DAVID GREENE, HOST:
When you walk into a bank and ask for a loan, a banker will pull up your financial history - you know, your bank accounts, your credit scores, your assets and liabilities. This is how the bank determines whether or not you're a good risk. Well, in poor countries, many people have no financial track record. And as a result of that, loans don't get made, businesses don't get started, homes don't get purchased. Recently, researchers discovered that there might be a way to help banks judge whether people with no financial records are a good credit risk. And to talk about that, NPR social science correspondent Shankar Vedantam is here.
Hey, Shankar.
SHANKAR VEDANTAM, BYLINE: Hi, David.
GREENE: So what exactly did researchers do here?
VEDANTAM: Well, there's an old joke, David, that if you drop something on a dark street, you can either look for it at the point at which you dropped it or you can look down the road under the streetlight because it happens to be bright there. But in some ways, this new research is about how sometimes looking under the streetlight might actually be a good idea. Here's what happened. Daniel Bjorkegren is an economist at Brown University. And along with Darrell Grissen, he realized there's very little financial information on people in poor countries, but there's billions and billions of mobile phone accounts. So Bjorkegren said instead of looking at financial records, which is the dark part of the street, let me look at phone records because there's lots and lots of light there.
DANIEL BJORKEGREN: There are a lot of behaviors that are visible if you look in this data. This would include things like do they keep their balance top-upped so that they can make calls in case there's an emergency? Or do they let their balance run low so that their friends would have to call them?
GREENE: Makes sense, but, I mean, this kind of information really is as helpful as other information to judge someone's creditworthiness?
VEDANTAM: So Bjorkegren wasn't sure about that, but he said it's worth looking. So what he did was he got data from a Caribbean bank on more than 3,000 loans. About 1 in 8 of these loans defaulted, meaning the borrower did not repay the loan properly. Bjorkegren then looked at mobile phone data during the 90 days leading up to when people received the loan. And then he crunched the data, and he had a computer examine whether there were patterns that connected the way people use their phones and the likelihood of them repaying the loan. And surprisingly, Bjorkegren finds that patterns in phone usage are remarkably accurate in predicting who's going to default. So for example, David, if I'm calling just a few people each week versus calling a hundred people each week, it actually gives you some sense about my potential as an entrepreneur, about how large my social network is and how many contacts I have.
GREENE: So a smaller social network might mean that you're not going to be as creditworthy?
VEDANTAM: Exactly because part of the reason people are taking these loans is they're taking these loans to start businesses. And if I'm starting a business and I have three friends, I'm likely to be less successful than if I'm starting a business and I have 3,000 friends. So Bjorkegren thinks that data like this can be remarkably useful in telling a bank who to give money to and who not to give money to.
BJORKEGREN: This particular bank could have continued to loan to 75 percent of its borrowers. And by eliminating those 25 percent who are most risky, they could have actually eliminated 43 percent of their defaults. So it works pretty well in identifying who among these borrowers was likely to default.
GREENE: Works pretty well - I mean, banks take this very seriously, obviously. I mean, does he really think that banks are going to buy into this and do it?
VEDANTAM: So I think this is really a pilot study in a way, David. This would have to be replicated. But Bjorkegren told me that the data in the phone records almost approached the accuracy of having a credit report in the United States and predicting who was going to default. So in other words, it's almost as good as actually having financial records on people.
GREENE: Shankar, thanks as always.
VEDANTAM: Thank you, David.
GREENE: That's NPR social science correspondent Shankar Vedantam. He regularly comes in to talk about social science research. You can follow him on Twitter at @HiddenBrain. This program is @nprgreene and @MorningEdition. Transcript provided by NPR, Copyright NPR.
