Policyholder Behavior secondo Solvency II

The introduction, in 2016, of the new European insurance framework, Solvency II, has led insurance companies to step up their modelling of all different risks to which they are subject. In particular, in this thesis, I treat the modelling of policyholder behavior, with a particular emphasis on surrender behavior and lapse risk, considered the biggest source of non-financial risks for insurance companies. After describing various type of modelling, I focus on statistical predictive modelling and develop a logistic regression and Random Forest model fitted on empirical data, comparing the effectiveness of the two.