Here’s today’s “The Tuition is Too Damn High” from the Washington Post, the second of a 10 part series. It asks if college is worth it, arguing from data on income.
I still think this project misses that which is most important. The only conceptual dependent measure is money, but surely education is about something else that is not counted here.
But let’s assume that money is all that matters. The author writes: “So does college raise incomes? Is it an investment good enough to make widely accessible? Yes, it is. Period.”
Is that conclusion warranted? No, it is not. Period.
Why not? Consider the types of data.
(a) The author cites various correlational studies. It is clear that those who go to college earn more money. But correlation is not cause. Perhaps some other variable (e.g., diligence, intelligence, social skills, having a four leaf clover) makes people both go to college and earn a lot thereafter. The author cites some studies that control for this or that variable. But do they control for all of them? Imagine that both diligence and intelligence increase (separately) the likelihood someone will go to college and make money. If you only control for diligence you’ll find that going to college still predicts income. If you only control for intelligence, same thing. You need to control for both before you find that college isn’t related to income after all. Now, imagine you do control for both. Maybe there’s a different variable you’ve not thought of, something that will be obvious in a century or millennium, but not now. Then you are back to mistaking correlation for cause. You just can’t know cause from correlational data.
(b) Experiments. Experiments do get to cause. But: “Truly experimental evidence is hard to find, as it’s difficult to design a study in which people are randomly assigned to go to college or to stay home,” writes the author. Just repeat that about 20 times and forget the rest of what he wrote (other than the bits about the marginal student and signaling, both of which I found interesting). Without true experiments you just aren’t going to get at cause, much as you might wish you could.
The author reports one vocational education program that was randomized and worked. But…is that the same as a liberal arts education? Do we really want to conclude that since vocational training raises income liberal arts training does as well?
The author reports on a GI Bill study for folks randomly assigned (by draft lottery) to the military or not. But…(a) there was a lot of draft avoidance which makes this not completely random, thus not a true experiment; and (b) as best I can tell there’s a confound–experience in the military. The author also reports on differences between programs in Ontario and Quebec, but folks are not randomly assigned to be from one or the other of those two!
There’s virtually no experimental evidence.
We don’t know the effect of college education on income. We really don’t. We won’t till we start doing randomized experiments and those aren’t going to happen. So we’ll be left with uncertainty.
Look. I love a college education. I think it’s a great thing. But the fact that I love it should not mean that I am more confident it boosts income than is warranted by the data.
And, to return to the original theme, we reduce the value of higher education, we reduce the experience of higher education, by turning it into a way to get high-paying jobs. I’ll repeat my Shea quote from the other day: “The sacred pearl on the forehead of the goddess has fallen to mere wealth.”