Over the last year the ONS has been focusing resources on its headline economic statistics and progressing improvements, particularly on our complex labour market data. In line with our commitment to transparency at the earliest opportunity, James Benford updates on an operational issue which will temporarily impact data quality for the labour force survey from next month. James describes the immediate action taken and work in train on a wider lessons-learned exercise to ensure we both correct course and act to strengthen our operations to further progress the recovery of our statistics.
We published labour market statistics for the three-months to April on Thursday. The labour market remained broadly stable, with further softening evident in some measures. Payroll numbers continued to fall over this period, with new recruits at their lowest level in five years. However, overall employment was little changed, with some signs of workers moving into self-employment. Meanwhile, regular wage growth in the private sector slowed to its lowest rate in five and a half years, though total earnings are growing faster because bonus payments in March and April are higher than a year ago, particularly in the financial sector. Public sector pay growth increased but is once again affected by the timing of pay awards varying this year.
This assessment was informed by our range of labour market data including surveys of households and businesses as well as administrative data from HM Revenue & Customs and the Department for Work and Pensions. Behind some of these numbers is a team of ONS interviewers in the field and working over the phone who are collecting data.
The purpose of this blog is to alert users of our labour market statistics to an operational issue with how telephone interviewers’ time was allocated, which occurred in collection of survey data during May and early June. This did not affect Thursday’s release but will impact on our next labour market release in July.
Our work to collect survey data on the labour market is complicated at the moment by the need to run two surveys in parallel – our Labour Force Survey (LFS) and our Transformed Labour Force Survey (TLFS). Across the two we have over 594 field interviewers and the majority of our 139 telephone interviewers dedicated to securing responses from samples of over 118,000 households per month. Running two surveys in parallel is costly and complex, which is why we are working hard to transition to the TLFS in 2027.
At the start of last month, a mistake in operational systems occurred which led to an over-allocation of telephone interviewers from the LFS to the TLFS. The direct impact was on LFS response levels between 3 May and 10 June, with some residual effects persisting until 17 June.
The issue primarily affects Waves 2 to 5 of the LFS. The Wave 1 achieved sample has remained broadly in line with previous performance. The impact on Waves 2-5 of the main sample has been to reduce the number of achieved household interviews by around 1,200 (19%) compared to the preceding period. July will be the first full month in which corrected allocations are in place allowing achieved sample sizes to return to previous levels.
I am sorry that this issue occurred and regret that it was not detected and acted upon at an earlier point. Once the issue became clear, we undertook a number of rapid actions.
Firstly, operationally, the issue has now been resolved. The measures we have put in place to ensure it doesn’t happen again include tighter daily monitoring, reprioritisation of resource and changes to shift patterns and work allocation.
Next, we have been assessing how much impact this will have on our labour market statistics. In simple terms, there will be a level of reduced quality for our labour market statistics in July, with a smaller effect on the subsequent releases, because we will be replacing missing data points with estimated values. To go into more depth, the first datasets to be affected are those for the March to May 2026 rolling quarter (MM26) which will be published on 21 July. We can already see that the size of this rolling dataset (UK including estimated values or ‘imputation’) is broadly similar to previous rolling 3-month periods at around 80,000. This is because the impact of this issue in data collection is mitigated by an increased level of imputation in the dataset. This increased level of imputation will to a degree likely artificially dampen measured changes between time periods. We will work to quantify the impact here and will report on it transparently with the next labour market release, including any effects on bias.
Importantly, this temporary setback does not affect how overall we are continuing to see a longer-term increase in LFS responses. The reallocation should allow data collected for the full month of July to return to an achieved sample size that is broadly in line with the pre-pandemic level. This reflects our recovery work over a number of years to sharpen the incentives in the survey and build our field community to 835 interviewers.
We will conduct a lessons-learned exercise within the ONS which will report to me, bring in external perspectives and seek to reach initial conclusions within the next month. We want to learn as much as we can from this incident on the LFS, while also taking a broader look across both social and business surveys at how we conduct, monitor and actively manage our survey operations. Drawing broad lessons will mean that we recover from this incident with more robust operational processes and a stronger culture around them. Further detail will be outlined in our quarterly Surveys and Economic Statistics Improvement Plan Update in July.
At a time when we have made such progress in improving our economic statistics, it is critical that we maximise the learnings from this temporary setback and I am confident that we will. Ultimately this will mean that the wider recovery in our surveys and economic statistics will be even stronger, with more robust operational processes and a supporting culture to interrogate and continuously improve them.
James Benford is Director-General for Surveys and Economic and Social Statistics at the Office for National Statistics.