RESEARCH DOSSIER
Household Bargaining and Intrahousehold Allocation
Abstract. The household combines a shared feasible set with multiple persons whose interests, resources, access rights, time constraints, and opportunities may differ. Household bargaining and intrahousehold allocation examine how these differences determine the distribution of consumption, labor, leisure, care, savings, risk, and access to collective goods. This monograph reconstructs the historical transition from common-preference demand to cooperative bargaining, collective rationality, noncooperative interaction, and limited intertemporal commitment. It develops transparent consequences of weighted utility maximization, income pooling, distribution-factor exclusions, and individual participation constraints. Comparative examples establish why aggregate expenditure, the identity of a transfer recipient, and reported decision-making authority each supply incomplete evidence about individual welfare. Research from the United Kingdom, the United States, South Africa, Mexico, the Philippines, and Malawi is examined in relation to the specific identification strategies and maintained assumptions that support its findings. Official survey guidance situates individual-level assets, decision participation, and time use in a coherent measurement program. The analytical synthesis establishes the distinction between changes in the household opportunity set, changes in the internal allocation of that set, and changes in the capacity to sustain mutually advantageous cooperation. The analysis is a critical theoretical and empirical literature synthesis with original worked examples; it introduces no newly named theory and reports no original microdata estimate.
Keywords: Household bargaining; intrahousehold allocation; collective household models; income pooling; Pareto weights; distribution factors; sharing rules; household production; limited commitment; individual poverty; gender economics; household finance; causal identification
Object and historical lineage
The person within the household
The economic household is a unit of interdependent decisions. It converts earnings, accumulated assets, transfers, time, market goods, and domestic production into individual and shared achievements. A person belongs to this unit through relations of residence, resource sharing, responsibility, or mutual dependence, and these relations may overlap imperfectly. A survey household defined by a common dwelling therefore provides a statistical boundary that must be examined before it becomes the boundary of an economic model.
Consider two adults who live together, one of whom pays rent and the other cares for a child. Their monetary expenditures may be recorded under one reporting unit even when legal ownership of wealth, access to cash, discretion over leisure, and authority over future choices remain distinct. An increase in total resources expands possible allocations; a change in the decision procedure determines which allocation becomes actual. These are separate analytical operations.
The central research object is the relation between personal preferences, individually relevant constraints, and the internal organization of joint choice. Bargaining power denotes influence over the resolution of decisions involving divergent interests. A person's resource share describes an allocation of command over resources under a specified accounting or structural representation. Welfare is a further object, involving what those resources and choices achieve for that person. None of these measures can be substituted for the others without additional assumptions.
The historical development of the question
Traditional household demand can be represented by maximization of one objective subject to a joint constraint. Such a representation can describe common preferences, centralized decision-making, or a stable weighted aggregation of individual preferences. Its practical value is that it permits analysis of household expenditure and labor supply without separately estimating every internal decision.
The research agenda shifted when economists explicitly incorporated differences in personal interests and alternatives to cooperation. Manser and Brown (1980) gave an early bargaining treatment of marriage and household decisions. McElroy and Horney's work in 1981 belongs to the same historical lineage, as documented by Lundberg and Pollak. A marriage can create a surplus through sharing, specialization, and risk pooling; the existence of this surplus leaves its division open. Bargaining analysis supplies an allocation mechanism tied to the consequences of failing to agree.
Lundberg and Pollak (1993) subsequently developed a model in which the relevant fallback is a noncooperative arrangement inside marriage. This extension matters where exit is difficult, undesirable, costly, or unnecessary for conflict to arise. The allocation can react to changes in the organization of responsibilities even when both partners intend to remain together.
The collective approach adopted a different level of abstraction. In Collective Labor Supply and Welfare, Chiappori (1992) represents household members separately and studies the consequences of efficient joint choice. A precise bargaining game need not be specified in order to derive empirically meaningful restrictions. The approach focuses on the properties of observed allocation and leaves the process selecting a point on the efficient frontier partly unspecified.
Browning and Chiappori (1998) extend this reasoning to a general demand system and discuss the empirical restrictions of collective behavior. Their evidence supports examining couples differently from single-person households, subject to assumptions required by the particular demand-system specification. The conceptual advance is testability: internal distribution becomes an empirical question rather than an unobservable residual that can always be absorbed into household preferences.
These developments establish the historical sequence relevant to this dossier: aggregated choice, explicit bargaining, collective efficiency, and the study of failures of enforceability and information. Each approach specifies a different relation between observed cooperation and the underlying decision process.
Definitions, assumptions, and institutional scope
The feasible set
For a two-adult household, let individuals be indexed by i = A, B. Each person has a private consumption vector c_i, market work h_i, domestic work d_i, and leisure l_i. Let T_i be available time and q a vector of shared goods or household services. Let w_i be the wage and y_i nonlabor income. Prices of private and shared market goods are p and r, respectively. In a one-period setting without borrowing or saving, a simple feasibility representation is
The first equation abstracts from taxes, transfers in kind, transaction costs, and nonmarket resources. The second treats time activities as mutually exclusive; simultaneous care and other activities require a different accounting convention. Household production can be added as q = F(d_A,d_B,m), where m is a vector of intermediate inputs. In that formulation the market expenditure on m and any purchased final collective goods must be counted once, rather than charging both the inputs and the imputed value of output.
The assumption of a joint budget requires the ability to transfer purchasing power. Legally segregated property, inaccessible bank balances, earmarked transfers, individual debts, or restrictions on travel can require separate constraints. In particular, a bank account balance observed by the researcher can overstate spendable resources for a person who lacks the authority to withdraw from that account.
