> Posted by Sonja E. Kelly, Fellow, CFI
Big data is sexy. It’s new, it’s hip, and we’re only beginning to explore its uses for increasing financial inclusion. McKinsey calls it “the next frontier,” CGAP puts it in its “trends to watch” category, and we’ve talked about it on our blog. Big data is here to stay, and it’s changing the way that the financial inclusion industry operates. But as we proceed with big data, let’s exercise the caution required to ensure consumer protection.
Big data is starting to be used as an alternative to standard credit bureau data, with new scoring methods being created to construct credit ratings for those with thin or poor credit history. Proxies for credit history can be anything from how frequently a person “tops up” their mobile phone credit to the number of minutes spent looking at a loan product online. To determine creditworthiness, analysts look at larger trends in the data in the same way an insurance company might, comparing the individual to the average and looking for factors that correlate with creditworthiness. For example, on average, people who spend longer reading and understanding the terms of a loan online might be more likely to pay the loan on time.
Two groups of people currently share the bulk of potential benefits of big data applied to credit products. First, there are those who have previously been considered a credit risk who should not be classified as such. Perhaps these people have an unfairly low credit score, or perhaps past mistakes do not indicate future credit behavior. Second, there are those who have “thin files,” or not enough information available on them to enable a lender to make a determination of creditworthiness. For these two groups, additional data points could provide more indication about future credit behavior.
While recognizing that big data is an industry game-changer, we do need to keep in mind some critical questions. Big data has a great deal of power to transform financial inclusion efforts, but what are its downstream effects? What are the consumer protection and legal implications? Does big data allow for implicit new discrimination? And as it’s being used now, is it making life better for consumers?
Recently, the National Consumer Law Center released a report with a status update on the promise of big data for loans in the United States, assessing the terms and conditions of loans made using big data as an alternative credit scoring method. The report focuses on the U.S. context where 64 million Americans have no credit history or poor credit history. It highlights consumer protection and privacy concerns surrounding big data in the industry. It offers the finding that the cost of loan products that incorporate big data is similar to rates offered by payday lenders, rather than being lower, as hoped. We must ask whether big data can offer better loan terms for those clients who are good credit risks but don’t pass standard credit scoring metrics. Surveying a number of startups, LendUp, ZestFinance Inc., and Think Finance Inc., the report finds that annual interest rates range from 134 to 749 percent, while the average payday loan rate is 400 percent.
Consumer protection becomes a major issue when we consider the type of “big data” information being collected and how it’s used for credit decisions. Federal Trade Commissioner Julie Brill recently gave an address in which she brought up the consumer protection concerns that accompany “becoming the sum of our digits.” Among these privacy concerns are whether consumers are aware that information is being collected about them (a question that recently painfully entered the mainstream public discourse after the revelation about the amount of data to which the NSA has access). Beyond this is the question of who “owns” the data—is it the consumers? The institution collecting it? The institution using it? If a data breech happens, who is to blame, and who is the victim? To date, we haven’t adequately addressed these concerns.
Another area specifically relevant to the U.S. legal context but with lessons for other countries is whether using big data exclusively for evaluation of creditworthiness is discriminatory. Today, because of the proprietary nature of algorithms that big data companies are using, there is no way of knowing how race, gender, or ethnicity are treated. With no public examination of the criteria used to determine creditworthiness, it is difficult to discern whether algorithms are using discriminatory standards that would be illegal in more visible circumstances.
Is the use of big data effective? Yes, it really does look like a game changer. For a host of tech startups, statistical analysis of big data to evaluate creditworthiness is actively used to make decisions much more efficiently, and big data applications have been able to identify and include an additional subset of people who might not have been offered credit at a non-payday lending institution. One organization recently disclosed to me that their evaluation of applicant suitability for a loan shrank from weeks to just minutes using a product developed by DeMyst Data. IBM reports that big data on average cuts loan approval time by 75 percent – an enormous benefit to both lender and borrower Anecdotally, big data is believed to identify about 15 percent of rejected loan applications who could actually qualify for a loan using big data.
It’s true that big data is changing the industry, though while we wait to see the extent of the change, we do need to continue to ask tough questions. Call me a curmudgeon, but I’m not one hundred percent on the big data bandwagon. I am optimistic about the role big data will play in financial inclusion, but cautiously so. While we wait to see, we do have the responsibility to think more critically about it.
Image credit: Infocux Technologies
Have you read?