Skip to content

Instantly share code, notes, and snippets.

@fasiha
Last active February 24, 2018 02:13
Show Gist options
  • Star 0 You must be signed in to star a gist
  • Fork 0 You must be signed in to fork a gist
  • Save fasiha/b78ab1f9e31e2c4a4f0a50b995d53a95 to your computer and use it in GitHub Desktop.
Save fasiha/b78ab1f9e31e2c4a4f0a50b995d53a95 to your computer and use it in GitHub Desktop.
Practical training for risk managers

From Aaron Brown, Red-blooded risk: the secret history of Wall Street, Chapter 20

Students must actually do things to learn. There are four things I think any risk management degree should require. …

The first requirement is extensive practical calculation of VaRs, in as many different applications as possible. Any sort of objective data that is available daily can be used: total receipts by the university parking service, inches of rain reported in New York’s Central Park, number of times the letter P appears in the USA Today lead story. Students should have an automated system that posts a VaR before midnight the night before the information becomes available. The system must post all the time, even if input data are missing or the student is away or sick or there is a problem with the university computer system (a box could be available for paper submissions in a power failure). The VaR should pass a standard back-test. For this purpose, an 80 percent VaR is practical compromise between the desire to focus on the tail of the distribution and the need for reliable validation over relatively short periods of time.

The second requirement is experience tracking down numbers. I have in mind here something like the proportion of tractors in the United States that run on gasoline. A good source is statistics cited blithely in news articles. Looking into things like this, students discover definitional issues in the most straightforward-seeming questions. There will be lots of data available, not all directly on point, and much of it wrong. Tracing through to sources and cross-checking in clever ways against objective information teaches both skepticism about all reported data and skills for verifying things.

Third, I believe it’s very important to have experience doing real research in models of risk other than coin flips. That means providing statistical assistance to projects in evolutionary biology, quantum physics, behavioral game theory or any other field that works with risk and progressed beyond the seventeenth century. I don’t think you can appreciate these ideas from lectures and toy examples. There’s a lot of difference between a homework problem and a genuine research issue—one whose answer is both unknown and important. I can see this as either a requirement prior to admission—many students will already have done this—or an integrated part of the program.

Finally, students must have some experience taking risk. Anyone who is not confident that he can walk into a casino and win has no business telling others how to manage risk. And the guy who is confident but wrong is the least qualified. I’m not just talking about beating the games here. That’s just a matter of figuring things out. I mean making real bankroll decisions with real meaningful stakes. That takes knowledge and faith in your knowledge, and discipline.

Okay, I know no university is going to make blackjack card counting, poker, and sports betting into graduation requirements. They’re not going to send their students to Macau with $10,000 and tell them, “Don’t come back until you double it.” But the schools have to find some way to make the computations real, to teach students what it means to bet aggressively on the calculations.

How can we do this, without betting real money and without tainting the program with cards, dice, and sports bets? Here’s a proposal. Each quarter each student is given $1,000 play money to bet at Iowa Electronic Markets. If you’re not familiar with that, it offers real money small bets on various issues like election results, weather, movie grosses, and so on. But I would allow students to bet play money at the mid of the bid and ask prices. This is a huge advantage; if you bet real money you either have to pay the ask or receive the bid, or else work your order, which opens you up to the possibility of adverse selection (that is, no one will hit your bid unless you set it too high). It’s not too hard to win playing this way.

Students would have to double the play money by the end of the quarter, or start over the next quarter with a new play $1,000. One successful doubling and you’re done. But if you haven’t succeeded by graduation, no degree. Okay, no school is going to deny a degree to a student for something as silly as being a bad bettor, even if the whole point of the degree is risk management. But there has to be some meaningful penalty, perhaps an additional paper requirement.

It would take some courage for any school to offer such a program. But that’s okay, because only a risk-taking school should offer it. Who wants a risk management degree awarded by an institution afraid to deviate from staid consensus? Even more, who wants to hire someone who wants that degree?

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment