ICSA: US Initial Jobless Claims Weekly Data from FRED (1967–2026)

The ICSA series, published weekly by FRED from the US Department of Labor Employment and Training Administration, tracks new filings for unemployment insurance in the United States since January 1967 — more than 3,000 weekly observations. Released every Thursday at 8:30am ET with only a 5-day lag, ICSA is the highest-frequency labor market indicator in the US data stack and the earliest available signal of deteriorating employment conditions.

Dataset: US Initial Jobless Claims (1967–2026) · Updated 2026-05-15

Source: FRED series ICSA · 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 ICSA:

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

Direct CSV Access — Eco3min Structured Dataset

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

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

Using the Dataset in R

library(readr)

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

head(df)
summary(df$icsa)

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

Methodology

The ICSA series counts the number of new initial claims for state unemployment insurance filed during the reference week (Sunday through Saturday). The source data is collected by the US Department of Labor Employment and Training Administration from the 50 state UI programs plus DC, Puerto Rico and the Virgin Islands, then transmitted to FRED. Claims under the Federal Employees and Newly Discharged Veterans programs (UCFE, UCX) are tracked separately and are not included in ICSA.

The series is seasonally adjusted by the US DOL using X-13ARIMA-SEATS methodology, with seasonal factors revised every five years against historical claims data. The non-seasonally-adjusted equivalent is published under FRED code ICNSA.

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

Data Quality & Provider Notes

The ICSA series is one of the most operationally reliable high-frequency US data points: weekly release, narrow revision band, and a continuous methodology since 1967 with a single seasonal-adjustment update window every five years.

• Release latency. The US DOL Employment and Training Administration publishes ICSA every Thursday at 8:30am ET for the week ending the prior Saturday — a 5-day lag. FRED mirrors the release within approximately one hour; the Eco3min pipeline pulls the FRED feed on a weekly cadence.
• Revisions policy. Each weekly release contains the new observation plus a single revision to the prior week’s reading. Magnitudes are typically small (a few thousand claims) and rarely change the directional read. There are no large back-revisions; the 4-week moving average is the standard reference for trend analysis.
• Alternative sources. The same underlying data is available directly from the US DOL ETA via the “UI Weekly Claims Report” PDF and CSV. Bloomberg, Refinitiv/LSEG and Haver Analytics terminals carry the series with identical numbers but different formatting and metadata conventions.
• Known gaps. Federal and state holidays falling within the reference week occasionally trigger state-level imputations. The 2020–2021 pandemic emergency programs (PUA, PEUC) ran in parallel to ICSA and inflated total claims; ICSA itself remained the regular state-program series throughout.

For analytical work, the convention is to monitor the 4-week moving average of ICSA rather than the latest weekly print, and to verify the most recent published date before drawing conclusions on direction.

Common Pitfalls When Using ICSA

The ICSA series is straightforward to download, but several recurring interpretation errors distort the signal it carries.

1. Reading single weekly prints as trend. Weekly volatility in ICSA driven by holidays, state-level processing delays and weather events frequently produces 20,000–40,000 swings that reverse the following week. The 4-week moving average is the analytical reference; isolated weekly readings are rarely informative on their own.
2. Confusing initial claims (ICSA) with continued claims (CCSA). ICSA measures new filings during the week — a flow. CCSA measures the stock of people still receiving benefits after the first week. The two series track different dynamics: ICSA inflects on layoff acceleration, CCSA on hiring difficulty. Misreading them as substitutes leads to wrong conclusions on the labor cycle.
3. Comparing nominal levels across decades without baseline adjustment. A 230,000 ICSA reading in 1990 and 2026 do not describe the same labor market: the US civilian labor force has grown from approximately 125 million to over 170 million, while the share of workers covered by state UI programs has shifted. Comparisons are more robust when expressed relative to the size of the covered labor force.
4. Treating the 400,000 threshold as universal. Pre-2008, the conventional rule of thumb was that ICSA above 400,000 signaled labor market weakness. This anchor reflects a different underlying labor force and a different unemployment insurance recipiency rate; structurally lower readings since 2014 mean the relevant benchmark today is the trajectory relative to the trailing 52-week low, not a fixed nominal level.

Historical Regimes

1967–1979 — Stagflation-era volatility. The ICSA series began publication during a period of large oil-shock-driven labor cycles. Claims swung between roughly 200,000 and 575,000, peaking during the 1973–1975 recession at levels that reflected the underlying volatility of manufacturing employment in that era.

1980–1982 — Volcker disinflation. As the Fed Funds rate reached 19% in mid-1981, ICSA spiked above 650,000 in October 1982, the highest non-pandemic reading on record at that point. The 4-week moving average remained above 600,000 for several months — the labor-market cost of the disinflation. See the initial claims weekly recession signal study for the full timing analysis.

1991–2000 — Long expansion and labor-market deepening. ICSA trended below 350,000 for most of the decade, reaching cyclical lows near 260,000 in 2000 before the tech bust. The narrow range despite a growing labor force reflected both a tighter labor market and the secular decline in manufacturing-sector layoff intensity.

2007–2010 — Global Financial Crisis. ICSA peaked at 665,000 for the week ending 28 March 2009. The 4-week moving average stayed above 600,000 for nine consecutive months, a duration unmatched outside the 2020 pandemic episode. The recovery was slow: claims did not return below 400,000 until early 2011.

2014–2019 — Multi-decade lows. ICSA fell below 220,000 consistently from 2017 onward, reaching a 49-year low of 193,000 in April 2019. This period coincided with one of the longest US expansions on record, alongside a falling unemployment rate trend.

2020 — Pandemic shock. Claims hit 6.137 million for the week ending 4 April 2020 — roughly ten times the previous all-time high. The series briefly lost its conventional informational content as state UI systems were overwhelmed and the federal Pandemic Unemployment Assistance program ran in parallel. The 4-week moving average exceeded 5 million for two months before normalizing.

2022–2026 — Post-pandemic normalization. Despite the Fed funds rate rising from 0% to a 5.25–5.50% range between March 2022 and July 2023, ICSA stabilized in a 200,000–260,000 band — a pattern consistent with labor hoarding rather than a typical late-cycle layoff acceleration. The Sahm Rule recession indicator nevertheless triggered briefly in 2024 on the basis of rising unemployment rate readings, illustrating that low ICSA and a rising UNRATE can coexist when labor-force expansion outpaces hiring.

Related Macroeconomic Datasets

ICSA is the highest-frequency input in a broader US labor-market data stack. The datasets below cover the complementary stock measures, activity indicators and recession-signal frameworks that ICSA naturally connects 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 Department of Labor — Employment and Training Administration (UI Weekly Claims Report)

Dataset Reference

Eco3min Research. (2026). US Initial Jobless Claims — ICSA (1967–2026).<br> https://eco3min.fr/en/us-initial-jobless-claims-dataset/