Diffusion indexes and a statistical test for predicting turning points in business cycles

Abstract

This paper presents a non-econometric statistical procedure for testing the significance of turning points in a sample diffusion index of leading indicators, allowing one to make inferences with respect to turning points of the corresponding business cycle. The procedure is applied utilizing 30 leading indicators in the prediction of peaks and troughs in the U.S. business cycle between 1969 and 1982. For this time period, the procedure correctly predicts the four peaks and the four troughs in the U.S. business cycle with a lead time range of from zero to twelve months, although false turns with respect to the prediction of peaks were recorded. This procedure compares very favorable with the Neftci non-econometric statistical test approach for predicting turning points in business cycles. Further, this procedure may be used in conjunction with smoothing and weighting techniques for improving predictions from leading indicators.