One country, one winter, two answers
In February 2021, the UK's Office for National Statistics (ONS) published two official numbers about unemployment. One said that about 1.74 million people were unemployed. The other said about 2.6 million. Both came from the same statistics office, about the same country, in the same winter.
So how many people were unemployed? It sounds simple: count the people without a job. But try these. Someone works three hours a week and wants thirty. Someone gave up looking for work months ago, because they believe there are no jobs. Someone's workplace is closed, but they are still paid and will go back. Is each of them unemployed?
The answer depends on the definition, and on the method used to count. This lecture shows the two main ways of counting, using the UK as the example, and why they disagreed in 2021. The last lecture promised to show how countries count the unemployed. Here it is.
Three groups, not two
The last lecture gave the definition. A person is unemployed if they have no job, are available to start work, and are actively looking for work. Most countries use the precise version agreed by the International Labour Organization (ILO), an agency of the United Nations. Under the ILO definition, a person aged 16 or over is unemployed if all three of these are true:
- 1. They did no paid work in the week in question. Even one hour of paid work counts as being employed.
- 2. They have actively looked for work in the past four weeks, for example by applying for jobs.
- 3. They are available to start work within the next two weeks.
There is one more case. A person who is out of work, and has found a job that starts within the next two weeks, also counts as unemployed. They are about to fill a place in the labour force.
This sorts every adult into three groups, not two. Employed people have a job, including people temporarily away from it, such as on holiday. Unemployed people meet the three tests. Everyone else is economically inactive: not working and not looking. Students, retired people, full-time carers and people who are long-term sick are usually in this group.
The labour force is the employed plus the unemployed. The unemployment rate is a share of the labour force, not of the whole population:
The definition is strict on purpose. It asks about actions (did you look?) and facts (could you start?), not about feelings. So two interviewers should put the same person in the same group. Try it on eight people.
Put each person in the box the ILO definition gives them. Check the three tests: no paid work, looked in the past four weeks, can start within two weeks.
Eight people, sorted by the ILO definition.
| Person | Group | Reason |
|---|---|---|
| Amara works 6 hours a week in a café and wants full-time work. | Employed | One hour of paid work counts. She is underemployed. |
| Joon lost his job last month, applied for ten jobs this week and could start on Monday. | Unemployed | All three tests are met. |
| Lucía is 70, retired and not looking for work. | Inactive | She is not looking. |
| Tendai stopped looking six months ago because he believes there are no jobs in his town. | Inactive | He has not looked in four weeks. He is a discouraged worker. |
| Sofia is a full-time student and is not looking for work. | Inactive | She is not looking. |
| Ravi's factory is shut for a month of repairs. He is still paid and will go back. | Employed | He is temporarily away from a job he still has. |
| Mei applies for jobs every week, but she is in hospital for the next six weeks. | Inactive | She cannot start within two weeks. |
| Kwame has no job, but has accepted one that starts in ten days. | Unemployed | Out of work, with a job starting within two weeks. |
Look at Amara, Tendai and Mei. Each of them wants work, but none of them is unemployed by this definition. Amara and Tendai come back later in the lecture.
Measure 1: count the people claiming benefits
The first way to count is to use records that already exist. In the UK, the claimant count is the number of people claiming benefits mainly because they are unemployed. Today that means Jobseeker's Allowance, plus people on Universal Credit who are required to look for work.
Its strengths come from the benefits system. The records already exist, so it costs almost nothing to produce. It is published every month, soon after the month ends. It covers every claimant, not a sample, so it gives figures even for small towns.
Its weaknesses come from the same place. It counts claimants, not unemployed people, and the two groups are not the same.
Many unemployed people do not claim. Some are not eligible, for example because they have savings or a partner with a good income. Some do not know they could claim. Some choose not to. All of them can be unemployed by the ILO definition and still be missing from the count.
Some claimants are not unemployed. Under Universal Credit, a person in a job with very low earnings or few hours can be required to look for more work. Those people are in the claimant count, even though they are employed.
The rules can change the number. If the government makes it harder to claim, the count falls. It falls because fewer people qualify, not because more people have jobs. This has happened in the UK. In October 1982 the monthly count switched from people registered at job centres to people claiming benefits. The government later estimated that this one change lowered the count by about 190 000 (Lord Young, House of Lords, 15 May 1987). In 1995, a minister told Parliament that nine changes since 1979 had had a noticeable effect on the count. Critics argued that changes like these made unemployment look lower than it was. A trend in the claimant count only means something if the rules stayed the same.
