Friday, October 22, 2010

B-School Bell Curve

Although curving grades is nothing new to most university students, there is something special about how the curve works in most graduate business schools. For starters every class is curved, the guaranteed average for any class is a B+, and in my school 8% of the students will be in the C range. Considering a B average is needed to graduate this seems a little harsh. But the law of probabilities works its magic and most b-school have retention rates above 90%.

For students not used to curving it takes a little getting used to. For instance I knew my intro to finance course in the summer would be just short of a cake walk and to top it off, it had a take home midterm. So with some mild studying the day before and a 8pm to 1am writing period I was feeling confident of an A+ test. And that's I got - on paper. On the curve my low 90 was curved to a B+; it looked like I was in the big leagues now (never fear, the final went swimmingly).Of course there was rumors of "team writing", and I did get a crucial fixed income question wrong, but the fact remained, if I was going to be grabbing the grades I wanted, I needed perfection.

But perfection is hard, so lets put our stats hats on and think about strategy!

The B-School Bell Curve
Using some of the metrics that I have heard thrown around I can approximate my schools curve to something similar to below.
The mean is a B+, it is normally distributed and 8% of the students get a below a B-. There is also a top line percentage quota which I hear is flexible. Regardless because there is a firm tail quota and a fixed mean, the school must have to apply different multipliers to different students grades. In other words if the curve is a bit bi-modal and there is a cluster near the top (as it often is), the penalizer for each grade off perfect would be large to push students down the curve, while weaker scores would receive boosting multipliers. This is where the strategy comes in.

The statistics tell us the curve is going to push you towards the mean whether you are above or below it, so you will be helped in courses you're weak, and hurt in courses your strong (relative to your peers). Thinking about this logically a student who was in 4 courses and got 4 paper grades of 80% would under-perform a student who got 4 paper grades of 60%, 70%, 90%, 100%, assuming their peers performed the same in each class on paper. This is because the 60% and 70% grades are going to receive a curve boost, while all 80% grades will be penalized. The conclusion: be an outlier.

Now we need to drop the assumption that your peers will perform the same in each class, because they won't. With some observation you should still be able to employ the statistical intuition developed above; you need to kill the exams where you think you already have an edge in the class, and this examination-annihilation can occur to some extent at the expense of studying for exams in classes you think you're not top dog. 

The other interesting facet of curves is they implicitly discourage peer-to-peer help, as it is in each persons interest for their peers to bomb tests. Thankfully we are all here for jobs at the end of the day, so there is a equal incentive to help out a classmate - they just might be giving you a job in 5 years. 

Unfortunately no-where in my stats output does it say anything about the usefulness of blogging in midterm season.

No comments:

Post a Comment