Abstract
This paper develops a new rule to detect US recessions by combining data on job vacancies and unemployment. We first construct a new recession indicator: the minimum of the Sahm-rule indicator (the increase in the 3-month average of the unemployment rate above its 12-month low) and a vacancy analogue. The minimum indicator captures simultaneous rises in unemployment and declines in vacancies. We then set the recession threshold to 0.29 percentage points (pp), so a recession is detected whenever the minimum indicator crosses 0.29pp. This new rule detects recessions faster than the Sahm rule: with an average delay of 1.2 months instead of 2.7 months, and a maximum delay of 3 months instead of 7 months. It is also more robust: it identifies all 15 recessions since 1929 without false positives, whereas the Sahm rule breaks down before 1960. By adding a second threshold, we can also compute recession probabilities: values between 0.29pp and 0.81pp signal a probable recession; values above 0.81pp signal a certain recession. In December 2024, the minimum indicator is at 0.43pp, implying a recession probability of 27%. This recession risk was first detected in March 2024.
Figure 2: Construction of the Michez rule, 1960–2024
Figure 3: Michez rule in the United States, 1960–2024
Figure 6: Michez rule in the United States, 1929–1959
Figure 7: Dual-threshold Michez rule in the United States, 1960–2024
Citation
Michaillat, Pascal, and Emmanuel Saez. 2025. “Has the Recession Started?” Oxford Bulletin of Economics and Statistics 87 (6): 1047–1058. https://doi.org/10.1111/obes.12685.
@article{MS25,
author = {Pascal Michaillat and Emmanuel Saez},
year = {2025},
title = {Has the Recession Started?},
journal = {Oxford Bulletin of Economics and Statistics},
volume = {87},
issue = {6},
pages = {1047--1058},
doi = {https://doi.org/10.1111/obes.12685}}
Related material
- Recession indicator in real time
- Recession probability in real time
- Follow-up paper (2025) - This paper searches over hundreds of millions of recession classifiers built from US unemployment and vacancy data and finds several that detect recessions even more rapidly and more precisely than the Michez rule itself.