Measure 2: ask a sample of households
The second way is a survey. The UK's Labour Force Survey (LFS) asks people in a large sample of households about their work in a particular week. In 2020, the ONS described the sample as around 35 000 households each quarter. The answers are put through the ILO definition, so the survey measures exactly the thing defined above.
This gives the survey three strengths. It counts the unemployed whether or not they claim anything. It does not move when benefit rules change. And because almost every country uses the ILO definition, its figures can be compared across countries. Two countries' rates measured this way are answers to the same question.
Its weaknesses come from being a survey. The figures are estimates from a sample, and a different sample would give a slightly different answer. Each month the ONS publishes an estimate for the latest three months together, such as June to August. Combining three months makes the sample larger and the estimate steadier, but it means the figure describes a period that is already in the past. The ONS publishes a margin of error for each estimate. For the UK rate, it has been about 0.2 percentage points either side, at 95% confidence (Fraser of Allander Institute, 2019, using ONS data).
A survey also depends on people agreeing to answer. In April to June 2023, only 14.6% of the households asked took part in the LFS (Office for Statistics Regulation, November 2023). With so few answers, the estimates became too uncertain, and in October 2023 the ONS suspended its headline LFS figures. They returned in February 2024, after changes to the method. A good definition is not enough. The data must be collected well too.
The UK's LFS unemployment rate was 4.7% for June to August 2026 (ONS, October 2026). Suppose next month's figure is 4.8%. A headline says: "Unemployment rises." Is that a fair conclusion?
Show the answer
Not yet. A change of 0.1 percentage points is smaller than the survey's own margin of about ±0.2 points. The true rate may not have changed at all. In people, the change is not tiny: the UK labour force was about 36.3 million (34.58 million employed plus 1.71 million unemployed), so 0.1 points is about 36 000 people. But the sample cannot measure a change that small with confidence. A rise that continues for several months is much stronger evidence than one month's movement.
The claimant count counts people claiming unemployment-related benefits. It is cheap, monthly and complete, but it depends on the benefit rules.
A labour force survey uses the ILO definition on a sample of households. It is internationally comparable and does not depend on the benefit rules, but it is an estimate with a margin of error.
A real case: what the pandemic did to the two numbers
Now return to the hook, and treat it like an experiment.
The policy. In 2020, the UK government responded to the pandemic in two ways. It made Universal Credit more generous, so more people became eligible, including people still in work whose earnings had fallen. And it paid most of the wages of workers who were sent home from closed workplaces, through the furlough scheme described in the last lecture.
The prediction. Use the definitions. The more generous benefit should push the claimant count up a long way, because it brings in new claimants, including some who still have jobs. The furlough scheme should hold the survey figure down. Furloughed workers still had a job they were temporarily away from, and the ONS classified them as employed, just like Ravi in the sorter. So the claimant count should rise far more than the survey figure.
What happened. Just before the pandemic, the two numbers were close. The LFS found 1.36 million unemployed for December 2019 to February 2020, and the claimant count was about 1.2 million in March 2020. By May 2020 the claimant count had reached 2.8 million. The ONS warned that this included people in work with low income or hours. In February 2021, the LFS figure for October to December 2020 was 1.74 million, a rate of 5.1%. The claimant count for January 2021 was 2.6 million, 1.4 million higher than in March 2020 (all figures ONS).
Before the pandemic, the survey number was the higher one. That fits the usual pattern: many unemployed people do not claim. Ten months in, the order had reversed. The claimant count rose by about 1.4 million, while the survey figure rose by about 0.38 million.
The verdict. The prediction was right, and neither number was wrong. The claimant count measured how many people needed unemployment-related support, which had risen enormously. The survey measured how many people had no job and were looking. That had risen much less, partly because furlough kept millions in their jobs. To understand 2021, you needed both numbers, and you needed to know what each one counts.
What both measures miss
Now return to Amara and Tendai. The ILO definition is clear, but it leaves out real problems.
