Initial Unemployment Claims
Initial claims count the people who filed for unemployment benefits for the first time in a week. It is the highest-frequency official measure of layoffs the United States publishes, arriving every Thursday with a one-week lag.
History
What it captures, and what it misses
A claim is filed after a job ends, so the series measures separations rather than hiring. It misses people who never file, independent contractors, and employees who are laid off but receive severance and delay filing. It also counts fraudulent claims, which have distorted the level in some periods.
Why the four-week average is the number
Weekly data is seasonal and noisy: a holiday, a storm or a plant shutdown can move the series by tens of thousands. The four-week moving average smooths that without losing the weekly frequency, and it is what the financial press quotes when it says claims are trending.
Watch the trend, not the level. A rise from 210,000 to 240,000 over six weeks is a more meaningful signal than a one-week spike to 260,000 that reverses.
Reading it against continuing claims
Continuing claims count people still receiving benefits, which measures re-employment rather than layoffs. When initial claims rise and continuing claims rise with them, hiring has stalled. When initial claims rise but continuing claims fall, laid-off workers are finding new jobs quickly.
How this reading is calculated
We pull ICSA from FRED, apply the transform shown above, and keep every observation the source publishes. Missing dates are dropped rather than filled, so holidays leave gaps instead of fake zeroes. The percentile is the share of observations since 1967 that sit below today's value.
The file is rebuilt daily. Derived series are computed by us from public-domain inputs and the formula is stated on this page, so anyone can reproduce the number from the original release.
How to read the percentile
The percentile compares today's reading with the indicator's own history, not with other indicators. A 20th percentile on this page means the current reading is lower than four fifths of all observations since 1967; it says nothing about whether that is good or bad for markets.
Extremes are the useful part. A reading above the 90th or below the 10th percentile is unusual enough to be worth explaining, and unusual readings tend to mean-revert. The middle of the distribution is mostly noise.
What is free and what is member-only
The current reading, the percentile and the full historical record are free. The most recent weeks of data are reserved for members, because that is the part that tells you what the indicator is doing right now rather than what it did years ago.
Members see every observation through the latest release, the same chart without the mask, and the same numbers on every indicator page. Nothing on this site is intraday.
Distribution since 1967
Each row is the share of historical observations at or below that level. A reading in the 90th percentile is higher than nine tenths of the record.
Important
Not investment advice. This page is published for information and education only. It is not a recommendation, solicitation, or offer to buy or sell any security. Market data may be delayed, incomplete, or revised. Indicators describe the past; they do not forecast returns. Past performance does not guarantee future results. Do your own research or consult a licensed financial professional before acting on anything you read here.
Questions
What counts as a high claims reading?
Context matters, but roughly 300,000 a week has historically been the line between a healthy and a softening labour market. Above 400,000 usually signals outright deterioration.
Why is the series seasonally adjusted?
Layoffs follow a strong annual pattern, with spikes after holidays and at the end of school terms. Seasonal adjustment removes that pattern so week-to-week changes reflect the cycle.
When is it released?
Every Thursday at 8:30 a.m. U.S. Eastern time, covering the week that ended the previous Saturday. It is the first labour-market data of each week.
Why is the recent data locked?
The current reading and the full history are free; the latest weeks are member-only, which is what funds the weekly refresh.