Labour Force Survey quality update: July 2026

• Response levels for the Labour Force Survey (LFS) have shown clear improvement since 2023 because of interventions, including increases to survey incentives and interviewer numbers, to improve quality, with response levels for Waves 2 to 5 slightly above their pre-coronavirus (COVID-19) pandemic level in January to March 2026.

• An operational error occurred in May 2026, which led to temporary under-resourcing in LFS telephone collection operations; our analysis to date indicates that the error had a minimal impact on the headline estimates, although, it does suggest there is a small but noticeable impact on the estimates of average hours and so the total actual hours worked.

• Coherence between the LFS and other labour market data sources, although improving, remains a challenge, and we continue to monitor this closely; you can find detail on the coherence of the latest estimates in our Labour market overview: July 2026 bulletin.

• Caution is advised when assessing change over time periods (particularly those affected by operational changes) and when analysing more detailed estimates; as the quality of the LFS has improved, it is likely that these quality improvements have affected reported statistics.

• While single month and wave-based estimates can provide valuable additional context, they are best used alongside the headline three-month estimates; comparisons over three-month intervals and analyses using the wave structure can offer further insight into underlying trends.

• Pay As You Earn (PAYE) real-time information (RTI) estimates relative to the population have been published for the first time and show additional insight into the labour market; we welcome feedback from users.

• The Transformed Labour Force Survey (TLFS) remains the long-term solution for collecting labour market data; updates on progress and plans are published quarterly in our Progress and plans article series.

Household surveys, in the UK and in comparable countries, have been facing the challenge of falling response rates for many years. There have also been challenges in collecting and processing survey data since the start of the coronavirus (COVID-19) pandemic, and there are substantial quality concerns about Labour Force Survey (LFS) data collected in 2023. This led to the suspension of labour market publications based on LFS data between October 2023 and February 2024, and the withdrawal of accredited official statistics status for publications based on LFS and Annual Population Survey (APS) data from 2024 onwards. As we reported in the latest Office for National Statistics (ONS) surveys and economic statistics improvement plan, quarterly progress update, we are working to ensure our priority statistics are ready for assessment against the standards required for accredited official statistics status. We will provide an update on the approach for labour market statistics derived from the LFS (and in the future from the TLFS) when we publish the next Labour market transformation update in August.

In response to these concerns, several changes were made to the operation and processing of LFS data since late 2023. These are detailed in our Labour market transformation – update on progress and plans article series.

Since October 2023, the ONS has made several changes to address quality concerns with the LFS. The five-wave structure of the LFS means that some of these changes take at least 15 months to fully feed through into survey estimates and at least 18 months to feed through into measures of quarterly change.

Estimates from January to March 2025 include the full effect of the improvements in LFS data collection and sampling methods, introduced from January 2024 onwards. However, some volatility remains, particularly for estimates for mid-2023 and throughout 2024, and for granular breakdowns, where sample sizes are smaller. Caution should be taken when drawing conclusions from short-term changes, and we advise users to focus on long term movements in the data. We recommend using as part of our suite of labour market indicators, alongside workforce jobs, Claimant Count and Pay As You Earn (PAYE) Real Time Information (RTI) estimates.

Throughout this article, we refer to LFS data for Great Britain and for the UK. The operational responsibility for data collection is split between the ONS for Great Britain, and the Northern Ireland Statistics and Research Agency (NISRA) for Northern Ireland. The ONS has responsibility for publishing statistics at a UK level. Data regarding the collection of the survey often refer to Great Britain, while references to published statistical measures often refer to the UK as a whole.

This article provides an update to our most recent Labour Force Survey quality update: April 2026 article. It also:

• provides an update on the impact of the LFS operational issue announced in our blog post in June 2026

• describes that analysis that is possible using the LFS single-month estimates

• presents RTI age analysis relative to the population

The article focuses on the LFS. However, given the linked design of the two surveys, changes to the LFS also affect the size and quality of the APS. We intend to publish an update on APS quality in September 2026.

Coherence between the LFS and other data sources has improved, but remains a challenge, and we continue to monitor this closely. Our latest views on interpreting LFS data and the labour market narrative will remain in our monthly Labour market overview bulletin.

In this section, we cover Labour Force Survey (LFS) data up to January to March 2026, to align with the latest data covered in our LFS performance and quality monitoring report: January to March 2026, published in May 2026.

The changes referred to in Section 2: Background to Labour Force Survey quality continue to have a clear positive effect on the number of achieved responses to the LFS. The overall level of response has increased steadily since the low point in July to September 2023.

Considering the number of responses for the UK, including imputation (which is the same as the number of cases available to analyse in a dataset), there were 80,078 responses to the LFS in January to March 2026. This is an increase of 35,840 from the low point in July to September 2023 but 3,984 responses below the figure seen in October to December 2019.

