Wednesday, September 7, 2016

Grading on a Curve?

Dan Drezner does a nice job of explaining grading on a curve and how it applies to the election, but I don't think it was clear enough.  What needs to be clear?

When one speaks of a normal curve or distribution, one is usually thinking of a relatively standard bell curve:
From http://www.hrwale.com/how-to-create-a-bell-curve-chart/
The essential assumption for grading (and for other stuff) is that there are some people who are better at the task than most, and that some people are worse than most.  One can set the curve--how one moves up or down the grade scale, so that the center is a C (the good old days that most people don't really miss) or a B or whatever. 

The thing about grading on a curve is that there is still a distribution with some high, some medium and some low.  Grading on a curve does turn a mediocre grade into a somewhat less mediocre grade.  With a sharp policy wonk like Hillary Clinton in the race, Trump will never get an A as she sets the top of the curve.  The real question is where does curving really put Trump?  And my problem with his analogy is this: no matter how much curving is done in a typical class, an F student who does not attend class, does not do the reading, does not complete assignments, fails the tests, and occasionally punches the teacher does not get curved up to a B or B+ or a C-.  They stay an F.  Because a hyper-low performer sets the bottom of the curve just as a hyper high performer sets the top. 

So, the press really has been doing something to keep Trump in a state of false equivalence with Clinton, but this pedantic prof says: it ain't curving the grade.  It is something else alright, but a student at the bottom of the class remains there, curve or no curve. 




No comments: