Can we predict recessions using only unemployment data? Applying the Sahm rule to OECD countries

The accurate and early identification of economic turning points is a longstanding challenge. While in some countries, recession periods are formally determined by business cycle committees – as is the case in the United States – these processes are complex and often take considerable time to conclude. Yet policymakers, economists and financial market participants require more timely signals of shifting economic conditions to support effective and well-informed decision-making.

The commonly used “technical recession” – usually defined as two consecutive quarters of negative GDP growth – provides a relatively timely indication. However, it comes with notable limitations. Quarterly GDP data are not frequent enough, and early estimates are often subject to substantial revision. As a result, this measure alone may not provide a sufficiently reliable basis for detecting recessions in a timely and accurate manner.

To overcome these limitations, analysts often rely on higher-frequency indicators to track business cycles. However, these indicators can be volatile, sensitive to non-cyclical shocks, and sometimes difficult to interpret in real time. Composite indicators – such as the OECD Composite Leading Indicators – help to smooth these fluctuations by combining information from multiple sources. While they are sometimes able to capture a more comprehensive view of economic conditions, they are typically complex, lacking in transparency, and may rely on data that are subject to revision.

Developing transparent and data-light recession-detection methods remains of practical interest. Once such example is the Sahm rule recession indicator, originally developed for the United States (Sahm, 2019). The rule identifies a recession when the three-month moving average of the unemployment rate rises by 0.5 percentage points or more above its lowest value over the previous 12 months. Its simplicity, reliance on timely and rarely revised labour market data, and strong historical performance in the United States make it a compelling candidate for broader application.

A recent OECD Working Paper assesses whether the Sahm rule can be effectively applied across 32 OECD countries (Miyahara and Betschka, 2026). To account for differences in labour market dynamics, the authors estimate country-specific Sahm rule thresholds, rather than relying on the original 0.5 percentage point threshold defined for the United States.

The analysis highlights two key findings. First, country-specific Sahm rule thresholds are superior to applying a uniform rule. When thresholds are calibrated to maximise a predictive performance measure across 32 OECD countries, substantial cross-country differences emerge in both the thresholds and their performance measures. Second, there is a trade-off between early detection and accuracy when applying the Sahm rule across economies. Using country-specific thresholds and focusing on the first month in a series where the threshold is exceeded, roughly two-thirds of recessions are identified at an early stage. However, this timeliness comes at the cost of precision: approximately one-third of the recession signals generated do not coincide with actual recession periods, reflecting a relatively high incidence of false positives.

The Sahm indicator is calculated following Sahm (2019), and its signals are compared with actual recession periods as defined by business cycle dating committees or by the “technical recession” definition.

The original Sahm rule threshold of 0.5 percentage points was designed to avoid false positives in the United States; however, differing labour market dynamics across countries mean that this value would not necessarily translate well to other economies. In addition, setting thresholds too high to avoid false positives can increase the risk of false negatives, where genuine recessions go undetected. To balance this trade-off, thresholds are selected for each country to maximise a standard performance measure: the F-score, which captures, in a single number, how well they avoid both false positives and false negatives. 

Despite the simplicity of the original Sahm rule, direct application across OECD countries is challenging. In some cases, the indicator crosses the threshold multiple times in quick succession or fluctuates around the threshold during periods of recession. In other instances, the indicator remains close to the threshold, generating repeated alerts within the same recession episode.

To address this limitation, the paper introduces an extension of the Sahm rule: the initial alert month (IAM). The IAM focuses exclusively on the first triggered month in a sequence of alerts, aiming to reduce false positives and improve the indicator’s performance as an early warning measure.

Formally, an IAM is defined as the first month in which the Sahm indicator reaches or exceeds its threshold following a period of at least six consecutive months without any alerts. Figure 1 illustrates this approach for Japan: the blue line represents the Sahm indicator for Japan, the green horizonal line denotes the Sahm rule’s threshold (0.1), and the shaded areas indicate recession periods. The dots on the blue line indicate the periods in which the indicator meets or exceeds the threshold. Among these, June 1969, August 1970, and June 1973 are classified as IAMs (orange dots). While both June 1973 and March 1974 satisfy the IAM definition, they are associated with the same recession period, so only the earliest – June 1973 – is retained as an IAM.

The dataset includes 154 IAMs across 32 countries. For each IAM, the analysis examines whether a recession occurred within a 12-month window before or after the alert. IAMs for which no recession was observed within the window are classified as false positives (FPs), representing 37% of all IAMs. Among the remaining IAMs, 23% precede recessions, 29% coincide with recessions (within one month lead or lag), and 48% lag recessions.

On average, 31% of recession episodes are not captured by an IAM – these false negatives (FNs) indicate missed signals. In some countries, however, the Sahm rule with IAMs works particularly well. Figure 2 shows the distribution of countries by the share of FNs within total recession cases. Ten countries record no FNs, indicating that all actual recessions in these countries are successfully flagged by the IAM.

There are substantial cross-country differences in performance. Some countries exhibit relatively high shares of FPs – such as Chile, Ireland and Mexico – with the IAM signalling recessions despite none occurring within the designated window. Others, particularly those experiencing more frequent or shorter recessions record higher FN rates, as seen in Hungary, Iceland, and Sweden.

Taken together, these findings show that simple, timely labour market indicators can offer valuable early signals of turning points, provided they are adapted to country-specific conditions. At the same time, the results also highlight important limitations: the Sahm rule with the initial alert month (IAM) extension does not capture all recession episodes, and the performances varies considerably across countries. While the Sahm rule indicator may serve as a useful complement to broader monitoring frameworks, it should be interpreted with appropriate caution in policy settings. Its utility may be enhanced when used in conjunction with other economic indicators.

• Miyahara, T. and Betschka, L. (2026), “Can the Sahm rule indicator signal recession in OECD countries?”, OECD Statistics Working Paper Series, 2026/01, https://doi.org/10.1787/c15f9c28-en.
• Sahm, C. (2019), Direct stimulus payments to individuals, https://www.hamiltonproject.org/publication/policy-proposal/direct-stimulus-payments-to-individuals/.