Using computers, expertise, and hunches to forecast US economy

''We came within a couple of billion dollars. Let's declare victory,'' Frank Cooper quips to his colleagues at Wharton Econometric Forecasting Associates. Mr. Cooper, a business activity expert, is referring to Wharton's February estimate of fourth-quarter 1983 corporate profits. Government data released in late March indicate the Wharton prediction was too high.

Every month Cooper and three other economists meet around a table in an office building here, gather their courage, and predict how the $1.6 trillion United States economy will perform during the next three years.

Using a combination of computer modeling, experience, intuition, and self-depreciating humor, they predict a staggering 1,400 different statistics for every three-month period. The items forecast range from the nation's total economic output to the demand for clothing and shoes. A computer model is a set of equations representing the economy.

The predictions cranked out at Wharton and other major forecasters like Data Resources Inc. (DRI) and Chase Econometrics have a serious impact on the lives of average citizens. For instance, administration and congressional tax plans are based in part on what the forecasting firms' models say the economy will be doing and thus how much tax revenue will be flowing into government coffers.

And many major corporations make sales projections, and thus hiring or firing plans, based to some degree on economic data the forecasting firms provide or help companies calculate in-house.

''We are trying to determine sales for the appliance industry'' using Wharton model data, says Frank Maly, supervisor of market analysis for Whirlpool Corporation, the big appliance manufacturer. ''You don't want to overshoot'' consumer demand or goods will back up in warehouses, he notes. But firms also want to avoid underestimating potential sales and thus not having enough products to meet demand.

The forecasting firms are far from infallible in predicting the economy. For instance, forecasts made in 1981 and early 1982 generally underestimated the recession and the slowdown in inflation that accompanied it. Year-end 1982 predictions for 1983 were better but tended to underestimate economic growth and overestimate inflation.

''There is almost no chance we will be correct'' in predicting the gross national product (the value of the nation's goods and services, or GNP) for 1986 , admits Robert Westcott, assistant director of Wharton's quarterly modeling service.

The reason is that there are many influences on the economy that can't be reduced to a mathematical equation. ''No model can predict how Tip O'Neill and Ronald Reagan are going to resolve the deficit problem,'' says James F. Smith, chief economist at Union Carbide Corporation.

Other unpredictable factors include changing oil prices and shifts in the Federal Reserve Board's monetary policy.

Critics say the forecasting firms' record is no better than that of than individual forecasters who operate without huge computer data banks and large forecasting staffs.

''I am not aware of anyone who is able to say even sophisticated econometricians at big forecasting firms do better than individual forecasters, '' says John A. Kaiser, associate director of the Center for International Business Cycle Research in New York.

The big forecasting firms say that is because other prognosticators tend to follow DRI, Chase, and Wharton.

''Independent forecasters are hesitant to be far away from the big three. The herd is always catching up to the big three,'' says Donald Straszheim, Wharton's vice-president for US services. Less-partial observers agree that most forecasters are wary of having forecasts that differ too sharply from the best-known firms.

Some experts also say the big three don't vary much in terms of accuracy. ''There is not too much difference'' among them, says Robert Eggert, editor of Blue Chip Economic Indicators, a forecasting newsletter.

And even in isolated cases in which a top performer can be identified, the difference between the best and worst performance ''is typically fairly small, too small to place much confidence in the assumption that it will persist in the future,'' Stephen K. McNees, vice-president of the Federal Reserve Bank of Boston, writes in the bank's magazine.

However, the forecasting firms and some clients say there are measurable differences among the economic modeling companies. Union Carbide executive Smith , who subscribes to DRI, Wharton, and Chase, says that Wharton's predictions ''are in general quite a bit better'' than those from the other two firms.

One advantage the firms offer is the consistent way their forecasts are produced.

While technical details of the forecasting process vary, the general approach is the same at all the modeling firms, combining computer simulations with a considerable amount of human judgment, says Lawrence Chimerine, chairman of Chase Econometrics.

At Wharton, the forecasting work in a given month starts with a meeting of a dozen or so economists who discuss the general economic outlook and hear reports from in-house experts on various sectors of the economy.

Then the smaller, four-man working group, all PhDs in economics, begin a detailed review of the economy using the latest available government data. At the March meeting, that data included revised fourth-quarter GNP numbers and the government's estimate of economic growth in the first quarter of 1984.

Some economic factors, like Federal Reserve action, have to be estimated without using the model. These are based on a variety of elements, including government data, contact with clients, and forecasters' experience and intuition.

The changes the group recommends in the previous month's forecast are written down by economist Cooper and later entered into a computer, which solves the equations to see what kind of forecast emerges. He makes additional changes until he is satisfied with the results.

These computer runs eventually are fine-tuned at three or more additional meetings until the forecast makes sense to the four-man team. Just before the forecast is sent to clients - both in written form and on disks that a personal computer can read - it is reviewed by Nobel economist Lawrence Klein, who founded Wharton Econometrics in 1963.