Credit institutions using the Internal Ratings Based Approach (IRBA) are permitted to estimate their own risk parameters and, in doing so, require significantly less capital than institutions using the standardised approach – provided that their models meet all regulatory requirements. One such requirement, paragraph 59 of EBA Guideline EBA/GL/2017/16, stipulates that credit reference agency information used in the rating must not be more than 24 months old.1 For a specialised institution with infrequent customer contact, such as a building society, this is no trivial detail: It determines whether a highly discriminatory external score, such as the SCHUFA score, can be used economically at all in the ongoing portfolio rating. Alte Leipziger Bauspar AG (ALB), with the support of msg for banking and in collaboration with SCHUFA, has developed a solution that meets this timeliness requirement without burdening the institution with uneconomical data queries. How does it work?
In short: The EBA Guideline EBA/GL/2017/16 stipulates that credit reference information used in the rating must not be more than 24 months old. Alte Leipziger Bauspar AG meets this requirement in its existing credit assessment by reporting to SCHUFA, on a quarterly basis, only those cases whose SCHUFA score would exceed the 24-month limit in the following quarter. This ensures that the SCHUFA score remains a highly discriminatory component of the credit assessment process without the need to regularly re-check the entire portfolio.
Advantages of the IRBA and regulatory challenges
Credit institutions using the IRBA have a decisive advantage over those using the standardised approach: They are permitted to use their own estimates for risk parameters and thus – provided their portfolio and model quality are sound – require significantly less capital than institutions that rely on the standardised risk weights of the standardised approach. This difference can be particularly significant in retail banking, where portfolios of consumer loans or mortgages typically account for a large proportion of the bank’s exposure. Even partial use of the IRBA on such portfolios provides a noticeable relief to the capital base.
The challenge is that IRBA models must not only be developed and authorised, but also continuously maintained, validated and refined. Furthermore, they must meet a wide range of regulatory requirements that are simply not covered by the standardised approach. This sometimes leads to conflicts between model quality and regulatory feasibility.
In practice, Alte Leipziger Bauspar AG has resolved this dilemma with the support of msg for banking and in collaboration with SCHUFA. msg for banking’s key contribution lay not only in its regulatory expertise, but also in its forward-looking and pragmatic approach: A model had to be developed in such a way that an external score, once calculated – the value of which may, under regulatory rules, be reused in the ongoing portfolio rating for up to 24 months before it needs to be recalculated – is statistically sound and regulatorily verifiable. The solution demonstrates how model development, process design and operational feasibility can be considered together from the outset.
The SCHUFA Score: too good to leave out
When the IRBA rating procedure was being redesigned, it quickly became clear that the new SCHUFA Score is a key component of the rating process, particularly for customers who have not attracted attention due to negative payment behaviour at the time of the rating. It consolidates external information on a borrower’s creditworthiness – data which the institution simply does not have at its disposal in this form. In univariate analysis, the score demonstrates a high degree of discriminatory power between good and bad debtors. Within the model, it significantly boosts overall performance. msg for banking supported Alte Leipziger Bauspar AG in ensuring that this technical strength was not viewed in isolation as a modelling issue, but was linked from the outset to the requirements for data availability, process design and subsequent regulatory traceability.
The SCHUFA score is standard practice in the application process anyway: before a loan is granted, the building society checks with the credit reference agency. The credit reference data is factored into the credit decision, and the score is thus available in the database.
The key risk drivers in the portfolio rating are based on individual payment behaviour. However, customers with no payment defaults are hardly differentiated at all. The obvious idea:
Why not use the SCHUFA score – which allows for differentiation even among borrowers with unremarkable credit histories – in the portfolio rating, i.e. in the ongoing risk monitoring of existing exposures?
The regulatory constraint
The EBA Guideline EBA/GL/2017/16 sets clear limits in this regard. Paragraph 59 stipulates that credit reference agency information used in the rating must not be more than 24 months old. This may sound manageable at first, but for a specialist institution which, due to the nature of its business model, has little ongoing customer interaction, it is primarily a question of cost-effectiveness.
Unlike a main bank, which keeps a constant eye on its customers via current accounts, cards and day-to-day payment transactions, a building society’s business model is geared towards the conclusion of contracts and the subsequent processing of loans. Data does not update itself in this context. There are few day-to-day transactions that serve as a natural source of information. A systematic bulk comparison with SCHUFA would be technically feasible, but would not make economic sense for a portfolio of long-term building society loans: the effort involved in carrying out quarterly or annual credit reference agency enquiries for tens of thousands of borrowers would be disproportionate to the information gained.
