Those who’ve been reading this series will recall that I didn’t think much of the “Baumol cost disease” explanation for why college costs are going up. That’s the theory that says that because you can’t get more productive at standing in front of a class and lecturing, to compete with other sectors where productivity is growing, higher education institutions are going to have to pay faculty ever-higher wages.

I don’t think there’s much evidence that that’s happening. But there’s a kernel of truth there. The way we teach hasn’t really changed in millenia. The best way to reduce college costs going forward would be to figure out a way to get better at making students smarter.

A growing movement is trying to do just that. By now, you’ve probably heard of Massively Open Online Courses, or MOOCs for short. Those are courses offered online, often for free, without admissions requirements or enrollment limits. Anyone who wants to can watch the videos and do the assignments. Some MOOC providers, such as Coursera, provide course-completers with a certificate. Five of its courses — offered through Duke University, UC Irvine and the University of Pennsylvania — have been recommended by the American Council of Education for course credit at its 1,800 member institutions.

This isn’t a wholly new trend. MIT’s OpenCourseWare has been around since 2002, offering online lectures and course assignments. UC Berkeley and Yale followed suit in 2006 and 2007, respectively. Academic Earth launched in 2009 as a way to collate lectures from Berkeley, MIT, Yale and other schools.

But MOOCs have taken off in recent years, with Stanford professors launching Coursera and its competitor Udacity, and Harvard and MIT banding together to create edX, all of which make courses at elite universities available to anyone for free. There's also Khan Academy, which started as a single man, Salman Khan, posting math lessons to YouTube and has expanded to include lectures on topics from literature to chemistry. Smaller companies offering specialized instruction, such as Treehouse, have sprouted up, as well.

If MOOCs are as good at imparting knowledge as brick and mortar institutions, that would be a game-changer for higher ed costs. Education would be detached from expensive amenities that drive up costs, and single professors could handle classrooms with hundreds of thousands of students. The cost of providing degrees would plummet, making college vastly more accessible to those who want it.

Some schools are betting on that already. San Jose State University has been experimenting with using edX to supplement its course offerings. They replaced the lectures in an electrical engineering course with edX videos and used class time for hands-on projects. Course pass rates increased by 46 percent.

But that’s not a randomized trial, and edX acknowledges that “the numbers of students were small and classes differed on many factors.” Maybe the students who signed up for the online class were more motivated. Maybe the ones who signed up for in-class lectures were poorer and had less access to computers to watch class videos. It’s hard to say anything definitive about the quality of the respective courses from that experiment.

The best study I’ve seen that puts MOOCs to that test was conducted by William Bowen — a former president of the Mellon Foundation and of Princeton University and one of the discoverers of the Baumol “cost disease” — along with Matthew Chingos of Brookings, Kelly Lack and Thomas Nygren. The authors randomly assigned students to either a fully in-person or “hybrid” statistics course developed by Carnegie Mellon, with 605 students from a variety of different institutions participating. That's a good size for a study of this sort.

The authors found that the students in the hybrid course performed just as well as traditional students on tests of the material covered. “We are persuaded that well-designed interactive systems have the potential to achieve at least equivalent educational outcomes while opening up the possibility of saving significant resources which could then be redeployed more productively,” the study authors conclude.

But that’s just one study, and it used a topic that’s arguably more amenable to being taught online than, say, humanities courses are. Moreover, the “hybrid” design means that in-person teaching was still involved, driving up costs. A Department of Education meta-analysis found that “students in online conditions performed modestly better, on average, than those learning the same material through traditional face-to-face instruction.” But students in blended classes did better still. What’s more, the studies in the meta-analysis focused on professional and mathematical training. One study that evaluated a Spanish education program found that students did worse on some measures if taught online.

Don’t get me wrong: the evidence shows that MOOCs have tremendous potential, especially for math and science and professional training. But it’s too soon to say if MOOCs are well-positioned to replace universities as a whole. To show that, one would need to somehow find out how to measure learning in topics as diverse as history or French language or theoretical physics, and conduct randomized comparisons of learning outcomes in all of them.

What’s more, there may be some benefits of college that are impossible to measure but that MOOCs by definition cannot deliver. Economists like to split the value of college into human capital building, signaling and “consumption value.” Human capital is just knowledge, or skills. If, before you started college, you couldn’t write iPhone applications, and, after going to college, you could, then college built your human capital. Signaling is the value in broadcasting your competence to potential employers and others. Graduating from college may help you get a job not just because it taught you particular skills but also because it makes employers confident that you’re the sort of person they want to hire.

Consumption value is another thing entirely. It’s the value one gets out of college because it’s fun. At a residential college, you typically get to work much less than 40 hours a week, live with your friends, eat food that other people make for you and drink a lot. It’s a pretty sweet deal, and it has nothing to do with imparting knowledge or signaling competence to employers. You pay for that because you like having fun, not because you want to learn stuff.

MOOCs may become as good as traditional universities at building human capital. But they have a long way to go before having the same signaling value. Taking an MIT course on edX doesn’t do as much to get you a job as actually going to MIT would. And, by their nature, they don’t have much consumption value.

Maybe we’re fine with that. A lot of public money goes to subsidize the fun that people in their late teens and early twenties have in college. Perhaps it could be better spent elsewhere, and perhaps MOOCs are the way to cleave the education function of college from its entertainment function.

But there are points where the two functions become hard to separate. Sure, college lets you drink with your friends, but it also lets you talk to them, and debate them, and learn from them. A growing literature on “peer effects” suggests that this social dimension has a real influence on academic outcomes.

Dartmouth’s Bruce Sacerdote exploited the school's random assignment procedure for freshman roommates to test what effects roommates have on each other’s college experiences. He found significant effects, not just in expected areas like choice of fraternity, but in grades: a 1-point increase in a student’s GPA is associated with a 0.11-point increase in his or her roommate’s.

Williams College’s David Zimmerman found the same thing. Williams students’ SAT scores had a real impact on their roommates’ freshman grades. UC-Davis’s Scott Carrell and the U.S. Air Force Academy’s Richard Fullerton and James West found that the grades of one’s fellow freshman squadron members in the Air Force Academy influenced academic performance not just that year, but sophomore, junior and senior years, as well.

Peer effects like those will be harder to gain in MOOCs, as the social dimension is significantly diminished. That may be a cost worth bearing, but it's a significant change in the nature of collegiate learning.