Different Denominators, Different Results: Reanalyzing CS Degrees by Gender, Race, and Ethnicity

Standard analysis of computer science degree data focuses on the percentage of degrees earned by women and men respectively, or on the percentage of degrees earned by racial and ethnic minorities respectively. This analysis is inaccurate because the groups involved are not the same size and their sizes have changed over time. Longitudinal comparisons are relevant only if statistics are computed separately for each cohort, examining the percentage of each cohort’s degrees earned in CS. The numerator should be the number of degrees earned in CS by a cohort, while the denominator should be all degrees earned by that cohort, not all degrees earned within the field.