SAHMREALTIME: Real-Time Sahm Rule Recession Indicator — Monthly Series Since 1959

This dataset covers the real-time variant SAHMREALTIME monthly since 1959 — built on the unrevised unemployment data available at each point, not the later-revised series. Every trigger since 1970 has coincided with an NBER-dated recession, though the brief 2024 crossing raised questions about that streak.

Dataset: Sahm Rule Recession Indicator (1959–2026) · Updated 2026-06-01

Source: FRED series SAHMREALTIME · Federal Reserve Bank of St. Louis

Macro Takeaway

Dataset Overview

Dataset Variables

The CSV and Excel files contain the following columns.

Column names match the CSV headers exactly.

Download the Complete Dataset

The full dataset is available in CSV and Excel formats.

FRED Direct CSV Access

The raw data is available via FRED under code SAHMREALTIME:

https://fred.stlouisfed.org/graph/fredgraph.csv?id=SAHMREALTIME

The Eco3min version provides a clean, analysis-ready format with consistent column names, pre-calculated derived metrics where applicable, and both CSV and Excel downloads.

Direct CSV Access — Eco3min Structured Dataset

https://eco3min.fr/dataset/sahm-rule-indicator.csv

This URL returns the complete dataset in CSV format. It can be used directly in pandas, R, curl, or any data tool.

Using the Dataset in Python

import pandas as pd

url = "https://eco3min.fr/dataset/sahm-rule-indicator.csv"
df = pd.read_csv(url)

print(df.head())
print(f"Latest value: {df['sahm_indicator'].iloc[-1]:.2f}")

Using the Dataset in R

library(readr)

url <- "https://eco3min.fr/dataset/sahm-rule-indicator.csv"
df <- read_csv(url)

head(df)
summary(df$sahm_indicator)

Both examples load the dataset directly from the URL — no download or API key required.

Methodology

The Sahm Rule was developed by economist Claudia Sahm during her tenure at the Federal Reserve as a fast, robust, real-time recession indicator that could be used to trigger automatic stabilizer policies. The rule is mechanically simple: subtract the prior 12 months’ minimum of the 3-month moving average of the national unemployment rate from the current 3-month moving average. When the difference reaches 0.50 percentage points, the indicator triggers.

SAHMREALTIME uses the unemployment rate as it was originally reported in real time, not the fully-revised series. This matters because BLS revises seasonally-adjusted unemployment in subsequent releases. The realtime variant reproduces what a policymaker actually saw at each historical point. FRED publishes a companion series, SAHMCURRENT, that uses the latest revised unemployment data — useful for academic backtesting but not for replicating contemporaneous decision-making.

This Eco3min dataset is updated monthly via automated pull from the FRED API following each BLS Employment Situation release (first Friday of each month).

Data Quality & Provider Notes

SAHMREALTIME is published by the St. Louis Fed using methodology documented by Claudia Sahm. Eco3min mirrors FRED with a daily pull, capturing each monthly update shortly after the underlying BLS unemployment release.

• Release latency. SAHMREALTIME is updated monthly after the BLS Employment Situation release (first Friday of the month), incorporating the new unemployment observation into the 3-month moving average. Latency is typically same-day or next-day to FRED.
• Revisions policy. SAHMREALTIME is, by construction, not revised: each historical value is locked to the unemployment data available at the time. This is the critical distinction from SAHMCURRENT, which is revised whenever BLS revises the underlying unemployment rate. For point-in-time backtests, SAHMREALTIME is the appropriate series.
• Alternative sources. The indicator can be reconstructed manually from the FRED unemployment rate (UNRATE) using the public methodology — but reconstruction will produce the SAHMCURRENT version, not SAHMREALTIME. Claudia Sahm’s own publications and the St. Louis Fed’s documentation remain the canonical references.
• Known gaps. The series is available monthly back to 1949 in FRED, though the data quality of US unemployment data pre-1959 is notably weaker (sampling frame changes in 1948, 1959, 1967, 1994). For applications requiring uniform data quality, restricting the analysis window to 1959 onward is the conventional choice.

Before using SAHMREALTIME in any analysis, verify the latest observation date against the BLS Employment Situation release calendar.

