Big data, big questions—insurers and advanced data analytics.
As with all other parts of the financial services sector, technological advances are transforming insurance.
This article first appeared in Fintech Law Report, published by Thomson Reuters.
Among the many facets of “InsurTech” are the use of artificial intelligence (“AI”) and advanced data analytics to analyze both traditional data and big data. For the purposes of this article, the term “AI” means technology-driven computational tools, including algorithms, to complete tasks that traditionally required humans. “Big data” lacks a uniform definition, but in this article, the term describes large and complicated sets of data from traditional and nontraditional/digital sources. Likewise, “advanced data analytics” describes the process for analyzing data, including big data, to assist in business processes and decisions.
As discussed in detail below, the use of advanced data analytics is very positive for insurers and policyholders because it can improve and transform the customer experience, increase efficiencies, and reduce costs. Insurers should be looking ahead to increased regulator interest in this space, both because of the positive effects of advanced data analytics, as well as potential policymaker concerns regarding their widespread use.
Current use of big data and advanced analytics by insurers.
The insurance industry has a reputation for being slower than other parts of the financial sector to innovate, and in a 2016 survey, only 11 percent of industry respondents agreed strongly that their organization was taking full advantage of advanced analytics. However, insurance is well poised to benefit from technological innovations, particularly AI and advanced data analytics. Even before the advent of new technology to gather and analyze data, insurers collected a great deal of consumer data via their underwriting, premium payment, and claims payment operations. Technological advances are now transforming business processes via the use of data analytics, and disruptive technology (wearables, telematics, etc.) have made more data available.
Advanced analytics are currently being used in insurance to enhance or replace back-office functions – notably, risk assessment and pricing (underwriting), fraud detection, and claims processing. In a 2017 survey, 51 percent of insurance companies surveyed used advanced analytics for claims modeling and in efforts to reduce claims, and 42 percent used analytics for actuarial model testing. These numbers are increasing rapidly, as technology advances and insurers realize the benefits of its use.
In the property and casualty industry, for example, before advanced data analytics, policies were historically priced based on fewer than 20 variables including, among other factors, driving record and age. A standard list of questions would further fine-tune pricing. Now, insurers are starting to use additional data from new and non-traditional sources, with more than 1,000 variables and ever-finer ratings classes.
To provide some specific examples of insurer use of big data and advanced data analytics, Zurich uses AI to process personal injury claims, including reviewing medical records. Some of the country’s largest auto insurers use Tractable, an AI review process, to process auto claims. Tractable uses an AI photo estimating system to provide appraisals, and the company’s founder estimates that the technology could replace human desk review on up to 70 percent of auto collisions. Allstate uses an online photo uploading process, QuickPhoto Claims Estimating System, to process auto claims.
Additionally, the commercial and homeowner insurers are exploring the use of telematics devices to feed data over the Internet of Things, and data from social media are being used to detect fraud by workers’ compensation carriers. Life insurers use advanced data analytics in accelerated underwriting, which can dramatically shorten the time between an initial application and approval.
State regulatory framework.
The regulatory oversight of the use of big data and advanced analytics falls to the states as the industry’s primary prudential regulator. As part of their broad regulatory authority, states regulate, among other components, the pricing of premiums/rates, product approval, advertising, and privacy protections. Each of these areas will be affected by the use of big data, and the state regulation and scrutiny of advanced data analytics by insurers is increasing quickly.
As one example, over a dozen states and the District of Columbia have restricted or prohibited “price optimization,” the practice of insurers analyzing and using data not correlated to predicted loss to establish pricing. A key regulator concern regarding using non-traditional data is that it can result in consumers paying different prices for identical coverage.
NAIC Activity. Individual states can address insurers’ use of big data through their broad supervisory powers and specific statutes, including consumer privacy statutes, and the National Association of Insurance Commissioners (“NAIC”) also is examining the issue. Specifically, the NAIC has created an Innovation and Technology (“EX”) Task Force to address technology in insurance broadly, and a Big Data Working Group to address issues around big data specifically.
The NAIC Big Data Working Group currently has three charges: (A) review and consider any needed changes to the existing regulatory framework; (B) propose a mechanism to provide resources, and allow states to share resources, to facilitate review of complex models used for underwriting, rating, and claims; and (C) assess data and tools required for regulators to appropriately monitor the marketplace. In 2018, the Working Group will consider modifying the use of consumer data in marketing, underwriting, claims settlement, and pricing for property and casualty, health, and life/annuity products. The Working Group also will consider model laws and regulations governing the use of consumer data during this process. In other words, insurers should keep very close track of the work of this group in general, and as a source for recommendations on future changes to existing regulations.
