Mack (1994) Flashcards
3 Assumptions of CL method
- Expected incremental losses are proportional to losses reported to date
- Losses in each AY are independent
- Variance of incremental losses is proportional to losses reported to date
Selected LDF for variance assumption alpha squared times c^2
Simple average
Selected LDF for variance assumption alpha squared * c
Weighted average
Selected LDF for variance assumption alpha squared * 1
Least Squares
Confidence interval of reserves, normal
R +/- Z * standard error
Confidence interval of reserves, lognormal
mu = Re^[-sigma^2/2 + Z*sigma] sigma^2 = ln[1 + (se/R)^2]
Graphing residuals (Mack 1994)
Testing assumption 3 – looking for residuals randomly around zero, no downward or upward trend
Calendar year test, mean and variance for n = 1 to 6
1: 0, 0
2: 0.5, 0.25
3: 0.75, 0.188
4: 1.25, 0.438
5: 1.563, 0.370
6: 2.062, 0.62
Test for correlation of adjacent development factors
1) Tk = 1 - 6S/[n(n^2 - 1)] n = # rank-pairs in column 2) Weight Tk(s) by n - 1 3) E[T] = 0 4) Var[T] = 2/(N-1)(N-2)
Formula for residuals
r = (A - E)/sqrt(E)
Graphing c(i,k+1) vs c(i,k)
Testing assumption 1: Linearity of factor