Academic Integrity in On-line Exams: Evidence from a Randomized Field Experiment

We study academic integrity in a final exam of a compulsory course with almost 500 undergraduate students (mostly in Economics and Business Management and Administration) at a major Spanish university. Confinement and university closure due to Covid-19 took place by the end of the last lecture week. As a consequence, the usual classroom exam was turned into an unproctored on-line multiple-choice exam without backtracking. We exploit the different orders of exam problems and detailed data with timestamps to study students’ academic integrity. Taking the average over questions that were part of both earlier and later “rounds,” we find that the number of correct answers to questions in the later round was 7.7% higher than those to the same questions in the earlier round. Moreover, the average completion time of questions in the later round was 18.1% shorter than that of the same questions in the earlier round. We estimate that between 13.4% and 22.5% of the students cheated due to information flows from earlier to later rounds. Nonetheless, since exam grades are positively correlated with previous continuous assessment, they can be considered informative. Finally, a mere reminder of the university’s code of ethics, which was sent to a subgroup halfway through the exam, did not affect cheating levels.