Art of Forecasting

October 5, 1959

Is “scientific” forecasting possible? Can we, by studying enough statistics, and inventing more ingenious statistical methods, tell just when business conditions are going to get better or worse, and by just how much? Is there, perhaps, some single indicator, if we could only find what it is, that would point the future course of the whole economy?

There are many professional forecasters who would like to give that impression—by the solemnity and confidence of their tone, if not by explicit assurance or the size of their fees. But a recent eighteen-page pamphlet by Henry M. Platt, Economic Indicators: Their Use in Business Forecasting, published by the Amos Tuck School of Business Administration at Dartmouth College, makes as lucid and intelligent an effort as one is likely to find to answer that question, and does not end with assurance.

Dr. Platt takes off from the work of the National Bureau of Economic Research, particularly in the revised form adopted by Geoffrey H. Moore in 1950. Dr. Moore selected 21 business indicators, and divided them into eight indicators that ordinarily lead, eight that tend more or less to coincide with, and five that generally lag behind cyclical turning points.

THE 21 INDICATORS

The eight “leaders” are: Number of new incorporations, new orders for durable goods, industrial stock prices, wholesale prices of basic commodities, commercial and industrial construction contracts, residential construction contracts, the average work week in manufacturing, and business failures.

The “coinciding” indicators are: The Federal Reserve Board index of production, nonagricultural employment, unemployment, bank debits outside of New York City, freight-car loadings, wholesale prices (except of farm and food products), corporate profits, and the gross national product.

The “lagging” indicators are: Personal income, retail sales, consumer installment debt, bank rates on business loans, and inventories.

Now though these indicators were chosen empirically, Dr. Platt tries to show that there are good reasons for each of them to act or turn as it generally does in relation to general business activity. He begins by briefly describing what he calls the “self-generating cycle,” according to which a revival, once begun, tends to be cumulative, to reach a peak, then to break, and to turn into a cumulative downswing or recession, which in turn reaches a bottom and generates forces that bring about revival again.

Dr. Platt tries to show why, for example, the eight “leading” indicators tend to lead. The first six all represent or reflect investment commitments. Investment commitments are present actions based largely on plans or expectations for the future—that is to say, on “economic forecasts.” Similarly, he tries to show why the coinciding indicators coincide and why the lagging indicators lag.

THEY CAN’T PREDICT

But when he is all through, he is obliged to conclude that “Economic indicators cannot predict when a business turning point will occur.” Further they “can give no idea of how intense a revival or recession will be, or how long it will last, until it is well under way”—i.e., until it is too late.

Dr. Platt also gives some of the reasons for this disappointing conclusion. “Genuine reversals in some series usually cannot be confidently distinguished from random fluctuations until some time has elapsed.” “It takes time to gather and transmit the data that show an indicator’s curve has turned.” Though the “leading” indicators reflect business expectations, those expectations may themselves turn out to be wrong.

One can think of additional reasons. It is not merely that all statistics reflect merely past events; or that they can never adequately cover the infinite number of economic developments. But business is often tremendously affected by developments from outside the purely “business” world, from government interferences to revolutions and wars.

This does not mean that it is useless to study, refine, or follow business statistics. It does mean that business forecasting is not a science but an art, depending heavily on personal judgment, hunches—and luck.