This paper reevaluates the inflation forecast performance of M2-based P* models relative to other competing models over the period of 1970–96. Included in the comparative study are newly developed monetary aggregates, including M2+, MZM, and M2*, and direct treatments of velocity changes associated with recent developments in M2. Out-of-sample rolling-horizon forecast exercises suggest that the predictive accuracy of alternative P* model specifications relative to traditional inflation models is not robust to different subsamples. The switching forecast performance between money-based and output-based models across periods highlights the extent of structural instability in the inflation generating process. (JEL E3, E4, C2)