Previous studies on gender wage discrimination have relied on OLS when estimating the wage equations. However, there exists a number of recent studies, devoted to estimating the return to education, that have shown that OLS may produce biased estimates for a number of reasons. Consequently, if we neglect this potential bias in OLS when estimating the gender wage gap, there is every reason to believe that the estimated wage gaps will also be biased. We first estimate wage equations using instrumental variables procedures and panel data, and then investigate the distributions of wage discrimination experienced. We find overwhelming evidence that OLS seriously overestimate the unexplained gender wage gap.