The constraint depends on the institutions of access. A household can possess a valuable asset while one member has low practical liquidity. Likewise, the opportunity cost of care is conditional on employment possibilities, wage rates, transport, hours restrictions, and the availability of substitutes for domestic production.
Preferences, public goods, and externalities
An individual utility function may depend on her own goods, leisure, and shared services. It may also incorporate concern for a partner or a child, expectations about obligations, or preferences regarding how choices are made. Writing utilities separately does not require that agents be selfish. Interdependent preferences are compatible with collective analysis, although they complicate the decentralization of efficient allocations.
A private good in the strict model benefits its recipient without directly entering another person's consumption. An assignable expenditure is an observed purchase whose intended consumer can be identified. The distinction is important: clothing purchased for one person may be observed but its value, actual use, or contribution to welfare can vary systematically. A shared service such as residential heating may benefit multiple members at once while producing unequal gains because they spend different amounts of time at home.
The number of relevant persons is an empirical question. Dependent children, older relatives, and nonresident contributors may affect both the feasible set and the objectives. An adult bargaining model can incorporate a child through parental utility or a separate welfare criterion. These representations have different implications when the child's own interests diverge from adults' choices.
Measures of influence and distribution
Jayachandran and Voena (2026) define household power through influence over collective choices and distinguish it from broader concepts of empowerment. A bargaining weight measures the relative coefficient on an individual's utility in a specified model. Self-reported participation measures reported authority in a domain. Ownership indicates a category of legal or reported property rights. Independent access to an account describes a practical means of controlling purchasing power. Each can be investigated separately.
Suppose a household allocates equal private expenditure to two persons with unequal health needs. Equality of expenditure is consistent with unequal achieved welfare. Suppose instead the spouses report joint decision-making but one controls the information about the household's debts. Joint reporting can coexist with unequal influence. Such examples motivate a research design that records resources, information, practical agency, and outcomes simultaneously.
Formal mechanisms of household bargaining
Unitary choice and the income-pooling restriction
Let total nonlabor income be Y = y_A + y_B, with a fixed price and wage vector. Under a single stable objective and a joint budget, demand for good k can be expressed as x_k = X_k(p,w,Y). At an interior differentiable point,
The equality follows from the chain rule, conditional on other arguments and the absence of source-specific constraints. A redistribution of income between adults at unchanged total income is predicted to leave demand unchanged. The restriction is about the effect of the origin of income, not the proposition that each adult spends the same amount.
This is a useful null hypothesis because it can fail in systematically interpretable ways. It also has clear limits. If an income source has a different tax treatment, comes with a designated use, changes eligibility for benefits, or alters working time, it may change the joint feasible set. Then a response to its source does not isolate bargaining power. Estimation must respect the exact conditions of the theoretical comparison.
Cooperative bargaining
Let S be the feasible utility set and let d_A and d_B denote the utilities attainable without an agreement. Under the standard cooperative Nash specification, with bargaining asymmetry represented by a parameter a in (0,1), the selected point solves
For an interior solution with strictly positive gains, maximizing the logarithm produces the same allocation. In a smooth allocation representation using a scalar z,
The expression equates weighted proportional marginal increments in gains from agreement. It concerns utilities relative to the disagreement point, not merely the total utilities received. If the feasible utility set is nonconvex, the feasible bargaining problem may have multiple local solutions; existence and uniqueness require attention to the shape and closure of the relevant set. The standard solution's axiomatic interpretation requires its usual conditions.
Consider a transparent transferable-utility example: u_A + u_B = 100, with d_A = 20 and d_B = 10. Let a = 0.5. Agreement generates a surplus of 70; equal Nash bargaining shares imply u_A = 55 and u_B = 45. Raise d_A to 40 while leaving the frontier and d_B unchanged. Now the surplus is 50 and the outcome is u_A = 65, u_B = 35. The fallback improvement raises A's level even though the gain from cooperation falls for both.
The result generalizes in this particular linear model to u_A = d_A + a(100 - d_A - d_B). Its derivative with respect to d_A is 1 - a. The improvement need not be one-for-one. The coefficient is illustrative and depends on the transferable frontier and fixed a; no universal numerical effect follows.
Collective efficiency and Pareto weights
The collective formulation selects an allocation by maximizing a weighted combination of individual utility functions:
Here F is the feasible allocation set, and mu may depend on prices, nonlabor income, institutional variables, and the household's circumstances. For given positive weights, every global maximizer is Pareto efficient. Otherwise there would exist an allocation increasing at least one utility without reducing the other, which would increase the weighted objective. Conversely, convexity of the attainable utility set and supporting-hyperplane regularity permit efficient points to be represented by nonnegative supporting weights; positivity and uniqueness need further assumptions.
The weights are coordinates of a welfare representation. An affine rescaling of one person's utility changes the numerical coefficient required to represent the same choice. For example, replacing U_A by 2U_A can be offset by dividing its unnormalized coefficient by two and renormalizing both coefficients. This invariance is essential when interpreting estimated weights as measures of bargaining influence. A weight comparison across populations requires a comparable utility normalization.
A collective representation can accommodate changes in bargaining circumstances without selecting a specific disagreement point. Its flexibility has empirical costs. A different weight may rationalize an altered decision even when the investigator has little information about the mechanism changing the weight. Credible restrictions on its determinants create identification.
Private consumption and sharing rules
Assume two adults consume only a private composite good, each unit costs one, income is M, and both have logarithmic utility. For positive consumption, maximize
The first-order conditions give mu/c_A = (1 - mu)/c_B. Strict concavity ensures the unique optimum:
At M = 120 and mu = 0.4, A receives 48 and B 72; at the same M and mu = 0.6, A receives 72 and B 48. Total expenditure is identical. The allocation is efficient in both cases and the shift improves A's utility while reducing B's.
