Time preference may vary across goods

When we model intertemporal choices, we usually go with a single consumption good, which means (up to some aggregation assumptions) that all goods have the same intertemporal elasticity of substitution and economic agents have the same risk aversion for each. Or that somehow any dispersion would not matter through aggregation. Is there any evidence of such dispersion?

According to Diego Ubfal, there is, and it is considerable. Using a quite large experimental survey of 2400 people in rural Uganda, he finds that the monthly discount rate goes from 110% (meat) to 66% (salt). And unlike this literature, this paper does not assume linear utility and goes through the trouble of estimating the curvature of the utility function as well (without reporting results, though). I wonder, however, how seasonality may play in this. For households that are that poor and who have to live from day to day, it is perfectly understandable that the rate may vary widely whether the good is in season or not, whether one is in a period of the year with relative slack on the budget or resource constraint, and add to this whether the size of the remuneration from the experiment may distort things. Still, this is an interesting and detailed study.