Chapter 5 Labour Market Statistics

Jonathan Athow

What do we care about when we try to measure the labour market?

How do we define employment, unemployment, economic inactivity and labour force participation?

How many hours do people work and how does that relate to underemployment?

What wages do people earn, and how does this affect issues like the gender pay gap?

How do we capture different forms of work such as the self-employed and those working in the ‘gig’ economy?

How do labour market outcomes affect household incomes, inequality and poverty?

While controversial at the time, the poster was seen as highly effective–so effective that it was revived the following year as part of the 1979 general election campaign, with the updated phrase ‘Labour still isn’t working’. Some politicians credited it with winning the election. In 1999, it was awarded the best billboard advertisement of the 20th century. The story of the iconic photograph makes interesting reading!

In 1978, the Conservative Party, then in opposition, wanted a billboard that would express feelings of discontent with the Labour government at the time. Saatchi & Saatchi, the advertising agency that the Conservative Party used, designed a poster showing a long queue of people outside the unemployment office. The headline: ‘Labour isn’t working’.

There was a reason why unemployment was a hot political issue. In 1979, Margaret Thatcher became prime minister after the Conservative Party won a 43-seat majority in Parliament. In the late 1970s the unemployment rate was almost 6%, compared to less than 4% in the early 1970s. Unemployment continued to climb during the 1980s, reaching a peak of just under 12% in 1984.

The labour market connects percentages to real people in a way that many other economic statistics do not. A job, and the earnings that go with it, are crucial for our welfare and that of our friends and our families.

The functioning of the labour market in terms of its economic efficiency and distributional effects is also central to political debate, but this is not something that started in the 1970s. In the UK, there were controversial political policies that affected the labour market in the 19th century. For example, in 1871, trades unions were legalised and in 1878 children under the age of 10 were prohibited from working in factories.

While the issues are different today, the labour market remains an important feature of political discourse and economic discussion.

ONS Resource

See Annex 1 of the ONS ‘Long-term trends in UK employment: 1861 to 2018’ for a list of key legislation affecting the labour market in the UK.

There are many ways in which the labour market is fundamental to our understanding of a modern economy:

  • It allocates a key factor of production (labour) to firms. This affects the output and productivity of the whole economy.
  • It distributes earnings to workers. This is the largest part of household incomes–meaning that the labour market affects the patterns of incomes across society. Living standards, inequality and poverty are all therefore affected by the movements of the labour market.
  • It shapes economic policy. Sir Jon Cunliffe, Deputy Governor at the Bank of England put it as follows:

    The Bank of England has a clear primary objective of price stability and a forward-looking inflation targeting remit … while neither unemployment nor pay is part of our target, the relationship between them is important to our understanding of domestically generated inflationary pressure in the economy and how it is likely to evolve over the 2 to 3 year period over which monetary policy can have influence.1

  • It affects long-term growth prospects of the economy. The economic potential of a country is determined by the factors of production available and the way in which those factors are combined. The most important factor of production for developed economies is labour, so increasing the number of people who can work and their productive potential is a key determinant of long-run economic growth. The more people in the workforce, the greater the potential prosperity of the country in question.

Many developed countries are seeing demographic changes including a slowing of growth, and in some cases falls, in their working age populations. This means the main way in which the labour input can be grown is through increasing the proportion of the population active in the labour market. Increasing labour market participation, therefore, is key to determining economic growth in the long term. Increasing labour market participation, therefore, together with improving productivity, is key to determining economic growth in the long term. The latter is discussed further in Chapter 6.

5.1 How we measure the labour market

The core statistical framework for the labour market needs to allow policymakers to understand both the short- and long-run issues. In addition, while many years of time series of data measured on a consistent basis are important for understanding labour market trends, there is also a need to ensure that the statistics cover a rapidly changing picture of work.

5.1.1 A complicated market

We would be wrong to assume the labour market is just like any other goods market.

It seems similar on the surface: there is a price (the wage) and quantity (the number of people in work and/or the hours they work), determined by the interaction of the supply of workers by households and the demand for those workers by employers. As with other markets, we can apply concepts such as excess demand and supply to think about how wages are set.

It is, however, important to bear in mind that the labour market is more complex and nuanced than many other markets. This means we need to be able to answer questions that move beyond some of the simpler concepts. For example, how many people are not employed and, importantly, why are they not working? How do people move between these different labour market states? What is happening to the distribution of wages in the economy, and—in particular—the number of people on high and low pay?

There are two main ways in which data on the labour market are collected: surveys of people and employers, and administrative data.

5.1.2 Labour market stocks: Levels

The basic and most widely used concepts in labour market statistics are the ‘stock’ measures, namely the numbers of people in and out of work. These can be expressed in levels terms (millions of people in work) or as a rate (the share of the population who are in work).

At any one point, individuals can be categorised into three groups: employed, unemployed and economically inactive. These are the key concepts underpinning many labour market statistics.

Those people in work, either as employees or self-employed, who receive earnings in return for their work.

People in the employed category are in work, either as employees or self-employed, and they receive earnings in return for their work. The category also includes other small groups, such as government-supported training and people known as ‘unpaid family workers’, who are discussed in Section 5.4.3.

For those employed, people may be working full-time or part-time. However, there is no agreed international definition over what constitutes full- or part-time work, and the International Labour Organisation simply describes part-time work as “regular employment in which working time is substantially less than normal”. This might seem unsatisfactory. However, we need to bear in mind that hours of work vary between different types of work, between different countries and over time, as countries become more developed.

ONS Resource

In the Labour Force Survey, people are asked to classify themselves as working part- or full-time, which avoids having arbitrary boundaries, but obviously adds a degree of subjectivity. In addition, as discussed in Section 5.3.6, we record people’s hours of work so that we can see the numbers of people working at or below particular thresholds.

Those people who are not working, but are looking for a job and available to take up work should they find it.
economically active
The sum of those people who are employed or unemployed.
economically inactive
These are people who are not working and are not looking for work, or not able to take up work.

The complementary category, those who were queueing up in the political poster, are unemployed people. They are not working, but are looking for a job and available to take up work should they find it.

Taken together, the employed and unemployed are described as the economically active. There is another category, people who are defined as economically inactive: not working, and either not looking for work or unable to take a job.

This can be because, for example, people are studying, caring for others or are too unwell to work. It also includes people who are retired, although many labour market statistics exclude people who are no longer of working age.

How are those categories defined? In a few cases, there are conventions in the presentation of these numbers that are not always obvious. Figure 5.1 shows how we divide up the labour market.

We begin with the number of people in the UK aged over 16, on the grounds that those below 16 are in (compulsory) education and therefore not of immediate relevance to the labour market. We can then divide that population in two ways:

  • We can break the population down into those who are employed, unemployed and economically inactive.
  • Those who are over 65 (and therefore more likely to be retired) and those who are aged between 16 and 64, who are described as the ‘working age’ population. Some countries use slightly different age cut-offs for working age.

This gives six possible combinations. These can be added up in various ways.

Figure 5.1 The different concepts and components of the labour market

The different concepts and components of the labour market

We can now put some numbers on the people in some of these different categories. Figure 5.2 shows the number of people in the main categories we are interested in.

Category Levels (thousands of people)
Population aged 16 and over 53,873
Population of working age (16 to 64) 41,516
Employed people 32,744
Of which employed people of working age 31,370
Unemployed 1,262
Of which unemployed people of working age 1,235
Economically inactive of working age 8,910

Figure 5.2 The main labour market aggregates

The main labour market aggregates

UK, 2022

5.1.3 Labour market stocks: Rates

The other concept we are interested in is the rates of employment, unemployment and inactivity. Again, these can be derived from the various components set out in Figure 5.1.

employment rate
The headline employment rate is calculated by dividing the employment level for those aged from 16 to 64 by the population for that age group.

