The fine and hazy art of economic forecasting based on early estimates
``Tough life,'' jokes Stephen K. McNees. What doesn't help for an economic forecaster is the frequent revisions of economic statistics. Revisions published last week on the United States' economic output for the first quarter made a flat spot in earlier estimates disappear.Skip to next paragraph
Subscribe Today to the Monitor
That sort of thing, notes Mr. McNees, an economist at the Federal Reserve Bank of Boston, causes a certain amount of ``personal anguish'' for those making a living by forecasting the economy.
An economist making a prediction must start from a numbers base for the recent status of the economy.
If those numbers change substantially in subsequent revisions, it can make their forecasts based on the earlier statistics look foolish.
Speaking of the first estimate of the gross national product by the Department of Commerce, he noted: ``You have the choice of throwing it in the wastebasket or using it.''
Actually, most professional forecasters have no choice but to use it. Their bosses - investors, businessmen, or government policymakers - need their best judgments now - not a year or two in the future - in making decisions.
McNees doesn't blame Washington statisticians for the revision problem.
Indeed, McNees praises them for the ``plethora'' of relatively accurate and timely data compared with that of most other nations.
One of McNees's specialties is analyzing the accuracy of economists' forecasts.
In his bank's publication, New England Economic Review, he looked at the record of economists in forecasting the turns in the three most recent business cycles.
This is especially relevant today with the current expansion already a long-lived 4 years old. Most economists aren't predicting a recession yet. But based on the average life spans of previous expansions, they are watching closely for signs of a slump.
Looking at a couple of ``consensus,'' or average, forecasts from large groups of economists and forecasts by some unidentified individual forecasters, McNees found that in both 1981 and 1983 these seers recognized the start of recessions only at the time they began. These were predictions in a statistical sense, since data to confirm them lag by a few months.
Predictions of the recession that began at the start of 1980 were made throughout 1979 and as early as '78.
But the warnings were so imprecise in dating the peak that they seemed to have been largely ignored, says McNees. There was an element of ``crying wolf'' in these forecasts when the expansion kept rolling along for so long.
``Whether undue pessimism but with a long lead time is more or less harmful than greater precision with no advance notice depends on the purpose of the forecast user,'' he notes.
Right now some economists are predicting a pickup in the economy as the year goes on. But probably a majority are talking of slow growth.
Last year Commerce Department officials dropped the ``flash'' quarterly estimate of gross national product based on only two months of a quarter's data. It was too inaccurate, they contended.
McNees says the ``flash'' was no more inaccurate than the first official estimate published soon after the end of a quarter.
In any case, economists will have to wait until next month to see who's right on the progress of the economy. And, they know, because of revisions in weeks to follow, that those who were right at first may well be last wrong.