Pub. 10 2015-2016 Issue 6

March/April 2016 15 Extraordinary Service for Extraordinary Members.  Counselor’s Corner — continued on page 16 experience information or sell predictive analytics based on non-traditional characteristics. 9 Is the bank complying with adverse action and risk-based pricing notice requirements where applicable? Are there circumstances where the bank itself could be considered a consumer reporting agency? Equal opportunity laws, including the Equal Credit Op- portunity Act (ECOA) 10 and its implementing regulation (Regulation B 11 ), Title VII of the Civil Rights Act of 1964, 12 the Americans with Disabilities Act, 13 the Age Discrimina- tion in Employment Act 14 , the Fair Housing Act, 15 and the Genetic InformationNondiscrimination Act 16 —Banks should consider whether their use of big data results in disparate treatment of, or a disparate impact on, a protected class. This could arise in a variety of ways, including by use of “innova- tive” underwriting techniques. The FTC reminds creditors that disparate treatment of a protected class is not permitted even if supported by a big data conclusion—for example, a lender cannot refuse to lend to a single person as such even if big data analytics show single persons are less likely to pay than married persons. 17 This may be a greater concern if the lender’s big data process results in systemic disparate treat- ment. Likewise, using big data (such as making credit deci- sions based on zip codes or social media usage) may violate the ECOA if it results in a disparate impact on a protected class that is not justified by a legitimate business necessity. 18 The FTC also suggests that creditors use caution when using big data to target advertising to particular communities. In addition to specific Regulation B requirements, advertising may impact subsequent lending patterns. 19 The Dodd-FrankWall Street Reformand Consumer Pro- tection Act (in particular, the prohibition of unfair, decep- tive, or abusive acts or practices) 20 —Since it was directed to a broader audience, the FTC Report references the Federal Trade Commission Act. 21 Dodd-Frank’s Consumer Protection Act and its UDAAP considerations are likely of more concern to bankers. Can a bank’s collection, analysis, and use of big data, or the representations made (or not made) about those practices, be considered unfair, deceptive, or abusive acts or practices? The FTC cites the example of a card issuer that touted the ability of customers to use their cards for cash advances but failed to disclose that such usage could result in reduction of their credit limits. 22 Is the bank keeping the consumer data it compiles reasonably secure? Is the bank being careful to vet those to whom it provides data; are they using it for fraudulent or discriminatory purposes? Other Laws —The federal consumer protection laws mentioned above are certainly not the only laws applicable to big data. State consumer protection, fair credit reporting, privacy, and data collection laws should not be overlooked. In addition, state and federal guidance on information security safeguards, breach notifications, data retention and destruc- tion policies, and vendor management (for big data vendors) alsomay be relevant, among other things. Some consideration of card, clearinghouse, and other network association rules also may be necessary, depending on the data in question. Fair Use of Big Data The FTC also suggests that stakeholders should think further, beyond existing regulatory compliance topics. Big data may provide very useful tools and information and may make it possible to reach out to new customers. But big data also may have detrimental impacts such as denying oppor- tunities based on the actions of others, 23 reinforcing existing disparities, 24 exposing sensitive information, 25 creating new justifications for exclusion, 26 selectively charging higher prices, 27 and weakening consumer choice. 28 According to the FTC, big data users should evaluate the good they can do and the harm they can cause when using big data. Whether these suggestions will find their way into more formal regulatory guidance remains to be seen. But for now, the FTC has posed some basic questions to consider when using big data: 29 1. How representative is your data set? Underrepresen- tation of certain populations in data can cause skewed results. A nonbanking example the FTC cites was a Boston smartphone app to collect pothole data. Once it was recognized that lower-income persons are less likely to carry smartphones, it could be seen that the data generatedmight not be fully representative of road conditions in low-income neighborhoods. 2. Does your data model account for biases? Large data sets may contain hidden biases. Those biases can be carried forward and reproduce existing patterns of discrimination. For example, if big data algorithms are based on applicants from certain colleges, will they sim- ply result in replicating previous biases that may have existed in the underlying college admissions decisions? 3. How accurate are your predictions based on big data? Big datamay be good at finding correlations, but it is not necessarily as good at determining which correlations are meaningful. 30 For example, using rental payment history to determine access to credit may not properly consider why individuals aren’t paying rent in a timely manner (the landlord may be withholding utilities or providing substandard living conditions). 31 Similarly, if big data analytics were to predict that people not partici- pating in social media were 30 percent more likely to be identity thieves, would a creditor be justified in treating them differently because they might be more risky? 32 4. Does your reliance on big data raise ethical or fairness concerns? Perhaps the hardest question of all, the FTC suggests that big data users consider whether their use is fair. There should still be a place for critical think- ing about fairness in deciding what to do with big data results. More positively, banks might find that big data can be used proactively to provide new opportunities to serve underrepresented and underserved communi- ties. 