Breaking this down by wave of response, as reported in our LFS performance and quality monitoring report, we have seen UK response levels for Wave 1 almost recover to pre-coronavirus (COVID-19) pandemic levels. There were 8,586 Wave 1 household interviews in January to March 2026, compared with 8,963 in October to December 2019. Waves 2 to 5 have increased to above pre-pandemic levels, with 27,285 responses (including imputed households) in the latest period, compared with 26,823 at the end of 2019.

We have also seen improvements to response rates to the LFS, although they are more moderate by this measure and are below pre-pandemic rates. Recovery in the level of responses is partly because of an increase in the size of the selected sample. Further information on survey response levels and response rates can be found in our LFS performance and quality monitoring report.

The higher LFS response level helps to increase the precision in our estimates. This is shown by the confidence intervals for our headline labour supply estimates, which are presented in Figure 1. Across all three headline rates, these have narrowed since mid- to late-2023, which suggests improved precision compared with the periods where response rates were at their lowest. Confidence intervals remain wider than before the pandemic for the unemployment rate, however confidence intervals are back to pre-pandemic ranges for both the employment and inactivity rates.

Detailed estimates will continue to see greater volatility compared with more aggregated measures. However, this has always been a feature of LFS-based analysis to some extent. It is also likely that variation in response rates has affected non-response bias in the survey results. For example, we can consider estimates of redundancies or estimates of young people not in employment, education or training. Both concepts focus on much smaller population subgroups compared with the headline rates, and therefore continue to see relatively large quarterly changes, despite improved sample sizes.

The size of the two-quarter longitudinal dataset has increased since its low point in October to December 2023 and reached 27,623 in January to March 2026, an increase of 15,116 over that period. This is now above the pre-pandemic level in October to December 2019, following an increase in the number of interviewers working on Waves 2 to 5 during 2025.

An operational error occurred in May 2026, which led to temporary under-resourcing in Labour Force Survey (LFS) telephone collection operations. The direct impact was on response levels between 3 May and 10 June, with some residual effects persisting until 17 June. The issue primarily affected Waves 2 to 5 of the survey. The error reflects the complexities of running two large survey operations in parallel, but should have been detected earlier. The issue has now been corrected and further information on the mitigations put in place were set out in our Learning from an operational issue temporarily impacting LFS data blog post. An exercise about lessons learned is being carried out and will be completed later this month. We will report on the lessons learned alongside the update on progress transitioning to the TLFS in August.

The first datasets to be affected are those for the March to May 2026 rolling quarter, which were published on 21 July. The size of the individual level rolling dataset (UK including imputation) is broadly similar to previous rolling 3-month periods at around 80,000. This is because the impact of the operational issue is an increased level of imputation in that dataset compared with the December to February 3-month period.

While it is difficult to quantify the precise impact of the issue, we have carried out analysis to help users understand the potential effect on headline estimates. An overview of this analysis and our overall assessment of the impact on estimates is outlined in the following sections.

Simulation using previous periods

We simulated what the impact would have been had a similar issue occurred in February to April 2026 (the most recent three-month period not affected by the issue) as well as March to May 2024 and March to May 2025 (the two most recent three-month periods subject to the same seasonal pattern as March to May 2026). The impact of the additional imputation depends both on the time of the year and the evolution of the labour market over the affected period. 

In each of the three study periods, we imputed the employment status and working hours of a random subsample of Wave 2 to 5 respondents, by carrying forward their previous-wave responses. The likelihood of respondents being included in the subsample was proportional to the estimated probability of someone with the same set of personal characteristics (including previous-wave employment status and working hours) not responding in March to May 2026.

The subsample was drawn from the final month of each of the three-month study periods and the proportion of respondents included was the same as that affected by the operational issue, this enabled us to replicate the additional non-response observed in March to May 2026.

We estimated the resulting employment, unemployment and economic inactivity rates, and average actual weekly hours worked, and compared them with the published non-seasonally adjusted (NSA) estimates. This subsampling, imputation and estimation approach was repeated 10,000 times for each of the three study periods, providing an indication of the range of impacts that might be expected, had different respondents been affected by the operational issue. This provides an indication of both the average effect of the issue on headline estimates, as well as the change in sampling variability resulting from the error.

In all three study periods, the average difference between the simulated and published quarterly NSA estimates across the 10,000 iterations are close to zero for headline employment, unemployment and inactivity rates, with the differences rarely exceeding 0.1 percentage points. Differences are slightly bigger for average actual weekly hours, which may reflect the affected period coinciding with the late May Bank Holiday and half term (with missing observations imputed from a period less affected by holidays), but the difference is mainly less than or 0.1 hours in either direction (Table 1). In all cases, the 95% confidence interval around the average of the simulated NSA estimates is small relative to the 95% confidence interval (the sampling variation) around the corresponding published estimate.

