Invest Skeptically

As my former schoolmates (engineering school) will tell you, I don’t pretend to be an expert in math. In fact, I don’t even have a Ph.D. or masters degree in any of the popular risk management backgrounds: math, physics, engineering. But in light of recent experiences in the credit markets I’ll give you my two cents on quantitative risk management. Hopefully there are some mathematical finance readers out there who can offer me their feedback.

Why it’s a valuable management tool

Measuring risk quantitatively is a very important exercise for the purpose of financial risk management. You need to be able to quantify your exposures for the purpose of (among other things): setting capital reserves, reporting risk to regulators and investors, understanding how you’re doing in terms of managing your risk, creating scenarios for contingency planning, and, rating structured financial assets as ‘AAA’.

My naive generalization of the approach

Using a plethora of assumptions about probability distributions, relationships between asset classes, and all the required input parameters for such, you create a correlated loss distribution. In practice this is generated through simulation but in theory it is a nice continuous function. This distribution can tell you things like: how much will be lost with X probability? (you do this to calculate things like VaR or finding the rated attachment points in a CDO).

The real dangers

The problem with any purely quantitative approach, at least my version as outlined above, is that the “real” dangers to any financial operation are typically very lumpy/chunky/discrete. Examples:

1. When liquidity runs out (as with the Canadian conduit market right now, or with LTCM’s collapse), no matter how well positioned you believe you are, the game is over.
2. When a company defaults or goes bankrupt, and you’re holding the credit risk, you’re in big trouble fast. A colleague of mine once taught me that diversification works differently with credit. In your stock portfolio, your winners can make up for your losers, but the best a bond can ever do is pay you back.
3. Changes in the rules (whether from regulators, legislators, or even your suppliers) can completely ruin the economics of a business model. Check the monthly growth of the ABCP market across 2006 versus YTD 2007. This is after the press release from DBRS, DBRS Revises CDO Criteria for Canadian ABCP Issuers (January 19, 2007 ), saying CDO transactions must be, “supported by liquidity facilities from DBRS-approved liquidity providers that contain conditions to draw that are not limited to market disruption and are not dependent on a confirmation of the then-current ratings” (emphasis added).

The conclusion

While modern risk management is a mathemagician’s game, there’s always going to be an important role for people that can understand and make prudent judgements about fundamental risks that can’t be accurately modeled.