Predicting US business-cycle regimes

This article examines the use of logistic classification methods for the identification and prediction of postwar U.S. business-cycle expansion and contraction regimes as defined by the National Bureau of Economic Research (NBER) reference turning-point dates. We present a coherent theoretical framework for this task using measures of discriminatory information. The analysis encompasses model selection, parameter estimation, and classification decision rules. We examine the performance of logistic procedures in reproducing the NBER regime classifications and in predicting one and three months ahead using leading-indicator variables. Our models are shown to provide substantially more accurate business-cycle regime predictions than Markov switching specifications.