Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
This study examines the theoretical and practical implications of ranking teachers with a one-dimensional value-added metric when teacher effectiveness varies across subjects or student types. We create a theoretical framework which suggests specific tests of the standard teacher input homogeneity assumption. Using North Carolina data we show that value-added fails to empirically meet these tests and document that this leads to a large number of teacher misrankings. Thus, critics of potential value-added teacher personnel policies are correct that such policies will terminate many of the wrong teachers. However, we derive the conditions under which such policies will improve student test scores and find that they will almost certainly be met. We then demonstrate that value-added information can also be used to improve student test scores by matching teachers to students or subjects according to their comparative advantage. These matching gains likely exceed those of a feasible, value-added based firing policy.