Underprediction of female performance from standardized knowledge tests: Afurther example from the knowledge of geography test

It has been documented that some tests of background knowledge underpredict the performance of female students in college. This study explored whether the underprediction phenomenon would also be found for a test that tapped four subfields of geography. Students (primarily White, N = 315) enrolled i n nine geography classes at a comprehensive, midwestern university complete d the Knowledge of Geography (KOG) test during the first week of the semest er and consented to release their first exam grades, final grades, and ACT scores. Replicating a previous study (Henrie, Aron, Nelson, & Poole, 1997), there were gender differences favoring males across all four subfields of the KOG test. KOG rest scores correlated with grades, but males and females achieved comparable course grades despite the lower performance of females on the KOG test. Examples illustrate how small differences between predict ed and actual grades can translate into large gender discrepancies whenever minimum scores from tests that underpredict the performance of a subgroup are used to qualify students for educational opportunities.