Falling pay, divergent data and a bulging middle.


The ONS has published its 2023 Annual Survey of Hours and Earnings (ASHE), taken in April this year. ASHE is a big survey of employers that provides a detailed picture of employee pay across different places, jobs, and types of workers. It’s also the only data we get on hourly pay. So what does this year’s data show?

Well (you guessed it) real pay is falling – but there’s more going on beneath the surface. There are some notable demographic deviations within the fall, not to mention some questions about the integrity of the survey data. Plus, there’s some good news about the impact of the minimum wage (keep an eye out for an in-depth look at what this all means for the 2024 minimum wage rate later this week…).

So how bad is it?

In April 2023, median weekly earnings were up 7.7 per cent on 2022 – the fastest growth in the survey’s history (which goes back to 1997). But even-higher inflation meant that earnings were still falling in real terms – by 0.9 per cent on the year. That hit to earnings in 2023 was less bad than the even steeper (-3.1 per cent) fall in 2022, but shows the cost-of-living crisis squeeze continuing.

The real-terms fall in median hourly pay was bigger – falling 1.2 per cent in 2023 (from £16.02 to £13.83, in April 2023 prices). The fact that the fall in weekly pay was smaller is partly related to a compositional hours shift this year, with the proportion of employees working part-time falling 1 percentage point between 2022 and 2023 in the ASHE data.

As a grim aside, these falls in the value of real earnings over the past two years come in the context of long-term pay stagnation since the financial crisis. The 2010s were worse for pay growth than any decade over the past century.

How does that break down for different types of people?

But the point of ASHE is to look beneath the headlines (we already knew from other data sources that the real pay squeeze continued up until the summer) and to get an insight into how pay varies for different groups. Since 2019, real-terms growth in median weekly earnings has been weakest for men (- 4.5 per cent), for workers in London (- 4.7 per cent), and for the youngest workers, aged 18-21 (- 6.4 per cent). Remember, this last group aren’t entitled to the National Living Wage (but many still receive it). People over 60, resident in Scotland or working in the private sector have by contrast done a little better. It’s important to note however that none of these demographics saw pay growth much above 2 per cent.

This survey also lets us look at how pay growth varies at different points of pay distribution.

The good news first – the minimum wage has had a huge impact on hourly pay and led to much faster pay growth for low earners, especially since 2015.

This pattern of stronger hourly pay growth for low earners, (thank you minimum wage), is why the rate of ‘low pay’ is falling, and falling rapidly. The ONS estimate that the proportion of workers earning less than 2/3 median fell to 8.9 per cent in 2023. That’s down from 22.8 per cent in 1995 – a real policy success story.

In the ASHE data at least (more on data questions below), the good news appears to extend to weekly pay as well, which is also showing stronger pay growth for low earners.

The latest data shows that as well as ‘low pay’ falling, the share of employees in ‘high pay’ – defined as hourly pay being at least 50 per cent above the median – has also fallen. The drivers of this change are less obvious, although high-paid employees switching to self-employment to reduce their tax bills might be part of the story. But the sum of all this is that Britain has a ‘bulging middle’ when it comes to hourly pay – with the proportion of employees who are neither low- or high-paid rising from 58 per cent in 2019 to 68 per cent in 2023.

Mind the (gender pay) gap

ASHE is also an important source of data on the gender pay gap, which was essentially unchanged in 2023. It can be easy to forget how dramatically the gap has narrowed over the past decades. In 1971 it stood at 36 per cent, but has now fallen to 8 per cent. This is a strong reminder of the progress that’s already been made, and the gap that can still be closed. ONS’s write-up shows there continue to be large gender pay gaps within occupations, so it’s not just a question of sorting effects.

What’s all this about dodgy data?

It’s vital to note at this stage that there are real questions about the quality of the data. There was lots of discussion about the problems facing the Labour Force Survey last week (which prompted ONS to delay their normal release and then only release ‘experimental’ revised data). ASHE suffers from some of the sample problems – linked to a falling sample size (especially among smaller businesses), and questions about how sample numbers should be ‘grossed up’ to produce whole-workforce estimates. Xiaowei Xu from the IFS has a good thread looking at these issues.

You can find some work from Felix Ritchie and his colleagues in the ‘Wages and Employment Dynamics’ project digging into the impact of that here, and showing that ASHE’s estimates of typical hourly pay are likely to be biased upwards as a result. At some point down the line this will give the ONS (who have to worry about data quality) and Low Pay Commission (who rely on estimates of median hourly pay from ASHE to set the minimum wage) a big headache.
Questions about ASHE’s estimates show up in a few places. One is differences between the ASHE data and data from other datasets. For example, the picture painted by ASHE data on pay growth across the weekly pay distribution (stronger at the bottom) does not show up in HMRC’s PAYE data. In fact, until the most recent months HMRC was showing stronger pay growth at the top.

There are also differences when it comes to sectors. In ASHE, pay growth looks strongest in some low paying sectors (retail, arts & recreation), whereas in HMRC’s data it’s the higher-paying sectors like finance and professional services that are seeing the steepest growth. This is consistent with the above in terms of HMRC showing stronger pay growth for high earners.

Clearly, it’s far from ideal having contrasting messages coming from datasets which are, ostensibly at least, measuring roughly the same thing. No doubt the ONS will be working hard to understand these differences.

More encouragingly, although there are questions about levels, and in how pay growth differs for high and low earners, ASHE seems to be more or less in line with other datasets when it comes to estimates of the overall rate of pay growth coming from other datasets – HMRC’s PAYE data, and the ONS’s Average Weekly Earnings dataset. For example, in April 2023, these datasets produced estimates of the growth rate of average (i.e. mean) earnings which fell within 0.4 percentage points of one another – 6.9 per cent (mean gross weekly pay in ASHE), 6.7 per cent (HMRC PAYE mean monthly pay), and 7.3 per cent (Average Weekly Earnings regular pay).

Nevertheless, there is definitely work to do across the board when it comes to the quality of the UK’s labour market data.

For now, today’s ASHE data is confirmation of two important facts on pay which have defined recent years: the recent sharp real-terms fall in earnings, alongside the more positive news that the minimum wage continues to boost hourly pay for the lowest paid workers.

2023 may not have brought the same blow to take home pay that 2022 did, but household budgets remain squeezed by stagnant pay packets.