Primary tabs

Finance Watch

Finance Watch strongly supports the application of the guidelines to ALL credit providers, regardless of the type of institution. It is important to distinguish between credit offered by banks or by non-banks.
It would be important to make sure that the guidelines apply to all newcomers entering the credit market, in respect of the creditworthiness assessment processes.
Finance Watch would also recommend (for example in section 4.3, but not exclusively) that the EBA should include requirements related to anti-discrimination practices in the guidelines.
Data access, data analysis and risk analysis are all potential areas of risk in terms of discrimination. It would make sense to require that institutions document the way processes have been designed to guarantee a fair and equal treatment of any clients, in line with paragraph 31. Credit scoring, for example, is not completely free from discrimination, because it is not built on samples that are representative of society.
Credit providers should document creditworthiness assessments and explain why they have offered a contract or not. The ability to explain is important. It would be critical if at some point algorithm driven decision-making, or another development, would no longer allow for proper explanation and the necessary ability to check their results.
In the field of the data collected for creditworthiness assessment, we suggest integrating a focus on the necessary quality (non-falsifiable) with a clear capacity to guarantee data accuracy and being up to date, as well as a reminder on the key requirements under the GDPR. In this case the necessary data to undertake the creditworthiness assessment should relate to the budget. Particular attention needs to be paid to this point in the context of big data. In other words, only the use of credit related data and possibly payment transactions data (through open banking, for example) should be allowed in the context of creditworthiness assessment. The use of non-financial data (social media, web browsing, and other tools to track consumers daily life) should be prohibited as they are extremely risky and controversial (intrusive, erroneous conclusions due to the wrong causality links, discrimination, shaping consumer behaviour). The EU industry (credit bureaus and lenders) have an increasing appetite to use this kind of data. In a way, the use of non-financial data could be compared to social scoring in China – a development that we certainly do not want to see in Europe.
Yes. Proper implementation of creditworthiness assessment will have a significant impact on the process during the pre-contractual phase and on the way credit decisions will integrate these new methodologies. Many credit providers (non-bank), specialised in revolving credit, car loans, hire-purchase, short-term high-cost loans (payday loans), personal loans have developed risk assessment (credit scoring) that limits data collection to the very minimum and quality checks related to budget information of the client.
As such, if properly implemented, creditworthiness assessment should impact the business model of a significant part of the credit providers. These providers maximize their profit based on a certain level of defaulted credit (explained, partially by economy of scales for mass credit, because of rather high levels of interest rate and fees and penalties and by the limited time spent on the pre-contractual phase), which might be beyond the level proper creditworthiness assessment should allow.
Very promising options might be proposed in a near future from aggregators (open-banking) which might develop proper- GDPR compliant- accounts analysis, possibly for creditworthiness assessment purposes. This innovation, if designed the right way, should allow access to low cost/ quick / qualitative and accurate information with no necessity to collect data in any kind of “private register”. The consumer’s consent should be very explicit and they should be the only one to be able to share this creditworthiness analysis with the credit providers that they might be willing to contact.
In order to build future-proof regulation and guidelines, Finance Watch underlines the importance of defining the problems encountered by consumers that should be properly measured and therefore monitored through the years.
In this respect, the number of defaulted clients, per category of product, the size of the debt and the duration of the arrears are key indicators that should allow regulators to keep an informed view on the quality of the credit market from a consumer perspective. An important point is therefore for this guidelines to apply to any newcomer, from innovative sectors to open banking and others.
Some additional issues with the requirements which may be inadequately future-proofed include:
- Paragraph 47(b): The fact that data integrity does not equate to data quality. A database may have been securely managed with no data breach or manipulation, but this does not mean that the underlying data is of high quality or suitable for automated decision making processes. For instance, data collected on user behaviour can be of very low quality as it may not accurately reflect the situation of a user. On social networks, people often misrepresent their lifestyles or habits in order to gain social approval. But there are also tools which can be used to manipulate user-generated data like GPS spoofing (virtually manipulating your location), etc.
