AN EMPIRICAL BAYES APPROACH TO ESTIMATING LOSS RATIOS

The empirical Bayes model is explored as a technique for estimating key financial variables that are meaningful to regulators, policyholders, and security analysts of property-liability insurers. The time series mean and two empirical Bayes models of the loss ratio are evaluated for four lines of business-auto liability, auto physical damage, medical malpractice, and fire. Tests were conducted of the forecasting ability of the three estimators. The empirical Bayes process performed well in the forecasting experiments for all four lines of business, especially medical malpractice. Empirical Bayes methods perform better for small firms, which have higher loss ratio variances than large firms. Forecasting tests using incurred losses suggest that the results might be improved by using a longer time series and recognizing the autoregression of the loss process.