Examining Predictors of U.S. Recessions: A Regime-Switching Approach

This study uses univariate and bivariate regime-switching models to compare the predictive performance of five popular business cycle indicators. The empirical results suggest that all considered series are sensibly modelled as following a two-state regimeswitching process. For each variable one regime more or less strongly indicates recession periods, while the other one is associated with phases of economic recovery or expansion. The yield curve spread is confirmed to be the most reliable recession predictor with an average predictive lead time of three quarters. As the most important result, our simple univariate model turns out to be a filter that transforms accurately term spread changes into turning point predictions.