Underprediction of performance for US minorities using cognitive ability measures

The purpose of this paper is to examine moderating effects of employee race on ability–performance relationships in a well‐controlled study.

The cognitive ability of a large sample (n = 972) of employees in a garment manufacturing organization was measured using a well‐validated instrument (the General Aptitude Test Battery). Relationships to objective measures of performance were assessed for differential validity.

Contrary to expectations, the correlation between ability and performance was found to be stronger for black employees than white employees. This results in underprediction of performance for black job applicants if a common cutoff score is used.

The near demise of research on differential validity may be premature. Subgroup differences are more likely to be detected when appropriate research designs are used.

Organizations may risk moral and legal problems if they use selection procedures without adequately addressing potential problems with differential validity.

This paper stimulates interest in examining potential race‐based differential validity effects when examining organizational selection procedures.