PAYEMS: US Total Nonfarm Payrolls Monthly Employment Data from FRED (1939–2026)

The PAYEMS series, published monthly by FRED from the Bureau of Labor Statistics, tracks the total number of paid US workers excluding farm employees, private household staff, non-profit organization employees and active-duty military personnel since January 1939 — over 1,000 monthly observations. PAYEMS is the headline measure of US payroll employment and the single most market-moving economic data release worldwide, used by the Federal Reserve, the IMF and the OECD as the reference series for US labor market conditions.

Dataset: US Nonfarm Payrolls (1939–2026) · Updated 2026-06-01

Source: FRED series PAYEMS · 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 underlying data is available from FRED under series code PAYEMS:

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

Direct CSV Access — Eco3min Structured Dataset

https://eco3min.fr/dataset/us-nonfarm-payrolls.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/us-nonfarm-payrolls.csv"
df = pd.read_csv(url, parse_dates=["date"])

print(df.head())
print(df["payems"].describe())

Using the Dataset in R

library(readr)

url <- "https://eco3min.fr/dataset/us-nonfarm-payrolls.csv"
df <- read_csv(url)

head(df)
summary(df$payems)

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

Methodology

The PAYEMS series comes from the BLS Current Employment Statistics (CES) program, a monthly survey of approximately 119,000 business and government establishments covering roughly one-third of total US nonfarm employment. Initial estimates are released approximately 3 weeks after the reference period (the pay period including the 12th of the month). The survey methodology was redesigned in 2003 to incorporate the birth-death imputation model, which estimates net job creation from business openings and closings not yet captured in the sample.

PAYEMS is seasonally adjusted by BLS using X-13ARIMA-SEATS. Each monthly release revises the previous two months’ figures based on additional sample responses, and an annual benchmark revision in February rebases the entire series to the Quarterly Census of Employment and Wages (QCEW) — UI tax records covering about 95% of US jobs. The non-seasonally-adjusted equivalent is published under FRED code PAYNSA.

This dataset is updated monthly via automated pull from the FRED API.

Data Quality & Provider Notes

The PAYEMS series carries an unusually wide revision footprint relative to other macro releases of comparable importance, which has direct implications for how the initial release should be read.

• Release latency. BLS publishes the Employment Situation report on the first Friday of the following month at 8:30am ET — a 3–4 week lag depending on the calendar. FRED mirrors the release within minutes. The Eco3min pipeline pulls the FRED feed on a monthly cadence.
• Revisions policy. Each monthly release revises the preceding two months. The annual benchmark revision, published every February and previewed in August, rebases the level to QCEW administrative data and can shift the trajectory materially: the August 2024 preliminary benchmark cut March 2024 employment by approximately 818,000, the largest downward revision since 2009. Earlier vintages are accessible via ALFRED.
• Alternative sources. The BLS publishes the same data directly via its Employment Situation release page. ADP National Employment Report covers private-sector payrolls only with a different statistical methodology (matched-sample regression on ADP-served firms) and is released two days earlier; it is not a substitute for PAYEMS but provides a complementary read.
• Known gaps. US federal government shutdowns can delay BLS releases (notably October 2013 and January 2019). The birth-death imputation model — which contributes a non-survey-based component to the monthly change — is particularly difficult to interpret near cyclical turning points, when business openings and closings diverge from their historical patterns.

For analytical work, the convention is to read the 3-month moving average of the change in PAYEMS rather than a single monthly print, and to consult ALFRED vintages when comparing reactions in real time to subsequent revisions.

Common Pitfalls When Using PAYEMS

The PAYEMS series is widely used but several recurring interpretation errors materially distort the labor-market read.

