This is the authors' abstract. We don't add key points for this paper.
The spread between the 10-year and 3-month US Treasury yields is the best-known financial predictor of recessions. Using monthly FRED data from 1959 to 2026 and NBER recession dates, I estimate a probit for a recession 12 months ahead. The slope is -0.560 (se 0.061): a spread of zero implies a 25.1% probability and a spread of minus one point a 45.6% probability. The area under the ROC curve is 0.8203, and I show that it is identical for the raw spread and the fitted probit because the probit is a monotone function of the spread; the probit changes the calibration of the signal, not its ranking. The relation has weakened: the slope is -0.651 in 1959–1985 and -0.434 in 1986–2026, with AUCs of 0.8311 and 0.7885. Since 1959 there have been eight inversion episodes. Seven were followed by a recession within 6 to 40 months. The eighth, from November 2022 to November 2024, lasted 25 months with a minimum of -1.57 points, the deepest since 1981; the fitted probability peaked at 0.582 in May 2023, and no recession had been declared as of the latest NBER data used here. A simple rule (recession within 12 months if the spread is negative) catches 40.0% of recession months with a false-alarm rate of 8.4%.
Answers are grounded in this paper and cite their sources.
No takes yet. Share an insight, caveat, or question.
Manoj Kumar Aguru (2026) studied this question.
Synapse has enriched 8 closely related papers on similar clinical questions. Consider them for comparative context: