Let’s not confuse discrimination with prejudice

I recently came across a Facebook post asking why we assume Hermione Granger is white when reading Harry Potter. The not-so-subtle implication was that it’s because of structural racism.  

Perhaps; but it might also be for the same reason we think Cho Chang is Asian: because statistically speaking, every time we’ve come across someone bearing a name like Cho Chang we’ve discovered they were of Asian heritage.

This is an example of a heuristic. In psychological parlance, heuristics are rules of thumb that people use to form quick judgements and make decisions based on probabilities drawn from previous experience. They were extremely important in evolution for, say, quickly deciding that the shadow you just saw is ‘probably’ a snake, or that rustle in the bushes is ‘probably’ a tiger about to eat you.

Heuristics are also useful for making other decisions where information is scarce. For example, when you’re hiring from a pool of 40 applicants you assume that the people with decent grades on their transcript have more ability and/or work ethic than those with weak grades or without a transcript. This isn’t always true, but it’s true enough that following such rules of thumb results in more net benefit to you than you would receive from engaging in the detective work to discover the actual highest ability/effort people in the bunch, because such detective work comes with a high cost in time.

I think we can all agree that there is nothing evil about this. If you want the job you need to send the appropriate signals. Anything more and we would be placing a positive moral duty on the human resources person in the sense that we expect them to hurt themselves for the benefit of others. That’s something we tend to avoid as part of liberal society.

Yet when such heuristics are extended to minorities we cry foul. For example, when firms promote men to middle management because statistically speaking, more women drop out of the workforce at 30 than men we think they’re being evil.

Before everyone adds me to their blacklist, let me emphasise that I want to see the gender and racial wage gaps shrink and I want firms to promote women to middle management and beyond. This article is about how to do that. Bear with me.

Someone may be substantially committed to women’s liberation and still not promote a woman to middle management if there is a strong chance that they will exit the workforce for several years to have children. This is because promoting them comes with a substantial risk that anyone with a concurrent responsibility to the fiscal health of their firm cannot bear.

If we as a society want to see changes in these areas of discrimination then we as a society need to bear the associated risks rather than externalising them onto individuals.

The key thing to understand is that there is a difference between prejudicial discrimination, which is irrational if not evil, and statistical discrimination, which is rational in the sense that people use probability to maximise their payoffs. The tools you need to fight prejudicial and rational discrimination are fundamentally different. A large part of the reason why we aren’t making massive progress on statistical discrimination is that many of the most vocal activists are applying the wrong tools.

Importantly, rage doesn’t help. You can write hundreds of thousands of words (indeed we have) lambasting the gender wage gap, but it won’t change anything (indeed, it hasn’t), because the fundamental cause of that wage gap is not predominantly about prejudice but about significant biological (pregnancy) and behavioural (sick leave because kids, flexible work hours because kids) differences between men and women that we as a society need to develop public policy tools to address (by way of the incentives that underlie them) if we want to see them change. Indeed, because rage encourages defensiveness and rhetorical flourishes it obviates against the kind of considered, meticulous thinking that is required to correct analyse and dissect such complex issues.

Second, guilt tripping people about how they are racist for assuming Hermione is white or that ‘Josh’ is a safer hire than ‘Jose’ won’t work either. Those who feel very guilty will take the risks mentioned above and subsequently go bankrupt in a larger numbers than those who don’t take the risks, and we will be left with dickheads in charge of most firms. The only way to prevent this from happening is to ensure that in all cases Jose really is statistically just as safe a hire as Josh.

The situation is obviously different when Jose is just as safe a hire as Josh. This is an example of a stereotype. The key difference between a stereotype and a heuristic is that stereotypes are by definition incorrect. Heuristics, on the other, have been found to have a good track record. They are a risk assessment done to the best of someone’s ability based on available information. In cases of misinformation what is required is not so much rage, as this won’t disabuse people of their presumptions, but an education campaign (which might effectively utilise a bit of teasing rather than the self-righteousness of the social justice set).

We must of course also acknowledge that there is a dangerous feedback loop present in the situation with Josh and Jose. For example, if women aren’t hired because they often drop out to have babies then it starts to become rational for households to have the woman drop out and have babies, which feeds the tendency of firms to not hire women, and so on. But again, this is an argument for a particular policy intervention (some sort of affirmative action) rather than for raging against the patriarchy, as the behaviour of all parties is rational rather than immoral.

At this point it is important to mention some of the academic literature on prejudicial discrimination. First, the Becker model, which suggests that prejudicial firms (note, not rationally discriminating firms) should go out of business over time because rational firms (remember that prejudice is irrational) will simply hire the best people regardless of minority status. Their resulting higher productivity will make them more competitive than the prejudicial firms in the long run and thus, over time, prejudice will be eroded from the labour market.

