Insurance Edge is always interested in new ideas, especially when those techniques have been studied and assessed in a real world scenario, via the Insurance Fraud Bureau. In this article Dr Tim Mitchell and Dr Benny Cheung take a look at fraud and how patterns of human behaviour can be modified. Prevention can sometimes be better than cure.
Dr Tim Mitchell is a Principal at Dectech, having joined in 2007. He received his PhD from Brunel University following degrees in Psychology (BSc) and Computer Science (MSc) at the University of Bristol and Bath University respectively. At Decision Technology he runs behavioural research projects for clients from a variety of industries including finance, telecommunications, retail and utilities.
Dr Benny Cheung is a Director at Dectech, where he has been since 2005. Prior to joining Decision Technology, Benny completed a PhD and a two-year research fellowship in behavioural science at the University of Cambridge. His areas of commercial expertise include retail, energy, financial services, advertising and telecommunications. He also heads the firm’s Brand Practice and personally oversees all related client accounts and internal R&D initiatives.
How Can Insurers Help to Prevent Opportunistic Fraud Attempts?
Opportunistic insurance fraud occurs when otherwise law-abiding citizens are dishonest or exaggerate when making insurance applications or claims in order to get a lower premium or higher claim. It is estimated to cost the insurance industry in the UK around £1 billion every single year 1. For many insurers it is therefore a multi-million-pound problem, and one which causes the premiums of honest customers to be higher than they otherwise would be. In turn, it leads to greater customer dissatisfaction and an erosion of trust between customers and insurers.
Opportunistic fraud differs from fraud perpetrated by organised criminal gangs, but is no less a headache for insurers to deal with. It will come as no surprise, then, that insurers put significant investment into tackling fraud. This investment typically goes into the detection of possible fraud, broadly with three key steps involved 2:
· Detection. This is typically automated with algorithms in place determined from approaches such as machine learning. These do the ‘heavy lifting’ of sifting through the vast volumes of applications, claims and other data to flag anomalies signalling potential fraud, which may then be investigated by a person.
· Investigation. Claims or applications flagged as potentially fraudulent are then typically investigated by a person. This step helps to avoid instances where an application or claim is legitimate and has incorrectly been flagged by the automated system.
· Resolution. Once a suspected case of fraud has been detected and investigated, the final step is for the case to be resolved, which in some cases can mean criminal prosecution. Sufficient evidence is needed for this step to take place, and it is inevitably a time-consuming process.
The issues with detection
Fraud detection and the steps that follow it are an important process in the battle against insurance fraud, but are not without their issues, and more needs to be done 3. The process of detection is not a simple one, with the following downsides:
· Implementation Complexity. Putting in place the various means of detecting fraud is not trivial. From implementing AI and machine learning techniques, to cross-provider data sharing, there are a number of processes that need to fit together seemlessly.
· Time-consuming. As well as having automated processes in places, detection of fraud will typically involve the downstream processes of human investigation and prosecution, mentioned above, in order to be successful. From flagging a potential case of fraud to successful resolution can take time.
· Cost. The process of detection and downstream investigation and prosecution is therefore costly to set up and maintain.
· Reputation Risks. Investigating potentially fraudulent applications and claims can mean interruptions to the service given to the customers in question. Particularly in instances where the application or claim proves to not be fraudulent, this presents some reputation risk to the insurer, in turn harming potential future sales.
There are difficulties, then, in detecting and tackling insurance fraud, and opportunistic fraud is especially hard to detect. This is because the spur-of-the-moment nature of opportunistic fraud (as opposed to the premeditated nature of organised criminal fraud) leaves few clues to the fraud taking place. However, the spur-of-the-moment nature of opportunistic fraud also makes it susceptible to behavioural ‘nudges’, which raises the prospect of whether customers can be influenced to behave more honestly. The insurance industry needs a new weapon in the battle against opportunistic fraud – is prevention the answer?
How to prevent opportunistic insurance fraud
Working with the Insurance Fraud Bureau (IFB), we conducted research to investigate ways of preventing opportunistic insurance fraud. In a carefully designed experiment, we replicated online motor insurance claims and applications processes. With over 12,000 participants, we tested ways of encouraging honesty when completing the forms by inserting behavioural science-inspired messages (behavioural interventions) in front of questions that customers may lie or exaggerate in response to, in order to get a better deal for themselves. The results were impressive, with the behavioural interventions reducing dishonesty by an average of 36% 4.
Replicating these findings in the real world could save the insurance industry hundreds of millions of pounds each year in the UK alone.
In summary, opportunistic insurance fraud is a big problem for the industry and an extremely costly one. Existing means of detecting such fraud, and the increasingly sophisticated ways of doing so are important. However, behavioural science offers an additional approach in the form of using behavioural interventions to prevent opportunistic fraud from happening in the first place. This approach may prove more effective, as well as much simpler, easier, less risky and less costly. As such, prevention is a process insurers need to take seriously.
In order to do so they should first undertake an audit of their existing operational communication platforms to identify where interventions could be applied. They should then field trial these to confirm their effectiveness, before rolling out a full-scale implementation across all platforms.
These steps can be taken alongside traditional detection processes, with the two complementing each other. In short, insurers that leverage the best of both worlds – prevention and detection – are likely to see better outcomes both for their own bottom lines and those of their honest customers.