That study’s results exemplify a danger highlighted by the great economist Thomas Sowell. Differential economic outcomes across sex or race groups do not automatically imply discrimination. In a genuinely tolerant world where we all lived up to Martin Luther King’s aspiration for judging the contents of individuals’ characters rather than the colour of their skin or sex, we might still see “average group differences” between sexes or races, because of different preferences, or benign cultural phenomena.
Holding up some “gap” in attainment or prosperity as self-evident structural racism or sexism is misguided. Crude data shows that Pakistani and Bangladeshi families have lower incomes than many other ethnic groups. Yet half of women in these communities are economically inactive, compared with a quarter of white women. It would clearly be wrong to assume all the familial income gap then was driven by racism by ignoring the cultural decision by many Pakistani and Bangladeshi women to stay home and forgo income.
Yet overt racism, or prejudicial biases, of course do exist too — the online abuse received by England’s footballers demonstrates that clearly, albeit with much of it originating overseas. Racial prejudice affects life chances and outcomes of some ethnic groups as it manifests in educational, hiring, or promotional decisions, or even “default settings” for goods and services in everyday life.
What are known as “job application” field experiments with different names on the same CVs have shown apparent racial bias from employers. Ethnic minority applicants have been found to have to send 1.6 applications for every 1 application from a majority White British name to receive a callback. These ratios are higher still for black or Indian applicants. At the very least, employers seem less receptive to ethnic minority names.
But assessing how big a problem this more overt racism is overall, how it manifests in shaping outcomes, and whether we can do anything about it is extremely difficult. Most social science research is not based on experiments where everything else other than a name is held constant. And in a world in which, even if all anti-black racism were eliminated, outcome gaps might remain, as they do between black Caribbean and black African ethnicities, how would we even know when real racism ceased to be a problem?
The Government’s Sewell report on race and ethnic disparities earlier this year examined how far non-race factors such as age, class, geography, job, and sex “explained” educational, employment, and health outcome gaps. The thinking was that what remained as “unexplained” might provide an upper bound of racism or discrimination that is harder to identify or model.
This type of analysis is useful in theory. The method helps control for the fact that, say, a younger ethnic minority population might have less wealth simply because the individuals within it have had less time in life to save, rather than because of nefarious racism. The report made headlines as it concluded that “geography, family influence, socio-economic background, culture and religion have more significant impact on life chances than the existence of racism.”
But using all these controls begs the question: what if racial biases themselves influence where people live, the structure of families, or jobs, in other less discernible ways?
The report found that black male senior managers, for example, earned just 83p for every £1 earned by their White counterparts. It suggested this was primarily a function of seniority level. Restrict the analysis to “very senior managers” and the gap shrinks to 97p earned by Black men for every £1 earned by Whites. Yet what if the decision not to promote more black men to very senior roles was shaped by prejudice? Too many control variables might mask real discrimination.
If disentangling this was not difficult enough, “average group differences” themselves can create feedback loops, as knowledge of them affects people’s actions or decisions by shaping expectations about people’s potential.
In the US, concern about young black males with criminal records being locked out of job markets led to “ban the box” reforms, with certain states banning employers from asking about an applicant’s criminal record. Unable to ask the question, evidence from some states suggested employers instead acted on their awareness of the higher probability of young black males having served time. They offered even fewer jobs to young black men after the reform, racially profiling candidates to compensate for the lack of knowledge about the individual applicant.
Unfortunately, this nuance about the difficulties of identifying actual racism — and the crudeness of government interventions designed to tackle it — is lost as debates become a dialogue of the deaf. In today’s culture wars, the extremes that dominate instead either hold up all major disparities as evidence of some ubiquitous societal racism, or else explain so much away that they deny evident injustices.