Our society focuses a lot on the individual, yet at the same time we are all constantly part of group processes and group influence. I believe that in the investigation into undesired behavior, which naturally also includes insurance fraud, still pays far too little attention to the influence of the social network.
What is social network analysis?
Our discipline involves charting social networks. This allows us to show the structures of the way we interact with each other: who communicates with whom and about what?
The intensive method to document such networks is by questioning people: with whom do you discuss your work, who do you ask permission for something, who do you ask for advice, with whom do you discuss personal affairs, with whom do you gossip? But you can also analyze networks through existing electronic communication methods such as email, chat and social media.
In the end, the analysis results in a graphic representation of the various points (people) in a network connected by their relationships in thick and thinner lines. This way, you can immediately see who in a group of friends or in a department is popular and influential, both formally and informally. Which people trust each other and where might there be a link between friendship and favoritism with regards to business opportunities or office politics?
Social network analysis for insurers
My expertise is primarily the charting of networks. There are several interesting studies of their application. Networks greatly influence how information spreads, as well as behavior. This also applies to undesirable, socially unacceptable behavior such as insurance fraud. Group behavior can both strengthen and delay how it spreads within a cluster (part of the network where nearly everyone is connected), but it is unlikely that it will be passed on to another cluster.
Building blocks of a network
In insurance, people and policies are the central elements in the network. We can connect these with companies and aliases, but also with vehicles, locations and addresses. And even with events such as making an insurance claim, having an accident or applying for a policy. All these relationships are the obvious building blocks of a network.
An example: we find that there are substantial regional differences in the prices charged by car repair companies. In some locations, for example, car windows are more frequently replaced than repaired. By using social network analysis it is possible to gain insight into the source of this behavior: does it concern specific car repair companies, or is there a connection between the cases and a certain middleman or is there a group of individually insured persons that are connected?
Another real-life example: on Malta there were a large number of suspicious fraud cases involving car damage. As insurers over there exchange their data, the collected data brought to light an informal network. Social network analysis was used to connect the patterns around the used cars, damages, insured persons and locations. Thus, the way these suspicious cases were linked was brought to light. As it turned out, there was a network of friends that faked accidents with each other in order to make claims for the damage while these repairs were not actually carried out. Thanks to the social network analysis, the fraud network was uncovered. What’s more, new claims now can be checked better as well.
Relationships are more important than the individual
If those who work in the prevention of insurance fraud only focus on individuals, they miss a great many opportunities to tackle the problem closer to the root. You also need to look at the structures and interactions and ask yourself how the behavior of individuals is influenced. Make an analysis why something or someone is suspicious, describe this in terms of behavior and practice and check whether it corresponds with the existing data. A statistically sound substantiation of this theoretical assumption allows you to see the network and thus the likely (potential) perpetrators. Based on further investigation and analysis specific evidence can then be collected.
Fraud prevention should focus less on individuals and more on groups and structures. In this way social network analysis is a powerful weapon in the battle against fraud.