Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
In this paper we extend the standard model of statistical discrimination to a multidimensional framework where the accuracy of evaluators depends on how knowledgeable they are in each dimension. The model yields two main implications. First, candidates who excel in the same dimensions as the evaluator tend to be preferred. Second, if two equally productive groups of workers differ in their distribution of ability across dimensions group discrimination will arise unless (i) evaluators are well informed about the extent of these differences and (ii) evaluators can take candidatesʼ group belonging into account in their assessments.