Grade inflation: What goes up must come down

In undergrad, the professor who taught intermediate microeconomics told us about how law and business school admissions officers knew how tough his course was (it was the weeder course), and that doing well would be a feather in our cap. I made the mistake of believing him. The savvy students took intermediate micro at nearby schools and transferred the credit along with a higher grade.

Graduate schools will look at rec letters and take the professor's praise into account, but they haven’t the slightest notion of whether a class you took is hard or not. Students who want to optimize their future prospects know this, and select their courses accordingly. Oh, and employers know/care even less than graduate schools.

One thing I actually really like about the big tech companies is they will ignore low grades if you interview well. I had a semester where I failed almost every class due to a life event and was too proud to admit to my parents I should probably just withdraw. This dropped my gpa from a 3.3 down to a 2.6. Two years later when I graduated my GPA was still below a 3. I remember I was sitting at an interview with John Deere and they told me that someone with my academic failings probably wouldn’t do well at John Deere. Broad rejections from General Motors, ford, Boeing, many others. But Microsoft and Google did not seem to care at all. I was told I got a solid yes from all my Microsoft interviewers and was actually able to negotiate a level bump up and a larger signing bonus. For me it showed that Microsoft/google are thinking holistically in their HR decisions, and only use GPA as a signal. I hope my success at Microsoft since then is an indication that they made the right call.

I went to a university in South Africa which had the opposite problem, grade deflation. The engineering school was masochistic in its obsession with lowering grades and failing students. Despite working extremely hard I went from 90%s to 60%s. But I was considered lucky because most students failed (and by most I mean +90% of the class) and repeated years. I did not. The lecturers repeatedly blamed the students but if even A students are failing I think it's not the students fault, it's the school's.

My university teamed up with an American University, Embray-Riddle, to offer a joint masters. Embray-Riddle required a minimum average grade of 80%; my university came back to them and said no one in the last 10 years qualified for the masters. Eventually they lowered the requirements just so they could have students.

Later I did my masters in the Netherlands. The university made me do extra courses to compensate for my low grades. But I ended up doing reasonably well with an 8/10 average and was the top student in a few subjects.

Grade inflation sounds bad but I can positively say that grade deflation is worse. It badly demotivates students and robs them of years of their life as they repeat courses.

My question is, why wouldn’t it be this way? For two equivalent students from equivalently prestigious schools, if one has a 3.5 GPA and the other has a 3.9 due to grade inflation, of course the second student will have more opportunities. And more opportunities means higher pay, and higher pay means more donations back to the school. Goodhart's law and all that.

Grade inflation and the grading idiosyncrasies between schools make it such that GPA is fairly meaningless as a metric to me when I’m conducting interviews. I simply don’t care.

I worked my butt off in undergrad to barely get a 3.1 at one school, and in grad school it was hard to not get an A in a course.

Instead, I care about what you as an interviewee know. Not what school you went to, not what contrived number you scored.

I think the premise of this article is flawed — aside from the (imo very valid) “who cares!” argument with regards to grades not actually mattering for really anything anymore except when it comes to some grad programs, the implicit assumption of this analysis is that the grade system will exist in its current form in the future.

I think it’s more likely Harvard does away with the traditional letter system altogether. Aside from the fact Harvard operated just fine without it for the first 2-3 centuries it operated, let’s take a step back and try to remember what the grades were to track in the first place: students’ absorption of the material (i.e. “learning”). This type of grade system is better for measuring things that have right/wrong answers— math, memorization of dates, anything mechanical, etc. Yet with calculators and the internet making things like historical timelines and formulas etc accessible even to payment, graders’ job of even grading these “objective” becomes more subjective— if you arrive to the wrong math question, but did all the right high level steps and made a small but cascading error in the beginning, I’d posit that student likely knows the material well and would avoid that error were it pointed out. But from a grading perspective, what is that? 7/10? Same situation to a history student laying out the historical context leading up to some event and mixing up some dates. Things get even more muddled when it comes to humanities.

In fact, off the top of my head, Pomona College and Brown University, both well-regarded schools within academia (and Pomona with I believe the highest undergrad acceptance rate into Harvard grad programs), have the option for every class to be taken as Pass/Fail, except when it comes to later-stage major requirements. Graduation accolades are awarded by professor recommendation, which fosters getting to know your professors out of the classroom better than grades. Anything that needs to be GPA-gated could instead be (likely better) gated by a placement test.

Knowledge is not a 0 sum game. Maybe instead of trying to adapt our current education circumstances to to an old grading system, it’s time to think of a new system altogether.

You can tell this was written by a student who doesn't really understand grades or what the problem is.

