Appropriateness of inferences: Test validity
In most countries the burden of persuasion for the defendant in an adverse impact case centres on the need to demonstrate the business necessity of a selection process. While the notion of providing evidence of the test’s job-relatedness is consistent across most countries, the evidence is often not specified (Sackett et al., 2010). An exception is found in the US where an employer must establish validity, or the appropriateness of inferences resulting from their selection process, to demonstrate job-relatedness. Validation is important for any assessment - selection processes specifically - as it involves providing empirical data to demonstrate that the inferences made from a test are appropriate. In the selection context validation efforts might focus on demonstrating that a test is an accurate predictor of performance on a job. Much like adverse impact, validity is not a feature of a test, but refers to the use of a test in a particular context. A test’s application may be more or less appropriate for assessing candidates depending on the goals of the testing process. For example, a test of computer programming may be a valid predictor of performance for a software engineer but will not be a valid predictor of performance for a kindergarten teacher. The very same test may not even be a good predictor of performance for software engineers in different organizations depending on the content of the job and the content of the test.
While there is often agreement between the Uniform Guidelines, Standards and Principles, in some situations these professional standards diverge, for example, in how the documents codify validity. The Uniform Guidelines have been interpreted as specifying three types of validity - criterion, construct and content - with each considered to be a different kind of validity requiring its own type of evidence, and with each seen as most appropriate only in certain situations. On the other hand, the Principles and the Standards consider validity to be a unitary construct with each of the three strategies supporting validity in a different way (see Landy, 1986). The three discrete validities of the Guidelines and the three strategies of the Principles and Standards are in fact the same, and therefore many of the principles for validation overlap across these documents. However, there are differences in how validity is treated in the documents given these different frames, and these have implications for employers and their validation processes. Here, our description of validity leans towards the Guidelines’ interpretations, as they are the ones most frequently cited in court proceedings (see Jeanneret, 2005); however, we encourage readers to review the Standards and Principles for a more comprehensive understanding of validity issues.