Adam’s post about rankings raises the general question of when it’s better to rank things using a granular scale and when it’s better to put them into bigger buckets. Granular scales give the appearance of (often false) precision, bigger buckets instead create big cliffs between one bucket and the next. I also share his intuition that law school rankings would be better — if we had them at all — on a bucket model than a scale model.
The same kind of question comes up in law school grading. Some law schools have a scale model with lots of available numerical grades (Chicago has 32). Others have more of a bucket model with 3 or so options (H/P/F, or sometimes H*/H/P/LP/F). Many are in between, with 10 or so options. My intuition is that grades are better on the scale model than the bucket model.
Why would it make sense for these to be different? I can think of two reasons.
1: Law students get lots of grades, but law schools don’t get lots of rankings. When the grades are going to be averaged over a bunch of courses, the resulting average number will already partake of a scale, so we may as well avoid cliffs in the inputs.
2: Trying to get better grades is good; trying to get better rankings is largely wasteful. Under a scale model, most people have a marginal incentive to do slightly better — to study a little more for a slightly better grade, or to lobby a little more for a slightly higher ranking. For grades, this is largely a good thing. The easiest way to get a slightly better grade is to study a little more and learn a little more law, which is valuable. For rankings, this is largely a bad thing. Much of what schools do in pursuit of slightly better rankings is zero-sum or rent-seeking activity.
This is just impressionistic, and I’m sure there are exceptions too. And of course, there may be times it makes sense for law school grades to instead switch to the ultimate bucket model — pass/fail. That might be a topic for next week.