Labour market transformation – update on progress and plans: July 2024

• In our last update we set out our plans to: analyse the latest available TLFS data alongside our expert peer reviewers; undertake an independent academic review of TLFS methodology; and initiate a new Stakeholder Advisory Panel for labour market statistics, so that we can best utilise external expertise.

• Collectively these actions provided a clear picture of the current quality of the survey; while we have made good progress with the TLFS, with a larger sample and higher response rates, there are issues that remain before we can transition and further steps are required so that the TLFS can reach the quality necessary for users.

• As a result, we will extend the dual run of the TLFS and LFS to provide further quarters of data for comparison, including accounting for any seasonal differences, which will help increase user readiness and confidence.

• We will test further design improvements within the TLFS; we recognise that any significant changes introduced to the TLFS would require an additional period of dual running with the LFS, to ensure stable and comparable data.

• During the next six months, further quarters of TLFS data, survey test results and ongoing user feedback will inform our decision-making on transition; we will keep users informed of progress and will report back in the first quarter of next year.

• While we further develop the TLFS, we will continue to use the LFS as our lead measure of the labour market.

• This article provides the latest information on the positive impact the LFS recovery plan has had; the LFS will be reweighted using updated population information by the end of the year to improve quality and respond to user needs.

We have been developing a transformed online-first version of the Labour Force Survey (LFS) as the long-term solution to falling response rates and quality challenges on the LFS. The ambition of the Transformed Labour Force Survey (TLFS) is to allow a more adaptive and responsive survey to meet user needs, enhance respondent experience and improve the quality of our labour market statistics.

High-quality labour market statistics are of strategic importance to the UK. We are committed to robust quality assurance of the new survey, and close engagement with users to ensure they will be ready and confident to transition onto the TLFS. This has been the focus of our activity since our previous update in April 2024. We shared a second quarter of data from the new TLFS with expert peer reviewers and gathered feedback, completed an independent academic review of the TLFS, and established a new external expert Stakeholder Advisory Panel for labour market statistics.

This recent work has provided evidence that we can have confidence in the original concept of the TLFS. Compared with the LFS, the TLFS provides a larger achieved sample, a higher response rate (with reduced variability in household response patterns), and the potential for more stable weighted outputs.

However, this review point has highlighted the challenges that remain before we are able to transition confidently and securely. It identified the further steps required, so that the TLFS can reach the quality necessary for users. Users have stressed the importance of an extended period of dual running of the TLFS and LFS to enable data comparison, and to allow assessment of the whole range of datasets, including flows, annual and household data. They have indicated that at least five quarters of stable data will strengthen quality assurance across the range of outputs, including accounting for any seasonal behaviour. This extended dual run was also recommended by the academic review.

This work has also identified some challenges with the TLFS data that require further investigation and analysis. These include a bias in the TLFS response towards older age groups, who are more likely to complete the survey, and also more likely to remain in the survey for the second and subsequent follow-ups.

There is also an increased level of partial TLFS responses, when compared with the LFS. This is because of the online nature of the survey, which takes two main forms. First, some individuals drop out of the survey before completion. While we obtain enough information to assess whether they are in employment, unemployed or economically inactive, there can be data missing about other characteristics such as level of education or whether they have a disability. Second, for some households with more than one person, we are only gaining responses for a single person when the TLFS aims to include information from everyone in a selected household.

Furthermore, collecting certain complex variables is challenging when moving from an interviewer led survey to an online collection. For example, to capture a responder's occupation, an interviewer may ask some follow up questions to accurately classify their role. For the TLFS we rely on a text response from responders, which can lack detail and lead to substantial differences between TLFS and LFS estimates by occupation. There is a similar challenge for establishing which industry to allocate a respondent. 

To ensure that we can fully investigate these challenges, and provide more comparable data for users, we will extend the dual run of the TLFS and LFS to enable longitudinal flows, and annual and household datasets to be shared with users. Additional quarters of data also allow any differences in the reporting of seasonal labour market behaviour between the two surveys to be understood and modelled.

We will also conduct a series of discrete online design tests in the autumn to assess the impact of a shorter labour market focused variant of the TLFS questionnaire, along with new questions to address bias, partial responses and collection of complex variables. These tests will be entirely separate from the main TLFS, which will continue to run without change during this period.

In the New Year, we will assess the results from the online tests and decide whether to implement changes into the ongoing dual run. The tests will be evaluated against specific criteria and compared with the performance of the main TLFS survey. The anticipated benefits of improved response and reduced bias will be weighed against the risk of de-stabilising the comparability of the dual run dataset. We recognise that following any new changes to the TLFS, users will wish to see a further extended period of dual running with the LFS to ensure data comparability. However, we will take opportunities to consult with users about timings, and whether we have sufficient stable data to enable transition.

The content for the shorter test has been informed by previous user engagement activity. While the testing is being carried out, we will engage users on what the final content of a shorter core labour market TLFS should be, so we are ready to implement this if the testing indicates the quality benefits outweigh the downsides. We will also ensure that data not covered by the shorter TLFS will continue to be collected through other means.

We have developed detailed readiness criteria for the TLFS, which will be regularly assessed. At a high level these are:

1. Statistical and data quality - the TLFS can produce all important labour market content within the agreed quality tolerances, with a sufficient stable time series and with explainable differences to the LFS.

2. User confidence and readiness - users are prepared for the transition to the TLFS and are confident with the data produced.

3. Operational readiness - a sustainable operational and statistical process to user accepted labour market publication schedules.

During the next six months we will keep users informed of progress on the next stages of this project and will report back in the first quarter of next year.

The Labour Force Survey (LFS) recovery plan is continuing to progress well. In the period from July to September 2023, before the recovery plan started, we achieved 14,180 completed LFS household interviews. This increased to 15,187 from October to December 2023, when we started implementing the recovery plan. This further increased to 17,053 in January to March 2024, when we boosted the main LFS sample.

In the latest data collection period from March to May 2024, the effects of the boost began to feed through to Wave 2, and have helped increase the overall number of completed LFS household responses to 17,885.

As we are continuing to publish the LFS, we plan to reweight the LFS data based on the population projections published in January this year. This has followed a strong steer from important stakeholders that this would improve the quality of the LFS estimates by bringing them into line with our latest assessment of population numbers, making them more comparable with other data sources. The same population projections will also be applied to the TLFS weighting, so that future data shares of LFS and TLFS will be comparable.

Reweighting is a complex process, which involves applying new weights to every response for each period of the data across a number of datasets. Balancing the need for data quality with available resources, we have decided to limit the scope of the reweighting exercise to the LFS person data. This reweighting is planned to cover the period back to 2019, at which point there will be a structural break in most series. We plan to introduce the reweighted series by the end of 2024. We will provide further updates ahead of this being implemented into the LFS-based statistics, including recommendations on how best to reconcile LFS outputs.

As more data become available, and in line with user requirements, we expect to reweight the LFS again in the future. We will continue to engage with users on their priorities for reweighting, communicate our plans in advance, and provide notice of likely changes to the outputs.