Abstract
The paper develops a new method for detecting US recessions in real time. The method combines recession classifiers selected from the anticipation-precision frontier, improving upon traditional approaches like the Sahm rule. The Sahm rule and similar threshold-based methods rely on simple but arbitrary triggers. The proposed method systematically constructs and evaluates millions of recession classifiers, constructed by combining unemployment and vacancy data to reduce detection noise. The classifiers are trained to avoid both false negatives (missed recessions) and false positives (nonexistent recessions). By selecting classifiers that lie on the anticipation-precision frontier, the method optimizes early detection without sacrificing precision. Using classifiers trained on 1929–2021, I find that the probability that the US is in recession in February 2025 is 71%. Backtesting confirms that the new method detects recessions in a timely manner and with great reliability, providing a robust tool for policymakers.
Anticipation-precision of 2,343,752 recession classifiers trained on 1929–2021 data
Recession probability in the United States, 2022–2025, obtained from classifiers on the anticipation-precision frontier
Recession probability in the United States, 2005–2025, obtained in backtesting on 1929–2004 data
Citation
Michaillat, Pascal. 2025. “Early and Accurate Recession Detection Using Classifiers on the Anticipation-Precision Frontier.” https://pascalmichaillat.org/17/.
@techreport{M25,
author = {Pascal Michaillat},
year = {2025},
title = {Early and Accurate Recession Detection Using Classifiers on the Anticipation-Precision Frontier},
url = {https://pascalmichaillat.org/17/}}