Grant: Discovery Grants Program - Individual

The overall goal of this research program is to develop novel multivariate statistical methods for modelling non-life insurance data. An insurer's solvency and profitability rely critically on their ability to accurately assess the financial risk associated with all lines of business. While there are several practices that ensure the proper functioning of an insurance company, ratemaking is a vital component. Fundamentally, ratemaking consists of developing pricing models to estimate the costs associated with the risk assumed by the insurer in exchange for the protection provided to the policyholder. Certainly, ratemaking practices have significant impacts on both profit margins and the overall operations of an insurer. Ratemaking involves a comprehensive evaluation of past claims, which are typically multidimensional as they reflect the multi-peril design of insurance products. In order to effectively manage and adequately price the underlying risk, it is imperative that the analysis of claims experience be carried out on a multivariate scale. In analyzing claims, it is insightful to consider a two-part approach wherein a model is defined for both the claim frequency (number of claims) and severity (loss amount per claim). A main goal of this research program is to develop a novel two-part modelling framework involving both a multivariate frequency and multivariate severity model. An important consideration in these models is the association between the claim counts and amounts themselves. Accordingly, an additional layer of dependence must be incorporated into framework to accurately capture the frequency-severity correlation. In order for the models to be appropriate for ratemaking, covariate effects must also be incorporated into the framework. Indeed, premiums are determined as a function of policy-holder characteristics which are predictive of their risk. As such, regression-based versions of the multivariate models will be developed. While profitability is directly related to pricing, customer retention also plays a key role. To this end, an additional facet of the proposal is to explore the use of personalized strategies for establishing optimal customer retention actions (such as a discount) to maximize profitability. The methodology developed through this research program will allow to gain better insights into the complex dependence structures inherent in claims data. Focusing on ratemaking, a main goal of this proposal it to develop novel pricing models which reflect the multi-peril nature of insurance products. In the proposed framework, profitability will be assessed not only in terms of pricing, but also in terms of customer retention through the development of novel personalized retention strategies. The overall modelling framework developed in this program will allow for a more accurate and comprehensive analysis of claims, which is imperative for effective risk management and strategic planning in the insurance industry.