von Braun and Webb (1989)’s work suggests, when new

technologies are introduced. Financial outlays for fertilizer, improved seeds, and pesticides raise

marginal returns to labor and increase the penalty for untimely performance of agricultural tasks,

especially if the household is risk averse. New agricultural technologies may also raise the

returns to education, and hence the comparative advantage of better educated household heads as

farm managers (e.g., Jamison and Lau (1982)). In these cases, technology adoption is expected to

trigger a concentration of control over household labor into the hands of a single individual, typi-

cally the head of household (e.g., von Braun and Webb (1989)).11 If concentration is efficient,

then it should be possible for household heads to improve everyone’s lot while at the same time

compensating other members for their loss of individual output. From an economic point of view,

such compensation can be regarded as a payment for the use of productive resources controled

by individual household members.

The structure of production is also likely to be affected by returns to experience (e.g.,

Becker (1981)). Suppose that production requires that various tasks be performed, e.g., planting,
________________
11 This of course begs the issue of who in the household is most suitable person to assume management
responsibilities. For instance, if schooling and experience raise effectiveness in management, the better schooled and
most experienced member of the household ought to be the one to take over management duties. Some of these issues
are addressed below.
 13

weeding, harvesting, threshing. Further suppose that, thanks to learning by doing, individuals

with a lot of experience in a certain task k are more proficient in that task, i.e., that:

∞
e tik = gk ( Σ Σ γ s L tijk−s ) (9)
s =1 j

where e tik denotes the labor efficiency of individual i in task k at time t, γ ≤ 1 is a depreciation fac-

tor, and gk (.) is some increasing function. In the presence of returns to experience, it is in

society’s interest to encourage individuals to specialize in particular tasks and to make sure that a

task is always performed by the person most experienced to undertake it. To achieve this result,

individuals may have to work on each other’s fields, e.g., men clearing women’s fields and

women weeding men’s fields. Alternatively, they may become exclusively responsible for partic-

ular activities involving closely connected tasks, such as food preparation or beer brewing, for

instance (e.g., Fafchamps and Quisumbing (1998)).

From the point of view of efficiency, how tasks are distributed among people is largely

irrelevant as long as gains from learning by doing are achieved. This may account for the

existence of extremely wide variations in the distribution of agricultural tasks by gender and age:

in northern Nigeria, for instance, women do not participate ia field work (e.g., Hill (1977),

Fafchamps (1986)); in much of Zaire, they perform almost all agricultural tasks with the excep-

tion of land clearing. Individual comparative advantage in learning and performing particular

tasks may nevertheless dictate who specializes in what. It would be unwise, for instance, to

assign children to tasks that require much physical strength, such as clearing fields. But they are

perfectly adapted to herding livestock. In countries where women bear many children and

breastfeed them for extended periods of time, one may argue that activities around the home con-

stitute their comparative advantage. In stable pre-industrial societies where the range of tasks and

activities to be undertaken does not change over time, the allocation of people to specific tasks is

achieved largely through age and gender casting. Gender-specific role models often reinforce

gender casting (e.g., Collier and Yanagisako (1987)). Certain societies push this principle even
 14

further by restricting particular tasks or activities to individuals belonging to a specific lineage or

caste (e.g., the caste system in India or Yemen; marabouts, griots, and blacksmith in Sub-Saharan

Africa.)

Age and gender casting to match individuals with particular tasks also dictates a particular

pattern of land allocation and, thus, of land inheritance and disposition inter vivos. For instance,

when men are responsible for clearing the land and do most of the land prepration while women

do most of the weeding, harvesting, and processing, as is true in parts of South Asia, there is no

need to allocate land by gender; it is optimal for men and women to farm together under the

management of a single head. In this case, inheritance customs are expected to ensure that house-

hold have a viable farm size, without much concern for gender. Of course if farm management

itself is a task that is the full responsibility of the household head, the devolution of land, either

by inheritance or inter vivos transfer, should be tightly linked to household formation: land will

be passed from heads of households to heads of households.12 To the extent that household heads

are married men, access to land by women will be de facto restricted.

In contrast, when women are responsible for growing vegetable crops while men cultivate

staple food crops, as is the case in Burkina Faso for instance (e.g., Matlon and Fafchamps

(1988)), most land should go to men; women should only receive small vegetable gardens --

which is precisely what we observe. When women take care of producing the staple food while

men produce a cash crop, as occurs in Central Ghana for instance, then one would expect women

to have larger plots and a more equal access to land. The contrast that Egyir draws between Cen-

tral Ghana and Northern Ghana where little cash crop production takes place is consistent with

this pattern: women have less access to land and are more subordinated to their husbands in the

North. Finally, if women are not participating to crop production at all but look after livestock in
________________
12 It is so natural to think of a farm as being managed by the household head that survey questionnaires often
assumes this to be the case. This stands in sharp contrast with off-farm activities such as crafts and trade which are
often undertaken by dependents and are not linked to household formation: while the creation of a new farming unit
occurs simultaneously with the creation of a new household, the creation of a new trading or craft enterprise is not.
 15

and around the house, as is the case in Ethiopia and Pakistan, for instance (e.g., Fafchamps and

Quisumbing (1998)), all cultivable land ought to go to men. Women may nevertheless hold graz-

ing rights as a result of their involvement in livestock production. The same is true for secondary

land rights that women derive from their responsibilities for firewood and water collection. What

gender casting cannot explain, however, is why male and female members of the same house-

holds would grow the same crops on different, individually managed plots -- as reported by Udry

(1996a).

In the presence of learning by doing, sorting people among multiple activities cannot easily

be achieved in a decentralized manner. Letting individuals decide what to specialize in may lead

to over or under provision of certain skills.13 Age and gender casting can thus can be seen as a

means of solving the coordination problem involved in achieving gains from specialization. It

nevertheless raises two serious objections. First, it seldom is equitable. Certain tasks are more

pleasant than others, and the return from certain skills higher than others. Those forced to under-

take unpleasant and undervalued tasks are likely to resent the allocation process and oppose it if

they can. Second, age and gender casting is unlikely to be fully efficient because it prevents many

individuals from expressing their full potential. The efficiency loss will be larger when new

economic opportunities arise and traditional casting rules no longer ensure a match between the

supply and demand for task specific skills. This may explain why age and gender casting have

been more successfully challenged in modern societies than in traditional ones.14

It would be erroneous to assume that economic growth necessarily leads to a loosening of

gender roles, however. In many societies, a sizeable gap exists between ideal gender casting as

specified in the dominant ideology, and the actual division of tasks by gender. Local culture may,
________________
13 See Matsuyama (1991) for an illustration of how occupational decisions can lead to multiple Pareto ranked
equilibrium paths in the presence of sunk costs.
14 Gender casting may be less rigid than sometimes assumed. von Braun and Webb (1989) showed, for instance,
that the introduction of modern irrigation method along the Gambia river led men to rapidly replace women in rice
cultivation in spite of strong taboos regarding male involvement with rice.
 16

for instance, stipulate that women should not work in the field because such work is thought

demeaning. Yet many poor households may fall short of the social ideal simply because they can-

not afford to abide by it (e.g., Hill (1977), Matlon (1977)). In such an environment, increased

prosperity is likely to result in a closer match between social ideals and social realities and thus

to reinforce gender casting in practice. Good examples of such processes can be found in many

Islamic countries where higher standards of living translate into a tighter separation of the sexes

(e.g., Saudi Arabia, Pakistan).15 Efforts by trade unions and other progressive political forces to

have strenuous jobs declared unsuitable for women in turn-of-the-century Europe can be seen in

a similar light. A complete discussion of these issues is left for future work.

