Application of Propensity Score Matching in Predicting Price Elasticity

Specialeforsvar: Caroline Ørgaard Dalager

Titel: Application of Propensity Score Matching in Predicting Price Elasticity

Abstract: In non-life insurance, understanding customer behavior and predicting price elasticity are crucial. The profitability of a portfolio can be adversely affected when rate increases cause policy lapses as customers migrate to other insurance companies. Conventional methods to calculate price elasticity from most insurance databases often prove inadequate due to the observational rather than experimental nature of the data. Specifically, historical rate changes are typically a result of risk-based pricing, and the specific rate change to which a customer is exposed is a deterministic function of their observed covariates.
In this master thesis, we propose an application of propensity score matching techniques to address this complexity. We present a causal inference framework to predict price elasticity within the non-life insurance industry. The foundation of our approach is a thorough exploration of the theoretical aspects of causal inference and matching methodologies. Our findings illuminate the
extent to which the data base can support causal effects from rate changes. We implement a simulation study to investigate the effectiveness of our methodology, further applying it to real-world data from Codan Forsikring. The results, while illuminating, also indicate the necessity for more prudent data selection and hypotheses closely aligned with business processes.
Our approach can lead to more reliable estimation of individual policyholder’s price-elasticity functions, acknowledging that the causal effect of a rate change varies across individuals. Accurate rate change choices at the individual subject level, informed by their characteristics, are essential to optimizing the overall expected profitability of the portfolio. Despite certain limitations and unexpected outcomes, this thesis strive to underscores the potential benefits of using matching techniques in predicting price elasticity. It serves as a stepping-stone for future research aimed at more accurate and applicable price elasticity predictions within the non-life insurance industry.

Vejleder: Munir Hiabu
Censor:    Martin Møller Svensson