Watch for recommender systems to disrupt and replace many of these rating systems.
As Gladwell pointed out in the Car & Driver example, the "scores" assigned in these rating systems are arbitrary linear combinations of factors chosen by "experts." You the car buyer will personally care more or less about certain factors than they do, but they've already chosen the coefficients for you, so too bad. Same for the college ratings assigned by the USNWR.
I think we're asking the wrong questions with college rankings. Prospective students should be asking, what education will maximize the expected value to me personally in the future? It's not a heterogenous and inexplicable "score" they're looking for, it's the probability that the match between student and college will be a productive one. Given data about individual students, colleges, and historical outcomes, surely a better predictive model is possible.
As Gladwell pointed out in the Car & Driver example, the "scores" assigned in these rating systems are arbitrary linear combinations of factors chosen by "experts." You the car buyer will personally care more or less about certain factors than they do, but they've already chosen the coefficients for you, so too bad. Same for the college ratings assigned by the USNWR.
I think we're asking the wrong questions with college rankings. Prospective students should be asking, what education will maximize the expected value to me personally in the future? It's not a heterogenous and inexplicable "score" they're looking for, it's the probability that the match between student and college will be a productive one. Given data about individual students, colleges, and historical outcomes, surely a better predictive model is possible.