Section 3. Missing Land and Labor Markets and Voluntary Exchange

In Section 2 we argued that land and other productive resources are allocated efficiently

within the household if outside markets exist and are competitive. In practice, however, factor

markets are notoriously subject to transaction costs, information asymmetries and enforcement

problems. Can the household efficiently allocate resources within its confines even in the absence

of outside markets? The answer is yes, provided traditional gender casting rules ensure a perfect

match between the demand and supply of skills and rewards for work on household enterprises

equate the marginal return to labor. In the absence of an outside market to set clearly identifiable

wage rate w and factor prices ri , household members may, however, disagree as to how gains

from resource reallocation ought to be shared among themselves.

To see this, consider the model presented in Section 2 and suppose for a moment that the

economy is limited to the household.16 Formally, nothing has changed. There exists, therefore, a

wage rate w and land price ri that equilibrate the household factor markets and ensure allocative
________________
15 I observed similar trends in rural Morocco and Northern Nigeria (see also Matlon (1977)).
16 The argument is by and large unchanged if certain goods can be bought from the market while others, such as
child care and food preparation, are produced directly by the household. That certain goods are not traded may be due
to transaction costs, imperfect information, or enforcement problems. The pattern of transactions costs would, in turn,
determine the optimal size and composition of the household (e.g., Binswanger and McIntire (1987)).
 17

efficiency within the household. Given the small number of people involved, however, it is

unlikely that the household will resort to explicit wage labor transactions. It can, in principle,

achieve the same outcome through implicit contracts. Furthermore, the household economy being

small, perfect competition is unlikely; there is scope for parties to bargain over the distribution of

the gains from resource reallocation. If bargaining is efficient, however, productive resources will

nevertheless be allocated such that their marginal returns are equalized across all activities

undertaken by the household.

Thanks to the second welfare theorem, any Pareto efficient resource allocation can be

represented as a price system combined with lumpsum transfers. Any efficient bargained outcome

can thus be written as the solution to a series of individual maximization problems of the form:

Max U i (ci ) + V i (Ti − ΣLij ) (10)

ci , Lij j

subject to a budget constraint

p ci = F(Ai + ΣÂ ji − ΣÂij , Σ e j Lji ) +

j j j

Σ (w ei Lij − w e j Lji + r Âij − r Âji + Gji − Gij ) (11)

j

and the usual non-negativity constraints Lij ≥ 0, Gij ≥ 0, and Âi ≥ 0. Variable Âij denotes a tem-

porary transfer of productive resource Ai from i to j and Gij is a lumpsum transfer from i to j. The

wage rate w and the rental price r can be thought as the shadow cost of labor and resource Ai that

individuals face in their bargaining over consumption with other household members.17

For a bargaining equilibrium to be both efficient and voluntary, individuals must internalize

the true opportunity cost of the land and labor they command. This is achieved by providing

rewards or transfers w ei Lij and r Âij for labor and land contributed to other members’ produc-

tive activities, and by demanding rewards and transfers equivalent to w e j Lji and r  ji for labor

and land contributed by others to i. In practice, of course, these transfers are netted out, so that
________________
17 Âi , Lij , w and r can be vectors without affecting the argument.
 18

they are not separately observed.18 What matters is simply that household members perceive

that, at the margin, an additional unit of labor or resource that they provide to others or receive

from others is matched by an increase or decrease in individual consumption equivalent to the

economic usefulness of these assets in household production. In an efficient bargaining frame-

work, differences in bargaining power only determine the size of lumpsum transfers Gij between

household members, not shadow prices w and r.19

Without specifying the bargaining process, we now examine the conditions under which

individual household members will voluntarily agree to an incentive scheme {w, r, Gij , G ji }. For

readers familiar with general equilibrium or cooperative game theory, this is akin to constructing

the core of the household economy (e.g., Hildenbrand (1974), Friedman (1990)). We consider

three possible threat points: (1) non-cooperation within the household; (2) separation from the

household; and (3) domestic violence (e.g., Lundberg and Pollack (1993), Kevane (1996)). In the

case of non-cooperation within the household, we assume that individual members can refuse to

exchange labor and land with others and choose to rely on their own labor and land Ai . In the

case of separation from the household, we allow for the possibility that laws and customs may

request that land Ai and other resources be redistributed20 and that compulsory transfers be

imposed on one of the parties.21 Finally, we examine how the threat of domestic violence may

induce someone to ’voluntarily’ agree to a distribution of the gains from resource reallocation

that they would otherwise reject.

Let W ai be the discounted utility of a non-cooperating household member i:

T
W ai = Σ δ ti U i (c ai ) + V i (Ti − Laii ) (12)
t =0
________________
18 Jones (1983, 1986) nevertheless reports that Cameroonian women who work in their husband’s field customarily
receive a share of the harvest that is proportional to the work they have contributed. In Papoua-New Guinea, it is
customary for all household members to receive a share of crop revenue that is commensurate to their contribution to
output (private communication from John Gibson).
19 The size of lumpsum transfers may nevertheless affect the supply of labor and thus the equilibrium wage rate w.
20 E.g., some customs specify that the bride price must be returned in case of repudiation.
21 E.g., alimony and child support.
 19

where c ai and L aij are the levels of consumption and labor that solve equations (1-2) and δi is the

discount factor of individual i. Let W ci be the discounted payoff of a cooperating household

member:

T
W ci = Σ δ ti U i (c ci ) + V i (Ti − ΣLcij ) (13)
t =0 j

where c ci and L cij are the solution to the individual’s optimization problem when facing incentive

scheme {w, r, Gij , G ji }. Individual i will voluntarily agree to the incentive scheme if:

W ci ≥ W ai (14)
This condition, called the household participation constraint, sets limits to the redistribution of

welfare that can credibly take place within the household.

Since cooperating households allocate resources more efficiently than non-cooperating

ones, individuals can be made to cooperate to incentive schemes that specify lumpsum transfers

to others. Lumpsum transfers can be thought of as an entry fee that individuals have to pay to join

the cooperating entity. Some household members may be able to capture most of the gains from

better resource allocation even if participation by others is voluntary. The maximum size of

voluntary transfers is large when gains from cooperation are large, e.g., in the presence of returns

to scale and specialization.

Individuals may object to transfers Gij that would make them worse off than what they

could guarantee themselves by leaving the household. For this reason, rules regarding access to

land are likely to have a paramount effect on the intrahousehold distribution of welfare. Let W li

represent the discounted payoff after separation. Depending on current asset endowments Ai

(themselves a function of patterns of inheritance and inter vivos land transfers) and on laws and

customs regarding the rights and obligations of separating households members (e.g., laws

regarding divorce, minors, support to the elderly; see Egyir (1998))), W li may be larger or smaller

than W ai . If W li > W ai , individual i can credibly threaten to separate from the household unless the

intrahousehold distribution of consumption is such that W ci ≥ W li . What happens if W li < W ai ? Can

 20

the household member credibly threaten to refuse to cooperate without leaving the household?

This depends on what happens upon separation. Suppose that separation only entails a redistribu-

tion of household land Ai and direct transfer obligations. In this case, W lj > W aj for some j when-

ever W li < W ai : taking resources away from one must make the other better off. Consequently, any

threat by household member i to refuse to cooperate without leaving the household will be met

by j’s threat to leave the household. In this case, all that matters are the W ai payoffs. Threats to

remain within the household while refusing to cooperate have no power.

