GLM_Validation4
Hi, in solution for b) "If the current rating plan works well then all quantiles should have the same loss ratio. This is not the case here." According to experience rating, I thought the manual LRs are supposed to increase w e-mod for each quintile. However, the actual LRs here are supposed to flat, is it b/c the LRs are equivalent to the standard LRs in experience rating? Thank you in advance!
Comments
With experience rating we assign each risk a modification which reflects how well we feel the risk fits its manual risk class. When we order the risks by ascending experience modification we should see the manual (actual) loss ratios are increasing. This reflects the best risks are overpriced and the worst risks underpriced. Applying the experience mod should correct this which is why we want the standard loss ratios to be flat.
In this GLM problem though we can think of the mod as always being 1 because we're not using experience rating. Instead we're comparing two rating plans, the current one and the proposed one. If the current plan is performing well, then all risks should have approximately the same loss ratio as measured using the actual loss cost / current premium because we would be appropriately matching rate to risk. Therefore, it shouldn't matter how we group the risks to assess this. Thus, we can choose to group the risks by ordering them by ascending modeled loss ratio = modeled loss cost / current premium.
Since when we do this we see the actual loss ratios aren't all equal, we conclude the current plan isn't better.