![\"\"](\"https:\/\/www.newsbeep.com\/au\/wp-content\/uploads\/2025\/08\/figure-1-scaled.jpg\")
<\/a><\/p>\nIf the LFS is right about the unemployment rate but wrong about the employment rate, there are two implications. First, the survey bias that has plagued the employment rate (people without jobs being more likely to answer the survey, in a way that the ONS can\u2019t control for) must be roughly the same, but in the opposite direction, for those out of work. Second, if the unemployment rate is rising more slowly than the employment rate is falling, then the inactivity rate must be rising too: the trends in employment and unemployment suggest that inactivity has risen by around 1 percentage point over the past two years, rather than fallen by 1 percentage point in the past year as the official data suggests (see the right-hand panel of Figure 2). If the inactivity rate is indeed rising rather than falling, this would be a major concern for policy makers trying to reduce inactivity and would make encouraging labour force participation more urgent than at any point since the height of the pandemic.<\/p>\n
Multiple indicators suggest the labour market is both loose and loosening<\/p>\n So, unemployment is rising in line with the official data. But labour market tightness is about more than unemployment alone \u2013 it depends not only on the size of the pool of under- and unemployed workers looking to work more, but also how many workers firms are looking to hire, and how easy it is for jobseekers and employers to find each other. For example, even if unemployment is low, the labour market can be loose (at least temporarily) if firms do not have very many vacancies to fill.<\/p>\n We can use the same broad range of data used to estimate the unemployment rate in Figure 1 to assess labour market tightness. Figure 3 plots each of these indicators normalised to their 2019 average level, and with signs reversed where necessary so that a higher number indicates more labour market slack. For example, wage growth measures have had their signs flipped, because slower wage growth is normally associated with more slack.) But for a full picture of labour market tightness we must look beyond labour utilisation, to how hard firms are finding it to retain workers and fill vacancies (the bottom two panels of Figure 3). The overall picture is of a labour market with more slack opening up \u2013 somewhat looser than it was both last year and in 2019, a point when the degree of labour market tightness appeared sustainable. Of the 12 indicators plotted in Figure 3, all but two (the redundancy rate and the quit rate) suggest the labour market has more slack than a year ago. Along with weak utilisation, hiring conditions (as measured by the Bank of England\u2019s agents), the vacancy rate, the quit rate and employment growth all suggest that the labour market has loosened over the past year. A minority of indicators suggest similar conditions to 2019, including the redundancy rate (0.36 per cent of workers reported were made redundant in the three months to June, compared to 0.38 per cent on average in 2019). Finally, the top-right panel of Figure 3 shows that nominal wages are still growing at a rate consistent with a hot labour market \u2013 something we return to below. In Figure 4, we combine the indicators shown above into a single measure of labour market tightness, using a statistical model that assumes that all the series are driven by both underlying tightness and their own unique set of factors, none of which we directly observe. It suggests the labour market has more slack now than immediately before the pandemic (the slack measure is 0.7 standard deviations above its 2019 level, based on the standard deviation of the series after 2000), but remains a long way short of the amount of slack that opened up during the financial crisis and Covid periods (when the measure was two standard deviations above its 2019 level, again based on the standard deviation of the series after 2000).<\/p>\n Figure 4 also plots the unemployment rate, scaled on the same basis, and this suggests that the unemployment rate currently gives the same steer on the amount of slack in the labour market relative to 2019 as the broader measure of slack. Reasonable adjustments to the statistical model can change the precise numbers that come out, but the key takeaway \u2013 that the labour market is looser now than last year or 2019, but not as loose as after the financial crisis or during the pandemic \u2013 remains.<\/p>\n Two other standard measures of labour market tightness, shown in Figure 5, paint a similar picture. The left panel plots vacancies and unemployment. A move towards the bottom right of this chart (higher unemployment and lower vacancies) represents a loosening of the labour market, and that is what we have seen since 2022. The right panel shows the implied time it takes to fill a vacancy \u2013 a reasonably direct measure of tightness \u2013 which has also fallen sharply since 2022 and is now well below its pre-pandemic level.<\/p>\n The strength of wage growth remains a puzzle<\/p>\n The vacancy-filling rate in Figure 5 directly influences the marginal cost of production: if it takes a long time to find a new worker, it is expensive and difficult to increase output. So, to sum up, the Labour Force Survey appears to be giving a good read on unemployment at the moment. This is good news for the credibility of UK statistics, but bad news for households, with around 200,000 more people looking for work than last year. And it is also concerning for policy makers. The implied rise in the inactivity rate is especially concerning for a Government with a target to boost employment and GDP growth. And wage growth that looks too strong to be sustainable is likely to have slowed any adjustment to interest rates that the Bank of England would have taken in response to labour market loosening.<\/p>\n \u00a0<\/p>\n Box 1: Measuring unemployment and labour market tightness using many data series<\/p>\n To estimate underlying unemployment, we specify a Dynamic Factor Model in which each of the panel of labour market series in Figure 3 is driven by the unemployment rate and a series-specific factor. The unemployment rate is observed perfectly (with LFS unemployment) up to 2020 and then with error after that. The model looks for the unemployment rate that best explains the full set of observed data, including the post-2020 LFS unemployment rate. The unemployment rate is from the LFS. \u00a0<\/p>\n Lifting the lid | Three trends in focus<\/p>\n Falls in the employment rate are no longer being driven entirely by population growth Administrative data shows a decline in the number of payrolled employees over the past year, but this is not reflected in other datasets<\/p>\n Since the onset of the Covid-19 pandemic, different data sources have painted very different pictures of what has happened to employment. Figure 8 compares employee figures from three key sources: the Labour Force Survey (LFS), which underlies the UK\u2019s headline employment statistics but has been beset by significant challenges; Workforce Jobs, which is based largely on a survey of businesses; and HMRC\u2019s administrative payroll data. According to HMRC, the number of payrolled employees has been gradually falling since last summer \u2013 down by 165,000 from its peak in October 2024 \u2013 driving the fall in overall employment shown in Figure 8. But this fall is not reflected in the other two data sources, including Workforce Jobs, which has tended to line up well with payroll data.<\/a><\/p>\n
The top-left panel of Figure 3 begins by looking at indicators of labour utilisation. While the LFS unemployment rate is an important indicator of this, it is not the only one: other indicators include the Claimant Count, the proportion of workers who want more hours (underemployment rather than unemployment) and the employment rate. A fall in labour demand will affect them all, but to a different extent, and they will all be driven by their own unique set of factors.
