Efficient Estimators Through Data Segmentation

Statutory Statement of Accounting Principles 55 of the NAIC states that “management shall record its best estimate of its liability for unpaid claims, unpaid losses and loss/claim adjustment expenses”. One of the considerations in analyzing the reasonability of these estimates will be the amount of variability in the estimates. Variability of reserve estimates can be reduced through data segmentation. From a statistical standpoint, among unbiased estimators, an efficient estimator is defined as an estimator with the same mean as another estimator, but a lower variance. The best estimator would have the same mean as other estimators, but the least variance overall. The efficiency of various reserve estimators (alternative data segmentations) can be compared by projecting the variance of the one year runoff of the loss and loss/claim adjustment expense data for each data segmentation alternative. A model has been developed to compare the efficiency of various reserve estimators and determine the impact of data segmentation on the variability of loss reserving estimates.