A Statistical Analysis of Adverse Impact of Employer Decisions

Abstract

Federal law prohibits discrimination in employment decisions against persons 40 years old and older. This article uses data from an actual case to illustrate several methods of showing adverse impact, a legal doctrine under which only the effects of the employer's acts are at issue and not the motives with which they were done. The strengths and weaknesses of the Fisher exact~est of significance, a.Bayesian analysis, and a method of paired observations inspired by the Mann-Whitney-Wilcoxon statl~tlc are assessed'. One Important conclusion from this analysis is that it is useful to have several different kinds of analysis beanng on the same Issue. To the extent that the analyses agree, credibility is added to each.

Optimal Experimental Design for Another's Analysis

“…They argued that inferences should convince those evaluating medical trials, despite the prior opinions of those performing the trials. A further example is that of statistical inference in legal contexts, where the prosecution gathers information to convince an impartial or objective court of the liability of the defendant (see, for example, Kadane 1990). …”

Statistical Evidence in Discrimination Cases

“…By analogy with race and sex discrimination claims, the most common method compared the rates for employees over 40 with those of similar employees under 40 years old. If occupation or divisions of the ®rm were important, strati®ed analyses, such as the Cochran±Mantel±Haenszel test, were used (Kadane, 1990). When the plainti was over 50 years old courts sometimes accepted comparisons of the rates for employees over 50 with those under 50 or under 40 years old.…”

Hierarchical Models for Employment Decisions

“…We use age discrimination in termination decisions to illustrate our ideas because age discrimination cases dominate our experience. Age discrimination over a short period of time, for example, when an employer makes a large reduction in the workforce over a matter of a few days or weeks as the result of a single policy decision, is comparatively easy to analyze (Kadane 1990) mainly because it is reasonable to assume that the odds ratio is constant; however, age discrimination over an extended time period is more dif cult to model, both because the same individual can over time move from the unprotected to the protected class and because, unlike gender or race, age is a continuous characteristic and consequently the hazard rate may vary within the protected class.…”