On Expenditure Indices in Revealed Preference Tests

Abstract

Recently in this Journal, Chalfant and Alston (1988) utilized revealed preference techniques to demonstrate that changes away from consumption of red meat may not be due to changes in tastes but rather to changes in relative prices. The authors stressed that previous studies that find evidence of changes in taste are based on estimation of particular functional forms. The revealed preference approach is therefore appropriate since consumers with the same tastes will yield observations consistent with revealed preference regardless of the particular utility function. The research is carefully done and the conclusions seem appropriately drawn. The authors found some violations of revealed preference, but they concluded that the violations are negligible on the basis of computation of an expenditure index. Without disagreeing with their conclusion, I believe that the expenditure index is an inappropriate metric of the seriousness of violations of revealed preference. Instead, I present a metric that captures differences in tastes not captured by the expenditure index. The use of this superior index is particularly important when consumers face very similar prices and have similar incomes. Afriat (1967, 1973) and Varian (1982) prove that for observations to be consistent with the hypothesis of no change in taste, satisfaction of the generalized axiom of revealed preference (GARP) is a necessary and sufficient condition. Knoblauch (1990) proves that when only two goods are in the consumption bundle, consistency with the weak axiom is necessary and sufficient for the existence of a common