Disclosure of Use of Advanced Analytics. Currently, there is no uniform mechanism for the disclosure by insurers of their data sources and use, though insurers are subject to routine and ad hoc data calls by state regulators. Consumer groups have suggested that the NAIC assist states in creating a template for disclosure of the use of big data and develop resources to help states analyze the use of big data models by insurers.
Regulation of Data Vendors. The NAIC and consumer groups also have raised the issue of the regulation of data vendors as worthy of additional scrutiny. One consumer group has recommended additional regulation/oversight of advisory organizations and statistical agents. To date, the regulation of vendors of big data has been raised as a concern but no clear regulatory consensus has emerged.
Disparate impact and discrimination concerns.
Another area where both federal and state regulators have raised initial questions about the potential positive and negative effect of the use of big data by insurers is in the area of discrimination.
Federal Law Regarding Discrimination by Insurers. In order to accurately price policies based on predictive data, insurers are permitted to differentiate between categories of consumers by risk profile in underwriting. However, this differentiation is subject to limits, including under federal law. For example, health insurers are prohibited from considering preexisting conditions, genetic information, or gender in underwriting.
In addition, the Obama Administration issued a rule, upheld by the Supreme Court, that provided that the Fair Housing Act (“FHA”) applied prohibited practices (regardless of discriminatory intent) that have a disparate impact on protected classes. The insurance industry is fighting the application of this rule to homeowners insurance. The rule and the subsequent cases did not specifically involve big data in arguing disparate impact, but future cases could certainly arise in which plaintiffs argue disparate impact under the FHA resulting from the use of big data in homeowners insurance.
State Regulation of Discrimination by Insurers. Other than the areas of health insurance and homeowners’ insurance, no federal laws specifically prohibit insurers from discriminating in underwriting. Rather, that issue is left to the states as the industry’s primary prudential regulator. Generally speaking, a rate structure is unfairly discriminatory if the premium cost between categories of insureds does not correspond to expected differences in actual costs. Risk discrimination is permitted, but rates that are unfairly discriminatory are not.
A 2013 in-depth survey of state law revealed that at that time, 28 states had general statutes prohibiting “unfair discrimination or “unfairly discriminatory” rates by insurers across all or multiple insurance lines. Regarding specific classes that are otherwise protected by federal civil rights law, the survey found that 33 states and the District of Columbia either limited or prohibited the use of race by property/casualty insurer and six states imposed limitations on the use of gender in life insurance. Said another way, states are not uniform in their prohibitions on insurers using race, religion, and national origin, and according to the survey, only seven states forbid the use of race, national origin, and religion across all lines of insurance; two states prohibit the use of race and national origin but not religion, and one state (Louisiana) explicitly permits the use of race in life insurance.
But how is this relevant to the use of big data? Going forward, as regulators and consumers become more aware of and knowledgeable about the use of big data in insurance, there is potential for increased concern that big data could unintentionally disproportionately harm protected classes. For example, the use of big data and the outcomes it produces could be challenged under general unfair discrimination statutes. In addition, if there is a high correlation between non-traditional data and the effect of the use of that data on protected classes, public and regulatory scrutiny could increase.
To date, no state has enacted a disparate impact standard that would apply to insurers. Indeed, as one commenter has noted: It is likely that the rate standard of unfairly discriminatory will be in direct conflict with the application of a disparate impact standard to insurance rates. This conflict will potentially exist for nearly every risk factor used to develop property/casualty insurance rates because protected classes, most if not all of the time, will not be evenly distributed throughout the various risk classifications. If a court or legislature were to order that all disparate impacts be eliminated from insurance premiums, it is likely that accurate risk assessment would be destroyed, resulting in unfairly discriminatory rates.
As the American Academy of Actuaries has noted, risk classification may be more palatable to the public if there is clearly causation between risk characteristics and cost versus merely correlation evidence. This may be especially true with the use of big data, where algorithms and machines that learn may produce results that extends beyond human programming of rules. Indeed, consumer groups have suggested that laws be amended to require more than correlation to avoid and minimize the disparate impact.
Federal government scrutiny of big data and advanced analytics.
