Friday, April 14, 2006

Exam Grades and Curves

I received a very good anonymous question in response to my post yesterday ("Of Politics and Law Schools"). The question was as follows:
Finally a blog by a law school professor. Do you grade according to a curve? Are you forced to because of your school? And if yes, do you find a forced curve is the best way to judge student performance?

In addition to brilliantly stroking my ego ("Finally a blog by a law school professor"!), this comment raises excellent questions about the use of grading curves in law schools, which I will try to answer below. Final exam season is upon us, after all, so this is a timely topic.

But first, in the interest of full disclosure I should say that there are a fair number of other law profs out there who blog. Some of these blogs are linked to on my blog. However, my focus is a bit different than most other law prof blogs: I focus on student and junior practitioner issues.

Anyway, back to the questions at hand.

Do I grade according to a curve? Does my school require me to use a curve? My school does not require the use of a curve, so this is left to my discretion. In small classes I do not grade according to a curve. In large classes I do. But here's the interesting thing about curves: in my classes at least, curves typically have no effect on the distribution of grades, or they work in students' favor.

Let me explain that a bit. In my experience, a properly constructed law school exam leads to a bell curve distribution of grades. A good exam will have easy issues on it, hard issues, and some issues in between. The easy parts show me who understands the most basic aspects of the course. The middling-hard aspects are mastered by fewer students, and this represents the middle of the class. And the hard stuff allows the top students in the class to shine.

So yes, I sometimes grade using a curve of my own devising. But it is usually grade-neutral or actually helps students (i.e., not everyone gets a C).

Am I in favor of a forced curve as the "best" way to judge student performance? The "best" way? Let's limit the question to whether I am in favor of its use in large classes, since there are lot of other ways students should be evaluated (e.g., writing papers, trial and appellate advocacy programs, clinics, etc.). I'm actually still undecided about the mandatory use of curves in large classes. One argument in favor of forced curves is that they help to harmonize grades across different sections of the same class. If you have a tough-grading professor for Contracts and the other section has a softie, why is that fair? Wouldn't it be better to have some forced consistency, so that the tough professor has to give a minimum number of As, and the easy grader can't give give everyone As?

On the other hand, it is a bit of a shame for a good exam paper to get bumped down by a curve. But again, in my experience a well-constructed exam largely avoids that problem. So perhaps I lean slightly in favor of mandatory curves for large classes. But I could be persuaded otherwise.

Any comments or thoughts?


Jason Wojciechowski said...

I think your argument from a consistency standpoint is the most important. Most law schools are entirely grade-driven. Graduation honors, law review, other law journals, moot court are great things to have on your resume, and (for the latter three) great experiences to have in law school. For some section to "luck out" and have a few professors who skew toward giving more A's would be completely unfair and would invite all manner of complaints ... even lawsuits?

biff said...

At my school, after the first year, student can play games by choosing seminar classes that are not on the curve. But the competition is over by then. The grades that truly matter are your first year grades.

Michael said...

There are a few problems with your justification of curving grades.
First, you said that one argument for curving grades is that they help harmonize grades across different sections of the same class; limiting the softie from giving too many A’s. However, the A is still 4.0 points, and the B is still 3.0 points. In a hard class, a student getting a B can be scoring higher/doing better than the person scoring the A.
Person A scores 85% with a mean of 76%, standard deviation of 3.0% (z = 3)
Person B scores 90.0% with a mean of 88.5%, standard deviation of 1.5% (z = 1)
Person A did outstanding, while person B did average. If we give both of them an A, then Person B did significantly worse than Person A, but got the same grade, while if we give Person B a C, then even though he mastered 90% of the material, he gets a C, just like the person in Person A’s class with a 73 to 79%!
Compound this over several classes and you get numbers that are simply not comparable.
In smaller classes this is even worse. Example: in legal research and writing, one person received a B by getting 74.85% while the other person got a C+ by getting 74.10%. The difference was 1 point on the final paper, or 3 points on the first draft.
My gripe with law professors is that they believe a properly constructed test leads to a bell curve distribution. This idea is nonsense. To get construct such an exam, law professors insert a weird “issue.” Those that catch it, end up in the top half of the class, those that don’t, end up in the bottom half. B+’s see the issue but write a bad exam, B’s don’t see the issue and write a good exam.
The only fair way to rank students is to have them take a standard test that none of the professor’s created; after all that is what will happen when they take the bar exam. Curving that will give you a ranking; no need to give out grades at all.