Detecting and Correcting Shift Errors in Standardized Exams

Hundreds of millions multiple choice exams are given every year in the United States. Form filling shift errors, where a single absent-minded mismarking displaces a long run of correct answers, can substantially reduce scores and invalidate the results of the exam. In this paper, we study the prevalence, impact and detectability of shift errors by analyzing over 100,000 Scholastic Amplitude Tests. We can effectively distinguish approximately 50% of the these shifts, and 80% of the substantial shift errors which reduce the score by 3 or more correct answers. Our resul ts suggest that approximately 1.8% of all SAT exams contain shift errors. Extrapolating these results over all multiple choice exams implies that students make millions of undetected shift errors each year. We describe our methods that accurately detect and correct the majority of significant shift errors while guaranteeing nearly no false detections. We argue that our methods should be integrated into automated testing procedures.