Table structure for Hurlimann Benktander
AY | Prem | %paid | %unpaid | Cum Loss | Rind | Rcoll | Zi | Benktander Reserve
Expected Incremental LR (mk) formula
Sum of Inc Loss / Sum prem for each incremental triangle column k
ELR
Sum of mk
% paid
Sum of dev periods to date mk / ELR
R ind - CL equivalent
%unpaid * Cumulative Loss / %paid
R coll - BF equivalent
%unpaid * Prem * ELR
Benktander Zi
%paid
Neuhaus Zi
%paid * ELR
Optimal Zi when Var(U) = Var(BC)
%paid / (%paid + sqrt(%paid))
Benktander Reserve
Zi * Rind + (1-Zi) * Rcoll
Optimal Cred when Var(U) differs from Var(BC)
f = Var(U) / Var(BC)
t_opt = (f - 1 + sqrt((f + 1)*(f -1 + 2p))) / 2
Zopt = p / (p + t_opt)
What does optimal credibility weight minimize?
The variance of credible loss reserves and MSE
Advantages of Hurlimann over Mack
Mack requires an ELR assumption so different actuaries might come up with different reserve estimates; Rcoll does not rely on an a priori value