Forecasting M2 Growth: An Exploration in Real Time,” Economic Review, Federal Reserve Bank of Dallas (Second Quarter

For several reasons, it is important that Federal Reserve policymakers have a good un-derstanding of the relationship between money growth, interest rates, and spending. For ex-ample, the Federal Reserve is legally obligated to provide Congress with annual money growth projections. Federal Reserve officials must be prepared to explain the basis for these projec-tions and to interpret deviations of actual money growth from past forecasts. Moreover, several studies have suggested that there is information on future spending and future inflation in the difference between the current money supply and the long-run demand for money (Hallman, Porter, and Small 1991; Feldstein and Stock 1993; Koenig 1994; Duca 1994). Successful extraction of this information requires that the long-run demand for money be accurately estimated. Unfortunately, many models have system-atically overpredicted money growth during the 1990s—often by very large amounts. The forecasting record of DRI/McGraw-Hill (DRI) is typical. Figure 1 shows actual annualized M2 growth from 1990 through 1994, along with the M2 growth forecasts published by DRI each January. In every quarter over this five-year period, DRI overpredicted M2 growth. The average error was over 3 percentage points. The M2 model developed in the late 1980s by the staff of the Federal Reserve Board also overpredicted money growth.1 In 1993, citing increased uncertainty in the relationship be-tween M2 and spending, the Federal Open Market Committee announced that it would de-emphasize M2 in the policy-making process (Greenspan 1993). Recent efforts to explain the unexpectedly weak M2 growth of the early 1990s have fo