The German Insurance Association (GdV) assumes that one of eight traffic accidents may be a fraud [1]. Furthermore, the annual financial damage is estimated to around 2 billion € [1]. These alarming figures prompted the insurers to develop measures and methods for preventing frauds.

A new approach for preventing frauds in motor-vehicle liability insurance cases is being developed and tested by the Fraunhofer IVI in Dresden. The aim of the method and this paper is the development of a new possibility to create an additional indicator which helps to identify frauds and thereby supports insurers and experts in in their forensic investigation.

The new approach for preventing frauds is based on a large number of geographically referenced accidents recorded by the police and provided to IVI. The statistical comparison between a newly reported accident and accidents with similar accident parameters at the same location predicts a probability measure of the accident. At this stage of the method, parameters like the number of participants, the height of the material damage, the technical condition of the vehicle, the daytime, the kind of accident and the type of accident are compared and give a cumulated index for the single accident in comparison to other accidents. After parametrizing the included variables and defining individual threshold values, this method provides a conclusive indicator for identifying possible frauds.

(EVU-members can download the full article)