For example, the employment rate depends on labour supply as well as labour demand, while the Claimant Count does not measure unemployed people who are not claiming benefits, and since the rollout of Universal Credit also includes people who are working but are required to look for more hours, which has likely distorted the trend. Nevertheless, all these indicators of labour utilisation suggest it is lower now than in both 2024 and 2019, although only marginally so for the underemployment rate.<\/p>\n
Overall, from the point of view of workers, these indicators are good news for those who have a job \u2013 the risk of job loss isn\u2019t rising, and pay growth remains relatively strong. But labour market conditions are getting worse for the growing number of unemployed (up by 206,000 over the past year according to the LFS), because jobs growth and hiring have slowed considerably.<\/p>\n<\/a><\/p>\n<\/a><\/p>\n<\/a><\/p>\n
But tightness is also a driver of another key determinant of production costs \u2013 pay. As we would expect, pay growth has been decelerating as the labour market has loosened. But Figure 6 shows that private sector pay growth remains about 1 percentage point higher than we would expect given the overall tightness of the labour market. Along the same lines, the Bank of England also thinks that there is a high \u2018unexplained\u2019 component to wage growth at the moment \u2013 i.e. not explained by productivity, slack or inflation expectations. ONS data suggests that changes in the composition of the workforce towards more highly-paid sectors may only be contributing around 0.2 percentage points to annual pay growth \u2013 so much of the puzzle remains. The relationship in Figure 6, which compares private sector wage growth in each period to labour market tightness a year earlier, suggests that wage growth is set to continue falling, consistent with the Bank\u2019s latest Monetary Policy Report and the Decision Maker Panel survey.<\/p>\n<\/a><\/p>\n
To estimate labour market tightness, once again we specify a Dynamic Factor Model with each of the observable series being driven by both zero-drift measurement errors, an unobserved, non-stationary common component that represents labour market tightness, and series-specific drift term that represents structural changes in the relationship between the observable variables and labour market tightness (e.g. trend inflation, the NAIRU).
The data sources plotted in Figure 3 and used in the Dynamic Factor Models are as follows:<\/p>\n
\nThe Claimant Count rate is the number of people claiming unemployment-related benefits compared to the number of jobs.
\n\u2018Want more hours\u2019 refers to under-employment, from the LFS.
\nEmployment rate is based on RF estimates using tax and population data.
\nWage growth measures are annual growth in private sector regular pay from Average Weekly Earnings.
\nBank of England estimates of underlying nominal wage growth are used during 2020-2022 when data affected by furlough.
\n Real earnings growth are deflated by CPIH.
\nThe vacancy rate is ONS vacancies divided by payrolled employees.
\n\u2018Redundancies\u2019 refers to the redundancy rate from the LFS \u2013 the number of people reporting having been made redundant in the past three months divided by the number of employees in the previous quarter.
\nEmployment growth is as estimated by RF based on tax data.
\nThe quit rate is the number of job moves where resignation is given as the reason, divided by the number of jobs.
\nHiring difficulty measures are from the Bank of England agents and the BCC\u2019s Quarterly Economic Survey.<\/p>\n
As set out in this Outlook\u2019s spotlight, the labour market is loosening, and one sign of this is a falling employment rate. The employment rate \u2013 calculated using administrative data due to issues with the Labour Force Survey \u2013 has been falling year-on-year since October 2023. But the nature of the fall has changed since early 2025 (Figure 7). Up until February, the employment rate was falling because the number of people in work and population were both growing, but the latter more quickly. So as the orange bars in Figure 7 show, the falling employment rate was driven by population change. But since then, not only has the population continued to grow quickly, but the absolute number of jobs has also started to fall. This has resulted in a negative contribution from employment (the blue bars in Figure 7) in addition to the negative population effect. This is worrying because absolute falls in employment are not normally seen outside of recessions.<\/p>\n<\/a><\/p>\n
One possible explanation could be more people taking on multiple jobs \u2013 which would raise the number of employee jobs (which Workforce Jobs measures) but not the number of people who are employees. (LFS data shows a rise in the number of workers with a second job, albeit measured with uncertainty.) It is also possible that the divergence between data sources is temporary. Workforce Jobs figures are currently only available up to Q1 2025, and the Q2 data, set to be published in September, could align more closely with payroll data.<\/p>\n