The labour market

Reviewed 14 October 2026

Okenitemode explains how UK indicators are built and read. Nothing on this page is a forecast, a recommendation, or a substitute for the official statistical bulletin. Worked numbers below are invented so the arithmetic is visible. They are not Office for National Statistics figures.

Three states, and only three

For people aged 16 and over, the labour-market framework used in the Labour Force Survey sorts everyone into employed, unemployed, or economically inactive.

• Employed. In work, including employees, the self-employed, and people temporarily away from a job.
• Unemployed. Without work, wanting work, actively seeking it in the last four weeks, and available to start. This is the International Labour Organization definition. It is a survey concept, not a benefits concept.
• Inactive. Everyone else: students who are not seeking work, people who are unwell, carers, and people who have retired. Inactive is not a polite word for unemployed.

The unemployment rate divides the unemployed by the labour force, and the labour force is employed plus unemployed. The employment rate and the inactivity rate divide by the whole population in the age group, not by the labour force. The denominators differ. That is why the three rates do not add to 100 in the way a casual reader expects: unemployment uses a smaller base.

A town of one thousand adults

Picture a fictional town with 1,000 people aged 16 and over. 620 are employed, 40 are unemployed, and 340 are inactive.

The unemployment rate is 40 divided by 660, which is 6.1 percent. The employment rate is 620 divided by 1,000, which is 62 percent. The inactivity rate is 34 percent. If 10 of the inactive start looking for work and do not find it immediately, unemployment rises even though nobody has been sacked. The rate moved because the labour force changed, not because payrolls shrank.

The claimant count is a different series

The claimant count measures people claiming unemployment-related benefits. Rules on who can claim change. People can be unemployed on the survey definition and not claim, and some claimants are not unemployed on the survey definition. When a minister and a critic quote different “unemployment” numbers, check whether one of them has switched to the claimant count.

Pay, vacancies, and the early reads

Average weekly earnings come in a regular-pay measure and a total-pay measure that includes bonuses. Nominal earnings are the cash. Real earnings deflate that cash by a price index. A 4 percent pay rise beside 6 percent inflation is a real pay cut. A 4 percent pay rise beside 2 percent inflation is a real pay rise. The headline that says only “wages are up” has not finished the sentence.

Vacancies, from the ONS vacancy survey, are posts for which employers are actively seeking people from outside the firm. They are not the same as online advert counts, though adverts are a useful faster hint. A fall in vacancies can mean firms are cautious, or that posts have been filled. Pair it with unemployment before you decide which story you are in.

PAYE real-time information, drawn from tax records, gives an earlier look at employees and their pay. Workforce jobs is another employment picture, counted from the employer side. They will not match the Labour Force Survey person-for-person, because one counts jobs and one counts people, and the sources differ.

Treat the survey with the caution the ONS already prints

Response rates to the Labour Force Survey fell sharply in the early 2020s. The ONS has reweighted results, flagged higher uncertainty, and been building a transformed survey. Until you have read the latest quality note on the release itself, do not treat a small move in the unemployment rate as a precise fact. Cross-check the direction with PAYE employees, workforce jobs, vacancies, and earnings. A story that survives three sources is sturdier than a story that lives in one noisy tenth of a percentage point.

Most labour-market headlines use a three-month average compared with the previous three months, or with a year earlier, precisely because a single month of survey data jumps. If a commentary uses a one-month change, ask why.