Testing for Rationality with Consumption Data : Demographics and Heterogeneity

In this paper, we introduce a new measure of how close a set of choices are to satisfying the observable implications of rational choice, and apply it to a large balanced panel of household level consumption data. We use this method to answer three related questions: (i) 'How close are individual consumption choices to satisfying the model of utility maximization?' (ii) 'Are there differences in rationality between different demographic groups?' (iii) 'Can choices be aggregated across individuals under the assumption of homogeneous preferences?' Crucially, in answering these questions, we take into account the power of budget sets faced by each household to expose failures of rationality. To summarize our results we find that: (i) while observed violations of rationality are small in absolute terms, our households are only moderately more rational than the benchmark of random choice; (ii) there are significant differences in the rationality of different groups, with multi-head households more rational than single head households, and the youngest households more rational than middle age households; (iii) the assumption of homogenous preferences is strongly rejected: choice data that is aggregated across households exhibits high levels of irrationality