“It is a truth universally acknowledged…” that credit risk assessment is not always either accurate or fair. Indeed, some would say that was putting it mildly, for those with little or no credit history they likely to be badly served by the current market. One way in which it has been suggested that banks and other financial institutions could improve assessment of credit risk is to use social media data. I had a chance to explore this further with our analytics expert, Colin Gray
How can social media help credit risk assessments?
Well, what you post on social media, and your friendships and connections, give a very good indication of your lifestyle. And this, by definition, gives quite good information about how reliable you are, and may give an indication of credit risk. It may be particularly helpful for young people who have not been working so long, and have not had a chance to build up a credit history. Most of them are using some form of social media. It’s effectively a form of character reference. Obviously, nobody is suggesting that it should form the whole of a credit risk assessment, but it might well make a useful contribution.
How would it work, then?
There are several possible ways. First of all, lenders could look at what an individual had posted in various social media networks, to make a judgement about whether they look reliable, and therefore creditworthy. LinkedIn, for example, would show whether they were working, and also whether on a temporary or permanent basis. This is probably one of the biggest indicators of financial risk. For new graduates, banks could also see from LinkedIn whether they had held jobs in the past, perhaps that they had worked part-time through college, which might make them a better risk. Facebook would show whether the individual was out and about doing expensive activities every weekend, which might make it harder for them to save up and repay a loan.
Alternatively, banks could look at the individual’s contacts and connections. They could be used as ‘virtual references’: knowing a large number of people in good jobs could be a sign that you too have good earning potential. With permission, contacts could also be approached via social media, and asked to provide character references. Traditional reference systems can be gamed, but this at least means that banks can pick someone at random, giving them more chance that it’s not the individual in disguise.
But would people be happy for banks to see their social media profiles?
This is a really big question. Obviously banks could only access the information if they had full consent. And that means documented, with clear parameters to say what they can and cannot access. That might be both in terms of social network, or within a particular network. For example, you might say that you agreed the bank could see every social network but Facebook. Or you might not want anyone to look at your friends list, but everything else is fair game. The question is perhaps more what any refusal could suggest about what you had to hide, and whether people would be happy for that implication to be drawn.
How would a consent process work?
Banks would need to set out clearly exactly what information they would look at, and how they would use it. This is likely to be pretty challenging in itself, because at this stage, nobody is really clear about the potential.
Wouldn’t it be possible to ‘game’ this? After all, you only have to find out how the scoring works
Depending on how any algorithm or selection criteria worked, yes, it might be possible to post all the right things. It does very much depend on how banks use the information. There is also the possibility that people might claim that particular data was misleading. The key is probably to use social media data only as part of the picture. The idea certainly needs a lot more work before it’s ready for full roll-out.
Is anybody doing this now?
Not in any organised way. But FICO, the analytics and credit risk scoring provider, is talking about incorporating social media data into its systems. This suggests that it won’t be very long before this type of data is considered a standard part of credit scoring. Within about ten years, it could be seen as completely normal, with everyone wondering why they ever worried about it. Alternatively, the whole idea may have been abandoned completely as an invasion of privacy. Only time will tell.
If you would like to get a deeper understanding, please read this whitepaper: The Changing Landscape for Credit Risk Management. From silos to convergence with enterprise risk.