- Paragraph 47(c): The requirement to be able to explain the outcome of an automated decision, while it is required by the GDPR, is a very utopian requirement which may not be possible in practice. Developments in machine learning and complex neural networks make it nearly impossible to justify what lead to a certain decision by an algorithm, as it can make links between millions and millions of data points which cannot be traced or understood. An easy example is the fiasco caused by the Google Images algorithm which mistakenly identified black people as gorillas. The only way that Google could solve the problem was to shut down the algorithm and change the underlying database which was used to “train” the algorithm with a hope that it would be able to tell the difference. In other words, a complex algorithm is very simple to set up: find links between two sets of data. In this case, it is any data gathered about a consumer, linking it to the likelihood of default. On the other hand, to explain why an algorithm made a certain decision can be extremely complex to explain, as algorithms can make links between data points that are not necessarily intuitive to humans or explainable in any meaningful way from the human perspective. Examples include using data collected by a users’ smartphone to carry out a credit risk assessment. Studies have shown that apparently, simply looking at a users’ call records/patterns is sufficient to predict repayment outcomes. While the algorithm may have found some complex statistical relationship between these two data sets (call records, likelihood of repayment), there is no way to explain it to an individual user who would ask a bank to justify why his credit risk is higher or why he/she needs to pay an extra premium based on his call records/patterns (in other words, explain the causality or rationale behind the decision other than to say that there is a statistical correlation to justify it, which is by no means a transparent or sufficient justification).
- Paragraph 47(d): Being able to compare outputs from automated decisions to traditional methods means that these traditional methods have to be used for a comparable client base. This would mean that the EBA would have to require banks to use traditional methods of creditworthiness assessment on a significant part of their clients (for example 50%). It is uncertain how realistic such a requirement would be.
- The use of such innovations should be limited to basic algorithms examining objective and neutral data that carries little or no risk of discrimination. An example would be examining the budgetary management of a customer over a certain period (1-2 years) (ability to save, spending patterns,…).
- Adding any other criteria may lead to discrimination over race, gender, etc. An algorithm which is trained and programmed to find correlations between likelihood of default and a set of data that it has access to will find any and all supposed links" between default rates and a certain piece of data such as sex, religion, age, race etc, regardless of whether it violates human rights or discriminates against certain people or not. Furthermore, algorithms can also be used to target vulnerable users (such as people with mental health issues, or over-indebted people looking desperately for money) in order to maximize the sale of certain financial products. Placing restrictions on the use of behavioural data is, therefore, key not only with regards to the creditworthiness process, but also in the sales and customer acquisition process."
First, it is important to underline that the paper does not clarify the difference between credit risk and creditworthiness assessments and seems to be using the terms interchangeably. This is not the case!
Credit risk assessments are ways for banks to hedge risk and to protect themselves and their profitability (linked to the prudential side of regulation, making sure a bank does not go bankrupt), by ensuring that within a certain risk pool (consumers with a similar risk profile), the price of the credit (interest rate, fees, credit insurance…) offsets the loss resulting from defaults of a certain percentage of that risk pool. However, it is worth underlining the irony in this risk hedging strategy: by increasing the price of the credit, vulnerable consumers may find themselves at even higher risks of defaulting given the extra financial burden. This creates a vicious circle where by trying to hedge risk by raising the price, banks create additional risk which then has to be hedged by raising prices even more etc. In short, hedging risk via risk based pricing in credit does not work and actually creates more risk (both in terms of consumer protection and prudential risk), contrary to other types of products, such as insurance where increasing the price of motor insurance based on a drivers’ accident history does not increase the risk of the driver having an accident. For instance, the Walloon Belgian Fund (Fonds du Logement Wallon) grants loans to poor/vulnerable families to help them acquire a home through mortgages with very low interest rates. The repayment rate is as high as 97% (on par with NPL ratios in mainstream banks), even if they only target a vulnerable population which would be considered high risk by the private banking sector. This goes to show that risk hedging via higher interest rates is counter-productive. A better strategy is 1) to revert back to creditworthiness as the main criteria for granting a loan (basically, if a consumer is deemed creditworthy, then they should have access to a loan at a similar/same price relative to other customers, in other words, a binary decision: either a consumer is deemed creditworthy and has sufficient revenue and left over income after expenses to cover the reimbursement of the loan or not, in which case the loan is not granted, rather than granted nonetheless but at a higher interest rate), 2) push for more socialization of risk between borrowers (all studies on over-indebtedness show that the main cause leading to over-indebtedness is financial shocks such as loss of employment, life accidents such as divorce or health accidents including death, rather than irresponsible budgetary management all consumers are equally susceptible to these life accidents, which justifies a socialization of risk rather than creating risk pools).
