Academic journal article Journal of Risk and Insurance

A Robust Unsupervised Method for Fraud Rate Estimation

Academic journal article Journal of Risk and Insurance

A Robust Unsupervised Method for Fraud Rate Estimation

Article excerpt

ABSTRACT

If one is interested in managing fraud, one must measure the fraud rate to be able to assess the degree of the problem and the effectiveness of the fraud management technique. This article offers a robust new method for estimating fraud rate, PRIDIT-FRE (PRIDIT-based Fraud Rate Estimation), developed based on PRIDIT, an unsupervised fraud detection method to assess individual claim fraud suspiciousness. PRIDIT-FRE presents the first nonparametric unsupervised estimator of the actual rate of fraud in a population of claims, robust to the bias contained in an audited sample (arising from the quality or individual hubris of an auditor or investigator, or the natural data-gathering process through claims adjusting). PRIDIT-FRE exploits the internal consistency of fraud predictors and makes use of a small audited sample or an unaudited sample only. Using two insurance fraud data sets with different characteristics, we illustrate the effectiveness of PRIDIT-FRE and examine its robustness in varying scenarios.

Introduction

Background and Motivation

Fraud is a fact of social behavior having increasingly important consequences including loss of revenues to businesses, government, and society. Fraud is also expensive, driving up cost for detection and fraud risk reduction. Boyer (2000) quotes a study by the Rand Corporation Institute for Civil Justice estimating that in automobile insurance claims, questionable medical claims added between $13 and $18 billion to the nation's total automobile insurance bill in 1993. More recently, it is estimated that in 2007 claims fraud added between U.S. $4.8 billion to $6.8 billion to automobile injury insurance claims paid, or around 15 percent of total claim payment (Insurance Research Council [IRC], 2008). Viaene and Dedene (2004) detail results of an Insurance Research Council (IRC) and Insurance Services Office (ISO) 2001 study of property and casualty insurers where it was found that half of the respondents felt insurance fraud was "a serious problem."

As a result, active fraud control has gradually become an integrated part of business decision-making processes. Insurance companies must deal with fraud perpetrated by consumers on the firm and spend money on fraud detection and monitoring. A lot of research has focused on the fraud detection efforts, that is, assessing and ranking the fraud suspiciousness of individual claims, currently most of which are parametric and supervised.1 For example, see Derrig and Ostaszewski (1995), Artis, Ayuso, and Guillen (2002), Viaene et al. (2002), and Caudill, Ayuso, and Guillen (2005) for insurance fraud detection. Also see Bolton and Hand (2002) for a review of statistical fraud detection methods. There are also limited unsupervised fraud detection methods available in the literature (cf. Brockett et al., 2002; Ai, Brockett, and Golden, 2009). However, in order to design an appropriate fraud management activity and assess its success, one must have a measure of fraud rate prior to intervention. As it is commonly said, you cannot manage something that you cannot measure. This article is devoted to this critical step in fraud risk management.

The present article investigates the fraud problem from a perspective distinct from detection of individual cases of fraud. It examines the rate of fraud in a population of claim files. Firms and governments need to obtain a fraud rate estimate to assess where and how much they are at risk for fraud (e.g., Caron and Dionne, 1999; Heron and Lie, 2009). In addition, some fraud detection techniques rely on fraud rate as an input to the analysis (e.g., see Durtschi, Hillison, and Pacini, 2004). Ultimately, an automated and continuous fraud monitoring system (Viaene et al., 2007) needs to include fraud rate estimation so that companies and regulators can ascertain if their fraud mitigation efforts are effective for internal control or peer comparison purposes.

However, obtaining a good estimate of fraud rate is very difficult (Caron and Dionne, 1999). …

Search by... Author
Show... All Results Primary Sources Peer-reviewed

Oops!

An unknown error has occurred. Please click the button below to reload the page. If the problem persists, please try again in a little while.