The ONS is transitioning to better methods of data collection fit for the digital age. The changes in progress mean some temporary disruption as traditional surveys are updated or replaced. In this post Darren Morgan explains how we are adapting our approach to maintain the accuracy of our key labour market statistics during this transitional phase.
The ONS Labour Force Survey (LFS) is our largest monthly survey, covering tens of thousands of households across the UK. Its sheer scale has traditionally given us enough data to produce reliable and detailed estimates of people who are in employment, unemployed and looking for work, or economically inactive for reasons like sickness, study or caring responsibilities.
This survey has been running successfully now for around five decades. But times – and people’s lifestyles – have been changing. In common with statistical colleagues in many other countries, we’re found it harder to engage with enough people in certain groups.
So, we have been busy developing and implementing a new transformed survey designed to better capture a fully representative sample of the whole population by making it online first, followed up by teams phoning households and knocking on doors in areas of low response, thus vastly increasing the response rates, the range of people replying and, ultimately, data quality. We are in the middle of our transition as we wind down the old-style LFS and build up the new transformed version ready to be launched in the Spring.
These radical plans are part of a wider long-term strategy to move away from a reliance on traditional surveys and embrace the full range of sources such as using VAT data to calculate GDP and checkout scanner and web scraped data to measure inflation.
Under normal circumstances this month’s statistics would have been published a week ago but we rescheduled them to allow us time to adapt our approach and ensure we are producing the best possible estimates of the labour market using the best available data sources.
Today, we have produced a new metric, adjusting our headline estimates using robust administrative data sources that we receive from other government departments. This maintains the accuracy of our key statistics.
Since before the pandemic, we have been publishing a timely and accurate measure of the number of people with a job on payroll, calculated using HMRC PAYE data. While this doesn’t include the self-employed, in the short term, changes in the number of employees are a reliable indicator of changes in the overall rate of employment. So for the recent two months where survey response have been particularly low, we have used these data to calculate the overall change in the rate of employment.
Likewise, we have used claimant count data to calculate changes in the rate of the number of unemployed (people out of work but actively seeking work). While during the changeover to Universal Credit and during the pandemic, when the eligibility criteria shifted, we urged caution with this dataset as a measure of change. However, over the last year it has proved an accurate measure of changes in those looking for work and so can be reliably used for this purpose.
This new adjusted metric gives a good picture of changes in the labour market. However, to avoid confusion we have not published the usual unadjusted LFS dataset alongside it, when we know there are challenges with these unadjusted data.
Falling survey response rates is a challenge facing statistics agencies worldwide, but here at the ONS we are working hard to ensure we can keep producing trusted labour market metrics.