When focussing on the third month of each of the three-month study periods (from which the subsamples are drawn), average differences between the simulated and published single-month NSA estimates are slightly larger, reaching:

• 0.10 percentage points for employment (February to April 2026)

• negative 0.08 percentage points for unemployment (February to April 2026)

• negative 0.04 percentage points for inactivity (March to May 2024)

• negative 0.16 hours for average actual weekly hours worked (March to May 2025)

However, single-month estimates are not our headline measure, as they are inherently more volatile and subject to greater sampling error than the three-month estimates. This is because sample sizes are smaller, and differences between the simulated and published single-month estimates should be interpreted in this context.

Full results can be found in our accompanying dataset.

Counterfactual reweighting approach

As it was not possible to identify the exact households affected by the operational issue, a counterfactual reweighting approach was used to assess how labour market estimates and associated measures of precision might have differed, had the additional roll-forward (RF) cases not been included in the March to May 2026 dataset.

To represent this counterfactual scenario, we identified a group of potentially affected RF cases and removed a random sample of approximately 4,000 cases, reflecting the estimated number of additional RF cases. This approach assumes that the affected households had similar characteristics to the overall sample and therefore may not fully reflect the operational issue in practice, if the households affected differed systematically from other respondents.

The remaining sample was then reweighted using the standard LFS weighting methodology and a revised set of labour market estimates was produced. We compared estimates of employment, unemployment and inactivity, together with associated measures of statistical precision, between the original dataset and the counterfactual scenario.

The analysis showed that the unweighted and weighted characteristics of the affected group were broadly similar to those of the overall sample, and that headline labour market estimates changed very little when the simulated RF cases were removed. Differences in precision measures were also small, suggesting that the additional RF cases are unlikely to have had a material effect on headline labour market statistics, although the results should be interpreted as an indication of potential impact rather than a direct estimate of the operational issue's total effect.

Overall assessment and next steps

In summary, our analysis to date indicates that the operational error has had minimal impact on the headline estimates of employment, unemployment and inactivity rates. The impact on average actual weekly hours is slightly larger and, when scaled up to produce estimates of total hours worked, the analysis suggests there is a small but noticeable impact.

We will undertake further analysis to assess the potential impact of this issue for the period April to June 2026, including repeating the counterfactual reweighting analysis for total actual hours worked across both periods. We will also assess the subsequent impact on hourly productivity estimates, and engage with users to explore whether there is merit in correcting for the estimated effect on average hours ahead of the publication of the next productivity estimates in August. Further details of the analysis completed to date, together with findings from any subsequent analysis, will be published in a future article.

It is also worth noting that while our analysis has focused on the quarterly datasets, the operational error will affect the longitudinal dataset, which covers two quarters and does not include the additional roll-forward cases. As a result, the longitudinal dataset will be smaller and is likely to lead to greater variability in the estimates.

The Labour Force Survey (LFS) is designed to produce three‑month estimates, which remain the most robust and reliable measure of individuals' labour market status. Alongside these, we also publish single‑month estimates and wave-based analysis as additional, more detailed views of the data. These outputs can provide useful insights, but they require careful interpretation. This section explains what these estimates represent, what their limitations are, and how they can be used appropriately.

The LFS operates using a rotational sampling design. Once a household is selected, it is interviewed five times at three‑month (13‑week) intervals, meaning it remains in the survey for around a year.

These interviews are referred to as Waves 1 to 5, with Wave 1 representing the first interview and Wave 5 the final one. Each quarter, a new group of households enters the sample at Wave 1, while existing households move on to the next wave, and those in Wave 5 leave the survey. As a result, at any point in time, around one‑fifth of the sample is in each wave, and there is approximately 80% overlap between samples in consecutive quarters.

This overlap is an important feature of the design, as it helps to stabilise estimates and makes comparisons over time more reliable. It also supports longitudinal analysis, allowing us to measure how people move into and out of different labour market statuses over time.

What this means for single-month estimates

This design has important implications for how single-month estimates should be interpreted. As households are interviewed at three-month intervals, those included in one month will not appear in the sample in the next two months. For example, households interviewed in January will not be included in February or March, but a large proportion of them, around 80%, will be interviewed again in April. This means that consecutive months are based on entirely different samples, with no overlap between them.

The sample design also means that, in any given month, not every address in the country is eligible for selection. This is to allow interviewer workloads to be me more geographically clustered in any period, allowing them to move more efficiently between selected addresses. This can lead to an uneven distribution of household types within any monthly sample, so single-month estimates should not be interpreted as fully representative of the UK population on their own.

Consequently, comparisons between consecutive months are particularly susceptible to volatility. Changes observed between one month and the next are likely to reflect sampling variability rather than genuine changes in the labour market. Given the relatively small sample size at a single-month level, this variability can be pronounced, and month-to-month movements should therefore be treated with caution. By contrast, comparing a month with the same month three months earlier, such as April with January, draws on samples that largely overlap. This provides a more stable basis for comparison and is more likely to reflect meaningful changes over time, albeit still with large volatility because of the small sample sizes.

