Insurance Claims - Model with Compound Distribution
This model shows how to get the uncertainty of Number of Claims and Cost Impact per claim using the RiskCompound functionality available both in @RISK and also in PCR (Palisade Custom Runtime).
The model shown here provides a better way to capture the uncertainty of this type of calculations as it provides a sample for the number of events occurred and instead of having one single sample for calculating the product of the frequency and impact, it gets multiple, different samples from the impact distribution according to the number of impacts sampled in the frequency distribution, and sum them together.
In a model where the product is calculated, a typical random amount was multiplied by the number of claims, and the total was completely dependent on this one random amount. If it was abnormally large or small, so was the total.
But in a RiskCompound model, the total isn't affected by a single random amount; it is the sum of many random amounts, some of which will be small and some large. So the total is much less likely to be extreme.