Using matched employer-employee data we examine firm-specific gender and ethnicity pay differentials in Britain. We estimate an econometric earnings model using the partially-observed pay variable provided in the data and test the normality assumption that underlies the usual interval regression...