On a more general level, if banks truly hedge default risk by increasing the interest rate, it begs the question of why consumers who default on a loan should be forced to repay it. If the interest rate has been raised to hedge or cover against the risk of default, that means that the consumers’ default has been priced in, meaning that the bank made sure that it would be profitable even assuming a certain number of defaults. In other words, it is as if a consumer paid for a health insurance, but in the case that the consumer broke their leg, they would still need to pay for the health expenses, which defies the very purpose of the health insurance.
The separation of consumers into different risk pools is only an artificial effect of competition, not a prudential imperative. In essence, it is as if health insurance companies only took on consumers with no (perceived) risk and made all other consumers pay extortionate rates for health insurance (as is the case in the US for instance). The European tradition, however, is one of solidarity where everyone pays, more or less the same health insurance premium, regardless of their personal condition, and the price set on the premium is set to cover the overall global risk of an entire population. The same reasoning can be applied to credit risk and risk pricing. This is not to say that anyone, even very poor/vulnerable consumers should have access to cheap loans, but simply that if they pass the test of having sufficient savings capacity (disposable income after all expenses), they should be granted a loan at a similar/same interest rate as other consumers. In that respect, interest rate caps play a role in ensuring that consumers, regardless of their profile, pay a similar/same interest rate. Since banks cannot impose outrageous interest rates, they have to socialize risk to some extent if they want to keep lending to the same consumer segment.
Creditworthiness assessment on the other hand is a way to protect the borrower (linked to consumer protection) and prevent situations of over-indebtedness by ensuring that a consumer that does not have the means to repay a credit given their financial situation is refused access to credit (1) . Indeed, poor consumers are very vulnerable, and at risk of exploitation by creditors. Precarity plays against a borrowers’ capacity to exert their free will and limits their capacity to refuse abusive terms and conditions. In this regard, one clear element missing from the guidelines is to include the definition of a creditworthy customer and what rate of default inside a risk pool qualifies as irresponsible lending. In the case of credit risk, there is no limit to the acceptable rate of default inside a risk pool so long as the revenue from the customers who keep paying offsets the losses from the defaults of the other customers. In the case of creditworthiness (or affordability), on the other hand, there should be a clear definition as to when a bank has engaged in irresponsible lending (for instance, if within a risk pool, over 10% of customers default on their loan, is it considered irresponsible lending? Or 15%?) One useful tool would be to measure an industry average in terms of defaults on all credit products, and impose penalties on banks which deviate significantly from that average.
In the paragraph 33. Finance Watch would recommend to envisage the design of an EU common typology (classification), in particular when we consider products and specific credit facilities. This should allow, in the mid-term, comparisons to be made between the EU member states.
Creditworthiness assessment should lead to an objective diagnostic on financial capacity to repay credit together with the other expenditures (other credit/ other contracts such as PPI) and the existing incompressible expenditures (rent/food/health/mobility…) which, in case of negative decisions, should allow a comprehensive explanation of the decision to the client. A particular focus should be made on the way automated decisions will comply with the obligation to provide a proper explanation.
The Finance Watch recommends that a suitability check be included, which should document credit provider’s information and advice duty, and information on the type of credit proposed considering its use and the circumstances of the consumer.
Finance Watch has some concerns over the compatibility of paragraph 36 of the guidelines with a less risky credit offer, after it includes proper creditworthiness check. As already mentioned, a creditworthiness check might lead to a negative credit decision which should have been positive otherwise. As a consequence, the risk appetite of the provider might be incompatible or at least not aligned with responsible lending practices.
(1) Please refer to the following FSUG paper on responsible lending :
Finance Watch recommends also including a requirement to document the situation of end-users in the paragraph 31 (so as to have an indicator of consumers at risk of over-indebtedness). EBA should at a second stage elaborate a standardised set of common EU indicators (cf. paragraph 33 - “products and specific credit facilities” should benefit from a common EU typology).
In paragraph 35(b) - and to be added in annex 1: credit granting criteria should guarantee sufficient remaining income to allow, beside credit and contract reimbursement, a decent living standard for the household.
In this respect, it should be compulsory to collect data on the composition of households, as it is key to being able to ensure sufficient income and guarantee decent living standards.
