Relative enrollment in online classes has tripled over the last ten years, but the efficacy of learning online remains unclear. While two recent Meta analyses report higher exam grades for online vs. traditional classes, this body of research has been marked by two recurrent limitations: (1) a possible problem of selection bias wherein students self select the mode of course delivery and (2) a relative lack of proctoring of exams in online sections. Both of these confounders contribute to observed differences in performance. The present study addresses these limitations. Data refer to 64 students enrolled in criminology classes at a Carnegie research extensive university. Due to an administrative error in the course schedule, which failed to list one section as online, students were unable to self select into the online section, creating a rare opportunity for quasi randomization of students into sections. Both sections were taught by the same instructor. The dependent variable is the score on the standardized final examination. All exams were proctored by the instructor. The central independent variable is method of delivery of content: online vs. the traditional classroom. Controlling for other constructs, there was no significant difference between exam scores. Also, student evaluations did not differ between sections. Controlling for selection effects and the proctoring of exams, the academic performance of online students was the same as that of traditional students. Future work is needed for other courses, other fields, and other types of academic institutions.