Is this true? Well, it’s very hard to test, but we have some studies. There is, for example, the work of Nobel Laureate Jim Heckman, going right back to the late 90s. See here, for example, and here. Heckman’s argument, based on data, of which he is a maestro, is that while differences in market outcomes between minorities are certainly abundant, this comes down primarily to a difference in skills. Prejudice has almost disappeared from American firms in a statistically significant sense, meaning that you might still find it but in the grand schemes of things there is very little of it. As such, the ‘solution’ to disparate labour market outcomes (note the emphasis only on the labour market) is not more civil rights legislation but policies to address the skills difference, like early childhood interventions, on which Heckman has done some positively awesome work. There is a similar study of a large, high profile French firm using payroll data going back decades, which finds similar results.

Now there are also studies that suggest some greyer space. For example, there was a very good randomised control trial of resumes recently in France testing for racial bias. You can get the paper here and the digest by The Economist here. This one found that people responded substantially to changes in the skills of applicants, but that there was also a degree of bias (mild) in the sense that people tended to select candidates from the same minority as them (including white-white, Islamic-Islamic and women-women, man-man). This is much more of a cause for concern, though again I suspect that a policy response in terms of acculturation (that would be hard when ‘cultural appropriation’ and assimilation are dirty words in social justice circles) and affirmative action is more appropriate than rage seeing as how it is a heuristic that people go with what they know because there are risks inherent in the unfamiliar.  This extends to why you don’t see a lot of Chinese characters in Hollywood movies and equally not a lot of white characters in movies from Hong Kong—people write about what they know about—but let’s leave that to a separate article.

A word now on statistical significance before everyone loses their shit. When you say that prejudice is not common in a statistically significant sense it does not mean that prejudice does not exist. It does not mean, for example, that you can’t find a few weird young white men in the American South ready to go on a shooting spree in a black church. It does not mean that you won’t occasionally come across a mentally ill woman raging against an Asian on Sydney buses, or an ex-Olympian telling Greeks to go home. It does not mean that you won’t occasionally come across a small business owner who won’t employ Hispanics. It does not mean that you won’t find people put out by discovering Rue from the Hunger Games is black. It does mean, however, that these incidents will be drowned out when you look at the entire society. For example, how many tweets were sent about the Hunger Games movie? How many of those were racist? Is the probability of selecting a tweet from among a sample of Hunger Games tweets greater than 1%? I doubt it.  

Part of what plagues social justice warfare is that it operates on the internet, where the loudest voices are the most prominent even when they are in the minority. Indeed, such voices flock to the internet because of its anonymity and ability to give you a voice when those around you don’t care for what you have to say. This dynamic becomes particularly dangerous in the context of social justice warriors’ obsession with ‘lived experience’, which is really just a way to trump up anecdotes into more than they’re worth. Now let me stop the rage brigade again. Narratives are important to provide depth, quality and colour to trends and other broader phenomena that are observed through a quantitative lens. But let’s not go pretending that anecdotes are a sensible evidentiary base upon which to enact public policy. 

So I’ve suggested that we shouldn’t see everything as ‘racism’ or ‘sexism’ because this implies prejudice when in fact we often have evolutionarily sensible tendencies going on. Note that I’m trying my best not to generalise here. Some things, like domestic violence, are largely prejudicial in nature. The wage gap not so much.

I’ve also suggested that rage and guilt are perhaps not very efficacious ways to reduce statistical discrimination. What might we do instead?

I can’t comment on most areas of statistical discrimination because I don’t have expertise. One of the few areas where I’d like to think I do have some knowledge though is the gender wage gap. There are a handful of policies that have been implemented in places globally over the past decade or two that seem to work, at least to a degree. They are child care, paternity leave and policies requiring firms to hold positions. I’ll tackle each of these briefly by way of a conclusion.

The key thing to emphasise is that differences in labour market outcomes between men and women are largely driven by one thing: childrearing. For a start, when women become housewives raising kids they stop contributing to superannuation and stop accumulating human capital in the form of workplace experience. When they then get divorced they become welfare dependent and when they reach retirement age they are entirely reliant on the pension. Getting more women onto boards is important, but it pales in comparison to the importance of reducing the gender division in welfare reception. Fortunately, the remedy overlaps with the other major issue, which is that the childbearing years coincide with the point in most careers where advancement takes off along an exponential curve. By the time a woman is ready to return to the workforce after 3 kids her husband is already knocking on the partner door. Her skills have also atrophied and her industry has left her experience behind. It’s hard to restart. This is why we typically observe not just an absolute wage penalty to motherhood but also a wage growth penalty to mothers if they do return to the workforce. So it’s key that we help women retain their connection to the workforce during their childbearing and childrearing years.

Step one: affordable child care. Part-time work is perhaps a possibility once your kids are old enough to have a measure of independence, but what about before then, and what about full time work? Indeed, what about full time work in professional services jobs that require enormous hours? How many women who want to remain connected to the workplace can afford to put their kids into child care for 6 hours a day? With rates as they are a lot of people are giving up their pay packets just to stay in the workforce. Obviously men could take care of the kids instead of the women, but given that a lot of men want careers as well, a more efficacious policy might be to make childcare affordable either through subsidisation or by loosening supply.

