An ‘Amazon’ of Online College?

Once upon a time, stores sold things. Some stores specialized in selling particular things, while “department” stores sold a wider range of things. While there were some scale economies in branding and distribution, they were mild enough to allow many different department stores. With the internet, however, much bigger stores have been favored. Not only can huge online stores hold more variety, with scale economies in storage and distribution a single store can dominate that industry. Hence, Amazon.

While the internet favors a few huge platforms for some types of products and services, the strength of this effect varies with the kind of product or service. For example, there seems to be room for many movie streaming services, as there seem to be fewer scale or scope economies there. Yes, one can better price discriminate by selling many movies rather than just one, but that still leaves room for many services each of which has many movies. Though perhaps the current variety of streaming services won’t last long.

What about college? In the past, students attended class in person, and so each college arranged to have many classes all close enough that one could live nearby and travel to all of its classes. So travel time between classes set a maximum feasible size for a college. But now there are (for many non-lab-or-hands-on topics) online classes which one can attend from anywhere in the world. In a future of online college classes (and tests), will we still have the same size colleges, or will much larger platforms take over?

Clearly there is a big potential for much larger individual classes. Instead of a thousand profs teaching the same class all over the world to thirty students each, maybe only ten profs will teach to three thousand students each. At least when individual grading and talk isn’t the main cost. And if students can choose from classes made all over the world, a far wider variety of classes can be made available to each student, classes on more topics, at more levels, and with more different teaching/learning styles.

Yes, the most elite colleges would probably be the last to contribute their courses to large online catalogs of courses. They’d say, “if you want the very best college experience, you should come here and limit yourself to our classes.” But that pitch wouldn’t work so well coming from mid-rank colleges.

Still I wonder: will the thirty or so classes on a future student college transcript be mostly from teachers who all produce their classes near each other at the same “college”, or will student transcripts instead contain classes from twenty or more different sources? And if the latter, how many distributors or platforms will there be? That is, will there be a single “Amazon” from which most all students select their classes, or will there be many different strongly competing distributors of classes, more like movie streaming services today. Another way to ask this question is: what are the scale and scope economies that might favor a few big college class distributors, instead of the thousands of colleges we have today?

As mentioned above, price discrimination offers one scope economy, but this runs out near the scale of a typical college, so won’t push for much larger units. A similar logic applies to other scale and scope economies that mostly run out near the scale of typical colleges today. For example, an online college class platform may want to select and evaluate the classes that if offers, to judge which classes could serve as prerequisites for which other classes, and maybe also to select and evaluate students for their suitability for various classes. And yes, these tasks look easier for larger platforms. But such effects still seem to allow a lot of room for many competing platforms.

However, here is a scale and scope effect that may push more toward a more Amazon-like scenario: giving students grades that are comparable over wide scopes. Today employers mostly look at a college graduate’s school and major, and sometimes also at their GPA. This works because schools have known reputations, and majors are pretty similar across many colleges.

This is in stark contrast to most jobs that students might take instead of going to college; it is much harder to know how to compare letters of recommendation based on typical job performance. Even US military veterans face this problem; employers find it hard to know what school/major/GPA record is comparable to 2 years as a “helicopter repairer”. Superior college comparability is a big reason many go to college instead of starting work (or the military) right after high school.

Imagine a college class platform that gives you a transcript showing what classes you took, and what grades you got in each class, but that doesn’t do much to help employers know how to compare the grades obtained from different sources. That wouldn’t be so valuable. In contrast, a college platform is much more valuable to future employers, and thus to students, if it can rank and categorize student performance in comparable and meaningful ways.

One simple way to do this is to sometimes randomize which classes students take. That is, flatter, pay, or cajole many students into letting the platform sometimes pick which particular class they take, out of a set of similar classes offered on the platform. With enough students taking enough classes that give enough feedback on student performance, standard statistical models could estimate individual student abilities and specializations. Which helps not only future employers, but also the providers of classes when deciding which students to admit into their classes. It also helps to estimate student satisfaction in particular classes as expressed by student evaluation of classes.

Yes, if all student class performance info were made available to all platforms, many of them could produce similar statistical estimates. But the largest platforms may use privacy excuses to successfully resist efforts to force them to share their customer info, as have social media giants today. Yes, platforms with less data might claim that they had found clever uses of machine learning etc. that give similar quality evaluations of students. But it isn’t clear why students and employers should believe such claims.

Scale economies in using customer data to make student performance comparable across a wide scope of classes may push toward a single huge Amazon-like catalog of online college classes.

From a conversation with Phil Magness.

Added 9a: College admissions and grading has recently become a political battleground. While today such battles are limited by the fact that colleges must compete with each other, such limits might be less when one of a few big orgs dominated the online college catalog market.