Regression Methods of Loss Reserving -Time Varying Models

This session is a sequel to Traditional Regression Methods in Loss Reserving. That session presented regression methods for modeling loss development, which assume the parameters to be estimated are constant over time. However, insurance data is impacted by the influence of a number of economic and social factors whose effects often vary over time. In this session, models of loss development with parameters that vary over time will be presented. Traditional applications which use weighted averages are an example of a development model with time varying parameters. Methods for explicitly including time varying assumptions will be presented and illustrated with realistic applications. The panelists will then look at time variant versions of the smooth chain ladder and other parameterized models.