The situation is different if household members must relinquish control over some of their

land Ai and other assets upon household separation. For example, upon repudiation the bride

price may have to be returned to the wife’s parents, not to the wife herself. Court fees in case of

divorce also subtract from the household assets.22 In these cases, it is possible that W ai > W li for

all i’s in the household. If this condition is satisfied, no household member can credibly threaten

to separate from the household; W ai payoffs are the relevant bounds to the household participation

constraints. If, however, W ai < W li for any household members, then the relevant payoffs are the

W li ’s.

Household members may also try to affect the distribution of gains from cooperation

through the threat or use of verbal and physical violence (e.g., Egyir (1998), Kevane (1996)).

Domestic violence is unfortunately too common for this possibility to be discounted. Let Pij

denote the pain and suffering that i can inflict to j and let W ai − ΣP ji be the payoff of a non-
j

cooperating household member who is ’punished’ by the others through domestic violence.

Whether or not the threat of violence is credible depends on social attitudes and on the legal pro-

tection (or lack thereof) that is granted to victims. Clearly, if the threat of violence is credible,

W ai − ΣP ji < W ai and the threatened household member is more amenable to intrahousehold dis-
j
________________
22 Another example would be if a young adult who has received land from the chief must return it to the
community in case of dispute with his father.
 21

tributions of leisure and consumption that are detrimental to himself or herself.

Depending on the social context,23 a household member may be able to escape violence by

leaving the household, provided of course that violence ceases once the victim leaves. This is not

always true, as numerous cases of ineffective restraining orders against violent spouses amply

demonstrate in the U.S. If the threat of violence after separation is credible, the payoff upon

separation becomes W li − Σ jP ji , thereby further reducing bargaining power. If violence against

runaway household members is not credible, perhaps because of physical distance, it is then pos-

sible that W li > W ai − ΣP ji . If the above inequality holds, the threat of violence to non-
j

cooperating household members is ineffective -- provided individuals can flee the household fast

enough to avoid ever incurring P ji .

Given that most domestic violence takes place behind closed doors, deterrence by outside

institutions is only feasible if victims choose to report it. Reporting is not problematic after

separation since, by definition, W li − ΣP ji < W li . In the case of violence to non-cooperating

j

members within the household, however, complaining to an external agency may not be indivi-

dually rational if it leads to the dissolution of the household (e.g., husband or father arrested) and

if life after separation is not an attractive option, i.e., if W ci > W li . As a corollary of this principle,

a poor wife might optimally choose not to report a beating rather than starve. If this is the case,

the threat of domestic violence is credible and effective even though external deterrence is avail-

able at no cost. In other words, household members can only be protected against the threat of

physical violence if their outside payoff W li is high enough. This reasoning explains why domes-

tic violence might be more readily tolerated by poor, illiterate women in environments character-

ized by widespread unemployment.

________________
23 In some societies runaway wives are hunted down and brought back to their husband (e.g., Kevane (1996)). The
same treatment is reserved to runaway children almost anywhere.
 22

We have seen that bargaining power partly determine intrahousehold transfers and thus the

relative influence that household members have on household decisions. Credible threats thus

affect whose preferences have the largest weight in consumption choices (what Chiappori calls

the sharing rule). This idea is implicitly behind much of the discourse about the potentially

beneficial consequence of empowering women. For instance, it is often argued that women are

more concerned about child nutrition and education than men. If true, this implies that child

nutrition and education can be improved by raising the bargaining power of women within the

household. Women have similarly be assumed to care more about vegetables and less about beer

and cigarettes than men. To the extent that policy makers identify more readily with (supposed)

female preferences, they may choose to increase women’s bargaining power even if they do not

care about female welfare per se.

In Section 2 we argued that age and gender casting are a convenient way of capturing gains

from specialization but that, judging from the wide diversity of rules encountered in practice, the

allocation of tasks by age and gender is largely arbitrary. In other words, in any given society,

there are many different ways of dividing tasks among men and women, the young and the old,

all of which capture returns to specialization and are equally efficient. We also pointed out that

the choice of age and gender casting rule dictates a particular set of rules regarding access to

land. The arguments presented in this section indicated that access to land (and other income

earning opportunities) is expected to have a dramatic impact on the intrahousehold distribution of

welfare. Putting the two sections together implies that combinations of rules of gender casting

and access to land that are equivalent from an efficiency point of view need not yield the same

intrahousehold distribution of welfare. An immediate policy implication is that the relative wel-

fare of a particular group, e.g., women or young adults, can be modified by redefining rules of

role casting and access to land (and other income earning opportunities). To be successful, how-

ever, efforts to challenge the status quo need to satisfy two requirements. First, rules regarding
 23

role casting and access to land must remain ’in sync’: there is little point insisting that women

inherit land if they are not allowed to cultivate it, and vice versa. Second, the new rules must

ensure that returns to specialization are achieved. In this respect, efforts to eliminate age and

gender casting altogether and replace it with a Jeffersonian ideal of equal opportunity for all may

be ill suited to traditional societies with little economic progress. Replacing one role casting sys-

tem by another that is more beneficial to the target group may be more acceptable by all than the

’chaos’ of equal opportunity for all.

Section 4. Discrimination and Collusion in Labor and Land Markets

Throughout the previous section, we assumed that markets do not exist for labor, land, and

other productive resources so that the only way efficiency gains from resource reallocation can be

achieved is for individuals to form households.24 In this section we investigate the effect of vari-

ous market configurations on intrahousehold allocation. We begin by noting that, if complete

markets exist and there are no household public goods, there are no efficiency gains to be made

from the formation of households.25 There is, therefore, no room for the cooperative payoff W ci to

be higher than the non-cooperative payoff W ai for all i. The only way W ci can be greater than W ai

for some i is for W cj to be below W aj for some j. In this case, if separation is costless26 and domes-

tic violence cannot be exerted on separated household members, cooperative payoffs must equal

autarky payoffs:

W ci = W li (15)
for all i’s. If separation is costly for the household, there is room left for negotiation on the distri-

bution of household leisure and consumption. But if it is not, individual welfare is entirely dic-
________________
24 Or other, looser structures such as co-residence units (e.g., compounds) and social networks.
25 This simple but powerful observation may account for the erosion of the family in developed countries. See
Binswanger and McIntire (1987) on vertically integrated households.
26 i.e., if household members do not have to incur costs to separate (e.g., legal costs), they do not have to relinquish
household assets to non-household members (e.g., brideprice returned to wife family), and they do not loose
entitlement benefits (e.g., tax break).
 24

tated by factor prices and the distribution of household resources after separation.

The above suggests that legal provisions governing the attribution of household assets and

providing for alimony and child support should have a direct effect on the relative welfare of

household members (see Egyir (1998) for a description of these provisions in Ghana). In a world

of complete markets, however, individuals should refuse to voluntarilly join households unless

they are guaranteed a payoff after separation that is equivalent to their expected payoff, should

they remain on their own. Once laws and customs affecting intrahousehold welfare are fully anti-

cipated, they should result in side payments (e.g., bride price, dowry) at the onset of the union

itself, making it difficult for an equity-minded policy maker to durably affect the intrahousehold

distribution of welfare (e.g., Becker (1981), Lundberg and Pollack (1993)).

