Sunday, May 23, 2010

Grade Inflation?

I know Harvard is supposed to be famous for its grade inflation, but that's not generally the case in my class (and, I think, for our CS classes in general).  Having recently turned in grades, and having already started getting concerned notes from students regarding grades, I'm curious:

For an undergraduate class ---

What's the median grade you give?

Does it vary significantly from year to year?

Does the class size matter?  (Do you tend to give higher average grades in smaller classes?)

Just curious.

11 comments:

Matt Welsh said...

One year in CS61 (intro to computer systems, ~80 students a year) I ended up giving nearly half the class A's because both the midterm and final were takehome exams, so most students did very well -- although the time they took to do the exam varied from 45 minutes to (reportedly) more than 8 hours.

This last term I had one student who got an A- and pestered me and the course staff for weeks to get back every possible point with the hopes of making an A. We took a close look at the grades but couldn't justify bumping the student's grade. I made sure the student understood that making an A- was not going to ruin their academic career. This was someone who had probably never made anything but A's in their entire life.

Harry Lewis said...

I have a related question for the crowd. Does the use of student course evaluations contribute to grade inflation? And do you think the complaints you are getting, even if you deny the appeals, are likely to push up your grade curve next time around?

David Andersen said...

In the new ugrad distributed systems class, the median grade was a B. However, I deliberately shifted the distribution high, not by inflating the curve, but because it was a brand new class -- so when I was designing the projects (and stealing/adapting parts of it from Frans Kaashoek's graduate-level class), I erred on the side of risking making it too easy, rather than introducing errors. The distribution was a little higher than I'd prefer (12 A, 25 B, 5 C, 1 D, 4 F), but it's not completely out of whack. Note that CMU does not report +- grades for CS undergrads.

JeffE said...

Low B-, every year, regardless of class size. (But the undergrad classes I teach are all required; enrollment varies between 80 and 180.)

D. Eppstein said...

When I teach our required undergraduate algorithms class my median is almost always in the B- range. But it's higher for electives, and much higher for graduate courses (effectively, the range of grades for graduate courses here runs from B to A; B- is a failing grade).

Anonymous said...

I know Harvard is supposed to be famous for its grade inflation, but that's not generally the case in my class (and, I think, for our CS classes in general).

What is the grade distribution in your class and for CS classes in general? If it differs from the overall Harvard distribution, does it create any problems for CS students when they graduate?

Michael Mitzenmacher said...

Anon #6: My grade distribution this year had the median at a B (although it was almost a B+). My median grade has slid a bit the last few years; it used to be a low B+ but has moved to a high B. (My grading practice has stayed the same, but the median has changed.)

In general there is always concern because science grades, historically, have been markedly lower than both social science and humanities grades -- at least at Harvard. It's not clear what the solution for this is, or even if it's an actual problem, though it has certainly been widely discussed.

I believe my median grade is somewhat lower than average even for CS. I don't believe this causes problems for CS students when they graduate -- I think their problems are much more dramatic if they don't take my class! However it is something I take in mind.

As DE points out, grad classes are a different beast, since there B- is something like failing.

fortyquestions said...

Just to put some numbers on this, in the natural sciences at Harvard in the academic year 2000, 49% of all undergraduate letter grades (excluding Pass) were A or A-. That compares with 41% in 1990. I have no idea what the numbers are today.

If Harvard's natural sciences are like elsewhere, the numbers should be up to about 54% A or A-. But it's certainly true that applying average results of all institutions to one school is hardly ever a good idea.

Paul Beame said...

Does the use of student course evaluations contribute to grade inflation?

At UW, course evaluation medians are reported in two ways, once using raw scores and the other as "adjusted medians" which are supposed to, in part, re-normalize things based on students' expected grades in a course. (Other parts of the adjustment include whether the course is required or optional.) I have no idea what formula they use but it is really easy to see how a tough midterm increases the gap between the adjusted scores and the raw ones.

And do you think the complaints you are getting, even if you deny the appeals, are likely to push up your grade curve next time around?

No. I have been using the same grade distributions since my first or second year teaching. My typical median is a little below the median of our students' GPAs overall. The median grade goes up or down a little bit based on how well I think the class as a whole has done relative to previous classes.

Because we have restricted admission to our major, the student population is more homogeneous in ability and we seem to get relatively few complaints about grades.

One other way that I may get reduced complaints is that I tend to give a relatively easier midterm and a relatively more difficult final. Students may be less likely to complain when the raw score on their final exam is low. (My goal on the midterm is to be able to identify those students who are struggling relative to the others and on the final to identify those at the top end (though the very bottom students are also identified).

Anonymous said...

I would guess smaller classes and electives would tend to get higher grades. Although this effect only may be very noticeable for really small classes (<10).

AnonBerkeleyite said...

Berkeley EECS's grading guidelines are online:

http://www.eecs.berkeley.edu/Policies/ugrad.grading.shtml

Rough summary: typical GPA for lower-division courses is in the range 2.5-2.9, for upper division courses is in the range 2.7-3.1.