Suppose you read a blog post that started like this:
It is well known that if you run a regression analysis of wages, you will find that if a person tortured kittens as a teenager, then he or she will have significantly lower earnings in the marketplace.
Now there are two schools of thought when it comes to explaining this outcome. One is the kitten torturing human capital theory, which says that people who torture kittens as teenagers degrade their inner moral compass, and thus reduce their productivity. Thus, it shouldn’t surprise us that employers pay kitten torturers less than they pay other workers.
The other explanation is the kitten torturing signaling theory, which says that kitten torturing per se doesn’t directly impact productivity, but the type of person who tortures kittens is also not likely to be a team player in the office. Thus, it shouldn’t surprise us that employers pay kitten torturers less than they pay other workers….
Isn’t it obvious that there is something odd going on here? Employers don’t know if someone tortured kittens as a teenager; that’s not on a criminal record (usually?) and nobody puts it on his resume. So clearly the second school of thought–the signaling model–can’t be what’s going on, because you can’t use a “signal” that is invisible to everyone.
On the other hand, the first school of thought doesn’t necessarily seem right, either. Maybe people start out with a screwed up moral compass (or maybe they are damaged by abuse), and then the kitten torturing as a teenager is just a symptom of something deeper.
It seems then that there is room for a third school of thought, which says: There is a strong negative correlation between kitten torturing and low productivity in conventional jobs, meaning that if employers have a tendency to pay workers their marginal product, then we will observe a strong negative correlation between kitten torturing and wages.
It’s the part I put in bold that seems absent from Bryan Caplan’s series of musings on this issue. For example, in his most recent post (drawing on his forthcoming book) he writes:
To weigh the power of human capital versus signaling, however, we must zero in on occupations with little or no plausible connection to traditional academic curricula. Despite many debatable cases, there are common occupations that workers clearly don’t learn in school. Almost no one goes to high school to become a bartender, cashier, cook, janitor, security guard, or waiter. No one goes to a four-year college to prepare for such jobs. Yet as Table 4.6 shows, the labor market comfortably rewards bartenders, cashiers, cooks, janitors, security guards, and waiters for both high school diplomas and college degrees.
But it seems to me Bryan has missed out on a better explanation: It’s not that their degrees are “signals” of their productivity. Rather, the college grads who get paid more as cashiers, cooks, janitors, security guards, and waiters are getting paid more because they are directly observable to be more productive (in general) than the ones who don’t have college degrees.
There are certain occupations where it’s not easy to assess “marginal revenue product” the way we assume in introductory textbooks. The employer can’t just draw his nice curves and find the tangency point.
But it’s pretty obvious after a week or two whether someone hired as a cook, or a janitor, is worth the wage he is getting paid. Maybe security guard or cashier is a grayer area, because you’re not sure if the former is sleeping on the job or if the latter is figuring out ways to let his friends buy stuff for free.
But as far as cook and janitor, there’s no “signaling” needed. It might help someone get his foot in the door, but that’s not really the explanation for the person’s consistently higher pay, any more than a fancy-looking resume could explain it. If it happens to be the case (and I have no reason to doubt Bryan’s data) that cooks with college degrees get paid more than cooks without college degrees, I endorse neither the human capital nor the signaling model to explain this result. Rather, I would say there is a correlation between the type of person who gets a college degree and a good cook, versus the type of person who doesn’t get a college degree, and that employers in this industry can readily observe productivity and thus pay cooks accordingly.
NOTE: I’m not “blowing up” Bryan’s post. He has obviously read way more of this literature than I have. My modest point is that his contrast between the human capital versus signaling models of education often seems to me, to artificially force the world into two categories that are not exhaustive of the possibilities.