Abstract
Against the backdrop of The Great Moderation, The Great Recession is a remarkable departure. Economic losses have been severe. The U.S. business cycle is decidedly not dead and the justifications for continued research into U.S. recession forecasting are myriad. In this study, I test whether targeting the business-cycle frequency band of explanatory data can, ceteris paribus, increase U.S. recession probability forecast model performance. This study is the first to combine frequency decomposed time series with probit models using the General-to-Specific search methodology. Using data from the NBER and The Conference Board, I compared test criteria across model estimation results derived from General-to-Specific specification selection methodologies, and concluded that, on average, those derived from the decomposed explanatory datasets outperform their full-spectrum dataset rivals. Additionally, I’ve concluded that useful explanatory information is left behind by the filtering processes utilized here and that further research into differing decomposition schemes is therefore warranted.