This paper shows the aggregate income (GDP) elasticity of money demand depends on the elasticity at the individual level and on population growth. As money regressions are usually formulated using US post-WW.II data, a low and stable aggregate income elasticity implies a negative and unstable elasticity at the individual level. Likewise positing a stable and non-unitary individual income elasticity implies wide variation in the conventional GDP elasticity over time. It has been possible to ignore these issues in empirical work due to collinearity between regressors and population, even for differenced data. Measuring money and income on a per household basis is a simple way to avoid these problems and to distinguish income from mere replication effects.