PREDICTING TURNING-POINTS IN BUSINESS CYCLES BY DETECTION OF SLOPE CHANGES IN THE LEADING COMPOSITE INDEX

Abstract

A Bayesian statistical method to detect turning points in the leading composite index is introduced. Under the assumption of causal priority of the leading composite index to the business cycle, the turning points in business cycles are predicted by detection of them in the index. The underlying process of the leading composite index is described by a dynamic linear model with random level and slope, where the random slope is distorted by a random shock at each turning point. The turning point is detected by obtaining a large value of the posterior probability that one of the previous slope components has undergone a major change. The intensity of the change causing a turn in the business cycle is quantified by estimating the size of the random shock. The application of the results to the US leading composite index are compared with results of earlier studies.