Underemployment. Amara works six hours and wants full-time work. She counts as employed, the same as someone with a full-time job. People who work fewer hours than they want are underemployed. In a downturn, many firms cut hours before they cut jobs. So underemployment can rise while the unemployment rate hardly moves.
Hidden unemployment. Tendai would work if he believed a job existed, but he has stopped looking. He is a discouraged worker, counted as inactive. People like him are called the hidden unemployed. In a long recession, their number grows, and the unemployment rate understates how weak the labour market is.
Discouraged workers also cause a strange effect when the economy recovers. Predict it before you read on.
An invented country has 28 million employed and 2 million unemployed, so its rate is ≈ 6.7%. A recovery begins. 600 000 discouraged workers start looking for work again. Within the quarter, 200 000 of them find jobs. The other 400 000 are still looking. Nothing else changes. What happens to the unemployment rate?
The worked answer. Employed: 28 + 0.2 = 28.2 million. Unemployed: 2 + 0.4 = 2.4 million. Labour force: 30.6 million. The rate rises from ≈ 6.7% to 2.4 ÷ 30.6 × 100 ≈ 7.8%, even though 200 000 more people have jobs. The 400 000 still searching were inactive before. Now they are looking, so they count as unemployed.
That is it. That is the main lesson about every unemployment figure. The rate depends on who is looking, not only on who has a job. A falling rate can be bad news, if people are giving up. A rising rate can be good news, if people are returning to search. Before you judge a change, ask what happened to the labour force.
Worked examples
An invented country has 50 million people aged 16 and over. 28 million are employed, 2 million are unemployed, and 20 million are economically inactive. Find the unemployment rate.
Labour force = employed + unemployed = 28 + 2 = 30 million.
Unemployment rate = 2 ÷ 30 × 100 ≈ 6.7%.
Check: dividing by the whole population would give 2 ÷ 50 × 100 = 4%. That is a different number: the share of all adults who are unemployed. It puts retired people and students in the denominator, as if they were in the labour market.
In an invented country, the claimant count is 1 000 000 and the survey finds 1 300 000 unemployed. The government changes the rules so that people with savings above $10 000 can no longer claim. 120 000 claimants lose their benefit but keep looking for work. Nothing else changes. What happens to each measure?
The claimant count falls to 1 000 000 − 120 000 = 880 000, a fall of 12%.
The survey figure does not move. The 120 000 people still have no job, are still looking and can still start. They are still unemployed by the ILO definition, so the survey still finds 1 300 000.
Verdict: a headline saying "claimant unemployment falls 12%" would be true about the count and false about unemployment. When the claimant count moves, first check whether the rules changed.
The country in Example 1 also has 0.5 million discouraged workers (now counted as inactive) and 1.5 million underemployed workers (counted as employed). Find a wider measure of people without the work they want.
Add the discouraged workers to the labour force, because they would work if they believed a job existed: 30 + 0.5 = 30.5 million.
People without the work they want: 2 + 0.5 + 1.5 = 4 million.
Wider rate = 4 ÷ 30.5 × 100 ≈ 13.1%.
Check: the wider rate is about twice the headline rate of ≈ 6.7%. Neither is "the true" number. They answer different questions, which is why some statistics offices publish several measures, from narrow to wide.
Three classic mistakes
The unemployment rate is the number unemployed divided by the labour force, not by the population. In Example 1, the correct rate is ≈ 6.7%, not 4%. Students, retired people and other inactive people are not in the denominator.
The claimant count can fall because the rules changed, or because people's benefit ran out. The number of unemployed people may not have changed at all. Before you use a fall in the count as evidence, ask whether the rules changed, and check it against the survey.
The rate can fall because discouraged workers stop looking and leave the labour force. No one found a job, but the rate went down. Always look at employment and inactivity alongside the rate.
Work each one on paper first. Open a hint only when you are stuck.
-
An invented country has 18.4 million employed, 1.6 million unemployed and 10 million economically inactive adults. Find the unemployment rate. Then find the percentage of all adults who are unemployed, and explain why the two numbers differ.
Hint 1
The labour force is the employed plus the unemployed.
Hint 2
Labour force = 20 million. All adults = 30 million.
