The real-time predictive content of money for output

Data on monetary aggregates are subject to periodic redefinitions, presumab ly in part to improve their link to measures of output. Money data are also revised on a regular basis. Taking these data imperfections into account, we reassess the evidence on the marginal predictive content of MI and M2 fo r real and nominal output. In particular, first by using the latest version of the data that is available, and then using sequences of historical time series that would have been available to forecasters in real-time, we are able to provide a comprehensive assessment of whether money is useful for p redicting output. We conclude that the generally significant marginal predi ctive content of M1 or M2 for output that is found using a recently revised data set is not duplicated in a realtime setting, although M2 is shown to remain useful when 1-year ahead forecasts are constructed using fitted vect or autoregressive models. (C) 2001 Elsevier Science B.V. All rights reserve d.