Big Forecasters Failed to Predict Own Misfortune
It was a $250-million business built around a process that only the computer age could have spawned: mathematically manipulating hundreds of statistics to produce forecasts of the future course of the economy. Who could have known that the computer would almost destroy the business, too?
Produced by three major firms, computer-generated projections of interest rates, inflation, housing starts, and hundreds of other important indicators became a fixture of corporate board rooms and cabinet offices throughout the 1970s. Private and government clients paid as much as $150,000 a year for the services of the big forecasting firms, which were growing at rates of up to 40% a year.
No longer. The econometric forecasting business is in the midst of change so radical and swift that some of its most prominent practitioners already may be threatened financially.
The seers have been critically affected by developments for which their own mathematical models failed to account. One was the advent of the personal computer, which cut sharply into the firms’ crucial revenues from computer time-sharing. The second was the economy’s penchant for confounding any attempt to view it coherently. Of the three major forecasting firms, one, Data Resources Inc., has been been split up and strewn in components around its parent company, McGraw-Hill Inc.; its profits have slid by more than 40% in two years, to $6 million in 1984 from $8.5 million in 1983 and $10.9 million in 1982.
Firm Up for Sale
The second-largest, Chase Econometrics, is for sale by its owner, Chase Manhattan Corp. One potential buyer was McGraw-Hill, which relished the smaller company’s client contacts and expertise in some specialized areas. Talks between Chase and McGraw-Hill have broken off, however.
The third firm, Wharton Econometric Forecasting Associates, founded as a nonprofit company in 1963, has maintained its record of red ink despite its sale in 1980 to a profit-oriented company, Ziff-Davis Publishing Co. Ziff sold the firm in 1983 to a computer firm owned by the French government, which is now looking for international partners willing to put up equity capital to help the firm grow, particularly in Europe and the Far East.
At their peak in the early 1980s, these three firms split as much as $170 million in annual revenues, insiders say. Yet they now face pressure to slash prices while confronting competition from much smaller firms that offer statistical databases at a cut rate from what the big firms once charged.
“In the last three years, there’s been hellacious competitive pricing,” says one former Data Resources economist now at a private corporation.
In retrospect, 1982 appears to be something of a watershed year for the major firms. It was a recession year, and many clients closely examined their consulting expenditures. As it happens, the big firms had done so poorly in forecasting that recession that many clients found their services expendable.
Affected by Computer
Simultaneously, the personal computer vogue took off, establishing the powerful desk-top machine as a competitor to the huge mainframe computers operated by the forecasters.
That is important because for Data Resources and Chase, the actual economic projections sold to clients were attractions for clients but money-losing propositions for the forecasters; the companies were really selling computer time. At Data Resources, time-sharing fees could come to as much as two-thirds of a client’s bill, and accounted for most of Data Resources’ profit. At Chase, the share of revenues was more than 40%. At Wharton, the smallest of the big three and the one most oriented toward research, the share was about 10%.
The personal computer allows individual users to do at their desks many of the myriad computations once executed by the big firms’ full-sized mainframes. Income from time-sharing, now only necessary for the most complex computations, accordingly has been steadily shrinking.
“In the 1970s, you would have hundreds of companies spending hundreds of thousands of dollars per year to hook on to these mainframes, just to plot the consumer price index,” says Donald Straszheim, who recently resigned as chief economist at Wharton to join Merrill Lynch Economics. “They were not doing sophisticated analysis, just massaging the numbers. Now the PC software allows people to manipulate data till hell freezes over--at essentially no marginal cost except electricity and a person’s time.”
Slide in Profits
The upshot was a slide in profits. After topping out at nearly $11 million in 1982, Data Resources’ pretax earnings dropped to about $6 million last year, according to analysts’ estimates. Time-sharing accounted for less than half of Data Resources’ revenues in 1984.
Data Resources is hardly the same company this year. As part of a company-wide reorganization effective Jan. 1 at McGraw-Hill, Data Resources has been cut up, its units strewn about McGraw-Hill’s newly-formed operating divisions. “It’s now more of a brand name than a company,” says a former Data Resources economist now at another company.