The result so far:
The SCHUFA score is typically only used at the application stage. Its potential remains untapped in ongoing credit assessments.
The approach: Only retrieve what is really necessary
The solution, developed in collaboration with SCHUFA, is essentially simple: Instead of periodically querying the entire database, only a specific set of cases is selected on a quarterly basis – those whose SCHUFA score would exceed the 24-month limit in the following quarter. The institution carries out this selection internally and transmits the relevant identification numbers to SCHUFA. SCHUFA enriches the data records with up-to-date score values and makes them available for download. The institution imports the data, and the score is then valid for regulatory purposes for a further 24 months.
The principle: Do not base the model on the most recent version; instead, retrieve the data specifically when the requirement for up-to-date information dictates. With a 24-month cycle and quarterly queries, in principle around one-eighth of the portfolio reaches the threshold each quarter. In practice, the figure is significantly lower, as cases are excluded that have since been repaid or have been re-queried for other reasons, such as a new application or a significant payment arrears. Costs remain manageable. And the regulatory requirement is fully met.
What this means in practice
The process resolves several issues simultaneously. It retains a component in the portfolio rating that has been proven to be highly discriminatory and which would otherwise be omitted. It keeps costs under control by limiting queries to the regulatory minimum. And it creates a documented, traceable data path, which is of considerable value in model documentation, internal validation and supervisory discussions.
The real crux of the matter lies in a modelling decision that runs counter to the usual instinct: The SCHUFA score is not treated in the model as a continuously updated value, but as a component of the rating process valid for up to 24 months. This is what enables the consistent integration of model development and operational use, but it must also be implemented accordingly in the data preparation for model development. In this way, the institution can demonstrate that a score treated in this manner retains its discriminatory power over the retention period and meets the regulatory requirement for timeliness, and that the process is designed to be robust from the outset with regard to model validation and supervisory communication. In practice, the key to successfully translating good modelling ideas into practice lies in thinking beyond the existing implementation and taking operational implementation into account.
A blueprint for other institutions
The situation is not unique. Many specialised institutions, such as car finance companies, mortgage lenders, sales finance providers and specialist consumer credit lenders, face the same challenges: They have little ongoing customer interaction, rely heavily on the data collected at the time of application when assessing creditworthiness, rely on external scores that provide genuine insight, and at the same time face growing regulatory requirements.
For such institutions, this approach could serve as a blueprint – not just as a process template, but as a methodology: the model, data acquisition and rating operations are considered together from the outset, rather than being tackled one after the other. The regulatory requirements are the same for all these institutions. The solution is transferable. Institutions can benefit from a more robust portfolio rating using a credit reference score (which is only updated when its validity expires), improved risk management and, ultimately, sound capital adequacy.
This also demonstrates that:
An IRBA model is only as good as the process that underpins its operation.
Those who take a holistic approach to model development and operational implementation from the outset not only create better models, but also establish the transparency and data foundation that are essential for sound risk management and robust communication with the supervisory authorities.
Support for model development and revision
More on SCHUFA and the IRBA rating method
Further insights into the interaction between SCHUFA data and IRBA models:
Frequently Asked Questions
What does EBA Guideline EBA/GL/2017/16 stipulate regarding the timeliness of credit reference data?
Paragraph 59 of the Guideline requires that credit reference information used in credit ratings must not be more than 24 months old. For institutions wishing to use a credit reference score, such as the SCHUFA score, in their ongoing credit ratings, this time limit is the key regulatory hurdle.
Why is the 24-month limit particularly challenging for building societies?
Because their business model is geared towards concluding contracts and processing loans, and they have very little ongoing customer interaction compared to a high street bank. Regular bulk checks with the credit reference agency would be technically possible, but would not make economic sense for large existing portfolios.
How does Alte Leipziger Bauspar AG’s approach work in practice?
On a quarterly basis, only those cases are selected whose SCHUFA score would exceed the 24-month limit in the following quarter. The institution sends the identification numbers to SCHUFA, which enriches the data records with current score values and makes them available for download.
What proportion of the portfolio needs to be updated each quarter?
Mathematically, with a 24-month cycle and quarterly queries, around one-eighth of the portfolio reaches the threshold each quarter. In practice, the proportion is lower because cases that have since been paid off or have been re-queried for other reasons are excluded.
Can this approach be applied to other institutions?
Yes. Specialised institutions with little ongoing customer interaction, such as car finance companies, mortgage lenders or sales finance providers, face the same requirement for up-to-date data and can adopt the methodology of quarterly selection and targeted follow-up enquiries.
