While the existence of an unexplained gap in wages is certainly consistent with discrimination, this measure suffers from two main drawbacks. First, most observational data inevitably lack information on some of the determinants of a worker's productivity, which are nevertheless observed and valued by employers. Thus the unexplained gap is contaminated by unobserved differences in productivity, and whether it provides an upward or downward bias of the true extent of discrimination depends on the sign of differences in such unobservables.
