---
ticker: "KRSAHM"
title: "South Korea Sahm-Rule Recession Indicator"
unit: "pp"
frequency: "Monthly"
source: "Sahm (2019)"
release: "Updated Monthly"
category: "Recession & Nowcasting"
country: "KR"
language: "en"
canonical: "https://kred.dev/en/series/KRSAHM"
license: "https://creativecommons.org/licenses/by-nc-nd/4.0/"
latest_value: 0.00
latest_date: "2026-08-01"
first_date: "2000-08-01"
observations_total: 313
observations_shown: 120
---

# South Korea Sahm-Rule Recession Indicator

## Overview

Measures how far unemployment sits above its recent low, flagging labor-market weakness and recession risk.

## Key Figures

|  | Value | Date |
|---|---|---|
| Latest | 0.00 | 2026-08-01 |
| Change from previous | -0.07 | 2026-07-01 |
| Change over one year | -0.10 | 2025-08-01 |
| Highest on record | 0.83 | 2021-01-01 |
| Lowest on record | -0.23 | 2001-06-01 |
| Period covered | 2000-08-01 – 2026-08-01 |  |
| Observations | 313 |  |

## Recent observations

| Date | Value | Change |
|---|---|---|
| 2016-09-01 | 0.27 | +0.10 |
| 2016-10-01 | 0.33 | +0.07 |
| 2016-11-01 | 0.27 | -0.07 |
| 2016-12-01 | 0.13 | -0.13 |
| 2017-01-01 | 0.00 | -0.13 |
| 2017-02-01 | 0.07 | +0.07 |
| 2017-03-01 | 0.10 | +0.03 |
| 2017-04-01 | 0.20 | +0.10 |
| 2017-05-01 | 0.07 | -0.13 |
| 2017-06-01 | 0.10 | +0.03 |
| 2017-07-01 | 0.03 | -0.07 |
| 2017-08-01 | 0.20 | +0.17 |
| 2017-09-01 | 0.20 | 0.00 |
| 2017-10-01 | 0.23 | +0.03 |
| 2017-11-01 | 0.13 | -0.10 |
| 2017-12-01 | 0.10 | -0.03 |
| 2018-01-01 | 0.00 | -0.10 |
| 2018-02-01 | 0.00 | 0.00 |
| 2018-03-01 | 0.13 | +0.13 |
| 2018-04-01 | 0.27 | +0.13 |
| 2018-05-01 | 0.30 | +0.03 |
| 2018-06-01 | 0.17 | -0.13 |
| 2018-07-01 | 0.17 | 0.00 |
| 2018-08-01 | 0.43 | +0.27 |
| 2018-09-01 | 0.60 | +0.17 |
| 2018-10-01 | 0.67 | +0.07 |
| 2018-11-01 | 0.43 | -0.23 |
| 2018-12-01 | 0.30 | -0.13 |
| 2019-01-01 | 0.27 | -0.03 |
| 2019-02-01 | 0.23 | -0.03 |
| 2019-03-01 | 0.13 | -0.10 |
| 2019-04-01 | 0.13 | 0.00 |
| 2019-05-01 | 0.13 | 0.00 |
| 2019-06-01 | 0.17 | +0.03 |
| 2019-07-01 | 0.17 | 0.00 |
| 2019-08-01 | 0.03 | -0.13 |
| 2019-09-01 | -0.07 | -0.10 |
| 2019-10-01 | -0.17 | -0.10 |
| 2019-11-01 | 0.03 | +0.20 |
| 2019-12-01 | 0.00 | -0.03 |
| 2020-01-01 | -0.03 | -0.03 |
| 2020-02-01 | -0.10 | -0.07 |
| 2020-03-01 | 0.13 | +0.23 |
