What benefits does the growing datafication of society present to the Insurance Industry, and what if any are risks to society ?

The emergence of modern technologies like artificial intelligence, internet of things and analytics has fundamentally transformed the role of data in the Insurance Industry. Huge investments are made in the digitalization of the processes and products by globally recognised insurers. While the advances of big data technologies in insurance will lead to greater economic and societal benefits but at the same time, it might impose new threats on public interest like customer privacy, individualization of insurance and products. While many insurers were slow to adopt analytical practices in the early years, a large part of the industry has now started assimilating new forms of data-intensive analytics into their core business operations. Consumers, organisations and policymakers are facing a sensitive trade-off for striking balance between the privacy protection of consumers and still allowing space for innovation. The essay aims to emphasise on the key economic and societal benefits of datafication and the regulatory risks involved in the same with an emphasis on a judicious trade-off to reap the benefits of datafication in Insurance Industry.

Data Analysis has been pivotal in transforming the insurance industry from intuitive decision making to rational & fact based decision making. Earlier Insurers mainly relied on the information obtained by the policymakers while they subscribe to a particular policy, so that the policyholders can be divided into specific risk groups. For instance, in case of car insurance, premium and risk class is derived from the age, car type and accident history. Insurers over the past few decades has started collecting data from the third party vendors the objective being the same, to understand the risk and provide policy holders the best of the services and products. With technological advances, artificial intelligence, analytics and internet of things has dynamically changed the role of data in Insurance Industry, the interconnected digital infrastructure platform continually generates real time data which is fed into self-learning algorithms based out of machine learning and artificial Intelligence to continuously improve the output and provide feedback loops.

There are two important data sources when it comes to Insurance Industry. The first is the personal information generated due to our browsing patterns, data shared on the social media platforms also data generated via e-commerce due to online shopping behaviours by consumers. This kind of data is an asset to insurance industry since it reveals a lot about an individual’s lifestyle and behaviour. The major contributors of such data are companies like Google, Facebook, Apple, Amazon, Alibaba, Baidu or Microsoft. These big players have the highest market shares and are prominently  based on the value of the consumer data. The second new data source originate from internet of things, the sensors installed in appliances, cars, wearables or smart homes.

                The emergence of large data analytics and artificial intelligence has triggered various eco-societal benefits throughout the value chain of Insurance Industry. Few of the specific areas are discussed below:

New distribution models: new applications are revolutionizing customer interaction by virtual assistants, automated brokers, chatbots and robotic consultants and using big data and artificial intelligence to optimize customer segmentation, targeted marketing and dynamic pricing.

System automation: These systems of big data and artificial intelligence seek to automate or improve the efficiency of internal processes. Straight-through processing allows for the integration of portions of the value chain or even the entire value chain, possibly involving underwriting, claim handling, risk management, finance and investment management, as well as regulatory reporting and enforcement.

Lower Premium Rates: Data Science allows insurers to look in much greater detail at the risk profiles of their applicants. This more granular risk assessment will require better informed underwriting decisions and allow insurers to more effectively set premiums in line with the required level of risk.

Risk Assessment: Continuous collection and analysis of behavioural data enables individual and dynamic risk assessment and the creation of a continuous feedback loop with no or limited human intervention for customers. Such digital monitoring not only improves the quality of risk assessments but can also provide policy holders with real-time insights into their risk behaviour and individual risk reduction incentives.

Fraudulent Claims assessment: Analysis of social media activity and connections can also be used effectively by groups of people working together to make a series of false or exaggerated claims to spot fraudulent claims.

New proposals: New data technologies allow for the development of new products and alternative business models, including peer-to-peer insurance, on-demand insurance, usage-based insurance, and new types of risk insurance products.

Stronger customer engagement: Data Science has a role to play in making the management of insurance claims and associated complaint processes more efficient, benefiting both insurer and policyholder. Data analytics can be used to prioritize claims: with straightforward claims fast-tracked for quick settlement, and more complex cases flagged by claims assessors for further analysis.

Customer Targeting: Insurers should use new data sources to better direct intended consumers to different, and potentially more suitable, goods. Analysis of web search history or posts on social media helps predict consumer preferences and behaviours. Those analyses, in turn, can be used to match consumers to specific products.

Reduction of Losses: By setting up a feedback loop for policyholders, digital monitoring enables them to reduce risk by adapting their behaviour. In addition, enhanced data facilitates the creation of advanced risk management and early warning systems that enable timely interventions to reduce losses and lead to additional benefits for policyholders. The combination of new data sources also paves the way for innovative risk management programs to be applied, using predictive analytics as a basis for early intervention and risk prevention.

