How do movements in the distribution of income affect the macroeconomy? Krusell and Smith (1998) analyzed this question in a neoclassical growth model, and their results show that the representative-agent assumption provides a good approximation for aggregate behaviors of heterogeneous agents. This paper extends their analysis to a cash-in-advance model with heterogeneous money demand. It is shown that movements in the distribution of monetary income can have significant impact on the macroeconomy. For example, the dynamic responses of aggregate output to monetary shocks behave very differently from those of a representative agent; the welfare costs of moderate inflation are much higher than previously thought, up to 20% of consumption when the inequality of cash distribution is sufficiently large. This is in sharp contrast to the findings of Cooley and Hansen (1989) and Lucas (2000) based on representative-agent models.