Under the relevant separability and private-goods assumptions, a sharing representation can assign individual budgets rho_i with rho_A + rho_B equal to household disposable resources. Individuals optimize within their assignments. In Chiappori's labor-supply formulation (1992), labor choices yield restrictions useful for recovering preferences and the sharing rule, with an additive-constant qualification. The recovered rule is not automatically a bank balance, a cash transfer, or a direct record of legal ownership.
With shared consumption, resource assignment is less direct. The financing incidence of a jointly used dwelling may differ from the benefit enjoyed by each occupant. A full collective model may require a consumption technology, personal demand equations, and a definition of shadow costs before a sharing rule can be interpreted.
Distribution factors and restrictions
Introduce a variable z that influences mu without changing the feasible set or personal utility functions directly. It is a distribution factor in the model. Let household demand be X(p,M,mu(z)); differentiation yields
For two or more such factors, the matrix of these demand derivatives has rank at most one wherever the smooth representation holds. Columns are proportional unless all derivatives vanish. This elementary implication explains why multiple factors can support restrictions beyond a test of income pooling.
For more than two decision makers, a normalized vector of n positive weights has n minus one independent coordinates, so the corresponding rank bound expands to at most n minus one under an analogous differentiable representation. Rank restrictions cease to have empirical force when the dimension of observables is too small relative to the dimension of the allowed distribution index.
Chiappori, Fortin, and Lacroix (2002) treat marriage-market sex ratios and divorce rules as candidate distribution factors and derive labor-supply restrictions. Their exclusion is an economic assumption: changes in divorce law could also change expected wealth, fertility, residence, and marriage composition. A demographic variable can shift wages through migration and local labor markets. Such pathways have to be assessed in the setting of a particular test.
Comparative models and counterexamples
The limits of observational equivalence
The unitary and collective models can generate identical measured choices if the Pareto weight is constant and the feasible set is correctly represented. In the logarithmic example, total consumption is M for every weight. Even observing changes in total consumption across income levels does not identify individual shares.
If observed data include one private expenditure category for each person, the information increases. Still, the ratio of spending on that category depends on individual preferences and the mapping from expenditures to welfare. A clothing ratio can change because a person's taste for clothing changes, because prices available to that person change, or because their command over resources changes. Structural interpretation requires explicit restrictions.
Conversely, source-dependent spending can occur even with stable personal preferences and full efficiency. The weight can depend on which individual receives income; different outcomes need not imply foregone joint surplus. Income pooling, Pareto efficiency, equality, and one-person authority are logically distinct hypotheses.
Noncooperative contributions to a shared good
Let each adult have one unit of a privately consumed resource, write g_i for their contribution to a common service, and let G = g_A + g_B. Their utilities are
With symmetric interior noncooperative choices, each adult treats the other's contribution as given. The first-order condition is beta/G = 1/(1 - g_i). In a symmetric equilibrium, g_i = beta/(beta + 2). A jointly selected symmetric optimum maximizing the sum of both utilities instead gives g_i = beta/(beta + 1).
At beta = 1, decentralized separate contributions are one third each, while the efficient symmetric contribution is one half each. The private consumption and common provision differ because each unilateral decision ignores the positive benefit contributed to the other person. Both adults obtain higher utility at the cooperative symmetric allocation in this example: each utility is log(1/2) + log(1), versus log(2/3) + log(2/3) under noncooperation.
This failure arises from the assumed external benefit and decentralized control, not from divergent preferences alone. Agreements, repeated interaction, monitoring, or institutions that internalize the benefit can change the result. With strategic substitutes and heterogeneous incomes, equilibrium contributions can also become asymmetric and some persons contribute nothing.
The separate-spheres model of Lundberg and Pollak (1993) proposes a related reason to examine fallback allocations inside the household. Traditional assignments of provision and responsibilities can determine the resources each person can use without obtaining agreement. A transfer can alter these practical alternatives rather than simply raising total spending.
Production efficiency and allocation
Suppose two plots produce output according to square-root functions of labor: f_A(L_A) = 4 sqrt(L_A) and f_B(L_B) = 2 sqrt(L_B), with ten units of homogeneous labor freely transferable. Efficient output maximization requires 2/sqrt(L_A) = 1/sqrt(L_B), hence L_A = 4L_B. With the total constraint, the efficient division is 8 and 2. Output is 4 sqrt(8) + 2 sqrt(2), approximately 14.142, compared with 6 sqrt(5), approximately 13.416, under equal labor division.
This calculation isolates the production implication of unequal marginal productivity. A researcher observing equal labor on empirically different plots cannot infer inefficiency from these stylized functions. Actual labor types, soil quality, transport costs, risk, private rights to harvest, credit restrictions, and measurement error may explain the allocation. The relevant welfare frontier is conditional on actual constraints, not the unconstrained maximum constructed by an analyst.
The production and consumption tests answer different questions. Efficient exchange of produced output can coexist with inefficient production, and formally efficient production can coexist with inequitable consumption. A complete household assessment must investigate both margins.
Intrahousehold allocation over time and under risk
Joint saving and limited commitment
A household uses current income for present consumption and accumulation of resources available later. Let a_t represent jointly usable wealth, R the gross interest factor, Y_t income, and C_t total expenditure. With a simple asset convention,
Borrowing restrictions, liquidity, asset-specific returns, and individually controlled accounts can create additional state variables. When a_t represents the balance at the start of period t, the expression assumes returns accrue before that period's choices and earnings. Other timing conventions change the equation without changing the conceptual role of saving.
Under full commitment with a common discount factor delta, an illustrative dynamic collective program maximizes expected discounted weighted utilities with an initial fixed pair of weights. In a deterministic interior one-good problem without binding financial constraints, each period's shadow value lambda_t satisfies lambda_t = delta R lambda_(t+1). Individual marginal utilities adjusted by the fixed weights equal that shadow value. The allocation of future consumption can therefore be linked to present decisions by the planned intertemporal structure.
