UK economy: Unemployment figures explained by ONS

Employment and unemployment are among the most closely-watched economic indicators. But how are the figures drawn up? Who exactly is counted as unemployed? And why does ONS quote unemployment figures for a three-month period, when the number of Job Seeker's Allowance (JSA) claimants is given for a single month?

Some have questioned the way the Office for National Statistics compiles the figures. Here, Caron Walker, director of collection and production from the ONS explains how and why it is done:

The answers lie with the Labour Force Survey, the huge continuous survey that ONS uses to measure unemployment (along with employment and economic inactivity).

The unemployment figures use an internationally-agreed definition.

To count as unemployed, people have to say they are not working, are available for work and have either looked for work in the past four weeks or are waiting to start a new job they have already obtained. Someone who is out of work but doesn't meet these criteria counts as "economically inactive".

Each quarter the LFS covers 100,000 people in 40,000 households chosen randomly by postcode. That's about one in 600 of the total population.

The results are then weighted to give an estimate that reflects the entire population. Our survey has a very large sample size compared, for example, with opinion polls, which often sample around 1,000 people.

Even so, in any survey there is always a margin of uncertainty, in this case around plus or minus 3% for the unemployment level.

People sometimes ask why ONS doesn't increase the sample size to give monthly figures.

Put simply, it's a matter of resources. Surveys on this scale take a great deal of work. The cost of doing it on a monthly basis was estimated as at least £7 million a year back in the mid-1990s. So a rolling three-monthly survey remains the best solution for efficiently producing accurate yet manageable data.

Obviously it's possible to look at the number of people unemployed, for example, in January-March and compare that with the February-April figure to see what the change is. But it wouldn't be a good idea: the February and March data are in both sides of the comparison, so effectively you're looking at January compared with April.

Quite apart from the smaller sample involved, the sample frame wasn't really designed for a monthly survey.

Prior to 1998, the LFS results were only published once a quarter, and the sample frame was designed around this. The country was divided into 212 interviewer areas, each of which is subdivided into 13 weekly 'stints', each randomly allocated across one of the 13 weeks of the quarter.

So while this means that across the entire three months we can be sure that the sample in each interviewer area is representative of the area as a whole, we would not necessarily expect it to be so with just one month's worth of interviews.

ONS has in fact been publishing some single-month LFS employment, unemployment and inactivity estimates on our website since 2004.

We did this in response to user requests for more information about the reasons behind movements in the three-month rolling figures. However, the monthly series are more volatile than the three-month ones - for the reasons noted - and so these are not of the quality to be designated National Statistics.

The other measure of joblessness - the claimant count - is published for each single month. It doesn't suffer from the limitations of sample size and sampling frame, because it derives from the numbers of Jobseeker's Allowance (JSA) claimants recorded by Jobcentre Plus, so a monthly figure is possible right down to local level.

But because many people who are out of work won't be eligible for JSA, it's a narrower measure than unemployment, typically about 1.5 million people recently, compared with about 2.5 million for unemployment.

And it's not comparable with other countries' data, unlike the LFS measures which use internationally agreed definitions.

So that leaves us with the three-month on previous three-month comparisons as the most robust and most complete measure of joblessness, which is why it's the one that ONS uses for the headline figures in the monthly labour market statistical bulletins.

It might not be perfect, but it has clear advantages over the other ways we might do it. We believe the UK figures are as reliable as any produced internationally. And because so much of their strength lies in the sheer scale of our survey, we owe many thanks to all the people who freely give their time to take part.

The opinions expressed are those of the author and are not held by the BBC unless specifically stated.