The ethnicity pay gap was highlighted in a report published last year by the Resolution Foundation, authored by Kathleen Henehan and Helena Rose. It was found that even after accounting for individual characteristics and background factors, large gaps in hourly pay persist between ethnic minority people and the white British ethnic majority. The headline finding was of a £3.2 billion differential in pay between minority groups and the ethnic majority.
Examples cited in a blog by Henehan were that non-graduate Pakistani and Bangladeshi men were paid £1.91 less per hour than comparable whites (14 per cent), once background factors were accounted for while black graduate men were paid £3.90 less per hour than comparable white men (17 per cent).
“We cannot be certain about what factors underlie these remaining differences – attributes not captured by the data may play a role – but the strong suggestion is that discrimination matters too.”
Note that she does not actually say her findings evidence discrimination but her choice of wording is sufficient for the reader to infer it as a key explanatory factor. It is rather reminiscent of Francis Urquhart’s famous line: “You might very well think that; I couldn’t possibly comment”.
Indeed, the tone of her article as well as the Resolution Foundation’s report itself is carefully noncommittal. While Henehan commends the Government’s diktat that companies be obliged to publish data on pay by gender, she does not explicitly commit herself to backing a similar measure on pay by ethnicity. The report concludes that it has:
“…shown that the ‘Burning Injustices’ the Prime Minister highlighted on the steps of 10 Downing Street are very real indeed… Government interest in this area is timely and welcome: policy makers and researchers have a challenge on their hands.”
While allowing for the fact that differences can and are narrowing, in some cases, on their own, it is not made explicit why differences in pay are a ‘burning injustice’ and if so, what is to be done. We are only told it is a challenge.
At heart though, there is a problem of method here which renders all conclusions, however succinctly put, about the role of discrimination questionable.
The report’s findings are based on regression analysis of the Labour Force Survey. This is a cross-sectional survey of households that forms the basis for much of our understanding of the labour market. The regression analysis works by presenting the difference in mean pay between ethnic groups and allows you to take into account other variables which might explain away differences. For instance, you might think that white people earn more because they are on average older than ethnic minorities – the regression analysis will allow you to ‘control’ for age, showing you the differences between groups, if any, with the same age profiles.
The Resolution Foundation’s analysis controls for the following statistical variables: age, region, qualifications, degree class, whether a person was born in the UK or not, full-time/part-time work, years of experience in the labour market, occupation, industry, public/private sector, permanent or non-permanent contract.
It is a good list of statistical controls. However, it does not control for everything, nor is this possible using these data. The most glaring omission is seniority – white people are more likely to occupy positions of leadership and therefore command higher pay. That in itself may be attributable to discrimination but there are other explanations, most notably that it takes time for ethnic minority groups to spread out from their original patterns of concentration. Increasingly, this is what we are seeing with more and more minority individuals taking up positions of leadership in middle class professions such as law, medicine, and the civil service, as evidenced in my report for Policy Exchange, Bittersweet Success.
Overlooked too is whether or not someone is responsible for managing other workers which would contribute to higher pay as well as vary by ethnic group. Other potentially confounding variables would be job prestige and occupational supply and demand.
That occupational rank might be a key variable is lent weight by recent analysis by the Greater London Authority of how it pays its own staff (GLA). It found a difference in median pay between non-white ethnic minorities and white people of 11.5 per cent. However, differences are null or small, within most of the different pay grades that make up the GLA’s hierarchy. As seen in the graph below, at grades 1-6 as well as 12, there are no differences. In the cases of grades 8, 10, 11, and 15 there is a difference in favour of whites of no more than 3 per cent. At grades 7, and 9 there are differences of no more than 2 per cent in favour of non-whites, while for grade 13 it is almost 6 per cent in their favour.
The same graph shows how the top-line pay gap can be accounted for by minorities clustering towards the lower grades.
Figure 1. GLA gaps in median pay between white and non-white, by pay grade (left-hand axis), plus percentage of staff who are non-white (right-hand axis).
The difference in median pay is largely accounted for by grade. Overall, whites earn more at the GLA but that is because there are more of them in the top jobs that are better paid. Within the different grades, that is comparing people with similar jobs, there are little, if any, differences. But you would not know such things from reading an interview with deputy mayor for social integration, Debbie Weekes-Bernard. On the matter of the GLA’s ethnic pay gaps, she said:
“We have moved on a great deal in that we know more about what those gaps are and what’s causing them, and we’re getting better at developing ways to tackle them.”
Tackle what? Why is that when officials have, by their own terms, a success story, do they not shout about it? This is especially frustrating given that Weekes-Bernard was being interviewed by The Voice, Britain’s leading black newspaper. Even more disappointingly, the London mayor Sadiq Kahn is quoted in the same article as saying: “We’ve made progress at City Hall and across the group but this data clearly shows there is more work to be done.”