Product Design: Similar analyses of preferences and behaviours trends also enhance the understanding of changes in consumer needs by an insurer, and this insight can be useful in developing innovative new products and designing associated features.

Such powerful new business model combination are already being launched or are clearly visible in the horizon. They include genuine peer-to-peer concepts and fully digital insurers (for example Oscar, InShared , Haven Life or Sherpa).This will inevitably improve the role of insurance from mere risk management to' predicting and preventing'

          However, the "datafication" of commercial activity will not come without cost, and many of the harms that this development causes are just beginning to be identified. Early efforts by scholars, interest groups and regulatory bodies have focused primarily on how these changes imperil the privacy and data security interests of individuals. However, Insurers and consumers could be impacted in new and unexpected ways, with potential unintended consequences on the insurance market and possible distortions can take place.

Insurance not for everyone ?

Some subsets of the general public may find that Datafication has an adverse effect on insurance costs or availability. If insurers have a clearer understanding of the risk characteristics of an individual, then individuals in certain segments of the market may find insurance more difficult or more expensive to obtain. For example: young drivers obtaining motor insurance with little driving experience. The changing premium rates as per the risk profile and  as a result of increased visibility to risk profile due to datafication can cause lack of affordability of cover for few, sometimes the risk factors are beyond the control of an individual. The lack of affordability of cover at reasonable price can cause market failure and avoiding it would require some degree of intervention from the government, regulators or the insurance industry itself.

Discrimination ?

On one hand, insurance customers are treated based on their individual risk but by doing so it implies that protected groups  have a disadvantage in case the risk is higher than that the average, at the same time not managing customers as per the risk would put questions on the risk classification being unfair. There is no easy solution to this dilemma. Consequently, a delicate balance must be struck between the precision of risk assessments and the potential for disparate impact on different social groups. How to balance this trade-off will depend, among other factors, on the cultural context and the type of risk being considered. For example, unequal factors may not be considered a problem if the risk is largely within the individual's control, or if all groups (including high risks) benefit from absolute premium reductions, although to varying degrees. In any event, insurers should test and evaluate potential disparate impact algorithms.

Less pooling of risk ?

Data Science could also influence the related concept of the degree of risk pooling. This risk pooling-where there is sharing of risks between policyholders with broadly similar risk characteristics-is a long-established insurance feature. The rise in data science is likely to steadily reduce the size of each pool, thereby reducing existing cross-subsidy levels between different policyholders. Discussions may be necessary to determine the degree of pooling that society deems appropriate and what actions government, regulators or the insurance industry need to take as a result.

Data ownership: individual or insurer ?

Another question concerns who owns the data a policyholder collects.If the insurer claims to own the insight gathered via a telematics or wearables device, this could restrict the right of the consumer to have access to a better deal elsewhere. Consequently if the customer change the provider will it still be necessary to share the pre-existing data with the insurer ?

Contextual Integrity & Privacy:

The concept of contextual integrity presupposes the flow of personal information in accordance with expected context-specific information standards. According to the concept of contextual integrity, existing contexts in which activities are grounded shape expectations that, if unfulfilled, cause mistrust. For example, customers typically expect personal information to be treated confidentially when interacting with their insurer, regardless of whether the interaction is face-to-face, over the telephone, or online. Likewise, individuals may not expect such personal information to be used to determine insurance premiums when engaging in social interaction on a social media platform.

Premium dynamics:

Through comprehensive risk assessment and increased individualisation of premiums, an individual's premiums can vary over time along with changes in risk. While such risk-based pricing increases actuarial fairness, increased premium volatility reduces an individual's insurance value, and hence their willingness to pay. Therefore, insurers will have to align the pace of rate changes with the demand of customers in a stable and predictable premium.

Threat to consumer autonomy ?

Advanced Analytics may threaten consumer autonomy, once insurers have all the information about the characteristics, actions and qualities related to risk, it will become increasingly convenient for insurers to impose onerous conditions and restrictions on consumers.

