Examines the credibility of inferences based on cross-group comparisons of regression coefficients using Black-White differences in the educational attainment process as an example. Shows that reanalyses of data from previous studies fail to produce consistent results even when major differences in samples, models, and methods are held constant. (Author/GC)