Returning now to the Resolution Foundation’s research, the chief problem is to infer the presence of discrimination – a pay penalty as they put it – from an unexplained difference. All the regression analysis does is to rule out potential explanations – it is evidence against something rather than evidence for something. Just because we know differences in pay between ethnic groups are not accounted for by things like age or region, does not mean they are necessarily accounted for by discrimination. While it may go on with regard to pay (and we know racial, ethnic and religious discrimination do exist in the labour market when it comes to job applications), the Resolution Foundation’s analysis has done nothing to evidence its existence nor extent. It is at best suggestive, but inconclusive.
The claim that differences in pay between ethnic groups are accounted for by experiences of discrimination is one of spurious correlation – ethnic minorities receive less pay because ethnic minorities are more likely to experience discrimination. Not all will experience discrimination however and if you hypothetically had data on such experiences, then were you to control for it in regression analysis and assuming discrimination was the key explanatory factor, the differences between groups would disappear. However, being discriminated against is impossible to measure in a survey such as the Labour Force Survey since most respondents will not be able to say for sure whether or not they have been discriminated against.
What is overlooked by the Resolution Foundation researchers is that their measure of ethnicity may also be reflecting variations between groups that will persist net of experiences of discrimination. Indeed, the one finding that undermines the entire tone of the report is that there is no ‘pay penalty’ for Indian graduate men. If discrimination is evidenced from gaps, how can it be that this particular subset has no penalty? Is this the first recorded case of philoindiangraduatemenism? (The RF analysis does not include the Chinese but were it to do so, I suspect it would probably show they did as well if not better than the ethnic majority in terms of pay.)
One might argue that just because there is no difference between graduate Indian men and comparable whites, it does not mean to say that this group is not discriminated against. It could be that they deserve more. That would be reasonable but it would undermine the whole inference – gaps evidence discrimination but equality between groups might also evidence discrimination. With this in mind, what is the meaning of the Resolution Foundation’s stated goal of closing gaps? Even if we reach a situation of equality of hourly pay between groups as measured by survey data, it does not necessarily mean there is no discrimination. There’s no way of knowing.
It is also amusing that the Resolution Foundation wants to close gaps between ethnic groups but the interventions mentioned as possibly corrective would benefit all – “helping young people onto better educational and career-related pathways or giving adults access to careers advice and development”. You can’t close gaps by helping everyone.
We also need to consider the moral complexity of the real world. Let us imagine an ethnic minority person. A graduate from a good university, British-born but the child of immigrants. It is possible that this person might be more willing to accept less given that they have less to fall back on and thus have weaker bargaining power (compared to some, though not all, white people), with their strategy being to establish themselves first before securing bigger pay-packets later? While this is not ideal, it is not the same thing as racial discrimination. The reality is that employers will always look to pay as little as they can. And sometimes being able to offer a commodity for less – labour in this case – can be an advantage.
By telling such a simple story of ‘ethnic pay penalties’ – the cost of being an ethnic minority – the Resolution Foundation will receive plaudits and will be invited to the top table as an ‘expert’ organisation, but more to the point, as the good guys. However, in truth, they don’t really know what is going on with regard to pay, equality, and fairness, and no regression model can tell you. All the data say is to remain agnostic about the extent of discrimination when it comes to pay.
Publishing pay data at company level is intended as a nudge to action – a wake-up call for the employer and a spur to action for women and minorities. It is overlooked that men and the ethnic majority will also respond, possibly driving up their own pay as they compete with each other as new information on pay levels comes to light. The complexities of such data are difficult to tease out and easily misunderstood. This is potentially bad news for companies as it may lead to unfair reputational damage and an unintended distortion of pay as well as the creation of more bureaucracy in order to monitor things. This costs money too. (Note that the current mooted proposal is for compelling companies with more than 249 staff, to publish their pay data by ethnic group, but Kahn wants to extend it to even smaller businesses.)
What bothers me the most however is that ambiguous research comes to be presented as fact in the wider conversation, through the media, commentators, and politicians. If you believe that ethnic minority people are being cheated to the tune of £3.2 billion this may lead people to opt out, to not apply for that dream job, thus entrenching separation as well as resentment.
The truth is that there are bad things that happen to minority people that do not happen to the majority, but there are also really good opportunities that Britain has to offer – note that the ethnic Chinese and Indians are much more likely to be in the most desirable white-collar jobs. A relentless and overly-pessimistic focus on the negative will only encourage a lowering of expectations and play into the hands of the grievance hucksters and extremists who would wish to exploit young people. Discrimination in pay based on race or ethnicity will occur – it should be part of any explanation – but that is not to say that an unexplained difference between groups is an accurate measure of its extent.
Finishing on a happy note, there are actually little differences in reported happiness between ethnic groups. When asked to rate their happiness yesterday on a scale of 0 to 10, there are actually no substantial differences. For instance, the average score for Pakistanis is 7.5 out of 10, for Indians 7.8, for black people 7.3, compared to 7.5 for white people. For all the gaps and burning injustices in the world, we would do well not to lose sight of this.
 Grades 3 and 14 are censored by GLA for privacy reasons.