Capped Loss unavailable in Teng/Perkins?

[lxj034000]

The paper mentioned that Table M is used to estimate CL/L because the capped losses development data are often unavailable. My question is why the capped data is not available if the uncapped data and policy parameters are known (They have to be known to make the PDLD method work)?

First, we can calculate the min and max loss based on the min and max premium because they have the relationship:

P = (BP + CL*LCF) * TM

Then we can cap the loss by the min and max losses.

This means using the (uncapped) loss development data and the parameters, we can derive the capped losses development easily.

Did I miss something?

[hellomath]

On Pg 614, it says

This method is intended to be applied to an aggregate book of business, or large segment of a book of business, rather than at the individual policy level.

On Pg 617, it says

These retro rating parameters may be computed as the average of the sold retro parameters.

It is definitely possible to calculate capped losses on an individual policy basis, but I think the paper is intended for aggregate level. In fact, PDLD using the retro parameter method is calcuated at an aggregate level.

[lxj034000]

So I can sum up the capped losses of the individual policies to get the aggregate capped losses, and them appliy it to the whole book of business?

What confused me was why I use a harder way to get a less accuarate estimate?

[hellomath]

Take a look at Feldblum's review on Pg 281. He explains why using individual data would not be a good idea.