CHAPTER 13Bayesian Networks for Risk Measurement


We will use a nonfinancial example to illustrate the application of Bayesian networks for risk management. We consider a risk manager addressing the road accident risk for the company's truck fleet. He wants to represent the three parameters (exposure, occurrence, and impact) of this risk using a Bayesian network.

For this particular risk, the appropriate exposure measurement is the distance covered by the fleet during one period of time. Each kilometer covered by a truck is exposed to the occurrence of a road accident. The number of trucks would not be an appropriate measurement for exposure, since buying a new truck does not increase the risk of accident until the truck is on the road. Similarly, a driver contributes to the risk exposure only when he is at the wheel.

This measure of exposure is generally used by national authorities, such as the Observatoire national interministériel de sécurité routière1 in France.

The number of kilometers covered per year for the considered company is the selected measurement for exposure. In a prospective risk management point of view, this is clearly a random variable since we cannot know in advance the number of kilometers that would be covered next year.

Using the distance covered as a measure of exposure is only acceptable because we will analyse further the risk for each individual kilometer. We will examine whether this kilometer is covered ...

Get Operational Risk Modeling in Financial Services now with O’Reilly online learning.

O’Reilly members experience live online training, plus books, videos, and digital content from 200+ publishers.