1. Reading the headline initial release as a stable signal. The first-print monthly change in PAYEMS has an average absolute revision of roughly 50,000 jobs by the third estimate, and benchmark revisions can move the level by hundreds of thousands. A single release should be treated as a noisy estimate, not a verdict on the labor market.
2. Conflating the establishment survey (CES, the source of PAYEMS) with the household survey (CPS, the source of the unemployment rate). CES surveys employers and counts jobs; CPS surveys households and counts employed persons. Multiple jobholders inflate CES relative to CPS, and the two surveys can diverge for extended periods — as they did in 2023 and 2024 — without one being “wrong”.
3. Ignoring the birth-death model contribution. The CES sample does not include newly opened firms, so BLS imputes their employment via a statistical model. In normal cycles, this model contributes modestly; near turning points, the model continues imputing job creation while actual business formation is contracting, which has historically biased PAYEMS upward at the entry to recessions — the central subject of our NFP revisions study.
4. Treating total nonfarm payrolls as equivalent to private-sector payrolls. PAYEMS includes government employment, which can grow countercyclically through hiring at the federal, state and local levels (notably during census years and pandemic-response periods). The private payrolls series (FRED code USPRIV) is the cleaner read for underlying business-cycle dynamics.

Historical Regimes

1939–1945 — Wartime mobilization. PAYEMS grew from roughly 30 million at series inception to 41.6 million at the wartime peak. The series began publication as the Department of Labor formalized statistical labor reporting under the 1939 Bureau of Labor Statistics reorganization.

1955–1973 — Postwar expansion. Steady 1–3% year-over-year payroll growth through three short recessions; PAYEMS doubled from approximately 50 million to 76 million over the period. The era featured a stable manufacturing employment share around 30% of total nonfarm payrolls.

1973–1982 — Two stagflation recessions. Cumulative payrolls fell during the 1973–1975 oil-shock recession and again during the 1980 and 1981–1982 disinflation recessions. The early-1980s episode pushed the unemployment rate to 10.8% while PAYEMS contracted by roughly 2.8 million from its July 1981 peak.

1991–2000 — Tech-led expansion. The economy added approximately 22.5 million payrolls over the decade, driven by services and information technology. Manufacturing share of total PAYEMS fell below 15% for the first time in series history.

2001–2003 — Jobless recovery. Following the tech bust, PAYEMS contracted by 2.7 million and took approximately 46 months to recover the prior peak — the slowest jobs recovery since the series began, until 2008. The way these payroll cycles surface in the unemployment rate is mapped in Eco3min’s reading of the unemployment rate across the cycle.

2008–2010 — Global Financial Crisis. PAYEMS contracted by 8.7 million from its January 2008 peak to its February 2010 trough — the deepest peacetime loss in the series. The 2010 employment recovery took six years to reach the prior peak, the slowest non-pandemic recovery on record.

2020 — Pandemic shock. March–April 2020 saw a cumulative loss of approximately 22 million payrolls in two months — the largest single-event labor collapse in PAYEMS history and roughly 2.5 times the entire GFC drawdown. The recovery to the prior peak took 28 months, faster than the GFC despite a deeper trough.

2022–2026 — Recovery and benchmark revisions controversy. Headline monthly prints averaged above 250,000 from 2022 through mid-2024 before the August 2024 preliminary benchmark revision cut the March 2024 level by approximately 818,000 jobs — the largest downward benchmark revision since 2009. The revision shifted the perceived underlying pace of hiring and re-opened the methodological debate on the birth-death model addressed in our five recession rules study.

Related Macroeconomic Datasets

PAYEMS is the establishment-survey anchor in the US labor data stack. The datasets below cover the household-survey counterparts, higher-frequency labor flows and downstream activity measures that PAYEMS directly relates to.

Macroeconomic Dataset Hub

This dataset is part of the Eco3min macro-financial data repository.

Explore the Eco3min Dataset Hub

Sources

• Federal Reserve Bank of St. Louis — FRED database
• US Bureau of Labor Statistics — Current Employment Statistics (CES) program, Employment Situation report

Dataset Reference

Eco3min Research. (2026). US Total Nonfarm Payrolls — PAYEMS (1939–2026).<br> https://eco3min.fr/en/us-nonfarm-payrolls-dataset/