Marking standards and the differential predictability of the first-year university performance of different demographic groups

Abstract

This study investigates the differential validity and differential prediction of the first-year university marks of two cohorts of students from historically white (HWS) and historically black high schools (HBS) with matriculation marks being the predictor. In agreement with results obtained elsewhere, a common regression line developed for the entire group underpredicted the university marks of HWS women and of HWS students as a group, and overpredicted that of HWS men and of HBS students as a group. To adjust for differences in marking standards, a procedure proposed by Elliott and Strenta (1988) was used. Because HWS women and HBS students as a group tended to prevail in leniently marked courses, as opposed to HWS men who tended to predominate in strictly marked courses, such adjustments reduced the over- and underpredictions for HWS men and women, respectively, but also increased the overpredictions for the 1996 HBS women and for the 1999 HBS men. For all groups, the correlation between matriculation marks and university marks were slightly higher following such adjustments. The effectiveness of the adjustment procedure was impeded by the very small number of students who took courses in more than one faculty. South African Journal of Higher Education Vol.15(1) 2001: 129-137