No Steps Forward, Two Steps Back: The Fallacy of Trying to “Eradicate” Adverse Impact?

Abstract

Within the context of the 50th anniversary of the Civil Rights Act of 1964, and specifically as it pertains to the tenets of Title VII, Lindsey, King, Dunleavy, McCausland, and Jones (2013) state: “This focal article raises and addresses critical issues regarding a yet unanswered question: How can organizational researchers and practitioners contribute to the ultimate goal of eradicating employment discrimination” (p. 391). We argue that in the context of employment testing and selection, at least as per the disparate impact theory of discrimination, this question is the wrong one—certainly as framed by Lindsey et al. To the contrary, instead of holding up the "eradication of employment discrimination" as our ultimate goal, perhaps we should continue to focus on the development, implementation, and support of the best (i.e., most job-related and valid) employment practices possible.

Using natural language processing to increase prediction and reduce subgroup differences in personnel selection decisions.

“…We conducted hypothetical adverse impact analyses to illustrate the effect sizes of these reductions in subgroup differences. Although impact reduction should be related to subgroup differences, adverse impact in practice is situationally specific, depending on the unique pool of candidates, skew in the distributions of scores, selection rates, and other factors, especially with small subsamples (Arthur & Woehr, 2013; Arthur et al, 2013). We conducted hypothetical analyses because the organization requested actual adverse impact ratios not be shared.…”

Adverse Impact in der Personalauswahl einer deutschen Behörde: Eine Analyse ethnischer Subgruppendifferenzen

“…Was kann die Wissenschaft einer Behörde praktisch anraten, bei der ein massiver Adverse Impact festgestellt wurde? Einfach durch Forcierung des Personalmarketings den Pool an Bewerbungen aus dem Migrationsmilieu zu erhöhen, wie das Lindsey et al (2013) neben anderen Maßnahmen vorschlagen, ist sicherlich keine zielführende Maßnahme: Die Wahrscheinlichkeit ist vielmehr hoch, dass der Adverse Impact weiter ansteigt (Arthur and Woehr, 2013). Aussichtsreicher könnte es sein, gezielte Werbemaßnahmen für die anvisierten Milieus zu entwerfen und schon im Marketingprozess auf eine bessere Passung der Bewerbungen zu den Arbeits- und Organisationsanforderungen Wert zu legen (Newman and Lyon, 2009).…”

Does the use of alternative predictor methods reduce subgroup differences? It depends on the construct

“…Nevertheless, a review of the literature failed to locate any studies that used a research design in which the predictor construct was held constant and the methods varied in an attempt to demonstrate the professed reduction in subgroup differences, that is, studies that recognized and maintained the construct/method distinction (Arthur & Villado, 2008). 1 Thus, there is very limited, if any, research that shows that the use of different methods (e.g., assessment centers, interviews, SJTs) to measure the same construct (e.g., cognitive ability) results in a reduction of subgroup differences or adverse impact (Arthur & Woehr, 2013; Oh, 2013). So the assertion or received doctrine that the use of alternative methods results in a reduction in subgroup differences and subsequently adverse impact potential is based on research that has failed to take into account the constructs assessed.…”

“…Thus, it is important to clarify that the focus of this paper is on the use of different or alternative methods to measure the same construct; specifically, a suitable alternative method under consideration is one that (a) taps the same or similar job‐relevant characteristics, and (b) produces smaller subgroup differences 2 . Consonant with the extant literature (Arthur et al, 2013; Arthur & Woehr, 2013; Ployhart & Holtz, 2008), we refer to this as the method‐change approach to reducing subgroup differences and adverse impact potential.…”