Nonrepresentative Representative Consumers

Single consumer models are often used to focus attention on economic e-ciency, leaving aside equity considerations. In general, these \representative consumer" models do not accurately portray the efiects of changes in policies, endowments or technology in the multi-consumer economies they are meant to represent; but even when they do, they are not necessarily adequate for evaluating e-ciency. The representative consumer can be Pareto inconsistent, preferring a situation B to A even though all the consumers in the represented economy prefer A to B. It is not clear from the literature how serious a defect this can be. The known examples of Pareto inconsistency are not robust. Small changes in the consumers' preferences remove the Pareto inconsistency. It has been an open question whether large robust Pareto inconsistencies are possible. This paper shows that they are. In one example, the actual consumers require 56% more income than the representative consumer requires in order to be compensated for the doubling of a price. But such large Pareto inconsistencies require that there is a Gifien good for the representative consumer. We argue that the inconsistencies of representative consumers in most macroeconomic applications are likely to be small and we give conditions ruling out inconsistencies entirely.