Intra-Household Efficiency: An Experimental Study from Ethiopia

Abstract

This paper uses different versions of a voluntary contribution mechanism in experimental games to test for household efficiency. Efficiency is decisively rejected in all treatments casting doubt on ‘unitary’ and ‘collective’ household models – significant amounts of potential surplus are not realised. Efficiency is rejected whether initial endowments are equally split between spouses or only one spouse takes all. Similarly, efficiency is also rejected with all variations in allocations, whether the household common pool is split equally between the spouses or allocated by the husband or the wife. Information on initial endowments of spouses improves efficiency in some treatments while not having an effect in others suggesting that the role of information is context dependent. Actual and expected contribution rates of spouses are systematically different; husbands’ expectations of their wives’ contributions are higher than actual contributions and wives’ expectations of their husbands’ contributions are lower than actual contributions. These systematic errors in expected and actual behaviour indicate that the attainment of equilibrium in a game theoretic framework is unlikely. Statistical tests indicate that instead of efficiency other considerations like norms are likely important. Overall, most of the empirical results cast doubt on cooperative models and provide some support for behaviour guided either by fairness or other norms. The data are collected from 1,200 married couples in three sites in Ethiopia.