Couret - Multi-Dimentional Credibility

Question 1

Describe the role of correlation between classes in the multi-dimentional credibility estimation method

Answer 1

1. serious injury types have low frequency
2. serious injury types are correlated. F, PT and Major claims are produces by similar physical circumstances
3. Multi-dimentional credibility uses correlation to improve the estimate of each type’s claim frequency

Question 2

What two tests were conducted to validate the credibility estimates?

Answer 2

1. Individual class SSE
new method showed slight reduction in SSE over using raw average

2. Ranked portfolio of classes SSE (Quintile Testing)
   significant improvement in SSE reduction

Question 3

How does this method differ from historical methods?

Answer 3

Historically the fraction of costs excess of retentions has been calculated for large groups of classes (GHs), not individual classes. C&V show that estimation by class does produce additional information.

Question 4

What is the purpose of holdout sample.

Answer 4

allows the model to be tested on real data which it hasn’t been fitted to.

Question 5

Justify using odd/even years as hold out

Answer 5

any emerging trends in the data will likely be present in both even and odd years