In this post, rather than a personal view or opinion, I am sharing the findings of a small study we did on impact of grading schemes. Grading schemes in universities vary with many systems in existence. To study the impact of grading schemes on student’s performance in terms of SGPA/CGPA, we have performed a small study to study two common grading schemes. Scheme 1 which uses letter grades A, B, C, D, and F with corresponding points as 10, 8, 6, 4, and 2. And scheme 2 which uses grades A, A-, B, B-, C, C-, D, and F, with corresponding points as 10, 9,…, 4, 2. 

Experimental Setup

To analyze the impact of the two grading schemes, we have used the data of students’ marks from three core courses: Computer organization, Data structures, and algorithms and Probability and Statistics, each with enrolment in of about 140. The data of the students from these three courses was first divided into two groups of equal sizes Group 1 with even numbered students and Group 2 contained odd numbered – i.e. both groups had about 70 students in each class with similar performance.

First, Group 1 was given to the six professors from three institutions, and they were requested to grade the students using the Grading scheme 1. They were given the entire spreadsheet of performance over the semester, but without the student names. After 2 weeks, Group 2 was given to the same professors for grading using scheme 2. The purpose of the two week gap is to make two grading exercises independent of each other.  At the end of this experiment, we had the grades provided by six professors using the two different grading schemes for sets of students whose performance was effectively similar. And we had this data for three different courses.

Analysis and Key Observations

  • Average grade point of students (after taking the average grade point of six professors) is approximately the same at around 6.5 (average of scheme 2 was higher by about 0.1, but given the small sample size it was not taken as statistically significant.) This is an useful  insight – it shows that professors do not simply take students falling in A category in Scheme 1 and divide them into A and A- (and similarly for B and C), as that would have resulted in significant reduction in average grade in Scheme 2. But, as is intended, by having finer grades in scheme 2, Professors put some A students (of Scheme 1) in A-, but also put some B students (of scheme 1) in A-. (It is worth pointing that some professors consistently gave higher average grade in one scheme, while others gave higher grade in the other scheme.)
  • Average grade point (after taking the average grade point of six professors) of top students is higher with grading scheme 1 across courses. This is to be expected as in Scheme 2, some students from A grade in Scheme 1 will get moved to A-. However, one Professor gave more As with grading scheme 2 in one course. (To study this, we determined the average grade of top K students, varying K from 1 to 15, i.e. up to about top 20% students.)
  • Average grade point (after taking the average grade point of six professors) of bottom students is higher with grading scheme 2 across courses.  In other words, bottom students would prefer grading scheme 2. This is also as expected – with finer grades; fewer students should end up in D and F.  Here also,  two Professors gave a larger no of Fs with grading scheme 2 for one course each, (For this also, we studied the grade of bottom K students, varying K from 1 to 15).
  • The average number of Fs reduces with grading scheme 2 to about half of those in Scheme 1. Even at individual professor level, number of Fs reduce –  4 out of 6 professors have given more or equal Fs with grading scheme 1 than grading scheme 2 across all the courses; in some courses, the same professor has substantially lower threshold for given F in Scheme 2 than in F.  One can argue that as there is no D- grade, there should be no impact on number of Fs. But data seems to suggest that overall, having a finer grades seems to reduce the need for failing students.
  • The data provided us an opportunity to analyze the consistence in the grading behaviour of professors. For this purpose, we postulated five consistency hypotheses:
    • H1: Equal or more As in grading scheme 1 than in grading scheme 2
    • H2: Equal or higher cut-off for grade B in grading scheme 1 than in grading scheme 2
    • H3: Equal or higher cut-off for grade C in grading scheme 1 than in grading scheme 2
    • H4: Equal or higher cut-off for grade D in grading scheme 1 than in grading scheme 2
    • H5: Equal cut-off for grade F in grading scheme 1 and in grading scheme 2

We then determined how many of these hypotheses were satisfied by different professors.  We found that 5 out of 6 Professors  satisfied 4 or 5 of these (four did not satisfy  H5, as mentioned above). However, one Professor satisfied only 2 of these.  In other words, most Professors are quite consistent in their grading behavior across schemes (except for F grade.)

Acknowledgements

I would like to thank Mayank Pundir for his help in analyzing the data and in writing the report, and Vidushi Chaudhary for her help in performing the experiment and initial analysis. I would also like to thank the professors who participated in the study and graded the students using the two grading schemes (not mentioning their names for confidentiality).  Details of the analysis are also available with the author.

Advertisements