CCSA: US Continued Jobless Claims Weekly Data from FRED (1967–2026)

The US Continued Jobless Claims series counts insured unemployment — people collecting state unemployment benefits beyond their first week — every week since 1967, from the Department of Labor. Where initial claims measure the flow into unemployment, continued claims measure the stock that stays: how quickly the laid-off find work again. The range spans 988,000 (1969) to 23.1 million (May 2020); at 1.81 million in mid-2026, the level sits above its 2019 average of 1.68 million — the gap that quantifies today’s slow-rehiring labor market.

Dataset: US Continued Jobless Claims (1967–2026) · Updated 2026-07-24

Source: US Employment and Training Administration · FRED series CCSA

Macro Takeaway

Dataset Overview

Dataset Variables

The CSV and Excel files contain the following columns.

A value of 1,814 means 1.81 million people were drawing state unemployment benefits that week.

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 CCSA:

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

Direct CSV Access — Eco3min Structured Dataset

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

print(f"Latest: {df['continued_claims'].iloc[-1]/1000:.2f} million on benefit rolls")
print(df.tail())

Using the Dataset in R

library(readr)

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

tail(df)
summary(df$continued_claims)

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

Methodology

Continued claims (insured unemployment) count persons who filed for benefits and certified continuing eligibility for at least a second week, under regular state programs. Data are administrative — a census of benefit rolls, not a sample — reported by state agencies and aggregated by the DOL Employment and Training Administration.

The published week refers to the reference week two weeks prior (one week behind initial claims). The Eco3min CSV serves the seasonally adjusted level converted to thousands.

This dataset is updated daily (Mon–Sat, 08:00 UTC) via automated pull from the FRED API; new observations appear with each source release.

Data Quality & Provider Notes

• Release latency. Nine days after the reference week — the fastest labor-market stock measure available.
• Revisions policy. The prior week revises routinely as late state reports arrive; seasonal factors update annually and can reshape recent months.
• Universe. Regular state programs only — federal pandemic-era programs (PUA, PEUC) were counted in separate series, so 2020–2021 totals here understate the true benefit rolls.
• Seasonal adjustment quirks. Holiday weeks and (post-2020) the additive-vs-multiplicative factor switch generate artifacts; multi-week averages are standard.
• Known gaps. None; continuous weekly since 1967.

Common Pitfalls When Using Continued Claims

1. Reading the level across decades without scaling. The labor force nearly doubled since 1967; 1.8 million insured unemployed today is a far smaller share than in 1975. The insured unemployment rate is the normalized companion.
2. Equating continued claims with unemployment. Coverage is ~30% of the unemployed; ineligible workers, exhaustees, and non-filers are invisible. Level gaps versus the household survey are structural, not errors.
3. Ignoring benefit exhaustion. In long downturns the count falls as benefits run out, not as people find jobs — the 2010–2011 ‘improvement’ partly measured exhaustion.
4. Treating single weeks as signal. State reporting noise and seasonal artifacts make the 4-week average the minimum readable unit.

Historical Regimes

1967–1982 — The high-volatility era. Recessions drove the rolls from under 1 million (the 1969 record low of 988k) to repeated peaks above 4 million (1975, 1982) in a labor force half today’s size — proportionally the deepest insured-unemployment cycles in the record.

1983–2007 — The moderation. Two long expansions kept the rolls mostly between 2 and 3.5 million, with muted recession peaks in 1991 and 2002.

2008–2009 — The GFC stock. Continued claims peaked at 6.64 million (May 2009), then declined only slowly — the era when exhaustion and emergency extensions made the regular-program count an undercount of long-term unemployment.

2010–2019 — The decade-long drain. A near-monotonic fall to a 1.68 million average in 2019 — the lowest insured unemployment relative to the labor force in the series’ history.

2020–2021 — Off the chart. 23.1 million (May 2020) in regular programs alone, with millions more in federal pandemic programs — a spike 3.5 times the GFC peak, unwound within two years.

2022–2026 — Low firing, slow hiring. The rolls bottomed near 1.3 million in 2022, then drifted to 1.81 million by mid-2026 while initial claims stayed low — rising duration, not rising layoffs.

Related Macroeconomic Datasets

Continued claims are the stock between the layoff flow and the rehiring flow; the adjacent series are below.

Macroeconomic Dataset Hub

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

Sources

• US Department of Labor, Employment and Training Administration — Unemployment Insurance Weekly Claims
• Federal Reserve Bank of St. Louis — FRED series CCSA

Dataset Reference