We provide a revealed preference characterisation in the tradition of Afriat (1967), Diewert
(1973) and Varian (1982) of the necessary and su¢ cient conditions for the existence of a norma-
tive representative consumer. These results are very simple and complement those of Gorman
(1953, 1961); they can also be applied to data very readily and without the need for auxilliary
parametric or statistical assumptions. We investigate the application of these methods to a
balanced microdata panel survey. The main empirical result is that there does not exist a
normative representative consumer for our data. Given this, we use our revealed preference
characterisation for identifying the unobservable heterogeneity across households (in terms of
their marginal propensity to consume) that makes the data reject a normative representative
consumer.
Acknowledgments: We are very grateful to Richard Blundell, Don Brown, Martin Browning,
Erwin Diewert, Peter Neary, Margaret Stevens, Kirshna Pendakur and seminar participants at
Oxford, CIREQ Montreal and Paris Dauphine for helpful comments and suggestions.