                        Comprehensive regulation of consumer insurance markets will have to limit how companies use data when performing the functions of underwriting, rate setting, policy building and claims management. For instance, within the context of underwriting, regulators should limit the flexibility of insurers to differentiate between consumers by imposing Community rating rules on all insurance consumer lines. The Affordable Care Act shows how community rating can strike a balance between the values discussed above.By stating that insurers can only take certain characteristics into account when setting premiums (age, smoking status, number of dependents), it effectively prevents insurers from engaging in prohibited discrimination and prevents them from placing additional burdens on classes that have been identified unfairly burdensome. Compliance with the General Data Protection Regulation (GDPR) in UK is a data protection legislation which is highly relevant to Data Science. The consent principle for the use of data is key, but users will need to be clear about the boundaries of data that are allowed to be used. For instance, where data is gained from social media activity and applied to insurance underwriting or marketing purposes, do policyholders and social media platforms necessarily consent to this? Also, the data collected should be relevant to the purpose for which it is being used. There is another question regarding ownership and access, as mentioned above. Hence Insurers need to follow a robust data governance system along with appropriate controls so that the evolving privacy protection regulations are adhered by. Datafication is continuously transforming the Insurance industry, providing clear benefits to policyholders and insurers. Consequently, increasing datafication is also causing wide range of public interest issues for the regulators, government, insurance industry and other involved stakeholders.  Balancing the various trade-offs discussed requires difficult value judgments on the part of consumers, firms, policymakers and regulators alike. What makes it difficult to balance these trade-offs is that they are context-specific, often ambiguous and sometimes intangible. However, finding an appropriate balance between protection of privacy and allowing innovation is essential.

References:

·      Franklin, J. (2001) “The Science of Conjecture: Evidence and Probability before Pascal”. Baltimore, MD.: The John Hopkins University Press. 

·      Kopf, E.W. (1927) “The Early History of the Annuity”. New York: Lawrence, p. 248ff. 

·      In this report, the terms ‘data’ and ‘information’ are used interchangeably. See the glossary for a definition of terms.

·      See appendix for an overview of data types traditionally used in insurance. There we also list additional types of data which become available in the context of big data. 

·      Clarke, R. and Libarikian, A. (2014): “Unleashing the value of advanced analytics in insurance”. McKinsey & Company, https://www.mckinsey.com/ industries/financial-services/our-insights/unleashing-the-value-of-advanced-analytics-in-insurance.

·     No Author, (2017) available at-https://www.actuaries.org.uk/system/files/field/document/Policy%20-%20Data%20Science%20in%20Insurance%20V08.pdf,

·      Helevstone,H. (2015) available at: https://clsbluesky.law.columbia.edu/2015/04/28/how-big-data-will-revolutionize-the-insurance-industry-and-challenge-regulatory-bodies/

·       Benno Keller, March 2018 Big Data and Insurance: Implications for Innovation, Competition and Privacy www.genevaassociation.org @TheGenevaAssoc BIG DATA AND INSURANCE: IMPLICATIONS FOR INNOVATION, COMPETITION AND PRIVACY

·       Zingales, L. and Rolink, G. (2017) “A Way to Own Your Social-Media Data”. New York Times, June 30, 2017. Market shares probably refer to the U.S. market, although this is not specified.

·      An example for such an integrated risk management system is the “Together for Safer Roads” coalition. This coalition, which includes private sector companies from different industries and insurer AIG, works with three cities (Atlanta, Sao Paolo and Shanghai) to identify and address the cities’ strategic road safety challenges. In this coalition, companies and public stakeholders share data and expertise to determine road safety challenges and potential solutions based on advanced analytics. See https://www.togetherforsaferroads.org/. 

·      Braun, A. and Schreiber, F. (2017) “The Current InsurTech Landscape: Business Models and Disruptive Potential”. Institute of Insurance Economics of the University of St. Gallen in cooperation with Swiss Re Institute. 

·      Zarsky, T.Z. (2014): “Understanding Discrimination in the Scored Society”., Washington Law Review

·      lock, W., Snow, N., and Stringham, E. (2008) “Banks, Insurance Companies, and Discrimination”, MPRA Paper 26035, University Library of Munich, Germany. https://ideas.repec.org/p/pra/ mprapa/26035.html. 

·      Pedreshi, D., Ruggieri, S. and Turini, F. (2008) “Discrimination-Aware Data Mining”, in Proceedings of the 14th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining, ACM, pp. 560–568; Calders, T., Kamiran, F., Pechenizkiy, M. (2009) “Building Classifiers with Independency Constraints”, IEEE International Conference on Data Mining Workshops; and Feldman, M., Friedler, S.A., Moeller, J., Scheidegger,

C., Venkatasubramanian, S. (2015) “Certifying and removing disparate impact”, arXiv:1412.3756 [stat.ML], Proceedings of the 21th ACM SIGKDD International Conference on Knowledge Discovery and Data Mining. 

·      Zarsky, T.Z. (2014) “Understanding Discrimination in the Scored Society”, Washington Law Review, 89 (4)