ABSTRACT
During the past four decades, as access to higher education has increased for virtually all segments of American society, the demographic composition of the pool of examinees taking college admissions tests has changed markedly. In particular, during this period, the test-taking population has become increasingly diverse and is now more representative of American society as a whole. As diversity increases, so do questions about the tests that are used in college admissions. Specifically, are these tests equally valid for all subgroups of test takers? Studies about test validity for different populations of examinees fall under the general topics known as differential validity and differential prediction. Historically, test-taker subgroups classified by race and sex have been the ones most studied since these groups are of greatest interest for educational, political, and social reasons, and because they are the groups most easily identified. The first studies of race and sex differences in college admissions testing date back more than 30 years (Cleary, 1968; Linn, 1973). Although other demographic characteristics, such as family socioeconomic status and students’ English proficiency, are of wide interest, few studies of differential validity and prediction using these variables have been conducted to date. The studies by Leonard Ramist, Charles Lewis, and Laura McCamley-Jenkins (1994, 2001) are two of the few studies that examined student language proficiency and parental education for group differences in validity and prediction. Previously, summaries of the research findings in differential validity and prediction have been written by Hunter Breland (1979), Robert Linn (1982), and Richard Durán (1983).
