This paper describes an experiment focused on measuring and explaining differences in students learning between online and face-to-face modes of instruction in college level principles of economics courses. Our results indicate that students in face-to-face sections scored better on the Test of Understanding College Economics (TUCE) than students in online sections. We find that failure to account for the self-selection of students into online or face-to-face sections biases toward zero the differential in TUCE scores between online and face-to-face students. Online students score a statistically significant 3-6 fewer correct answers, out of 33 questions, than face-to-face students in the selection-corrected model. However, an endogenous switching model finds that students who select into the online classes perform better than they would, all other things constant, in a face-to-face class. Other results suggest caution in using the web to teach underclassmen.