Common Pitfalls When Using SAHMREALTIME

SAHMREALTIME is mechanically simple but several recurring interpretation errors weaken downstream analysis — particularly around what the indicator does and does not claim to measure.

1. Confusing SAHMREALTIME with SAHMCURRENT. FRED publishes two Sahm Rule series. SAHMREALTIME uses the unrevised unemployment data available at each historical point; SAHMCURRENT uses fully-revised unemployment data and is therefore retroactively rewritten when BLS revises UNRATE. For policy-relevant or contemporaneous-decision analysis, SAHMREALTIME is the correct series. For pure historical pattern analysis, SAHMCURRENT may be preferred — but the two will not match exactly.
2. Reading the 0.50pp threshold as mechanical. The 0.50pp trigger is an empirical regularity that has historically aligned with NBER recessions since 1970, not a structural law. Recessions can occur without a SAHMREALTIME trigger (rare but possible in fast-moving shocks), and elevated SAHMREALTIME readings can occur without a recession when the unemployment rise reflects labor-supply expansion rather than demand collapse — as the 2024 episode demonstrated.
3. Assuming a Sahm trigger means recession is happening now. NBER recession dating typically lags the start of a recession by 6 to 18 months. The Sahm Rule was designed for real-time policy activation precisely because the official recession start is known only retrospectively. Reading a SAHMREALTIME trigger as “we are in recession” conflates the indicator’s signal with the NBER’s definitive call.
4. Ignoring labor-supply effects. The Sahm Rule was constructed during decades when unemployment rate changes were almost entirely demand-driven. Episodes where the labor force expands rapidly (immigration waves, post-pandemic re-entry) can inflate the unemployment rate without the demand-collapse mechanism that underlies the rule’s recession signal. The 2024 trigger is the textbook example of this scenario.

Historical Regimes

1959–1969 — Low volatility. SAHMREALTIME remained near zero during the longest post-war expansion. The 1960–61 recession triggered the rule, but the signal was mild relative to later episodes.

1970–1982 — Repeated triggers. Four recessions in 12 years produced clear SAHMREALTIME signals, each time with the indicator surging well above 0.50pp. The 1973–75 and 1981–82 recessions produced readings above 2. The mechanics of that 0.50-point trigger are detailed in our breakdown of the Sahm Rule recession threshold.0pp, the highest sustained levels of the modern series.

1983–2007 — Great Moderation. Only two triggers (1990, 2001), both relatively mild by historical standards. SAHMREALTIME spent most of this period near zero, reflecting historically low unemployment volatility and the structural shift to a more services-oriented economy.

2008–2009 — GFC. SAHMREALTIME surged to 4.0pp — the highest reading in the modern dataset — reflecting the catastrophic labor market deterioration of the Great Recession. The trigger preceded the NBER-dated recession start by a few months but lagged the equity market peak.

2020 — COVID. The sharpest and most extreme trigger in the SAHMREALTIME history (well above 10pp), but also the fastest reversal. The unemployment spike was artificially induced by lockdowns rather than organic economic deterioration, and reversed mechanically once mobility restrictions lifted.

2024 — Ambiguous trigger. SAHMREALTIME briefly exceeded 0.50pp in mid-2024, sparking recession fears. The trigger occurred without the underlying demand-collapse dynamics that historically accompanied previous triggers. Whether this represents a genuine recession signal, a delayed signal, or the first false positive in the indicator’s history remains an active empirical question — and is precisely the regime tested in the five recession rules study.

Related Macroeconomic Datasets

SAHMREALTIME is built on a single input (UNRATE) and a single rule. Cross-referencing with adjacent labor-market and financial signals turns a binary trigger into a structured read of where the cycle actually stands.

Frequently Asked Questions

Macroeconomic Dataset Hub

This dataset is part of the Eco3min macro-financial data repository. Explore all available datasets including inflation, interest rates, equity returns, credit spreads, and currency indicators.

Sources

• Federal Reserve Bank of St. Louis — FRED series SAHMREALTIME
• Bureau of Labor Statistics — Employment Situation (underlying unemployment rate)

Dataset Reference