Score contribution per author:
α: calibrated so average coauthorship-adjusted count equals average raw count
The Covid-19 pandemic resulted in a dramatic move to online education. Although schools and colleges have returned to in-person classes, student and professor interest in online testing in university contexts remains high, given concerns about testing anxiety as well as the considerably lower administrative costs associated with online testing. The modality of testing may have significant consequences for student performance, learning, and integrity. We vary the modality of testing for midterm exams in two large, introductory courses at a state university. We find that students perform substantially better on online exams, but that the premium largely disappears if never-before-seen questions are used. The online premium for low-performing students is particularly large, exceeding a full letter-grade, which is likely to have considerable implications for course pass rates. These results have significant implications for instructors seeking to gain the logistical simplicity of online testing and the benefits of increased student satisfaction, without encouraging dishonesty in testing.