Applications of the Offset in Property-Casualty Predictive Modeling

Generalized Linear Model [GLM] theory is a commonly accepted framework for building insurance pricing and scoring models. A helpful feature of the GLM framework is the “offset” option. An offset is a model variable with a known or pre-specified coefficient. This paper presents several sample applications of offsets in property-casualty modeling applications. In addition, we will connect the offset option with more traditional actuarial techniques such as exposure and premium adjustments. A recurring theme of the discussion is that actuarial modelers have at their disposal several conceptually related techniques that can be used to eliminate the impact of variables that (for whatever reason) are not intended for inclusion in a model, despite the fact that they might be correlated with both the target variable and other predictive variables. Examples discussed in this paper include a class plan analysis as well as a tier scoring application. Sample SAS code for fitting GLMs will be provided in the body of the paper.

Keywords: Offset, Residual, Generalized Linear Models, GLM, Predictive Modeling, Ratemaking, SAS