The Numbers Game

Financial forecasters work in a fishbowl, and they seem to want it that way. Open a newspaper, and there's noted economist Claire Voyant prophesying where interest rates, growth, or inflation is headed. That's free publicity for Voyant's employer, whether a money-management firm, commercial bank, or her own eponymous consultancy, says Owen Lamont, an assistant professor at the GSB. "The company approves of and encourages its economists to get quoted."

Unfortunately, no one in the fishbowl has a pipeline to the truth: It's well known that the average, "consensus forecast" is more accurate than any individual's opinion. But if the consensus is common knowledge, and accuracy is a forecaster's goal, why stand out from the crowd? Why don't the forecasting fish always swim in a school?

Lamont, who himself spent two years working for a forecasting firm, has an unsentimental answer to this old economic question. Forecasters, he believes, aren't aiming for accuracy. "What they get paid for is their reputation--whether they help the firm's image, or whether they help the firm get clients."

That's too bad for clients, who, with a faulty forecast, might pay too much for a loan, hire too many workers, or build a new plant when a recession threatens. One consequence of the prognosticators' self-interest may especially surprise trusting customers: The older and more established a forecaster, Lamont finds, the less accurate the forecast.

A specialist in corporate finance and macroeconomics, Lamont tested his belief by analyzing two decades of Business Week's year-end forecasts, in which dozens of economists and econometric models predict next year's unemployment rate, gross national product, and consumer price index. The 1971-92 sample included 118 economists and 15 models.

A forecaster's age shouldn't matter, Lamont argues, if accuracy alone were the goal. But over time, he says, the reputation-seeking economist should deviate either less from the pack (what he calls "herding") or more ("scattering"). The impetuous youth, for instance, might make some attention-getting forecasts, garner a name after a few lucky calls, and then turn conservative. Alternately, he or she may wait until safely established to issue unusual projections.

In the Business Week data, Lamont found, forecasters "scatter" as they age. In the case of GNP growth--a number that's often just 2 or 3 percent--they stray from the consensus by an extra 22 basis points, or 0.22 percent, for every ten years' additional experience. Are their unorthodox hunches less accurate, too? "The answer," Lamont writes in a National Bureau of Economic Research working paper, "is a clear yes."

Yet the underlying cause of the scattering probably wasn't age but rather a need for good press. While forecasters who stuck with one employer changed little over their career, scattering surged among the 11 people who founded eponymous consulting firms during the 20-year period. "When you own your own firm, you just want publicity more," explains Lamont. "By making radical forecasts, you get more publicity and seem more like a dynamic, exciting type of guy."

Reputations, he concludes, last longer than track records. Bad predictions don't hurt the way good ones help.

While the plain-vanilla consensus forecast offers reliability, "part of what some people pay for when they buy a forecast is not just the number, but the story that goes with the number," Lamont notes. "You could be a bad forecaster but still be a very popular guy."--A.C.

More Investigations:

Forecast--Cloudy: Astronomers Priscilla Frisch and Daniel Welty map our interstellar neighborhood.

Also in Investigations:

• Family income is the number one factor in determining who attends college, according to a National Opinion Research Center study.

• Encryption technology would protect individual privacy and prevent economic espionage, according to a National Research Council study led by Law School Professor Kenneth Dam.

• Graeme Bell and Nancy Cox of the BSD have narrowed the search to a region on human chromosome 2 for a gene that plays a major role in non-insulin-dependent diabetes melitus.