Fed says it no better at forecasts than others

WASHINGTON (Reuters) - There is not much difference in the forecasting track records of the U.S. Federal Reserve, the U.S. government and the private sector, according to a Fed study of historical forecasting errors published on Tuesday.

The hefty working paper, which examined forecasts going back 20 years, was prepared in conjunction with the debut release of a summary of Fed policy-makers' economic projections and was intended to give a sense of "normal" degrees of forecasting error.

Beginning with the projections released on Tuesday, Fed policy-makers will be providing quarterly forecasts to the public, as well as judgments on the degree of uncertainty around the estimates, relative to past experience.

"The Federal Reserve staff performs about the same as other forecasters," the paper found from a comparison with the U.S. Congressional Budget Office, the White House, and private sector projections by the Blue Chip consensus forecast and the Survey of Professional Forecasters between 1986 and 2006.

"Average differences in predictive performance across the forecasters in our sample are small, suggesting we can use information gathered from a range of sources to help gauge the average magnitude of past uncertainty," the paper noted.

On the other hand, no one does particularly well and the study's main observation was to emphasize what forecasters have always been happy to admit.

"Historical forecast errors are large in economic terms, indicating that - if past performance is a good guide to future accuracy - uncertainty about the economic outlook is considerable," according to the papers' authors, David Reifschneider and Peter Tulip.

The record led the authors to conclude that about 70 percent of forecasts had fallen within a fairly predictable band, defined as one root mean squared error of forecast.

"This result implies that historical prediction errors provide a reasonable basis for making explicit probability statements about the accuracy of future projections," they said.

"If current uncertainty is similar to that seen in the past, we can expect that actual economic outcomes will fall about 70 percent of the time, on average, within a range equal to the baseline forecast plus or minus the RSME (root square mean error)," they said.

In August the Congressional Budget Office forecast unemployment would rise to 4.8 percent next year.

The authors calculated that this meant "a probability of about 70 percent that the unemployment rate could be between 4.0 percent and 5.4 percent in 2008."

There was also wide uncertainty for consumer price index (CPI) inflation, which the CBO expected to slow from 3.6 percent this year to 2.2 percent, on average, over 2008 to 2010.

"Our analysis suggests a probability of about 30 percent that total CPI inflation could run below 1-1/4 percent or above 3-1/4 percent in 2010," the authors said.

(Reporting by Alister Bull, Editing by Andrea Ricci)