Academic journal article Federal Reserve Bank of St. Louis Review

FOMC Forecasts: Is All the Information in the Central Tendency?

Academic journal article Federal Reserve Bank of St. Louis Review

FOMC Forecasts: Is All the Information in the Central Tendency?

Article excerpt

Since 1980, there has been a substantial improvement in the performance of monetary policy among most of the industrialized countries. Bernanke et al. (1998) attribute this success to the adoption of inflation targeting, but they are somewhat vague in defining what it means to have an inflation target. Even countries that do not have numerical inflation targets are said to be inflation targeting if they focus on a long-term objective for price stability (low inflation) and publish inflation forecasts. By this broad definition, even the Federal Reserve may be considered to have an inflation target.

This article describes and documents 23 years of FOMC forecasts. Congress required the Fed to begin preparing and reporting forecasts in 1979. Section 108 of the Humphrey-Hawkins Act explicitly required the Fed to submit "written reports setting forth (1) a review and analysis of recent developments affecting economic trends in the nation; (2) the objectives and plans...with respect to the monetary and credit aggregates...; and (3) the relationship of the aforesaid objectives and plans to the shortterm goals set forth in the most recent Economic Report of the President." In order to satisfy the third item, the Federal Reserve Chairman began reporting a summary of Fed policymakers' forecasts to Congress in July 1979. (1) Since then, similar summaries of forecasts have been reported every February and July. Forecasts are made of annual, fourth-quarter-over-fourth-quarter growth rates for nominal gross domestic product (GDP), real GDP, and inflation. (2) Fed policymakers also forecast the average level of unemp loyment for the fourth quarter of the year. In February, the forecasts pertain to the current calendar year (referred to below as the 12-month-ahead forecasts). In July, forecasts are updated for the current calendar year (6-month-ahead forecasts) and preliminary projections are made for the next calendar year (18-month-ahead forecasts).

Separate forecasts are made by each of the FOMC members (and nonvoting Federal Reserve Bank presidents), but the individual forecasts are not published. (3) Rather, the Fed reports two summary statistics: the low and the high forecast among all the policymakers and the central tendency, which has been reported since February 1983 and is a smaller range that omits extreme forecasts.

This paper evaluates the information content of the two summary statistics, the full range and central tendency, and defines two FOMC forecasts based on those statistics: the midpoint of the full range (FR) and the midpoint of the central tendency range (CT). The paper compares forecast accuracy, tests for bias and efficiency, and uses encompassing tests to see whether either of these summary statistics encompasses the other. Admittedly, the two forecasts calculated from the FR and CT are quite close. In many of the tests, we cannot distinguish between the two. However, where statistically significant differences exist, they generally favor the use of the FR.


If the outlying forecasts for the FR and CT are symmetric, then the two will be the same. The FOMC has generated a large number of forecasts since 1979. There are two alternative ranges for each of four forecasted economic variables made at three different horizons (24 separate forecast series ranging from 18 to 23 years each). To compare the forecast with the outcome, we measure actual output, inflation, and unemployment using real-time data rather than the latest vintage data.

There is always an issue about how to define the "actual" data because the data are regularly revised as we get more information about the past. The analysis in this paper is based on the real-time data that are maintained at the Federal Reserve Bank of Philadelphia. They created vintages of data that were available in the middle of the following months: February, May, August, and November. (4) Except where explicitly noted, the actual values are taken from the first report for a calendar year that occurs at the end of January and is stored in the "February" files by the Philadelphia Fed. …

Search by... Author
Show... All Results Primary Sources Peer-reviewed


An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.