The money demand function has traditionally been estimated with income and interest rates, typically employing quite lengthy time series. Controversy, however, surrounds the importance of heterogeneous agents in monetary economics and throughout macroeconomics more generally. In particular, if proportions of agents with different traits (and hence, different money demands) are changing over time, ignoring those changes may bias estimated income and interest elasticities. This concern, as well as that of appropriate functional form, is explored here. Controlling for consumer heterogeneity has surprising effects on the estimated elasticities of traditional variables.