Disparate Impact
(April 30, 2012; rev. June 24, 2012)
Disparate impact is the concept originally developed in employment law whereby a practice that disproportionately disadvantages a protected group can be found unlawful unless justified by a sound business purpose. The concept has been given much attention lately in light of the decision of a number of federal agencies to apply the concept in the context of the enforcement of fair lending laws. See my June 5, 2012 American Banker article “’Disparate Impact’: Regulators Need a Lesson in Statistics.” There have been a number of problems with standard interpretations of disparate impact theory over the years. The problem to which various pages of this site are most pertinent, and one I have highlighted with respect to the enforcement of fair lending laws, involves the failure to recognize the pattern by which the rarer an outcome the greater tends to be the relative difference in experiencing it and the smaller tends to be the relative difference in avoiding it, particularly in the context of an inquiry into the existence of a less discriminatory alternative to a challenged practice. Lenders have long been encouraged to relax their lending criteria because minorities were less able to meet standard criteria. No one involved in the matter understood that relaxing lending criteria, while tending to reduce relative differences in rates of securing mortgages, tends to increase relative differences in mortgage rejection rates. Thus, as regulators and others enforcing fair lending laws continued to appraise the degree of disparity in terms of relative differences in mortgage rejection rates, the lenders most responsive to federal encouragements to relax lending criteria became the most likely targets for litigation. See the Lending Disparities page of this site and the articles cited there.
That reducing adverse outcomes tends to increase relative differences in experiencing them raises a host of difficult issues in any disparate impact setting where the impact is measured in terms of relative differences in adverse outcomes. For all of the alternatives to measures that cause disparities in adverse outcome rates that would typically be deemed less discriminatory (commonly, measures that reduce adverse outcome rates) will tend to increase relative differences in adverse outcome rates. See the Discipline Disparities page of this site. Section A of that page addresses the false perception that stringent public school discipline policies lead to large racial differences in discipline rates, which is the exact opposite of reality. But Sections B and C address the same issue in employment and criminal justice settings and Section D addresses misperceptions about the racial impact of National Collegiate Athletic Association academic eligibility requirements. See the Less Discriminatory Alternative – Substantive sub-page.
But other aspects of disparate impact theory have been poorly analyzed as well. One of the most incongruous aspects of any Supreme Court ruling of which I am aware involves the holding in Wards Cove Packing Co. v. Atonio, 490 U.S. 642 (1989), that liability would attach when, following the plaintiffs demonstration of a less discriminatory alternative at trial, the employer failed to adopt it. That is, the employer’s liability would turn on what it did after trial. Yet Congress proceeded to codify this interpretation in the Civil Rights Act of 1991. See my "Slip-Up in the Civil Rights Bill," Legal Times (Dec. 11, 1991), and “Wards Cove Packing Co. v. Atonio, 490 U.S. 642 (1989)” in Affirmative Action, An Encyclopedia (James A. Beckman ed.) Greenwood Press, 2004, 930-32. See the Less Discriminatory Alternative – Procedural sub-page.
One important issue that has never been adequately treated involves the challenge to subjective decision making on a disparate impact theory. One aspect of this matter involves treating subjectivity as a common practice across many units of an organization in order to satisfy the typicality and commonality requirements for treatment as a class action. That has been often treated and I do not presently think I have anything useful to add to those treatments. But the other involves whether and how the subjectivity causes discrimination. The reality of such claim is that to allow decision-makers to make subjective judgments leads to disparate treatment of individuals – that is, that the persons engaged in the subjective decision-making consciously or unconsciously treat persons differently because of race, age, gender, etc. I have not studied whether there exists sound research indicating subjective decision-making leads to larger (properly measured) differences disparities in outcome rates than objective decision-making. One must keep in mind that there exists virtually no sound research comparing the size of disparities in outcome rates because all research that has attempted do so has failed to consider the prevalence of the outcomes in settings being compared. See, e.g., the Disparities – High Income sub-page of the Lending Disparities page. Research that involves continuous measures may be more sound.
Such issue will be explored taking into account considerations set out in the Employment Tests sub-page of the Scanlan’s Rule as well as the evidence from simulations in EEOC v. Sears (subject of The Sears Case page of this site) indicating that error in the selection process tends to work to the advantage of the group with lower average qualifications.
The above and other issues may eventually be addressed at some length on this page or its sub-pages. Currently the Four-Fifths Rule sub-page addresses reasons why, contrary to the belief or many, the four-fifths rule of the Uniform Guidelines on Employee Selection Procedures does not provide a plausible measure of effect.