In February and July of each year since 1979, the Federal Reserve has provided a set of forecasts of the future state of the U.S. economy in the Monetary Policy Report to the Congress. While these forecasts have always been examined closely by the public, it is likely that the forecasts this February will be examined in even greater detail as markets struggle to determine when the cycle of quantitative easing will end and monetary policy (as well as the economy) will revert to normal.

Unfortunately, the report does not provide forecasts per se, but rather lists two ranges constructed using the forecasts made by each individual member of the Federal Open Market Committee (FOMC): the "full range," which includes the highest and lowest forecasted values, and a "trimmed range," which removes the three highest and lowest values from the full range. Here I look at two common approaches to using these ranges to construct a single forecast that could be considered representative of the majority of the FOMC members. One approach uses the midpoint of the trimmed range as representative of the FOMC forecast; the other uses the midpoint of the full range. Given these two forecasts, I choose between them based on their accuracy using the mean square error (MSE) as the relevant metric.

The table shows the MSEs associated with these two forms of forecasts for U.S. real gross domestic product (GDP) growth, the unemployment rate, and inflation.¹ For each variable I look at three distinct forecast horizons: current-year forecasts starting in February, current-year forecasts starting in July, and next-year forecasts starting in July. For simplicity, I refer to these as 12-month-, 6-month-, and 18-month-ahead forecasts, respectively.

In almost every case, for the real variables (such as GDP and the unemployment rate), the midpoint of the full range is more accurate than the midpoint of the trimmed range. In contrast, for inflation, in each instance the midpoint of the trimmed range is more accurate than the midpoint of the full range. While the magnitudes of improvement are not always large, the pattern is consistent enough across forecast horizons to suggest using the midpoint of the full range for the real variables and the midpoint of the trimmed range for inflation.

One possible explanation of this result relates to the underlying strategic motives behind these forecasts. Specifi­cally, FOMC members are directed to construct their forecasts conditional on "appropriate monetary policy"—an admittedly vague concept left to the discretion of the individual FOMC member. Insofar as FOMC members believe that their preferred policies should be implemented, and they want to persuade other FOMC members of their views, they may have an incentive to "forecast" dire consequences if that policy is not enacted. For example, an inflation hawk has an incentive to forecast very high inflation regardless of whether that outcome is the most likely, and an inflation dove has a similar set of incentives to forecast lower inflation. If the bulk of this strategic behavior is related to views of inflation, and less to the real variables, trimming the largest and smallest three forecasts may lead to more accurate inflation forecasts but less accurate forecasts of the real variables. As such, when choosing a forecast that is representative of the majority of the FOMC members, I suggest using the midpoint of the trimmed range for inflation and using the midpoint of the full range for both real GDP and the unemployment rate.

Note

¹ The forecasts are for gross national product from 1983 to 1992 and GDP thereafter. The inflation forecasts are for the GNP deflator from 1983 to 1988, consumer price index from 1989 to 1999, personal consumption expenditures (PCE) from 2000 to 2003, and core PCE from 2004 to present.

© 2010, Federal Reserve Bank of St. Louis. The views expressed are those of the author(s) and do not necessarily reflect official positions of the Federal Reserve Bank of St. Louis or the Federal Reserve System.