Statistical Proof of Discrimination: Beyond "Damned Lies"

Evidence that an employer's work force contains fewer minorities or women than would be expected if selection were random with respect to race and sex has been taken as powerful—and often sufficient—evidence of systematic intentional discrimination. In relying on this kind of statistical evidence, courts have made two fundamental errors. The first error is assuming that statistical analysis can reveal the probability that observed work-force disparities were produced by chance. This error leads courts to exclude chance as a cause when such a conclusion is unwarranted.

The second error is assuming that, except for random deviations, the work force of a nondiscriminating employer would mirror the racial and sexual composition of the relevant labor force. This assumption has led courts inappropriately to shift the burden of proof to employers in pattern-or-practice cases once a statistical disparity is shown. Recognition of these two errors suggests that the role of statistical evidence in discrimination cases should be greatly reduced.