Most studies that have examined the effect of race on earnings in the
United States have used aggregated data. This often obscures as much a
s it reveals. This article computes white-black earnings differentials
for a number of relatively narrow groups, and finds that the earnings
gap varies considerably across them. By using recent data from 1987,
the study also updates earlier estimates of racial differentials. The
use of Cotton and Neumark's modified version of the widely used Oaxaca
technique, and Heckman's corrections for selectivity bias strengthen
the methodological underpinnings of the model, and constitute improvem
ent over previous work.