Statistical discrimination and committees

We develop a statistical discrimination model where groups of workers differ in the observability their productivity signals by evaluation committee. assume that informativeness depends on match between potential worker and interviewer: when both parties have similar backgrounds, signal is likely to be more informative. Under this “homo-accuracy” bias, group most represented committee generates accurate signals, and, consequently, has greater incentive invest human capital. This trap. If one initially poorly evaluated (less into committee), translates lower investment capital individuals such group, which leads representation future, generating persistent process. explore dynamic process show quotas may effective deal with introducing “temporary” quota allows reach steady state equilibrium higher welfare size population. instead discriminated underrepresented workers’ population (for example because his race), restoring efficiency requires implement “permanent” system quotas.