The generic assumptions underpinning traditional risk assessment – often barely fit for purpose as it is – have been further undermined by Covid. Credit risk in particular requires a significant overhaul to help protect net interest margin (NIM) and repair banks’ balance sheets, reports Douglas Blakey
Credit bureaus and banks are innovating but arguably the most meaningful developments emanate primarily from new players leveraging new types of technology, stressing the importance of a partnership-led approach.
The market leaders for alternative credit risk comprises a vast ecosystem of provider types and business models. Category leaders include big tech players such as Alibaba, Tencent, Amazon, Baidu, Mercado Libre and Microsoft.
Alternative lenders in the fintech sector include LendUp, Rocket Mortgage, OnDeck and Prosper.
And then there are the credit bureaus, FICO, Equifax and Experian and major vendors Temenos, Finastra, SAS, Fiserv and FIS.
So, the market for alternative credit scoring comprises a vast ecosystem of provider types and business models. Much of it exists, by definition, outside of traditional financial services in the form of alternative lenders, niche fintech providers, and big tech. However, incumbent credit bureaus and banks are also innovating, while challenger banks and credit bureaus are bringing a variety of new data-driven decision-making techniques to market – especially in emerging markets.
Orthodox credit risk models penalize and/or exclude large segments of society
Incumbent financial services providers have traditionally relied on a limited range of data types when assessing the credit risk of individuals and SMEs. In the US, Canada, the UK, and Germany, creditworthiness is determined primarily based on credit scores provided by large credit reporting agencies such as Equifax, Experian, and TransUnion.
Other countries such as Brazil and Australia are gradually transitioning towards similar systems based on “positive” credit scores. Meanwhile, in France and Japan lenders primarily focus on employment history and income, with repayments/defaults used as detracting factors.
These incumbent credit scoring systems typically work well for financially conventional, well-off consumers, but large segments of the population worldwide are penalised when accessing credit or are denied access altogether.
For instance, in the US those with a limited credit history – such as first-time borrowers and noncitizens/non-residents – often struggle to demonstrate sufficient credit history. Similarly, in France and Japan, those who are not employed full-time or who do not rely on a single employer for their income (eg. the self-employed, gig economy workers) often struggle to obtain credit. In developing economies, where large segments of the population remain unbanked or underbanked, these credit scoring systems can exclude millions of people.
Now, amid Covid-19, the millions of otherwise careful, frugal customers, who scored well under traditional models have become non-creditworthy. Continued reliance on traditional risk measures could prove procyclical, choking off credit supply for those struggling small businesses that will be so critical to economic recovery.
Government schemes to provide life support, while helpful, complicate credit risk further as they obscure true economic indicators such as current delinquencies typically used to project future losses.
Traditional credit models rely on inputs about the presumed macroeconomic forecasts, which typically use traditional economic theory concepts of general or partial equilibrium. These forecasts may be completely unreliable as the artificial shutdown of many consumer goods and services markets has pushed the economy into a state of disequilibrium.
The need for alternative data
Against this backdrop, the need for alternative data not necessarily directly related to applicants’ financial history is more important than ever to help enhance the predictive power of credit risk modelling. In turn, traditional lenders and credit bureaus are extending the data pools relied on for scoring to include utility and rent payments, for example.
Alternative lenders and big tech, meanwhile, have developed fundamentally new approaches to credit assessment, incorporating a wider variety of data sources, such as history of utility bill payments (electricity, gas, telecoms), rent payments, repayments to payday lenders, social media data, browsing and search history, employment history, educational background, and even psychometric testing.
Key technology trends
A number of key technology trends are impacting the alternative credit scoring theme. For example, there is a growing use of advanced analytics for sub-sector and individual-specific risk assessment.
Meantime, there is a growth in performance disparities between machine learning algorithms and traditional scorecard approaches. New credit scoring models used by fintech lenders rely on machine learning techniques rather than traditional loss and default models.
The prediction capability of machine learning models has been reliably demonstrated in stationary external environments, but performance under dual structural shocks –say Brexit and COVID in the UK – remains far more contested. And there has been an exponential increase in the volume and variety of data usefully analysed by non-traditional providers.
Big data technology allows alternative lenders to collect and use a larger volume and variety of data points. For example, Ant Financial and Mercado Libre claim that their credit quality assessment typically involves more than 1,000 data series per loan applicant.
GlobalData’s Thematic Research team’s report entitled ‘Alternative Credit Scoring’ ranks all of the major players based on overall leadership in the ten themes that matter most to the industry, generating a leading indicator of future performance. Given the size of the market and the potential for challengers to disrupt the market, it will be one of the most read reports in the first half of the year on the GlobalData Banking and Payments Intelligence Center.