My first (and alas, last) cover story for Newsweek was on whether there's a college bubble. I got a lot of pushback from college professors, as you might imagine. Now Peter Orszag is asking essentially the same question: "why are so many college graduates driving taxis?"

Tyler Cowen suggests the answer is not that the value of the degree is falling, per se, but rather that the variance of the returns to education is rising.

Quick tutorial for those of you who didn't major in finance: the expected value of an investment that returns exactly $100 every year is the same as that of an investment which has a 50% chance of delivering $50 in any given year, and a 50% chance of delivering $150. But the second has much more variance than the first.

In terms of college education, what that means is that some people get very high returns, while others get low or negative returns. Looking at the mean or even the median can thus be misleading. If Bill Gates had walked into my office back when I was a tech consultant, both the mean and the median would have gone up, but that wouldn't necessarily have made it a good idea to go into technology in the summer of 1999.

Here's one way to think about this story: as the country urbanizes and mechanizes, in the 19th and 20th century, the return to learning reading and math rises dramatically. Someone who can do basic math has a big advantage in the labor market; they can be a store clerk, or manage all sorts of businesses. But someone who can do very advanced math isn't necessarily that much more employable. To be sure, there are professorships, and ballistics experts in the Navy, but these aren't actually all that well paid.

Fast forward to the 21st century. Cash registers don't even require you to make change. For that matter, they barely need to read; the cash registers can have pictures on them. So the returns to basic math ability have fallen dramatically. Meanwhile, the return to advanced mathematical ability have risen to the point where someone who has the math skills to, say, work in the tech or finance industries, can average many many multiples of what those with more basic skills can do.

As automation is taking over more and more of the basic tasks, having a middling set of skills is declining in value, and having very high level skills is rising. Those will show up as a high average return to a college degree. It may even show up as a high median return to a college degree, if the very basic skills taught in high school are being completely devalued. But that conceals a great deal of variance in the outcomes: for some people, the expense isnt' worth it at all.

As I said a few weeks ago, this is troubling in the face of all the problems people are having in the lower reaches of the labor market. Education has long seemed like the uncontroversial cure-all for these sorts of issues. If education only entrenches inequality, instead of helping to increase it, what do we have left?