Automobile insurance fraud detection in the age of big data – a systematic and comprehensive literature review
| DOI | https://doi.org/10.1108/JFRC-11-2021-0102 |
| Published date | 08 April 2022 |
| Date | 08 April 2022 |
| Pages | 503-523 |
| Subject Matter | Accounting & finance,Financial risk/company failure,Financial compliance/regulation |
| Author | Botond Benedek,Cristina Ciumas,Bálint Zsolt Nagy |
Automobile insurance fraud
detection in the age of big data –a
systematic and comprehensive
literature review
Botond Benedek,Cristina Ciumas and B
alint Zsolt Nagy
Department of Economics and Business Administration, Babes
,-Bolyai University,
Cluj-Napoca, Romania
Abstract
Purpose –The purpose of this paperis to survey the automobile insurance fraud detectionliterature in the
past 31 years (1990–2021) and present a research agenda that addresses the challenges and opportunities
artificialintelligence and machine learning bring tocar insurance fraud detection.
Design/methodology/approach –Content analysis methodology is used to analyze 46 peer-reviewed
academic papers from 31 journals plus eight conference proceedings to identify their research themes and
detect trends and changes in the automobile insurance fraud detection literature according to content
characteristics.
Findings –This study found that automobileinsurance fraud detection is going through a transformation,
where traditional statistics-based detection methods are replacedby data mining- and artificial intelligence-
based approaches.In this study, it was also noticed that cost-sensitive and hybrid approachesare the up-and-
coming avenuesfor further research.
Practical implications –This paper’sfindings not only highlight the rise and benefits of data
mining- and artificial intelligence-based automobile insurance fraud detection but also highlight the
deficiencies observable in this field such as the lack of cost-sensitive approaches or the absence of
reliable data sets.
Originality/value –This paper offers greater insight into how artificial intelligence and data mining
challenges traditional automobileinsurance fraud detection models and addresses the need to develop new
cost-sensitivefraud detection methods that identify new real-worlddata sets.
Keywords Literature review, Data mining, Automobile insurance fraud detection
Paper type Literature review
1. Introduction
Insurance fraud is an issue that has significantconsequences in both the insurance industry
and everyday life. Fraud can reduce confidence in the industry, destabilize economies and
affect people’s cost of living. The relevance and importance of insurance fraud detection is
underlined by the many reports published related to the topic, which state that insurance
fraud affects 10%–20% of all the contracts even in the most developed countries. For
example, a report from the Insurance Information Institute states that the global cost of
insurance fraud is between $38 and $83bn per year. This means that an averageUS family
has $800–1,400 extra expenses eachyear caused by insurance fraud (Insurance Information
Institute, 2021). If we look at motor insurance fraud, in the USA and Western Europe 7%–
10%, in the Central and Eastern European regions10%–20% and in China 18%–20% of the
policies are affected (Association of British Insurers, 2021;Insurance Information Institute,
2019).
Automobile
insurance
fraud detection
503
Received30 November 2021
Revised2 February 2022
Accepted15 February 2022
Journalof Financial Regulation
andCompliance
Vol.30 No. 4, 2022
pp. 503-523
© Emerald Publishing Limited
1358-1988
DOI 10.1108/JFRC-11-2021-0102
The current issue and full text archive of this journal is available on Emerald Insight at:
https://www.emerald.com/insight/1358-1988.htm
Surveys and studies since 1990 have attemptedto quantify the proportion of fraudulent
policies, identify key fraud indicators and suggested methods and models that can indicate
or forecast the potential fraudulentclaims in the automobile insurance industry; however, as
of today, there is no widely acceptedlist of fraud indicators and the proportion of fraudulent
claims in most countries is unknown. Most of the authors in the academic literature used
their own data sets with different potential fraud indicators and applied them on a self-
developed model. Therefore, in this study we summarize the most important fraud
indicators, and similarly to Ngai et al. (2011),Abdallah et al. (2016) and West and
Bhattacharya (2016),we present a systematic and comprehensive academic literature review
strictly focusing on theautomobile insurance fraud detection. We analyze 46journal articles
and eight conference proceedings on the subject indexed by the Web of Science database
between 1990 and 2021, based on key aspects such as data set, fraud indicators, detection
algorithm or performance of the detectionmethods. Our aim is to provide valuable tools for
both scholars and practitionersof automobile insurance that can be used forfraud detection.
Our study is by no means exhaustive, but we hope that it will certify as a helpful resource
for researchersand practitioners of automobile insurance frauddetection.
The research gap of this study stems from the fact that there is no general and
widely accepted definition for automobile insurance fraud and the existing definitions
do not cover all cases. Decision-makers from insurance companies can use the
research to build up a common, public, representative and up-to-date database for
their special investigation units and for researchers. By doing this, they can still
guarantee the privacy/anonymity of their data/company but at the sam e time they can
learnfromeachother.
The rest of the paper is organized as follows: Section2 presents the difficultiesof the field
of research, and in Sections 3 and 4, we discuss the evolutionof fraud detection methods and
the fraud indicators used in the literature. Section 5 providesdifferent classifications (based
on data set, detection algorithm, performance, etc.) of the existing automobile insurance
fraud detection methods. Finally, Section 6 presents the limitations of the study and future
directions, and Section7 concludes.
2. Difficulties in the research of automobile insurance fraud
Before we proceed to the detailed analysis of the literature, we emphasize certain general
difficulties the authors faced and thatare essential for understanding this literature. First of
all, studies have shown that there is no generally accepted definition of automobile
insurance fraud. The most common definition worldwide is the Massachusetts Regulation
(211 CMR 93.03) which defines fraudulent claims as “claims submitted with the intent of
receiving a larger payment from the insurerthan the amount, if any, to which the claimant is
entitled under the policy, includingclaims for:
nonexistent losses;
amounts in excess of actual losses; or
incidents which the claimant has arranged in an effort to receive an insurance
payment”(Massachusetts Regulation, 1993).
Nevertheless, this definition does not contain other types of fraud, such as the so-called
misrepresentation or the recklessness of the insured person (which is a consequence of
the very existence of insurance). As presented in Wilson (2009), the so-called
“misrepresentation”is less deliberate. “The defrauders may rationalize that there is
nothing wrong with their actions. An example is a parent whose child just received a
JFRC
30,4
504
To continue reading
Request your trial