Federal Reserve Board research finds low national recession risk while highlighting state-level pockets of vulnerability

The Federal Reserve Board published a FEDS Note extending a Bayesian Markov-switching framework that uses national and state nonfarm payroll employment data to estimate recession probabilities and distinguish between full-recovery U-shaped recessions and L-shaped recessions associated with hysteresis. The analysis suggests national recession risk has been low in 2023–2025 and classifies the COVID-19 recession as a U-shaped episode with a low likelihood of hysteresis, while state-level estimates indicate localized areas of higher risk. Recession probability estimates broadly line up with National Bureau of Economic Research recession dates, and the employment-based approach yields higher L-shaped recession probabilities than a version using real GDP growth, implying stronger hysteresis effects in labor markets than in aggregate output. At the state level, the model shows substantial cross-state heterogeneity and identifies idiosyncratic downturns such as Louisiana in 2005 and North Dakota in 2015; it also finds U-shaped recessions became less common after the 1990s, with L-shaped patterns becoming more prevalent. In 2023–2025, state recession probabilities range from zero to 10 percent, with Massachusetts and Rhode Island showing U-shaped probabilities near 10 percent and New York and New Jersey exhibiting somewhat elevated L-shaped probabilities; the note also cites Beige Book evidence of softer conditions in New England. While emphasizing that predicting recessions in advance remains challenging, the note reviews additional macro and financial indicators and heuristic or model-based signals and finds they generally point to low near-term recession risk, including low initial claims and layoff indicators, a positive yield curve slope, low credit spreads, and low readings from measures such as the Sahm Rule and various sentiment- and labor-stress-based models.