| 2020-04-01 | 0.30 | +0.17 |
| 2020-05-01 | 0.60 | +0.30 |
| 2020-06-01 | 0.70 | +0.10 |
| 2020-07-01 | 0.80 | +0.10 |
| 2020-08-01 | 0.67 | -0.13 |
| 2020-09-01 | 0.67 | 0.00 |
| 2020-10-01 | 0.70 | +0.03 |
| 2020-11-01 | 0.77 | +0.07 |
| 2020-12-01 | 0.73 | -0.03 |
| 2021-01-01 | 0.83 | +0.10 |
| 2021-02-01 | 0.83 | 0.00 |
| 2021-03-01 | 0.67 | -0.17 |
| 2021-04-01 | 0.23 | -0.43 |
| 2021-05-01 | -0.10 | -0.33 |
| 2021-06-01 | -0.10 | 0.00 |
| 2021-07-01 | -0.17 | -0.07 |
| 2021-08-01 | -0.13 | +0.03 |
| 2021-09-01 | -0.20 | -0.07 |
| 2021-10-01 | 0.00 | +0.20 |
| 2021-11-01 | -0.10 | -0.10 |
| 2021-12-01 | 0.03 | +0.13 |
| 2022-01-01 | 0.00 | -0.03 |
| 2022-02-01 | -0.07 | -0.07 |
| 2022-03-01 | -0.13 | -0.07 |
| 2022-04-01 | -0.13 | 0.00 |
| 2022-05-01 | 0.00 | +0.13 |
| 2022-06-01 | 0.03 | +0.03 |
| 2022-07-01 | 0.10 | +0.07 |
| 2022-08-01 | 0.07 | -0.03 |
| 2022-09-01 | 0.07 | 0.00 |
| 2022-10-01 | 0.03 | -0.03 |
| 2022-11-01 | 0.03 | 0.00 |
| 2022-12-01 | -0.03 | -0.07 |
| 2023-01-01 | -0.07 | -0.03 |
| 2023-02-01 | -0.03 | +0.03 |
| 2023-03-01 | 0.00 | +0.03 |
| 2023-04-01 | -0.03 | -0.03 |
| 2023-05-01 | -0.07 | -0.03 |
| 2023-06-01 | -0.03 | +0.03 |
| 2023-07-01 | 0.10 | +0.13 |
| 2023-08-01 | 0.17 | +0.07 |
| 2023-09-01 | 0.23 | +0.07 |
| 2023-10-01 | 0.13 | -0.10 |
| 2023-11-01 | 0.17 | +0.03 |
| 2023-12-01 | 0.17 | 0.00 |
| 2024-01-01 | 0.23 | +0.07 |
| 2024-02-01 | 0.20 | -0.03 |
| 2024-03-01 | 0.20 | 0.00 |
| 2024-04-01 | 0.20 | 0.00 |
| 2024-05-01 | 0.23 | +0.03 |
| 2024-06-01 | 0.23 | 0.00 |
| 2024-07-01 | 0.10 | -0.13 |
| 2024-08-01 | 0.00 | -0.10 |
| 2024-09-01 | -0.07 | -0.07 |
| 2024-10-01 | 0.03 | +0.10 |
| 2024-11-01 | 0.10 | +0.07 |
| 2024-12-01 | 0.27 | +0.17 |
| 2025-01-01 | 0.27 | 0.00 |
| 2025-02-01 | 0.27 | 0.00 |
| 2025-03-01 | 0.17 | -0.10 |
| 2025-04-01 | 0.13 | -0.03 |
| 2025-05-01 | 0.10 | -0.03 |
| 2025-06-01 | 0.07 | -0.03 |
| 2025-07-01 | 0.07 | 0.00 |
| 2025-08-01 | 0.10 | +0.03 |
| 2025-09-01 | 0.10 | 0.00 |
| 2025-10-01 | 0.07 | -0.03 |
| 2025-11-01 | 0.03 | -0.03 |
| 2025-12-01 | 0.23 | +0.20 |
| 2026-01-01 | 0.33 | +0.10 |
| 2026-02-01 | 0.37 | +0.03 |
| 2026-03-01 | 0.17 | -0.20 |
| 2026-04-01 | 0.10 | -0.07 |
| 2026-05-01 | 0.07 | -0.03 |
| 2026-06-01 | 0.07 | 0.00 |
| 2026-07-01 | 0.07 | 0.00 |
| 2026-08-01 | 0.00 | -0.07 |