> You may have gotten a C in Math 55 but your potential employer could see the average grade was a B+.

This never happens. No one cares if you were a C student at Harvard or an A student. All they care about is the degree. With one exception: grad school. Bad grades can shut off doors for more education.

From he point of view of people who give grades:

1. High grades are free, low grades are expensive.

I can mint as many As as I want. No one will ever complain.

Give a student a B or a C? I have had students flip. Their parents can call. They can appeal. Some people just cause endless problems.

2. I don't need people posting hate about me online

When you give students good grades they say great things about you on ratemyprofessor and on surveys. When you give bad grades.. they say bad things. No thanks.

This has even resulted in lecturers losing their jobs.

3. Cheating distorts grades anyway

Can you imagine the disaster when you catch most of a class cheating? In my experience a significant fraction of students cheat. Guess how many hours of your life you will waste if you pursue every clear cheater?

With that in mind, if you grade on a curve and set a low average grade, you're just punishing the honest stuents.

4. Who cares!?

So what if everyone gets an A? What's the point? Given that no employer cares and that your wealth is largely a function of your parent's wealth, why are we wasting so much time on this. Any reasonable effort should merit an A and then we leave people alone. Students who want to excel will go on to do research, side projects, etc. regardless.

A couple of year ago I applied for a graduate position at a prominent Canadian university, and was accepted, based in numerous factors (like publication history, work experience, and academic program). But a few months later, I got a notification that I had been cut: my GPA was too low. My university in Australia didn't even have GPAs back then 1990s-2000s), and it was a very hard marked program (maybe 5% of the class would get over 90%). H3 was a solid pass. But I was competing with people who had just finished undergraduate and master's studies in North America, all of whom had almost perfect A+ GPAs.

The worst part is that it reveals that universities believe their own bullshit. In fact, believing the bs seems to be a strategy for career success.

"Harvard students, who can be very risk-averse and afraid of failure..."

An interesting statement, which seems worthy of its own article.

(Giving some context into this as founder in edtech -- running a 4 yr CS undergraduate program at kalvium.com)

One reason for grade inflation is that this is currently the only filter provided to the industry, who would like to filter students they review. This has ended up as a perverse incentive, and students and the university want to be above the "cutoff" GPA.

One of the solutions we're going with is to provide a 5 point grade per course, and provide this full breadth of data as a filter to the industry during the placements/internships season. Our bet is that the industry will have more diverse options to go by and we end up with better match-making.

Another solution is to have publicly verifiable artifacts of coursework (plagiarism is a significant problem in higher education).

I'd also like to hear HN's recommendations on this topic.

In the UK, at the university level, grade inflation is often related to the high annual cost of tuition.

Students are increasingly treating university as a paid service with universities treating students as customers.

Failing too many students is a failure in the service provided - bringing the university into some sort of liability.

Further, students are working hard for the £50,000+ worth of student debt they accumulate. Which wasn't the case generations prior.

At my uni the grade transcript from the university includes the grade distribution for the course next to each grade. So it's easy for a company to see that "a B in this course was comparatively good" for instance. It was almost always a curve. Where about 10% got an A, and the median of course was a C etc.

I've always found it so weird how in the US the scale isn't used. By doing all course work you're often assigned enough credits to guarantee a B (the challenge is having enough time to do it all, I guess), and by doing some more you'll get a A- or whatever. I've had friends that wee C/D students here take a year in the US and get straight A's.

Anecdotally I think some gpa inflation is related to modern course design.

What was the design and pedagogy (workload too) for a biology course at harvard in the 1920s versus today? Without this info the graph tells us less than it might.

What classes were being offered in 1920s harvard?

At my university we dont offer classes that were at one time considered essential

Want to learn about virgil? Well, theres the internet

Grade inflation is always and everywhere an admission rate problem.

The US college-applying population has grown (a lot). The Harvard class size remains the same. Harvard continues to be a (the?) top tier school. This means that Harvard classes are probably getting better year after year. Even if Harvard is evaluating performance exactly the same (no new cushiness), if the admitted are better, they’ll be getting better grades on average, and those grades will be more compressed.

Maybe the problem is the way we average grades to get a GPA. If your course scores were 3.98, 3.99, 3.99, 2.10, the linear average of ~3.5 doesn't capture the obviously high ability. Perhaps taking the average like:

  4.01 - exp(avg(log(4.01 - x_i)))

would capture high grades without such a heavy penalty for a flubbed course.

Article doesn't explain why it "must come down" only offers a few side effects of it being high.

I've been always talking that grades are irrelevant because there's too much external factors (easy/difficult profs, biases) and standardized tests are the way to go

Grades could at best be proxy for effort, but still grades talk about the past, not what's now.

