We need to have a conversation about how we evaluate academic performance. This conversation should begin by recognizing that American higher education is at least doing better than most other places. Consider Oxford University, for example, which arguably assesses its students in the worst possible manner. There, most undergraduates are evaluated solely on the basis of examinations taken during their final month of university, resulting in a single number grade. Clearly, this system is utterly bonkers. It is relentlessly unforgiving, has awful consequences for mental health and disadvantages students who would fare better under a more holistic evaluation framework. As one Oxford residential college administrator recently said to me about this archaic tradition, “No one would ever design a university this way anymore.” In sum, universities like Oxford have got academic evaluation all wrong, letting their students pay the price.
While the more holistic approach of the U.S. is slightly better, significant room for improvement remains, both at Yale and the U.S. at large. In particular, there is a strong case to be made that we allow GPA to have far too great of an impact on student well-being and postgraduate opportunities, considering how simplistic of a measure it truly is. You’re probably expecting me to rehash a somewhat cliché argument: that numerical metrics of academic performance are simply no good at all and that we need an entirely different system. But while this notion certainly has some merit, I have to concede that there is utility in being able to concisely summarize academic performance. Instead, I want to make a simple statistical argument.
Specifically, we should reconsider the choice of the grade-point average as the one all-important statistic to care about, as it has considerable weaknesses as a measure of central tendency. One such weakness of GPA is that it weighs each grade equally. When considering GPA, your worst distributional requirement matters as much as the centerpiece of your major. Even in a liberal arts model, this strikes me as odd. But an even greater disadvantage is the general weakness of averages — they are highly affected by extreme values. This is especially consequential in a grading system that doesn’t allow for outliers in the positive direction. You could write a publishable thesis, but you still won’t get an A+++ to boost your GPA. But the grading scale certainly allows for negative outliers, meaning that one bad semester can hurt you more than it should. It only takes one personal tragedy or health crisis to attach a heavy anchor to your GPA, preventing you from being able to apply to certain internships, fellowships or get into the graduate school of your choice. I know students who are geniuses by any reasonable definition of the word but remain haunted by a single tumultuous semester. In short, GPA is a very crude way of summarizing a grade distribution and can be very misleading. In a way, it’s surprising that I even have to raise this point. Generally, we recognize that we shouldn’t rely solely on averages in discussions of, say, income distribution, where extreme positive outliers skew the statistic significantly.
So why don’t we consider the alternatives in the statistical toolbox? We could mitigate the impact of negative outliers by dropping the lowest grade from calculations of GPA, or begin a tradition of also showing grade-point medians or modes on transcripts. Or maybe we should stick with GPAs but allow transcripts to reflect students’ positive outliers — the A+++ performances. Another possibility is adopting a measure with weighted averages to better capture the relative importance of different courses. Clearly, there are several ways to preserve the utility of GPA as a concise metric while reflecting academic performance more accurately and honestly.
At the end of the day, it would probably be better if we all cared a little less about grades and a little more about learning, but I’m not too optimistic about this general obsession going away anytime soon. Still, if we begin to recognize the shortcomings of how we define success, there are small things that each of us can do, short of sweeping cultural change. The next time you need to evaluate a candidate for something, ask for their GPMedian as well as their GPA. When your friend winces at the one class they bombed, remind them of their positive outlier. Most importantly, the next time you look yourself and your GPA in the mirror, remember that this one statistic doesn’t define you. And I don’t just mean that in the cliché sense. I mean that GPA is literally an arbitrary method for evaluating academic performance. There is no fundamental law of statistics saying that arithmetic means are the best way of summarizing 36 data points — and trust me, my median stats grade is pretty sick.
Joshua Monrad is a junior in Saybrook College. Contact him at firstname.lastname@example.org.