So today I was talking to someone who works in admissions for an economics PhD program. The way they do admissions is that they go through every folder and give it a numerical score that’s as objective as possible, based on grades, GRE scores, quality of recommendations, undergrad research, etc. Then they go through all the folders, looking at the scores, and decide who to admit.
In the second step they always practice explicit affirmative action in favor of women; the score required to get in as a woman is significantly lower than the score required to get in as a man. And yet, consistently their female students do better on average than their male students.
I thought that was really interesting.
I don’t suppose anyone took those scores and tried to figure out a system that would better predict how well people do? It seems likely that some of those criteria are anti-correlated with success.
Yeah, economics PhD admissions seems like kind of a hard problem. Of course one thing you want is people who will do well in classes, and that’s easy–just get smart people from top schools with good grades and stuff.
But the main thing you want is people who will be good at research, and that’s not as correlated with all the obvious things, and they haven’t found a really good predictor. It seems like success in research involves a lot of intangible factors–broad range of knowledge and interests? curiosity about the world around you? creativity? ability to work in a self-directed way?–that are hard to measure from reading a folder about a college senior.
The one thing they do know is that, conditional on having 99th percentile GRE scores or whatever, these intangible factors seem to be positively correlated with being female.