If somebody takes grades seriously then for me that's huge red flag and I think that person is naive and focuses on not really important things.

I wish there were standardized exams after graduation, so you could compare students from top schools and bottom more easily, that would allow students from bottom schools to bootstrap themselves purely on their skills and knowledge

The GPA system has been a bit of a mystery to me since I discovered its existence. As a European from a country that has no such system, it seems like a bit of a rat race, where people giving grades have bad incentives to pad them. Where I'm from, you graduated from a program or you didn't ; the only distinction was "with merit", or "valedictorian" and that's it. Your distinction comes from the institution you receive the sesame from, and the ranking it receives from random industry rankings and simple reputation. Not a perfected system, but a bit more straight forward.

Now in Canada, I receive resumes from new grads who have either >3.8 or no GPA posted. I'm yet to see a correlation between those GPA numbers and superior performance, but the university they come from is a stronger predictor of fit and effectiveness. Therefore I tend to ignore the GPA (or worse, consider the boasting as a negative) and put emphasis on university. I'm dubitative what the value of the score actually is.

> When grades weren’t as inflated, a stray C+ or B- was to be expected. As A-’s become the new default and grades stabilize at the upper end, any lower grade is seen as a failure. Because most students receive consistently high scores, the cost of experimenting with hard classes or new subjects substantially rises.

Not if the hard classes or new subjects have the same level of grade inflation. On the contrary, if students have nothing to fear about hard classes messing up with their grades, they are more likely to take them.

Most kind of grading schemes are subject to manipulation. Students always find ways to game the system. Teachers are often biased toward giving grades that are correlated with past performance of students.

Its better to have harder courses with pass/fail. The top students will have more motivation to pursue subjects they are interested in. Grades are not great motivators for students at the top. Richard Feynman, Albert Einstein (or insert your favorite physicist here) were never motivated by getting grades that look better than their classmates.

Only major issue is that how to make the admission process fair when there are no grades. Despite all their downcomings, standardized tests solve this to a large degree. When grades are not standard, students who come out of a fairer grading system suffer compared to the ones with inflation. Good luck trying to explain this to the admission committee or whoever else.

I've pondered this for a while. One solution would be, imo:

1. Maintain a large set of questions that are mechanically markable (e.g. no room for bias/teacher choice in marking). You can expand this set by a small percentage each year.

2. Randomly assemble each student's exam every year by sampling questions without replacement.

3. Use trueskill to simultaneously estimate question difficulty and student skill.

4. Report skills as the primary metric.

I think this would work, although of course writing many questions is challenging.

One question is whether you want to use grades for validating knowledge or for ranking.

You can say that A or B means pass (much easier for professors), or that 8/20 means pass and you have the whole spectrum from 0 to 20 to rank candidates if needed.

Nowadays, with international recruiting, I think grades becomes insignificant (graduates come from too many institutions, and grades aren't comparable) and companies need their own ranking metrics (hence the infamous leetcode questions).

The one thing that annoys me the most about modern grading is the fact that so many courses heavily weight homework and projects. You can do exceptionally well on tests and quizzes and objectively understand the material in the course but still get a terrible grade not based on knowledge but based on how much busy work you did. Grades are supposed to be just that, grading someone's understanding of the knowledge of the course. That should be the guiding principle for any kind of grading policy and yet it's so rarely the case when courses have 40% of the grade made up of tests and quizzes and then 60% based on homework, reading assignments, group projects, even attendance. I've had courses where a student could get a failing grade on every single test and quiz and still walk out with a B. Effort !== Knowledge and GPA should not just boil down to a measure of participation.

This also seems like a great way for Harvard to colonize as many institutions with graduates easily. If employers don't know about the compression at the top and they do value GPA and the reputation of the school, it likely creates a situation where a Harvard graduate gets employed at a company even when doing poorly on the interview because "Hey, we'd be stupid to not employ a Harvard graduate with a 3.8 GPA. They're top of their class from a very reputable institution."

Over time this puts enough Harvard people in institutions and Harvard people have a reputation for alma mater nepotism (that I've personally observed multiple times).

Almost all major US institution from media companies to banks to government have their upper ranks colonized by Harvard grads.

As echoed before in the discussion, I suspect for most jobs the GPA one receives is not relevant not even disclosed in job applications. It might matter for folks going for graduate degrees / law / med though. A 180 on the LSAT or 528 on the MCAT speaks quite a bit, too.

At West Point, until 2020 the big thing was class rank. Graduates could request their three top choices of branch of service (infantry, artillery, logistics, intelligence, etc.). Graduates with the highest class rank got their first choices. (This has changed a bit recently.[1]) So only the order of grades matters, not the values. Especially since everyone takes roughly the same set of courses.

