A nonparametric approach to analyzing operational risk with an application to insurance fraud

Catalina Bolance, Mercedes Ayuso and Montserrat Guillen

Abstract

ABSTRACT

Nonparametric methods combine most of the advantages of parametric alternatives for assessing risk. We illustrate a new method for addressing quantile estimation with no distribution assumptions. In our case study, the operational risk related to external fraud in an automobile insurance database is analyzed. Alternative antifraud strategies followed by an insurance company are evaluated. Scenarios of plausible fraud detection rates are presented to assess the impact on operational risk reduction. Finally, we discuss the evidence that fraud detection systems effectively mitigate operational risk in the context of non-life insurance.