Full answer
Labour force = 18.4 + 1.6 = 20 million. Unemployment rate = 1.6 ÷ 20 × 100 = 8%. All adults = 20 + 10 = 30 million, and 1.6 ÷ 30 × 100 ≈ 5.3%. The second number is smaller because it includes the 10 million inactive people in the denominator, even though they are not looking for work. The unemployment rate is the share of people in the labour market who cannot find work. Check: 8% of 20 million is 1.6 million. ✓
-
Classify each person using the ILO definition, and give the reason.
(a) Imran did two hours of paid tutoring last week and wants more work.
(b) Olga has no job and applied for three jobs this week. She could start in three days, but she is moving abroad next month.
(c) Diego has no job and has not looked for work since last year. He says he would take a job if offered one.
Hint 1
Use the three tests in order: any paid work? looked in the past four weeks? available within two weeks?
Hint 2
Plans for next month do not matter. Only the tests do.
Full answer
(a) Imran is employed, because he did at least one hour of paid work. He is underemployed. (b) Olga is unemployed: no job, looked in the past four weeks, and available within two weeks. Her later move does not change her status this week. (c) Diego is economically inactive, because he has not looked in the past four weeks. He is a discouraged worker, part of the hidden unemployed. Check: each answer follows from one of the three tests, not from how the person feels about work. ✓
-
An invented country has a labour force of 40 million and 2.4 million unemployed. In a long recession, 0.8 million unemployed people give up looking and become inactive. No one finds a job. Find the unemployment rate before and after. What would a politician say, and what is really happening?
Hint 1
The 0.8 million leave the unemployed and the labour force.
Hint 2
After: 1.6 million unemployed out of a labour force of 39.2 million.
Full answer
Before: 2.4 ÷ 40 × 100 = 6%. After: unemployed = 2.4 − 0.8 = 1.6 million, labour force = 40 − 0.8 = 39.2 million, so the rate is 1.6 ÷ 39.2 × 100 ≈ 4.1%. A politician might say "unemployment has fallen by almost 2 percentage points". In fact no one found work. 0.8 million people have stopped believing they can, which is a sign the labour market is getting worse. Check: if the discouraged workers are counted back in, the rate is 2.4 ÷ 40 = 6% again, unchanged. ✓
-
A minister announces: "The LFS unemployment rate fell from 4.8% to 4.7% this month, and the claimant count fell by 30 000. Our policies are working." Using what you know about both measures, write three questions you would ask before you accept this claim. Use the UK labour force of about 36.3 million to say what 0.1 percentage points means in people.
Hint 1
Think about the margin of error, the benefit rules, and what happened to employment and inactivity.
Hint 2
0.1% of 36.3 million is 0.001 × 36 300 000.
Full answer
0.1 percentage points of 36.3 million is about 36 000 people. Three good questions: (1) Is the 0.1-point fall bigger than the survey's margin of error? At about ±0.2 points, it is not, so one month's change could be sampling noise. (2) Did the benefit rules change? A fall of 30 000 in the claimant count means little if fewer people now qualify. (3) Did employment rise, or did inactivity rise? If the rate fell because discouraged workers stopped looking, the news is bad, not good. A fourth good question: did the trend continue over several months? Check: each question tests a different weakness, one for the survey, one for the claimant count and one shared by both. ✓
Every economic statistic is a definition plus a method. GDP leaves out unpaid work because of its definition. The claimant count follows the benefit rules because of its method. The survey has a margin of error because it asks a sample. None of this makes the numbers useless. It means a number is only as good as your understanding of how it was made.
This matters because these numbers guide real decisions. Central banks and governments use unemployment figures to set interest rates and plan spending, and voters use them to judge governments. The same habit works for inflation, for poverty and for exam results: before you trust a number, ask who was counted, how, and what was left out.
- Under the ILO definition, a person is unemployed if they have no paid work, have looked for work in the past four weeks and can start within two weeks.
- The unemployment rate is the number unemployed divided by the labour force, which is the employed plus the unemployed, never the whole population.
- The claimant count is cheap, monthly and complete, but it counts benefit claimants and changes whenever the benefit rules change.
- A labour force survey is internationally comparable and ignores benefit rules, but it is an estimate from a sample with a margin of error.
- Underemployed workers count as employed and discouraged workers count as inactive, so the headline rate can understate how many people lack the work they want.
- The rate can rise when discouraged workers return to searching, and fall when they give up, so changes must be read alongside employment and inactivity.