Should consumer expenditures be the scale variable in empirical money demand equations?

Traditionally, real GNP or permanent income or wealth have been the scale variable of choice in empirical money demand equations. Recently, Mankiw and Summers (1986) argue that consumer expenditures are an ideal proxy for permanent income in money demand, and they provide evidence that total consumption expenditures or consumption expenditures on non-durables and services are better scale variables in money demand than current GNP. This result is odd because consumer expenditures reflect only the desires of the households, and a significant proportion of money balances is held by firms. This paper shows the difficulties in using consumer expenditures as a proxy for permanent income, shows that, properly estimated and compared, consumer expenditures are no better as a scale variable than real GNP and provides evidence that permanent income is a better scale variable than either consumer expenditures or GNP.