Website CV Paper
Fields
Econometrics, Macroeconomics, Urban & Regional, Finance
Job market paper
Output Gap Measures, Data Revisions, and Predictive Performance
This paper compares eight output gap methods and four forecast combinations on U.S. data from 1980 to 2025. Each gap is estimated ex post, in quasi-real time, and in real time, which separates the effect of later observations from that of data revisions. The estimates differ in level, volatility, and cyclical timing. Ex post, the HP gap signals recessions a year ahead and, before 2020, improves growth forecasts, as do simple combinations. These gains largely disappear once the sample is truncated at each forecast date. In real time, neither individual gaps nor combinations significantly improve on current growth for recessions or on an ARMA benchmark for growth, even when they revise little.