1. The headline rate vs. time spent unemployed
Show data table
2. Hiring and layoffs
Show data table
3. By sector
- 2019 average
- Last 12 months
Tech and biotech employment
Unemployment by industry of the person's last job (CPS, not seasonally adjusted, so shown as 12-month averages; people with no previous job are not assigned an industry). Hiring and layoffs from JOLTS, seasonally adjusted, 12-month averages. "Trade, transportation, utilities" combines two CPS categories weighted by labor force. Mining and agriculture are omitted: small CPS samples, and agriculture is not in JOLTS. BLS did not collect October 2025 household survey data, so averages that span it use 11 months.
Show data table
4. When state unemployment benefits run out
Shaded blue: federal programs paid extra weeks after state benefits ended (2008-2013 and 2020-2021), so in those windows people using up state benefits did not necessarily lose all benefits. Each state sets its own maximum duration, from 12 to 26 weeks in most states, so exhaustion rates are not directly comparable across states.
Show data table
5. What happens after benefits end
Show data table
Sources and methods
Why the headline rate can stay low
- Narrow definition. You count as unemployed (U-3) only if you had no work at all and searched in the past four weeks. Someone who stopped searching is "not in the labor force." Someone who did a few hours of paid work, including consulting or gig work, counts as employed.
- Low hiring, low layoffs. When few people are laid off, few enter unemployment, so the rate stays flat. When employers also hire slowly, the people who are unemployed stay unemployed longer. The rate measures how many people are unemployed, not how long they stay unemployed.
- Averages across very different job markets. The national rate combines every sector and occupation, so a downturn in a few fields barely moves it.
- Benefit records stop when benefits end. Unemployment benefit (claims) data covers only people currently receiving benefits. When benefits end, people drop out of those numbers whether or not they found work.
Data sources and latest month available
Definitions
- Exhaustion rate (DOL formula): people receiving their final state benefit payment over the last 12 months, divided by people receiving a first payment over the 12 months ending six months earlier. Computed here from the DOL ETA 5159 report. Computed values match DOL's published quarterly rates.
- People starting or using up benefits: monthly first payments and final payments under regular state UI, shown as a 12-month moving average to remove seasonal swings. The table shows raw monthly counts.
- U-6: U-3 plus people who want work but stopped looking ("marginally attached," including "discouraged workers") and people working part-time because they could not find full-time work, as a share of the labor force plus the marginally attached.
- Recessions (gray shading): NBER business-cycle dates.
Section 5 method
The Census Survey of Income and Program Participation (SIPP) re-interviews the same people every year, asking about each month of the prior year. That lets us follow someone from the month their benefits stopped to what they were doing afterward. SIPP does not record why benefits stopped. "5 or more months of benefits" is used as a stand-in for running out, since most states pay at most 26 weeks. Results are recomputed when Census releases a new year of SIPP data (usually in summer, covering the year before last).
Downloads
- labor.csv: national monthly series (wide format)
- benefits.csv: first payments, final payments, and exhaustion rate by state and month, 1971 to present
- sectors.csv: unemployment, hiring, and layoff rates by sector and month, 2000 to present
- industries.csv: tech and biotech payroll employment (thousands) and index, 2007 to present
- sipp.json: Section 5 estimates with sample sizes
Code and pipeline: github.com/yottachem/unemployment-gap. Data refreshes automatically each weekday.