McGraw-Hill executives stress that the reorganization was company-wide and not a result of Data Resources’ performance, and that they aim to maintain the firm’s identity. “One of my responsibilities is to preserve the integrity of the DRI analytical system,” says Joseph E. Kasputys, McGraw-Hill’s executive vice president and chief economist.
At the second-largest forecaster, Chase Econometrics, there has been a belt-tightening; branch offices have been closed and some employees laid off. The suspension of sale talks with McGraw-Hill has left Chase in an uneasy position as a unit of Chase Manhattan.
“Fundamentally the kind of business we’re in is not consistent with the strategic direction of the bank,” said Lawrence Chimerine, the chief economist at Chase Econometrics. On the other hand, he added, “The company would have gained very little from an association with McGraw-Hill.”
That leaves Wharton Econometric Forecasting Associates, founded in 1963 as a nonprofit research firm by Lawrence Klein, then a University of Pennsylvania economist and later a Nobel economics laureate.
Impact Is Less
Of the big three, Wharton was always least dependent upon time-sharing. While it grew more modestly than Chase or Data Resources, it has not suffered the same decline. Corporate and government economists say the firm’s reputation for forecasting accuracy as well as reasonable prices is rising.
Still, Wharton has remained unprofitable. Under the ownership of CISI Group, a computer-services firm largely owned by the French government, Wharton has stepped up its marketing worldwide, but still does not expect to have a profitable year in 1985.
All three have come far since the great days of the 1970s, when they rode swelling reputations as economic seers. In the vanguard was Data Resources, founded by the late Harvard professor and Johnson Administration economist Otto Eckstein, who in 1969 saw the potential in collecting the disorderly Babel of government financial data and ordering it into a semblance of coherence for sale.
The company projected the air of a Gucci boutique on Fifth Avenue, where the customers were expected to appreciate the cachet of paying top-dollar.
“Every time you’d turn around they’d be jacking up their prices again,” says John Qualls, chief economist for St. Louis-based Monsanto Corp. The company was also reluctant to customize its products, that is, sell clients only those features they needed.
When Monsanto approached the company about buying its data but not its computer time, says Qualls, “they as much as told us they weren’t interested.” Monsanto, a Data Resources client since 1974, dropped the firm in 1983 to work more closely with Wharton; its time-sharing bill dropped from about $40,000 a year in 1982 to less than $10,000 the following year, Qualls says.
Reconsidered Pricing
Now Data Resources has reconsidered its pricing philosophy. In April, Kasputys says, the company cut time-sharing rates by 12%. But at the same time it “unbundled” its services, meaning that a client will henceforth pay separately for the data, consulting, and computer time that was once sold in a package.
Dissatisfaction with Data Resources’ services and those of the other firms grew as clients came to the conclusion that the econometric models have great shortcomings, not the least of which is a certain sameness. This perception is generally accurate, says Stephen K. McNees, an economist at the Federal Reserve Bank of Boston who has tracked the records of 12 prominent forecasters, including Data Resources, Wharton, and Chase, over several years.
“No one seems to be able to document a systematically better record than the group as a whole,” McNees says.
The entire discipline’s record was marred by two particularly embarrassing blunders--its inability to predict the oil-inspired recession of 1974-75 and the quite different 1981-82 recession.
McNees calls the first episode the “darkest hour” of econometrics; in that period, the forecasters made their largest errors in the 15 years of his study in predicting such fundamental variables as gross national product and the rate of inflation. The reason is that none anticipated the staggering impact of the run-up in oil prices. “No one came close to getting the magnitude of the effect,” he says, “so the forecasters way overestimated economic growth and underestimated inflation.”
Overestimated Impact
In the second case, he says, forecasters grossly overestimated the impact of the Reagan Administration’s fiscal policy--that is, the 1981 tax cuts, which were phased in so gradually their effect was all but dissipated. So none anticipated that policy’s failure to bring the economy promptly out of recession.