As noted above the primary responsibility to oversee the use of advanced data analytics by insurers lies with the states, with several notable exceptions. However, federal scrutiny of the issue could inform and influence state activity.
Obama Administration Big Data Working Group. The most recent significant scrutiny of the issue of big data occurred under the Obama Administration. President Obama created a Big Data Working Group that released several reports, the latest in 2016. That report and prior reports specifically identified potential positive and negative effects of the use of big data in the financial services sector (specifically, banking). Among the concerns raised were privacy issues and the potential for intentional and unintentional discrimination in the use of big data. The Working Group report focused on potential disparate impact in the provision of credit, but the concerns are the same across banking and insurance – that the use of big data could result in a disproportionately negative effect on low-income and underserved consumers. On the positive side, the report noted the potential use of big data to detect and prevent discrimination.
The Federal Trade Commission. The Federal Trade Commission (“FTC”) does not have direct jurisdiction over the business of insurance, but insurers are subject to the Fair Credit Reporting Act when they use credit scores in underwriting. In addition, the FTC is an influential federal regulator generally and could spark congressional and other regulator interest.
The FTC has examined big data brokers, and in a 2014 report, concluded that most data brokers collect vast amounts of consumer data from numerous sources, largely without their knowledge. In this report, the FTC made a number of legislative recommendations to Congress to require additional disclosure from data brokers, and in some cases, to permit consumer opt-outs. While the current Administration may be unlikely to adopt any Obama-era recommendations, and the current Republican Congress may be unlikely to implement these recommendations, the FTC’s report is significant in that, like the NAIC interest in data brokers, it points to additional scrutiny being likely in the future.
In addition to its focus on data brokers, in January 2016, the FTC issued a report entitled “Big Data: A Tool for Inclusion or Exclusion?” which, among other concerns, raised the concern that the use of big data could have a disparate impact on low-income and underserved communities. While insurance was not specifically analyzed in this report, the concerns raised would apply to the use of big data analytics in pricing, claims, and many other insurance processes. As with the 2014 data broker report, the 2016 FTC report is significant because it highlights another area of state and federal overlapping interest (potential discrimination concerns) rather than because it signals immediate federal action.
The Federal Insurance Office. The Federal Insurance Office (“FIO”), housed within the Treasury Department, does not have regulatory authority over the insurance industry but does have the statutory authority to “monitor all aspects of the insurance industry.” The FIO issued a 2016 report on the Protection of Insurance Consumers and Access to Insurance, which opined that “the increasing use of big data…may…present risks for consumers, creating an increased need for policymakers and insurance regulators to guard against unlawful discrimination.”
The FIO report also noted that brokers of big data to insurers are outside of the regulatory purview of state insurance supervisors, but are supplying critical pricing inputs, including developing pricing formulas, which may have a material effect on the accessibility and affordability of insurance.
Financial stability board interest.
In addition to the Federal Government interest in the use of big data and advanced data analytics, international regulatory groups have also taken an interest, from the perspective of protecting the global economy. Specifically, the Financial Stability Board (“FSB”) released a report in November 2017 on AI and machine learning in financial services and implications for global financial stability.
While the report identified positive effects from the more widespread use of AI and machine learning, it also noted that it could create systemically important institutions outside of traditional regulatory channels, and, notably, stated that a lack of understanding of how AI and machine learning methods arrive at conclusions (and lack of auditability) could give rise to macro risks, and result in unintended consequences.
Implications of the Report. The FSB, which reports to the G-20, does not have regulatory authority itself but does identify systemic risk issues and can provide input to other international regulatory groups, including the International Association of Insurance Supervisors. The United States is a party to the FSB and the IAIS. Ultimately, any regulatory agreements flowing from FSB-identified concerns would have to be agreed to by the US as a member of the IAIS, and, more importantly, voluntarily implemented in the United States by insurance regulators. Under the current Administration, FSB scrutiny and recommendations are unlikely to result in any action in the United States; nonetheless, the FSB’s concern is worth watching.
Our bottom line.
The use of big data and advanced data analytics by insurers is swiftly transforming the business of insurance and the customer experience. These technological advances have enormous power to improve outcomes for consumers and the industry. Still, the industry should anticipate increased scrutiny by regulators, primarily at the state but also federal and international levels, and ensure that their use of technology, and the benefits to consumers, are well understood and supported by their regulators.