Co-blogger John Lee has written an interesting post on the place premium — the extent to which a person’s current location affects that person’s earnings — and how it compares with various estimates for the “gender premium” — the estimates for how much less females earn than males with the same skills and qualifications. John focused on the gender premium, but pretty much the same observations can be made about the “race premium” — for instance, about the wage differences between whites and blacks in the United States with the same skills and qualifications. I was planning to write this current post before I saw John’s, but since he’s already written his, I’ll stick to a few points that I think are important but that I think were not salient in John’s post.

My main point is this: race and gender are intrinsic to a person’s identity. Sex change operations are rare, though not unheard of. Race change isn’t possible — though a person can change the race he/she self-identified with, the person’s race as perceived by others cannot be changed through an act of fiat. This is important in two ways.

Measurement uncertainty

The ideal way to measure a “premium” (place, race, or gender) is to take a fixed person, start that person in one place (respectively race, gender), then change the person’s place (respectively race, gender) and then measure how much this affected that person’s earnings. Now, in the case of race and gender, this type of ideal “controlled experiment” is almost impossible. In the gender case, it’s impossible unless you can persuade people to undergo sex change operations — but people who undergo sex change operations are likely to not be representative either before or after their operations. In the race case, it’s also impossible or nearly so.

There is considerable debate on the gender premium in the United States, for instance, largely because it’s very difficult to figure out what exactly it means for a female and a male to be identical in all respects other than gender. Are we accounting for various unobserved characteristics? Different choices of how to break the data down (by profession, sub-profession, years of experience) yield different estimates. Which of these choices is justified? The Atlantic interview that John linked to from his blog post highlights some of the points of contention. Bryan Caplan, who is skeptical of the existence of a huge gender premium or race premium in the United States, has lecture notes on the key points of contention.

Now, for the place premium, this kind of experiment is relatively easier. It’s difficult to do it in a completely controlled manner, because countries are unlikely to allow immigration or emigration just for the sake of a social science experiment, and individuals don’t make migration decisions simply to satisfy the curiosity of social science researchers. But, there is enough natural migration, and enough data on migrants’ earnings pre- and post-migration, to at least give estimates that are far more reliable (in order of magnitude terms, though not necessarily in terms of digit precision) than for race or gender. An obvious criticism of natural migration is that it is prone to self-selection for at least one obvious reason: the people who choose to migrate are precisely the ones who have the most to gain from it. And there are probably other selection factors.

Nonetheless, various convoluted immigration laws allow for “natural experiments” in this area that closely resemble randomized controlled trials. The canonical paper on the place premium by Clemens, Montenegro, and Pritchett mostly engages in educated guesswork on migrant selectivity, but it does have one reference that explicitly discusses a lottery “natural experiment”: this paper by McKenzie, Gibson, and Stillman (if you can’t access the PDF, check out this ungated version). And that finds the same order of magnitude of place premium. In other words, while migrant selectivity definitely does exist, the place premium survives, to roughly the same order of magnitude, even after controlling for it.

I think this puts the existence of the place premium on far more solid footing than the existence of race and gender premia.

I would also add here a bit about economic theory: market forces would tend to reduce gender and race premia, and we theoretically expect the magnitudes of the race and gender premium to be relatively small because the profit incentive punishes people’s urge to discriminate. Now, market forces would also tend to reduce the place premium (down to a factor of 2 or less, as John says in his post), if the market were allowed to operate freely, i.e., immigration restrictions were removed. Thus, the persistence of a place premium is not in conflict with the existing research literature that suggests that huge wage differentials for identical workers cannot survive market competition.

We can fix this

This is a point that John Lee makes in his blog post, but it deserves reiteration, perhaps from a somewhat different perspective. To the extent that a gender premium exists in the United States, it’s quite unclear what can be done about it. The solution offered by many is more legislation, such as the Lilly Ledbetter Fair Pay Act. This may or may not work. The point I am trying to make is that such legislation necessarily involves a trade-off, in that it adds more layers of complexity and regulation (and litigatory uncertainty) to the employment process in return for promises of pay equity. Whether the trade-off is worthwhile is a moot point. But the point I’m making is that there are diminishing returns to heavy-handed laws that aim to address a gender or race premium, if these laws add to uncertainties in hiring and litigation. At some point, even with a positive pay gap, it might be the case that more efforts to address these through legislation do not pass a cost-benefit test. There are obviously non-legislative approaches to tackle pay gaps, but it’s unclear that a lot more can be done with the stroke of a pen than is currently being done.

The case of the place premium is strikingly different. Governments do not need to work to protect the rights of foreigners against wage discrimination. What they need to do is literally get out of the way and end the practice of affirmative action for citizens. This is a solution that should simultaneously please libertarians, utilitarians, and egalitarians — governments enhancing efficiency, ensuring equality of opportunity, and helping the poorest people by getting rid of regulatory red tape that prevents employees and employers from entering into voluntary arrangements. What’s not to like about that?