Peter D. England and Steven Haberman
Abstract
This paper describes a complete framework for the statistical modeling of excess mortality, with particular reference to the experience of insured, impaired lives. The principal measure of excess mortality considered is the standardized mortality ratio. The modeling approach, based on the theory of generalized linear models, allows us to build models containing several explanatory variables. The statistical significance of these variables can be tested, and the effect of interactions between the variables can be assessed rigorously. The paper uses data drawn from the extensive, continuing investigation into the mortality of insured, impaired lives conducted by the Prudential Assurance Company (UK). The methodology has close connections with the traditional actuarial approach to the measurement of excess mortality and can be regarded as a generalization of this traditional approach.
Key words and phrases: impaired lives, generalized linear models, multiplicative hazard
Peter D. England