The UK Unemployment Rate Is Not a Monthly Number
The UK unemployment rate is 4.6%. The Office for National Statistics published that figure on November 17, and it is not a rate for November, or for October, or for September alone.
It is one estimate for July, August and September together. ONS labels it July to September 2026, and the clearest way to say it out loud is "the three months to September."
That sounds like a technicality. It isn't. A headline rate that comes out every month is easy to hear as a monthly reading, the way a US jobs report is heard, and a monthly reading is the one thing it can't give you.
What the UK unemployment rate actually covers
The rate comes from the Labour Force Survey, which ONS runs on households. It is the share of people who are working or looking for work who are unemployed. ONS is explicit that it is not a share of the whole population.
Each headline covers a three-month window, and ONS publishes a new one every month. So every release moves the window forward by one month and keeps the other two.
The timing matters too. By November 17, the newest month in the window, September, had been over for about seven weeks. The oldest, July, had started more than four months earlier. A rate described as the latest one was already a description of the third quarter.
Last month's figure shares two of the three months
The release before this one, on October 20, put unemployment at 4.7% for June to August.
Line the two windows up:
- October 20 release: June, July, August. 4.7%.
- November 17 release: July, August, September. 4.6%.
July and August are in both. The only month in November's figure that wasn't in October's is September, and the only one that dropped out is June.
Subtract one headline from the other and you get a fall of 0.1 points, which reads like unemployment dropping a tenth in a month. ONS doesn't report it that way. Its own comparisons are on the quarter and on the year, and on both, the rate was down 0.3 points.
So the 0.1 is my arithmetic on two figures that share two-thirds of their calendar. The 0.3 is ONS's. When they differ, the 0.3 is the one to repeat.
Repeat it with the warning ONS attached this time. The bulletin says the quarter's fall in unemployment partly reflects "particularly large differences in employment status between the groups entering and leaving the survey." The respondents finishing their last quarter in the survey had lower employment and higher unemployment than the group that replaced them. Part of the 0.3 is who was asked, not what changed.
You can see the method most clearly in vacancies. ONS put them at 690,000 for August to October, down 22,000 on May to July. Three months against the three months before them, with no month counted twice. That is the comparison to make with any figure in this release. (What the vacancy numbers say about UK hiring is its own story, and I wrote it up in September: The UK Hasn't Started Shedding Jobs. It Has Stopped Advertising Them.)
One release, several calendars
The unemployment rate isn't the only figure in the release, and the others don't share its window.
| Figure | Period it covers | Where it comes from |
|---|---|---|
| Unemployment, employment and inactivity rates | July to September | Labour Force Survey, households |
| Vacancies | August to October | Vacancy Survey, businesses |
| Payrolled employees | September; October as an early estimate | HMRC tax records |
| Claimant Count | October | Benefit claims |
The payroll count is the one most worth a second look. It comes from HMRC's tax records, one month at a time, and ONS says the newest month is provisional and likely to be revised when more data come in.
What ONS does with it is the instructive part. Before it sets the payroll count beside the survey, it lines the two up. For July to September, which ONS calls "the period comparable with our Labour Force Survey (LFS) estimates," payrolled employees fell 90,000 over the year and 32,000 over the quarter.
If the people who produce the figures line the windows up before comparing them, so should anyone quoting them.
The Claimant Count needs a warning of its own. It covers a single month and counts people receiving a benefit mainly because they are unemployed, and ONS has published a separate article on how it differs by definition from the survey's unemployment. In this release the two moved in opposite directions: the Claimant Count for October rose on the month and on the year, while the survey's rate for July to September fell. That isn't a contradiction. They measure different things over different months, and a sentence that blends them into "UK unemployment" is describing neither.
Why it can't sit next to a US rate
The US unemployment rate from the Bureau of Labor Statistics is published for a single month, and its household survey asks about a single week in it: in BLS's words, "generally the calendar week that contains the 12th day of the month." A UK headline covers three months. Put one beside the other and you are comparing a week with a quarter before you have even checked whether the two surveys count unemployment the same way.
I'm not going to print the two side by side here. If you see a chart that does, ask which window each bar covers.
How to read the next one
ONS publishes the next figures on December 15. Four habits make them more useful.
Name the window. "4.6% in the three months to September," never "4.6% in September." If a report doesn't say which months, it hasn't told you what it measured.
Use the changes ONS prints. On the quarter and on the year, rather than subtracting last month's headline from this month's. The first is a comparison. The second mostly compares a quarter with itself.
Don't move on one quarter. ONS's own advice on the survey is to "focus on long term movements in the data," and this release shows why: part of the quarter's fall came from who rotated through the survey. For a long-term movement, compare the same window a year apart, or watch the quarterly change over several quarters.
Remember how old it is. The newest month in the headline ended weeks before you heard about it. If you are job hunting in the UK, the replies you got last week are more current than anything in the release. Use the national rate for context and your own numbers for decisions.
A quarter is not a month. Read the window first, then the number.
Sources: Office for National Statistics, Labour market overview, UK: November 2026 (released November 17, 2026; next release December 15, 2026): overview, headline rates, vacancies, payrolled employees, Claimant Count, glossary and current quality considerations. Labour market overview, UK: October 2026 (released October 20, 2026), for the June to August unemployment rate. ONS's methodology guide to its labor market statistics, for how the Labour Force Survey is run and how often its estimates are published. Bureau of Labor Statistics, Employment Situation Technical Note (last modified November 6, 2026), for the US reference week.
Related posts
Veteran Unemployment Was 3.5% Last Year. For Women Veterans Aged 25 to 34, It Was 8.1%.
The government's annual report on veterans shows them out-performing nonveterans in 2025. Underneath that average are the groups where the search is hardest, and a disability figure that looks better than it is.
Is It Too Late to Job-Search in Q4? You Have This Backwards.
Everyone tells you hiring stops in December. What slows is the process, not the need, and waiting for January puts you in the queue with everyone else who took the advice.
Job Openings by Industry: Why Tech, Healthcare and Education Are Not the Categories You Think
The August JOLTS release breaks job openings down by industry. Three of the labels people reach for most do not contain what the name suggests, and reading them literally will send you looking in the wrong place.
Ready to take control of your job search?
Free for 4 weeks. Cancel anytime.
Get Started FreeGet Job Market Intelligence by email
I write about the job market from primary sources, and I send two emails a month with what published.
Two emails a month at most. Unsubscribe in one click from any email. I do not sell or share your address.
Already get this on LinkedIn? Then you have it. This is the same writing by email, for people who are not there.