Adverse (disparate) impact
Adverse impact (also referred to as disparate impact) pertains to situations where an organization treats all applicants equally (e.g., administers everyone the same tests in the same way) throughout the hiring process, but the process yields differential outcomes according to subgroup (e.g., racial, gender, religious). If this occurs, even if there is no intent to discriminate, the personnel procedure results in a disparate impact on a subgroup. If the subgroup constitutes a protected class, there may be important legal ramifications for the organization and affected class. Adverse impact provides prima facie evidence of discrimination.
Identifying adverse impact Several techniques can be used to determine if the differences between subgroups are substantial enough to justify labelling them as adverse impact. Perhaps the best known is the four-fifths or 80% rule. The four-fifths rule has been codified in the Uniform Guidelines as a possible guideline for establishing adverse impact. Based on the rule, adverse impact is present when the selection ratio for a minority group is less than 80% of the selection ratio in the comparison majority group. For example, assume 10 White applicants take a selection test, and 5 Black applicants take the same test. All 10 of the White applicants pass and are offered a hire, compared to 3 of the Black applicants. Therefore, the hiring rate for White applicants (the majority group in this example) is 100%, whereas the hiring rate for Black applicants (the minority group in this example) is 60%. The minority/majority ratio hiring rates are 0.60/1.00, or 0.60. Since 0.60 is less than 0.80, adverse impact would be considered evident in this example under the four-fifths rule.
While Whites and/or males are often the majority group in these comparisons, they could in some circumstances be considered as the minority group. While there are protected classes of individuals under Title VII (e.g., race, gender), there are no specific subgroups (e.g., Blacks, women) that are considered as always favoured or discriminated against. Therefore, the majority and minority classification in adverse impact analyses vary and are based on the composition of the job, organization or industry (Hanges et al., 2013).
The four-fifths rule is frequently used to identify adverse impact and is easily understood, particularly by practitioners. However, there is evidence that it leads to false-positive indications of adverse impact (Roth, Bobko & Switzer, 2006). In addition, the four-fifths rule is strongly influenced by the way the test is used (e.g., how severe the cut score or fail point is).
Fortunately, there are other ways to identify adverse impact. The Uniform Guidelines reference using statistical tests to identify adverse impact, though the Guidelines also stress the need to have an adequate sample size. While many types of statistical analysis might be used to identify adverse impact, Fisher’s exact test, the chi-square test and the Z test for difference in proportions are the most frequently implemented techniques (Hanges et al., 2013). Murphy and Jacobs (2012) also argue for the use of effect sizes, such as standardized difference and the percentage of variance explained, as indicators of adverse impact that do not rely on statistical significance testing to make this determination. This recommendation fits with best practices in academia as effect sizes help clarify the magnitude of an effect above and beyond its statistical significance.
Using more than one type of adverse impact test (e.g., using the four-fifths rule, an effect size estimate and a statistical significance test) can help reduce the number of false- positives and provide more context to the presence or absence of adverse impact. The rich information provided by multiple types of adverse impact tests is important for those making decisions in the legal process as it offers a more detailed understanding of the test and its potential effects on protected groups. Because these statistical tests do not overcome some of the problems that can also create challenges for the four-fifths rule, such as small sample sizes, they should not be seen as replacements for the four-fifths rule or other adverse impact tests, but rather as complementary to them.
These methods for identifying adverse impact are established by guidelines and enforcement agencies in the US, but other countries often do not specify how adverse impact should be demonstrated (Sackett et al., 2010). Within the European Union alone a number of member states allow for statistical tests as an indicator of adverse impact whereas others may perceive the same tests to be unacceptable or insufficient (Hanges & Feinberg, 2010).
Regardless of how adverse impact is measured, it is important to recognize that evidence of adverse impact is not a feature of the test, but a feature of how the test is being used in a given context. A test may cause adverse impact for one job in a specific context but not for other jobs or the same job in different contexts. Using a given test in combination with other selection tools or with cut scores will impact the percentage of people passing the test, which in turn will influence pass rates for different subgroups. Therefore, tests of adverse impact need to be understood in relation to both the features of the test itself and how it is being applied in a specific context (Hanges et al., 2013).
It is important to remember that establishing a prima facie case for adverse impact does not mean a test is discriminatory. As discussed above, in most countries there is a shifting burden of proof that must be addressed by plaintiffs and defendants in turn (Sackett et al., 2010). A plaintiff meets the burden of production requirement to show that discrimination may exist by presenting evidence of disparate treatment or adverse impact. It then falls to the defendant to demonstrate a burden of persuasion. That is, the defendant has to show that a qualified applicant was not hired for non-discriminatory reasons in a disparate treatment case or show that the selection tool that was used is job-related and meets the organization’s business necessity in an adverse impact case (Hanges et al., 2013). Finally, if the defendant establishes the job-relatedness of the test in question, the burden of proof shifts back to the plaintiff. In this final step, the plaintiff must offer alternative selection procedures that show reduced adverse impact but equivalent jobrelatedness.
In summary, the Uniform Guidelines outline two forms of discrimination - disparate treatment and disparate (adverse) impact - which may be seen in an employment context. Demonstrating disparate treatment requires showing that a member or members of a protected group were treated differently from other applicants in a way that disadvantaged them. Demonstrating adverse impact requires showing that the same personnel process applied consistently across applicants results in differential outcomes for individuals from different protected groups. Various measures can be used to assess adverse impact, including the four-fifths rule, statistical significance tests and effect size measures. Ideally, multiple measures are used to provide triangulation of adverse impact evidence. Finally, adverse impact cases rely on the shifting burden of proof model, where the plaintiff must first establish adverse impact, followed by the defendant establishing the job-relatedness of a personnel procedure and concluding with the plaintiff suggesting alternatives that show less adverse impact but serve the business’s purpose. Next, we provide an overview of a key component of the defendants’ arguments establishing job-relatedness: the psychometric properties of the test.