Making a New Grade

UW considers participating in Duke study of alternative GPA

Mary is one of a group of four students who have taken some of their classes together. Any time Mary took a class with one of her three friends, she received the highest grade.

John, another member of the group, always received the lowest grade of the four when he was in courses with any of them. Sam and Linda, the other two group members, received various grades and rankings, but Sam always outscored Linda in any course they took together.

One might expect that Mary would have the highest grade point averages of the four, followed, respectively, by Sam, Linda and John.

However, it is quite easy to construct course and grade scenarios in which their GPAs are in the exact opposite order. If most of John’s courses are graded on easy scales — awarding mostly A’s and B’s — while most of Mary’s courses are curved to generate an average grade of C, it is quite possible that John’s GPA would be higher than Mary’s.

Similarly, if most of the students in John’s classes are of average ability, while most of the students in Mary’s courses are honor students, it may be inherently easier for John to get higher grades than Mary.

Issues like these have long troubled both faculty and students and have led to periodic calls for reforms. Just such a call is now being issued by a faculty committee at Duke University, where work is underway on an achievement index that could supplement or even replace the traditional GPA.

Duke has asked 15 to 20 universities, including UW–Madison, to participate in a study of a new method of ranking grades, called the achievement index.

“The problem with the GPA is that it is based upon the erroneous assumption that grades from different courses are comparable and can therefore be added and averaged,” says a report from Duke’s Committee on Grades titled “Incentives To Excel.”

The achievement index, or AI for short, is calculated by a statistical technique that involves estimating grade cutoff levels — the lowest percentage that could receive an A, for example — for each class, given the ability levels of the students in the course. The process is then repeated until it converges on stable values for the grade scales and relative student abilities.

That means that if student A usually performs better than student B in courses they take together, then it is overwhelmingly likely that the achievement index of student A will be greater than that of student B.

This, proponents claim, removes any incentive for students to seek out courses they believe to be graded on easy scales and reduces the pressure on faculty to inflate grades. One disadvantage (and the reason the AI is unlikely to completely replace the GPA), is that neither students nor faculty would be able to compute the AI values for individuals, as the computations involve all grades for all students in all courses.

The request for UW–Madison’s participation in studying the index came from Valen Johnson, an associate professor of statistics at Duke, in a letter dated Aug. 6 to Brent McCown, professor of horticulture and chair of the University Committee, and David Musolf, secretary of the faculty.

McCown brought Johnson’s request before the University Committee on Sept. 8, which referred the request to the Undergraduate Education Committee for consideration.

That committee will begin reviewing the request this month, says Robert Miller, professor of business and statistics and chair of the committee. Once recommendations from the undergraduate-education committee are heard, the University Committee, the executive committee of the Faculty Senate, will decide how to proceed.

“I think it’s a very clever system that has some advantages,” says Provost John Wiley. “But there isn’t any one number that adequately summarizes a student’s academic aptitude or achievement.”

Johnson wants UW–Madison and other universities to send him course and grade data from an entire semester or academic year for an entire cohort of students — the sophomore class, for example. Johnson would then analyze the data using his statistical model and assign adjusted GPAs, or AIs, to each student.

In a recent study at Duke that used the achievement index, students at the top and bottom of their class generally stayed close to their traditional GPA and class rank, but some students in the middle changed greatly.

Grade inflation is, in part, fueling the desire at Duke for the achievement index.

The percentage of A’s given at Duke rose from 21 percent to 46 percent of all grades between 1986 and 1994, according to the “Incentives to Excel” report. If that trend persists, 97 percent of all grades at Duke will be A’s in 25 years, the report says.

Wiley says UW–Madison has seen a steady trend of fewer F’s and more A’s on campus over the last 10 years. But he adds that each incoming freshman class over the past decade has entered with better academic credentials than the previous class.

The Duke report is available on the Web. A summary of the report is also available.

	Standard GPA	AI GPA	Standard Class Rank	AI Class Rank
Student A	3.97	3.95	8	2
Student B	3.59	3.21	296	822
Student C	3.21	3.54	822	340
Student D	2.45	2.68	1280	1209