What Does the Unemployment Rate Tell Us? | Economics for the Curious

Start Here: Who Counts as Unemployed?

Five people, none of them working full time. Before reading on, decide which of them the government counts as unemployed:

Who Counts as Unemployed?: Person
Person	Counted as	Why
Maria was laid off in August and has sent out applications every week since.	Unemployed	She has no job, is available to work, and has looked in the past four weeks.
DeShawn is a full-time college student who is not looking for work.	Not in the labor force	He is not working and not looking, so he is neither employed nor unemployed.
Linda, 67, retired from teaching last spring.	Not in the labor force	Retirees are not looking for work.
Tom works 20 hours a week at a warehouse but wants full-time work.	Employed	Any work for pay in the survey week counts as employed. He shows up in a broader measure instead.
Ray looked for months, found nothing, and stopped looking in the spring.	Not in the labor force	He has not looked in the past four weeks. If he still wants a job, he is a discouraged worker.

WATCH Defining the Unemployment Rate (Marginal Revolution University)

Define It: Three Tests

1 Without a job, available, and looking

The Bureau of Labor Statistics (BLS) uses a short definition: “People are classified as unemployed if they do not have a job, have actively looked for work in the prior 4 weeks, and are currently available for work.” People on temporary layoff who expect to be called back also count, even if they are not looking.

Everyone 16 and older who is not in the military or living in an institution, such as a prison or nursing home, falls into one of three groups:

Employed. Did any work at all for pay or profit in the survey week, or had a job but was out sick, on vacation, or on leave.
Unemployed. Passed all three tests: no job, available, and looking.
Not in the labor force. Everyone else: students, retirees, people caring for family, people with disabilities that keep them from working, and people who have given up looking.

The labor force is the employed plus the unemployed. The unemployment rate is the share of the labor force that is unemployed. It is not the share of all adults without a job.

2 Not a count of benefit claims

Many people assume the number comes from unemployment insurance claims. It doesn’t. The BLS explains why: “Because unemployment insurance records relate only to people who have applied for such benefits, and since it is impractical to count every unemployed person each month, the government conducts a monthly survey.” Many jobless people never qualify for benefits, some never apply, and others are still looking after their benefits run out. As the BLS puts it, claims data “cannot be used as a source for complete information on the number of unemployed.”

Measure It: Two Surveys, One Report

3 The household survey and the payroll survey

On the first Friday of most months, the BLS releases the Employment Situation report, often called the jobs report. It draws on two separate surveys that answer different questions:

Measure It: Two Surveys, One Report
	Household survey	Payroll survey
Who is asked	About 60,000 households, roughly 110,000 people	About 119,000 businesses and government agencies at about 622,000 worksites
What it measures	Whether each person is employed, unemployed, or not in the labor force	How many jobs are on payrolls, plus hours and pay
Headline number	The unemployment rate	The monthly change in jobs
October 2026	4.1 percent unemployed	165,000 jobs added

The two can disagree in a given month. The household survey counts people, so someone with two jobs counts once. The payroll survey counts jobs, so the same person counts twice. The household survey also includes farm workers and many self-employed people, while the payroll survey does not.

4 The October 2026 report

In October, the unemployment rate “changed little at 4.1 percent,” and it has stayed between 4.1 and 4.3 percent since March. Payrolls rose by 165,000, well above the average monthly gain of 48,000 over the prior 12 months. Gains came mostly in professional and business services, health care, and transportation and warehousing. Of the 7.1 million unemployed, 2.0 million, or 27.6 percent, had been looking for 27 weeks or longer.

5 Noise and revisions

Because both numbers come from samples, they carry a margin of error. The BLS says a monthly change in the unemployment rate has to be larger than about 0.3 percentage point before it is statistically significant. For payroll jobs, the margin is about 122,000. A report of “50,000 jobs added” could mean anything from 72,000 jobs lost to 172,000 gained.

