In the coming weeks, throngs of new freshmen will descend on campuses across the country. For many traditional students, going off to college means living with a roommate for the first time. That got us wondering: How are colleges assigning roommates these days, and does it really matter?
We caught up with Bruce Sacerdote, a professor of economics at Dartmouth College who has studied the effects roommates have on each other, to find out. The following conversation has been edited and condensed.
Q. How do colleges usually pair roommates, and how has that changed over time?
A. The traditional thing was to ask students a set of questions about their interests, and if they had a request for dorms or particular roommates, and use those to put people together. And then, about 20 years ago, they hit on the idea that maybe we should just randomize, maybe that would be more fair and more fun. And so a number of places started doing that.
I’m not claiming that Dartmouth was an innovator, but it certainly began doing that because it wanted to have a more diverse mix on hallways and within rooms. And the military academies that I’ve studied a bunch and my coauthors have studied a ton — they’ve done it perfectly randomly for a long time.
Q. Has there been some shift in how roommates are assigned more recently?
A. Any number of institutions have called me and talked about the fact that they’re going toward randomization rather than away, because they see the benefits. It actually stimulates cross-geographic, cross-race, cross-cultural interaction.
Q. I’ve also heard of new systems that are supposed to let freshmen pick their own roommates. That’s still uncommon?
A. That’s a relatively new trend. That could be going on at the same time that the trend toward randomization is, so some colleges may be allowing you to pick your own, and then some are going more toward randomization. I don’t know which one’s going to win. You’ll probably end up with half doing pure randomization and half doing pure picking based on a Match.com algorithm.
Q. Tell me more about your research on college roommates.
A. I started out just simply noting that they were randomly assigned, and then seeing how much they would influence each other with regard to their GPA and their study habits. Then I branched out into things like drinking and joining fraternities and sticking with their varsity athletic team. And now it’s been broadened not only by me, but by other folks, to think about how does it influence your attitudes. If you come from a very white place and you room with an African-American student, how did that influence your attitudes toward affirmative action, toward people of different races, and actually your willingness to room with those folks in the future.
And interestingly there’s where you see the biggest development, on drinking behavior and socializing behavior and, to some degree, study behavior. You see very modest effects on grade-point average and such. There’s also a lot of influence on job choice, so if you happen to get someone who’s interested in finance, it makes you significantly more likely to pursue that both in internships and as a career.
Q. Based on those findings, what method of roommate selection do you think is best for students?
A. My coauthors, Scott Carrell and Jim West, and I decided that pure randomization is ideal because you get that unexpected mixing that you would not otherwise have.
At one point, Scott and Jim convinced the Air Force Academy — since there is this influence of roommates — we could reduce the number of cadets who drop out if we explicitly assigned students to squadrons in a way that would maximize their GPA and minimize their chance of dropping out.
It failed spectacularly because the students resisted the groups we tried to form, and we ended up with a clump of very super academically prepared students only talking to each other, and the least academically prepared students in the same squadron only talking to each other. We actually had a negative impact on overall GPA.
Because predicting the outcomes from true social engineering is pretty difficult, and because randomization has these nice properties of being fair and creating a ton of mixing, there’s no reason that we shouldn’t do that.
Q. It also seems as if randomly assigned roommates are a boon for researchers.
A. That’s super helpful, too. Most of my work — I do a lot of other things besides roommates, of course — has some element of randomization, so you can try to make some causal statements about what’s going on, as opposed to just selection.
Q. What was your freshman-roommate experience like?
A. I just saw my roommate last month at my 25th reunion. A guy named Henry Todd. He had a tremendous impact on me, encouraging me to travel, try different things, branch out, consider joining a Greek house. He definitely made me more social. And I know a whole group of friends that I would not know at all were it not for my connection to Henry.
Beckie Supiano writes about college affordability, the job market for new graduates, and professional schools, among other things. Follow her on Twitter @becksup, or drop her a line at beckie.supiano@chronicle.com.