[1] https://www.armytimes.com/news/your-army/2019/09/13/west-poi...

This is a serious problem that I think is more widespread than just Harvard or the US. Schools giving higher votes necessarily compress the variance with mean toward the high bound. The real problem is that everything becomes too much sensible to noise: getting a vote a bit higher or a bit lower than the mean doesn't reflect how much I know more, but may make a significant change in society, i.e. who gets ahead in academics or jobs (thus perpetuating this high grading mechanism, probably).

Add to this that professors seeking tenure benefit from high enrollment in their classes, and nothing spurs enrollment like the reputation of an easy A.

The only solution I can think of to all of this is to maintain a second number along with GPA, which is a weighted average of averages in every class the student took.

Earning an A- in a course where the average is a 2.5 becomes more valuable than getting handed an A in a course where the average is a 3.9.

Over time, the competition to get into these schools have gotten tougher and competition is now on a global scale for schools such as Harvard. If you don't grade on a curve, would it not be normal to see students do better? There are more past exam papers, more prep before you even join the school and more onus on the students to do well because they attend such a great school.

i don't think I've ever understood the point of letter grades and especially of GPA. what is it even meant to measure? clearly not mastery over material since we can see from this article we've seen a steady increase (while i don't think anyone is claiming students have gotten that much better). relative rank, then? again no because of the change over time.

i believe we should have two separate measures. one, did you pass the class? and two, what was your percentile within that course? of course percentile isn't necessarily relevant to employers. but then I would argue that should already be the case with GPA.

at least the benefit with percentile is it can't experience inflation like this, and gives you a comparable value across colleges. one would be able to make judgements on, would i want a 95th percentile state school candidate or a 5th percentile ivy? with the current system this is impossible.

The UK has grade inflation in secondary schools (11-18yrs) to the point where year-on-year results kept 'improving' (because each year was more intelligent than the last, clearly) then we introduced A* above an A (because too many kids were getting As...), and finally, replaced that with 1-9.

sooo theres no possibility students are trying harder these days or are just plain better students? The acceptance rate at harvard in 1969 was around 20%, compared with 4.6% today. It kinda makes sense that if you're accepting people who are better students, they're gonna get higher GPAs.

I've been considering applying for Georgetown/nyu law.

What is the acceptance rate like? Do they consider your actual grades or your references or both or neither?

I graduated high school in 2009 and the top schools definitely weren't taking in students with a 3.6 GPA. In fact, I'm not aware of any school in the top 50 that asked for anything under a 4.0 (AP/IB programs enable GPAs above 4.0). Right now I see sites saying that the average Harvard GPA is 4.18. Berkeley is lower but I know plenty of people that got rejected with high GPAs (I know two people that had a 4.8 -- the top you could get in my school -- that were rejected from several top schools).

I say this because two things. First, it is clear that there are things other than GPA that admissions are accounting for. It is a bit unclear what these things are. It would be nice to know how they are extracting signal from noise (and it is clear that this is a very noisy system, though we often don't admit this). Second, if these are actual averages, there's clearly some other aspects that are heavily weighted in that process. The four that would be the most obvious to me are legacy, affirmative action, sports, and students who don't have access to AP/IB programs. These are clearly not meritocratic aspects (lack of access to better programs like AP/IB would imply that these signals aren't meritocratic because meritocracy isn't exactly equitable in accessibility. Or do all these schools know exactly how good each high school is and their inflation rates and variance rates?).

I really think as we're learning more about how people learn that we learn that the signals are extremely noisy and may not even exist. Students from poorer backgrounds may need extra care[0] but they still do pretty well[1]. So how do you measure someone that has potential from a poor school? How does this differentiate from measuring someone that has potential but not the opportunities from median or rich schools? (environment is a big factor in success)

I really just have a lot of questions here and I really do think we should be discussing how noisy this system is. A question I want to ask is: if I randomly select a person from the population and give them a Harvard (or top tier) education, will they be more successful than if they went to a college they could have gotten into based on their merits? If the answer is "yes" (and I think there's good reason to believe this is true) then maybe all these measurements don't really mean much and we're just reading tea leaves. (I think there's some signal but I just don't know how much. I'm also not an expert. Thanks for listening to my ted talk)

[0] https://www.thecrimson.com/article/2022/11/16/klapper-privil...

[1] https://www.nytimes.com/interactive/projects/college-mobilit...

The obvious choice at this point seems to be just abandoning grades at Harvard. Any reason not to?

Perfect timing, make it harder. I’ve done graduated and moved on from the hustle and bustle.

never had a company look at my highschool grades...