There are two main ways of calculating an employment rate:

  • The most common measure in the UK looks only at the working age population, where the employment rate is the number of employed people of working age divided by the population of working age.
  • The overall employment rate including those aged over 65, so this is the total employed divided by the population aged over 16.
unemployment rate
The headline unemployment rate is calculated by dividing the unemployment level for those aged 16 and over by the total number of economically active people aged 16 and over. Economically active is defined as those in employment plus those who are unemployed.

The unemployment rate is calculated in a slightly different way. The standard approach is to take the total population of unemployed people, and divide it by the number of employed people of working age plus the number of unemployed people of working age. There are two features of this approach to the unemployment rate which are notable:

  • The numerator includes unemployed people above working age, but the denominator is working age only. (Although, in reality, the numbers of unemployed people aged 65+ is small in comparison to the overall population).
  • The denominator is not the whole population, but rather the unemployed and employed of working age, also known as the economically active population.
activity rate
The sum of the employed people of working age and the unemployed people of working age, divided by the total population of working age. Also known as: participation rate.
economic inactivity rate
The number of people of economically inactive people of working age divided by the population of working age.

Just as with our definitions of the numbers of people who are economically active and inactive, there are rates of activity and inactivity. You may also hear the activity rate described as a “participation rate”. By definition, the activity and economic inactivity rates must sum to 100%.

Populations vary between countries and over time, and so much of our economic analysis focuses on rates. Figure 5.3 shows the main labour market data expressed as rates for the year 2022.

Rate Value (%)
Employment rate of working age 75.6
Unemployment rate 3.7
Activity or participation rate 78.6
Economic inactivity rate 21.5

Figure 5.3 The main labour market rate

The main labour market rate

UK, 2022

5.1.4 Labour market stocks: Time series

Using rates, we can also look at data through time. Figures 5.4 and 5.5 show the economic activity rate and the unemployment rate since the early 1970s. There has been a steady, small structural rise in activity (participation) over this period, from around 75% to around 80%. A key driver of this has been the increasing participation of women in the labour market.

Unemployment rates are more varied, reflecting the fact that unemployment responds strongly to the economic cycle. The last three downturns—the early 1980s, early 1990s and late 2000s—are quite obvious, with large increases in unemployment. One interesting feature of the chart is that the peak of unemployment following the late 2000s downturn—around 8%—is much less than in previous two downturns. This was despite the fact that the 2000s downturn saw a greater fall in GDP compared with the two previous downturns. The rise in the unemployment rate during the coronavirus pandemic was constrained by strong government support to the labour market, in particular the job furlough scheme which sought to preserve jobs in industries adversely affected by restrictions on economic and social activities. Labour markets have numerous margins of adjustment, so workers may see a fall in wages or reductions in hours rather than a reduction in the number of jobs.

Figure 5.4 Economic activity rate

Economic activity rate

UK, January to March 1971 to October to December 2022

Figure 5.5 Unemployment rate

Unemployment rate

UK, January to March 1971 to October to December 2022

5.1.5 Labour market flows

The concepts of employment and unemployment reflect an individual’s labour market status at a distinct point in time, and often we see little change in these measures between quarters. Beneath these headline aggregates, we see a significant amount of churn in the labour market. These are important to understand, as it is the cumulative effect of labour market flows that determine levels of employment and non-employment.

Figure 5.6 shows labour market flows between the third quarter (July to September) and the fourth quarter (October to December) of 2022.

Figure 5.6 Labour market flows

Labour market flows

UK, Quarter 3 2022 to Quarter 4 2022, thousands of people

Figure 5.6a All labour market flows

The figure shows flows between employment, unemployment, and inactivity.

The figure shows flows between employment, unemployment, and inactivity.

Figure 5.6b Gross flows between labour market statuses

This is the number of people moving between different statuses. So, for example, 384,000 people moved from unemployment to employment between the two quarters.

This is the number of people moving between different statuses. So, for example, 384,000 people moved from unemployment to employment between the two quarters.

Figure 5.6c Net flows

These adjust for the flows in the other direction. So, while 384,000 people left unemployment to become employed, 270,000 people left employment to be unemployed, meaning that the net flow from unemployment to employees was 114,000.

These adjust for the flows in the other direction. So, while 384,000 people left unemployment to become employed, 270,000 people left employment to be unemployed, meaning that the net flow from unemployment to employees was 114,000.

Figure 5.6d Did not change status

The numbers indicate the number of people in each labour market status who did not change status between the last two quarters of 2022.

The numbers indicate the number of people in each labour market status who did not change status between the last two quarters of 2022.

Figure 5.6 shows a much more dynamic labour market than the headline figures would suggest.

Starting with employment, we see that 856,000 people left employment, of whom 270,000 became unemployed and 586,000 became economically inactive. Going the other way, 1,018,000 people became employed, of whom 634,000 were economically inactive and 384,000 were unemployed. The total change in employment between quarters was only 61,000, but under the surface there was much more of a churn.

The flows in and out of unemployment are even more dramatic. Some 648,000 people left unemployment between quarters, nearly the same as the number of people staying unemployed between quarters. The change in the stock of unemployment between quarters was relatively small—a total inflow of 46,000 people—but around half of people who started the quarter as unemployed left unemployment by the next quarter.

As this is a dynamic system, relatively small differences in flows can accumulate over time. So, for example, there only needs to be a relatively small percentage change in the in- and out-flows to lead to a much larger change in overall unemployment over the course of a few years.

Another way of representing these flows is through ‘hazard ratios’. These measure the proportion of people who are in the same or different status the following quarter compared to the total number of people in original status in the previous quarter. For example, a hazard ratio can show the ratio of people who are still in employment in Quarter 4 compared to the number who were in employment in Quarter 3.

Figure 5.7 shows the hazard ratio for different labour market states, although unlike Figure 5.6, we have put self-employed and employees together as the employed. This reinforces the point that the labour market shows considerable fluidity. It is worth remembering that these are figures for just one quarter, and over the course of a year they are much more dramatic.

Labour market state Hazard ratio
As percentage of those in employment in previous quarter  
Still in employment 97.2
Employment to unemployment 0.9
Employment to inactivity 1.9
As percentage of those in unemployment in previous quarter  
Still in unemployment 45.1
Unemployment to employment 32.0
Unemployment to inactivity 22.0
As percentage of those in inactivity in previous quarter  
Still in inactivity 85.6
Inactivity to employment 7.1
Inactivity to unemployment 4.7

Figure 5.7 Hazard ratios

Hazard ratios

UK, October to December 2022

It is striking that even among the group of people who are economically inactive, who one might see as a more settled group, more than 10% move out of the category in one quarter. All of these statistics show the importance of understanding the dynamics of labour market flows.

We can also look at these hazard ratios over time. Figure 5.8 shows the hazard ratios for staying in the same category compared to the previous month. In Figure 5.8 and Figure 5.9 we have shaded an area to indicate the recent economic downturn. The shaded area corresponds to the period when the unemployment rate was above 7.5% (the period March to May 2009 to July to September 2013). This gives an indicator of an economic downturn.

Figure 5.8 Hazard ratios

Hazard ratios

UK, October to December 2001 to October to December 2022

For both those staying in employment or economic inactivity, there is very little cyclicality. Even at its lowest point, the hazard ratio in Figure 5.8 for staying in employment was at 95.9%. But the picture for unemployment is completely different. The hazard ratio peaked at just over 66% when the labour market was weak, compared to around 50% before the economic downturn.

This rising hazard ratio for remaining in unemployment during the economic downturn does not mean that people stopped moving into employment. Even at the labour market’s weakest points, around one-third of unemployed people were leaving unemployment in any one quarter. Moreover, at the peak of the hazard ratio for remaining in unemployment—October to December 2011—around 900,000 people left unemployment, of whom 500,000 moved into work. This shows that, even when unemployment is high or rising, people continue to flow into employment in large numbers.

