Labour Market

BBC study illustrates the pitfalls of ethnic pay gap reporting

The BBC has published a thought-provoking article by Rianna Croxford on pay gaps in academia, based on months of freedom of information requests and interviews. The striking headline is that ‘black and Arab academics at the UK’s top universities earn an average 26% less than white colleagues’. But a closer look quickly reveals some of the familiar problems with pay gap reporting.

For a start, the conclusion many others have drawn is that ethnic minorities are paid 26% less for doing comparable jobs. Indeed, while the article itself is relatively careful, the repeated use of the word ‘colleagues’ appears to imply that a black or Arab Professor should expect to earn much less than a white Professor with identical responsibilities and seniority. While that may happen in some cases, it is unlikely to be common (let alone legal), especially in a profession where fixed pay scales are the norm.

The same problem bedevils the reporting of gender pay gap data, which some campaigners persistently misrepresent as evidence of unequal pay. Kate Andrews and I have written on this before in the context of ‘Equal Pay Day’.

Academia appears to fit the usual pattern of fewer ethnic minorities in more senior roles. Of course, that may still reflect ongoing discrimination of one form or another, and I would not deny that this exists. But it could also largely be a legacy of what academia was like 10, 20 or 30 years ago, rather than today, or a reflection of a myriad of other factors. The raw data in the BBC article simply cannot tell us.

Again, there are parallels with the gender pay gap, which is now close to zero for men and women in full-time employment between the ages of 18 and 39. It would be interesting to see the same data specifically for academics, but it would be surprising if the picture in universities were very different.

Indeed, the BBC report appears to support this. Both Warwick and Glasgow universities pointed out that a much larger proportion of their ‘early career’ staff came from ethnic minorities, who are of course paid less than more senior staff, regardless of ethnicity.

The BBC study also illustrates the problem of small sample sizes. In particular, only 1% of academics surveyed are ‘Black’, and just 0.4% are ‘Arab’. In this case the absolute numbers are probably still large enough to be fairly reliable. But if you required firms with as few as 250 employees to report their ethnic pay gaps, the figures are much less likely to be statistically significant.

What’s more, for any such data to be meaningful, you’d probably want to divide the groups further. The 2011 census had 18 standardised ethnic categories, with ‘Asian’ (for example) broken down into Indian, Pakistani, Bangladeshi, Chinese and ‘other’.

This is important, because we already know from economy-wide data there are large differences in average pay even between particular ethnic minority groups. But if you leap straight to ‘discrimination’ as the explanation, you soon run into difficulties.

For instance, data published by the Equality and Human Rights Commission show that Black Caribbean, Chinese and Indian women earn more, on average, than White British women. Should we be demanding action is taken to correct this? Similarly, looking at the gender pay gaps for part-time employment, women earn more, on average, than men. Is this a problem?

This isn’t just a pedantic complaint about the abuse of statistics (annoying though I find that to be). If done badly, ethnic pay gap reporting could have all sorts of unintended and undesirable consequences. I doubt that the headlines from the BBC story will encourage many people from ethnic minority backgrounds to apply for academic posts. A narrow focus on aggregate ethnic or gender pay gap data could even deter some employers from appointing young black women to junior roles, for fear that this would drag down their averages.

In sum, I’m sure that many people continue to face discrimination in workplaces of all kinds, including on the grounds of sex and race. Rianna Croxford’s experience of Cambridge was clearly different from mine. But we need to be aware of the limitations of pay gap data, whether based on gender or ethnicity, and interpret them with more care.

Julian Jessop is an independent economist with over thirty years of experience gained in the public sector, City and consultancy, including senior positions at HM Treasury, HSBC, Standard Chartered Bank and Capital Economics. He was Chief Economist and Head of the Brexit Unit at the IEA until December 2018 and continues to support our work, especially schools outreach, on a pro bono basis.

2 thoughts on “BBC study illustrates the pitfalls of ethnic pay gap reporting”

  1. Posted 08/12/2018 at 16:27 | Permalink

    From the BBC article-

    “When Meenakshi Sarkar landed a job as a teaching fellow at Leeds University after a successful career in business in India”

    Some Googling brings up her profile, her last job in India 013–2014: Director Learning & Development, Earth Infrastructure Ltd, New Delhi, some more Googling and we find that Earth Infrastructure Ltd is a complete fraud, here are just a few examples –

    Court dismisses bail pleas of 3 directors of real estate group Earth Infrastructure

    Earth Infrastructures – Investor Alert ( regarding it being a fraud )

    Earth Infrastructure Employees Are Fraud — EARTH INFRASTRUCTURE EMPLOYEES ARE FRAUD

  2. Posted 08/12/2018 at 17:41 | Permalink

    Right on the money; although I would say that they were not careful in their conclusion “The new data suggests that ethnic minority academics are less likely than white men or women to be promoted to better paid, senior positions.”

    In epidemiology if one was trying to make an analysis of this type we would try to control for all the possible confounding variables (experience, role, citizenship, etc.), and to see if we’d done enough we have a technique for checking whether biases remain: negative controls. Don’t just look at a breakdown by the target of your analysis (black vs white), but also look at data on a non-target thought not to be affected by the same mechanism (racial bias): such as in this case, perhaps white non-British vs white British. If the analysis suggests discrimination against white non-British then maybe you’re just seeing a dominant effect from the factors other than racial bias

Leave a Reply

Your email address will not be published. Required fields are marked *