Thursday, April 01, 2010
Can risk be modeled?
If you are talking about Life Insurance and the risk of death, more than one hundred years of experience has proven that, short of a war or pandemic, mortality is predictable. Law of large number will reduce the variability and you can model to a degree of confidence. If you are talking about probability of default on debt, then you are talking about modeling people's behavior that will be driven by incentives and the performance of the entire economy, which as many under currents. Everyone probably still have the credit crisis and financial panic of 2008 in mind. One of the source of the panic is that financial engineers assume mortgage default rate will behave about the same as the past 10 years, and priced the CDO/ CDS accordingly. When law of large number does not stabilize the default rate and it becomes clear that trillions of derivative assets are priced on wrong assumptions, the run starts. The disaster was a magnificent view. You put your trust on the model you don't fully understand at your own peril. And some short sellers are wise enough to profit handsomely on it. The lesson I learned is that I should examine the assumptions modelers use carefully, before embracing "financial weapon of mass destruction" (Buffett) as a panacea to reduce funding cost.