Individual Risk Predictive Modeling
Edward W. (Jed) Frees
University of Wisconsin - Madison
In this talk, I consider individual risk models that are commonly encountered in short-term insurance policies. In this context, “individual risk modeling” refers to a model that is specific to a policy, policyholder or claim, in contrast to a portfolio. With ever-increasing access to data at a micro-level and computing capabilities to handle large datasets, individual risk modeling is taking on a greater role in managing risks. To provide grounding, I draw on examples from four practice areas: homeowners, personal automobile, health and term life insurance.
One objective of this presentation is to underscore the rich set of options that an analyst has when analyzing short-term coverages. It is well-known that when working with individual risks, principles of competition and adverse selection dictate that industry analysts employ many covariates to explain outcomes of interest. This presentation focuses on outcomes that may: (1) come in two parts, corresponding to a frequency and severity, (2) be multivariate, such as different cause of loss or outcome type, or (3) be longitudinal, such as repeated contracts or a process of payments. Further complicating model specification is that insurance outcomes are often skewed, heavy-tailed and censored.
Another objective is to emphasize that predictive modeling can be used to meet important risk management goals. Risk management goals include pricing risks in a competitive market, managing claims, detecting fraud and capital allocation. Individual risk, or “micro-level,” modeling can help to understand the distribution of a risk as well as a portfolio of risks. Knowledge of a distribution of outcomes leads to informed decision-making in meeting risk management goals.
Date received: March 1, 2010
Copyright © 2010 by the author(s). The author(s) of this work and the organizers of the conference have granted their consent to include this abstract in Topology Atlas. Document # cbak-23.