By John Forth
The government has tried to tackle low earnings through the National Minimum Wage (NMW) –first introduced in 1999, and the National Living Wage (NLW), which came into force in 2016 to replace the NMW for employees aged 25 and over.
The Annual Survey of Hours and Earnings (ASHE) is a cornerstone of UK labour market statistics, providing invaluable insights into wages, hours, and employment conditions. ASHE is based on a 1% sample of employee jobs, and statistics are generated using official weights designed to make the achieved sample in each year representative of the population of employee jobs in Britain. ASHE is the main source of information on the incidence of low pay.
A new ESCoE discussion paper from John Forth, Alex Bryson, Van Phan, Felix Ritchie, Carl Singleton, Lucy Stokes and Damian Whittard takes another look at estimates of low pay in Britain from 2004-2021, using ASHE. We use additional weighting techniques to adjust for biases in the ASHE sample and generate new estimates of the incidence of low pay. Our findings suggest that the bite of the National Living Wage is greater – and progress toward the government target for eradicating low pay has been faster – than previously understood.
How did we do it?
Our analysis used data developed as part of the Wage and Employment Dynamics project.
ONS (the UK Office for National Statistics) generates weights for each annual ASHE dataset which make the achieved sample representative of the population of employee jobs in Britain in terms of gender, age, occupation and religion. We linked data from ASHE to the Business Structure Database – a research-ready version of the ONS business register. This allowed us to investigate whether certain types of organisations remain under or overrepresented in the achieved cross-sectional ASHE sample, even after applying the original ASHE weights.
We also investigated rates of longitudinal attrition in ASHE, after using the Annual Population Survey to account for an employee’s likelihood of moving out of employee status (and hence out of scope for ASHE).
What did we find?
- Jobs in smaller organisations, younger organisations and those in the private sector were under-represented in the annual weighted samples from ASHE.
- Younger employees, those on low wages and those working relatively few hours were less likely to remain in the ASHE sample over time. This was even after accounting for their likelihood of leaving employment.
Source: ASHE
We constructed new weights to remove these biases. Our new estimates suggest that:
- The percentage of jobs paid at or below the National Minimum Wage is under-estimated by around 1/5 (see Figure 1).
- The bite of the NLW is also under-estimated, such that the Government’s targets for this measure have been reached more quickly than previously thought.
- In contrast, the share of employees moving off the minimum wage to higher-paid employment each year is not substantively affected by sample bias.
Why it matters?
The findings indicate that there are observable response biases in existing estimates produced from ASHE. These biases have a significant impact on our view of the bottom end of the wage distribution.
This has clear policy implications, as the share of jobs paid at or below the NMW or NLW and the incidence of low pay are key indicators for the Low Pay Commission that advises the government on the NMW and NLW.
What’s next?
The Wage and Employment Dynamics project is also now linking ASHE to data from HMRC’s Real Time Information (RTI) system. The RTI data can tell us whether a person sampled for ASHE is in employment at any given point in the year. The link between ASHE and the RTI data will therefore enable us to evaluate the representativeness of the longitudinal ASHE samples in a different manner than has been possible to date.
This links to ESCoE work on LEED (Linked Employee-Employer Data).
ESCoE blogs are published to further debate. Any views expressed are solely those of the author(s) and so cannot be taken to represent those of the ESCoE, its partner institutions or the Office for National Statistics.
This work was produced using statistical data owned by ONS and accessed through the ONS Secure Research Service. The use of the ONS statistical data in this work does not imply the endorsement of the ONS in relation to the interpretation or analyses of the statistical data. The work uses research datasets which may not exactly reproduce National Statistics aggregates.