We can also analyse the number of people who move between jobs in one quarter compared with the previous quarter. Of the 97% or so who stay in employment between one quarter and the next, hundreds of thousands of people move jobs.

Figure 5.9 shows the proportion of people who move jobs in one quarter as a share of employment in the previous quarter. As before, the shaded area corresponds to the period when the unemployment rate was above 7.5%, an indicator of the economic downturn.

Figure 5.9 Job-to-job moves as share of employment

Job-to-job moves as share of employment

UK, October to December 2001 to October to December 2022

This is another statistic that is strongly cyclical, with a large reduction in job-to-job moves during the economic downturn. In general, it is thought that people are less likely to move jobs in a downturn as there are fewer opportunities and greater risk aversion among employees.

As we will see, higher rates of job moves are associated with greater pay increases. This makes intuitive sense: those who change jobs often do so to get a pay rise or promotion.

5.2 Do the statistics correspond with labour market theory?

The distinction between the economically inactive and the unemployed seems arbitrary. These concepts, however, reflect economic thinking about how labour markets operate.

5.2.1 Labour supply and demand

Most economic thinking about labour markets is based on the idea that the number of people who have jobs and wage rates is determined by the interplay between the demand for labour and the supply of labour. There are different models that describe the interactions of labour supply and demand. Despite their differences, the concept of labour supply—the number of people willing to work for a given wage—is often key.

People who are economically inactive—those who are either unable to work and/or not looking for work—are not part of labour supply and, therefore, the labour market. This means they do not affect the setting of employment and wage. A corollary of this is that people who are either working or unemployed are said to be ‘economically active’. The other terminology here is that those who are either employed or unemployed are active or participating in the labour market.

Figure 5.10 shows a simple competitive supply and demand model of the labour market, a common tool that economists use.

Figure 5.10 Labour supply and demand

Labour supply and demand


If you have studied economics, you have probably seen something like this many times. We have the wage (W) on the vertical axis, and the amount of people in work (L) on the horizontal axis. The labour supply curve—LS—is upward sloping, reflecting higher wages that could tempt more people into work. The labour demand curve, LD, is downward sloping. This reflects that every additional worker an employer takes on is likely to contribute less in terms of additional output than the previous worker. In this model labour market, the equilibrium wage is W*, and equilibrium level of employment is L*.

This model is very simple, but we can use it to map the concepts of employment (E), the unemployed (U) and the economically inactive (I). The total number of people active (or participating) in the labour market A is the sum of those in employment (E) and those who are unemployed (U).

5.2.2 Mapping labour force statistics to the model

Many people would argue that the idea that the labour market is competitive is unrealistic, and others may argue about the appropriate shapes of the supply and demand curves. Nonetheless, it helps us to understand how some of the concepts apply to the labour market models that are taught. In addition, the same categorisation of employment, unemployment and economic inactivity can be applied to models of imperfectly competitive labour markets. The economic interpretation of unemployment varies between an imperfectly competitive and competitive labour market, but the statistical concept—that people are available and looking for work—is the same.

The economically inactive are outside the labour market and, because they are not looking or available for work, there is no wage level that would get them into work. This includes groups who have previously looked for work and could in theory move into employment—such as ‘discouraged workers’ who have given up job seeking—but because they are no longer looking, there is no wage that could tempt them back into a job.

The economically inactive also includes people who are sick or disabled and people caring for children or adults. These people may face particular barriers to moving into work, for example, finding suitable childcare. In this case, offering a higher wage may not be enough to get people back into work.

In principle, one could argue that there would be a wage that could tempt people, such as the retired, back into the labour market. But remember that the definition of inactivity is that they are neither able to take up work, nor looking for work. If you are not looking for work, you are not likely to be aware of the wages on offer.

The framework of employed, unemployed and economically inactive people is a coherent and exhaustive approach to categorising people in and out of the labour market. It makes strict logical sense, but—as with many models—it is something of a simplification, for two main reasons.

  • Unemployed people are not a homogenous group. People who have been unemployed for some time may lose skills or become more detached from the labour market. In terms of the labour market, their labour supply might be less effective than, say, someone who has only recently become unemployed. Putting this another way, just looking at the level or rate of unemployment might not tell you about how much slack there is in the labour market.
  • People move between categories. The framework looks at behaviour at one point in time. In reality, people change and move between categories. For example, the person who is long-term unemployed might in time give up looking for work and then become economically inactive. Likewise, active labour market policies might help people who are economically inactive to start looking for work, increasing labour market activity and the supply of labour. It would be wrong to think that the level of economic activity or inactivity shown in Figure 5.1 is fixed.

5.3 What is work like?

5.3.1 The Labour Force Survey

ONS Resource

You can browse the data from the Labour Force Survey.

The principal source of labour market statistics in the UK is the Labour Force Survey (LFS). This is an extensive survey of households, covering around 80,000 households every quarter. Most other countries run a similar survey to produce their labour market statistics.

One point to note about the LFS is that it is a survey of private households. This means it does not cover what are known as ‘communal establishments’, which include care homes, student halls of residence and similar. In the 2011 Census, around 1.7% of the population of England lived in communal establishments. In terms of the labour market, however, only around 185,000 people in communal establishments in 2011 were economically active. This is relatively small in relation to the overall economically active population, which is currently around 34 million.

The LFS allows the Office for National Statistics to collect a wide range of data, including both ‘hard’ information—such as whether someone is in or out of work—and ‘soft’ behavioural or attitudinal information—such as whether people are looking for work and if not, why not, and their preferences over hours. As we discuss in Section 5.4.4 administrative data from government can complement or replace some of the ‘hard’ data, but some of the other wider aspects of labour market behaviours—such as the ‘soft’ characteristics—always require some form of survey.

The other feature of the LFS is that it is designed to follow people for a year. People join the LFS in ‘wave 1’ and they are followed up on a quarterly basis until ‘wave 5’, when they drop out of the sample. This rolling panel approach allows some aspects of longitudinal behaviour to be explored— such as labour market flows described in Section 5.1.5—while continuing to provide a good picture of the current state of the labour market.

As with all surveys, there are some challenges with the LFS and some areas where it performs less well. A major issue is that the sample size does not always allow very detailed analysis to be undertaken. For example, if we are interested in youth unemployment in a particular part of the UK, the sample size is often too small to produce reliable statistics.

5.3.2 Counting jobs and vacancies

The discussions so far have focused on the number of people in work, unemployed or economically inactive. As discussed in the introduction, people in the first two categories can be thought of as labour supply, and the economically inactive can, and do, become part of labour supply over time.

The number of jobs in the economy can be seen as an indicator of labour demand: how many jobs are firms willing to make available at a given wage rate. In turn, this can be thought of as having two components: the number of filled roles, and the number of vacant posts.

Even when considering the number of filled roles, it is also important to remember that the number of people currently in employment is not the same as the number of jobs that employers may be offering, for two reasons:

  • Some employed people are self-employed. They are not captured in measures of employee jobs. This can be compensated for by adding estimates of self-employment from the Labour Force Survey to estimates of jobs obtained by surveying employers.
  • Some people may have more than one job.

The number of vacancies is often seen as a leading indicator of the labour market. Increases in vacancies might lead to rising employment and falling unemployment in the future, and vice versa. Arguably, employers stop advertising for new posts before reducing their existing workforce. As we saw in Section 5.1.5, a reduced flow into employment will, in time, feed through into increased unemployment and/or economic inactivity.

Figure 5.11 shows the total level of vacancies graphed against the inverted unemployment rate. We can see some evidence of the leading indicator nature of vacancies, with increasing vacancies usually feeding through to a fall in the unemployment rate and vice-versa.