When future agreements cannot be enforced, each person's continuation utility must be compared with alternatives available at that date and state. If the promised utility is below the available alternative, renegotiation or exit becomes relevant. The resulting weights can vary with shocks and history. A household may be efficient conditional on current participation restrictions and still fail to implement the ex ante allocation attainable under full commitment.
Mazzocco (2007) derives a test distinguishing full-commitment and relaxed-commitment Euler implications. Using the Consumer Expenditure Survey, he reports rejection of the full-commitment hypothesis in his specification. The conclusion is conditional on the data and economic restrictions in that test; it does not establish that all households frequently renegotiate or that every observed Euler-equation residual measures a bargaining change.
Two-period illustration
Consider a no-interest, no-borrowing example with a fixed resource endowment of 100 over two periods and identical discounting equal to one. Let both persons have logarithmic utility of their own consumption in each period. With fixed equal weights, the unconstrained efficient plan gives each person 25 units each period.
Now imagine that period-two assets are controlled jointly but A can enforce a claim to at least 35 units in that period from the remaining resources. Keep total period-two resources fixed at 50. The equal sharing outcome violates A's minimum claim; the constrained solution is 35 for A and 15 for B. An anticipated claim could alter period-one savings and division, so fixing the period-two envelope is a partial-equilibrium example of renegotiation rather than a solved dynamic equilibrium.
A policy changing the ownership of savings can change the feasible claims and thereby change the allocation of resources over time. A one-period measure of the recipient's spending is incomplete when the main effect is a change in accumulated assets, future exit options, or longer-run labor supply.
Insurance and the identity of the insured
Risk pooling can create substantial value in a household. Suppose two individuals have incomes (100,0) or (0,100) with equal probability. Perfect pooling with symmetric log utility gives each 50 in both states. Without a transfer, zero consumption makes log utility undefined; replacing zero with a small positive endowment gives a finite but much lower expected utility. The illustration clarifies the value of sharing when income shocks are imperfectly correlated.
The implementation of insurance requires communication, trust, enforceability, or accessible common assets. If the person with positive income can retain all proceeds in each state and has no enforceable obligation, the proposed pooling rule may violate a state-specific participation constraint. If the two persons instead face the same aggregate income shock, internal transfers cannot restore the lost total resources. External insurance, credit, savings, or public support becomes relevant.
A common response to uncertainty is the accumulation of precautionary resources. When ownership and control are unequal, the motive for one person's saving may also include securing access to resources after a disruption in the household. Aggregate saving can conceal opposing individual motives.
Labor, specialization, and bargaining tomorrow
Household members can specialize in paid work or care according to relative productivity and institutional possibilities. The specialization may enhance current joint resources and at the same time shape future alternatives. A person reducing market work for care may accumulate less individual earnings history, lower personally controlled pension rights, and fewer opportunities upon dissolution. Current compensated cooperation and future dependence can therefore coexist.
A dynamic welfare analysis needs to record the division of accumulated claims, the durability of sharing arrangements, and the time at which compensation becomes enforceable. This consideration explains why static Pareto efficiency alone cannot determine the welfare consequences of a long-lived specialization decision.
Empirical indicators and identification
Matching the observable to the hypothesis
Empirical analysis begins by distinguishing three objects: the resource constraint, the mechanism of selection within it, and person-specific achievements. Income, assets, and prices characterize the first; source of income, control of accounts, legal options, and decision reports may inform the second; separately measured consumption, health, leisure, education, and financial security inform the third. Each data source covers only part of this chain.
| Indicator | Directly observed or constructed object | Main interpretive limitation |
|---|---|---|
| Individual earnings | Labor income received or reported | Endogenous labor supply and incomplete control |
| Legally held assets | Ownership or title status | Disposal rights and actual access can differ |
| Assignable expenditure | Purchase attributed to a person | Prices, quality, sharing, and need |
| Decision participation | Reported own or joint final say | Preference agreement, reporting norms, domain specificity |
| Time diary | Recorded activity duration | Simultaneous activities and category comparability |
| Child nutrition | Anthropometric outcome | Disease, sanitation, care, measurement and lagged inputs |
| Structural sharing rule | Estimated resource assignment | Identification and functional assumptions |
The World Bank's LSMS+ survey guidance concerns the collection of individual-disaggregated asset and labor information. It emphasizes the value of asking individual respondents about their own circumstances. An interview with one household representative may be an efficient way to measure total rent but less informative about concealed savings, individual ownership, or each person's own employment arrangements.
The World Bank's major-purchase decision indicator identifies currently married interviewed women ages 15–49 who report having final say alone or jointly. The denominator is the interviewed currently married women in that age range. A comparison with an indicator covering all adult women or one measuring sole authority requires a population and definition adjustment.
The OECD Time Use Database uses harmonized broad activity groups for ages 15–64 and warns about differences in diary design and activity classification. Time spent on unpaid work should be analyzed with participation, intensity, timing, simultaneous care, and leisure. A daily average over all respondents is a different statistic from an average conditional on participation in an activity. Harmonization improves comparability without eliminating measurement differences.
Reduced-form identification
For an allocation outcome Q_h, an empirical income-source test might estimate
With exogenous variation in the two income sources, stable prices, and relevant controls, the null beta_A = beta_B corresponds to the linear version of income pooling. An alternative uses total income and the income of one person; the coefficient on the latter then tests the source difference at fixed total income. The parametrization must reflect the units and support of the variation.
