Caught between the risks of inflation and economic recession, the health of the global economy is likely once again to rely heavily on the health of the American consumer. And, as we know, American consumers calibrate their consumption based on their ability to access credit. This is what inspired us to present the revolution that artificial intelligence has made possible in the credit-scoring of American consumers. Its main societal benefit is in helping to grant credit to consumers who were excluded in the past, due to the limitations of traditional credit-scoring models.
The limitations of traditional models for scoring American consumers’ creditworthiness...
Consumer access to credit is governed by a higher level of formalism in the US than in Europe. Most credit-granting banks and organisations in the US use a FICO Score. This is a three-digit score (between 300 and 850) that is monopoly-held by the listed company Fair Isaac Corporation (FICO US). FICO Scores, which date back to 1989, are calculated on the basis of third-party data provided by the three national credit reporting agencies ( TransUnion , Equifax and Experian ). They are based on data such as your borrowing history (regularity of your repayments, composition of your loans, etc.) and the gross amount of your liabilities. They are the pillars for evaluating consumers applying for new loans and for analysing risk on loans already taken out.
These traditional models for assessing consumer solvency are now reaching their limits, which are: 1) societal (a large proportion of millennials do not use credit cards and therefore have not created a borrowing history, although they may decide to save, making them borrowers with good repayment capacities); 2) holistic (credit- scoring of American consumers in the 21st century must integrate many more parameters than those initially provided for by the FICO score); and 3) financial (traditional FICO models focus on consumers' ability to repay their loans on time, rather than on their overall cash flow at a time when it is possible, for example, to repay loans with other loans). The ultimate goal is to reintroduce more fairness into the system so that good borrowers stop paying undue premiums on the cost of their credit and that, conversely, more dubious debtors can continue to access credit.
...are now being overcome by artificial intelligence
In a world where 80% of American consumers have never defaulted on a loan and yet only 48% of them have access to credit at optimal cost, the fintech Upstart is introducing an AI-based technological disruption to traditional methods of credit-scoring American consumers. Over the past eight years, Upstart has built a dataset of 1,500 variables (compared to about 20 for the FICO score). Artificial intelligence algorithms have been trained on more than 20 million repayment sequences, establishing rules and correlations that did not exist in the past and creating a barrier to entry for Upstart’s would-be competitors. Indeed, the use of hundreds of thousands of datapoints would be impossible for credit-scoring without sophisticated machine learning models. This allows Upstart to introduce credit-scoring criteria, most of which do not exist in the FICO score, such as the nature of your job, your academic background, your banking history, your standard of living and even your interactions with consumer credit websites. In short, artificial intelligence enables a much more holistic approach to credit-scoring American consumers, resulting in greater fairness in everyone's ability to borrow.
Upstart's artificial intelligence platform is available to partner banks and credit institutions via a cloud-based application. Consumers can access its scoring service (upstart.com) on its website or the websites of partner banks.
