Abstract
The present paper gives practical examples involving motor car insurance portfolios which we believe provide a natural setting in which collateral data should be used. In passing, the problem of assigning partial credibility is solved and an application of the empirical linearized Bayes estimators to experience rating is discussed.