The promise of credit: creditworthiness assessment in the age of big data

Combining traditional information from credit history with utility bills, social media tracks and cookies, big data are predicting our ability to repay our last car loan and our future mortgage. Policymaking actions are still not in sight. A few reflections on credit scoring and creditworthiness assessment in the age of FinTech.

Creditworthiness assessment lies at the heart of the financial service industry and is part of bankers’ daily life. From the private individual seeking a facility to refurbish its apartment, to the large corporate customer in need for financing its brand new production line, the analysis of the effective probability of repayment is the bulk of the circle linking short-term deposit-taking and long-term lending. The estimated probability of default of a borrower leads to a “Yes” or a “No” by the lender, and – when a positive answer is on the pipeline – to a certain labelling in terms of pricing/risk based on credit scoring techniques and decision trees models.

Given their importance for the money industry, poor credit assessments can have a substantial negative impact on the overall smooth functioning of the financial system: on the one hand, these can lead to the building up of substantial risks in loan portfolios, thus potentially prompting financial turmoil at institution-specific or market-wide level; on the other hand, they can cause credit discrimination, financial exclusion and inefficient pricing policies. Under an individual customer perspective, positive assessments can also exacerbate the risk of over-indebtedness: consumers might be potentially incentivized to overestimate their capacity to repay when facing a positive albeit flawed credit decision, under the assumption that a lender’s evaluation is expert-based and unbiased.

Credit scoring can (and should) trigger the decision to provide credit on the basis of a number of relevant factors, including debt history, other customer financial commitments, as well as stable sources of income. The factors taken into account considerably evolved over years: while in the early ages of the modern financial system information were limited, paper-based and often costly to retrieve, at present we find ourselves in an age where relevant data are way more easily and readily available. In a digital environment, as all data are “relevant data” when it comes to marketing of products and services, the financial domain has discovered that all data is “credit data”.

As one may guess, FinTech and the big data society have thus substantially changed the approach to credit scoring at least in two ways.

On the one hand, credit assessment is increasingly relying on a larger and larger amount of diversified and paperless data points and sources, including the digital footprints that we leave behind us whenever navigating the web, using social networks, and buying our favourite shoes on e-commerce shops. Credit institutions are also starting to make wise use of data from text message activity, mobile phone use and utility bills payments to capture additional information.

On the other hand, credit scoring has been enhanced by the sophistication of algorithms, which are increasingly able to factor in a wider variety of elements to refine credit worthiness assessments and establish more reliable predictive models. As noted by the Fsb, machine learning and artificial intelligence allow massive amounts of data to be analysed very quickly. As a result, these could yield credit scoring policies that can handle a broader range of credit inputs, lowering the cost of assessing credit risks for certain individuals, and increasing the number of individuals for whom firms can measure credit risk.

The Fsb also remarked that the application of machine learning algorithms to this large set of new data has enabled assessment of qualitative factors such as consumption behaviour and willingness to pay. The ability to leverage additional data on such measures allows for greater, faster, and cheaper segmentation of borrower quality and ultimately leads to a quicker credit decision. However, notes the Fsb, the use of personal data raises other policy issues, including those related to data privacy and data protections. Questions hence arise on the drawbacks and risks of this transformation, as well as on the effectiveness of such sophisticated techniques to predict our future credit history.

A key element to highlight is that the increased digitalization and data-driven transformation of credit scoring give rise to a substantial loss of control and understanding of the dynamics underlying a lending decision. As the credit scoring industry growingly relies on a combination of data retrieved from different sources and combined in an overly complex manner, individuals miss the opportunity to effectively influence their financial destiny, facing valuations that are everything but transparent. At the same time, advanced credit scoring models have the potential to perpetuate spirals of indebtedness, by predicting the behavioural patterns of consumers and exploiting their vulnerabilities: aggressive lending policies might hence target certain groups of vulnerable consumers by tracking their actions, pushing them to subscribe piles of renewable debt on a recurring basis.

In this respect, it has been pointed out that big-data credit scoring tools may lead to the establishment of a system of «creditworthiness by association» in which consumers’ interpersonal and social affiliations determine their eligibility for a certain credit facility. These would also result in potentially obscure discriminatory and subjective lending policies, eventually not emerging as a result of deliberate choices by the lender or the credit service providers, but as an ineffective combination of neutral data points by underlying algorithms. The latter might hence give rise, for example, to either the systematic denial of access to credit or the application of higher interest rate and pricing bands to certain groups of borrowers.

In the European Union, these problems are far from being addressed by policymakers. In its recent Report on Big Data and Advanced Analytics, the European Banking Authority (EBA) took note of the growing use of big data and analytics tools for risk-scoring by financial institutions, and highlighted that these might benefit both credit institutions’ customers and non-customers: while the first group might access pre-approved instant loans on the basis of data directly held by their institution, non-bank customers might more easily obtain credit facilities through the integration with the access to payment data through the APIs (application programming interfaces). Nonetheless, no specific policy actions seem to be at sight.

Credit scoring is hence another small albeit intriguing area in which the questions and problems on the table create an inextricable node between different conflicting interests. As in many other fields, technology pushes and blurs the lines between the private individual domain and the publicly available sphere, while the market evolves along the subtle line of legitimate exploitations of technological efficiency-driven advancements and increasing intrusions in private life. On its turn, regulation seems to be unable to keep pace of these evolutions, as the usual tools of command, control and enforcement are ill-equipped to cope with overly complex challenges: to the detriment of vulnerable individuals, missed actions are bound to easily turn into missed opportunities.