Hello, I have a few questions on April 2019's paper:
Q6(ii): The distribution of the number of claims for an individual is given, and the second table gives the aggregate claim amount for a risk class.
- In this case, using the Buhlmann-Straub model, we obtain the credibility estimator by weighting Beta (found using the given number of claims distribution) and the historical average claim amount per individual (from the tables given).
- Wouldn't this make the estimator inconsistent somehow? Since Beta is based on the number of claims but the other part of the credibility weight is applied on the average claim amount.
Q8(i): Liability claims development to ultimate
a. Why do we need to use BF instead for 2017, when the ultimate loss ratio under chain ladder is 26%; comparable to 2016's ratio?
b. It is mentioned in the solutions that there is an increasing trend for liability claims. However, aren't we essentially 'removing' that trend by imposing the 19% IELR on 2017 and 2018 claims using BF?
Q6(ii): The distribution of the number of claims for an individual is given, and the second table gives the aggregate claim amount for a risk class.
- In this case, using the Buhlmann-Straub model, we obtain the credibility estimator by weighting Beta (found using the given number of claims distribution) and the historical average claim amount per individual (from the tables given).
- Wouldn't this make the estimator inconsistent somehow? Since Beta is based on the number of claims but the other part of the credibility weight is applied on the average claim amount.
Q8(i): Liability claims development to ultimate
a. Why do we need to use BF instead for 2017, when the ultimate loss ratio under chain ladder is 26%; comparable to 2016's ratio?
b. It is mentioned in the solutions that there is an increasing trend for liability claims. However, aren't we essentially 'removing' that trend by imposing the 19% IELR on 2017 and 2018 claims using BF?