Evaluation of the U.S. Employment Service Workplan for the GATB Improvement Project (1995)

This report evaluates the research plan on the General Aptitude Test Battery (GATB), proposed by the U.S. Employment Service of the U.S. Department of Labor (see Appendix for the workplan). In particular, we were asked to answer three questions:

1. Does the workplan adequately address the major technical issues raised in the report of the National Research Council (NRC) Committee on the General Aptitude Test Battery?¹

2. Are there any significant gaps/flaws in the basic approach reflected in the workplan?

3. In particular, does the workplan reflect an adequate research response to recommendations in the 1989 NRC report in the key areas of reducing score differences and adverse impact and testing persons with disabilities?

The NRC committee's analysis demonstrated that the modest predictive validities of the GATB cause selection errors that weigh more heavily on minority than on majority job applicants. It was the committee's conclusion that this disproportionate impact of selection error provides grounds for adjusting the scores of minority applicants, in order for able minority applicants to have the same chances of referral as equally able majority applicants. In reaching this conclusion, the committee explicitly linked the analysis of predictive validity to the question of adverse impact that occurs when the GATB is used for job referrals. As to questions of whether that method was legally, politically, or socially acceptable, the committee noted that it was not the appropriate group to answer.

The committee's recommendation concerning score adjustments became moot with the 1991 Civil Rights Act. With the score adjustment method recommended by the 1989 report now prohibited by law, fundamental questions about the appropriateness of GATB in its current form must be addressed. Although within-group score adjustments were found by Congress to be an untenable means for compensating for the weaknesses of selection tests like the GATB, this does not imply that the legal (or social) environment now supports the use of a selection tool that both has low predictive validity and produces severe adverse impact.

Indeed, the 1991 Civil Rights Act presents a challenge--and an opportunity--for concerted attention to underlying assumptions about employment testing, in general, and for comprehensive evaluation of the GATB research and development program, specifically. The fundamental question now facing the Department of Labor is whether modest