Figure 5.11 Vacancies (thousands) and unemployment rate (%)

Vacancies (thousands) and unemployment rate (%)

UK, 2001 to 2022

The extent to which aggregate vacancies is a good leading indicator depends on the relative distribution of those vacancies. At times of significant shifts of relative labour demand—for example, when some sectors are shrinking and others growing—we might find that high vacancies are a sign of labour mismatch rather than strong aggregate labour market demand. This idea of mismatch occurs when the skills of the unemployed do not match well with the skills needed to fill the unfilled jobs.

Another common metric is the unemployment-to-vacancy ratio, sometimes called the u/v ratio. This is simply the number of unemployed people divided by the number of vacancies. The ratio provides an indicator of the level of slack in the labour market: when the u/v ratio is high, there are many people for every vacancy, indicating lots of scope for employment to grow. A low ratio arises when there are few unemployed people for each unfilled job.

5.3.3 Wages and earnings

Wages or earnings are the payments people receive for work. In theory, it encompasses both employees and the self-employed, but due to data limitations, most official statistics only cover employees.

There are two main sources of data on wages in the UK:

  • The Average Weekly Earnings (AWE) estimates. This is a monthly indicator. The AWE provides high-level estimates with less detail, but is available much more quickly than other sources. It is based on the Monthly Wages and Salaries Survey of businesses.
  • The Annual Survey of Hours and Earnings (ASHE). This provides a detailed set of data on wages, covering around 1% of all employees, again with data collected from businesses. While less timely, the richness of data allows for detail on issues such as the gender pay gap and measures of high and low pay.

ONS Resource

You can study the AWE data.

You can study the ASHE data.

The LFS also collects some information on wages, but this is not as strong as the earnings data from the AWE or ASHE. The LFS relies on respondents’ recall. In some cases, it even relies on someone responding on someone else’s behalf. The AWE and ASHE collect information directly from employers, who are likely to use their own pay records to respond to the survey. On average, we would expect employers to be able to provide more accurate information on wages than employees’ own recall.

On the other hand, the LFS includes a more detailed set of characteristics on the individual, such as ethnicity and education. This allows a richer analysis to be undertaken, based on those characteristics. It is, for example, the only source that can be used to calculate an ethnicity or a disability pay gap. In practice, LFS is used where those characteristics that it captures are important, and it is not generally used as a measure of wages in the economy.

When looking at earnings or wages, we tend to look at hourly pay and weekly pay. The difference between the two is the number of hours worked. This might seem trivial, but as we shall see, the changing patterns of hours can give rise to some statistical issues, as described in the box, Measuring wages and compositional change using AWE.

5.3.4 AWE: A snapshot of earnings in sectors

The two main sources of earnings data can help to understand different aspects of the labour market and have different properties. The AWE is an aggregate measure that gives us the (arithmetic) mean of earnings at a particular point in time. It is calculated by asking a sample of employers their pay bill and number of employees for a specific time period. This data can then be grossed up to the whole economy.

This approach allows data to be gathered at low cost and with minimal burden on employers. We can show the overall economy average and the average for different sectors of the economy, but not much more. Figure 5.12 disaggregates some of the data we have.

Figure 5.12 Annual growth in nominal average weekly earnings (total and regular)

Annual growth in nominal average weekly earnings (total and regular)

Great Britain, October to December 2021 and October to December 2022, seasonally adjusted

The data from AWE are a good measure of wages at an aggregate level, but have some limitations. Notably:

  • The AWE estimate is affected by the composition of the workforce. An increase in the share of highly paid jobs or an increase in the share of full-time workers increases the AWE measure. This is because the pay bill and employee numbers collected is a mix of full- and part-time workers, and reflects a changing mix of employees over time. It is possible for the AWE to rise or fall when the hourly or weekly wage of everyone in the sample is unchanged. The box, Measuring wages and compositional change explains this in more detail.
  • The AWE does not tell us about the number or proportion of people seeing rises in hourly or weekly earnings.
  • The AWE contains no information on the characteristics of the people in question. We cannot infer from the data what is happening with the gender pay gap or whether wage increases are accruing to those on high or low incomes.

How it’s done Measuring wages and compositional change using AWE

Let us take an economy that employs 100 people in two sectors. Sector 1 is low paid and part-time,with each person paid £10 an hour and working 20 hours a week. Sector 2 is higher paid and full-time, with a wage of £20 an hour and each person working 35 hours a week. For simplicity, let us assume that 50 employees work in each sector 1.

It is very straightforward to calculate the average (arithmetic mean) weekly earnings for the economy as a whole, as it is the weighted average of the two sectors:

  • for Sector 1, each employee earns/the average wage is: 20 × £10 = £200
  • for Sector 2: 35 × £20 = £700
  • for the whole economy, the average wage is: (0.5 × £200) + (0.5 × £700) = £450

Say that, the following year, 20 of the people in the low pay Sector 1 lose their jobs, but for everyone else their hours and hourly wages remain unchanged. We now have 30 people in Sector 1 and 50 people in Sector 2. The whole economy’s average wage is now: (0.375 × £200) + (0.625 × £700) = £512.50.

So, in this example, average wages have risen by around 14%, but note that no individual has seen their hourly or weekly earnings rise. In this case, the change in average weekly earnings is not a good indicator of changes in individuals’ hourly or weekly earnings.

We could have constructed the example to show a fall in average weekly earnings, or a change in hours worked that increased or decreased weekly earnings while leaving hourly wages unchanged. This effect, reflecting compositional change in the population in question, happens a lot in statistics where you are averaging across a diverse and changing population.

You might take the example to suggest that this measure of average weekly earnings is flawed in some way. That is, however, the wrong conclusion, as there is nothing incorrect with this average weekly earnings statistic. It is what it purports to be, but changes in it need to be understood. Put another way, there are limits to what you can infer from such a high-level aggregate.

To mitigate this effect, or at least decompose the change in an average weekly earnings change, you need to do a more ‘like for like’ comparison. This could mean, for example:

  • focusing on hourly wages rather than weekly earnings, so that changes in working hours are removed from the comparison
  • breaking down the aggregate to different sectors
  • controlling for other factors, such as the full-time and part-time split or the occupation of the employees
  • following the same individuals between years

5.3.5 ASHE: Longitudinal measurement of earnings

The other main source of earnings data, ASHE, addresses many of the limitations of AWE, but at the cost of timeliness. ASHE has a sample of around 300,000 employees and collects details on each individual—such as their gender and occupation—in addition to their hours and hourly wage rate in April of the year in question. This greater level of detail is more costly in terms of data collection and burdensome for employers, and is only really practical as an annual survey.

The way the sample is created has a ‘longitudinal’ element, which allows some employees to be followed between years. In turn, this allows us to see if hourly or weekly earnings are increasing between years for those people.

This mix of individual characteristics and the ability to follow an individual between years allows us to see—to a large degree—through the compositional change of the workforce. However, it should be noted that ASHE does not include all potentially relevant information on individuals, such as skills or qualifications.

Figure 5.13 shows the distribution of hourly wage growth. This is produced by ASHE by comparing individuals who were in employment in both years. The graph shows real terms changes in hourly earnings, i.e. after adjusting for the effects of inflation.

Figure 5.13 Cumulative distribution of real hourly wage growth (%)

Cumulative distribution of real hourly wage growth (%)

UK, 2011 to 2019

The cumulative distribution allows us to read off what proportion of people saw a fall in their real terms hourly wage rate. This can be done by looking down the 0% real growth line, marked on the chart. In 2011, around 60% of people saw a real-term fall in their hourly wages, while in 2015 only around 25% saw a fall in real-term hourly wages.

