A Criterion of Business Activity

Abstract

THIS IS A REPORT on a new experiment in the use of economic indicators. What is new about it is the method of combining a number of such indicators into one usable figure. The components themselves are mostly well-known, although several employed here are original. Analysts who use them, must know the practical difficulty of forecasting business from a multitude of individual indicators, because of their diverse and often conflicting behavior. Ideally-for the busy analyst, or one frankly unable to reach a decision from many indicators the solution would be to have, instead, just one useful composite indicator. This would violate no statistical theory. To become a practical tool, the composite would have to meet certain minimum requirements with respect to (a) an optimum duration of the composite lead (not too few but not too many months); (b) a certain minimum extent of movement between the composite indicator's cyclical turning point and the beginning of a general recession or recovery; and (c) a certain minimum degree of consistency or similarity of behavior of the composite in different business cycles. Computers would be a logical medium for experimenting with the construction of such a tool. Different combinations of indicators, possibly with different weights, could be tested until the desired composite emerged. From a practical standpoint, there would arise, of course, considerable difficulties with such an undertaking. To begin with, there is no definitive universe of indicators with which to work: it is not closed but potentially expanding. There is no reason, for example, to assume that it is limited to the thirty-odd leading indicators of the National Bureau of Economic Research currently published by the Department of Commerce. Another inconvenience is the fact that whereas aggregate economic activity has a pronounced rising trend, most of the widely known leading indicators move roughly horizontally on average over an extended period, and some (such as inverted large business failures and liabilities of business failures) have a strong downward long-term trend. Finally, it is doubtful that any possible combination of leading indicators could produce a composite with a recognizable lead at the bottoms of recessions, simply because hardly any individual indicator gives a practical clue to the imminence of recovery-in fact, some lag behind recession bottoms. Diffusion indexes do not consistently overcome this difficulty. A clue to an upturn must be provided mainly by other means, such as the relative position of leading indicators rather than by their movement. In view of these technical difficulties, and for othermore basic-reasons, the approach here described does not rely on purely mechanical means of selection of components. Instead, it utilizes the fact that a specially constructed composite of empirically chosen individual indicators moves not only earlier but at a different rate from that of aggregate economic activity (to the trend of which it is roughly geared), and thus serves as a criterion of general business. Figure 1 shows our composite, which we call the Business Criterion, together with the FRB Index of Industrial Production (used here to represent aggregate economic activity), both by months back to 1953. It may be seen that the Criterion is useful in two ways: it not only tends to anticipate the main cyclical swings of Industrial Production, especially downward, but also -by reference to its position vs. that of industrial production-suggests the relative strength or weakness of business conditions. Thus, when the Business Criterion is significantly below Industrial Production, and especially when it is falling, the position of business is presumed to be potentially weak; and when it is significantly above Industrial Production, and especially when it is rising, the indicated trend of business may be judged to be upward. In this manner, forecasting need not rely solely on leads but gets additional help from the relative positions of the two lines. This is the general idea, and in the case of the few cycles here covered, it has worked quite satisfactorily. Note that before a general recession the Business Criterion goes into its sharp phase of decline about three to four months before the onset of such recession, and always from a position below Industrial Production.*