Payroll numbers are also revised as late reports come in. In October’s report, the August gain was revised up from 133,000 to 186,000, and September’s from 29,000 to 60,000. Once a year, the BLS checks its payroll estimates against tax records that cover nearly every employer. Over the past ten years, those annual benchmark revisions averaged 0.2 percent of total employment. Revisions are not evidence that the numbers are made up. They are the system correcting itself in public.

6 Six measures, not one

The official rate is called U-3. The BLS publishes five other measures that run from narrow to broad:

Measure It: Two Surveys, One Report: Measure
Measure	What it counts	October 2026
U-1	Unemployed 15 weeks or longer	1.9%
U-2	Job losers and people who finished temporary jobs	1.9%
U-3	All unemployed (the official rate)	4.1%
U-4	U-3 plus discouraged workers	4.4%
U-5	U-4 plus all other people marginally attached to the labor force	5.1%
U-6	U-5 plus people working part time who want full-time work	7.9%

Back to the opening: Ray, who stopped looking, is a discouraged worker and shows up in U-4. Tom, who works part time but wants full time, is one of 4.8 million people working part time for economic reasons and shows up in U-6. All six measures tend to rise and fall together. U-6 is not the “real” rate that the government hides. It is published in the same report, on the same day, and it answers a different question.

WATCH Is Unemployment Undercounted? and What Is a Discouraged Worker? (Marginal Revolution University)

7 Who is unemployed?

Unemployment is not spread evenly. October’s rates for several groups:

Measure It: Two Surveys, One Report: Group
Group	Rate	Compared with 4.1% overall
Teenagers (16 to 19)	13.9%	More than three times as high
Black workers	6.6%	Higher
Hispanic workers	4.7%	Slightly higher
Adult men	3.9%	Slightly lower
White workers	3.7%	Lower
Adult women	3.6%	Lower
Asian workers	2.8%	Lower

Education matters, too. Among adults 25 and older, the October rate was 5.8 percent for those without a high school diploma and 2.6 percent for those with a bachelor’s degree or more. High school graduates (3.8 percent) and people with some college or an associate degree (3.9 percent) traded places from September, when the rates were 4.4 and 3.6 percent. A single month is noisy. Over many years the pattern is clear: more education, lower unemployment.

Since 1972, the Black unemployment rate has averaged about twice the white rate. Federal Reserve economists Tomaz Cajner, Tyler Radler, David Ratner, and Ivan Vidangos found that “observable characteristics can explain very little of this differential,” which is driven mainly by “a comparatively higher risk of job loss.” They also found that Black unemployment is “more cyclical”: it rises more in recessions and falls more in recoveries. Another Fed study, by Isabel Cairó and Avi Lipton, models employer discrimination as one cause of the gap. The Hispanic-white gap is smaller, and the first study found it is “largely explained by lower educational attainment of (mostly foreign-born) Hispanics.”

8 How many jobs is enough?

A gain of 165,000 jobs sounds good, but good compared with what? Economists use the idea of breakeven job growth: the number of new jobs needed each month just to keep the unemployment rate steady as the population grows. For years it was well above 100,000. Federal Reserve economists Seth Murray and Ivan Vidangos estimate that in 2026 it could be “nearly zero, requiring less than 10,000 new jobs per month.” The main reason is that “a precipitous drop in net international migration may cause population growth to be the weakest it has been since 1951,” on top of an aging population. That helps explain how the unemployment rate stayed near 4 percent even as monthly job gains averaged only 48,000.

Explain It: Why People Are Unemployed

9 Three kinds of unemployment

Frictional. People between jobs or just entering the job market. Matching workers to jobs takes time, and some of this is healthy: a worker who quits to find a better fit is frictionally unemployed.
Structural. A mismatch between the skills or locations of workers and the jobs available. When a plant closes and the new jobs need different skills or are in another state, unemployment can last.
Cyclical. Unemployment caused by a recession, when businesses cut back because customers are spending less. This is the kind that rises and falls with the business cycle.

