Choice, Rationality and Welfare Measurement

We present a method for evaluating the welfare of a decision maker, based on observed choice data. Our method can be used whether or not the observed choices are rational. We parallel the standard revealed preference theory, which does require rationality, by modeling choices "as if" they were the result of a speci…c decision making process. However, in place of the usual preference relation whose maximization induces the observations, we explain choice as arising from a compromise among a set of simultaneously-held, con‡icting preference relations. We use these preference relations as the basis to measure the decision maker’s welfare. In general our method does not yield a unique set of explanatory preferences. Thus we characterize all the explanatory combinations of preferences any one of which could generate the data. We use this set of explanations to compute bounds on welfare changes. We show that some standard results of rational choice theory can be extended to irrational decision makers. The theory can be used to explore a number of context-dependent choice patterns found in psychological experiments.