Till innehåll på sidan

Lily Melmoth: Predicting mortality through survival models

MSc Thesis Presentation

Tid: On 2020-09-09 kl 09.45

Föreläsare: Lily Melmoth

Plats: Zoom, meeting ID: 901405086

Ämnesområde: Actuarial mathematics

Handledare: Mathias Lindholm

Abstract

The insurance industry needs to be able to forecast mortality in order to value insurance and pension policies. Survival modelling predicts time until an event (for example death) and is therefore well suited to predicting mortality, however it is not a commonly used method in the UK. This paper describes and analyses common sur- vival models, highlighting practical considerations that should be made when fitting survival models. The paper is aimed at, but not limited to, predicting mortality and should enable the reader to fit survival models. The advantages of survival modelling over commonly used methods in the UK, such as mortality tables for a general population, include:
• Ability to predict mortality for any specific population, allowing insurers to reflect their own insured population rather than a general one
• Ability to incorporate explanatory effects from multiple variables, without having to create separate models or tables for males and females, different socioeconomic indicators or any other type of variable
While tools are available to assess how well survival models predict data, it can be difficult to fully assess the predictive ability due data containing incomplete observations. Survival models also require an adjustment in order to reflect longevity trends; a few options are discussed in this paper.

Innehållsansvarig:webmaster@math.kth.se
Tillhör: Institutionen för matematik
Senast ändrad: 2020-09-02