We just saw that the existence of perfect markets makes it harder for a particular household

members, e.g., the male head of household, to appropriate the efficiency gains from the redistri-

bution of land and productive resources. It may, therefore, be tempting for, say, male heads to

collude and deny their wives and dependents the right to get land of their own and to work out-

side the household (e.g., Kevane (1996), Egyir (1998), Fafchamps and Quisumbing (1998)). So

doing, they lower females’ W ai and W li and more easily extract transfers from them. The prohibi-

tion to get land and work outside the home or farm further raises the bargaining power of house-

hold heads if the members who specialize in domestic or household-based chores become less

well informed about possible exit option, less able to secure a job through job market contacts,

and thus be more fearful of the consequences of dissent.27

The allocation of labor resulting from collusion among male heads need not be inefficient,

however, provided that heads of household spend at least part of their time on their own farm or
________________
27 Formally, the lack of information about job market opportunities translates into flat (i.e., high variance) priors
which, if the person is risk averse, lower the expected utility of exit. The lack of network contacts has the same effect
by reducing the probability of finding a job. See also Schaffner (1995) for evidence that servility shifts preferences
against exit.
 25

business.28 Indeed, as long as captive and non-captive workers work together on the same pro-

duction activity, the opportunity cost of captive labor is the same as the market wage. Aggregate

labor resources can thus be allocated efficiently by ensuring that captive workers are given an

implicit incentive to work equivalent to w ei . Only when non-captive workers spend all their time

outside the household is the allocation of labor inefficient: the household no longer is at an inte-

rior solution and the marginal return to captive labor falls below the market wage. In this case, all

could be made better of by letting underemployed captive workers rent in land or hire themselves

out for a wage. If Pareto gains from letting women and children work for other households are

large enough, the redistributive gain that household heads obtain by colluding not to rent land or

hire each others’ dependents is smaller the efficiency gain that they could achieve by letting them

work as laborers or independent farmers. Such a situation is likely to arise in villages and regions

with lots of landless workers and sub-marginal farmers. Collusion to keep dependents out of the

land and labor markets then loses much of its attraction and can reasonably be expected to col-

lapse or take another form. Heads of household may, however, continue to extract transfers from

dependents by colluding to restrict access to the best land and to pay women and dependents a

low wage. Formally, the situation is akin to forming a cartel of employers and is discussed in

Akerlof (1976).

Both types of collusive arrangements among heads of household -- captive labor or lower

wage and bad land -- are, of course, vulnerable to free riding. In both cases, it is in the interest of

a single employer or landlord to circumvent the collusive arrangement. Employers could attract

captive workers by paying them a wage below the market rate or by renting them land at higher

than market price because, by securing an independent source of income, captive workers would

avoid paying the lumpsum ’tax’ imposed by the household29 and could thus be better off.
________________
28 As Basu and Van (1996) argues in the case of child labor, prohibiting certain household members from working
outside the household may also be efficient if the withdrawal of their labor raises total wages. See also Anderson and
Francois (1996).
29 Either directly by leaving the household, or indirectly by renegotiating their obligations within the household.
 26

Similarly, by raising the wage paid to women and children slightly above that paid by others,

employers could attract a larger number of low wage but equally productive workers and thus

make more profit. Cartel-like arrangements are thus not self-enforcing and require a meta-

punishment to be sustained. The form that the meta-punishment may take is unclear, although

one can imagine things like social ostracism by other male heads, sabotage, violence, poisening,

etc (e.g., Platteau (1997), Kevane (1996)). Women and children who work in violation of social

prohibition may also face various forms of harrassment, including sexual harrassment by co-

workers.30

Another, more subtle way to lower the wage of dependents and thus their exit payoff is to

lower their ei by restricting their access to education and by inciting them to learn tasks with low

potential for wage employment, such as household chores. As we have argued in section 2, age

and gender casting of this type need not be inefficient,31 even though they clearly have detrimen-

tal effects on equity. Finally, in places where income earning opportunities mostly take the form

of farming and self-employment, access to individual income requires personal finances. In that

case, another way of lowering the compensation for dependents’ labor is to restrict their access to

credit and hence to remunerative uses of their time other than collaboration to household produc-

tion (e.g., Egyir (1998), Groote and Kebe (1995), Goetz and Gupta (1996), Smith (1994)). Udry

(1996a) gives yet another example in which it is access to organic fertilizer which is the key fac-

tor to which women have unequal access.

Section 5. Time and Commitment

In the two preceding sections we argued that individual household members will object to a
________________
30 In such a setting, non-discriminatory employment opportunities for women can be created by setting up rural
public works programs open to all (e.g., Deolalikar and Gaiha (1992) and the references cited therein show that
women often constitute an important proportion of public works employees).
31 With the same caveat as before, namely that age and gender casting are restrictions to the freedom of
opportunity and prevent individuals from achieving their full economic potential, e.g., as entrepreneurs, leaders,
researchers, inventors, and the like.
 27

distribution of gains from resource reallocation that makes them worse off than what they could

guarantee themselves by leaving the household or refusing to cooperate. This idea was captured

in a household participation constraint of the form W ci ≥ W ki where W ki denotes the credible threat

point of household member i, which is either W ai or W li , with or without violence, depending on

circumstances (see Section 3). We now examine how participation constraints are affected by the

passage of time between the reallocation of land and other productive resources, and the enjoy-

ment of consumption. Throughout this section we assume that household first allocate labor,

land, and other productive resources to production (period one). Output is then realized and the

proceeds used to finance consumption (period two).

Consider the following arrangement between household members: i agrees to work on j’s

field in period one in exchange for a share of output in period two -- i.e., a "gift" at harvest time

(e.g., Jones (1986), Boe (1996)). As we have seen in the preceding section, it is rational for i to

accept such an arrangement if W ci ≥ W ki . For j to voluntarily compensate i, however, the agree-

ment must be self-supporting, that is, it must also be rational for j to fulfill his or her promise in

period two:

U j (q cj ) − U j (q cj − ei Lij + Gij − G ji ) ≤ δ (W ci − W ki ) (16)

In other words, for j’s promise to be credible, it must be in j’s interest to share output with i after-

wards. As the above equation shows, j will do so only if the short term loss from having to com-

pensate i is smaller than the long term gain from a continued relationship.32 This is unlikely to be

true if the promised compensation ei L ji + Gij − G ji is large, the gain from cooperation W ci − W ki

is small, or j’s time horizon is short, i.e., δ is low. If equation (16) is not satisfied, j is unable to

credibly commit to compensate i and the agreement will not be fulfilled. Anticipating this, i

should refuse to cooperate. There is commitment failure: a Pareto improving transaction does not

take place because one party cannot credibly commit to honor the contract (e.g., Coate and
________________
32 We have implicitly assumed that if j fails to pay i, individual i responds by not transfering Gij and refusing to
cooperate in the future.
 28

Ravallion (1993), Balsvik (1995)).

Clearly, the possibility of commitment failure puts additional constraints on the type of

resource reallocation and compensation schemes that are self-supporting (e.g., Pollack (1985)).

First, for a household member to credibly commit to any ex post payment, that member must not

have been pushed to his or her exit payoff; he or she must derive some benefit from participating

to the household. Put differently, for the household head to credibly commit to compensate

dependents for participating to farming or household chores, the head must be strictly better off in

a functional household than outside of it. If the welfare gains from household formation are

small, this may require that most if not all these gains go to the head for intrahousehold labor

transfers to take place. When this condition is not satisfied, members are likely to insist on

operating in strictly separate spheres with little or no labor exchange -- a feature that many would

associate with dysfunctional households. Of course, the refusal to collaborate on a common pro-

ject further reduces the gains from household formation and fragilizes the household even more.

