Abstract Diversifying science, technology, engineering, and mathematics (STEM) requires a critical examination of institutional structures at every educational level. In higher education, performance in core introductory courses required for STEM degrees is strongly associated with degree completion. Leveraging a multi-institutional database, we examine nearly 110,000 student records from six large, public, research-intensive universities in order to assess whether these introductory courses disproportionately weed out underrepresented minority (URM) students. We find that the association between low performance in an introductory STEM class and failure to obtain a STEM degree is stronger for URM students than for other students, even after controlling for academic preparation in high school and intent to obtain a STEM degree. To facilitate interpretation of our multivariate logistic regression model, and to highlight the dire situation in higher education, we also calculate predicted probabilities of STEM degree attainment for students of various demographics. The probability of obtaining a STEM degree for a STEM-intending white male student with average academic preparation who receives grades of C or better in all introductory courses is 48%. In contrast, for an otherwise similar URM female student, the probability is merely 35%. If these students receive less than a C in even one introductory STEM course, the probabilities drop to 33% and 21%, respectively.