Printer Friendly

Peer pressure: a new analytics model helps insurers identify how bad--and good--group behaviors spread throughout society.

Insurers spend a lot of time looking at linear, historical data to anticipate claims and loss severity. But what if instead of focusing on the past, insurers could examine group behavior to gain an understanding of changing risk trends--and see the impact that human behavior has on the progression of risk throughout society?

Assuming teenagers change some of their behaviors by adjusting to their peers as addressed in an article from the Journal of Health Economics, how long, for instance, would it take for the safer driving behavior of teens in Maine to spread to teens in North Dakota? Answers to questions such as this could give insurers information that could help them to not only gain an edge over traditional competitors but over new disruptors as well.

To that end, Accenture and researchers at the Stevens Institute of Technology developed an advanced analytics model designed to demonstrate how behavioral trends that affect risk--such as drunk driving, obesity or smoking can diffuse, or spread throughout society over time.

The Spatial Risk Diffusion Model enables insurers to build population groups--based on geographic, economic, and social components--and examine how they influence other group's behavior, for better or worse.

The basic premise: societal norms, which determine the level of risk taking, are social constructs, i.e. they are formed through the interactions of a group over a period of time. The theory of network propagation of risky behaviors has been supported for smaller networks and at an individual level, by research in other fields, such as a Harvard Medical School and University of California-San Diego studies on the spread of obesity among friends, and smoking cessation, that were published in the New England Journal of Medicine.

The insurance industry, however, has not yet applied the theory to understand the propagation of risk trends across the larger population, in part because of the sheer size of associated social networks and unavailability of individual level data.

To keep in step with competitors--and fend off incursions from digitally adept companies that could make their initial foray into the insurance industry-- insurers should embrace advanced analytics and explore a new data-driven approach to assess risk for a competitive advantage while addressing concerns related to network size and data access.

The Model in Action

The Spatial Risk Diffusion Model analyzes network nodes that represent a cluster of people located in a particular geographic area, rather than individuals. This enables the model to cover a larger population.

To test the model, the research team studied 12 years of National Highway Traffic Safety Administration data on risky teenage driving behavior, as measured by fatal and severe non-fatal accidents. Applying the model, the research predicted with meaningful certainty which statewide populations of teen drivers would be led by the behavior of young drivers in other states, as well as how long it would take for that behavior to propagate.

For example, the model accurately predicted that the safer driving behavior of teens in North Dakota would consistently lag behind the same behavior of teen drivers in Maine by more than a year.

The model's potential impact on insurers is significant. In the United States, teenagers' risk attitudes are becoming more conservative by about 4% per year. The analysis suggests that auto insurers could improve their loss ratios by 60 to 100 basis points by accounting for risk diffusion effects gleaned from state-level data. The loss-savings opportunity could be larger if the analysis is refined to ZIP code or county level, both are natural next phases of this study.

There are other high-potential areas to consider as well. One example is workers' compensation. Insurers are concerned about the length of time claimants remain out of work after suffering an injury. The decision whether to stay out of work as long as possible or return to work as soon as possible is also a social construct, driven by localized social norms in an area of collective influence; the Accenture/Stevens Institute of Technology study, Spatial Risk Diffusion--Predicting the Propagation of Risk Linked to Human Behavior, indicates this can be measured and predicted through spatial diffusion models.

Marketing, Public Policy Impact

The Spatial Risk Diffusion Model could provide tremendous support to insurers' marketing efforts and public policy campaigns. The model identifies groups that are leaders and laggards in behavior; insurers could allocate some marketing resources to regions the model has identified as having a meaningful influence over behavior in target markets.

The model could work similarly in public policy campaigns. A traditional campaign designed to discourage drunk driving in a specific neighborhood would focus on the target area. But, as the model demonstrates, there is a high likelihood the campaign would be more effective if some resources were allocated to a region from which drunk driving behavior diffuses and ultimately influences behavior in the target neighborhood.

To apply the spatial risk diffusion concept when selecting and pricing risk within the existing insurance regulatory framework, insurers will need to build specific behavioral assessment capabilities. This can be accomplished by taking three initial steps.

Identify data points of most value and with the highest predictive potential. When internal data and external data are combined with new analytical techniques, the resulting insights can become even more powerful. Insurers are already experiencing the risk assessment benefits of some digital data collection efforts. Auto insurers, for example, are using wireless devices in vehicles to monitor factors such as time spent driving, acceleration and braking. In the health insurance market, Humana has integrated its consumer health app with Apple's HealthKit to provide consumers with an overview of their health and fitness data.

Similarly, customers who own homes in which various systems and devices can be controlled remotely, such as the thermostat with a smartphone, can share data with insurers, utility companies and security system providers. Besides mitigating risk by leveraging data generated by connected homes, insurers can identify the progression of risk as a function of human behavior in other lines of coverage, including workers' compensation, burglary and health.

Incentivize customers to share personal data that would provide additional insights into behavior. Accenture's Consumer-Driven Innovation Survey showed that 78% of insurance customers globally would be willing to share personal information with insurers in return for lower premiums or quicker claims settlement. Insurers that can build a reputation for tailoring valuable services while safeguarding personal data they collect for this purpose will have a meaningful competitive advantage.

Consider partnering opportunities. Now that the spatial risk diffusion concept has been validated through the model, and the merit of applying it in an insurance context has become clear, the next step is developing a full-scale model that marries granular geographic and socio-economic networks with claims data. That model would be for a specific line of coverage, such as workers' compensation or auto insurance.

It is becoming increasingly evident that insurers who do not look beyond their traditional data sources and analytical methods will be left behind by rapidly emerging disruptors, as well as by traditional competitors who adopt innovative risk assessment methodologies.

As the Spatial Risk Diffusion Model results exhibited, advanced analytics can offer insurers the ability to see the impact human behavior has on the progression of risk throughout society. These are key insights that can help a company further explore its risk assessment strategy for a competitive advantage.

With a better understanding of how risk spreads throughout society, insurers would be able to improve their positioning against new market entrants as well as current competitors. They would see new opportunities for growth along with the chance to become the disruptor, rather than the disrupted, in an industry that is overdue for change.

Key Points

The Challenge: Insurers typically look at risk as it relates to past events.

The Means: A new analytics model allows insurers to better understand behavioral trends and the spread of risk.

The Outcome: Insurers who better understand how risky behaviors spread throughout society could tighten loss ratios and be more competitive.

Best's Review contributors:

Babak Heydari, Ph.D., is an assistant professor of systems science and engineering and director of the Complex Evolving Networked Systems Lab at Stevens Institute of Technology. Sharad Sachdev is a managing director, Accenture Analytics innovation and insurance industry lead. They can be reached at; The authors would like to acknowledge the contributions of Martin Spit, managing director, Accenture Strategy, Insurance who provided industry insights and Mohsen Mosleh, Ph.D. student at Stevens Institute of Technology, who was responsible for much of the data gathering, implementation of the model and visualization of the results.
COPYRIGHT 2016 A.M. Best Company, Inc.
No portion of this article can be reproduced without the express written permission from the copyright holder.
Copyright 2016 Gale, Cengage Learning. All rights reserved.

Article Details
Printer friendly Cite/link Email Feedback
Title Annotation:Analytics
Author:Heydari, Babak; Sachdev, Sharad
Publication:Best's Review
Date:Jul 1, 2016
Previous Article:Emerging from murky waters: the outlook for capital standards is clearer, but there is still work to be done.
Next Article:What A.M. Best has said about medical professional liability.

Terms of use | Privacy policy | Copyright © 2022 Farlex, Inc. | Feedback | For webmasters |