This study compared the effectiveness of several existing and proposed methods for statistically adjusting college GPAs for course and departmental differences in grading standards, using first-semester grades from an entire entering class at a large state university. Most of the adjusted GPAs produced by these methods functioned similarly and, despite high correlations with actual GPA, had greater internal-consistency reliability than actual GPA and were more predictable from SAT scores and high school rank (HSR). Most of the adjusted GPAs also functioned similarly with regard to sex differences in over-underprediction. The adjusted GPAs and actual GPA exhibited the same small but significant sex differences in over-underprediction by SAT scores, but the adjusted GPAs displayed smaller differences than actual GPA in overunderprediction by SAT scores and HSR.
Adjusting College Grade-Point Average for Variations in Grading StandardsCollege grade-point average (GPA) , though originally intended for administrative purposes (Smallwood, 1935), is widely employed in educational and psychological research, particularly as a criterion for validating admissions measures (e.g., see the reviews by Breland, 1981;Fishman & Pasanella, 1960;Lavin, 1965).Despite the popularity of GPA, it is generally recognized that this is a fallible index of academic performance (e.g., see the reviews by Milton, Pollio, & Eison, 1986;Warren, 1971;Willingham, 1990). A major problem is that GPA is based on a different set of courses for each student, and the grading standards are not uniform from course to course, a phenomenon that has been observed for many years (e.g., Meyer, 1908). Hence, GPA is not comparable for students who take courses with severe grading standards and students who take courses with lenient standards, and its reliability and validity are attenuated.Differences in grading standards have been rigorously documented among departments (Anderhalter, 1962;de Nevers, 1984;Elliott & Strenta, 1988;Frisbee, 1984: Gamson, 1967Goldman & Hewitt, 1975;Goldman, Schmidt, Hewitt, & Fisher, 1974;Goldman & Widawski, 1976;Juola, 1968;Prather & Smith, 1976; Prather, Smith, & Kodras, 1979; Ramist, Lewis, & McCamley, 1990;Sabot & Wakeman-Linn, 1991;Strenta & Elliott, 1987; Willingham, 1985), as well as within departments (Garrison, 1979;Juola, 1968).The consequences of variations in grading standards on the reliability and validity of GPA are suggested by studies that attempted to adjust GPA for differences in these standards. The adjustments increased the median correlation between yearly GPAs from ,67 to .72 (Elliott & Strenta, 1988).The adjustments also generally boosted the correlations of admissions measures with GPA: the multiple correlation of the Scholastic Aptitude Test (SAT; -2-Donlon, 1984) scores and high school GPA with four-year GPA increased from .58 to .64 (Young, 1990b), and the correlations of the total SAT score (combining the Verbal [V] and Mathematical [M] scores) with four-year GPA went from .43 to .50 (Strenta...