UK job vacancies averaged an estimated 691,000 in July to September 2026, the first reading below 700,000 since the three months ending April 2021. Yet the unemployment rate edged down in the ONS household survey. Someone looking for work could reasonably ask how both things happened. Part of the answer lies in who gets counted, and part in the dates printed above the figures.
The Office for National Statistics published its UK labour market overview on 20 October 2026. It was the latest monthly overview available on 12 November. Its vacancy estimate runs through September, while its unemployment and earnings figures cover June to August. The next release is due on 17 November. In the meantime, the October tables offer a way to distinguish employers advertising less work from people actually losing their jobs.
Where the vacancies disappeared
Employers reported an estimated 21,000 fewer openings than in April to June, a drop of 2.9%. Across the year the decline was 37,000. Fourteen of the 18 industries recorded quarterly falls. Construction accounted for the largest quarterly decline, with roughly 7,000 fewer vacancies. That is about a third of the overall net fall, even though construction is only one of the industries in the survey.
Size mattered too. Firms employing one to nine people recorded 21,000 fewer vacancies between the two quarters. Some smaller employers told the ONS that rising labour costs could be holding back recruitment. The industry and size breakdowns overlap. A builder employing six people belongs in both, so adding those two declines would count some of the same openings twice. The vacancies bulletin gives both breakdowns.
A vacancy leaves the count when somebody fills it. It can also disappear when an employer withdraws the offer. Imagine a building firm searching for a site supervisor. Hiring a supervisor and postponing the project would each reduce the vacancy count by one. Only the first outcome puts somebody to work. The survey measures openings still available, and its headline is an average over three months.
For someone choosing where to look, the industry ranking depends on the comparison. Over a full year, professional, scientific and technical activities recorded the largest drop in openings, at 9,000. Construction headed the quarterly ranking. A person scanning the results for their own occupation also needs to check who employs them. A bookkeeper working for a surveyor and one working for a supermarket appear in different industries despite doing similar work.
What a small fall in unemployment can tell us
For June to August, the estimated UK unemployment rate was 4.7%, down 0.1 percentage points from March to May. That rate covers economically active people aged 16 and over. The employment rate, by contrast, covers the whole population aged 16 to 64. It stood at 75.3%, up 0.2 percentage points. Because the two rates use different populations, subtracting the unemployment rate from 100 cannot give the employment rate.
Consider two people who finish temporary contracts on the same Friday. One starts looking for work straight away and could take a job within a fortnight. The other begins a full-time course and is unavailable for work. Both have left a job. The first meets the survey definition of unemployment, while the second falls into economic inactivity. Likewise, someone who stops looking to care for a relative can leave unemployment without taking a job.
There is also sampling uncertainty. The employment bulletin puts sampling variability around the quarterly change in the unemployment rate at plus or minus 0.3 percentage points. That is larger than the estimated fall. The range includes a modest rise as well as a fall. Describing the survey as proof that finding work became easier would give one decimal place more authority than the interviews support.
A payroll figure that changed direction
Tax records offer another test. Comparing June to August with March to May, the estimated number of UK payrolled employees fell by 37,000. The household survey includes people who work for themselves, whereas payroll records cover employees. Someone moving from a salaried post into self-employment can therefore leave a payroll while continuing to count as employed in a household interview.
The July to August payroll comparison illustrates what happens as more records arrive. September’s release had put the monthly fall at 26,000. By October, additional submissions from employers turned that into a rise of 1,000. The payroll bulletin says the first estimate typically uses about 85% of the information. In the next edition, coverage usually reaches between 98% and 99%. That difference matters when the movement is small.
For September itself, the early reading put 30.3 million people on payrolls, up 13,000 from August but down 66,000 from a year earlier. Keep both comparisons attached to the provisional label when using those numbers.
What employers paid the people already working
For employees in Great Britain, average weekly earnings excluding bonuses were 3.7% higher in June to August than a year earlier. Adjusted for inflation using the index that includes owner occupiers’ housing costs, regular earnings rose by 0.6%. These figures cover England, Scotland and Wales. The employment and vacancy figures cited above also include Northern Ireland.
An individual comparing an offer with an existing salary needs a closer match. Average weekly earnings can change when the mix of jobs changes. If a business with ten equally paid full-time staff adds two people working half the hours at the same hourly rate, average weekly pay falls. Every existing worker can still take home exactly the same wage. The earnings bulletin explains how workforce composition affects the measure.
Reading the figures from a local jobs board
Suppose a qualified electrician finds twelve apparently suitable adverts within commuting distance. Four of the adverts cover just two agency assignments, leaving ten distinct openings. Three require overnight travel that the applicant cannot manage. Seven remain. This exercise measures something the national figures cannot supply directly, the number of jobs this person could actually take. The result also depends on looking beyond construction firms, since an electrician might work for a factory or a hospital.
The ONS calculated 2.4 unemployed people per vacancy for June to August, down from 2.5 in March to May. Applying that ratio to the seven openings would be a mistake. It compares national totals for a common period. People already working can apply, some vacancies require particular licences, and a candidate can apply for several jobs. An employer interviewing for one opening could see many more candidates or considerably fewer.
An applicant can repeat the local search a fortnight later using the same radius, hours and job titles, recording which vacancies have closed and which employers have appeared. If the seven feasible jobs have become four, checking their closing dates and whether the roles were filled gives that change some meaning. For a hiring manager, comparing qualified applications and accepted offers for the same role across successive rounds answers a different question about competition for staff.
The November release will bring a fresh set of overlapping periods and a revised September payroll estimate. The most revealing comparison will pair the new figure with the revised history printed beside it. August’s reversal shows exactly how much difference an extra month of employer returns can make.