Today the ONS has published the latest data and insights from its Annual Survey of Hours and Earnings (ASHE), which provides a huge amount of detail about the wages of employees in the UK. Liz McKeown takes a look at the latest data, explores how they compare with other data sources on earnings and outlines the steps we are taking to improve these important statistics.
Each year ASHE provides an in-depth picture of what employees are earning and how that varies by their individual characteristics, their location, industry and type of job. It’s an important survey which provides key data to underpin policy decisions, including informing the setting of the National Minimum Wage and Living Wage. ASHE provides detailed insights on employee earnings, low and high pay and the gender pay gap. At a headline level, this year’s results show the proportion of low-paid employee jobs falling to the lowest level since the series began in 1997, while the gender pay gap continued to narrow. Beneath these headlines we can explore the ASHE data in detail to see how these – and other trends – differ for different groups.
The granularity that ASHE provides is possible because it is one of the largest surveys the ONS runs; sampling around 1% of all employees paid through HM Revenue & Custom’s (HMRC) Pay As You Earn system. This year’s estimates are derived from some 174,000 employee returns sent to us from businesses of every shape and size from right across the country. However, despite its large size, ASHE has faced some challenges since the pandemic. We began to address these last year (as I outlined in this blog) and have continued to make progress with them this year, with further improvements planned.
Exploring recent progress
Our overall approach to making improvements to ASHE was set out in the Economic Statistics Plan we published in June and highlighted further in James Benford’s blog last month, which looked more broadly at how we are working to improve labour market statistics. Today I want to provide a little more detail on two areas of focus where we have made progress in the last year:
- Modernising production of ASHE data We have taken a number of steps to improve the efficiency and effectiveness of our production processes, such as redeveloping our table production system and expanding the use of electronic data collection. We have also introduced a new, more efficient, tool – ClassifAI – which uses Artificial Intelligence to code occupations within ASHE. This is the first time that the ONS has applied AI directly into a statistical process and, in doing so, we have improved the accuracy of occupation coding while saving hundreds of hours of work, which we were able to invest into helping achieve a quality dataset. This tool is now being deployed to other surveys, with ASHE pioneering and benefitting from our recent pivoting of data science and innovation resource within the ONS to focus on improving our core statistics.
- Connecting with users and experts. Last year we recognised that there was scope for improvement in how we engaged with users on the changes that needed to be made to ASHE. To help address this we launched a new quarterly Earnings User Group, attended by around 20 different government departments and other public sector bodies. In April we co-organised (alongside the Wage and Employment Dynamics project, Administrative Data Research UK and the Economic Statistics Centre of Excellence) and hosted an inaugural Earnings symposium to engage with a wider user base, including researchers and academics. We also continue to draw on expertise from the Stakeholder Advisory Panel for Labour Market Statistics.
Comparing sources of earnings data
An area that continues to be important for users is the comparability and coherence of the different sources of earnings data. ASHE sits alongside our other key sources of earnings data, notably Average Weekly Earnings (AWE) obtained from a separate survey of businesses and administrative data from HMRC’s Pay as You Earn (PAYE) Real Time Information (RTI) system. In broad terms, while ASHE provides detailed granular information annually, AWE and RTI provide the best view monthly on economy wide trends and changes.
This year we continue to see some differences between sources with the pattern differing somewhat depending on whether we focus on median or mean measures. This reflects that these different sources have some differences in their design and methods, including, for example, in terms of coverage, frequency, reference periods and definitions and can consequently provide results which are a little different to each other. In last year’s blog we noted that these differences had become more apparent since the pandemic with a divergence in trends between ASHE, and AWE and RTI, emerging. This informed our decision to implement the improvements last year to the way we process information about higher earners in ASHE, in order to help address this issue. Continuing to assess and, where it is appropriate, address differences between sources forms part of our wider review of statistical methods in ASHE, which is already underway.
Addressing remaining challenges
This review represents one of our current key priorities in our work to improve the quality of our earnings statistics. Within that an important area of focus is reviewing our sampling and weighting approach, to ensure that the ASHE methods continue to be up-to-date. We have already undertaken extensive user engagement around our sampling approach, which helped confirm that the current ASHE sample design is still meeting the needs of our users and we are now shifting our focus to reviewing the weighting approach.
We recognise that some recent studies have pointed to the potential impact on estimates that might result from updating weighting methods. However, we are still assessing whether – and if so how – these weights need to change and as such it is too early to say what impact, if any, there will be on headline measures. Once the review is complete, we will look to provide an update, including sharing any assessment of indicative impacts at that stage.
Over the next two years we are also planning on introducing new processing systems, which will allow us to compile and quality assure these data more efficiently. These include new systems to collect and clean the ASHE microdata and to process ASHE results. Using modern systems will also provide more flexibility to implement any methods changes and allow us to adapt to emerging user needs more quickly in the future.
Given the importance of ASHE data in informing critical decisions, we are committed to continue working closely with key users and experts as we develop and implement any changes. We welcome feedback both on the steps we’ve already taken and on our key priorities for making further progress – please send this to earnings@ons.gov.uk.
Liz McKeown is Director of Economic Statistics at the Office for National Statistics.