Methods of Creating Model Point Files for Actuarial models
Method 1, Seriatim data MPF. Each model point represents one data point. Hence the MPFs are usually huge, and may be only used in valuation. Long model runtimes make it impractical for use in any ALM projections.
Method 2, Grouping based on selected criteria. Multiple data points are group into one cell based on the similarity of data characteristics. Commonly used criterion is the issue age. For example, only modeling issue ages of 15, 25, 35 and so on. Grouped MPFs generally use a weighted averaged issue month or all July issue month. The other criteria include premium mode, risk class, second life age and others. To further compress the MPFs, it is normal to set a minimum of policies or face amount for each model point.
Method 3, Result driven. Data points are grouped based on the results from the original or seriatim MPF. Generally, this method achieves higher compression ratio than other methods and also be able to produce acceptable results for the given metric. The drawback is that it needs to create and run a different model repetitively. Moreover, additional tools need to be deployed to create the new MPF.
Recommendations for MPF creation. Actuaries need to balance efficiency and accuracy. The MPFs with the highest compression ratio that also produce acceptable results should be used. For different projections, different sets of MPF should be used in small-scale or large-scale runs.