Finance Watch is also in favour of including a suitability check of the type of credit proposed. It should be used to document the information and advice consumers should have received in the pre-contractual phase.
In paragraph 41 an addition should be included to ensure that only quality and “non-falsifiable” information is used in the process (not unstructured data as found through the internet and social media).
Section 4.3.3
Institutions should demonstrate that the way they are using data is compliant with the GDPR and antidiscrimination regulation. Providers should be able to clearly demonstrate their compliance with principles such as necessity and proportionality and show how they are operationalised.
In paragraph 59 it is not clear what the rationale is behind using the borrowers’ geographic location in the credit-decision making framework. This could also lead to discrimination based on postcode lottery of currently better-off areas.
In paragraph 639(b)ii it should be clarified if the principle of independence and minimisation of conflict of interest, which references economic interest, covers sales incentives.
Paragraph 76(f) again talks about credit risk and creditworthiness as if they were similar processes. (See above) The following point should be added to address this:
Ensure that creditworthiness and credit risk assessments do not lead to contradictory lending decisions".
To put it in other words: while a client may not have sufficient funds to repay a loan (negative creditworthiness assessment), the credit risk associated with lending at a certain interest rate may still be viable for the financial institution and in line with prudential regulation (creating a risk pool with borrowers of a similar profile and ensuring that overall the borrowers that manage to repay cover for the loss of those who default).
In paragraph 76(g) the definition of "independent" needs to be better clarified. For instance, would credit rating agencies be considered "independent"? Would the Schufa score be considered independent? It is key to define what an independent or second opinion” looks like. From our perspective, it should be defined by law and supervised by public authorities. For instance, using a methodology for creditworthiness assessment which has been validated by a public institution and based on data which is also validated and approved by a public institution. However, “second opinion” are only needed where creditworthiness checks are not properly regulated or undertaken by financial institutions. The right approach must be to make sure that creditworthiness checks are of high quality and independent from the start rather than relying on them as a fall back plan.
At the end of paragraph 76 the following addition is needed:
“Put in place preventive mechanisms for early detection of financial problems and set up a specific unit to explore solutions with customers in difficulty such as putting loan reimbursement on hold, helping the customer with legal and administrative proceedings (obtaining social benefits, any benefits they may be entitled to given there difficult financial situation such as unemployment benefits etc), liaising and cooperating with not for profit or independent, recognized, high quality debt counselling and debt advice services.”"
Clarification is needed in paragraph 82(a) on how this would work in practice. The use of the term high quality is too vague and should be properly defined.
Firstly, given staff turn-over, it may not be possible to remunerate staff based on long-term criteria.
Secondly, does quality refer to compliance with prudential requirements or does it refer to a specific number of non-performing contracts or defaults from borrowers? For instance, in payday lending institutions, while their default rates are high, they still comply with prudential requirements.
The same question can be raised for point (b), where it is again not clear how “credit quality” is defined in that context.
In point (c), the best interest of the consumer also needs to be properly defined.
In order to make these recommendations or requirements operational, rigorous definitions need to be put in place to define when a credit is deemed to be high quality. As outlined above, setting a benchmark such as the level of NPLs for certain products and how they compare to other products could be an objective way to measure credit quality. If the level of defaults for a specific product or in a specific institution is significantly higher than this benchmark, that would signal where there are poor quality products/credit.
Overall, the above requirements are very focused on credit risk and not on consumers or protecting their best interest (protecting them against loan sharks, payday lenders, debt collectors etc). More focus should be put on creditworthiness and what can be done to mitigate negative consequences of defaults and NPLs from the point of view of consumers.
The 5.1.1 section should include a more precise definition of the creditworthiness- it should focus on household budgets (income, contracts, liabilities, incompressible expenditures, remaining income for a decent standard of life).
It should also underline the necessary objective of creditworthiness assessment to guarantee sufficient income to allow a decent standard of life considering a household’s composition (children or other family members - parents...). If this is the intention behind the definition of disposable income, mentioned in the paragraph 98, then it must be made very clear in the guidelines. In this regard, paragraph 98 should not only refer to the borrower’s income, but rather to the borrower’s budget.