## Definition

KRSAHM is the continuous gap, in percentage points, by which the three-month moving average of the Korean seasonally-adjusted unemployment rate sits above its minimum over the previous twelve months. The Sahm rule (Sahm 2019) reads a gap of 0.50 percentage points or more as a real-time recession signal, but the headline for KRSAHM is the continuous gap itself rather than that binary trigger.

The 0.50 pp threshold is a U.S. calibration, so its transferability to Korea's lower-amplitude labor market is an empirical question. It is therefore drawn only as a reference line, tested against the official Statistics Korea business-cycle reference dates, and not treated as a Korean decision rule. Because the reference minimum excludes the current month, the gap turns negative whenever the current moving average sets a fresh twelve-month low.

A reading near zero means the unemployment trend sits near its twelve-month low, while a rising reading means the labor market is moving away from that floor and deteriorating. The Korean sample does not validate the rule as a timely recession signal, however, so KRSAHM is a descriptive continuous gauge of labor-market stress that shows how far unemployment sits above its own recent floor rather than a Korean recession-timing indicator. The timely recession channel is the recession probability (KRRECP), and KRSAHM separately reports the level of labor-market stress.

## Methodology

With $U_t$ the seasonally-adjusted unemployment rate, the indicator is computed in the following steps.

**(1) Moving average.** Compute the three-month moving average:

$$\text{MA3}_t = \tfrac{1}{3}\left(U_t + U_{t-1} + U_{t-2}\right)$$

**(2) Gap and trigger overlay.** Define the continuous gap by subtracting from the moving average its minimum over the previous twelve months excluding the current month:

$$\text{gap}_t = \text{MA3}_t - \min_{\tau \in [t-12,\,t-1]} \text{MA3}_\tau$$

The overlay trigger is $\mathbf{1}\{\text{gap}_t \geq 0.50\}$. The reference window is the previous twelve months excluding the current month, matching the U.S. official gauge's construction, so the gap can go negative at fresh lows.

**(3) U.S. reproduction check.** The window definition is proven rather than assumed. Applying the identical algorithm to the U.S. seasonally-adjusted unemployment rate series reproduces the official U.S. real-time Sahm-rule gauge with a mean absolute difference of 0.002 percentage points and a maximum of 0.07 over the 926 common months, where the residual arises because the official gauge computes from the unrounded unemployment rate while the published unemployment-rate input is rounded to one decimal. The first valid month also matches exactly, becoming 1949-03 for a U.S. unemployment-rate series that starts in 1948-01, which only the exclusive window reproduces. An inclusive window fails this comparison by two orders of magnitude. This U.S. match is re-verified on every pipeline run and a structural test gates on it.

**(4) Frequency, sample, and loading.** The series is monthly with no daily projection, since unemployment is observed monthly. The burn-in is 14 months. The Korean seasonally-adjusted unemployment rate starts in June 1999, so the gap is available from August 2000 onward. Seasonal factors are re-estimated by the statistical agency and restate the full history, so the model recomputes the entire sample on every pipeline tick and the loader replaces the table outright.

## Applications in Economics

In the United States the Sahm rule is an exceptionally clean real-time recession signal. Sahm (2019) designed the rule as an automatic trigger for direct stimulus payments, exploiting the empirical regularity that U.S. unemployment rises quickly and unambiguously once a recession begins, and it has since become the reference automatic-stabilizer switch in the U.S. policy debate. The 0.50 pp threshold has fired within the first months of every recession since 1970, as dated by the National Bureau of Economic Research (NBER) Business Cycle Dating Committee, and has produced essentially no false alarms outside them.

