Chapter 1:
Financial Inclusion

As Insurers Embrace Big Data, Fewer Risks Are Shared

Since 2011, Progressive has offered Snapshot, a small monitoring device that drivers must install in their cars to receive the company’s best rates.^[1] The company offers discounts when the device reports that a driver brakes smoothly, keeps off the roads late at night, and drives infrequently—behaviors that correlate with a lower risk of future accidents.

Low-income individuals, many of whom are people of color, are more likely to work the night shift, putting them on the road late at night, and to live further from work.^[2] Devices like Snapshot reduce rates for some drivers by reducing the overall amount of risk sharing among drivers on the road, which means relatively higher costs for those with long car commutes or graveyard shift jobs. At the same time, such systems put responsible late shift workers into the same small category with late-night party-goers, forcing them to carry more of the cost of intoxicated and other irresponsible driving that happens disproportionately at night. Statistically speaking, this added cost does not simply reflect the risk that the late night commuter may be hit by a drunk driver. It also reflects the possibility that, as far as the insurer can tell, the late responsible night worker may be a drunk driver.

“Big data” allows for a new level of specificity in underwriting, changing how risk is allocated.

Insurers and lenders have long relied on statistics to help them assess the risks of prospective customers. But the deluge of “big data” allows for a new level of specificity in underwriting, changing how risk is allocated. Spreading risk among the insured population is a fundamental purpose of insurance. Some forms of price differentiation, such as charging more to drivers who accelerate or brake suddenly, may provide valuable incentives for the insured to drive more carefully—incentives to which drivers can respond by changing the way they drive. But for people who have to drive at night in order to reach their jobs, this differential pricing provides no benefit. It is simply an added cost.

A person’s future health, like their driving behavior, can also be predicted based on personal tracking to set insurance prices. At an annual conference of actuaries, consultants from Deloitte explained that they can now use thousands of “non-traditional” third party data sources, such as consumer buying history, to predict a life insurance applicant’s health status with an accuracy comparable to a medical exam.^[3] Models based on these data can “predict if individuals are afflicted with any of 17 diseases (e.g. diabetes, female cancer, tobacco related cancer, cardiovascular, depression, etc.) which impact mortality.” Deloitte’s model also incorporates the health of an applicant’s neighbors, at scales as small as two city blocks.

More individualized insurance pricing promises lower rates for those with the lowest risk. At the same time, however, this underwriting means less sharing of risk. Healthy people in low-income neighborhoods will pay more for their life insurance than will healthy people in healthier neighborhoods (because they are saddled with the health costs of their less healthy neighbors).^[4] Responsible night drivers will pay more for car insurance than will responsible daytime drivers (reflecting not only the night driver’s risk of being hit by a drunk driver, but also the risk that, as far as the insurer knows, the night driver might be a drunk driver). Insurance prices that are more accurate for most people may, by the same token, be less fair to those nearest the most vulnerable.