Paragraph 99 should also mention the remaining income needed for a decent standard of life, considering the composition of the household. This is aligned with paragraph 109 that mentions “appropriate substantiation and consideration of the living expenses”. If these expenditures are not connected with household composition then they lose all meaning (circumstances will differ if the amount of expenditures considered covers one adult or one adult and two children, for example). Using automated analysis of inflow and outflow from a consumer's accounts via PSD2 may be relevant in this case, rather than trying to manually assess the expenditures of the borrower.

In this respect, 5.1.2 paragraph should include the following points:
- Employment should be understood more broadly as all types of professional activity. The current trend of moving away from traditional labour contracts (no end date) to temporary contracts should be taken into account as banks may consider that temporary contracts are more “risky” than other labour contracts, which would significantly impact on many workers possibility of accessing decent credit conditions.
- Household composition.
- Amount of remaining budget (after incompressible expenditures such as contracts- rent, energy, water,..., as well as other financial commitments, liabilities) needed to ensure a decent standard of life (food, health, education, mobility…).
- Point (f) is too vague and should be removed as it is open ended and could lead to abuse such as using non-conventional data from social networks and other sources.
Several key relevant points have already been mentioned in the response to question n°7.
There is, however, an important concern with paragraph 110. Financial institutions should indeed provision for difficult to predict and sudden events such as the case of the Swiss Franc loans, and to insure against them. This is essential to ensure that consumers are able to keep reimbursing under the same conditions, instead of transferring all the risk to consumers, whilst also expecting them to repay (which introduces prudential risks in the case of mass defaults). Hedging, in this case, would be purchasing special insurance products at the level of the financial institutions and not counting on hedging at the level of the consumer (for instance, counting on the increase in value of property to hedge the risk in case of default and liquidation).
Given the unpredictability of products in a foreign currency and their exposure to currency movements, even with proper hedging Finance Watch considers foreign currency loans by their very nature render it impossible to undertake proper creditworthiness assessment. Indeed, such with products they may be significant changes to the total cost of the product that might affect the capacity of the borrower to meet their liabilities, even if their income and budget remain unchanged.
These circumstances and the major issues encountered by a large range of EU borrowers (1) , need to be addressed here by introducing a number of changes:
- The risk related to changes in exchange rates should not only be covered by the consumer but also by the credit provider.
To ensure meaningful creditworthiness assessment takes place, provisions in the credit contract should at least: mention maximum increases/ decreases in interest, in duration, in monthly repayments. These predefined scenarios should then be included in the creditworthiness assessment, along with documented reasoning of suitability of the product for the borrower at the time of the agreement).
- These predefined scenarios can also be used to defined what the worst case scenario could be and present to the consumer the maximum costs they could be exposed under different difficult or unexpected negative circumstances. This is the case for so-called ‘accordeon’ variable rate credit in Belgium, which bring not only flexibility but also security by defining the maximum possible cost of the credit in advance.
Under paragraph 118, it would be more relevant to assess the consumers' saving capacity since this reflects the income that remains after all current expenditures. It also crucially does not and should not assume that the consumer can compress one or more of their current expenditures to service a loan due to unexpected developments.
Data regarding missed payments should be treated with great care. Missed payments may reflect many different situations: an unfortunate circumstance (forgetting to repay, missed correspondence or mail...), or a conflictual situation with a provider (for instance, an error in the amount asked for utility bills and a consumer withholding payment until the problem is resolved, which may appear as a missed payment where the consumer is at fault, but is actually directly due to the provider).
In section 5.2.3, there may be a need to include a general remark on the fact that consumer credit origination should not only be backed by a guarantee but should be based on evidence of creditworthiness. In this respect, if the decision to offer credit is first and foremost based on the creditworthiness of the borrower, we should expect proportionality from the provider for requirement on the borrower to also guarantee the loan.
In paragraph 117 and more broadly in the guidelines as a whole, a key aim to foster must be for credit providers to design and adapt credit offers to new circumstances (type of jobs, incomes,…). This is important to bring more agile and user-friendly products that can help to build consumer trust.
(1) The recent experience in Central and Eastern Europe is an excellent demonstration of massive negative impact on consumers as they are unprotected against currency fluctuations. As we know, all regulators have by now prohibited (mostly not by law) the issuance of new FX loans. But this should be clearly and explicitly provided for at the EU level policy documents. Responsible product design should be part of the concept of ‘responsible lending’.
