Could you please explain how the following below help validate the model? To me model validation help evaluate whether the chosen model is appropriate.
- Comparing tail losses with market price reinsurance
- Use global data to complement Canadian limited earthquake data
Comments
When utilising an EQ model, the output that is usually desired would be the tail losses (i.e. the 1in 200, 1in 500, 1 in 1000 year events, etc.) These losses are very large and particularly sensitive to changes in assumptions/model inaccuracies/defects.
A reinsurer who would reinsurance very high layers of EQ coverage would also be using some form of EQ model to price their coverage. By comparing results of our own model to the market price of reinsurance which is basically almost a weighted average of all reinsurer's indicated earthquake losses (ignoring target profit and operating expenses), you can get a good sense check as to the reasonability of your own model. For example, if your model is indicating that it cost 20M to self-insure your EQ exposure while you can buy reinsurance for 10M, there is probably something wrong with your model.
This is just credibility weighing and looking at losses from other regions to help provide more colour to possible Canadian losses