Loss Development Using Credibility

Actuaries use development techniques to estimate future losses. Unfortunately, real data is subject to both random fluctuations and systematic distortions; only in textbooks can we expect smooth, stable development patterns. To correct for this, developed losses are often weighted with a prior estimate to stabilize the results. This paper describes a method that applies credibility directly to the loss development process. The approach appeals to our intuition, but it also has a sound theoretical base. While it requires little more data than the familiar link ratio method and is almost as easy to use, it responds more gracefully to situations in which the data is thin and random fluctuations are severe.