Saturday, 22 July 2017

Do You Know Good Discrimination From Bad Discrimination?

Economics tends to distinguish between two types of category discrimination. The first category is what's called animus discrimination, which is discrimination motivated by aversion to a particular group of people based solely on things like skin colour or sex. An example would be something like not employing a woman because you're a misogynist, or charging a black man more for a car than a white man because you're racist.

The second category of discrimination is what's called statistical discrimination, which is discrimination motivated by statistical trends associated with groups. An example would be an insurance company charging women a lower premium than men on the basis that they are statistically safer drivers, or counter terrorism authorities targeting more Muslims at airports because there is a higher risk that they will be terrorists.

The first type of discrimination is rightly frowned upon by everyone most of the time. Although even animus discrimination has its place in society. At school I actively discriminated in favour of girls over boys in the dating market on grounds that I was only attracted to girls.

The second kind of discrimination, statistical discrimination, is often frowned upon too. But not always wisely, as people regularly miss the fact that statistical discrimination is often a very wise kind of discrimination. Why wouldn't we target more Muslims at airports if there is a higher risk that they will be terrorists? It's a no brainer.

Statistical discrimination is simply a case of someone making what they believe is a rational distinction between good and bad outcomes, like when blind people aren't recruited as lifeguards, and when snake-handling shamans are not hired as primary school teachers. Other times people think unfair discrimination is going on when, in fact, there is a perfectly rational explanation for it.

A huge proportion of the debates we see in the media about so-called unfair discrimination are really enquiries about whether the discrimination is animus discrimination or whether it is statistical discrimination - that is, whether the discrimination is coming from a bad place or a good place. Sometimes, like in this example, it is not always easy to tell.

Recently I read about a couple of natural field experiments of significant interest in this matter. John List and Uri Gneezy are to behavioural economics as Jimi Hendrix is to guitar playing, and they have an interesting case study on what appears to be animus discrimination against the disabled when it come to getting their car fixed:

"First, we recruited several men between the ages of twenty-nine and forty-five to act as our secret agents. Half these men used wheelchairs and drove specially equipped vehicles. The other half were non-disabled, but in all cases the individuals hopped into a specially equipped vehicle for the disabled with a fresh ding on the side and headed to Chicago-area repair shops. When our secret agents got to an auto repair shop they simply asked for a price quote to fix their car. What we found initially was shocking. The disabled were given quotes 30 percent higher than the quotes given to non-disabled for the exact same repair!

Curious about the extent to which car repairman were motivated by hatred or just profit motive, though, we did one run of the experiment where both types of our secret agents got a quote and told the repairman that they were, “getting three price quotes today. What did this extra sentence do? Well the figure shows that for the able-bodied subjects, their price quotes didn’t change at all, but for the disabled they plummeted. Furthermore, the difference in prices for the disabled and abled disappeared."

Whichever way we cut the cloth, it seems like the repair men were engaging in blatant economic discrimination by charging more because of a customer’s disability. Here's another one: if you were looking for an iPod, would you care if the person holding the device in a catalogue was black or white? Surely not, right?

But economists Jennifer Doleac and Luke Stein suggest otherwise. Over the course of a year, they placed hundreds of ads in local online markets, randomly altering whether the hand holding an iPod for sale was black or white. Here's what they found:

"Black sellers do worse than white sellers on a variety of market outcome measures: they receive 13% fewer responses and 17% fewer offers. Conditional on receiving at least one offer, black sellers also receive 2-4% lower offers, despite the self-selected-and presumably less biased-pool of buyers. In addition, buyers corresponding with black sellers exhibit lower trust: they are 17% less likely to include their name in e-mails, 44% less likely to accept delivery by mail, and 56% more likely to express concern about making a long-distance payment. We find evidence that black sellers suffer particularly poor outcomes in thin markets; it appears that discrimination may not “survive” in the presence of significant competition among buyers. Furthermore, black sellers do worst in the most racially isolated markets and markets with high property crime rates, suggesting a role for statistical discrimination in explaining the disparity."

Is this kind of discrimination coming from a bad place or good place? Maybe many white people still harbour an evolutionary hardwired animus-type of negativity towards black skin, even though in terms of outward behaviour most don't act as if they do. Or maybe the black skin is serving as some kind of misguided proxy for black associations like higher crime rates (not that that's an excuse). It's difficult to know. One test might be to measure this against whether black sellers do particularly badly in high crime cities. Another test might be to use only white people, but with half the models having a neck tattoo and half not. Both of those tests might give more of an indication about the discrimination was animus or statistical discrimination.

Whenever we look at discrimination, our first job is to ask whether the discrimination is unfair and socially noxious (in which case carry on doing our best to stamp it out) or socially desirable, as much of it is (in which case carry on promoting the benefits of discrimination). Because remember, in the second kind of discrimination, trade and competition depend on it. If movies do better with attractive leads, then it's in the filmmakers' interest to hire attractive actors; if a 7 ft tall man is going to have problems being a fireman, he should not be hired; and if a shopkeeper doesn't want to hire a young man because he is a member of the BNP, he is perfectly entitled to make that call.

Remember too that imprudent animus discrimination by buyers or sellers in the marketplace penalises in terms of personal costs to the discriminator, as I explained in this blog post. On the other hand, statistical description is usually perfectly fair and desirable because as the above examples show, it starts with everyone on a level playing field and then looks to preclude those that are no longer objectively on that field.

Employers want to hire those who are most productive, just as the counter terrorism organisations want to investigate the people who are most likely to be terrorists. There is nothing unusual about that. The only thing that's unusual about this matter is how many people there are in society that do not really understand basic statistical principles.