Objective: This study analyzes how ethnicity, gender, and non-traditional student characteristics relate to differential online versus face-to-face outcomes in science, technology, engineering, and mathematics (STEM) courses at community colleges. Method: This study used a sample of 3,600 students in online and face-to-face courses matched by course, instructor, and semester from a large urban community college in the Northeast. Outcomes were measured using rates of successful course completion (with a "C-" or higher). Multilevel logistic regression and propensity score matching were utilized to control for unobserved heterogeneity between courses and for differences in student characteristics. Results: With respect to successful course completion, older students did significantly better online, and women did significantly worse (although no worse than men) online, than would be expected based on their outcomes in comparable face-to-face courses. There was no significant interaction between the online medium and ethnicity, suggesting that though Black and Hispanic students may do worse on average in STEM courses than their White and Asian peers both online and face-to-face, this gap was not increased by the online environment.