Second, for any distribution of the gains from collaboration W ci − W ai , there are limits to the

ex ante redistribution of productive resources. If equation (16) is binding, production efficiency

may be impossible to achieve: i may refuse to work more than a certain number of days Lij on j’s

field because any promise of payment in excess of w ei Lij is not credible. Of course, it may be

possible to redistribute the gains from cooperation differently and raise W cj − W kj so as to satisfy

equation (16). But this would mean taking something away from i, a prospect i is likely to object

to.

Third, commitment failure may be mitigated by allowing for an external enforcement

mechanism. One possible mechanism is for household members to pass among themselves

market transactions that are at least theoretically enforceable by courts.33 One conceivable
________________
33 Market transactions among spouses are common in parts of Sub-Saharan Africa, notably the coastal areas of
Ghana and Nigeria. It is, for instance, frequent for women traders to purchase the output of their husband farmers or
fishermen.
 29

mechanism is reliance on legal institutions, such as nuptial agreements, business partnership

agreements, joint ownership of assets, joint holding of bank accounts, etc. When small amounts

are involved, however, court action is seldom attractive because it is too costly and time consum-

ing. Alternative legal institutions at the village or community level -- such as chiefs and mara-

bouts -- may exist that have jurisdiction to adjudicate matrimonial disputes and can put pressure

on a recalcitrant spouse to honor a promise. Proving that a promise has been breached and that

one spouse is at fault is often difficult to verify, given that households often engage in complex

implicit arrangements. If breach of promise cannot be established, external agents may choose to

rely on general standards of fairness instead. Given the paucity of data on these issues, providing

empirical evidence of what these standards are and how matrimonial disputes are adjudicated in

practice is a promising area of research.

Fourth, agreements are more difficult to enforce if compensation is promised only in a dis-

tant future (i.e., δ low). In some cases, compensation may be delayed until the death of the

spouse or until dissolution of the marriage (e.g., Egyir (1998)). In this case, rules and customs

about the devolution of assets limit the forms that compensation may take, while at the same time

making some forms of credible commitments possible. For instance, if laws and customs stipulate

that the wife inherits half of the household assets upon death or divorce, it will be easier for the

husband to convince his wife to help build up joint assets than if the wife has no right upon them.

In this case, asset devolution rules can be seen as providing a crude commitment mechanism. For

such a commitment mechanism to be effective, husbands must have no discretion to change the

devolution of joint assets; efforts to circumvent rules must be regarded as a form of fraud and

punished accordingly.

Credit constraints also matter. Household members who are credit constrained may be

unable to promptly pay for labor services and may have to wait until sufficient output has been

generated before doing so (e.g., Egyir (1998)). Their production activity may also be subject to
 30

shocks, in which case they may be unable to pay in period two. One would therefore expect com-

mitment problems to be more severe and the intrahousehold allocation of resources to be less

efficient in poor, credit constrained households. Finally, commitment failure is more likely when

j’s time horizon is short, as would be the case if j expects to leave the household in the near

future. Implicit agreements among household members are thus less likely to be made and more

likely to be broken in households that are dysfunctional or are breaking apart.

For the same reason, commitment failure is more problematic for individuals who expect to

migrate, exit, or marry out of the household: these individuals are more likely to free-ride and

less likely to take a long-term interest in the prosperity of the household. Laws and customs

about inheritance, brideprice and dowry, etc, can be regarded as an effort to mitigate this form of

free riding. Societies in which daughters marry before they are old enough to rationally choose to

free-ride, also typically deny daughters a right to inherit. In contrast, sons often continue to con-

tribute to household production until they are much older and are typically given, through cus-

toms regarding inheritance and inter vivos transfers, a stake in their father’s farm (e.g., Egyir

(1998)).34

Another possible way that commitment problems can be solved is by trading labor against

land or other productive resources.35 Suppose j tries to convince i to work full time on his field

but i objects that she does not believe she will be fully compensated for her effort -- i.e., equation

(16) is violated. Given this state of affairs, it may be in j’s interest to let i cultivate part of his land

and keep the proceeds in exchange for a payment after harvest r ji . The transfer of user right

over land can then be seen as a way of giving an immediate payment and thus of bypassing com-

mitment failure (see Egyir (1998) for an illustration). Since the land transfer partly or fully can-
________________
34 For a discussion of how the concentration of land into the hands of the elder affect family organization and
sectoral employment, see for instance St-Amour and Vencatachellum (1996).
35 Other give-and-take arrangements can be found as well, e.g., letting women take some of the grain to brew it
into beer and sell it (e.g., Haggblade (1987)).
 31

cels out the implicit wage payment wei Lij , equation (16) is more easily satisfied and the

exchange of labor can take place.

Whether or not this arrangement is efficient depends on returns to scale and specialization:

if returns to scale are increasing, it is inefficient to divide the farm into distinct units. Even so, the

efficiency gain from a better allocation of labor may compensate the efficiency loss from a worse

allocation of land, so that the net effect of the arrangement is beneficial. Udry’s (1996a) finding

that women work harder on their fields than on their husband’s is consistent with the existence of

incentive problems such as the ones generated by imperfect commitment. Other features

described by Udry are also more consistent with commitment problems than role casting. For

instance, Udry finds that dependents often grow, on their individual fields, the same crop as the

head on the household fields; this is inconsistent with a pure role casting rationale for depen-

dents’ plots. Descriptive accounts also point out that dependents who cultivate their own plots

are entitled to use the money as they wish -- provided they contribute to household expenditures.

This is in line with the role of dependents’ plots as financial compensation devices: capturing

gains from specialization does not require financial independence but commitment failure does.

Finally, the fact that dependents’ plots are only granted on a temporary basis, mostly by the

household head himself, is further indication that commitment failure is the motivation. If role

casting was the rationale, there would be little reason for dependents’ plots not to be allocated on

a permanent basis. All in all, the preliminary evidence is thus highly suggestive that the practice

of dependents’ plots, which is widespread in much of Western and Central Africa, is primarily an

effort to overcome commitment failure within the household. More work is nevertheless needed

before we reach a definitive conclusion.

Conclusion and Prospect for Empirical Work

This paper has examined the conditions under which land, labor, and other productive

resources might not be allocated efficiently within (and across) households. We noted that,
 32

depending on returns to scale, specialization and experience, efficiency may require the centrali-

zation of production -- and thus land -- under the authority of the head, and the specialization of

individual household members in certain tasks and activities. In this context, age and gender

casting can be understood as an institution meant to resolve the allocation of tasks among house-

hold members. Because gender and age seldom confer a strong comparative advantage in any

particular task, role casting is largely arbitrary, as indicated by the wide variety of task allocation

rules that one finds in poor rural areas.

Role casting can thus be seen as an efficiency enhancing institution that attempts to capture

gains from specialization by providing a focal point for intrahousehold task allocation. So doing,

however, it introduces a potential mismatch between the tasks allocated to particular individuals

and their innate talent. The welfare loss from such a mismatch is probably low in unspecialized,

undifferentiated societies such as poor rural areas of the Third World. But it is likely to be high in

highly diversified, rapidly evolving societies. Consequently, we expect role casting to become

increasingly costly and ill-adapted as societies become more differentiated -- and thus to be pro-

gressively repealed as development takes place. More work is needed on this issue.

A good grasp of role casting is necessary before one can make sense of rules of access to

land because role casting often dictates a particular pattern of land allocation across individuals

and, thus, of land inheritance and disposition inter vivos. For instance, when women are respon-

sible for certain crops such as vegetables or food crops, it is natural that they be granted access to

a comensurate piece of land. In contrast, when women do not participate in crop production -- or

where their participation is restricted to certain agricultural tasks, not certain crops -- then female

access to land is not required to capture returns from specialization. In that case, one would

expect only household heads to receive land.

