D.B.A. Baker College (Michigan), 2016.
Specialization: Business administration
Exploring the potential association between key insurance indicators and automobile fraudulent claims
121 pages. UMI #: 10146768
Citation, Abstract & Full text in ProQuest Dissertations & Theses Database
Automobile insurance fraud is a significant problem affecting society, insurance companies, and consumers with costs of detecting fraudulent claims while delaying payment of legitimate claims. One approach used by insurance companies is to scan incoming claims to predict potentially fraudulent claims, which are submitted to Special Investigation Unit (SIU) for further examination, while continuing to process legitimate claims. However, a majority of claims submitted to SIU are found to be legitimate. One emerging technique to predict fraud is the use of insurance indicators, which are suspicious events related to a claim, but there has been limited research on whether these indicators are any better at predicting fraudulent claims. Using binary logistic regression analysis, Wilson determined there was an association between some insurance indicators and fraud. His study involved a small number of insurance indicators. The purpose of the current mixed-methods study was to extend Wilson’s work to wider list of insurance indicators and other sources of referrals. Theoretical support for this study was based on Wilson’s analysis and the work of Cressey, Wolfe, and Hermanson on the fraud triangle/diamond. The researcher worked with a major insurance company to analyze a sample of claims forwarded to SIU because of insurance indicators. Chi square analysis was used to test the null hypotheses, and Lamda statistic was used to determine predictive values. Results showed that both insurance indicators and manual referrals were associated with fraudulent claims, but manual referrals over performed the insurance indicators. The study confirms the importance of exploring the knowledge base of key individuals reviewing claims to capture their knowledge.