Although the size of the error was greater in 1974-75 than in 1981-82, McNees argues that the second mistake was the more frightful.
“In 1973, we understood the oil shock was an unprecedented event, so we were all humble about what we knew of its impact,” he says. “But the bad news is everybody thought they understood (the impact of fiscal policy). So I think of the 1982 error as particularly humiliating and discouraging, because everyone was talking as if they knew and no one really did.”
These errors, critics say, exposed the fundamental flaws of the econometric models, the strings of equations that are the core of the econometrician’s forecast.
Although a model might comprise equations involving a thousand variables (Data Resources has used one with 1,200 elements), it could never accommodate the one critical variable: “Human activity is the key to the economy, and the models don’t capture that,” says Lawrence A. Kudlow, a former chief economist for the Bear Stearns & Co. brokerage and for the Office of Management and Budget. “If the government changes its policy, human actions are going to change overnight.”
The models are most effective at extrapolating from past trends, a feature that contributed to their celebrated failures in 1974-75 and 1981-82, when the historical cycles were in turmoil.
Different Situation
“The models are all basically formulated on the ‘50s and ‘60s steady-growth years,” says A. Gary Shilling, a private economist. “It’s very different to deal with the volatility of the ‘70s and ‘80s.”
The big firms now are redesigning their products to aim at a new tier of customers--not only the Fortune 500 companies, but smaller firms and divisions of larger firms. McGraw-Hill executives hope the corporation’s reorganization will help Data Resources by placing those products that serve specific industries in the same divisions. Services and products aimed at the computer and information industry, for example, will be marketed out of a division that also encompasses McGraw-Hill’s Future Computing Co., a computer-business market-research firm, and the magazines Byte and Personal Computing.
“DRI was impacted fairly broadly (by the restructuring) because it was in more markets than most,” Kasputys says. “It was serving 16 markets, so a lot happened to DRI.”
Still, the PC revolution caught Data Resources and, to a lesser extent, Chase in a vise. McGraw-Hill had paid $103 million for Data Resources in 1979, a price that some insiders say provoked the big publishing conglomerate to “always question whether it got its money’s worth,” in the words of one departed economist.
Although Data Resources has produced only about $40 million in pretax revenues for McGraw-Hill since the purchase, some analysts argue that the acquisition has paid off in other ways that may be enhanced in the reorganization.
A Subtle Side
“There’s a subtle side to DRI,” says John S. Reidy, a media analyst for Drexel Burnham Lambert Inc. “You can’t just add up the earnings. DRI’s certainly been useful in corporate research-and-development.” Reidy argues that some analytical products marketed by certain McGraw-Hill divisions “were only possible because of synergies with DRI.”
Now all three firms say they are busily developing software for customers to use on PCs, although customers say the products are of widely varying quality and for the most complicated forecasts still less useful than mainframe-generated packages. Wharton’s software is considered by many customers to be the most refined, a verdict the company says is the result of its small stake in the time-sharing business.
For Data Resources and Chase, “time-sharing is such a large revenue stream that they don’t want to accelerate its demise,” says Wharton President Bruce R. Lippke. “It’s in their best interests to claim they are pushing PC software and not really do it.”
Data Resources and Chase say they are working just as hard to design new “delivery vehicles,” as forecasting and database software is called in their argot.
Chase, contends Chimerine, marketed more than six months ago the first software allowing PC users to shift data from the big main-frames to their own machines. “We look at the growth in microcomputers to be a market opportunity,” he says.
Moments of Lethargy
“Any organization has its moments of lethargy,” acknowledges George F. Brown Jr., executive vice president of McGraw-Hill’s Financial, Economic & Information Co., the umbrella division that includes Data Resources. “Like everybody, we made a few wrong decisions.” Among them, he says, is failing to anticipate “how IBM would dominate the PC line.”
But he adds, “I see us remaining at the forefront of delivery technology. There’s an incredibly rapidly growing number of clients now using DRI in the PC environment. In six months, you’ll find our entire capital base oriented toward the PC; my guess is it will allow our business to grow much more rapidly than time-sharing.”
