Take a group of 20 college students taking a four credit introductory physics class. Let’s assume they pay the average tuition at a non-profit college of $34,740 in 2017, and each student takes 16 credits per semester. This would mean that each student is paying $4,342.5 for this class, and the total cost of the class is $86,850¹.

The Experiment

Divide the class randomly into 5 groups of 4 students. No lectures. Only 4 two-hour exams. Every group has the possibility to take one exam every week, when they feel they are ready. The catch, if anyone in a group fails the exam, the whole group has to retake it. Every member of the first group to complete all exams, receives $500, the members of the second group, $250, the third, $150, and the members of the last two groups, each receive $50 once all exams are completed. Furthermore, each member of the group with the best group grade point average receives $300, the second best $200 and the third best $100.

At the beginning of the semester, all students are pointed toward all the introductory physics material that MIT has put online together with video of all their lectures, Khan Academy’s physics videos as well as one of the many free online physics textbooks. If that is not enough, each group also has access to 20 hours with a professional tutor. That shouldn’t be necessary, as the strongest students in any group have a great incentive to help the others out. But just in case.


The cost of all this? All the prize money amounts to $6,400. The cost of proctoring a two-hour exam every week for 13 weeks? Given a proctor paid $50 an hour as well as an examination space fitting up to 20 people rented at $50 an hour, this would amount to a total of $2,600 for the entire semester. For a grader paid $25 an hour (average TA salary: $11/hour), taking 30 minutes to grade one exam, this would amount to a total of $1,000 to grade all exams. Finally, a total of 100 hours of tutoring (20h/group) from a professional tutor, at a rate of $50 an hour, would cost a total of $5,000. I proceed on the assumption that creating the exams would entail no extra costs. All one needs is a large enough database of problems, and creating an exam becomes as simple as picking a few problems out at random.

In summary:

  • Prize money: $6,400
  • Exam proctoring: $2,600
  • Grading: $1,000
  • Extra tutoring: $5,000

    TOTAL: $15,000

Doing this would, therefore, cost less than 20% of the average cost of an introductory physics class. It should be noted that, if the students are the ones paying for this, the prize money is actually only redistributed on a meritocratic basis. The cost of the class without the prize money is $8,600 or less than 10% of the average cost of this college class. This amounts to $430 per student. It should also be noted that many of the estimates I make here do seem high, and if done efficiently, such a system could be even cheaper.


My hypothesis is that with this setup, students would learn just as well, and probably better, than in the traditional setting. There are a couple of reasons for this:

  1. I believe that free online educational material is almost always superior in teaching a subject matter compared to traditional, live, in-person lecture. Even just a video of a lecture is more useful than the lecture itself, since a video can be replayed at will at any time, played faster, paused… Furthermore, it is online that “The state of the art in explanation of a concept” is advanced, not in university, as online, there is competition between content creators for the best explanation.
  2. There is a financial incentive for working hard and helping one’s group mates.
  3. Inter-group competition. This can be a great motivator. As Jonathan Haidt argues in “The Righteous Mind”, human psychology has evolved for inter-group competition.
  4. Both the tutoring between students and the tutoring from an outside professional tutor can greatly help learning. A 1984 study found that students taught one-on-one performed better than 98% of their peers taught in a regular classroom.
  5. Peer pressure. Those who are most motivated to win will pressure those who are less motivated to work harder. One important assumption here is that the main value college provides with respect to learning is just the incentive it creates for learning. And the only incentive college provides is grades and the diploma. If something is “on the test”, students figure out a way to learn it, even if the professor was completely unintelligible during lecture.


If my hypothesis is correct, it may have wide-ranging implications. The most important of which is that the college affordability crisis could be solved by simply embracing modern technology and by changing the incentive structures.

It would also be further evidence that the main value college provides is not learning, and would lend credence to Bryan Caplan’s argument in “The Case Against Education” that it is mostly about signaling.


There are a lot of variables in this experiment that can be played around with. The size and the distribution of the prizes, the size of each group, the number of exams… My proposed distribution of prizes puts a greater emphasis on speed than performance. That is because I believe that most students, if properly incentivized, could learn much faster than they are doing today. And theoretically, if students finish one class before the end of one semester, they could start a new one right away.

Students do not only spend a lot of money on college. They also spend a lot of time. Making college cheaper should be just as much about lowering the price tag as about reducing the opportunity cost of spending months and years not working. This being said, my proposed prize distribution is probably not the most optimal.

Another variable that would be interesting to look at is the composition of the groups. It is important for there to be not too much disparity between groups, so that they remain competitive, and all groups have a shot at winning. However, it is important for there to be some disparity, but not too much, within a group, so that the stronger students have the opportunity to help those struggling.

There also needs to be a process by which a group can expel a student if they are really not pulling their weight. This, however, would need to carry a penalty, like making it impossible for the group to win anything just for finishing all exams, or taking a percentage of any other of their winning. This should only be a last resort though.

Finally, this model can be adapted to any class. There are free online resources available for almost any subject matter. Grading for some classes might be a little more expensive than for introductory physics, particularly for classes involving lots of writing. But, if this experiment is successful, then there is little doubt that universities are astonishingly inefficient at teaching.

I challenge any university to try this out. I doubt students would be opposed, so why not?

¹ 20 students, $34,740 for one year, 32 credits per year and a four credit class: 20*34,740*4/32=$86,850

This post was originally published on Medium in 2018.