Predicting the Cyclical Phases of the Post-War U.S. Leading and Coincident Indicators

A bifactor model of the unobserved common leading and coincident indicators with Markov switching, introduced via the common factor intercept term, is examined. The model has four regimes and the lag between the leading and coincident factors is reflected in transition probabilities matrix. Three hypotheses concerning the relationship between the two factors are evaluated: (1) cyclical dynamics of the two factors are independent (2) cyclical dynamics are common for both factors (3) dynamics are interrelated, with coincident factor lagging behind the leading factor. The models are estimated using US monthly macroeconomic time series. The estimated recession probabilities reveal close correspondence to NBER business cycle dating. Moreover, model 3 shows that the leading factor is entering the recession 5 months and the expansions 9 months earlier than the coincident one. This permits timely forecasting of the future evolution of the coincident economic indicator.