WATCH Frictional Unemployment, Structural Unemployment and Cyclical Unemployment (Marginal Revolution University)

10 The natural rate

Even in a strong economy, unemployment never reaches zero. People are always moving, changing careers, and entering the job market. Economists call the rate that holds when the economy is neither booming nor in recession the natural rate of unemployment. It is made up of frictional and structural unemployment. The Congressional Budget Office estimates it at about 4.4 percent for 2026. At 4.1 percent, the current rate is close to that estimate. The natural rate is an estimate, not a fixed number. It can shift as the workforce ages, as job-search technology improves, or as rules change.

WATCH What Is the Natural Rate of Unemployment? (Marginal Revolution University)

11 Openings, quits, and the Beveridge curve

A second BLS survey, called JOLTS, counts job openings, hires, quits, and layoffs. In September 2026, employers had 7.5 million job openings, close to the 7.1 million people unemployed: roughly one opening for each person looking. Workers quit 3.4 million jobs, and 1.6 million were laid off or fired. Quits are a useful signal. People quit more often when they are confident they can find something better.

Plotting the job openings rate against the unemployment rate traces out what economists call the Beveridge curve. When openings are plentiful, unemployment tends to be low. When the curve shifts outward, it means there are more openings at every level of unemployment, which suggests that workers and jobs are having a harder time finding each other.

12 An early warning: the Sahm rule

The economist Claudia Sahm proposed a simple early warning. When the three-month average of the unemployment rate rises half a percentage point or more above its lowest three-month average of the past year, a recession has likely begun. The rule sums up past patterns. It is not a law of nature. In September 2026, the Sahm indicator was 0.00, far from the 0.50 warning level.

13 Unemployment insurance

Unemployment insurance (UI) is a joint federal and state program, paid for mostly by taxes on employers. The federal tax is 6.0 percent on the first $7,000 of each worker’s wages, but employers who pay their state UI taxes on time get a credit of up to 5.4 percent. That leaves a federal tax of generally 0.6 percent, at most $42 per worker per year. State taxes use experience rating: an employer whose former workers collect more benefits pays a higher tax rate. In effect, layoffs cost the employer more.

Each state sets its own benefits. In South Carolina, weekly benefits range from $42 to $350 before taxes, for up to 20 weeks. Thirty-three states offer up to 26 weeks, while Florida and North Carolina offer 12. Not everyone who is unemployed gets benefits. In 2025, 26.5 percent of unemployed people nationwide received UI, and 14.4 percent in South Carolina, about one in seven.

UI does two jobs. For a family, it replaces part of a lost paycheck. For the economy, it is an automatic stabilizer: when a recession hits, benefit payments rise automatically, without any new law, which helps keep spending from falling as far.

Economists study whether UI also makes people search less hard. A Richmond Fed summary of the research reports that a 10 percent increase in benefits is associated with spells of unemployment that last 4 to 8 percent longer. Raj Chetty found that about 60 percent of that effect comes from what he calls a “liquidity effect.” Workers without savings use benefits to search for a better match instead of taking the first job offered, rather than simply choosing not to work.

Fair Challenges

Decide what you think of each challenge first, then open the response.

“The real unemployment rate is way higher than they say.”

It depends on the question. The official rate (4.1 percent) counts people who want work and are looking. The broadest measure, U-6 (7.9 percent), adds discouraged workers and part-timers who want full-time work. Both come out in the same report. Neither is hidden, and each answers a different question. Some people who are not in the labor force at all, such as men in their prime working years who have stopped looking, are a separate concern covered in the reading on labor force participation.

“The jobs numbers are made up for political reasons.”

The BLS publishes its methods, sample sizes, margins of error, and every revision. Revisions get headlines, but over the past ten years the annual benchmark revisions have averaged 0.2 percent of employment. Private payroll data and state UI records give outside checks. A fair concern is that fewer households answer the survey. The BLS reports that response rates have fallen from the low 90 percent range to the upper 60 percent range over the past 15 years, which makes the numbers noisier. That is a real problem, but it is not the same as fraud.

“Unemployment benefits just pay people not to work.”