Role casting and access to land do not only affect intrahousehold efficiency, they also are

important determinants of intrahousehold welfare through their effect on exit options and bar-
 33

gaining power within the household. Combinations of rules of gender casting and access to land

that are equivalent from an efficiency point of view need not yield the same intrahousehold distri-

bution of welfare. An immediate policy implication is that the relative welfare of a particular

group, e.g., women or young adults, can be modified by redefining role casting and rules of

access to land (and other income earning opportunities). To be successful, however, efforts to

challenge the status quo need to ensure that returns to specialization are still achieved.

We also investigated the relationship between intrahousehold land allocation, voluntary

participation constraints, efficiency, and equity and showed that commitment failure within the

household may lead to inefficiency. Granting dependents temporary rights to cultivate their own

plot of land, as discussed for instance in Udry (1996a), may be used to mitigate intrahousehold

commitment problems. If this is the case, women and other dependents are expected to apply

more labor to their farm and business than is efficient. Udry (1996b)’s results are consistent with

this prediction. Second, the head of household is expected to mitigate commitment problems by

allocating individual plots and other productive resources on a discretionary basis, as docu-

mented by Egyir (1998). Third, inefficiency is expected to be fostered by factors that exacerbate

commitment failure, such as short time horizon (sons and daughters who expect to leave the

household; dysfunctional families), low assets (low gains from intrahousehold exchange),

unequal stakes in the household (some members have low welfare gains from household partici-

pation), and poor external enforcement. This is consistent with the observation that adult depen-

dents work in general less hard than the household head and his wife (e.g., Cleave (1974),

Fafchamps (1986), Fafchamps and Quisumbing (1998)). Finally, our framework predicts that the

allocation of consumption and leisure should reflect participation constraints. It therefore pro-

vides the basis for a test of the effect of determinants of exit options -- such as access to land --

on the intrahousehold allocation of leisure and nutrition. Providing empirical verification for

these predictions is the object of future research.

 34

We finish with a few words about a topic that was ignored here, that of female headed

households. As pointed out by many authors (e.g., Egyir (1998), Gladwin and McMillan (1989),

Jiggins (1989), female headed households have unequal access to land and credit. Similar prob-

lems, although not as severe, afflict female operated households. The question is: why? One pos-

sible answer is that denying women access to land and credit is a way to lower their exit payoff.

Making exit less attractive for women has two effects. First, it reduces their bargaining power

within the household, and thus is likely to affect women’s welfare even in male headed house-

holds. This issue was discussed in Section 4. Second, it also makes commitment failure less

likely and thus mitigates incentive problems and inefficiency within households. This is because

allocating most of the welfare gains from household formation to male heads gives them a stake

in the preservation of the household. As a result, household heads have an incentive to honor

their commitment to redistribute part of the joint output to dependents in exchange for their parti-

cipation in household activities.

If the above interpretation is correct, denying women access to land and credit indirectly

serves to stabilize households -- although it does so, arguably, at women’s expense.36 This opens

another question, namely, what is the reason for the existence of female headed households and

for the apparent rise in the proportion of female headed and female operated households in some

African countries? Is it a result of efforts to empower women? Is it a consequence of male migra-

tion to cities, mines, and plantations? These issues deserve empirical investigation as they

undoubtedly have serious implications on the welfare of children and other social issues (see

Akerlof et al. (1996), Korenman and Neumark (1991), and Akerlof (1998) for a similar emphasis

in the U.S.). Before encouraging equal access to land and credit for women, one would like to

know what impact this will have on household stability.

________________
36 Because households with two or more adults benefit from cost savings in the provision of household public
goods, it is conceivable that women benefit in the absolute sense from being denied access to land if without it there
would be many more households headed by a single adult, male or female. A thorough treatment of this issue is left
for future research.
 35

To make this clear, consider the following stylized example: suppose that all land goes to

male heads of household, that heads marry to get access to female and child labor, and that they

can credibly commit to compensate their dependents because they gain from marriage. Suppose

further that a reform grants women half of their husband land upon divorce. While this reform

would probably raise the welfare of wives in existing households, its net long term effect may be

null or negative. For instance, men may respond by refusing to marry and by hiring women and

children as wage workers instead. In this case, the long-term net welfare impact of the reform on

women is essentially zero,37 but its effect on out-of-wedlock birth and thus on illegitimate chil-

dren is likely to be very negative. Only by simultaneously ensuring that daughters gain equal

access to land through inheritance and transfers inter vivos can this negative effect be avoided.

This example serves to illustrate that tampering with family law should not be attempted without

considering the effect on household formation.

________________
37 In equilibrium, the wage women get as workers is equal to the intrahousehold compensation they get when their
only asset is labor.
 36

References

Adams, R. H. and He, J. J., Sources of Income Inequality and Poverty in Rural Pakistan,
Research Report 102, IFPRI, Washington D.C., 1995.
Akerlof, G. A., ‘‘The Economy of Caste and of the Rat Race and Other Woeful Tales,’’ Quarterly
J. Econ., 90: 599-617, 1976.
Akerlof, G. A., ‘‘Men Without Children,’’ Economic Journal, 108(447): 287-309, March 1988.
Akerlof, G. A., Yellen, J. L., and Katz, M. L., ‘‘An Analysis of Out-of-Wedlock Childbearing in
the United States,’’ Quarterly Journal of Economics, 111: 278-317, May 1996.
Alderman, H. and Christi, S., ‘‘Simultaneous Determination of Household and Market-Oriented
Activities of Women in Rural Pakistan,’’ Research in Population Economics, 7: 245-265,
1990.
Alderman, H., Chiappori, P., Haddad, L., Hoddinott, J., and Kanbur, R., ‘‘Unitary Versus
Collective Models of the Household: Is It Time to Shift the Burdern of Proof,’’ World Bank
Research Observer, 10(1): 1-19, February 1995.
Anderson, S. and Francois, P., Unequal Access and the Detrimental Impact of Development on
Women, Department of Economics, University of British Columbia, Vancouver, May 1996.
(mimeograph).
Balsvik, R., Production Decisions in Nonpooling Households, Department of Economics,
University of Oslo, Oslo, January 1995.
Basu, K. and Van, P. H., The Economics of Child Labor, Department of Economics, Cornell
University, Ithaca, March 1996. (mimeograph).
Becker, G. S., ‘‘A Theory of the Allocation of Time,’’ Economic Journal, 75(299): 493-517,
September 1965.
Becker, G. S., A Treatise on the Family, Harvard U.P., Cambridge, Mass., 1981.
Bergstrom, T. C., A Survey of Theories of the Family, Department of Economics, University of
Michigan; CREST Working Paper, Ann Arbor, June 1995. (mimeograph).
Binswanger, H. P. and McIntire, J., ‘‘Behavioral and Material Determinants of Production
Relations in Land-Abundant Tropical Agriculture,’’ Econ. Dev. Cult. Change, 36(1): 73-99,
Oct. 1987.
Boe, T., Access Regimes, Loss Avoidance and Institutions: The Mandinka Household Economy,
Chris. Michelsen Institute, Bergen, Norway, June 1996. (mimeograph).
Boserup, E., Women’s Role in Economic Development, Allen and Unwin, London, 1970.
Bourguignon, F., Browning, M., Chiappori, P., and Lechene, V., ‘‘Intra Household Allocation of
Consumption: A Model and Some Evidence from French Data,’’ Annales d’Economie et de
Statistique, 29: 137-156, 1993.
Bourguignon, F., Browning, M., and Chiappori, P., The Collective Approach to Household
Behaviour, Centro Studi Luca d’Agliano -- Queen Elizabeth House, Development Studies
Working Paper No. 79, Oxford, April 1995. (mimeograph).
 37