The chart also allows us to see median hourly wage by looking along the 50% line, also marked. In 2011, the median wage growth was negative 1.2%, while in 2015 it was positive 2.2%.

The ASHE dataset also allows us to look at wage rises for those who move jobs and those who remain in the same job between years. People who move jobs typically experience higher year-on-year growth in hourly wages compared with those who do not move. This is not really surprising, as higher pay is likely to be one reason motivating people to change in jobs.

Figure 5.14 shows median real hourly wage growth for those who change jobs and those who remain in the same job. On average over the period, the median rise for job changers was five percentage points higher than the job stayers. It also appears to be slightly cyclical, with a lower premium when the labour market was at its slackest in around 2010.

As discussed in Section 5.1.5, the rate of job-to-job moves also appears cyclical—at least in recent years—and this interacts with job-movers’ pay premium to further depress wage growth during a downturn. When the economy is weak, fewer people move to a new job—weakening pay growth—and when they do move, the pay premium they receive is lower than when the economy is stronger.

This behaviour reinforces the point earlier that, to fully understand the labour market, you need to understand labour market flows. Flows affect employment and unemployment rates, but also appear to have effects on wage growth.

Figure 5.14 Annual percentage change in median full-time real gross hourly earnings for job stayers and job changers

Annual percentage change in median full-time real gross hourly earnings for job stayers and job changers

UK, 2002 to 2019

5.3.6 Hours of work and underemployment

We have so far focused on whether people are in or out of work. In the terminology of labour market economists, this is known as the ‘extensive’ margin. There is, however, another element of the quantity of labour in terms of the hours of work or the ‘intensive margin’. Employers may change the number of hours they offer and employees might adjust the number of hours they are prepared to work, reflecting a wide range of factors.

Statistics of the number of hours people work are also available from the Labour Force Survey. Figure 5.15 shows the share of employment by usual weekly hours of work, divided into five time bands.

Figure 5.15 Percentage share of employment by usual weekly hours

Percentage share of employment by usual weekly hours

UK, 1992 to 2022

The share of people usually working very short hours—fewer than six hours a week—is very low, around 1.5% or around 500,000 people out of the more than 32 million people in work. This is down from around 2% in the early- to mid-1990s. The next category—from 6 to 15 hours a week—has also shrunk as a share of employment over the same period of time.

Just over half of employed people usually work between 31 and 45 hours a week. About 20% of people work between 16 and 30 hours a week, while another 20% work more than 45 hours a week. The pattern seems to be that hours of work at both extremes—fewer than six hours or more than 45 hours—have decreased over the past 25 years or so.

An issue connected to hours of work is something called ‘underemployment’, which is the concept that some people might be working fewer hours than they would like. This relates closely to the concept of slack in the labour market—that is, the amount of spare capacity in the labour market. This can arise either through having lower employment, or by having the same employment with people working fewer hours than they would like.

There is no single agreed definition of underemployment, but there is much data collected by labour market surveys such as the UK’s Labour Force Survey. This survey asks people about their current working patterns and also their desire to work more or fewer hours.

While there are a number of measures of underemployment, two important ones are:

  • The Labour Force Survey (LFS) underemployment measure. This is the sum of all the people who either wish to work more hours in their current job, wish to change jobs to work longer hours, or who would like an additional job. This can be expressed either as a number of people, or as a proportion of the work force or working age population.
  • The Blanchflower-Bell hours-based approach. This is a net calculation of the number of extra hours people would like to work, less the number of hours more than they want to work. The Blanchflower-Bell measure is normally expressed as a proportion of some measure of hours worked.

The Blanchflower-Bell method to calculate underemployment is used to dig deeper into ONS data on underemployment and the lack of wage pressure in the UK.

The authors argue that their data shows that what politicians often described as ‘full employment’ in 2019 did not provide the full picture of underemployment in the US and Europe.

How it’s done Computing the Blanchflower-Bell measure

The LFS measure is a simple count, while the Blanchflower-Bell is more complex. The basic calculation for Blanchflower-Bell is as follows:

desired hours worked less actual hours worked, for those people who would like more hours
actual hours worked less desired hours worked, for those people who would like fewer hours
desired hours for unemployed people.

One challenge with the Blanchflower-Bell method is that an estimate of people’s desired hours of work needs to be made. It is usual practice to assume that desired hours are equal to a measure of average hours worked.

Figure 5.16 shows these two measures of underemployment alongside unemployment. In 2008, we saw a large rise in underemployment at the same time as unemployment rose. This suggests that simply relying on unemployment understated the amount of spare capacity in the labour market.

Figure 5.16 Underemployment vs employment, using LFS and Bell-Blanchflower indicators

Underemployment vs employment, using LFS and Bell-Blanchflower indicators

2002 to 2017

As can be seen, underemployment moves with unemployment suggesting that the labour market adjusts both through the intensive (hours) and extensive (employment/unemployment) margins.

5.4 Work—and workers—need careful measurement

The labour market brings with it lots of discussion on some of the key concepts. In addition, the ways in which people work are changing, and the extent to which this is fully reflected in the labour market statistics is often debated. Labour market statistics therefore need to continually evolve to ensure they capture all the salient features of the labour market.

5.4.1 Economic inactivity

Those people who are described as economically inactive are those who are not working but who also do not count as unemployed, because they are either not available or not looking for work.

This terminology makes sense to economists, but it gives an inaccurate impression of the people it describes. Many people who are economically inactive are very active in other respects. For example, many people are not working because they are caring for children or adults.

This raises a further point that some of the activities undertaken by the economically inactive—namely caring for others—could be categorised as work if they were paid. If, for example, you stayed at home to care for a child, you would not be counted as employed. But if you were to employ someone else to look after that child, that person would be counted as employed.

There is a parallel with how we measure the economy through Gross Domestic Product (GDP). The economic activity from paid work—such as caring for others—is counted in GDP, but where this work is done free of charge within the household, it is not counted, as discussed earlier in Section 2.3.3.

This approach of not counting unpaid work nor the output of that work ensures consistency between the different measures, but obviously means that these ‘household production’ activities are missed from both GDP and the labour market statistics.

Changes in technology and working patterns are raising questions about whether this boundary is drawn in the correct place. This question of what is included within the boundary of GDP and the labour market is explored further in Chapter 12 on hard to measure statistics.

The economically inactive are a heterogenous group as there are many other reasons why people might not be looking for work. Some people are wealthy enough not to need to work, while at the other end of the spectrum, are people who may have given up looking for work in the belief there is none to be found. They are often called ‘discouraged workers’.

Figure 5.17 shows the number of people in the UK who are economically inactive, breaking the total down into the main reason or the inactivity.

Figure 5.17 Inactivity by reason

Inactivity by reason

UK, April to June 1993 to October to December 2022, seasonally adjusted, thousands

In October to December 2022, the biggest category was long-term sick (2.5 million) followed by students (2.3 million), and people looking after family or home (1.7 million, predominantly women). There were also 1.1 million people in this age group who were already retired. The ‘discouraged workers’ group, however, was only 31,000 strong.

It would be wrong to assume this group of people has been permanently removed from the labour market. As we have seen, around 10% of people who are classified as economically inactive in one quarter become economically active in the following quarter.

5.4.2 Zero-hours contracts and the definition of employment

zero-hours contract
Workers are on call to work when their employers need them, but the employers do not have to give them work. Employees are also not obliged to work when asked.

One question that is frequently asked is whether the statistical definition of work distorts the employment figures. Are there large numbers of people working very short hours, perhaps those on zero-hour contracts or similar, which exaggerates our impression of the overall level of employment?

In the Labour Force Survey, each person surveyed is put into one of the three categories described in Figure 5.1: employed, unemployed or economically inactive. This is done using internationally agreed definitions set out by the International Labour Organization (ILO).