The evidence is mixed. Benefits can lengthen job searches, and research summarized by the Richmond Fed links a 10 percent rise in benefits to unemployment spells 4 to 8 percent longer. Much of that effect comes from workers without savings who can afford to search for a better match. Benefits are also temporary, require an active job search in most states, and reach only about one in four unemployed people nationally. How generous they should be is a judgment about trade-offs.

“Low unemployment means everyone who wants a job has one.”

Not quite. A low rate can hide people who have stopped looking, people stuck in part-time work, and large gaps between groups and places. The teen rate was 13.9 percent in October, and the Black rate 6.6 percent.

Economists also disagree among themselves. Three live debates:

Can openings fall without a jump in unemployment? In 2022, Federal Reserve economists Andrew Figura and Chris Waller argued that a “soft landing” was “a plausible outcome,” with job openings falling while unemployment rose by about a percentage point or less. Olivier Blanchard, Alex Domash, and Lawrence Summers replied that vacancies were “very unlikely to normalize without a major increase in unemployment.” Since then, the job openings rate has fallen from 7.5 percent in March 2022 to 4.5 percent, while unemployment rose from a low of 3.4 percent to 4.1 percent. Economists still debate how much of the outcome reflects the labor market and how much reflects other forces, such as supply chains healing.
How much did expanded benefits slow the recovery? Peter Ganong and coauthors found that the pandemic expansion of UI, “the largest increase in unemployment benefits in US history,” had “large spending impacts and small job-finding impacts.” Andreas Hornstein, Marios Karabarbounis, and coauthors at the Richmond Fed found larger effects once they controlled for the spending boost, estimating that the programs “held back the aggregate employment recovery by 3.4 percentage points” in 2020. Ryan Bourne and Erin Partin of the Cato Institute argued that “extending UI benefits right through this recovery will mean fewer jobs.”
Will AI cause mass unemployment? In a March 2026 review for Brookings, Jed Kolko found that the evidence on how AI is affecting jobs today is “inconclusive” and that claims of harm to particular groups of workers are “premature.” A study of the United States and Europe by Alexander Bick and coauthors found “no clear evidence” so far that AI adoption is linked to changes in employment by industry. Tyler Cowen, who is optimistic about AI, says, “I do not think we are headed for anything like mass unemployment.” Others expect larger effects as the technology spreads. The data so far are early.

Notice what evidence can settle and what it can’t. How many people are looking for work, how much benefits change search behavior, and whether AI is reducing jobs are questions evidence can address, even when studies disagree. How generous benefits should be, and how to balance help for people out of work against the incentive to search, are questions of values.

What Economics Adds

LISTEN AI, Employment, and Education (with Tyler Cowen) on EconTalk, hosted by Russ Roberts; and How Can We Improve Noisy Labor Market Data? with Ayşegül Şahin on the Brookings Podcast on Economic Activity, both with transcripts

The Normative Question

Economics can show who is counted as unemployed, how reliable the numbers are, and how benefits change behavior. It can’t decide how much support people out of work should get, how long it should last, or how to weigh security for workers against the cost to employers and the incentive to keep searching. Those are judgments about values, and reasonable people weigh them differently.

Key Terms in Plain Words

Labor force: Everyone 16 and older who is employed or unemployed.
Unemployed: Without a job, available to work, and actively looked in the past four weeks.
Unemployment rate: The unemployed as a share of the labor force.
Not in the labor force: Neither working nor looking: students, retirees, caregivers, and others.
Discouraged worker: Wants a job but stopped looking because they believe none is available.
Part time for economic reasons: Working part time but wants and is available for full-time work.
U-6: The broadest BLS measure: unemployed, marginally attached, and involuntary part-timers.
Breakeven job growth: New jobs needed each month to keep the unemployment rate steady.
Frictional unemployment: Time spent between jobs or searching for a first job.
Structural unemployment: A mismatch between workers’ skills or locations and available jobs.
Cyclical unemployment: Unemployment caused by a recession.
Natural rate of unemployment: The rate when the economy is neither booming nor in recession.
Beveridge curve: The relationship between job openings and unemployment.
Automatic stabilizer: Spending or taxes that rise or fall with the economy without new laws, such as UI.