Ordinary least squares generally does not isolate causal effects if earnings are chosen in response to consumption needs, time availability, fertility, or unobserved preferences. Nonlabor income may also correlate with wealth, age, or past employment. A policy-driven income shift or randomized assignment can provide independent variation, but its interpretation depends on what the intervention changes.
When treatment assignment is randomized across households, the intention-to-treat effect compares assigned groups, regardless of actual take-up. This identifies the causal effect of assignment under the design assumptions, including adequate handling of attrition and interference. It does not by itself identify the causal effect of a latent bargaining weight. A transfer may change both resources and influence, and the corresponding pathways can have different effects.
Structural identification and distribution factors
Structural estimation combines observable choices with restrictions on individual preferences, the consumption technology, and the allocation process. For example, if assignable private goods obey comparable Engel-curve restrictions, variations in total resources and family structure can inform latent person-specific shares. The identifying restrictions need to be separated from functional forms used only for estimation.
Dunbar, Lewbel, and Pendakur (2013) use a collective model with semiparametric restrictions to infer resource shares from person-specific expenditures. This strategy gives information about children who lack individually recorded comprehensive consumption. It highlights the fact that identifying allocations is a joint exercise in demand theory and measurement.
Browning, Chiappori, and Lewbel (2013) incorporate a technology of shared consumption to investigate resource assignment and indifference scales. The approach distinguishes the cost savings of living together from the distribution of the resulting gains. Comparing an individual's standard of living alone and in a couple requires accounting for both economic scale and personal allocation.
An identification argument should state which parameters are point identified, identified up to a normalization, bounded, or unidentified under the maintained assumptions. Standard errors conditional on a model do not convey uncertainty about whether those assumptions hold. Sensitivity analysis should vary the structural assumptions as well as the sampling specification.
Research implementation
Before using a particular survey, register the statistical unit, geographic coverage, dates, sampling design, weights, restrictions imposed by nonresponse, and the way out-of-household members are recorded. Obtain the household roster, questionnaire, codebook, and the exact version of the microdata. Link person records to expenditure or asset records using verified identifiers and document unmatched observations.
Convert nominal expenditure to suitable real purchasing-power units before making comparisons across dates or locations. Distinguish spatial cost-of-living adjustment within a country from conversion for international poverty analysis. Report the price assumptions, recall intervals, treatment of durable goods, imputation of rent if used, and the treatment of home production.
In panel research, use the temporal ordering of incomes, policy exposure, ownership, and outcomes. Within-household fixed effects can absorb time-invariant unobserved attributes, but not changing preferences, anticipation, or reverse causality. Household dissolution and formation create meaningful selection rather than ordinary missingness. A panel restricted to continuously married respondents may exclude the individuals for whom constraints and fallback options matter most.
No microdata were downloaded or estimated for this monograph. Its numerical examples are deterministic constructions; the empirical claims below report the identified authors' results within the scope of accessible evidence.
International and historical evidence
A change of payment recipient in the United Kingdom
In the late 1970s a change to United Kingdom child benefits redirected resources toward wives. Lundberg, Pollak, and Wales (1997) exploit this institutional change using Family Expenditure Survey data. They report a relative shift toward women's and children's clothing compared with men's clothing around the reform.
The study is important because the payment arrangement, rather than an ordinary variation in earnings, supplies the comparison. Its interpretation nevertheless requires attention to other contemporaneous changes and to the extent that the total disposable resources and payment characteristics were comparable. The result challenges the stable common-preference income-source restriction. It does not directly measure the utility of wives or children, and it cannot be read as an estimate of a universal propensity to spend on children.
South African pensions and multigenerational allocation
Duflo (2003) analyzes the expansion of pensions in South Africa in the early 1990s. The reported estimates associate pensions received by women with gains in the anthropometric status of girls; comparable gains are not found for pensions received by men, and effects for boys are limited in the reported findings.
The multigenerational setting expands the household bargaining question beyond spouses. A grandmother's access to a transfer can affect resource flows to grandchildren through co-residence, food, care, or other activities. The health outcome is downstream of these mechanisms, and the contrast between recipients does not uniquely identify relative preferences or a bargaining coefficient.
The inference should remain within the program, population, and outcome of the study. Pension eligibility, household composition, and decisions about living with relatives can all structure exposure to the transfer. No general statement about the behavior of all grandmothers or grandfathers follows.
Mexico and collective responses to cash transfers
Bobonis (2009) combines randomized program variation in women's income with variation connected to rainfall to examine restrictions associated with efficient allocation. The reported findings favor compatibility with Pareto efficiency in the studied setting. Attanasio and Lechene (2014) analyze rural Mexican conditional transfers with conditional-demand tests and candidate factors including the relative characteristics of family networks. They report nonrejection of the efficiency restrictions and improved predictions of income responses relative to their unitary comparison.
These studies clarify why transfers paid to one spouse can change spending without establishing household inefficiency. A change in the distribution of decision influence may move the household along an efficient frontier. The empirical challenge is to justify the exclusions needed to represent candidate factors as affecting distribution rather than preferences or the feasible resource technology.
The two studies refer to related policy settings. They supply complementary research designs; they should not be counted as independent evidence from different countries. Nonrejection means that the data, specification, and test do not provide sufficient grounds to reject the restrictions; the interpretation must allow observationally equivalent mechanisms and limited power.
Private financial control in the Philippines
Ashraf (2009) varies whether individual financial decisions are private, observable, or communicated to a spouse. The experiment finds that observed allocation choices respond to the information treatment and that patterns depend on which partner ordinarily controls savings. The comparison connects the organization of information to financial behavior.
A person may prefer a private saving instrument because it permits the accumulation of resources under their own control. If a decision becomes observable, the same person may choose an expenditure that is more difficult to reallocate. Communication can alter the feasible agreement. These mechanisms are distinguishable in principle even when expenditures in a household budget are unchanged in total.
