Discover more from Overcoming Bias
Last week I posted on how decision markets might help students pick a college. Now consider colleges picking students. The obvious approach is to bet on student success given starting at a certain college in a certain season (perhaps also at a certain school of that college).
A college could ask applicants to submit an essay, which does not identify them individually, and agree to let the college post this and other anonymized details for all bettors to see. And the college could commit to later publishing success stats on these students.
Compared to students picking colleges, here success criteria are easier. Whether the student graduated at all, and their GPA and eventual major if they do graduate, encompass a lot of what colleges now look at. If one could wait longer, colleges also care about donations students might make as alumni. And if we had worked out student success measures for college choice futures, college admissions could also look at these.
The main problem with this proposal, other than the many years needed to evaluate a trial, is that it seems hard to start small – one would have to get an entire school of some college to sign on to a trial, and they might reasonably fear how this would effect their reputation.
College Admission Futures
Things like US news and world report do a fairly good job of conveying the purely academic information about a school despite how much people dislike them because it hurts their pride. The reason they do so is not that they use a particularly succesful methodology but because people believe them. If you got all the students who attend harvard to all switch to some decent similarly sized state school you would probably find that in 10-15 years that school would be the best place to go in the country. Not only do professors like working with smart students but the people who have the biggest impact on students at college are other students, even academically.
---
The big problem with college admissions is that the things that students need to know other than academic performance are deliberately obscured. Students need to know whether they are going to be happy, have fun and enjoy themselves (if they don't they probably won't perform either) but all colleges try to cover up the partying and student socialization aspects of campus life because it looks bad to parents and older folks.
I mean I went to caltech and it's hardly a hard partying drinking school but you can be sure that the official campus information didn't mention the fact that the students picking their own RAs meant no stupid attempts to get the people smoking a bit of pot or doing some drinking in trouble. Nor would they make a point in their literature of the fact that (despite a significant christian minority) it was probably one of the more atheist student bodies around. And caltech (because it's student body is so extreme) is a place that is least able to disguise the nature of the student body.
It's not that it's about drinking and drug use. It's about the fact that every college tries to portray it's student body as the perfect adult image of young studious fun meaning that students can't trust any information about the social scene from the college and short visitation weekends aren't long enough to convey it.
"but what are college admissions boards maximizing?"
This is the real issue. Admissions officers with access to essays and personal interviews and test scores and GPA have, in studies, done a worse job of predicting student GPA's than those with access to just test scores and GPA's. If the outcome is GPA or graduation rate, intelligent bettors would rely on these statistics and, thus, probably provide no new information (or very little new information).
The exact problem is that colleges are trying to maximize intangibles, such as how much a person will contribute to the "experience" of going to that school. Without a way to quantify intangibles, you cannot have a futures market in them, and thus no futures market can provide you with the information you want.
Unless of course there is some way to structure a futures market around unquantifiable characterstics; I can't imagine how, but that hardly means it cannot be done.