In this paper we analyze US money demand stability and the indicator properties of derived money overhang measures of various monetary aggregates for predicting inflation over a sample from 1987Q1 to 2008Q2. In contrast to a large part of the literature, we find evidence of a stable money demand function for M2 in the framework of the cointegrated VAR (CVAR) model without resorting to ’exotic’ determinants or redefinitions of M2. Previous evidence suggesting instability of the M2 money demand function may have been related to two kinds of misspecification: First, with regard to the specification of the deterministic components, and secondly, with regard to the imposition of theoretically plausible but empirically rejected restrictions imposed on the model from the outset. Using formal stability tests, we find that stability of the long-run coefficients cannot be rejected, while stability of the short-run parameters is doubtable. Inference is not only based on asymptotics, but also on small-scale parametric) bootstraps. We find some evidence that money overhang is a useful information variable for predicting changes in the inflation rate. First, our estimates obtained from the CVAR model suggest that money overhang Granger-causes inflation. Secondly, recursive out-of-sample forecasts which we conducted over a hold-back period show that taking account of derived money overhang measures significantly improves forecasts of the change in inflation over long horizons (about 3 years). Finally, we provide some evidence that the importance of money overhang for predicting (changes in) inflation may have increased over time.