Risk-Sharing

"Covariate" shocks such as droughts or floods may affect everyone, but not everyone is affected equally. We devise ways to test the extent to which the effects of these shocks are shared. Download Paper (forthcoming in the American Economic Review) Abstract Efficient risk-sharing implies a simple factor structure for marginal utilities of expenditure (MUEs): Pareto weights divided by a common price. The standard approach infers MUEs from total expenditures, implicitly assuming homothetic preferences, unitary income elasticities, and identical price elasticities. Risk-sharing tests using total expenditures work for idiosyncratic shocks (budgets change, but not prices), but not ``covariate’’ shocks (prices change). We describe all preferences which permit one to infer MUEs from expenditures, and estimate nonhomothetic MUEs to test whether covariate shocks are shared efficiently in Uganda. This delivers sensible results; the standard approach suggests that droughts, floods, and pests are beneficial. ...

What stops villagers from sharing risk efficiently—limited commitment, moral hazard, or hidden income? We design lab experiments in rural Paraguay whose predictions differ across models, and find that no single friction explains all villages. Download Paper Abstract We conduct dictator-type games in rural Paraguay with different treatments that manipulate players’ information and choice sets. From individuals’ choices in the games, we draw inferences regarding impediments to efficient risk sharing in the larger village supergame. Outcomes from the experimental games suggest that players in most villages are reacting to the kinds of incentives that would be predicted from a private information model with hidden investments, while in others players act in a manner consistent with the predictions of a model with limited commitment. No single one of our models can explain outcomes in all villages, but outcomes in nearly every village are consistent with one or more of our models. ...

Which model of risk sharing fits village data—full insurance, limited commitment, or self-insurance? We devise tests that work without interest rate data and find that Tanzanian villages look like self-insurance. Download Paper Abstract This paper studies informal risk-sharing regimes in a unified framework by examining intertemporal consumption behavior of rural households in Tanzania. We exploit a theoretically-consistent link between interest rates and cross-sectional consumption moments to test alternative risk-sharing models without requiring data on interest rates or assuming a restriction to eliminate the need for such data, which are often unavailable in developing economies. We specify tests that allow distinguishing among models even with temporal dependence in income shocks. Our analysis shows that the consumption pattern in rural Tanzania is consistent with the self-insurance regime, and that risk aversion varies substantially across districts. Imposing a strict condition on interest rates, as often done in prior literature, misses their intertemporal heterogeneity and biases the estimation of risk aversion. ...

What motivates people in rural villages to share—altruism, reciprocity, or fear of sanctions? We use dictator game variants to measure each motive, and find that reciprocity best predicts real-world gift-giving. Download Paper Abstract What motivates people in rural villages to share? We first elicit a baseline level of sharing using a standard, anonymous dictator game. Then using variants of the dictator game that allow for either revealing the dictator’s identity or allowing the dictator to choose the recipient, we attribute variation in sharing to three different motives. The first of these, directed altruism, is related to preferences, while the remaining two are incentive-related (sanctions and reciprocity). We observe high average levels of sharing in our baseline treatment, while variation across individuals depends importantly on the incentive-related motives. Finally, variation in measured reciprocity within the experiment predicts observed ‘real-world’ gift-giving, while other motives measured in the experiment do not predict behavior outside the experiment. ...

How does a household decide who eats how much of what? We show that when nutrition affects future productivity, investing more in one member's nutrition can paradoxically reduce the quality of her food, and test these predictions with individual-level consumption data from Philippine farm households. Download Paper Abstract We construct a model of food demand which distinguishes between the nutrients supplied by a particular food bundle and the quality of that bundle, as measured by its cost. We show that when nutrition affects only utility then under rather general conditions it will be optimal for all members of the household to consume food bundles of identical quality. This is true even when household members have private information about their actions—in this case the quantity of food given may provide incentives, but quality remains common within the household. When nutrition affects household resources our finding that quality is constant is overturned—in this case when the household invests more in the nutrition of one member it will simultaneously reduce the quality of her food bundle. ...

Do households allocate food efficiently among their members? Using individual consumption data from the Philippines, we test whether nutrition's role in future productivity explains why consumption responds to earnings shocks within the household. Download Paper Abstract Using data on individual consumption from farm households in the Philippines, we construct a direct test of risk-sharing within the household. We contrast the efficient outcomes predicted by the unitary household model with the outcomes we might expect if food consumption delivers not only utils, but also nutrients affecting future productivity. The efficiency conditions which characterize the within-household allocation of food under the unitary model are violated, as consumption responds to earnings shocks. If productivity depends on nutrition, this explains some but not all of the response, as earnings “surprises” have some effect on the cost and composition of diet. ...

Cooperatives help members manage marketing risk within a season, but not production risk across years. We show how the multi-year cooperative relationship can also provide limited crop insurance. Download Paper Abstract Agricultural cooperatives have long played an important role in helping their members manage risk, though they do a much better job of helping members manage some sorts of risk than others. In particular, co-ops are good at helping members manage marketing risk, but seem not to be particularly good at helping their members manage production risk. This paper argues that by taking advantage of the multi-year nature of most members’ relationship with the cooperative, the cooperative can also provide a useful, though limited, form of insurance against crop shortfalls. ...

Poorer households grow faster on average but bear much more risk. Using Lorenz curve data from 53 countries over 50 years, we find no evidence that this greater risk is related to globalization. Download Paper Abstract “Globalization” implies change, and uncertainty over future change may affect household welfare. We use 50 years of Lorenz curve data from 53 mostly developing countries. Treating each country-quintile-year combination as an observation, we first account for variation in consumption expenditures, finding that global shocks matter less than country-level shocks in explaining consumption growth variation. While poorer households experience faster consumption growth than wealthier ones, they also face significantly more risk. However, we find no evidence linking this greater risk to globalization. ...

