Department of Mathematics

Auto-Owners Insurance Company

Project 1:

Proposed Project for the MSU Industrial Math Students

Modeling Policyholder Longevity (Phase 2).

Project Overview

Insurance companies monitor the length of time their policyholder insurance policies are in force or retained. Insurance customers non-renew their policies for various reasons such as competitors offering a better price, policyholders moving to another state, or poor agency service. One of Auto-Owners goals is to develop a long-term relationship, by providing super outstanding service, with the hope that the policyholder will become a lifetime customer. Auto-Owners is interested in learning what current policyholder characteristics are indicative of longer retention rates. By using existing data files, Auto-Owners wants to build a predictive model which projects expected longevity at policy inception.

Project Description

Data files containing insured characteristics such as agency, age, numbers of cars, family structure, etc., will be provided to do initial data mining for relationships and create an expected longevity model. Modeling work will entail determination of significant prediction variables, evaluation of interactions between those variables, and utilization of non-linear statistical modeling techniques. Auto-Owners plans to receive regular updates on the project’s progress.

Top of Page

Auto-Owners Insurance Company

Project 2:

Proposed Project for the MSU Industrial Math Students

Creating Territory Boundaries using Competitor Pricing Data.

Project Overview

As Insurance companies grow and expand into additional states, gaining information to set initial rates is primarily limited to statutory rate filings. All insurance companies split the states into smaller geographic areas or territories and then charge different premiums where the policyholder is located. The goal of this project is to use statistical techniques, such as spatial smoothing or clustering, to determine optimal personal auto territory boundaries by coverage. Examples of coverages would include but not be limited to comprehensive, collision, bodily injury, etc… Competitor insurance company characteristics such as market share, distribution system, and pricing sophistication will need to be taken into consideration.

Project Description

Data files containing policyholder characteristics and associated premiums for 10 insurance companies will be provided at the ZIP code level to complete the territory boundary analysis by coverage. Statistical software is required to determine the optimal ZIP code clusters, territory boundries, and rating factors. Mapping software can be used to increase understandability of the presentation results. Auto-Owners plans to receive regular updates on the project’s progress.

Top of Page