Should cash transfer programmes restrict consumer choice? For example, should food assistance delivered in cash be restricted to food and exclude temptation goods? Theoretically, restrictions induce (1) a substitution effect away from restricted goods and (2) a negative wealth effect if transfers are extra-marginal and the resale of goods is costly. The welfare impact on transfer recipients is negative. We test these predictions by exploiting a natural experiment in a refugee settlement in Kenya, where some refugees receive monthly cash transfers restricted to food while others receive unrestricted transfers. In line with theory, we find that restricted transfers increase participation in a shadow resale market and negatively affect non-food expenditure, temptation-goods spending, and subjective well-being. Consistent with theory, restrictions have no significant effect on food consumption. Our results show that policy-makers should avoid restrictions to maximise positive impacts on transfer beneficiaries, especially when extreme poverty implies that transfers are extra-marginal.