A lab-in-the-field or structured decision experiment typically identifies a treatment contrast for the participants and stakes studied. Extrapolating it to a lifetime household allocation requires assumptions about changes in trust, monitoring, real financial access, and repeated choices.
United States divorce law and expectations
Chiappori, Fortin, and Lacroix (2002) use 1988 Panel Study of Income Dynamics data to investigate household labor supply in relation to marriage-market and legal variables. Their reported results are compatible with the estimated collective restrictions.
Voena (2015) studies how divorce law relates to intertemporal behavior. The paper reports that unilateral divorce in states with equal property division is associated with higher household savings and lower female employment in its analysis. The combination of dissolution rules and property division matters because the consequences of exit depend on the distribution of accumulated claims.
An empirical comparison of legal reforms must account for differences in their timing, transition rules, other state policies, trends, anticipation, and selection into or out of marriage. The study's setting supplies evidence of an intertemporal institutional mechanism. Its reported direction is not a universal prediction for every property regime.
Individual child poverty in Malawi
Dunbar, Lewbel, and Pendakur (2013) apply their resource-sharing identification strategy to Malawi and report that standard household poverty indices underestimate the incidence of child poverty in their analysis.
The problem emerges because household-wide expenditure adjusted for size does not necessarily capture how much each person can use. An equivalence scale can correct for shared costs and some needs; an allocation model is needed to investigate the internal distribution associated with those costs. The Malawi result depends on the authors' maintained demand restrictions and empirical population. It motivates individual-level analysis without supplying a universal correction factor to poverty statistics.
Comparative conclusion
The British reform studies variation in who receives resources, South African pensions reveal intergenerational incidence, Mexican transfers supply collective-demand tests, Philippine experiments vary information, American legal changes connect present choices to future claims, and the Malawi research investigates unobserved resource division. These settings address different estimands.
Comparability resides in the underlying questions, not in a common numerical measure of bargaining power. The source and timing of variation, the outcome, the definition of who belongs to the household, and the constraints imposed by institutions must be aligned before findings are compared. A policy that changes the recipient may operate predominantly as an income increase where resources are shared freely, and as a change in influence where control over spending is enforceable.
Measurement, robustness, and contested interpretations
Poverty accounting and household size
A stylized example illustrates how an aggregate poverty threshold can conceal low individual command over resources. Suppose two adults consume a private composite good, have household expenditure 100, and have an individual poverty threshold of 45. If expenditure is divided equally, both are above the threshold. If the assignment is 70 and 30, one person is below it. Per-capita household expenditure remains 50 in both cases.
Suppose the threshold changes to 35. The same person remains below it in the unequal allocation; at a threshold of 25, both are above it. The estimated rate therefore depends on the welfare measure, threshold, and internal division. This example has no common goods or economies of scale; extending it to actual poverty measurement requires a justified individual consumption metric.
For households with a common rent payment, total resources cannot be distributed by counting the same rent as separately purchased by each person. The assignment of benefits or equivalent expenditure should reflect the model of consumption technology. The distinction between expenditure equality and equal standards of living is particularly important when people have differing health or care needs.
Unobserved preferences and reporting
Consider survey questions about who makes decisions. Two respondents may report joint choice because they agree, because both participate meaningfully, or because one proposes options and the other accepts them under pressure. If their preferences coincide, observing the final choice may be almost uninformative about which person has influence under disagreement. A bargaining study benefits from variation in decision domains and knowledge of each person's preferences and alternatives.
Information about household balances may differ between partners. Recall error and asymmetric knowledge can produce conflicting reports without deliberate deception. Sensitive interviews should protect privacy, make participation voluntary, and avoid creating household risks through disclosure. The statistical meaning of a response depends on the questionnaire and collection method.
Interpreting observed gender patterns
The empirical comparisons in the earlier chapters involve gendered institutions and relationships. They do not establish fixed or universal preferences based on gender. The roles of paid work, care, inheritance, property ownership, and safety can change across populations and over time.
The 2026 synthesis by Jayachandran and Voena discusses mixed effects of transfers targeted to women and more consistent evidence for some interventions that increase control over earnings. It also emphasizes the difficulty of inferring the consequences of power from policies that change resources and bargaining opportunities simultaneously. This evidence supports examining the specific intervention rather than using gender of recipient as a complete description of its mechanism.
The relevant policy objective may be a person's agency, child development, gender equality, joint consumption, or protection against economic harm. These goals can be connected, but empirical success on one indicator does not automatically establish success on all others.
Robustness and research audits
A valid investigation should first establish that assignment to a policy or other identifying variation has the claimed properties. Balance and attrition tests, pre-intervention outcomes, checks of sample construction, and appropriate sampling or assignment-level inference matter. For a policy adopted in particular jurisdictions, the identification strategy must address differential trends and changes in population composition. For an instrument, the researcher should specify relevance, exclusion, and the population affected by the variation.
Sensitivity analysis must go beyond changing the control variables in one regression. For a collective model, estimate or bound the implications of plausible changes in preference restrictions and assignment of shared goods. For an income-source test, examine whether source differences remain when eligibility, in-kind restrictions, and transaction costs are included. For a poverty application, vary the welfare metric, resource-sharing assumptions, price deflators, and threshold.
If a study has multiple outcomes, distinguish prespecified outcomes from exploratory comparisons and consider the inference consequences of multiple testing. Quantitative syntheses require effects with compatible units, target populations, uncertainty estimates, and treatment definitions. The studies surveyed here do not provide a common estimand; averaging their coefficients would not yield a defensible measure of household bargaining power.
