This paper compares the results from two procedures for estimating wage dif
ferentials. Using Brazilian data, the standard procedure of Krueger and Sum
mers (Econometrica, 56, 259-193, 1988) is shown to provide an overestimatio
n of standard errors and a biased estimation of the measure of wage dispers
ion, in comparison to the methodology proposed by Haisken-DeNew and Schmidt
.