Should lenders use social media data to assess creditworthiness?
Extending loans to the 64 million or so Americans with either no or sparse credit histories compels lenders to reference alternative credit data. This information consists of utility, rent and subscription payment data, all of which indicates whether certain individuals are financially responsible.
However, some analysts also include social media and mobile app usage information under the alternative credit umbrella. Utility and telecom payment data has demonstrated its ability to predict mortgage delinquencies, but can the same be said of Facebook and LinkedIn data?
Searching for patterns that aren't there
For every successful innovation there are a dozen failures. Using big data analytics to correlate social media habits with creditworthiness is an ambitious endeavor, but according to a study from the National Consumer Law Center, such a pursuit hasn't proven effective.
The NCLC sought to determine whether using Internet search, mobile app usage and social media data enabled lenders to develop accurate financial profiles of borrowers. One of the first problems the NCLC discovered was that some companies, in an attempt to use as much information as possible, may include hundreds of variables that may be irrelevant to risk. Nassim Taleb, a risk engineering professor at New York University, reinforced this point.
"If I generate ... a set of 200 variables - completely random and totally unrelated to each other - with about 1,000 data points each, then it would be near impossible not to find in it a certain number of 'significant' correlations of sorts," said Taleb, as quoted by the NCLC.
Essentially, a realistic connection between two factors - say, how often a person posts on Twitter and the number of pages he follows on Facebook - may not exist. The problem is, an algorithm designed to identify relationships may make a mistake and believe two factors are not only related, but also conclude that the data suggests a person won't make payments on time.
Social media and creditworthiness
The social media data NCLC collected from reports contained several inaccuracies. For example, in one report from Spokeo, nine out of the ten sample reports the NCLC analyzed contained either added or omitted social media accounts. This meant the organization used either included data from inactive social media profiles or information from profiles that did not exist.
Ultimately, the NCLC concluded that "while there might be some correlation between web searches, IP address, or social media posts and the likelihood of repayment, there have been no definitive understandable reason provided as to why those data points are a good measure of creditworthiness."
There's also the issue of data access. The Wall Street Journal reported that in May 2015, Facebook started restricting what information third-parties could pull from people's profiles. So even if there was a strong correlation between consumer creditworthiness and social media usage, companies don't have a lot of data to work with.
The bottom line: Social media activity isn't a good indicator as to whether your customers are going to pay back their loans. Stick with what's proven, and don't get caught up in the hype.