A Simulation Analysis: The Impact Of Grading On The Curve

International Journal for the Scholarship of Teaching and Learning http://www.georgiasouthern.edu/ijsotl Vol. 2, No. 2 (July 2008) ISSN 1931-4744 @ Georgia Southern University

The Impact of Grading on the Curve: A Simulation Analysis George Kulick Le Moyne College Syracuse, New York, USA [e-mail] Ronald Wright Le Moyne College Syracuse, New York, USA [e-mail] Abstract Grading on the curve is a common practice in higher education. While there are many critics of the practice it still finds wide spread acceptance particularly in science classes. Advocates believe that in large classes student ability is likely to be normally distributed. If test scores are also normally distributed instructors and students tend to believe
However we can assume some possible rationales. Supposedly these institutions and their professors feel compelled to distinguish performance among these outstanding students, to identify which students performed the best in the class and which were the poorer performers. Perhaps such rankings are designed to heighten the reputation of the institution by sending forward only the very best. Perhaps they are attempting to avoid grade inflation so that their best students will be clearly identified for the best medical schools. In many cases it is the very best institutions that are the most concerned about grade inflation (Gordon, 2006). However, any of these (or other reasons) must be based on the assumption that the best grades are going to the best students. It is this basic assumption that our work is designed to investigate. Does grading on the curve always, or even frequently, result in the best students getting the best grades? Simulation Model Investigating the assumption that the best students get the best grades would be difficult with samples of actual students. First defining what we mean by the best student is difficult. Do we mean the student with the best ability? Do we mean the student who is best prepared for the exam? Do we mean the students who know the most? And how would we assess the best by either of these measures other than by giving them an exam? How do we take into
Clearly the means and standard deviations in our model were selected somewhat arbitrarily. But the relationship between the standard deviation and correlation is unmistakable. Specific instructors can argue that as they make their exams more difficult they are also somehow making them such that the standard deviation of student preparedness also increases. However they cannot continue that argument forever. Clearly there is no ability to design an exam that will distinguish ability in the extreme case of a standard deviation of zero. And again, just as clearly, a normal distribution of test scores, by itself, provides no evidence of the exam’s capacity to correlate grades with ability. Just as importantly, the model suggests that even when there is more variation in student ability, luck still plays a role that can affect some students significantly. The primary conclusions of these simulations are best illustrated in the context of the extreme case of the outstanding students in large science classes. However the results can contribute to discussions about student assessment in all disciplines. In many fields text books come with test banks of multiple choice questions and instructors can randomly select the questions for their exams based on the chapters they have covered. There is certainly