The only other group to have done a revealed preference study – a set of professors from Yale, Harvard, and Stanford – also found Notre Dame rising surprisingly high above where they expected to find it in the rankings. One explanation offered for this is that Notre Dame recruits a particularly dedicated applicant pool that is willing to choose them over what are otherwise more notable institutions. Another explanation could well be that Notre Dame is undervalued as an institution.

Also, consider the different competition that Cornell and Notre Dame face: Cornell competes with the rest of the Ivies all day long (and loses about 80% of the time according to our data).

My theory on this situation is as follows (and some day I’ll look deeply into the data to see if this is actually the case): Presumably, many people want to get into an Ivy; they’re the same people who are aware the Cornell lacks the cachet of Yale or Harvard, so they’re less likely to attend Cornell if they get into a school with a “better name”. Notre Dame applicants are much less likely to be in the “I want to get into any Ivy” group of people. Notre Dame is more likely to be the big name school for them.

Are these raings by student preferences, as opposed to academic standards? That could explan it.

The Texas issue is interesting. To a large extent, with more data, I doubt this would be a problem. However, there are theoretical reasons that UT Dallas could rank above UT Austin: UT Dallas has a huge grant program for National Merit Scholars. Additionally, perhaps UT Austin competes on a national stage, whereas the other schools compete on a regional stage. In this case, Austin would lose out frequently to the heavy hitters (HYP, etc). However, UT Dallas might not, since it would not be competing for the same students. This would be reflected in the rankings.

The SUNY schools seem to be misranked; I am not familiar enough with them to say why. I would look to the same 3 culprits: (a) not enough data; (b) competition on different stages (regional v national); and (c) lures (perhaps Brockport uses scholarships).

You mention that these rankings will be swayed by factors such as proximity to home. On average, these factors will cancel out each other. Some Texans might choose UT Austin over UC Berkeley due to proximity; some Californians might choose UC Berkeley over UT Austin for the same reason; etc. These will lead to some slight reduction in Elo points for the higher ranked school, but otherwise should have a minimal effect.

Finally, you talk about the possibility of a better school losing to a worse one. Can you explain how one should decide what the “better school” is? In my view, the better school for Student A might be the one with the best overall academics, regardless of cost or location. The best for Student B might be one with a great education department in an underserved community in the Midwest. The best school for you may not be the best for me. Our ranking system incorporates all of that data in aggregate. When appropriately paired to the source data, the tool can also peer into the preferences of particular subsets of students to get more personalized rankings. (That tool is currently unpublished.)

That’s a good point, and brings up a few follow-up questions. What constitutes the most favored college? Should ranking all of the Ivies highly be the sine qua non of a valid ranking system? What constitutes gaming in a preference-based college ranking system? I will write another post soon that addresses these.

In the meantime, consider this: the lowest ranked Ivy is 14th in our 2009 results for National Universities. If other colleges are bribing students with scholarships to steal them away from the Ivies, it doesn’t appear to be working all that well.