Intra-household efficiency: An experimental study from Ethiopia

An experimental design using treatments of a voluntary contribution mechanism is used to test household efficiency. Efficiency is decisively rejected in all treatments contrary to the assumption of most household models. Information on initial endowments of spouses improves efficiency only in some treatments suggesting that the impact of information is context dependent. Actual and expected contribution rates of spouses are systematically different; husbands' (wives') expectations of their wives' (husbands') contributions are higher (lower) than actual contributions. These errors imply that equilibrium in a game theoretic framework is unlikely. Statistical tests indicate other considerations than efficiency are likely important.