Analytical synthesis and welfare implications
Separating opportunity, distribution, and enforceability
The author's synthesis treats three logically distinguishable aspects of household allocation. The first is the economic opportunity set, which contains the consumption, production, saving, and time combinations available under actual institutional constraints. The second is selection from that set, which depends on preferences and the decision procedure. The third is enforceability, which determines whether the selected arrangement can be implemented over time and across states of the world.
This distinction is an analytical organization of existing household economics rather than a new model. Its purpose is to prevent an observed change in one category from being incorrectly attributed to another.
A transfer of 20 in the private logarithmic example illustrates the point. Start at M = 100 and mu = 0.4, so A consumes 40 and B 60. If the transfer expands the budget to 120 while the weight remains unchanged, A consumes 48 and B 72. If the transfer simultaneously raises mu to 0.5, each consumes 60. Compared with baseline, A gains 20 and B is unchanged. The same budget increase can thus yield different personal gains.
The budget effect evaluated at the original weight is 8 for A and 12 for B. The distribution effect evaluated at the expanded budget is 12 for A and negative 12 for B. The sum of both components gives the total change. An alternative decomposition evaluates the weight at the initial budget and the budget at the new weight; the separate components then differ. Accordingly, the decomposition depends on the reference path, even though the total change is invariant.
This worked illustration is an algebraic exercise, not an empirical estimate or an identified mediation effect. It explains why a measured increase in individual consumption cannot be labeled a bargaining effect until the resource pathway has been addressed.
Welfare and distribution are distinct judgments
A Pareto-efficient allocation exhausts mutually feasible utility gains. It does not settle distributional justice. The private-good example with logarithmic preferences and unequal consumption lies on the frontier because increasing one person's consumption at fixed resources requires reducing the other's. Ethical ranking of frontier points requires a further welfare criterion.
A policy could expand feasible resources and allow both individuals to gain in principle while actual resource assignment favors one. The aggregate surplus created by a policy and the personal material improvement realized by each affected person must be assessed separately. This distinction holds even if the decision procedure is fully cooperative.
A public evaluator can define a social welfare function over individuals, thresholds for basic needs, or constraints on individual rights and agency. Each is a normative choice to be stated explicitly. The Pareto weights inferred from household behavior are descriptive coordinates within the maintained representation and do not supply an independently justified social welfare ranking.
Implications for transfer design and labor policy
A targeted transfer is characterized by its value, recipient, account of payment, conditions, timing, and the rules governing continued access. A payment into a recipient's individual account may raise household resources but provide little sustained personal control if the money can be appropriated immediately. A transfer with meaningful personal access changes a different set of feasible individual actions.
Evaluation should measure take-up, individual control, effects on other household members, time use, and future financial security. Where coercion or retaliation is a concern, privacy and safety are part of the design constraints. Reporting average household income alone can miss distributional costs.
Labor-market policies influence wages, employment access, working hours, and the value of unpaid time. These changes can affect both the household budget and personal fallback opportunities. The total employment response depends on the interaction of the substitution effect, income effect, redistribution, and constraints on care. A single directional prediction across household institutions would require assumptions that are not generally warranted.
A childcare service may change household production opportunities and release time for paid work or leisure. Its benefits and costs can be assessed by changes in time allocation, service use, earnings, child outcomes, and decision influence. The household's total monetary income provides only one dimension of the evaluation.
A decision procedure for interpreting findings
An analyst can organize a study by stating the household boundary and recording who controls the relevant margin. Next, the analyst specifies the feasible set and which variables change it. For the hypothesized allocation mechanism, the relevant comparative static is derived under explicit assumptions. Identification then establishes what variation can distinguish that mechanism from changes in preferences, resources, technology, information, or membership.
The findings can be interpreted at their justified level: descriptive association, causal assignment effect, conditional structural inference, or normative welfare assessment. Distinguishing these levels prevents statistical precision from being mistaken for institutional generality.
The principal contribution of this monograph is an integrated account of the relationship between static income pooling, distribution-factor rank restrictions, shared consumption technology, intertemporal enforceability, and person-specific welfare. The formal examples demonstrate observational equivalence and institutional contingency using internally transparent assumptions. The synthesis introduces neither an original empirical estimate nor a claim of priority for a new theory.
Research frontiers
More than two decision makers
Households with older relatives, multiple earners, guardians, or adult children may contain several decision participants. A two-person representation can omit channels through which care, remittances, housing, and property claims are distributed. The modeling problem grows with the number of independently varying weights and with the possibility of shifting coalitions.
Research should distinguish resident membership from economic participation and should investigate the conditions under which an additional decision maker changes the observable rank restrictions. Data identifying individual transfers across households can clarify the analytical boundary in multigenerational and migrant families.
Changes in preferences, information, and norms
A long-run change in education, legal rights, or employment opportunities can influence not only bargaining positions but also aspirations, expectations, and sharing norms. The strict distribution-factor exclusion becomes more demanding over long horizons. This motivates designs that combine short-run randomized variation with longer-run measures of preference, participation, and institutional change.
Relevant research questions include whether the allocation process changes after shocks, how partners learn about each other's opportunities, which forms of private financial access improve security, and how changes in family organization affect the transfer of resources across generations.
Individual welfare in social statistics
Official household income and consumption aggregates serve important comparative functions. Individual-disaggregated observations can supplement them by identifying ownership, time burdens, and uneven access to resources. Survey design needs to balance respondent burden, confidentiality, sampling precision, and the cost of obtaining person-specific information.
Future work can compare direct individual measurement with structurally estimated allocations on datasets where both are available. This would help determine where sharing-rule estimates are robust, where they depend strongly on assumptions, and which survey additions offer the largest information gains.
The long-run research frontier concerns the measurement of household cooperation as an institution with a history. The division of current expenditure is one observed outcome in that history. The assignment of resources and claims before formation, during cooperation, and after dissolution shapes individual economic possibilities in ways that cross-sectional totals can only partially reveal.
