This weary planet might not need yet another set of college rankings, but new models keep popping up. The latest is a rating of colleges based on their “desirability,” as determined by the choices applicants make.
In a new paper published by The Quarterly Journal of Economics, four researchers propose a method of ranking colleges according to students’ “revealed preferences”—the institutions they choose to attend over others that have accepted them. Using survey data from a national sample of high-achieving students, the researchers determined the winners and losers of each applicant’s “matriculation tournament.” They then used those outcomes to rank about 100 selective colleges. (Harvard University topped the list, but you already knew that; the University of Notre Dame nearly cracked the top 10.)
This model enabled the researchers to approximate the odds that an applicant would choose one college over another. For instance, there was a 59-percent chance that a student considering only Harvard and the California Institute of Technology would choose Harvard. If the student were choosing between only Harvard and Wellesley College—10 spots below Harvard on the revealed-preferences list—there was a 93-percent chance that she would end up heading to Cambridge, Mass.
The paper, “A Revealed Preference Ranking of U.S. Colleges and Universities,” proposes an alternative to the two most prominent measures of desirability—admission rates (the percentage of applicants accepted) and yield (the percentage of accepted students who enroll). Although valued by college guides, those metrics are subject to manipulation. For instance, rejecting applicants whom a college deems overqualified and, therefore, unlikely to enroll can make that college look more desirable.
Yet the revealed-preferences model, the authors suggest, does not reward such strategies. “Our method removes the incentive to reject overqualified applicants,” they write. “This is because although overqualified applicants are unlikely to enroll, a college gains tremendously in our ranking when they do enroll and it loses only trivially in our ranking if the overqualified applicants enroll instead at a much higher ranked institution.”
The authors also challenge the assumption that an admission rate is an indicator of desirability. Half of the top 20 colleges in the revealed-preferences list, they found, would fall outside the top 20 if one ranked them only according to their admission rates (the lower the rate, the better, conventional wisdom holds). Notre Dame, for instance, placed 13th on the desirability list, but its admission rate was only the 58th lowest. The University of Virginia placed 20th on the desirability list, but it had only the 76th lowest admission rate.
How much might other variables—such as varying financial-aid packages—explain the desirability ratings? Very little, the authors found. The revealed-preferences ranking “is not sensitive to whether we control for college characteristics that vary among students, such as net cost or distance from home,” they write. “Students self-selecting into applications based on their chance of admission has no effect on the ranking, and students strategically selecting into Early Decision causes the ranking to misrepresent colleges’ desirability only very slightly.”
The desirability model, the paper concludes, ”eliminates incentives for colleges to adopt strategic and inefficient admissions policies to improve their rankings.” At the very least, the findings suggest that many colleges are wasting their time and money trying to look prettier than they really are, at least in the eyes of applicants.
The paper’s authors are Christopher N. Avery, a professor of public policy and management at Harvard’s Kennedy School of Government; Mark E. Glickman, a professor of public-health policy and management at Boston University’s School of Public Health; Caroline M. Hoxby, a professor of economics at Stanford University; and Andrew Metrick, a professor of finance and management at Yale University’s School of Management.
The paper is available to the journal’s subscribers. Read the free abstract here.