The ILO definitions means that everyone aged 16 or over is in one of three statistical categories:

  • Employed. To be in this category, the person only needs to have worked at least one hour in the last week.
  • Unemployed. Someone in this category has been actively seeking work in the past four weeks and is available to start work in the next two weeks, or is out of work but found a job and is waiting to start in the next two weeks.
  • Economically inactive. Someone in this category does not have a job, has not been looking for a job in the past four weeks, or is not available to start work in the next two weeks—in other words, everyone who is not in the first two categories.

The ILO’s ILOSTAT data offers statistics based on international standards set by the International Conference of Labour Statisticians (ICLS). Its data means you can compare these numbers internationally; it is also free to download.

The LFS measures “usual hours of work” rather than contracted hours, and so people on zero-hour contracts—those that do not offer a guaranteed number of hours a week—are included in our employment figures on the basis of their usual hours of work. We have seen far more attention paid to these contracts recently. If a large number of people were in this type of work, our count of employed people would be narrowly correct, but would not reflect economic activity.

We saw in Section 5.3.6 that the proportion of people working between one and six hours a week is low—around 1.3% of workers—and has changed little in recent decades. So including them as employed does not distort our data.

5.4.3 Who is unemployed and who is in work?

While some economic statistics are fairly abstract, terms like employment and unemployment are readily understood, but the statistical measures that the ILO—and, by extension, the ONS—uses, are sometimes misunderstood.

Remember that the definitions of unemployment and employment are set internationally, and statistical bodies like the ONS apply these definitions. Moreover, these definitions have been in place for many years; we can trace a consistent measure of unemployment back to 1971. This is an advantage when making a time series: we can be clear that we have been using the same statistical definition for unemployment for nearly 50 years.

Misunderstanding 1: Because the government has tightened eligibility for benefits, the ONS undercounts unemployment.

It is true that the criteria to be eligible to claim for unemployment benefit have changed over time. But this is not the definition the ONS uses to produce its unemployment figures.

Misunderstanding 2: There is a group of ‘unpaid family workers’ included among the employed, so people caring for children or family members are counted as employed.

The people captured here are not those caring for children or family members, but people working in the family business. As an example, if a relative such as a spouse helps out in a family-owned shop, that is counted as an ‘unpaid family worker’. This makes sense as the earnings from family-owned businesses are likely to be shared within the family.

It is also important to remember that this is a small group—typically around 100,000 people—compared to over 30 million people who are employees or self-employed. So, even if you thought unpaid family workers should not be counted as employed, they would have very little effect on the overall labour market figures.

5.4.4 Administrative data

Information collected by government or its agencies as part of their day-to-day operations can be very useful for understanding the economy, and the labour market is no exception. But we must interpret this data with care.

Perhaps the best-known administrative data for the UK labour market is the ‘claimant count’. You are part of the claimant count if you met the criteria to receive an unemployment benefit as part of the system of Universal Credit.

This, however, uses a different definition from the unemployment statistics produced using the Labour Force Survey. The different definitions mean that the numbers of unemployed people and the claimant count have often been quite different. Moreover, the relationship between them has not been stable over time. This means that the ‘claimant count’ is rarely used as a measure of unemployment, and the Labour Force Survey measure—based on a standardised international definition—has become the headline measure.

There are, nonetheless, two other ways in which the administrative data can be of use.

  • More detailed information on labour market status. For example, we may be able to put together data on someone’s current benefit status with their history to estimate whether someone is unemployed or economically inactive.
  • Administrative data as a proxy for a labour force statistic. For example, data from the UK’s income tax system corresponds well with measures of employment (of employees) and the wages they are paid as measured in our surveys. From December 2019, the Office for National Statistics has reported information on wages using administrative data from the UK’s income tax system alongside survey-based measures of earnings.

ONS Resource

You can find the administrative wage data here.

5.4.5 Measuring the wages of the self-employed

The self-employed are counted as employed just like employees, although concepts such as hours of work are sometimes more challenging to apply than for many other employees. Many employees have very specific contracts listing their hours of work; this is not the case for the self-employed.

The greatest challenge for the self-employed is, however, measuring their earnings, which is effectively profits. Self-employed earnings are defined as the revenue earned from their business less business expenses. So for a self-employed taxi driver, earnings are the fares earned less the costs of running the taxi, such as fuel and insurance. The need to track both incoming and (relevant) outgoings makes it difficult for some self-employed people to calculate their earnings.

In the UK, self-employed people must report their earnings for the purposes of income tax annually, and their tax liabilities are measured on the basis of their earnings over a year. The annual administrative process and challenge of separately tracking revenue and expenses means that it is very difficult to get more frequent measures of income.

Currently, the best measure of self-employed earnings is produced from the Family Resources Survey, an annual household income survey produced by the Department of Work and Pensions. This survey is focused on an in-depth understanding of households’ incomes and has sufficient sample size to provide a reliable overall estimate for the self-employed. The problem is that only annual information is available, with a lag of around 12 months between the year in question and the publication of the statistics.

ONS Resource

The Family Resources Survey has many interesting insights into earnings.

There is no obvious alternative data source, such as administrative data. The income tax data has a similar lag, as the self-employed are not required to report their income until 10 months after the end of the tax year in question. Some benefits require reporting of self-employment income on a more regular basis, but are not likely to be a representative sample of the whole population.

A further challenge with the self-employed is that many people are both an employee and self-employed, so to get an accurate picture of their earnings, you need to combine two data sources.

5.4.6 Zero hours and the ‘gig’ economy

The main concepts used to explain the labour market are long established, which allows us to compare economic trends over many decades. On the other hand, the economy and the labour market are constantly changing and new employment concepts and practices arise over time.

In Section 5.4.2, we discussed zero-hours contracts. In theory, any trend among employers to using more of them can be measured by asking new questions in the Labour Force Survey. But there is a practical problem: we know some people do not realise they are on zero-hours contracts.

Figure 5.18 shows the number of people working on zero-hours contracts as recorded in the Labour Force Survey. It is believed some of the increase was due to greater awareness of this type of contract, rather than an actual increase in the number of contracts on offer.

Figure 5.18 Level and rate of people aged 16 and over on zero-hours contracts

Level and rate of people aged 16 and over on zero-hours contracts

UK, October to December 2000 to 2022, not seasonally adjusted

Another trend is people working in what is known as the gig economy. This involves people using internet platforms to find someone willing to pay for a service they offer. At one end, this can be highly structured, with people offering taxi-like food delivery services through well-known brands. At the other end, there are smaller scale platforms that match people with specific skills—such as video editing—to those who need the particular expertise.

People who work in the gig economy are likely to be self-employed, and at one level this is nothing new. Most taxi drivers in the UK have always been self-employed, and there have always been freelance video editors and similar. So today’s gig economy workers—just like their ‘analogue’ forebears—show up in the self-employment statistics and are therefore properly counted. On the other hand, the growth in this sector and its effects on the wider labour market and the economy are of great interest to policy makers.

The challenge with the gig economy is how to ask questions in the Labour Force Survey to find out if people are participating in the gig economy. The term, while widespread, is not necessarily understood, and its definition—using electronic platforms to find work—is not particularly easy to grasp either. The Office for National Statistics has tested questions on the gig economy, with mixed results. As yet, we do not have a good measure of the role of the gig economy in the labour market.

5.5 The labour market and the macroeconomy

There is a strong link between the labour market and other macroeconomic indicators. Clearly it affects measures of household income, and also income inequality and poverty (Chapter 7). Chapter 6 shows how labour market statistics combine with measures of GDP to help us understand productivity.

But perhaps one of the most interesting relationships is the link between levels of unemployment and measures such as inflation or wage growth.

