Can Statistics Predict?

October 23, 1961

In 1957 the Council of Economic Advisers suggested that the Census Bureau, with the cooperation of the National Bureau of Economic Research, a private statistical organization of high standing, “develop methods of appraising current business fluctuations in a monthly report that would take advantage of new findings about . . . economic processes over time, the availability of a great many economic time series, and large-scale electronic computers.”

As a result, Julius Shiskin took a leave of absence from his job as chief economic statistician at the Bureau of the Census to make a study for the NBER. The results now appear in a report entitled: “Signals of Recession and Recovery: An Experiment With Monthly Reporting.”

To estimate what Shiskin and his associates have accomplished, let us look first at the problem they were given to solve. As an aid in predicting the future of business, or even in ascertaining its present state, statisticians, private and governmental, have developed over the years an increasing number of statistical “series.” Shiskin estimates that in Economic Indicators, the Survey of Current Business, the Federal Reserve Bulletin, and magazines and newspapers, there are now well over 2,500 aggregate and component monthly or quarterly economic series. They cover an immense range of economic activity.

ROLE OF COMPUTERS

Obviously “this enrichment of the data available,” as Shiskin points out, “was not an unmitigated boon for the analyst. . . . It was beyond the resources of any individual . . . to carry out the calculations, such as seasonal adjustments, needed to make the series useful for current business cycle analysis.” Fortunately, “a major breakthrough came in the early 1950s, with the advent of the large-scale, general-purpose electronic computer.” A computer program can “adjust and summarize” statistics quickly, and give “early-warning signals” of business recessions and recoveries.

So Shiskin takes 65 economic indicators and presents elaborate tables showing their “behavior in past business cycles.” It will take time to assess the value of his result. They will begin to be tested on Oct. 23, when the Department of Commerce will initiate a new monthly report, showing how some 70 different “economic indicators” have acted during the preceding month. These will include at least 29 indicators that usually lead the business cycle, fourteen roughly coincident with it, an seven that have usually lagged behind. A few preliminary observation may be made:

1—Electronic computers are amazingly quick, but there is no scientific magic in them. They can give only the answers that are fed into them. If statisticians feed in the wrong kind of “adjustments,” weights or averages, computers will make the wrong kind.

2—Statistics, even statistics that “lead” the business cycle, such as stock prices or length of the average work week, are always the record of pastevents. As Ludwig von Mises has put it, statistics are history.

3—Even if we assume that the future is inevitably determined by the past, statistics give too partial a picture. They are mere abstractions and averages. They leave out, among other things, all the political decisions or events, domestic or foreign, that profoundly affect business.

4—Shiskin and his associates think better business indicators will help the government to fight depressions and keep business on an even keel. But free markets can do this better than anything else. Governments always seek perpetual boom.

5—It is not correct foresight itself that yields profits, but only better foresight than the majority.

6—Business predictions themselves affect the future they predict In fact, as Albert Hahn has pointed out, if the business cycle were predictable, there would be no cycle, as everybody would hedge against it.

Let us welcome the new statistical series. It should help us collectively, and no businessman can afford to be ignorant of the results. But those results can never be conclusive or completely “scientific.” The future will always be uncertain. The individual businessman, as always, will have to rely also on his knowledge of special situations and on his own hunches.