Gender, Achievement, and Horizontal Stratification in Higher Education: Field-Specific Evidence from Large-Scale National Assessments in Colombia

Background Research on gender and field of study commonly treats science, technology, engineering, and mathematics as a single category. The fields grouped under that label differ in history, occupational destination, and gender composition, so an estimate computed on the aggregate need not describe any of its parts. Methods We linked two compulsory Colombian national assessments, one taken at the end of secondary education and one taken at the point of undergraduate graduation, for 126,366 students. Bayesian logistic models estimated enrolment in each of eight knowledge areas as a function of gender, secondary school achievement quintile in mathematics, natural sciences, and language, socioeconomic level, and cohort. Because both assessments are compulsory, the analytic population was benchmarked against the full national cohort of 2,792,844 secondary examinees, and estimates were reweighted to that population. Results The pooled difference of about 16 percentage points was concentrated in one field. In engineering, architecture, and urbanism, women were 11.5 to 20.6 percentage points less likely to enrol than men with identical achievement, and the difference widened as achievement rose, reaching 20.6 points in the top mathematics quintile. In mathematics and natural sciences, enrolment was practically equivalent, with posterior probability of equivalence within one percentage point above 0.99 in every comparison. In health sciences the difference favoured women by 3.9 to 5.5 points. Women who outperformed men by a full quintile remained 7.5 to 14.6 points less likely to enter engineering. Reweighted estimates differed from the primary estimates by at most 1.5 points. Conclusions Differences observed at the level of the aggregate category reflect mainly the pattern in engineering. The equivalence found in mathematics and natural sciences is positive evidence rather than an absence of evidence, and initiatives framed around the aggregate category may not correspond to the field-level pattern.