The Representative Agent Hypothesis: An Empirical Test

This paper empirically tests the validity of using only /mean /income as a representative variable in the aggregate consumption relation and of assuming time-invariance of the coefficients in this relation, as done in macromodels. We use a statistical distributional approach of aggregation to test these properties on the UK-Family Expenditure Survey [1974-1993]. The time-invariance assumption is rejected in most cases. A bootstrap test also suggests that in addition to mean income, the /dispersion /of income matters significantly for the commodity group /services /in several years and for /total/ /nondurable/ in some years, thus invalidating the /representative agent hypothesis/.