By Dr Doron Samuell
The most appealing aspect of the Masters in Behavioural Science program for me, was the opportunity to undertake research that was relevant, impactful and could contribute to the evolving discipline of behavioural economics. I was fortunate to be working with a progressive financial services company who were willing to join me on this journey.
My research was a field study to determine the impact of behavioural economics interventions on real customers applying for life and disability insurance. I wanted to see if we could nudge customers, at the time of underwriting, into better meeting their disclosure obligations. Insurers are reliant on the truthful disclosures of customers. Customers who engage in false negative disclosures may get lower premiums but run the risk of having their policies voided if the inaccurate disclosure is identified. Correct disclosures allows accurate pricing and security of contract for the customer.
Around 2000 participants were randomly allocated to a control or one of three treatments and their disclosures were captured and statistically analysed. For confidentiality reasons, I am limited to stating that the treatments involved social norms, loss aversion and moral priming. With two of our treatments we achieved significant elevations in self-reported smoking, illicit substance use, medical disclosures and weight. It was a positive experience for the company who participated in the research, generated no customer complaints and helped customers meet their obligations.
The insurer who I partnered with gave me permission to do extensive exploratory data analysis (EDA) on their existing customer data. I compared the prevalence of customer disclosures with known population data; the biggest differences were found in community levels of smoking with customer self-reported smoking. On that basis, I chose smoking as the most appropriate dependent variable, as my intuition was that it had the highest levels of false-negative self-disclosure. The EDA also gave me a strong basis on which to make power calculations so that I could make an informed sample-size estimate.
Graph comparing prevalence of smokers in insured and general populations:
The next step was to choose the appropriate behavioural interventions. When I looked at the behavioural literature, there was no precedent for this kind of intervention, so I drew on research from similar domains. There were many practical issues to take in to account as I was working with real customers and a client who has commercial and reputational considerations. This meant that I needed to be mindful of what could be effective and acceptable in this specific context. What may work in a lab or university environment, such as Ariely’s recitation of the Ten Commandments, was clearly inappropriate in this context. We went through an iterative process involving several insurer internal stakeholders before we chose the treatments.
Introducing treatments into an underwriting process required training of call-centre staff and system changes for the capture and analysis of the data. I was provided with weekly data batches to analyse and I closely liaised with the insurer about any unintended consequences.
The experiment itself took only three months to execute, but the planning and EDA took a further 6 months. Getting a co-operative research partner is a process that involves trust, legal and commercial considerations; it requires a passionate and effective internal champion to navigate through a large, complex organisation. It was worth the effort, however, in addition to better disclosures, the company was left with a robust scientific basis for further research.
This blog post was written by Dr Doron Samuell as a summary of the research undertaken for their dissertation as part of the Executive MSc in Behavioural Science at LSE, 2015-16. Doron is currently researching Risk Modelling at Sydney University. Follow him on linkedin https://www.linkedin.com/in/doronsamuell