The "college wage premium"—the earnings gap between graduates and nongraduates—has been rising for long enough now that people sometimes think of the economy's increasing demand for highly credentialed workers as something like a law of nature. But there have been times in the past like the 1960s and 1970s when it's been flat or falling, and Peter Orszag has a good column rounding up some evidence that this may be happening again.
An idea I wanted to introduce into this, however, is that we shouldn't be so blithe about identifying formal education with "skills"—it's possible for the economy to change in ways that simply start rewarding a different set of skills than the ones colleges teach. Think about the spinning class at your local gym. You can be reasonably confident that the job of leading that class isn't going to be replaced by a machine because it would already be trivially simple to replace the instructor with a recording. People opt for the live instructor because they want a live instructor, not because it's impossible to conceive of a workable alternative to the live instructor. So if there's more and more automation, more and more job opportunities will end up having that quality where it's not that you couldn't be replaced by a machine (because soon everything will be in the category "can be replaced by a machine") but because sometimes people just strongly prefer to interact with an authentic human being.
At that point your wage is going to be determined primarily by your customer service skills. Are you pleasant to deal with? When people think to themselves that they'd rather interact with a human being, they typically don't have a grumpy and dyspeptic human being in mind—they're thinking of a nice, cheerful, helpful human being. And obviously this is a real skill. We've all had really good customer service experiences in our lives and also had really bad ones. But I'm not sure this is a skill that's well-acquired by getting a high SAT score and then hanging out with other people who had high SAT scores for four years while listening to lectures from very distinguished academic researchers. Obviously it's already the case that a lot of what's taught in college has very little direct application to people's jobs. But skilled academic researchers are necessarily an example of what Robert Reich calls skilled "symbolic analysts" and the skills of symbolic analysis that one learns from them plausibly have some transferability into the kinds of symbolic analysis that are relevant to high-wage jobs more generally.
But computers are getting much better at symbolic analysis than they are at having a warm smile or a firm handshake or an appropriately encouraging tone of voice. Those kinds of skills aren't great skills to learn from hanging out with college professors. So if which skills are valuable shifts, you would see a decline in the college wage premium not because "skills" are less valuable but because colleges teach a particular kind of skills.