How general are time preferences? Eliciting good-specific discount rates

This paper tests the commonly-used assumption that people apply a single discount rate to the utility from different sources of consumption. Using survey data from Uganda with both hypothetical and incentivized choices over different goods, we elicit time preferences from about 2,400 subjects. We reject the null of equal discount rates across goods; the average person in our sample is more impatient about sugar, meat and starchy plantains than about money and a list of other goods. We review the assumptions to recover discount rates from experimental choices for the case of goodspecific discounting. Consistently with the theoretical framework, we find convergence in discount rates across goods for two groups expected to engage in or think about arbitraging the rewards: traders and individuals with large quantities of the good at home. As an application, we evaluate empirically the conditions under which goodspecific discounting could predict a low-asset poverty trap.