Can modern credit markets crowd out informal insurance? Using data from an Indian village in transition, we find that "traditional" households exchange more in-kind transfers, while "modern" households rely more on formal markets. Download Paper Abstract Contractual interlinkages in rural developing communities may provide informal insurance when information is incomplete, potentially shaping resource distribution. The introduction of modern credit markets might undermine these informal arrangements, reducing their effectiveness as insurance. Using data from a transitional Indian village comparing households operating under “modern” and “traditional” regimes, we find that traditional households receive more frequent in-kind food and clothing transfers while making more frequent in-kind crop output payments to others. ...

Standard poverty measures miss the welfare costs of risk. We construct a vulnerability measure that decomposes welfare losses into poverty and risk components, and find they play roughly equal roles in Bulgaria. Download Paper Abstract We construct a measure of vulnerability which permits us to decompose the welfare loss associated with living in a risky environment. This decomposition lets us separately measure the welfare loss due to poverty and the welfare loss due to risk from different sources. Applying this decomposition to a panel dataset from Bulgaria, we find that poverty and risk play roughly equal roles in reducing welfare. Aggregate shocks are more important than idiosyncratic sources of risk, but households headed by an employed, educated male are less vulnerable to aggregate shocks than are other households. ...

Nash Bargaining predicts Pareto-optimal allocations, but forward-looking household members should smooth away income shocks. We replace Pareto optimality with limited commitment and show how this resolves the puzzle of women borrowers voluntarily surrendering Grameen Bank loans to their husbands. Download Paper Abstract Much recent empirical work on intra-household allocation uses the axiomatic Nash Bargaining model to make predictions about how the distribution of consumption within the household will respond to individuals’ income shocks. However, one of the basic axioms underlying this approach is that allocations will be Pareto optimal, so forward-looking, risk adverse household members ought to be expected to smooth away any such response to income shocks—Pareto optimality seems to be too strong in a dynamic setting. In this paper we use explicitly dynamic framework and replace the axiom of Pareto optimality with a weaker notion of efficiency. We give a simple algorithm for computing allocations, and construct an extended example, meant to model the effects of Grameen Bank lending on intra-household allocation in Bangladesh. The model resolves a puzzle in the literature, namely, it predicts that women borrowers will often voluntarily surrender control (“pipeline”) their loans to their husbands. ...

Why don't villagers fully pool risk? We show that limited commitment—the option to walk away—constrains informal insurance, and test the model with ICRISAT village data from India. Download Paper Abstract Recent work on consumption allocations in village economies finds that idiosyncratic variation in consumption is systematically related to idiosyncratic variation in income, thus rejecting the hypothesis of full risk-pooling. We attempt to explain these observations by adding limited commitment as an impediment to risk-pooling. We provide a general dynamic model and completely characterise efficient informal insurance arrangements constrained by limited commitment, and test the model using data from three Indian villages. We find that the model can fully explain the dynamic response of consumption to income, but that it fails to explain the distribution of consumption across households. ...

When farmers produce under contract, there is a tension between risk and incentives. We provide a simple characterization of optimal risk and show that "less risk" need not mean "smaller variance." Download Paper Abstract When farmers produce under contract, there may be an important tension between risk and incentives. We provide a simple characterization of the optimal risk in any production system. We also show that “less risk” need not imply “smaller variance,” and that while at the margin the behavior of risk-averse farmers may appear to be nearly risk-neutral, it does not follow that one can generally treat such producers “as if” they were risk-neutral without being greatly led astray. ...

When households can both save individually and insure informally, how do the two interact? We show that an enhanced storage technology can either improve or diminish welfare, and that ex ante transfers are replaced by differential storage. Download Paper Abstract We examine a dynamic model of mutual insurance when households can also engage in self-insurance by storage, assuming there is no enforcement mechanism so that any insurance is informal and must be self-enforcing. We show that consumption allocations satisfy a modified Euler condition and that an enhanced storage technology can either improve or diminish welfare. We also show that ex ante transfers introduced into dynamic informal insurance models are only used in the first period, with the role of ex ante transfers being replaced by differential individual storage. ...

How does contracting change a farmer's supply response to price changes? When intermediaries absorb risk and farmers have private information, supply response is dampened relative to spot-market farming. Download Paper Abstract We consider four environments in which agricultural producers might operate, studying the role of price and production risk in shaping farmers’ supply-response decisions. In the first two environments, farmers market their own produce and are risk neutral and risk averse, respectively. In the third, farmers are risk averse, but risk neutral intermediation predicts that farmers should not face any production or price risk. In the final environment, risk neutral intermediation continues but the possibility of private information for farmers is admitted, which rationalizes exposure to production and price risk and suggests that a farmer’s response to a change in expected price will be less pronounced than in other environments. ...

Full risk sharing is rejected in Indian village data, but why? This paper tests whether private information about income explains the pattern of consumption allocations in ICRISAT villages. Download Paper Abstract Efficient risk sharing implies that consumption allocations depend only on aggregate resources, not on individual income. This prediction is rejected by data from ICRISAT villages in India. This paper considers the possibility that these departures from full risk sharing are due to private information about income. We derive the implications of private information for risk sharing and test these restrictions using data from three Indian villages. We find evidence that private information constrains risk sharing, and that an informationally constrained efficient allocation fits the data significantly better than either full risk sharing or autarky. ...