Poor households in developing countries are subject to considerable risk. Although the Permanent Income Hypothesis (PIH) predicts that households will fully adjust consumption to changes in permanent income and smooth consumption over changes in transitory income, evidence suggests that both of these predictions may fail (Paxson, 1992; Fafchamps, Udry, and Czukas, 1998). The goal of this project is to identify the specific factors that might prevent households from smoothing consumption over changes in transitory income and to evaluate the effectiveness of two interventions designed to facilitate consumption smoothing. This project will focus specifically upon income shocks caused by household illness. As credit is rare in developing countries, one obvious reason that households may be unable to smooth changes in transitory income is that they lack a secure place to save. The first goal of this project will therefore be to provide a formal savings account to a randomly selected subset of daily income earners. Even with formal savings vehicles, however, households may be unable to save transitory income if their preferences for consumption are time-inconsistent. The second goal of this project will be to provide one of two types of commitment savings accounts to two randomly selected treatment groups. The first group will be offered target accounts that will feature a strictly enforced selfchosen weekly savings goal while the second group will be offered health accounts that will encourage withdrawals for health expenditures but discourage withdrawals for all other purposes by introducing relatively large withdrawal penalties for non-health purposes and eliminating such penalties for health withdrawals. This project will collect information on a variety of outcomes, including the difference in savings between the treatment and comparison groups, the difference in the labor supplied between the groups, and the extent to which the accounts serve to substitute for other, costlier types of income smoothing (such as choosing to invest in lower-yielding but less risky assets). The most important outcome, however, will be the differential responsiveness of consumption to transitory income between the treatment and comparison groups. As this experiment is randomized at the individual level, the effect of the standard savings accounts can be estimated directly from the differential responsiveness between the standard savings and comparison groups and the effect of the commitment savings accounts can be estimated from the differential responsiveness between the commitment savings and standard savings groups. Intellectual Merit This project will be of academic interest for several reasons. First, it will provide experimental evidence on the specific mechanisms by which the Permanent Income Hypothesis tends to fail. Second, it will add to the growing literature on the empirical relevance of commitment savings accounts. Third, it will test the responsiveness of consumption to transitory changes in income due to health shocks, in contrast to most studies which focus upon shocks to agricultural income caused by changes in rainfall. Broader Impact The results obtained from this research could potentially also have valuable policy implications. First, providing such accounts could represent a novel approach in designing savings accounts in developing countries and in developing risk-coping mechanisms for poor households. Second, to the extent that the accounts allow households to medically treat illness and so increase their labor income, the accounts could represent a specific mechanism for increasing income by improving health.
