Methodological Assurance Review Panel minutes: Transformed Labour Force Survey, 16 July 2026

The Methodological Assurance Review Panel (MARP) received an update from ONS on progress with the Transformed Labour Force Survey (TLFS), focusing on survey design, response rates, current development of weighting and estimation approaches, and coherence with administrative data.

This was the only item. 

• Sir Bernard Silverman (Chair) 
• Ana Basiri 
• Oliver Duke-Williams 
• Jakob Schneebacher  
• Natalie Shlomo 
• Leone Wardman (Presenter) 
• Tom Evans (Presenter) 
• Tom Tarling (Secretariat) 
• Dale Williams (Secretariat) 

Introduction

Sir Bernard Silverman welcomed the panel. Nik Lomax gave apologies. 

Transformed Labour Force Survey (Leone Wardman, Tom Evans)

ONS gave an overview and started the discussion with response rates. 

ONS noted there was a trade-off to consider with increased response, where more responses from similar groups may lead to an increasing impact from non-response bias. ONS uses weighting to compensate for non-response and attrition across survey waves and imputation to address within response drop-out at the item level.

Like most online surveys, response drop-off across TLFS survey waves has presented challenge. ONS noted that methods have been designed to account for drop-out that is correlated with socio-economic status.

The panel agreed this type of biased missingness was a challenge to the TLFS, and agreed with the development of multi-phase weights and imputation to help address both non-response and survey drop-out biases. The panel also suggested including geography within the weighting factors and to consider residual geographic bias in the sample frame.

ONS noted that performance targets are on unweighted responses, while published estimates are monitored against coefficient of variation (CV) targets. ONS commented that it engaged widely with stakeholders including the Bank of England, HM Treasury and industry experts to determine measures of success.

Comparison with other National Statistics Institutes (NSIs) was discussed, with ONS noting a 40% response rate occurred elsewhere, though noted in some countries response was mandatory rather than voluntary.

The panel discussed technical requirements and applications of the online survey to encourage participation. It suggested nudging, AB (experimental design) testing of questionnaire and materials and analytics of online usage could be done. ONS commented it has done analysis on the impact of incentives and there was ongoing development work.

Partial responses were discussed, with ONS commenting it kept individual responses and scaled them for incomplete households. The panel said survey fatigue was important to keep in mind, and interventions mid-survey and adaptive survey design could be useful. 

ONS outlined ongoing work on reweighting, seasonal adjustment and discontinuities. A challenge for methods development is that that it takes 15 months for any change to feed-through all the response waves.

The panel asked whether administrative data could be used. ONS responded it was looking into sources such as Pay-As-You-Earn (PAYE) data for coherence and to potentially support estimation methods. The panel suggested that HMRC Self-Assessment data be considered in addition to PAYE. ONS agreed, and noted that Self-Assessment data was becoming quarterly rather than annual in the near future.

ONS presented further details on weighting and adjustments affecting address bias, calibration and effects on estimates. The knock-to-nudge (KtN) approach was presented, with the impacts on observed employment rates considered.

The panel questioned the overall effect of KtN on the employment rate. ONS responded there was expected to be some impact on employment and inactivity when a KtN adjustment was introduced. ONS commented the effect of KtN was different across age-sex-region bands, and the revised approach and corresponding adjustment would ensure levels of KtN achieved over time were accounted for in estimation.

ONS presented quarterly changes to the employment rate estimation between the LFS and TLFS, improvements in response rate following changes, and decreased median completion times for respondent cohorts. This was compared to administrative data estimates from PAYE Real Time Information measures from HMRC, showing the relative movement over time. However, TLFS is still maturing and this picture is subject to change.

Further discussion was had on harmonisation as well as the interactions between imputation and weighting. ONS noted development and assurance of the methods were on-going, and would be optimised with data maturity.

The panel thanked ONS for the detailed presentation. ONS agreed to bring future updates to the panel for consideration.

The meeting concluded. Following the meeting, the panel issued a statement.