A clarification needs to be included in paragraph 182 on the credit decision making process between acceptance and refusal thresholds, linked to default rates. For instance, at which rate of default inside a specific risk pool do offers of credit start to be considered predatory lending or irresponsible lending? For instance, if inside a risk pool over 20% of consumers are considered to be likely to default on their loans, is this considered predatory lending?
Taken from the creditworthiness perspective, it is important to define how much of the borrowers' current savings capacity can be taken up by the reimbursement of a credit without creating a substantial risk, especially in the case of a financial shock (loss of employment, health problem, divorce etc).
A further point to consider in paragraph 182, is the importance of factoring in the constraints set out in macroprudential policy, where this is the case. Several EU Members States have put forward macroprudential policy tools (1) , some more binding than others, which restrict the provision of loans by setting limits on ratios such as debt-to-service ratios, loan-to-value measures, and on maturity. The guidelines should reflect the fact that decisions to offer credit are also bound by macroprudential measures (in the jurisdictions where such measures are in place).
In paragraph 183 the information on the key features of a loan being offered to the borrower should include the APR. The APR is a key indicator that helps to reflect the actual cost of a loan and is a fundamental piece of information to be able to compare different proposals (as referred to both in the CCD and the MCD).
(1) See ESRB’s report “A Review of Macroprudential Policy in the EU in 2018”, in particular Annex 2
Pricing of consumer credit should exclude (by design) any risk of discrimination, based on protected characteristics. There may be a risk of discrimination arising from creating “homogeneous risk groups” for example that should be considered. Pricing policies should be documented in order to allow a compliance check by the relevant authority to take place.
Paragraph 187(b) mentions that creditors should take into account behavioural assumptions in determining the cost of funding. This needs more clarification.
All of the provisions in this section have a limit: that of pricing with the context of current experimental monetary policy (QE), which artificially inflates the prices of real estate and the stock market. This is completely disconnected from fundamentals and the impact it may have in case of a massive recession or collapse of the “everything bubble”, which refers to the inflation of several assets among which, mostly, the price of stocks (thanks to massive corporate buybacks facilitated by QE) and the inflation in the price of (existing) property and real estate. Both of these assets are completely disconnected from economic fundamentals (for instance, stock prices relative to company earnings) and any reversal in the policy of central banks would precipitate their depreciation, risking the triggering of a massive crisis (the alternative being a “Japanification” of Europe- 30 years of economic stagnation with baseline central bank interest rates at negative or 0%).
In paragraph 204 the reference should be to paragraph 200 (not 2000).
In paragraph 229(e) the monitoring of credit risk and especially of NPLs across comparable consumer segments could be a basis for defining predatory lending. If any financial service provider has an NPL ratio which significantly negatively deviates (from a statistical point of view) from its competitors (average), then their lending practices should be closely investigated and considered inappropriate.
In general in this section there was no mention of prevention measures such as identifying consumers in financial difficulty before they default (1) on their existing financial commitments and a dedicated unit which deals with helping consumers in distress.
In paragraph 263 the early warnings seem to apply mostly to professional clients and not individual consumers. It would be important to add indicators such as a drop in the consumers' ability to save, or a lack of financial buffer (living from salary to salary).
In paragraph 266 the plan should also include the write off of part or all of the debt. This must also be a part of the solutions proposed.
Credit refinancing procedures (especially in the case of mortgages) should be properly considered. If the same strict standards are to be applied as they should with any entirely new credit or higher credit amount, too strict conditions could cause borrowers to default instead of preventing it. Especially if consumers were able to repay their previous instalments before the refinancing, despite having lower assumed creditworthiness. This would as well be in line with Article 28 of Directive 2014/17/ЕU (Mortgage Credit Directive).
In line with Article 28 of Directive 2014/17/EU, lenders should exercise reasonable forbearance and try to prevent credit contacts from becoming non-performing.
(1) French consumer credit regulation has introduced the notion of “fragile client” that should be identified before default occur, and for who adequate solution should be proposed to avoid a deterioration of its financial situation. – legal reference:
Arrêté du 9 mars 2016 pris en application de l'article R. 312-13 du code monétaire et financier et fixant la liste, le contenu et les modalités de transmission des informations transmises à l'Observatoire de l'inclusion bancaire – available on :
Finance Watch