Authors: Hans-W. Micklitz and Giovanni Sartor

 

I. Introduction

The idea of linking lending to a standardised assessment of the debtor’s creditworthiness is roughly one hundred years old and goes along with the rise of lending in the industrial age. In the consumer society, the creditworthiness test has gained importance and has reached ever broader societal areas. The ‘credit score’ has turned into a decisive element for purposes far beyond lending. It plays a role in the relationship between landlord and tenant, employer and employee, insurer and insured, etc. It has also turned into a benchmark that determines access to many potential economic and social services. Commercial banks have outsourced the creditworthiness test to private companies, and States have decided to entrust their central bank with creditworthiness assessments. That is why there are two types of so-called credit bureaus in Europe: public and private. The collapse of Lehman Brothers and the euro crisis have put household debt in the limelight of central banks and the World Bank. The two events have shown that household debt may endanger the financial stability of the euro and national currencies. However, they have not yet led to the adoption of comprehensive EU legislative measures aimed at ensuring greater transparency in the credit domain, and at regulating the granting of loans, and in particular creditworthiness assessments. The respective criteria are essentially in the hands of national states.

The 2008 revision of the Consumer Credit Directive (2008/48/EC)[1] introduced in Member States the obligation of creditors to test the creditworthiness of consumers and to inform the latter about the result. The respective national databases must be accessible in case of cross-border credit (Art. 8 and Art. 9). The 2014 Mortgage Credit Directive (2014/17/EU),[2] adopted after Lehman and the euro crisis, went one decisive step further and introduced the principle of responsible lending by requiring Member States to ensure that lenders only make credit available to consumers when the creditworthiness assessment indicates that the obligations incurred under a credit agreement are likely to be met in the manner required under that agreement (Art. 18 para. 5). At some point in time, it seemed as if the ECB would set standards for credit scoring and supervision over credit bureaus. However, this initiative has not yet been adopted. Recently the Comm has put forward the proposal for new Directive on Consumer Credit, which has not yet been adopted at the time of writing.[3] Art. 18 of this proposal establishes the lenders’ “obligation to assess the creditworthiness of the consumer,” specifying the kind of data to be processed for this purpose (“relevant and accurate information on the consumer’s income and expenses and other financial and economic circumstance”) as well as consumers’ rights relative to automated decision making (according to Art. 22 GDPR). Art. 19 grants cross-border access to databases used for assessing the creditworthiness of consumers.

Credit scoring is mainly motivated by the prospective creditors’ interest in not entering risky loans. However, it has normative significance from multiple viewpoints.[4] It involves considerations of allocative efficiency, i.e., ensuring that capital is allocated to the most economically valuable projects, i.e., to borrowers who are most likely to repay in time, to minimise inefficiencies due to nonperforming debts. It also involves issues of distributive fairness, which point to two, potentially divergent, goals. On the one hand, credit should only be provided to borrowers who can afford it, i.e., who can repay their debt without suffering financial distress (as happens when vulnerable individuals are exploited by unscrupulous lenders). On the other hand, credit should be provided at adequate conditions to vulnerable individuals, where this does not entail excessive risk for the lender. Finally, the assessment of creditworthiness involves issues pertaining to data protection and privacy in relation to prospective borrowers, since an assessment of borrowers’ conditions may involve the processing of a large amount of personal data.

The amount of borrowers’ data being processed is increased owing to the greater availability of personal data and of technologies for processing such data. In the era of local community-based banking, the assessment of prospective borrowers’ creditworthiness was mainly performed by local officers, whose judgments were based on personal knowledge, shared experiences with similar borrowers, and general information. The process has subsequently become computerised, leading automation of creditworthiness scoring. The borrowers’ scores were initially computed using relatively simple statistical methods, such as linear regression, and were based on a limited set of data, mostly consisting of financial information, such as the applicants’ credit history. Even today, many credit bureaus often use relatively simple mathematical models.[5] However, a significant evolution is now taking place, thanks to the increasing amount of personal data being available and to advances in computer technology. More complex technologies – including advanced machine learning techniques, among which are deep neural networks – are beginning to be applied to larger and more diverse datasets, not limited to financial data. The data being processed may include, for instance, education and employment histories, as well as behavioural data, i.e., digital footprints6 that have been collected by observing data subjects’ online behaviour, such as internet browsing, mobile phone usage, or records of fitness activities.

It has been claimed that expanding traditionally used data with digital footprints enables lenders to reduce default rates while expanding credit to groups of people that would otherwise not have accessed credit due to low credit scores assigned to them by credit bureaus working with more limited data.[6] While the use of behavioural data for credit scoring is still limited, such data are already being used by several lenders in the EU – including FinTech start-ups, payment service providers and even insurance operators –usually in combination with the data provided by credit bureaus. The use of non-financial data for the purpose of credit scoring raises concerns about the freedom of prospective borrowers, since they could be induced to adapt their choices and habits in everyday life as needed to obtain favourable credit scores, to the detriment of their autonomy.

 

 

 

[…]

 

 

 

[1]Directive 2008/48/EC of the European Parliament and of the Council of 23 April 2008 on credit agreements for consumers and repealing Council Directive 87/102/EEC, OJ L 133, 22.5.2008, pp. 66-92.

[2]Directive 2014/17/EU of the European Parliament and of the Council of 4 February 2014 on credit agreements for consumers relating to residential immovable property and amending Directives 2008/48/EC and 2013/36/EU and Regulation (EU) No 1093/2010 OJ L 60, 28.2.2014, pp. 34-85.

[3]Proposal for a Directive of the European Parliament and of the Council on Consumer Credits COM/2021/347 final. The EU institutions came to a political agreement in the trialogue, however, there are still many so-called technical questions open.

 

[4]Aggrawal, ‘The norms of algorithmic credit scoring’, Cambridge Law Journal, 80, pp. 47-73, 2021.

[5]SVRV, Consumer-friendly scoring, Sachverständigenrat für Verbraucherfragen, 2018, available online at https://www.svr-verbraucherfragen.de/en/wp-content/uploads/sites/2/Report-2.pdf.

[6]Berg/Burg/Gombovic/Puri, ‘On the rise of fintechs: Credit scoring using digital footprints’, Review of Financial Studies, 33, pp. 2845-2897 (2020).