1. Background

The Labour Force Survey (LFS) is ONS’s premier social survey. However, in common with many similar social surveys round the world, particularly in countries without a legal compulsion for households to respond, the response rate to the LFS has been falling steadily since the turn of the century. The rate of this decline accelerated steeply with the onset of the COVID pandemic, and there was no clear recovery post-COVID. The LFS is an interview only longitudinal survey (Wave 1 in person, then telephone for Waves 2 to 5) and the respondent burden, in terms of time required to provide all the information requested, can be high. The ONS recognised this issue early on, and in 2014 published a National Statistics Quality Review of the Labour Force Survey that recommended a move to an online data collection. Following a period of development and testing, and in response to the COVID pandemic, the ONS went live with a replacement for the LFS in March 2020. This is the TLFS, with a much larger Wave 1 sample and an online first data collection methodology.

Unfortunately, the changeover from the LFS to the TLFS has not gone smoothly, with analysis revealing significant differences in important labour force estimates output by the two surveys, and TLFS response levels that were not as good as had been anticipated. Both these issues were identified as most likely a consequence of the fact that the TLFS was attempting to collect online essentially all the information that the LFS currently collects via personal interview. To break this deadlock, the ONS is now proposing to scale back the information collected in the longitudinal Core component of the TLFS to headline labour force data, and to use the cross-sectional Plus component of the TLFS as the main vehicle for collecting the more detailed data that is used in annual LFS outputs like the APS. But these changes now need to be trialled, and at this stage it appears unlikely that the revised Core version of the TLFS will be ready for public release before 2026, with the Plus component afterwards.

The main impact of this delay has been the requirement to continue running the LFS into 2026 at a minimum, while at the same time implementing the two surveys that make up the Core and Plus components of TLFS, and then bringing them online to replace the LFS. This has major implications for the use of ONS social survey resources. Ongoing LFS programs aimed at “maintaining” the LFS are currently under review, with a focus on only carrying out essential maintenance of the LFS to ensure its sustainability until the TLFS takes over. It is in this context that we have been presented with an overview of this LFS “recovery” plan and asked to comment on it from a methodological perspective, considering the steps that have already been implemented as well as those proposed and remaining to be implemented.

2. Recovering Survey Response

The main methodological issues associated with ensuring the viability of the LFS until potentially into 2027 are changes to the survey collection methods to at least maintain the current response levels (if not improve them) and the need to integrate more current population benchmarks into LFS survey weights. The work on improving response is focusing on re-investing in interviewer capacity and recovering to pre-COVID capacity. This has already delivered an increase in response levels, which has the benefit of improving sample size to support the required estimates as well as potential reduction in non-response bias. We believe that this is the most practical way of ensuring the LFS continues to provide “acceptable” labour market data until it is replaced by the TLFS. That is, the ONS should continue to make incremental updates to LFS design, operations and engagement procedures, as well as associated changes to the survey collection, in order to not only halt the drop in its response rates but also to return to pre-COVID operations and performance.

3. Updating Benchmarks Used In LFS Weights

Re-weighting of LFS to revised population benchmarks has occurred post-census since the LFS’s inception, and it typically happens around this time in the inter-census cycle. However, in this period there is a natural delay with Scotland’s Census occurring in 2022 as opposed to 2021. As such, while it is a major use of methodological resources to compute new weights to reflect new information about population benchmarks, this activity is more about ongoing LFS maintenance than recovering from dropping response rates. Unlike the clear work targeting improved response in fieldwork, here we see a more confusing picture, partly a consequence of there not being a full UK census for 2021 but a product spread over 2021 and 2022. So far there have been two “interim” reweighting efforts (2022 and 2023) that have already been carried out as well as a proposed “full” reweighting exercise using 2024 benchmark data that will require substantial ONS methodological resources from mid-2025 through mid-2026 (and perhaps beyond).

Although we understand the drivers for these benchmark exercises, and that new benchmarks can level-shift estimates, the fact that UK benchmarks are age-sex-geography based will not provide the same improvement in underlying LFS quality as the focus on recovering response. We are also aware that implementing them detracts significantly from the resources that the ONS can use to get the TLFS up and running. We are therefore of the opinion that the commitment to the full post-census round of LFS reweighting be scheduled for 2026 and be integrated with the creation of the weights produced for the first set of official TLFS headline estimates. This will ensure that LFS estimates produced at the end of its life cycle are consistent with those output by the TLFS at the start of its life cycle. We understand that this could lead to some temporary lack of coordination between the benchmarks used by the LFS and the APS, as well as in the historical time series generated by these surveys. However, this may be a necessary cost for the ONS to free up methodological resources for the TLFS.

4. Concluding Remarks

Finally, it is worth emphasising that our comments above assume that the TLFS implementation programme proceeds without any major hiccups, in which case a changeover is a realistic option from around mid-2026. Obviously if there are further delays then the need to keep the LFS going longer becomes a real possibility and this will require resources and significant investment, particularly in terms of putting in patches for many of the technical systems currently employed by the LFS that will no longer be supported. This is a risk that the ONS will need to take on board, keeping in mind that the full LFS/TLFS reweighting will be required in 2026 whether or not ONS is moving to TLFS.

Professor Ray Chambers and Professor James Brown
2 March 2025