What is Credit Scoring
Everyone will face Credit Scoring in their life, even if they never have to take out a loan. Credit Scoring originally came out of the need for banks and other lenders, to assess how likely customers are going to repay them. To simplify their internal processes and be leaner, they outsourced this task to credit scoring companies that maintain customer credit track records and complete customer evaluations based on proprietary formulas. With this, they can provide a number from 1-1000(850) to banks and other lenders, Indicating levels of reliability.
Despite being originally intended for assessing one’s loan-taking capacity, right now credit scoring is used in all financial activities, from signing a new mobile contract, to renting out an apartment. It is now a way to know if customers are financially responsible (Sean LaPointe), with it all resting on a 3-digit number. Therefore, having a good credit score is often more lucrative than having an excellent job (Experian).
Yet, despite having a wider adoption than ever, the way credit scores have been calculated and the data used to do so have hardly changed.
What is wrong with credit scoring?
Right now, there are three major credit-scoring organisations: Equifax, Experian, and TransUnion. Together they do most of the credit scoring for the USA and UK and are the main trusted source of information about you, the customer, for lenders. To calculate the score, these companies use several models, FICO being the most popular. In it, they mostly assess how well you have repaid previous loans as well as what types of loans have you had and when.
What is surprising in this model is that it only uses past loans to assess future ones. This results in situations where a person with a well-paid job and savings who lives without credit has a lower score than someone who spends all their income on repaying credit for previous loans. This situation has recently caused an uprise of financially stable people taking loans, despite having plenty of funds, only to boost their credit scores (Emma Woodward).
We can do much better
This is obviously a worrying sign. Not only do we have entry barriers for financially stable people to get a loan, but people are now generally incentivised to get deeper into debt. Of course, it should not be like that. Thankfully, there is something we can do about it.
Every day any given consumer generates data that can be used as a clear indicator that they are trustworthypayers. From the way one spends money, activity participation in free time and even social media activity. Allthis paints a much better picture of whether you will be responsible for your debt or not. Moreover, this information can quickly adjust to new life conditions, compared to old credit scores which are mostly static unless you have an active line of credit.
It has also been shown that the usage of alternative data, like the ones indicated above, can drastically improve the quality of credit scoring, with reports of over 50% improvement. (Credit Scoring with Social Network Data, Retail credit scoring using fine-grained payment data). And in the days of Big Data, there are no limitations in on building new scoring systems —it is very much possible.
The new approach would be a great win for many people, especially young individuals who have not yet taken out loans but already have a strong profile. However, we have still not yet seen systems harnessing these benefits, and there is a reason for that — Privacy.
The privacy dilemma
Clearly, there is an abundance of data that can be used to obtain more accurate credit scores, however, this data is usually very sensitive. For example, would you be okay with sharing information about every phone conversation had with an external party so that they would calculate a better credit score? Probably not, especially if you consider that they might also eavesdrop on you and extract information to sell onwards to advertisers. What about sending over your Apple Watch health and location data? Or all of your bank transactions?
This privacy concern has been the major roadblock. Furthermore, even though there are models that could extract credit scores from this data, we are still living with old rusty credit scores. Nevertheless, there is hopeful light on the horizon.
In the last 10 years, there has been a rapid rise in the development of privacy-preserving computation tools. These are tools that allow execution algorithms over private data without ever risking exposing the data.
In our case, it would work as follows:
You would instruct your phone provider to share your encrypted call details with the credit scoring agency. They will then be able to run the credit scoring over the encrypted data, not knowing who you ever called. But as a result, they will get a vastly improved credit score. A win-win situation for both parties. And this can be done with any kind of data, and even any type of analytics models. Most crucially yet, you can be sure that the personal data you submit always remains private.
Today, there are two main directions of doing such private computations — software and hardware-based. The software approach is based on cryptographic techniques, including solutions such as Multi-Party Computation (MPC) and Fully Homomorphic Encryption (FHE), still very early in development. The hardware approach consists of special chips called Confidential Computing units which have already been used in the real world to secure sensitive data during computation. The latter technology is currently the most promising candidate to be used in building the required improved credit scoring model, fully fit for the modern day.
What will be our future?
There is emerging and convincing evidence (Credit Scoring in the Era of Big Data) proving that the new era of credit scoring is not too far away, hopefully seeing changes in the next decade.
A lot of banks and private lenders have realised that credit scores still provide too little information. Because of this, they are actively seeking access to the data themselves. Data privacy, again, becomes a big problem.
However, it is reasonable to assume that with private computation technologies, this will also change, and we will see an increase in data exchanging-related activity. With our consent, our encrypted data could be anonymously shared between services, so they provide better insurance quotes, mortgages, buy-now-pay-later offers, and much more.
Living in the Big Data age, the more data we get access to, the better the services we will receive. And privacy, the only major bump in the road, seems to have been smoothed out.