Text-Based Recession Probabilities

This paper proposes a new methodology based on textual analysis to forecast U.S. recessions. Specifically, the develops an index in spirit of Baker et al. (2016) and Caldara Iacoviello (2018) which tracks developments real activity. When used standard recession probability model, outperforms yield curve forecast, method recessions, at medium horizons, up 8 months. Moreover, contains information not included data that are useful understand episodes. as additional control slope curve, it improves accuracy by 5% 30% depending horizon. These results stable number different robustness checks, including changes estimation method, definition recessions controlling for asset purchases major central banks. Yield also outperform other popular leading indicators business cycle such PMIs, consumers' surveys or employment data.