In developing countries, unexpected income shocks are common but informal insurance is typically incomplete. An important question is therefore whether risk-sharing within the household is effective. This paper presents results from a field experiment with 142 married couples in Kenya in which individuals were given random income shocks. Even though the shocks were small relative to lifetime income, men increase private consumption when they receive the shock but not when their wives do, a rejection of efficiency. Such behavior is not specific to the experiment—both spouses spend more on themselves when their labor income is higher.