The Big Data revolution has begun to have a transformative impact on commercial markets. Businesses are continually developing new ways to capitalize on the increased availability of analytic services and drastic reductions in the costs of data collection and storage. Industry groups, regulatory bodies, the media, and academics have attempted to stay abreast of this seemingly endless torrent of data-centered activity. It is clear that the Big Data movement has the potential to significantly improve markets. By providing businesses with actionable information about consumers and the world, it will enable companies to increase the efficiency of their operations and kick start product innovation. The “datafication” of commercial activity will not come without costs, however, and many of the harms caused by this development are just beginning to be identified. Early efforts by scholars, interest groups, and regulatory bodies have predominantly focused on how these changes imperil individuals’ privacy and data security interests. While important, these two issues are only a small part of the dangers posed by commercial applications of Big Data.

It is clear that different industries’ uses of Big Data analytics will pose unique threats to individuals’ interests. Insurance, particularly lines sold to individual consumers, is one commercial sector where expansions in the use of analytics could have substantial detrimental effects. Three qualities distinguish insurance from other commercial goods and services in this context. First, insurers have always had a greater hunger for data than other entities, as data informs the predictions that determine their profitability. Second, many forms of coverage (particularly those offered in consumer lines) can be considered necessary goods. Individuals are often compelled (by the law or by third parties) to procure certain types of coverage, yet the availability of coverage has been left to the discretion of private companies. Finally, the benefits provided by insurance often cannot be acquired through other means. Especially in the consumer context, insurance is often the only way that an individual can protect themselves against risk.

While many insurers were slow to adopt Big Data practices in the movement’s initial years, a large part of the industry has begun incorporating new forms of data-intensive analytics into their core business operations. For instance, auto insurers have begun to directly monitor policyholders’ driving practices and use this information to calibrate personalized premium rates. Similarly, many casualty insurers are using data culled from social networking sites to inform their sales, advertising, and product development practices. Finally, some insurers have begun using data culled from individuals’ digital activities to inform determinations about whether claims are likely to be fraudulent.

One of the tacit assumption of current regulatory schemes is that cost considerations are sufficient to deter certain behaviors. If the datafication of the world becomes as pervasive as many have projected, there will eventually be a point where insurers no longer encounter any economic constraints on the amount of data they use to predict and price risk. Even under more conservative assumptions, decreasing costs will allow insurers to greatly expand upon their traditional practices. Inter-firm competition ensures that insurers will seek to use data and analytics to improve their operations to the greatest extent possible.

Many of the data-centered practices that insurers will adopt will have beneficial effects on insurance markets. The primary use for such data will be to enable insurers to price individuals’ risks more accurately than they have in the past and charge actuarially fair premiums. Insurers being able to accurately match price to risk on a policyholder-by-policyholder basis will generate a number of secondary benefits. First, at least under most conceptions of fairness, the insurance system will better advance fairness interests by charging individuals amounts that reflect their personal level of risk. Second, once premiums are individualized, traditional concerns about adverse selection disappear—low risk individuals will have no incentive to leave the insurance pool if their premium rates stop being affected by higher risk individuals. Finally, if we presume that data can predict whether an individual is likely to reduce their level of care after getting insurance coverage, then an individual’s moral hazard risk could simply be priced into his premium rates.

The widespread availability of data and the lack of restraints on how insurers use it could, unfortunately, also harm markets. While actuarial fairness is considered a primary virtue for insurance markets, it is not the only criterion for evaluating the health of an industry. For instance, there would be significant support for regulatory intervention if an insurance market operated in ways that conflicted with established antidiscrimination norms. Other values that could be used to judge the health of an insurance market are consumer autonomy, the mitigation of luck-based advantages, privacy, utility maximization, and good faith contractual norms.

There are reasons to believe that each of these values would be harmed in a data-saturated, minimally regulated environment. Why would data-saturated insurers pose a threat to antidiscrimination norms? In an unlimited data environment, insurers’ underwriting decisions will be made by an algorithm that takes into account a near infinitude of characteristics and composites of characteristics. It is easy to see how discrimination against protected classes could be facilitated by such a decision-making environment—insurers could systematically disadvantage members of a protected class by having their underwriting and pricing decisions be affected by qualities that are highly correlated with the protected characteristic.

Giving insurers free rein over data would also impair society’s efforts to mitigate the advantages and disadvantages that people have due to luck or their starting positions in life. By vastly expanding the number of qualities that factor into underwriting, the use of advanced analytics will inevitably lead to immutable or luck-based characteristics assuming a role in insurers’ issuance and pricing decisions. This will confer benefits to the fortunate and cause the unfortunate to suffer further.

Finally, advanced analytics will threaten consumer autonomy. Once insurers have more information about what acts, characteristics, and qualities correlate with risk, the incentives for impinging upon policyholder autonomy by influencing their behaviors will increase dramatically. This is particularly so given that it will become increasingly economically feasible for insurers to impose more onerous conditions on consumers and enforce more restrictive covenants.

Ideally, insurance markets would optimize all of the values—actuarial fairness, antidiscrimination, etc.—discussed above. Such a market cannot exist, however, as some of these goals irreconcilably conflict with others (actuarial fairness and antidiscrimination norms being the most obvious example of this).  The regulatory challenge posed by the introduction of Big Data to insurance markets, then, is to balance prohibiting particularly harmful uses of data and fostering market-enhancing innovations.

Comprehensive regulation of consumer insurance markets will have to limit how companies use data when performing underwriting, rate setting, policy construction, and claims management functions. In the context of underwriting, for example, regulators should limit insurers’ ability to distinguish among consumers by imposing community rating rules on all consumer lines of insurance. The Affordable Care Act demonstrates how community rating could strike a balance between the values discussed earlier. By stating that insurers can only take certain characteristics (age, smoking status, number of dependents) into consideration when setting premiums, it effectively prevents insurers from engaging in prohibited discrimination and stops them from putting additional burdens on classes that have been deemed to be unfairly burdened. It also limits the impact that insurers’ premium setting practices could have on individuals’ autonomy.

The preceding post comes to us from Max Helveston, Associate Professor at DePaul College of Law. The post is based on his recent article entitled “Consumer Protection in the Age of Big Data,” forthcoming in the Washington University Law Review.