Forecasting Money Demand with Econometric Models

Among the many troublesome econometric relationships, the demand for money has proved especially recalcitrant, as evidenced by a long history of tinkering with basic specifications, always in response to some recent perceived forecast failure. The shortcomings of this approach and an alternative econometric methodology based on time-varying coefficient models are the subject of this paper. For comparison, we introduce a class of tentatively plausible models of money demand and evaluate their forecast performance. When these models are re-estimated allowing all coefficients to vary over time, forecasting performance improves dramatically