This cleanness sits inside a longer tradition of treating unemployment as the most reliable real-time coincident-recession variable. Stock and Watson (1989) construct a dynamic-factor coincident index from a four-variable real-activity panel in which the unemployment-related component carries decisive weight at cycle turns. Hamilton (1989) reaches a related conclusion through a regime-switching specification in which the recession state is identified primarily from output- and labor-market dynamics together. Chauvet and Piger (2008) and Berge and Jordà (2011) evaluate real-time recession-dating procedures and find that the U.S. labor-market signal is unusually sharp by international standards, which is the empirical foundation Sahm (2019) builds on.

The Korean evidence does not support reusing the rule as a recession-timing tool, and this page reports the test outcome rather than re-labelling the gauge into the role it failed. Against the five Statistics Korea reference contractions that overlap the sample, the canonical 0.50 pp trigger fires in only two, namely the 2003 credit-card crisis (first firing 7 months after the reference peak) and the long cycle-11 contraction (first firing in September 2018, 12 months after the 2017-09 peak). It misses the 2000–2001 contraction (maximum gap 0.40 pp), the global financial crisis within the reference window (maximum 0.37 pp before the 2009-02 trough), and the 2011–2013 slowdown almost entirely (maximum 0.17 pp). A lag of 7–12 months at the peak is too late to function as a coincident dating signal, and the 0.37 pp and 0.17 pp misses are failures of the threshold rather than near-misses. The gap then keeps firing after the official troughs, in 2009-05 through 2010-03 and again in 2020-06 through 2021-03, because Statistics Korea dates the output-cycle bottom early relative to the labor market and Korean unemployment keeps deteriorating for several quarters past it. Of the 19 triggered months outside the reference windows, 17 are these post-trough continuations and one isolated month (November 2014, at exactly 0.50 pp) is a genuine false alarm.

No Korean threshold separates hits from false alarms cleanly. Lowering the trigger to 0.15 pp catches all five contractions in-window but fires in 68 months outside them, and intermediate thresholds trade hits against false alarms without a clean break, so no in-sample recalibration rescues the binary-trigger interpretation. The Berge and Jordà (2011) ROC-style evaluation of recession classifiers makes this trade-off transparent, with the Korean Sahm classifier sitting far from the corner where U.S. Sahm performance lives. The continuous gap is therefore the headline, and KRSAHM ships as a descriptive labor-market-stress gauge rather than a Korean recession switch.

Korean unemployment is low-amplitude by international standards. This partly reflects labor-market institutions that adjust hours and participation rather than headcount, which the Statistics Korea (KOSTAT) Economically Active Population Survey and Employment Trends document at monthly cadence. The self-employment share that Statistics Korea reports also sits well above OECD averages on an international comparison basis. The Korean seasonally-adjusted unemployment series starts in June 1999, so the 1998 Asian-financial-crisis episode that would presumably dominate the sample is untestable. With a structurally low-amplitude unemployment cycle combined with a left-censored sample, even an attempted Korean recalibration along the Sahm template would have a sparse positive sample to work against.

## Applications in Financial Markets

The gap is best read as a labor-market stress meter that records how far the current unemployment trend sits above its own twelve-month floor. A reading near zero says the trend is at or near its twelve-month low, while a reading approaching the U.S. reference line indicates deterioration of the scale that in the United States accompanies the start of a recession. The 0.50 pp line is drawn only as a benchmark for visual orientation rather than a Korean decision rule, given the documented misses at 0.37 pp and 0.17 pp in genuine Statistics Korea contractions and the 7–12 month lag where the trigger does fire.