5.5.1 The Phillips curve: A trade-off between unemployment and wage growth?

The exact relationship between the labour market and the wider economy is complex, nuanced and widely debated. One way of summarising it is through the Phillips curve, which is an empirical relationship between prices or wages on one hand, and unemployment on the other.

This relationship is based on the reasoning that as unemployment gets lower—and therefore labour becomes more scarce—wages should rise more quickly, and in turn this feeds through to inflation.

As this chapter covers the labour market, we will focus on the wage Phillips curve. Figure 5.19 plots wage growth against the unemployment rate, with the series split between the period 2001 to 2012, 2013 to 2020 and 2021 to 2022.

Figure 5.19 Wage Phillips curve

Wage Phillips curve

UK, 2001 to 2022

In 2001, unemployment was around 5% and wage growth 4%; in the years immediately following, the picture was broadly the same. As the economy went into the downturn following the financial crisis, wage growth fell and unemployment rose, peaking at just over 8%.

The pattern in the most recent period is different from that which held in the early 2000s. Unemployment has fallen, and in 2019 was around one percentage point lower than in 2001, yet wage growth was also lower.

This is described as a ‘flattening’ (or perhaps more accurately ‘lowering’) of the wage Phillips curve and suggests that the trade-off between unemployment and wage growth—and, by extension, inflation—was less during the 2010s than in the previous decade. However, during the two years 2021–2022 wage growth has accelerated in line with the rapid increase in consumer price inflation, giving the impression that the wage Phillips curve has shifted upwards. Whether this shift is temporary or permanent remains to be seen but has reignited interest in the Phillips curve which some proponents argued had become a misleading or outdated concept.

5.5.2 The Phillips curve and the NAIRU

One answer to the flattening of the Phillips curve might be in a concept called the non-accelerating inflation rate of unemployment, or NAIRU.

This the lowest rate of unemployment that can be maintained without allowing wage growth or inflation to increase unsustainably. The NAIRU is an economic concept, therefore we cannot capture it in labour market statistics. It must be estimated, using those statistics, employing complex econometric techniques.

In the short term, the NAIRU is viewed as fixed. The issue for macroeconomic policymakers is how to get close to the NAIRU. Policymakers consider how much slack there is in the labour market to help determine this. In fact, one measure of the amount of slack in a labour market is the difference between current unemployment and the NAIRU.

The NAIRU is, however, not fixed in the long run. There is considerable debate over why this is the case, which is beyond the scope of this chapter. Nonetheless, the reduction in the NAIRU is seen as a worthwhile policy objective, as it would help to expand the productive potential of the economy.

An Organisation for Economic Co-operation and Development (OECD) study from 2008 listed the different factors that have been considered in explaining the NAIRU:

The list of recurrent candidates identified in the literature, albeit not always easy to measure properly, includes the unemployment benefit replacement rate, the tax wedge, union density, the level of the minimum wage, product market regulations, employment protection legislation, measures of skill mismatch, and the efficiency of active labour market policies and of the job matching process.2

Whatever the exact influences on the NAIRU are, the intuition is that a flexible labour market will be able to match workers to jobs quickly and efficiently. The more bottlenecks or rigidities there are in the labour market, the more likely it is that increases in labour demand will feed through to wages rather than increased employment.

Take for example, the unemployment benefit replace rate which measures how high out of work benefits are compared to earnings in work. If benefits out of work are relatively high, then the ‘reservation wage’—that is the lowest wage that would tempt someone to take a job—is also likely to be high. In this case it is argued that increases in labour demand are unlikely to feed through to lower unemployment, but just higher wages and inflation.

A falling NAIRU is therefore one explanation for the flattening of the Phillips curve discussed in Section 5.5.1.

5.6 Labour market statistics drive policy

Section 5.5 provided an excellent example of how labour market statistics affect macroeconomic policy discussions. Central banks and finance ministries grapple with issues such as Phillips curves and the NAIRU to understand what to do with monetary and fiscal policy.

The labour market also heavily affects other policy discussions. Sections 5.6.1 and 5.6.3 discuss two areas where labour market statistics inform wider policies. The first deals with the gender pay gap, and how the pay of men and women differs. The second is understanding low and high pay, which is closely connected with debates on wage and income inequality.

5.6.1 The gender pay gap

The gender pay gap captures the difference in pay rates between men and women. There is no single measure, and there are three separate statistics that are usually reported:

  • the hourly pay difference between men and women who work full-time
  • the hourly pay for those working part-time
  • the hourly pay difference for full- and part-time workers combined

The Office for National Statistics reports all three, although tends to lead on the pay gap for full-timers. This controls for the difference in pay rates between those, and relative proportions of the men and women, who work part- or full-time. All gender pay gap data are based on ASHE, as this has data on gender as well as pay rate and occupation of the person in question. Figure 5.20 shows the gender pay gap using these three measures.

Figure 5.20 Gender pay gap for median gross hourly earnings (excluding overtime)

Gender pay gap for median gross hourly earnings (excluding overtime)

UK, April 1997 to 2022

ONS Resource

More on the gender pay gap in 2022.

The chart shows that the gender pay gap for full-time workers in 2022 was positive 8.3%, which means that the median hourly wage rate of women is 8.3% lower than the median for men. On the other hand, for part-time workers the gap is negative 2.8%, meaning that the women’s median is 2.8% above that for men. When we pool full-and part-time workers together, we find a pay gap of positive 14.9%—the median for women is 14.9% lower than for men.

It might strike you as odd in that we have two sub-groups with pay gaps of positive 8.3% and negative 2.8% giving an overall pay gap of positive 14.9%. How can the average of two sub-groups give a higher figure than either of the sub-groups? This is an example of something known as Simpson’s Paradox, which is explored in ‘Decoding the gender pay gap: how a Bletchley Park codebreaker helped explain a strange paradox’.

The gender pay gap does not necessarily mean that a man and a woman doing the same job are paid differently, so it is not a measure of discrimination. Gender pay gaps can arise where more senior posts are filled by men, or where men are more likely to work in higher paid industries. Of course, these factors could arise from discrimination but could also arise for other reasons such as years of experience or skill level.

5.6.2 Measuring the gender pay gap

The focus that the Office for National Statistics puts on the gender pay gap for full-time workers means a better possibility of comparing like-for-like jobs. This position can be criticised from two opposing views.

  • More women than men work part-time. If an occupation becomes less well paid as the proportion of women in the workforce grows, then we should not necessarily separate part- and full-time workers. In econometric terms, we have a problem of endogeneity. This line of argument would suggest that the gender pay gap for full-time workers is an understatement of the overall gender pay gap. Alternatively, you might argue that those working in full-time posts gain more experience more quickly and therefore have higher human capital. In this case, you could argue that the full-time pay gap is a good measure.
  • Unobserved variables. On the other hand, some people would argue that the gender pay gap for full-time workers overstates any degree of discrimination or disadvantage. This could happen if there are characteristics that are objectively relevant in determining wage rates—such as years of experience—that are not controlled for.

Ultimately, this argument comes down to a decision about which factors are legitimate in explaining differences in wages, and which are not.

We can also argue that the focus on hourly pay also misses the point that working part-time and having time out of the labour market combines with the hourly pay gap to mean that women’s lifetime earnings are restricted. This possibly leads to lower pension income in retirement. In this way, some of the factors that we might control for are just wider aspects of women’s disadvantage in the workforce.

Studies have been done to see how much of the gender pay gap can be explained by various characteristics, such as occupation, hours of work and tenure. One such analysis suggested that around one-third of the pay gap could be explained by characteristics available on the ASHE dataset. This suggests a sizeable unexplained pay gap that could be discrimination or disadvantage, although some characteristics—such as educational attainment—have been excluded because of a lack of data.

