A Kenyan insurance fraud scoring system: life insurance

Abstract

The aim of this study was to determine claim characteristics that are correlated with the occurrence of insurance fraud and to, additionally, developing a fraud scoring model. The model's purpose is to provide a probability score on the possibility of a claim being fraudulent. The scope of this study was the Kenyan insurance sector particularly life insurance business with two types of cover: funeral plans and personal accident cover. The data used is honest/paid out claims and simulated fraudulent claims for a period of seven years (2007 - 2013). The findings of this study are that some characteristics such as the claim amount, and the reporting delay are significant indicators of the presence of fraud in a claim. The fraud scoring model developed serves as a handy tool in the detection of fraudulent cases with nearly 90% accuracy depending on the probability threshold chosen by the insurance company.