Scope of evidence
The evidence reviewed in this dossier is deliberately distinguished by access level. Some original articles are represented by publisher abstracts or author-supplied records; selected full-text passages were examined for the collective-demand paper, the British child-benefit study, and the 2026 literature review. No original estimates, microdata replication, exhaustive systematic review, or coefficient-level meta-analysis are asserted. Empirical numerical parameters are therefore not imported into the worked examples. The theoretical derivations are illustrative and conditional on their stated functional forms and feasibility assumptions.
Conclusion
Household bargaining and intrahousehold allocation form an essential part of microeconomic analysis because a shared resource constraint does not determine the allocation among the persons who live and cooperate within it. The identities of beneficiaries, the internal decision mechanism, access to property and information, and the enforceability of intertemporal agreements influence the choices ultimately realized.
Unitary household choice supplies the restrictive benchmark of income pooling. Cooperative bargaining relates division to the surplus of agreement and alternatives to agreement. Collective theory studies the restrictions of efficient joint choice without specifying a complete bargaining protocol. Noncooperative and dynamic approaches explain why the organization of responsibilities, information, and future rights can alter feasible cooperation.
International research demonstrates that these mechanisms have observable implications. Its conclusions remain tied to particular interventions, data, institutions, and identifying assumptions. A shift in the source of income may reveal distributional effects even if total resources and efficiency are unchanged. A measured change in consumption, labor, or a child's welfare may express several mechanisms acting together.
The rigorous analysis of households therefore requires separately identifying resources, the rules of allocation, personal outcomes, and enforceable claims. This distinction gives policy and welfare research the capacity to investigate the persons who constitute the household as well as the aggregate economic unit they sustain.
References
- Nava Ashraf. 2009. Spousal Control and Intra-household Decision Making: An Experimental Study in the Philippines. American Economic Review 99 (4): 1245–1277. DOI: 10.1257/aer.99.4.1245. Source.
- Orazio P. Attanasio and Valérie Lechene. 2014. Efficient Responses to Targeted Cash Transfers. Journal of Political Economy 122 (1): 178–222. DOI: 10.1086/674968. Source.
- Gustavo J. Bobonis. 2009. Is the Allocation of Resources within the Household Efficient? New Evidence from a Randomized Experiment. Journal of Political Economy 117 (3): 453–503. DOI: 10.1086/600076. Source.
- Martin Browning and Pierre-André Chiappori. 1998. Efficient Intra-Household Allocations: A General Characterization and Empirical Tests. Econometrica 66 (6): 1241–1278. Source.
- Martin Browning, Pierre-André Chiappori, and Arthur Lewbel. 2013. Estimating Consumption Economies of Scale, Adult Equivalence Scales, and Household Bargaining Power. Review of Economic Studies 80 (4): 1267–1303. DOI: 10.1093/restud/rdt019. Source.
- Pierre-André Chiappori. 1992. Collective Labor Supply and Welfare. Journal of Political Economy 100 (3): 437–467. DOI: 10.1086/261825. Source.
- Pierre-André Chiappori, Bernard Fortin, and Guy Lacroix. 2002. Marriage Market, Divorce Legislation, and Household Labor Supply. Journal of Political Economy 110 (1): 37–72. DOI: 10.1086/324385. Source.
- Esther Duflo. 2003. Grandmothers and Granddaughters: Old-Age Pensions and Intrahousehold Allocation in South Africa. World Bank Economic Review 17 (1): 1–25. DOI: 10.1093/wber/lhg013. Source.
- Geoffrey R. Dunbar, Arthur Lewbel, and Krishna Pendakur. 2013. Children's Resources in Collective Households: Identification, Estimation, and an Application to Child Poverty in Malawi. American Economic Review 103 (1): 438–471. DOI: 10.1257/aer.103.1.438. Source.
- Seema Jayachandran and Alessandra Voena. 2026. Women's Power in the Household. Journal of Economic Literature 64 (2): 447–497. Author manuscript dated January 14, 2026. Source.
- Shelly Lundberg and Robert A. Pollak. 1993. Separate Spheres Bargaining and the Marriage Market. Journal of Political Economy 101 (6): 988–1010. DOI: 10.1086/261912. Source.
- Shelly J. Lundberg, Robert A. Pollak, and Terence J. Wales. 1997. Do Husbands and Wives Pool Their Resources? Evidence from the United Kingdom Child Benefit. Journal of Human Resources 32 (3): 463–480. Source.
- Marilyn Manser and Murray Brown. 1980. Marriage and Household Decision-Making: A Bargaining Analysis. International Economic Review 21 (1): 31–44. Source.
- Maurizio Mazzocco. 2007. Household Intertemporal Behaviour: A Collective Characterization and a Test of Commitment. Review of Economic Studies 74 (3): 857–895. DOI: 10.1111/j.1467-937X.2007.00447.x. Source.
- OECD. Time Use Database. Data and methodological notes, April 30, 2026 update. Accessed November 17, 2026. Source.
- Alessandra Voena. 2015. Yours, Mine, and Ours: Do Divorce Laws Affect the Intertemporal Behavior of Married Couples? American Economic Review 105 (8): 2295–2332. DOI: 10.1257/aer.20120234. Source.
- World Bank. Gender Statistics Metadata Glossary: Women Participating in Making Major Household Purchase Decisions, SG.DMK.PRCH.FN.ZS. Accessed November 17, 2026. Source: Demographic and Health Surveys. Source.
- World Bank. LSMS+ Program: Overview and Recommendations for Improving Individual-Disaggregated Data on Asset Ownership and Labor Outcomes. Institutional guidebook landing page. Accessed November 17, 2026. Source.