Brown, L. and Haddad, L., Time Allocation Patterns and Time Burdens: A Gendered Analysis of
Seven Countries, International Food Policy Research Institute, Washington, D.C., 1995.
Browning, M., Bourguignon, F., Chiappori, P., and Lechene, V., ‘‘Income and Outcomes: A
Structural Model of Intrahousehold Allocation,’’ J. Political Econ., 102(6): 1067-1096,
1994.
Carter, M. and Katz, E., ‘‘Separate Spheres and the Conjugal Contract: Understanding the Impact
of Gender-Based Development,’’ Intrahousehold Resource Allocation in Developing
Countries: Methods, Models and Policy, Lawrence Haddad, John Hoddinott and Harold
Alderman (eds.), IFPRI, Washington D.C., 1993. (book manuscript).
Chiappori, P., ‘‘Rational Household Labor Supply,’’ Econometrica, 56(1): 63-90, January 1988.
Chiappori, P., ‘‘Collective Labor Supply and Welfare,’’ J. Political Econ., 100(3): 437-467, 1992.
Chiappori, P., ‘‘"Collective" Models of Household Behavior: The Sharing Rule Approach,’’
Intrahousehold Resource Allocation in Developing Countries: Methods, Models and
Policy, Lawrence Haddad, John Hoddinott and Harold Alderman (eds.), IFPRI, Washington
D.C., 1993. (book manuscript).
Cleave, J., African Farmers: Labor Use in the Development of Smallholder Agriculture, Praeger,
1974.
Coate, S. and Ravallion, M., ‘‘Reciprocity Without Commitment: Characterization and
Performance of Informal Insurance Arrangements,’’ J. Dev. Econ., 40: 1-24, 1993.
Collier, J. F. and Yanagisako, S. J., Gender and Kinship: Essays Toward a Unified Analysis,
Stanford University Press, Stanford, 1987.
de Janvry, A., Fafchamps, M., and Sadoulet, E., ‘‘Peasant Household Behavior with Missing
Markets: Some Paradoxes Explained,’’ Economic Journal, 101(409): 1400-1417,
November 1991.
Deolalikar, A. B. and Gaiha, R., Targeting of Rural Public Works: Are Women Less Likely to
Participate?, University of Washington, Seattle, February 1992. (mimeograph).
Dercon, S. and Krishnan, P., Risk-Sharing within Households in Rural Ethiopia, Oxford
University, Oxford, April 1997. (mimeograph).
Egyir, I., Intra-Household Access to Land and Sources of Inefficiency: A Case Study of Ghana,
Department of Agricultural Economics, University of Ghana at Legon, Accra, March 1998.
(mimeograph).
Eswaran, M. and Kotwal, A., ‘‘Access to Capital and Agrarian Production Organization,’’
Economic Journal, 96(382): 482-498, June 1986.
Evenson, R. E., ‘‘Time Allocation in Rural Philippine Households,’’ Amer. J. Agric. Econ., p.
322-330, May 1978.
Fafchamps, M., Labor Use and Productivity and Technological Change in African Smallholder
Agriculture: Synthesis Report, Jobs and Skills Programme for Africa, International Labor
Organization, Addis Ababa, 1986.
Fafchamps, M. and Quisumbing, A. R., Social Roles, Human Capital, and the Division of Labor:
Evidence from Pakistan, Department of Economics, Stanford University, Stanford, April
1998. (mimeograph).
 38

Feder, G., ‘‘The Relation Between Farm Size and Farm Productivity,’’ Journal of Development
Economics, 18 (2-3): 297-314, August 1985.
Filmer, D., Gender Differences in the Intrahousehold Allocation of Schooling and Preventive
Medical Care in Developing Countries: Estimates from Ghana, Department of Economics,
Brown University, July 1996. (mimeograph).
Fortin, B. and Lacroix, G., A Test of the Neoclassical and Collective Models of Labor Supply,
Université Laval, Québec, 1993.
Friedman, J. W., Game Theory with Applications to Economics, Oxford University Press, 1990.
Gavian, S. and Fafchamps, M., ‘‘Land Tenure and Allocative Efficiency in Niger,’’ American
Journal of Agricultural Economics, 78: 460-471, May 1996.
Gladwin, C. H. and McMillan, D., ‘‘Is a Turnaround in Africa Possible Without Helping African
Women to Farm?,’’ EDCC, 37, no.2: 345-369, Jan. 1989.
Goetz, A. and Gupta, R. S., ‘‘Who Takes the Credit? Gender, Power, and Control Over Loan Use
in Rural Credit Programs in Bengladesh,’’ World Development, 24(1): 45-63, 1996.
Groote, H. d. and Kebe, D., The Impact of Credit on Women’s Income, or: Is Credit Targeting in
South Mali Justified, Paper presented at the Annual AAEA Conference in Indianapolis,
Indianapolis, August 1995.
Haddad, L. and Kanbur, R., ‘‘How Serious Is the Neglect of Intra-Household Inequality?,’’
Economic Journal, 100: 866-881, September 1990.
Haddad, L., Hoddinott, J., and Alderman, H., Intrahousehold Resource Allocation in Developing
Countries: Models, Methods, and Policy, Johns Hopkins University Press, Baltimore, 1997.
Haddad, L. and Kanbur, R., ‘‘The Value of Intrahousehold Survey Data for Age-Based
Nutritional Targeting,’’ Annales d’Economie et de Statistique, 29: 65-81, 1993.
Haggblade, S., ‘‘Vertical Considerations in Choice-of-Technique Studies: Evidence from Africa’s
Indigenous Beer Industry,’’ EDCC, 35, no.4: 723-742, July 1987.
Hildenbrand, W., Core and Equilibria of a Large Economy, Princeton University Press,
Princeton, 1974.
Hill, P., Population, Prosperity and Poverty: Rural Kano 1900 and 1970, Cambridge University
Press, Cambridge, 1977.
Jacoby, H. G., ‘‘Productivity of Men and Women and the Sexual Division of Labor in Peasant
Agriculture of the Peruvian Sierra,’’ J. Development Econ., 37: 265-287, 1992.
Jacoby, H. G., ‘‘The Economics of Polygyny in Sub-Saharan Africa: Female Productivity and the
Demand for Wives in Côte d’Ivoire,’’ J. Polit. Econ., 103(5): 938-971, 1995.
Jamison, D. T. and Lau, L. J., Farmer Education and Farm Efficiency, Johns Hopkins University
Press, Baltimore, 1982.
Jiggins, J., ‘‘How Poor Women Earn Income in sub-Saharan Africa and What Works Against
Them,’’ World Development, 17, no.7: 953-963, 1989.
Jones, C., ‘‘The Mobilization of Women’s Labor for Cash Crop Production: A Game Theoretic
Approach,’’ Amer. J. Agric. Econ., p. 1049-1054, December 1983.
 39

Jones, C. W., ‘‘Intra-Household Bargaining in Response to the Introduction of New Crops: A