The relationship to the recession probability (KRRECP) is a division of labor rather than redundancy. The timely recession channel on this site is KRRECP, which extracts a 3-month-ahead signal from the ACM-decomposed yield-curve slope and the near-term forward spread along the Estrella and Mishkin (1998) and Rudebusch and Williams (2009) tradition, and the Korean Sahm test reported above shows that KRSAHM cannot share that role. KRSAHM instead measures the level of labor-market stress relative to its own recent floor. A rising KRRECP probit signals an approaching contraction, while a rising Sahm gap separately signals that the labor market is deteriorating now relative to its twelve-month low, a stress diagnostic worth reading even when no recession is imminent, as in 2014–2015.

Two operational caveats apply. First, Korean unemployment is low-amplitude by international standards, so as long as the labor-market institutions the Statistics Korea Employment Trends release documents run adjustment through hours and participation rather than headcount, level comparisons with U.S. readings overstate calm. Second, seasonal-factor revisions restate the entire history each month, so recent gap values can move by a few hundredths of a percentage point between vintages, and the full-replace loading discipline shared with KRADS, KRFCI, and KRGAR keeps them consistent across vintages.

The U.S. Sahm signal is silent outside NBER-dated recessions, but the underlying Stock and Watson (1989) coincident-index logic remains informative as a continuous reading. KRSAHM's value to the desk in the periods between contractions is the same. Read through the Chauvet and Piger (2008) real-time dating frame, a rising gap points to a deteriorating labor market that Statistics Korea may or may not later date as a recession. The gauge sits in the dashboard as the labor-stress level reading, paired with KRRECP for the timing channel and KRADS for the broader coincident momentum reading.

## Statistical Tests

KRSAHM is the Sahm (2019) recession indicator, the three-month moving average of the unemployment rate minus its minimum over the preceding twelve months, a rolling-window transform whose output is range-bounded by the window and therefore mean-reverting by construction, so its integration order is fixed by the definition rather than estimated from the data. The augmented Dickey-Fuller test of Dickey and Fuller (1979), the Phillips and Perron (1988) test, the KPSS test (Kwiatkowski et al. 1992), the DF-GLS test of Elliott, Rothenberg, and Stock (1996), and the Ng and Perron (2001) refinement were deliberately not run, since on a bounded-memory rolling-minimum construction a unit-root rejection confirms the definition and not the economy, and the Hodrick-Prescott filtering ruling does not apply because the indicator is a rolling minimum rather than a two-sided filter, whose break asymptotics require a stationary object in any case (Sahm 2019; Perron 1989).

As a by-construction-stationary object the indicator is summarized by persistence and amplitude. Its first-order autoregressive coefficient is 0.901 over 309 monthly observations from 2000-08-01 to 2026-04-01, a half-life of about 6.6 months reported as a descriptive property and not as an order of integration and read with the small-sample downward bias of Andrews (1993) in mind, since a level portmanteau such as Ljung and Box (1978) would measure the rolling-window smoothing rather than the data-generating process. The indicator has a standard deviation of 0.22 percentage points and ranges from −0.23 to 0.83 percentage points.

The Bai and Perron (1998, 2003) multiple structural-break procedure, admissible on a by-construction-stationary object, locates two breaks in the mean at 2014-01-01 and 2021-05-01, the latter coinciding with the pandemic-era labor-market swings. Because the trailing window uses only past data, the indicator is non-revising by construction and reproduces the reference United States Sahm rule to a mean absolute deviation of 0.0023 percentage points (Orphanides and van Norden 2002).

## Frequently Asked Questions

### What does the Sahm-rule recession indicator compare?

The three-month moving average of the seasonally adjusted unemployment rate against its lowest value over the preceding twelve months, in percentage points. It measures how far unemployment has risen from its own recent floor.

### Why does the Sahm rule anchor to a recent unemployment floor?

The level at which unemployment becomes concerning differs by country and period, so absolute levels do not compare. Anchoring to the economy's own recent low makes readings comparable across episodes.

### Does crossing the Sahm-rule threshold declare a recession?

It does not. The KRED headline is the continuous gap rather than a binary signal, and it is published as a descriptive measure carrying no predictive claim.