Skip to main content

Consumer Credit Assessment in the Age of Big Data

Technology is adding a twist to the creditworthiness game. The good news: it may improve financial inclusion

How banks and financial institutions evaluate prospective borrowers is changing, thanks to big data, machine learning, analytics and artificial intelligence. Lynnette Purda, professor & RBC Fellow of Finance at Smith, says all this technology can actually improve financial inclusion. That’s especially good for young adults and newcomers to Canada, who have traditionally found it difficult to get credit. 

But are there drawbacks to alternative credit assessment information? 

In this video, Purda explains how technology is shifting the way creditworthiness is gauged. “There’s a lot of data that’s just residing in an individual’s smartphone,” she says, and lenders can use this data and machine learning to determine patterns in an individual’s willingness and ability to repay debt. Lenders can even look to a person’s behaviour and their social network to assess their credit quality. 

Purda also sheds light on some of the concerns with big-data credit checking — from data privacy and security issues to potential bias. By harnessing vast amounts of data, machine learning can identify key patterns, ultimately creating a credit quality assessment model that has the potential to be more inclusive. But understanding the challenges is crucial.  


[Music playing]   

How has a person’s credit traditionally been assessed? 

00:07: Lynnette Purda: Traditionally, when banks and other financial institutions have looked at prospective borrowers, they’ve looked at their past history. So, they’re looking at things like whether they borrowed in the past, how long have they borrowed for, what kind of credit instruments have they used and what is their capacity to service those credit instruments. So that’s things like income, whether they’re a homeowner, etcetera. Those are all great data points, but they do look back historically, and what we’re finding is that for individuals that don’t have that history, they really don’t have access to obtaining a credit score. So, for those individuals, which includes young adults or includes newcomers to Canada, it’s very difficult for them to obtain credit.  

How is technology changing the way creditworthiness is gauged? 

00:54: Lynnette Purda: Technology is really changing how we assess credit quality. One of the big factors is that we just have so much more data that’s available. There’s a lot of data that’s just residing in an individual’s smartphone, for example, and we’re starting to use some of this data and the tools of machine learning to see what it tells us about patterns in an individual’s willingness and ability to repay. That means that now we can look at characteristics like behavioural patterns and even your social network that are contained on your smartphone.  

There are some companies that will actually assess your credit based on a score provided by your social network. What this means is that an individual needs to provide what they call a trusted network. These are individuals that are willing to vouch on the credit quality and character of this person. What’s nice about this approach is that everyone’s got some skin in the game. If an individual defaults on a payment, the actual credit score of your trusted network goes down as well.  

How is technology improving financial inclusion? 

01:56: Lynnette Purda: One of the things that we see when we’re moving away from traditional credit assessment and traditional banks is that we’re growing financial inclusion. If you think about the kinds of things that I’ve talked about in terms of traditional assessment, things like your credit history, it’s estimated that about 75 per cent of Americans do have a credit history, but that leaves a huge proportion that don’t. They’re really underserved in terms of the financial services that they can access.  

Now, if we start to introduce technology and we start to introduce different methods, we can serve a greater portion of the population in terms of allowing them access to credit. That means better housing for them. It means perhaps the ability to get a loan, to get more education, etcetera.  

We’re looking at things like peer-to-peer lending. Now you’ve completely taken out the bank or the financial intermediary and you have individuals lending to one another. They’re providing a narrative as to what they want to do with the funds, and then others are coming forward and providing those funds. It reduces overhead because we’re relying on technology instead of actual physical buildings and it allows a far greater proportion of the population to access credit.  

We also see that there are alternative ways that people can measure their own credit quality. A great example that I like to use is Borrowell that counsels individuals on how they can change their credit quality. So, for example, anyone can create a free account and they can get a snapshot of what their creditworthiness is today, but then it goes one step further and it provides suggestions as to possible tools or ways that you could increase that credit quality. And what this is doing is educating people on the importance of their credit quality and how to improve it bit by bit.  

Are there problems with using alternative credit assessment information? 

03:40: Lynnette Purda: As we use different sources of data, one of the challenges we face is definitely privacy and security of that data. So you can imagine a scenario where I have a fintech app that’s asked to access data on my smartphone. Now I’m providing them with that information. I want to ensure that it is secure and that they’re keeping my data private. I also want to ensure that I have the ability to opt out of that at any specific time.  

So, it used to be that I’d provide information on my income. Now, however, I’m providing information on my contacts, on my network, on my behavioural patterns. So people are providing information themselves on, for example, a peer-to-peer lending platform. They might provide a picture, they might provide a narrative, they might provide discussion of why they’re using these funds. One of the things we’ve seen is that there can be bias on the part of potential lenders depending on all of that information. 

We see that there’s bias when individuals post a picture. Sometimes then, lending decisions are more likely to be made on physical appearance and characteristics, things like gender, perhaps ethnicity.  

What we also need to consider is: How do we analyze this data? Traditionally in economics and finance, we created a model and then we used data to make predictions based on that model. With machine learning, we don’t necessarily need to do that anymore. What we do is we take the tremendous amounts of data and allow the data and machine learning techniques to identify certain patterns. When they can identify those patterns, it then creates essentially a model for assessing credit quality.