Ability Bias and the College Premium

I recently finished The Case Against Education by Bryan Caplan. It's great. I highly recommend it, so pick up a copy.

A big part of Caplan's story is that the "college premium," the bump in income that people earn by finishing college, is exaggerated. You can't just compare salaries of "people who finish college" to "people who don't finish college" and say that the college education explains the difference. Maybe people who attend and finish college are different from people who don't?

Statisticians have an obvious way of fixing this problem: measure ability, and control for it in a regression analysis. What remains after controlling for "ability" will be the true college premium, eliminating the selection bias whereby high earners disproportionately earn college degrees. Well, sort of. Caplan has an old blog post fleshing this out. Read that for the full details; the links within that post are also excellent. If you believe the numbers in Caplan's post, ability bias accounts for ~40% of the college premium. A raw comparison of "high-school graduates" to "college graduates" grossly overstates the return to education.

Or maybe not! Labor economists (who Caplan accuses of being incredibly dogmatic) come up with excuses. "Ability is hard to measure." "IQ doesn't fully measure ability." All true enough. These labor economists even come up with sophisticated models that make the ability bias correction shrink or go away completely. I've been following this discussion for years. From Caplan's description, it sounded like these labor economists were performing some kind of statistical sleight of hand to make an uncomfortable result go away. And then I picked up Mastering Metrics by Joshua Angrist.

The following passage stood out as being relevant to Caplan's disagreement with labor economists:

...our London School of Economics and MIT students tend to be high ability, at least in the sense of having high test scores and good grades in high school. On the other hand, some people cut their schooling short so as to pursue more immediately lucrative activities. Sir Mick Jagger abandoned his pursuit of a degree at the London School of Economics in 1963 to play with an outfit known as the Rolling Stones. Jagger got no satisfaction, and he certainly never graduated from college, but he earned plenty as a singer in a rock and roll band. No less impressive, Swedish épée fencer Johan Harmenberg left MIT after 2 years of study in 1979, winning a gold medal at the 1980 Moscow Olympics, instead of earning an MIT diploma. Harmenberg went on to become a biotech executive and successful researcher. These examples illustrate how people with high ability—musical, athletic, entrepreneurial, or otherwise—may be economically successful without the benefit of an education. This suggests that ... ability bias, can be negative as easily as positive.

 Then, a discussion of using IQ as a control in various regression studies to measure ability bias, done by one Zvi Griliches:

Although intriguing, it’s hard to see Griliches’ findings as conclusive. IQ doesn’t capture Mick Jagger’s charisma or Johan Harmenberg’s perseverance, dimensions of ability that are rarely measured in statistical samples. The relevant notion of ability here is an individual’s earnings potential, a concept reminiscent of the potential outcomes we use to describe causal effects throughout the book. The problem with potential outcomes, as always, is that we can never see them all, we see only the one associated with the road taken. For example, we see only the “highly educated” potential outcome in a sample of college graduates. We can’t know how such people would have fared if they’d followed Johan and Mick out of college. Attempts to summarize potential earnings with a single test score are probably inadequate.

I read this passage from Angrist's book and thought, "Okay, that's not totally nuts." So maybe high-ability people don't just disproportionately finish college. Maybe some of them anticipate that they can earn high incomes without going to college. Think Mick Jagger or Mark Zuckerberg or Bill Gates or Steve Jobs or Michael Dell. Ability bias can be negative. If the Zuckerberg/Gates/Jobs/Dell effect is large, the college premium can actually be larger than the raw average difference between college graduates and non-graduates. I think that an average reader can probably understand the passage from Mastering Metrics quoted above; it's in plain English without any abstruse mathematical modeling (though those details are in Angrist's book). So I made a note to keep it in mind and bust it out later when this "ability bias" discussion comes up again. And then I saw this exact quote from Mastering Metrics in The Case Against Education, right in the middle of the discussion on ability bias. Good on Caplan for this. He's not just saying, "Labor economists use econometric magic to make ability bias disappear, don't let them snow you." and then leaving the details to the readers imagination. No, he's spelling out in plain English the arguments used to deny ability bias. He wants his reader to understand what critics of ability bias are actually saying, so he can explain why they are probably wrong. He and I both independently identified this quote as the best summary of the "denying ability bias" argument.

I am not qualified to settle this argument. But the "ability bias is negative" story seems pretty implausible. Sure, there are a few celebrity superstars who quit college to pursue extremely lucrative opportunities. But by and large it is the individuals with higher ability who enter and complete college. By and large it is the lower-ability people who leave college early or avoid it entirely. I was there. I saw it happening in real time. Think about your high school and college cohorts. Think about who attended college and who didn't, or who dropped out and who finished. Does the "High ability individuals skip college because they know they can earn more on their own merits" story hold water? Is it even remotely plausible? Based on the hundred or so people I knew in high school and their subsequent education or lack thereof, the "ability bias explains much of the college premium" story makes a lot more sense.

__________________________________________

Here is a great post with an early excerpt from his book; it's on his disagreement with labor economists over signalling and ability bias. A nice taste of what's in the full book.

Signalling and ability bias are completely different topics. I think some readers of Caplan's book might be confused by this, or might conflate these different items. Since I've been reading Caplan's blog for years, this came easy to me. But I remember a post where he lays out these very different explanations of the college premium: signalling vs. human capital vs. ability bias. It was confusing the first time I saw it. I've had years to digest these issues, so maybe it came easier to me than it would to an uninitiated reader. Or maybe the distinction is simple and I'm just dense, and the average reader will have no problem at all? A recent conversation with another reader of Caplan's book suggests that these really are difficult concepts. I suspect a lot of reviews will conflate signalling and ability bias.