In the past decade we’ve gone backwards on this front, partially out of good intentions. Evidence has emerged that early childhood interventions are a very useful policy for breaking intergenerational disadvantage. But providing this service requires substantial skills on the part of child care workers. Unfortunately, you can’t expect people to get skilled if you don’t pay them appropriately. So we pay skilled child care workers more. But the union doesn’t like that, so now you have to pay all the child care workers the same high wage and they all need to get qualified. Supply is restricted. The fuck-up in this outcome is that early childhood interventions are only important for people from disadvantaged backgrounds. ‘Privileged’ kids get the associated benefits from their home environments—they don’t need special child care programs. For more information, read Heckman here or watch this talk. Meanwhile, the mothers of those privileged kids are now locked out of child care services because they can no longer simply pay a clean cut university student $20/hour to take care of 6 kids. The uni-students (mostly girls in this case) are also locked out of a pretty standard job for uni-students. What’s obviously needed is a two-track system where subsidies are provided to relevant parents to enrol their kids in early childhood programs with qualified teachers while other parents are still able to pay a level-headed young person to look after their kids. It’s not rocket science and there is no reason why child care workers would suffer from such an approach.

Step two: paternity leave. One of the most recognised success stories in gender equality is the Scandinavian maternity leave system of daddy-bonuses. The basic idea is that the state provides generous parental leave payments to citizens provided that both the mother and father take the leave. The mother can be eligible for an extension of an additional 6 months but only if the father also takes at least 3 extra months. How does paying men help women? First, it exposes men to childrearing, which they might enjoy, and it encourages a transformation in social attitudes to see male parenting of small children as totally normal. Second, and arguably more importantly, it means that when firms look at their early career staff to consider who to promote they can’t tell whether it is the man who is about to drop out to have kids or the woman. Both of them is an equal risk and so the firm decides entirely based on ability. This breaks the feedback loop mentioned above.

Finally: holding positions. This is an idea principally out of Germany. Firms above a certain size (who can therefore fudge task allocation a lot more than small businesses) are required to hold an employee’s position for up to three years if they go on paternity leave. I am struggling to find the most specific reference (I have sent out for it), but this one is pretty good. Such employment protections have been found to have only negligible impacts on the motherhood wage penalty but have also be found to have dramatic effects in terms of reducing the motherhood wage growth penalty. After returning to work women are able to quickly resume the career track they were on before children and experience the same wage growth over time as their compatriots who did not have children.

So there you have three solid policies with extensive impact evaluations ready to roll out in Australia. Have you heard much about them? Tony Abbott wanted to introduce a generous maternity leave scheme to get high flying women out of the workforce. That’s the opposite of what we want. Have you seen anything from Labor? Nope. But they better get to it because I suspect Malcolm Turnbull, that filthy technocrat, might be firming on this stuff. I can’t help but think that part of the reason we haven’t heard anything is because all the math in the academic discussion of these things is too much for both ministerial staffers and social justice warriors. We can do better. These apparently passionate activists could pick up the phone and have a chat with the experts, for example.

Meanwhile, Daily Life, that bastion of feminism in Australian media, has a total of 4 articles on child care only one of which involved policy rather than personal stories. Jezebel had 1, about Amazon’s policy. Daily Life also had 1 article on parental leave for men and it was basically just a re-hash of what Annabel Crabb said. Yet they have dozens of articles on the gender wage gap. This is what I mean by rage not doing anything. Our public discourse is littered with rage and anger at statistics people barely understand instead of a more sophisticated discourse around what to do about it. I don’t understand why we fight statistical phenomena like the wage gap with vague cultural arguments like ‘Dear White Men: you don’t always get to speak first’ instead of with statistical tools. Again, perhaps it’s because the wage gap is a single number (though even just the fact that it’s an average across all women and all men seems to escape most commentators), which is easy to understand, while reading an academic article on child care inevitably involves a regression, which no Australian humanities graduate is equipped to understand.

Granted, there are some ways the ‘patriarchy’ does figure here, and so in fairness, I will end with some passing comments on it. These policies are all quite expensive. If we want to fund them we need to express our cultural values in favour of women’s agency and in support of motherhood through the democratic process and lobby for greater fund allocations in this direction, and potentially more taxes to boot. As a society, we should recognise that motherhood is fundamental to the health of humanity but that careers are also a huge part of agency for a lot of mothers (and fathers).

As I said earlier, as a society we must bear the costs of bridging these two tracks so that mothers can also be high flyers and boardroom bulldogs can also be ever-present fathers, rather than trying to externalise these costs onto individuals. There is a sense in which our resistance to such a movement comes down to the fact that we are more patriarchal than say, Norway. There is a role here for cultural discourse, but it is a nuanced discourse informed by the rigorous policy analysis that takes place in our universities and think tanks. Critically, it is not about good vs evil, but about transitioning from a cultural paradigm that has served a purpose up to this point but is now increasingly unsuitable to our survival, agency and development. There are certainly people who maliciously adopt the tropes of patriarchy (like the cretins on Return of Kings). But for the most part, culture is the product of heuristics and often very loving, well-intentioned childhood environments involving carers from all sexes/genders. People aren’t fundamentally evil, and behaving as though they are and this is the root of the problem gets us nowhere. We need calm discussions free from guilt, shame, aggression and self-righteousness to foster consensus on how to proceed. This is the opposite of the kind thing social justice warfare fosters. We can do better.