Note added October 10, 2013: A lengthy treatment of the measurement of discrimination in hiring and other employment contexts may be found in my paper “The Mismeasure of Discrimination,” which was presented in a faculty workshop at the University of Kansas School of Law on September 20, 2013.
The articles discussed several paragraphs below involve some complex issues about how to measure and appraise selection disparities in a tester study. A more important matter concerning the measuring of hiring discrimination involves the fact that the relative difference in selection rates is a flawed measures of such discrimination, just as the relative difference is a flawed measure of association with regard to all other matters in the law and social and medical sciences, as discussed in the Measuring Health Disparities(MHD)and Scanlan’s Rule(SR) pages of this site.
This point is best illustrated with respect to the measuring of hiring discrimination on the Relative Versus Absolute sub-page of MHD, which uses the employment setting to illustrate that there can be only reality as to the comparative size of a differences in the between the status of two groups from setting to setting. The sub-page, which is directed at refuting a notion in health disparities literature that various measures of differences between outcome rates may each provide a valid appraisal of the size of a difference between pairs of rates even when they yield opposite conclusions as to the comparative size of differences in different settings, uses as an example a situation where the question to be answered involves which of several employers is the most biased. The page demonstrates that the notion is unfounded. There exists only one reality as to the comparative size of the difference in the circumstances of demographic groups reflected by two or more pairs of rates of experiencing an outcome. That is so with regard any type of outcome and regardless of the nature of the forces driving the difference in outcome rates. But it is most obviously so when the force may be bias against a demographic group.
The Representational Disparities sub-page of SR explains that, since one must know the actual selection rates to effectively appraise the size of a selection disparity, it is not possible to appraise the size of selection disparity when all that is known is the proportion a group comprises of the persons eligible to be selected and the proportion it comprises of persons selected. See also the Case Study and the Case Study Answers sub-pages of SR.
The two articles below relate to a tester study that sought to quantify the extent of hiring discrimination based that apparent favored treatment of one tester applicant over another without consideration of the extent to which the employer gave serious attention to either applicant. That is, the denominator in the fractions used to quantify disparate treatment included, for example, all tests rather than the tests where either of the tester applicants received an offer. There are other problems with the study as well.