Wave analysis

Wave analysis offers an additional way to understand movements in the data by focusing on how the same households change between interviews. For example, households first interviewed in January at Wave 1 will be interviewed again in April at Wave 2. By examining responses across these waves, it is possible to track changes in labour market status for the same group of individuals over time. This helps to distinguish real changes from those resulting from differences in sample composition. However, it should be remembered that there is also an element of attrition between waves, with some movement reflecting those households dropping out of the survey.

Looking at the data from a wave perspective also provides insight into the effects of the survey's rotation. Each quarter, new households entering the survey replace those leaving after Wave 5. Comparing the characteristics of these incoming and outgoing groups can also help identify whether changes in headline estimates are being influenced by this rotation, rather than underlying shifts in the labour market itself.

Overall, while single‑month and wave-based estimates can provide valuable additional context, they are best used alongside the headline three‑month estimates. Single‑month movements should be interpreted carefully, particularly when comparing consecutive months, whereas comparisons over three‑month intervals and analyses using the wave structure can offer further insight into underlying trends.

Further detail about single-month estimates can be found in our Labour Force Survey Single Month Estimates, September 2011 article.

HM Revenue and Customs (HMRC) and the Office for National Statistics (ONS) publish estimates of Pay As You Earn (PAYE) Real Time Information (RTI) payrolled employees monthly, including a breakdown by age group. These estimates are presented as levels.

The early estimate of payrolled employees for June 2026 decreased by 71,000 (0.2%) on the year to 30.3 million. The 25-to-34-years age group has seen the largest annual falls in payrolled employees, while the 65-years-and-over age group has seen the greatest increase. Figures for June should be treated as provisional estimates and are likely to be revised when more data are received next month. 

To fully understand changes in payrolled employees by age, it is important to look at the data relative to the population, to account for the difference in the size of the population by the different age groups.

We have used ONS population estimates and adjusted projections to produce RTI estimates of payrolled employment as a percentage of the population – referred to here as RTI rates. The population estimates are based on mid-year population estimates for 2011 to 2024 (as published in September 2025) and mid-year population projections for 2025 to 2030.

The RTI rate differs from the LFS employment rate, which includes other types of employment, such as self-employment, not just payrolled employees. There are other differences between LFS and RTI, which are summarised in our Comparison of Labour Market data sources article.  

When looking at payrolled employees for those aged under 65 years as a proportion of the population, as shown in Figure 2, the rate has generally been falling since March 2023 and stands at 66.3% in June 2026. It was last lower in August 2021 when it was 66.2%.

Notes Payrolled employees for those under 65 years old as a percentage of the population aged 16 to 64 years

When looking at rates for the different age groups, as shown in Figure 3, the 65 years and over age group shows the greatest annual increase, similar to the pattern shown in the levels. However, when looking at the largest falls, it is the younger age group (those aged under 25 years) that has seen the largest annual falls in the rates in more recent periods.

We are publishing these RTI rates for the first time, following feedback from stakeholders, but would welcome feedback before making this information available more regularly.

We plan to work closely with the ONS population team (to ensure timely updates to the population data we use) and with the HMRC RTI team (to ensure we meet user needs).

The Labour Force Survey (LFS) remains the lead source of labour market data and continues to run alongside the Transformed Labour Force Survey (TLFS) while development continues. We aim to transition from the LFS to the TLFS for our published headline labour market statistics in 2027. Updates on progress and plans will be published quarterly in our Labour market transformation article; the next update is planned for August 2026.

We are continuing to monitor a range of characteristics to understand how the composition of respondents has shifted over time. In the latest quarter (January to March 2026), we are continuing to see the modest changes shown in our previous LFS quality updates. In general, there is ongoing convergence towards Census 2021 estimates, although this varies across the characteristics analysed. We will review the analysis quarterly and provide updates in our future articles. 

As highlighted in Section 6: Claimant Count of our Labour Force Survey quality update: April 2026 article, the Claimant Count has developed a recent pattern of downward revisions – usually around 20,000 per month. This is because new Universal Credit claimants are initially placed in the "Searching for Work" group but removed from it once their claims have been fully assessed.

In May 2026, the Department for Work and Pensions (DWP) published new information on the employment status by Universal Credit conditionality regime, as part of its Universal Credit quarterly statistics, 29 April 2013 to 12 February 2026. These data allow users to see whether those in the "Searching for Work" conditionality are in any form of Pay As You Earn (PAYE) employment and/or self-employment. We are continuing to work with the DWP to develop a closer reconciliation of the Claimant Count with LFS unemployment data, and will publish the findings in a future article.