Knowing the Score

This report describes two important ways that new data and technologies are changing the American consumer credit marketplace.

First, new kinds of data are flowing into the computerized decisionmaking systems that determine who gets access to credit, and on what terms. This “alternative data” has the greatest impact on financially underserved consumers, whose creditworthiness is not well described in traditional credit reports. Credit bureaus have begun to receive certain kinds of mainstream alternative data — such as a consumers’ monthly bill payments — that are similar in kind to the monthly credit payments that have long been commonly included in credit files. This data has the potential to expand the accessibility of mainstream credit. But at the same time, a much wider spectrum of alternative data, which we refer to as fringe alternative data, is being used by some companies to underwrite financial products (especially when a full credit report is unavailable). The predictive value and fairness of fringe alternative data is unproven.

Second, credit data is being modified and sold for unregulated marketing purposes, despite the fact it was originally gathered for underwriting purposes (and thus originally subject to the regulatory strictures of the Fair Credit Reporting Act). To avoid regulatory limits, credit bureaus sell slightly aggregated information, such as the financial circumstances of a household, rather than an individual. This data can be used to target products to groups of consumers with great precision, based on the financial health of their household or neighborhood. Credit bureaus have sold such marketing data for some time, but the data’s use in online contexts is a more recent development.

This report offers three recommendations. First, we recommend that advocates seriously consider advancing the inclusion of mainstream alternative data into credit files. This new data can make more consumers “scorable” within the mainstream financial system, but must be handled carefully so as not to undermine important public policies or cause disproportionate harm in non-credit contexts. Second, we advise caution concerning credit scoring models that rely on fringe alternative data, whose predictiveness and fairness has yet to be publicly demonstrated. Finally, we urge regulators to intensify their scrutiny of online marketing practices that rely upon credit data. Technological constraints make it difficult for outside observers to understand the impact of online marketing on financially underserved individuals; regulators must play a fact-finding role.