The question, over at the Risk, Regulation and Reporting area of linkedin.com:
Can you really measure and model risk?
Risk is not a physical quantity and as such doesn’t exist in Nature. It is not subject to any laws of physics. The question, then, is: can it be measured and managed? We think it cannot.
What do you think?
I repost my 2¢:
There isn’t a lot of question that risk CAN be modeled, as mathematical statistics has done that for more than a century. Up to now the modeling has been pretty simple; we know a lot about the risk involved in flipping throwing two dice for example. The person who understands this particular risk model shoots craps better than the guy who doesn’t.
Today’s capital models are much more complicated, of course. And it’s not surprising that some are having well-documented problems.
Keep in mind that in the financial meltdown, some of the models failed, but not most of them. More frequently, the interpreters of the models misunderstood them. The models themselves were made to forecast how likely the company was to run out of cash in the next 24 hours, but the people interpreting the model were using it to forecast how likely the company was to run out of cash in the next six months or a year.
It was like a weather model forecasting sunny skies tomorrow and the weatherman predicting a perfect summer.
Weather forecasting is a useful way to think about capital modeling, because a weatherman bases his prediction on his interpretation of two or three complex computer models. (This of course, is another example of successful risk modeling.) Capital models are really just big weather forecasting models, but they predict financial storms.
Obviously, the models aren’t always right. Sometimes the forecast is for rain but it doesn’t rain. And that’s the case for capital models, too.
However, weather forecasting has gotten better over time. As a schoolkid 40 years ago, I knew that the weatherman might forecast snow overnight, but I still better have my homework done, because like it as not he was wrong. Today, baseball games get canceled well in advance of the first raindrop, because the models have become more reliable.
Capital models will get that good one day, but with a discipline in its infancy, you can expect infantile errors.
That doesn’t mean that we need to wait till the models are perfect before using them any more than we ignored the weatherman 40 years ago. Even recognizing that the weatherman got it wrong a lot, it was helpful to know that it might rain tomorrow, because then you could make backup plans. We need to understand that the models can help us but understand that they have limitations.
The entire discussion is here. (reg. req.)