Statistical discrimination is externality deliniation

Discrimination based on real group average characteristics is a kind of externality within groups. Observers choose which groups to notice, then the behaviour of those in the groups alters the overall reputation of the group. We mostly blame those who choose the groups for this, not those who externalize within them. But if  we somehow stopped thinking in terms of any groups other than the whole population, the externality would still exist, you just wouldn’t notice it because it would be amongst all humans equally. If someone cheated you, you you would expect all people to cheat you a little more, whereas now you may notice the cheater’s other characteristics and put most of the increased expectation on similar people, such as Lebanese people or men.

Does this perspective change where to lay blame for the harm caused by such discrimination? A bit, if the point of blame is to change behaviour. Changing the behaviour of the category makers is still useful, though we probably try to change them in the wrong direction sometimes. But another option is to deal with the externalities in the usual fashion: subsidise positive externalities and tax negative ones. This is done via social pressure within some groups. Families often use such a system, thus the derision given for ‘bringing shame to the family’, along with the rewards of giving parents something to accidentally mention to their friends. Similar is seen in schools and teams sometimes I think, and in the occasional accusation ‘you give x a bad name!’, though that is often made by someone outside the group. I haven’t heard of it done much in many other groups or via money rather than social pressure. Are there more such examples?

One reason it is hard to enforce accountability for such externalities is that boundaries of groups are often quite unclear, and people near the edge feel unfairly treated if they fall on the more costly side. The less clear is the group boundary the more people are near the edge. Plus people toward the edge might only be seen as in the group a quarter of the time or something, so they aren’t externalizing or being externalized to so much. Families are a relatively clearly bounded group, so it is easier for them to punish and reward effects on family reputation. Gender is a relatively clear boundary too (far from completely clear, but more so than ‘tall people’), so I would expect this to work better there. Could women coordinate to improve the reputation of women in general by disrespecting the ones who complain too much for instance? Should they?

Of  course in a few areas making one group look better just makes another group look worse, so if all the externalities were internalized things would look just as they are. I don’t think this is usually the case, or the entire case.

8 responses to “Statistical discrimination is externality deliniation

  1. This implies that a good affirmative action program would be one that subsidised reputation-enhancing activities. The current system of lowered standards may be detrimental to the targeted group, as Larry David refers to in Curb Your Enthusiasm to a friend with a black dermatologist – “You let him treat you even with affirmative action?” Larry claims it was only a joke, but he makes a salient point: you should rationally lower your expectations for a group with lower average quality that can’t be observed directly. It is thus possible that affirmative action could create the kind of discrimination it was supposed to help stop. Subsidies for college might be a better alternative, although since college attendance can be observed, the biggest externalities would arise for completely unobservable activities. I am not sure what they be, perhaps some kind of civic activities.

  2. Great analysis, Katja. I feel this is one of the things Western societies understood in the past, but have now mostly forgotten.

    An interesting social pressure example of “subsidising positive externalities and taxing negative ones” concerns feminists and the issue of women in the workplace. When an employer hires someone, they’re gambling on the employee’s stream of future productivity. In the cases where the gamble takes several years to pay off, young women have traditionally been suspect because of the risk of maternity leave. Since employers don’t have perfect information on which women will have kids in the future, this casts a shadow on the entire gender. (Many governments attempt to prohibit gender discrimination, and I would argue that there are significant positive externalities to successful enforcement, frequently enough to overcome the employers’ individual losses from refraining.)

    Feminists try to combat this by trying to convince women to prioritize work over family. Basic evolutionary logic suggests that this approach is doomed to fail in the medium term–the next generations of women will largely be descended from those who are currently disregarding feminist exhortations, and the tendency to do so is partly heritable, both genetically and memetically. (Widespread adoption of transhumanist-flavored technologies like artificial reproduction could change this in the long term.)

  3. Groups almost always overlap, thereby muddling attempts to pinpoint the sources of externalities. Sometimes groups are even conflated (e.g. primarily Jamaican criminals are regarded as “black” criminals, when African immigrants as underrepresented in crime, and the communities don’t intersect much), again rendering the externality model useless. Some supposed externalities really aren’t (e.g. the idea that black people like fried chicken, grape soda and watermelons). Subsidise/tax doesn’t have a uniform effect on the perceptions of every other group. Some subsocieties in the Middle East, Afghanistan and South Asia are notorious for honour killings (usually committed by family members). These can be regarded as attempts to tax “shameful” behaviour, and although such acts raise their esteem among a few neighbouring tribes, they make the group look like brutal savages to other groups.

  4. Internalizing externalities incurs costs. Are these costs lower than simply the category-makers seeking out further information on the particular individual being judged? If we assume category-makers are rational and self-interested, they would have a greater incentive and ability to improve the accuracy of their judgment. This is because for individual group members, their censure ain’t going to have much of a marginal effect.

  5. Pingback: On behalf of physical things | Meteuphoric

  6. Wow, I hadn’t even thought of discrimination in terms of statistics before. It’s intriguing to note that the examples you use are men and Lebanese people – probably for their tendency to be negatively represented in the media (?). Do you think this is statistical discrimination is relevant to the status of women in Australian society?

    Also, @Henry, it’s interesting to think about affirmative action in terms of lowered expectations. It’s pretty solid in theory, and you can rationally see the impacts that it could have, but do you think that people actually operate that way in society? When I go to the doctor, I wouldn’t tend to question his or her expertise unless there were something glaringly awry in the way that they were acting (though this has never happened to me (in Australia) and I probably wouldn’t notice professional misconduct in a doctor until it was too late). Aren’t people more likely to trust the professional for his/her position before they question gender, race or sexuality? Perhaps I just haven’t had enough exposure to affirmative action in Australia.

    There’s some pretty interesting discussion of discrimination (particularly against women in the workforce and refugees) on but it doesn’t touch at all on statistical discrimination, and there isn’t much solid argument against affirmative action. It would be great if you guys could check it out and post your opinions. :)

  7. Pingback: Does SI make everyone look like swimsuit models? | Meteuphoric

  8. kudos for this purpose page, i just uncover new elements.


Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s

This site uses Akismet to reduce spam. Learn how your comment data is processed.