Risk modeling

Perhaps one contributing factor to the current economic crisis is that the models used by risk managers and employees did not model the correct risk. With so many methods to model one thing, you could possibly get two different outcomes by using different models. Considering a continuous number of methods and probability distributions, there is a good chance that the model you are implementing does not match the distribution. Even though the choice of method depends significantly on the amount and type of historical data available, each method has its advantages and disadvantages. Each method also requires varying analytical skill and experience.

There are 3 broad categories of modeling methods presented by the CAS:
Methods based primarily on the analysis of historical data
Methods based on a combination of historical data and expert input
Methods based primarily on expert input

I believe that models that represent both historical data and expert input should be more accurate. You don’t always want to base assumptions on models, since they don’t always represent real life situations, and if they did we wouldn’t be in an economic crisis.

One process that falls into this category is stochastic processes. This process expresses the difference in the value of a variable at time t and the value one time period later (t+1). This process is probably one of the more accurate processes since its probabilities consider the past, but future probabilities depend on the present state. Thus, stochastic processes predict random change looking into the future. However, there could be thousands of possible outcomes, which may get complicated to decipher.


Links:
http://www.ucop.edu/riskmgt/erm/documents/overview.pdf