May 6, 2017    5 minute read

Is It Time to Move to Alternative Credit Scoring Models?

Assessing Risk    May 6, 2017    5 minute read

Is It Time to Move to Alternative Credit Scoring Models?

One-third of humanity does not have access to credit, is not reported to any credit bureaus and has no credit history. New technology players use new data sources to assess credit risk. This will arguably not replace traditional credit scoring models, for now.

At least 2.5 billion people in the world and several hundred million small businesses still do not have access to financial services. This is primarily the case in emerging markets. In Africa, 80% of the 1.2 billion-strong population has not established any contact with a bank, mainly because they do not receive a fixed income, and have no savings or assets that can be a desirable collateral. In China, half a billion people have no credit history, and out of 340 million of the adults, only 20 percent was reported to have a credit capacity in 2014.

The Unbanked Are Everywhere

This problem, however, does not affect exclusively developing countries. The American Consumer Financial Protection Bureau released a study in 2015 that estimates that 26 million Americans are not covered by credit bureaus, and 19 million more are with outdated data and hence excluded from lending. Moreover, 66% of millennials are defined as subprime, that is, with dubious creditworthiness.

Unmarried or poorly-housed Americans along with immigrants make a total of about 64 million, but a large portion of them – up to 40% are homeowners, entrepreneurs, and retired people who are earning a high enough rating for FICO, the largest American credit bureau.

That is why many companies that sponsor consumers to build customer risk rating models already use or test new data sources. They include the TransUnion credit scoring agency, which believes that with the new approach, the positive credit verification rates may increase by up to 20%. FICO already uses data from telecom and cable networks to build new customer pricing models.

China’s Approach

When it comes to building a universal rating base the Chinese appear to be closest to the target. The private data gathering business became allowed in 2013, but Alibaba’s e-commerce giant, Sesame Credit, has already built a powerful commercial database containing information such as Internet payments, purchased goods (passive video game buyers may have lower ratings), tax settlements and unpaid tickets.

The Chinese government has even more Orwellian intentions – it wants to create a database within a few years to rate everyone’s reputation. A poor rating will have consequences not only for credit or, more broadly, for financial reasons (for example, the need to leave a deposit for a car or bicycle rental), but may even disqualify people when applying for public and managerial positions, reception of social benefits, and even booking a train ticket.

Ironing Out Risks

The use of technology by various technology companies to build credit algorithms has been taking several years. Of most predictive value continues to be historical credit data, then transactional, and at the bottom the social one. The latter are most easily manipulated by fraudulent customers who create false identities. So it boils down to identifying promising data sources, providing access to them and converting them into credit scoring values.

Such data can come from many sources like payments for media, rent, taxes, and even penalties for late return of books to the library and tuition at the university. One Asian lender found that the prediction of payday loans was not much worse than using credit history.

A good financial mirror of the customer is the settlement of online shopping and their form. With the consent of the company, they also refer to transaction data from other financial institutions (volatility of funds on the bank and savings account, use of debit, etc.), on the basis of which models of the financial behaviour of customers arise. For others, data from employers showing stability and employment history are important.

Broad-based Approaches

Data obtained from social networks are generally complementary and not basic in the course of the mobile loan procedure. One can also verify that the client statements in the application match the data in the respective portals. The method and the time of completing an application form may be of importance. TrustingSocial, for example, uses a credit scoring system based on three main groups of parameters:

  • Authenticity – that is, verification of digital identity and customer credibility;
  • Customer relationship quality (intensity and range);
  • Financial credibility based on the history of education and employment of the borrower.

Refusing to share one’s social profile and surfing history on the web does not necessarily disqualify the customer, although it lowers their credit score.

Using information about customers on social networks, even with their permission, can face problems. Last year, Facebook severely restricted technology lenders from accessing customer data. The US Federal Trade Commission has signalled that social media providing data for loans can be treated as agencies with consumer data, and this could change their regulatory status, which could mean a lot more responsibilities and costs.

Experimenting with new methods of assessing loanability takes just a few years. It is too early to prove their credibility and effectiveness. FICO has been used for 60 years. However, new risk evaluation models will continue to develop and improve – such possibilities include technology, including machine learning, which require new mobile consumers. But the real test will be a crisis that once pops up and will verify the real value of the loans given on a non-standard basis.

Get articles like this straight to your inbox each morning with our Breakfast Briefing. Sign up by clicking here!

Log in with your details


Forgot your details?

Create Account

Send this to a friend