Retention Modeling with Survival Analysis

Abstract

We utilize methods from a branch of statistics called survival analysis in order to assess factors that influence the cancellation of house fire and mandatory vehicle insurance policies. The data was obtained from an insurance company in Iceland and is based on insurance policies that were issued whithin a three year horizon, namely in 2013, 2014 and 2015. With statistical methods from survival analysis we were able to detect when policies are being cancelled, not just whether they were cancelled. A cox proportional hazard model was used to analyse factors that were predictive in terms of insurance policy cancellation. We provide a comprehensive review on the literature of insurance, customer relationship management as well as the practical applications of survival analysis when it comes to insurance policy cancellation.