For the same three months, one official measure of output per hour shows a fall and another shows a rise. The gap says less about the economy than about how unsettled the underlying survey still is.
ONS's headline productivity measure, output per hour worked, is built from the Labour Force Survey (LFS) — the traditional, survey-based method. In its most recent bulletin ONS reported that this measure fell 0.5% in the year to Q4 2025.1 In the same release, ONS also published an alternative estimate built instead from PAYE Real Time Information (RTI) tax data combined with LFS inputs, which it is trialling as a possible future lead indicator. That measure rose 1.6% over the identical quarter.1 Output per worker shows the same pattern in miniature: −0.2% on the survey-based measure, +2.0% on the administrative-data measure.1 Both are ONS's own numbers, for the same three months, and they do not agree on whether productivity rose or fell.
The LFS-based measure rests on a household survey whose response rate collapsed after the pandemic. By July–September 2023 the achieved response rate for Great Britain had fallen to 12.7%, with just 44,238 individual responses — a level ONS judged too uncertain to publish from, and LFS-based outputs were temporarily suspended that October.2 By October–December 2024, after a recovery plan that reintroduced face-to-face interviewing and expanded the sample, the response rate had recovered to 19.6%, with 63,069 responses — a 54% relative improvement in the response rate, though still well short of pre-pandemic norms.2 The RTI/PAYE-based alternative sidesteps this problem by drawing on employer tax records rather than a household survey, which is precisely why ONS is trialling it.
The Transformed Labour Force Survey (TLFS), a digital-first redesign, has been in development since March 2020.3 The current LFS was suspended for publication in October 2023 over quality concerns, then rebuilt through 2023–2024.2 ONS now plans a full assessment of the TLFS with users in July 2026, aiming to transition headline labour-market statistics in November 2026 — six years after the TLFS began, and explicitly caveated as data-led and liable to slip further.3
Which of the two measures in Exhibit 01 will prove closer to what a fully recovered, fully weighted LFS would eventually show cannot be answered yet — that is exactly what ONS's own consultation and assessment process, running to July 2026, is meant to establish. This piece could not independently re-derive earlier Blue Book vintages of output per hour from ONS's published archive within this session and so relies on ONS's own account of the response-rate collapse and recovery, rather than a self-computed revisions triangle comparing successive published vintages figure-by-figure. Separately, a small ONS-acknowledged data error affected some productivity-bulletin figures released in August and November 2025; ONS describes the resulting changes as small, and this piece has not attempted to quantify them independently.
That the two current measures disagree is a measured fact. Whether that disagreement is alarming is a live argument among users of the statistics.
The Office for Statistics Regulation has kept Labour Force Survey-based labour-market outputs outside full accredited-statistics status since the 2023 quality crisis, and ONS itself has pushed the TLFS transition back repeatedly, most recently to a "data-led" November 2026 target that could slip again — a five-plus-year repair job for a headline economic indicator.23
Revision and methodological parallel-running are how national accounts are built everywhere; ONS has been transparent about the response-rate problem, published the numbers, and is running two methods in the open specifically so users can see the disagreement rather than being handed a single falsely precise figure.