Left truncation in the As-If Markov Model

Specialeforsvar: Mads Kjøller Pedersen

Titel: Left truncation in the As-If Markov Model

Abstract: The traditional approach in multi-state life insurance is Markov modeling, where the state of the insured is cast as a Markov jump process. These models might, however, not reflect reality if the Markov assumption is violated. To address this other models have been proposed including the as-if Markov models. In this thesis the as-if Markov setting is presented and examined. The main contribution of this thesis is the implementation of left truncation for non-parametric estimation in the as-if-Markov setting. Specifically, three approaches to implementing left truncation are derived and compared and it is discovered how landmark estimation in the as-if Markov setting allows for effective estimation utilizing data from before the time of entering the portfolio. Numerical studies are carried out to compare the performance of the as-if Markov model to the classic Markov model and, further, to examine the effectiveness of the implementation of left truncation in the as-if Markov setting.

Vejleder:  Christian Furrer
Censor:    Lars Frederik Brandt Henriksen,, PFA