Job market discrimination has been at the centre of the last few months’ furore about the lack of transformation at universities across South Africa. On the one hand, those appealing for faster transformation believe, I surmise, that black candidates are not hired or promoted because of discrimination by the panels making the appointments. In contrast, those citing practical difficulties of hiring black staff believe, I surmise again, that the constraint rather lies with a shortage of black candidates.
I suspect both these premises are true. To solve the second – i.e. to increase the supply of black candidates – is a time and resource-intensive process that universities have neglected for too long but which the events of the last few months have certainly hastened. Perhaps that is the topic of a future blog post. Instead, I’ll focus on the former: discrimination at the hiring and promotion level. If this is indeed happening, and I think there is some evidence to suggest that it is, what can be done to forestall such discrimination and speed the much-needed transformation of universities? I should note that discrimination in the workplace may be overt through an undisguised preference for a specific race or gender. When and where this happens, there should be immediate action through the correct channels. But my suspicion is that it more commonly manifests unconsciously – a preference for the culturally familiar (“he went to the same school as I did”, or “she speaks the same language”) which reduces transaction and decision-making costs. Discrimination is then borne through a rational desire to hire the best candidate in a world of asymmetric information.
This type of discrimination was most famously demonstrated by Claudia Goldin and Cecilia Rouse in a 2000 American Economic Review paper. They showed that a change in the auditioning procedures of symphony orchestras – adopting a ‘blind’ assessment with a screen to conceal the candidate’s identity from the jury – increases the probability that a woman will be hired. Because conductors are often male and orchestras male-dominated, male musicians were considered (by the judges, also male) to be better than their female counterparts. Anonymous assessment (playing the violin behind a screen so that the candidate’s gender could not be seen) reduced the bias and meant that more females joined the orchestra.
There is a long history of studies showing similar biases for race. In the US, studies have repeatedly found that résumés with traditional white names are substantially more likely to lead to job interviews than identical résumés with distinctively black names. Roland Fryer and Steven Levitt have found that these effects, though, are not causal for later life outcomes: “we find no compelling evidence of a causal impact of Black names on a wide range of life outcomes after controlling for background characteristics”. Instead, black names better predict life outcomes for the parents than the children.
But given the evidence that people discriminate on race or gender, how does one mitigate the possibilities of such (unconscious) discrimination? Of course, it would help to have more diverse appointment and promotion panels, but given the much-highlighted current racial profile of staff, this will either not be feasible or it would impose heavy costs on black staff who now would have their days filled by sitting on appointment and promotion committees. Another option is to have anonymous résumés and promotion applications. Anonymous résumés or CVs would, theoretically, allow all candidates to be judged according to the exact same criteria, with the top five chosen for the interview process. It is difficult to imagine how to conduct anonymous interviews, but it is not entirely impossible: technology may allow each interviewer and interviewee to create a digital avatar, with the conversation through a Skype session and a mechanical voice transformer. It sounds silly, but it may be the only way to rid the selection committee of any biases (good or bad) they may have towards the candidate. Similarly, promotion applications could be done anonymously, with theoretically only the best candidates making the grade.
So why aren’t we seeing more of this? Well, a new paper just published in the American Economic Journal: Applied Economics suggests we should be cautious. The authors
evaluate an experimental program in which the French public employment service anonymized résumés for firms that were hiring. Firms were free to participate or not; participating firms were then randomly assigned to receive either anonymous résumés or name-bearing ones. [They] find that participating firms become less likely to interview and hire minority candidates when receiving anonymous résumés.
They ascribe their surprising results to two things: 1) self-selection into the voluntary programme and, 2) anonymization prevents the attenuation of negative signals when the candidate belongs to a minority. The latter may be most revealing for South Africa: it simply means that making things anonymous makes it harder for the adjudicators to account for the poorer performance of some candidates at a younger age. Here’s one example: the high school marks of a black South African who attended a poor school may be lower than a white South African who attended a better school, despite the two of them having the same capabilities. An anonymous panel – where school names are removed – would not allow adjudicators to see the high school marks in the context of a poor-performing school. (Education economists in South Africa know that a 80% math score for a kid in a poor school is a much stronger signal than a 80% math score for a kid in a wealthy school.) Instead, anonymous procedures do not factor in past inequalities. Anonymization perpetuates past inequalities; everyone is treated equally but unfairly.
The French government, having hoped that the system would lead to greater representation of minorities, actually abandoned the system as soon as the results became known. It just shows again that what might sound plausible in theory is not always the policy reality. The equal treatment that anonymity provides does not result in equal opportunity. That’s why policy evaluation is so important, and why addressing inequality so difficult.