Case Study from North Cameroon,’’ Understanding Africa’s Rural Households and
Farming Systems, Joyce L. Moock (ed.), Westview Press, Boulder and London, 1986.
Kanbur, R. and Haddad, L., ‘‘Are Better Off Households More Unequal or Less Unequal?,’’
Oxford Econ. Papers, 46: 445-458, 1994.
Katz, E. G., Gender and Trade Within the Household: Observations from Guatemala, The World
Bank, Washington D.C., 1994. (mimeo).
Kevane, M., Determinants of Women’s Time Allocation Between Home, ’Family’ Field, and
Market Production in South-Western Burkina Faso, Department of Economics, Santa Clara
University, Santa Clara, 1996. (mimeograph).
Khandker, S. R., ‘‘Determinants of Women’s Time Allocation in Rural Bengladesh,’’ Econ.
Development and Cultural Change, p. 111-126, 1988.
Korenman, S. and Neumark, D., ‘‘Does Marriage Really Make Men More Productive?,’’ Journal
of Human Resources, 26(Spring): 282-307, 1991.
Kumar, S. K., Adoption of Hybrid Maize in Zambia: Effects on Gender Roles, Food Consumption,
and Nutrition, Research Report 100, IFPRI, Washington D.C., 1994.
Lilja, N., Sanders, J. H., Durham, C. A., De Groote, H., and Dembélé, I., Factors Influencing the
Payments to Women in Malian Agriculture, Department of Agricultural Economics, Purdue
University, Lafayette, July 1996. (mimeograph).
Lilja, N., Technological Change in Agriculture and the Welfare of Women in Southern Mali,
Department of Agricultural Economics, Purdue University, Lafayette, December 1996.
(unpublished Ph.D. thesis).
Low, J. W., The Effect of Household Structure and Inter-Household Income Transfers on Child
Nutritional Status in Northern Malawi, Department of Economics, Cornell University,
Ithaca, 1994. (mimeograph).
Lundberg, S. and Pollack, R. A., ‘‘Separate Spheres Bargaining and the Marriage Market,’’ J.
Polit. Econ., 101(6): 988-1010, 1993.
Lundberg, S. and Pollack, R. A., ‘‘Noncooperative Bargaining Models of Marriage,’’ Amer.
Econ. Rev., 84(2): 132-137, May 1994.
Manser, M. and Brown, M., ‘‘Marriage and Household Decision Making: A Bargaining
Analysis,’’ Internat. Econ. Rev., 21(1): 31-44, February 1980.
Manser, M. E., ‘‘The Allocation of Consumption by Married Couple Families in the U.S.: An
Analysis Conditioning on Labor Supply,’’ Annales d’Economie et de Statistique, 29: 83-
108, 1993.
Matlon, P. J., Income Generation in Three Villages of Northern Nigeria , University of Cornell,
1977. Unpublished Ph.D thesis.
Matlon, P. J. and Fafchamps, M., Crop Budgets in Three Agro-Climatic Zones of Burkina Faso,
ICRISAT Progress Report, Hyderabad, 1989.
Matsuyama, K., ‘‘Increasing Returns, Industrialization, and Indeterminacy of Equilibrium,’’
Quat. J. Econ., CVI(425): 617-650, May 1991.
 40

McElroy, M. B. and Horney, M. J., ‘‘Nash-Bargained Household Decisions: Toward a

Generalization of the Theory of Demand,’’ International Econ. Rev., 22(2): 333-349, June
1981.
McElroy, M. B., ‘‘The Empirical Content of Narsh-Bargained Household Behavior,’’ J. Human
Resources, 25(4): 559-583, 1990.
Pitt, M. M., Rosenzweig, M. R., and Hassan, M. N., ‘‘Productivity, Health, and Inequality in the
Intrahousehold Distribution of Food in Low-Income Countries,’’ Amer. Econ. Rev., 80(5):
1139-1156, December 1990.
Platteau, J., Traditional Sharing Norms as an Obstacle to Economic Growth in Tribal Societies,
CRED, Facultés Universitaires Note-Dame de la Paix, Namur, Belgium, June 1996.
Cahiers No. 173.
Pollack, R. A. and Wachter, M. L., ‘‘The Relevance of the Household Production Function and
Its Implications for the Allocation of Time,’’ J. Polit. Econ., 83(2): 255-277, 1975.
Pollack, R. A., ‘‘A Transaction Cost Approach to Families and Households,’’ J. Econ. Literature,
23: 581-608, June 1985.
Quisumbing, A., Haddad, L., and Peña, C., Gender and Poverty: New Evidence from 10
Developing Countries, FCND Discussion Paper No. 9, IFPRI, Washington D.C., December
1995.
Rosenzweig, M. R. and Evenson, R., ‘‘Fertility, Schooling, and the Economic Contribution of
Children in Rural India: An Econometric Analysis,’’ Econometrica, 45(5): 1065-1078, July
1977.
Rosenzweig, M. R., ‘‘Program Interventions, Intrahousehold Distribution and the Welfare of
Individuals: Modeling Household Behavior,’’ World Development, 14(2): 233-243, 1986.
Rosenzweig, M. R. and Schultz, T. P., ‘‘Fertility and Investments in Human Capital: Estimates of
the Consequences of Imperfect Fertility Control in Malaysia,’’ J. Econometrics, 36: 163-
184, 1987.
Rosenzweig, M. R. and Wolpin, K. I., ‘‘Specific Experience, Household Structure, and
Intergenerational Transfers: Farm Family Land and Labor Arrangements in Developing
Countries,’’ Quarterly J. Econ., 100, Supplement: 961-987, 1985.
Schaffner, J. A., ‘‘Attached Farm Labor, Limited Horizons, and Servility,’’ Journal of
Development Economics, 47(2): 241-270, August 1995.
Shapiro, D., ‘‘Farm Size, Household Size and Composition, and Women’s Contribution to
Agricultural Production: Evidence from Zaire,’’ J. Dev. Studies, 27 (1): 1-21, Oct. 1990.
Singh, I., Squire, L., and Strauss, J., Agricultural Household Models: Extensions, Applications
and Policy, World Bank, Washington, D.C., 1986.
Skoufias, E., ‘‘Labor Market Opportunities and Intrafamily Time Allocation in Rural Households
in South Asia,’’ J. Development Economics, 40: 277-310, 1993.
Smith, L. C., Structural Adjustment and Welfare in Rural Africa: The Role of Resource Control
in Households, Department of Agricultural Economics, University of Wisconsin, Staff
Paper Seriew No. 368, Madison, January 1994.
St-Amour, P. and Vencatachellum, D., Family Organization, Retirement and Sectoral
Employment in Developing Agricultural Economies, Department of Economics, Laval
 41

University, Quebec, June 1996. (mimeograph).

Thomas, D. and Chen, C., Income Shares and Shares of Income: Empirical Tests of Models of
Household Resource Allocation, Labor and Population Program Working Paper 94-08,
Santa Monica CA, 1994. The Rand Corporation.
Thomas, D., ‘‘The Distribution of Income and Expenditure within the Household,’’ Annales
d’Economie et de Statistique, 29: 109-135, 1993.
Udry, C., ‘‘Gender, Agricultural Production and the Theory of the Household,’’ Journal of
Political Economy, 104(5): 1010-1046, October 1996a.
Udry, C., Efficiency and Market Structure: Testing for Profit Maximization in African
Agriculture, Department of Economics, Northwestern University, Evanston, June 1996b.
(mimeograph).
von Braun, J. and Webb, P. J., ‘‘The Impact of New Crop Technology on the Agricultural
Division of Labor in a West African Setting,’’ EDCC, 37, no.3: 513-534, Apr. 1989.