Let's talk about means and curves

Discuss comparisons of various school choices and the various metrics that inform them, including rankings, student life, location, etc.
Post Reply
User avatar
Stranger
Posts: 3149
Joined: Fri Jan 26, 2018 8:19 am

Let's talk about means and curves

Post by Stranger » Fri May 17, 2019 12:17 pm

So, there's a lot of reference made to the curved nature of law school classes, but before I went, I really didn't understand exactly how it would work (I had visions of a normal distribution around either a B or C). Some schools, I reckon, do a forced distribution. Some schools (my own included) don't.

Rather than a curve, we have a mandatory mean (with a small allowable variation) of 3.3 (for all 1L classes, upper level classes not based on an exam have a 3.5, and smaller upper level classes have a larger allowed variation). F's don't get counted into the mean. If you were to use a normal distribution around this mandatory mean, you'd wind up with something like this:
A: 6%
A-: 24%
B+: 38%
B: 24%
B-: 6%
(Yes, I'm aware that leaves 2% unaccounted for, but you could drop those pretty much in any grade band and still hit the mean with that distribution.)

But a professor doesn't have to turn in that distribution. An entire class being awarded B+ would meet the mandatory mean's requirements (and supposedly, it's happened). A professor could also choose a bimodal distribution, like this:
A: 30%
A-: 15%
B+: 10%
B: 15%
B-: 30%

There's also the maxim that "every C means two A's" - in order to hit the mean, a single person scoring four steps below the mean indicates that other students have to score four steps above collectively - four A-'s, an A and two A-'s, or two A's. To demonstrate this, let's imagine how the normal distribution would deform if you moved 5% from B+ to C (distributing half the opportunities as bumps from B+ to A and half as B to A-):
A: 11%
A-: 29%
B+: 28%
B-: 19%
B: 6%
C+: 0%
C: 5%

No one wants that C, but the gentler distribution it allows sure would be nice to get dropped into as one of the remaining 95%. The effect of a C is also quite pronounced if we let it deform the bimodal distribution:
A: 30%
A-: 15%
B+: 15%
B: 15%
B-: 20%
C+: 0%
C: 5%

Of course, the underlying reality here is that distributions with a large number of A's also mean that the odds of getting the mean or better shrink. The all-B+ scenario let everyone hit the mean but no-one got an A. The normal distribution has 68% of students at or above mean, but only 6% at A. The bimodal distribution has 30% A's, but only 55% at or above mean. The deformation of the normal distribution kept the high percentage at or above mean, but allowed 11% to get A's. And the deformation of the bimodal distribution got 60% at or above mean while retaining the 30% A's.

Real distributions are more granular than this, but it's illustrative of how the mean (and a curve) work. One more thing that I never thought to look for during my law school application cycle but should have, a table of law school means/medians/curves, has been collected by our good friends over at Wikipedia: https://en.wikipedia.org/wiki/List_of_l ... GPA_curves

Hope this helps.

Post Reply

Who is online

Users browsing this forum: Flietly and 14 guests