Predicting inflation: Does the quantity theory help?

Various inflation forecasting models are compared for the period 1979-2003 using a simulated out-of-sample forecasting framework. Our findings are (1) M2 has marginal predictive content for inflation; (2) it is necessary to allow for the possibility that money, prices, and output are cointegrated; and (3) cointegration vector parameter estimation error is important when making out-of-sample forecasts. Consistent with previous work, we find a structural break in the early 1990s, but the break was easily detected and would not have affected out-of-sample inflation forecasts. Two Monte Carlo experiments that lend credence to our findings are also reported on.(JEL E31, C32).