The ASHE dataset allows us to explore how the gender pay gap varies between different groups of people. One particular area of interest is age. Figure 5.21 shows the pay gap for all full-time and part-time employees of different age groups. From the age of 18 to 39, the full- and part-time gender pay gaps are close to zero. The only reason that the overall pay gap exists for women under 40 is because more women work in part-time roles.

Over the age of 40, the full-time (and, to a lesser extent, part-time) gender pay gap increases notably. People have hypothesised that this might be to do with childbearing and the split of caring responsibilities in couples. It is argued that women are more likely than men to take time out of the labour market, take a part-time role or otherwise affect their careers when they have children. This then leads to lower earnings later in life. Elsewhere, for example, we see that women tend to commute shorter distances, perhaps evidence that they are prioritising their child caring responsibilities over work to a greater degree than men appear to do.

All employees 16–17 18–21 22–29 30–39 40–49 50–59 60+
Proportion of women who work part-time 40 91 64 25 36 39 41 61
Proportion of men who work part-time 15 79 48 14 10 10 12 28

Figure 5.21 Gender pay gap by age

Gender pay gap by age

UK, April 2022

We need to be careful with this interpretation, as the data used in this analysis are cross sectional: they look at people’s ages now rather than tracing individuals over their lifetimes. So the pattern we see is also consistent with a ‘cohort’ effect. Perhaps the women who are in the 40 to 49 age group now had a similar gender pay gap when they were in the 20 to 29 age group. In which case, the idea that the gender pay gap opens up around the time of childbearing is just an illusion created by the data.

Figure 5.22 shows the full-time gender pay gap for different age groups since 1997. There is some evidence of a cohort effect here as, for example, women in the 40 to 49 age group have seen a 10 percentage point fall in their gender pay gap since 1997. The effect is, however, relatively modest, and for some age categories—for example those aged 50 to 59—the effect is very muted.

On balance, therefore, there is evidence consistent with the view that a large part of the gender pay gap arises from childbearing, and the consequential losses of hours and pay. However, with something as complex as the gender pay gap, there is never likely to be a single explanation.

Figure 5.22 Gender pay gap for full-time median gross hourly earnings (excluding overtime)

Gender pay gap for full-time median gross hourly earnings (excluding overtime)

UK, April 1997 to 2022

5.6.3 High and low pay

Earnings form an important part of household incomes and therefore overall living standards. In turn this means they affect the distribution of incomes between different households. Wages therefore affect measures of inequality and poverty.

While earnings are important in determining household incomes, we need to bear in mind two very important points:

  • Income is shared. When we look at wages or earnings, we look at individuals, but when we look at household (or ‘family’) income, we are calculating the combined income of a number of people who may or may not be working. This is because some degree of income sharing is assumed within households; to understand the living standards of people within the household you should look at the incomes of all the people in that household.
  • Other sources of income. For some households, their main source of income may not be earnings. For the non-working households or retired households, wages or earnings are unlikely to be important in determining living standards.

These two points taken together mean that the lowest paid people are not necessarily living in the poorest households. One way of looking at this is to consider which individuals are paid at (or close to) measures of the minimum wage, and where they appear in the overall household income distribution.

In the UK, there are two types of minimum wage: the National Living Wage covers those workers aged 25 or older, and the National Minimum Wage those aged 16 to 24.

While those on the various minimum wages are to be found towards the bottom of the household income distribution, they are more spread out than one might imagine. Research by Brewer and Agostini, as shown in Figure 5.23, shows that around one-third of households where someone works on the National Living Wage are in the top half of the income distribution. This could arise where there is a couple with one person earning well above the National Living Wage, and the other person working on the National Living Wage.

Household incomes and inequality are discussed in full in Chapter 7.

Figure 5.23 Working-age families across the working-age income distribution, by employment and NMW/NLW status

Working-age families across the working-age income distribution, by employment and NMW/NLW status

UK, 2017 to 2018

Mike Brewer and Paola De Agostini (2017), The National Minimum Wage, the National Living Wage and the tax and benefit system, Institute of Social and Economic Research, University of Essex, using data from the Family Resources Survey, 2014 and 2015.

low pay
Per the OECD: where someone earnings below two-thirds of the median hourly or weekly wage rate.
high pay
Per the OECD: where someone earns more than 1.5 times the median hourly or weekly wage rate.

While the relationship between low pay and low household income is imperfect, low pay contributes to overall measures of inequality. There is, therefore, considerable interest in data on the distribution of earnings, which is available through ASHE. As with household income, there is a distinction between relative measures of high pay and low pay and absolute measures. Relative measures describe how well people are paid compared to others in the population, while absolute measures gauge pay against some specific standard of living, such as a measure of a ‘living wage’.

In terms of hourly pay, the convention is that you are low paid, if you earn less than two-thirds of the median UK wage and high paid, if you receive more than one-and-a-half times the median wage. In 2019, the UK median wage was £13.27. So you would be deemed to be low paid if you earned less than £8.85 (i.e. £13.27 × 2/3) per hour or high paid if your wage was above £19.91 (i.e. £ 13.27 × 1.5) per hour.

Figure 5.24 shows the proportion of employees who were classified as high or low paid on an hourly or weekly basis over the last 20 years or so. It is important to note that the low or high pay calculations are relative to the median in the year in question; high and low pay in 1997 are determined relative to the median in the year, and so on.

Figure 5.24 Proportion of high and low paid employee jobs for hourly pay and gross weekly pay, whole economy

Proportion of high and low paid employee jobs for hourly pay and gross weekly pay, whole economy

UK, 1997 to 2022

There has, until relatively recently, been considerable stability in the proportions of the workforce who are high and low paid. This is despite a rise in employment rates and considerable structural change in the labour market. In recent years, however, as the minimum wage rates have risen so there has been a quite dramatic fall in the proportion of employees who are low paid on an hourly basis.

Looking at the weekly measure of low pay, the fall is much more muted, reflecting that average hours for the lowest paid have fallen. This is an important reminder that the labour market can—and does—adjust hours as well as employment levels and wages. In that sense—and hopefully a message that has been clearly delivered in this chapter—it is very important to consider the labour market in totality.

5.7 Summary

This chapter covered the main concepts of measuring the labour market. We began with understanding the way in which unemployment, employment and economic activity are measured. Importantly, the concepts are anchored in theories of the labour market.

We then looked at labour market flows, that is people moving between labour market states. The numbers here are large relative to the respective stock measures, reminding us that the labour market is remarkably dynamic. A large number of people regularly move between work, unemployment and economic inactivity. Many people also regularly move between jobs.

The chapter also covered the measurement of wages, including how flows of people between labour market states affect overall wages. The final set of issues concern hours of work—the so-called intensive margin—and related concepts, such as underemployment.

These concepts link to wider issues. There remains some controversy over how the labour market is measured, so we discussed the nature of economic inactivity, how unemployment is defined, the role of administrative data, self-employed earnings and phenomena such as zero-hour contracts. These are also areas where labour market statistics need to adapt to ensure we are capturing relevant developments.

Finally, we looked at how important labour market statistics are in terms of relationships to other concepts measured in this book and also to wider policy. The trade-off between inflation and unemployment is captured by the Phillips curve and the closely related concept of the NAIRU. These are issues that macroeconomic policy makers consider regularly. But labour market statistics are also issues of interest in wider debates on issues, such as the gender pay gap and measuring inequality in pay.

There are perhaps two main messages from this discussion. Firstly, that the labour market and related statistics are hugely important and salient in their own right. Secondly, that the labour market is intimately connected with other statistics and economic concepts, such as inequality, productivity and macroeconomic policy.

5.8 Further reading


  1. Speech at the Oxford Economic Society, 14 November 2017 

  2. Christian Gianella, Isabell Koske, Elena Rusticelli and Olivier Chatal (2008), ‘What Drives the NAIRU? Evidence from a Panel of OECD Countries’, Economics Department Working Paper No. 649