This article examines the effect of preference heterogeneity on nonparametr
ic and parametric tests of the rank of demand systems. Using samples of hou
seholds drawn from the U.K. Family Expenditure Survey, we find that prefere
nce heterogeneity increases the rank of demand systems. When the effects of
household characteristics on demand are removed by semiparametric techniqu
es, a rank-3 demand system appears to be an adequate empirical specificatio
n for all samples.