Analysis of insurance attriction with Accelarated failure time model

Abstract

Survival analysis can be defined as a set of statistical methods for analyzing data where the variable of interest is the time until a known event occurs. In the insurance sector, one such event of interest is policy lapse or cancellation. Policy cancellation affects the profitability of an insurance company, and its quantification leads to the optimization of managerial decision-making in the company. The aim of this article is to analyze the impact of individual variables on the duration of life insurance policies and to identify a group of policyholders who cancel their policies the fastest (i.e., are most risky for the insurer). To achieve this goal, we will use parametric regression models, focusing primarily on the accelerated failure time (AFT) model. The advantage of survival analysis methods over other regression models is that they not only provide information on whether a policy is cancelled, but also when this cancellation occurs. The AFT model is one alternative for comparing the survival time of two or more groups of objects. The article compares three types of AFT models - exponential, Weibull, and log-logistic - along with an interpretation of the results obtained.