Proposed Project for the MSU Industrial Math Students

Segment Data and Fit Loss Distributions to Claims Information

Project Overview: Enterprise Risk Management (ERM) and loss reserve analysis relies upon a fundamental understanding of the claims reporting process.  In particular, predicting and modeling prospective frequency and severity loss activity for the company is important. One of the key considerations is identifying homogeneous data subsets while taking into account data credibility.  Characteristics to be considered are coverage, size of loss, deductible, limit, state, company, etc.

The goal of this project is to use statistical techniques to identify homogeneous segments of data that provide the best loss distribution fit in to frequency, severity, and pure premium data, taking into consideration the characteristics previously listed.

Project Description:Data files containing exposure, number of claims, and incurred losses will be provided along with associated data characteristics. The first step is to group data into optimal segments to produce the best model fit. The second step is to fit frequency, severity, and pure premium loss distributions to the selected data segments. The third step is to simulate future losses using the loss distributions generated in step 2. Statistical software is required to determine the optimal data segments and loss distributions. Auto-Owners plans to receive regular updates on the project’s progress.

Benefit to Auto-Owners: Interaction with Michigan State Industrial Math Students would enable Auto-Owners to determine more optimal data groupings for loss reserving and learn new techniques to fit loss distributions to the data.  These techniques can be used to enhance Auto-Owners ERM modeling initiatives.

Top of Page