Couret and Venter Flashcards
Mapping type of claims to variable letter
Fatal: V
PT: W
Major: X
Minor: Y
Types of letters, Couret and Venter
Capital letters: computed from sample data
Lowercase: estimates of the population means
Multi-dimensional credibility
Matrix equation for vi
Expected excess loss formula
3 options to test the results, Couret and Venter
- Hazard group injury type rations (Vh)
- Raw sample data injury type ratios (Vi)
- Injury type ratios resulting from credibility procedures (viest)
Initial SSE test, Couret and Venter
Initial SSE test results, Couret and Venter
Credibility procedure best, but not much of an improvement over hazard group ratios
Attributed to the fact that class data by year is volatile, estimators are designed to fit that data
True improvement of credibility procedure masked
Quintiles Test procedure
- Sort injury type relativities in increasing order
- Group into quintiles so that size roughly equal to the same number of TT claims
- Calculate weighted average injury type relativity across all classes within each quantile and within the hazard group
- Use results from step 3 to calculate SSE for each of the 3 methods
SSE calculation
Hazard Group Method
Raw Sample Data Method
Credibility Procedure Method
Credibility procedure yields substantial reduction in SSE
Why credibility procedure does not show an improvement for Hazard Group A
Group is already very homogenous, so little value in trying to create subgroups to improve estimation accuracy