New ranking system based on choice

In their proposal, the economists sidestep the tricky question of what makes a good college. Instead, they assume top high school students know best, and they simply report their choices. Of the students admitted to, say, both Brown University and the University of Pennsylvania, how many choose each place? It is the same principle in the Zagat restaurant guides: Don't try to grade the food, just say whether a lot of people like it or not.

The authors -- Caroline Hoxby and Christopher Avery of Harvard, Andrew Metrick of the University of Pennsylvania, and Mark Glickman of Boston University -- have been working on their model for years. Their most detailed results yet were published recently by the National Bureau of Economic Research.

Here's how it works: Imagine two students, one choosing between Stanford and Harvard, another among Stanford, Berkeley, and Pomona. The statistical model views each one of those students as a ''tournament" between the colleges involved; if Stanford ''wins" either student, its ranking rises in relation to the schools it beat.

Colleges often compete against the same schools over and over and may never compete against others. But with enough data, Stanford's place in relation to all schools begins to emerge and the rankings take shape.

The model, which resembles the one used to rank professional chess players, adjusts to balance out influencing factors such as a big financial aid offer from one school.

The authors offered a preliminary demonstration by tracking the college choices of 3,240 high-performing students from 396 high schools nationwide. They say that it works well for top schools but that more data are needed to improve their confidence in the rankings lower down.