Tailoring analytics to your needs

Internal models validation services

Modern banking has become a technical field, and getting more so. Big data, automated decision-making and sophisticated statistical models are common themes in the industry. Credit risk management is at the forefront of this trend, bringing banks’ long tradition of prudency and conservatism to the era of data analytics and modern technology.

A review of internal models cannot be thought of as a panacea of an internal models process. Especially since it is such a complicated process of interactions between technology, data and experts. Instead the review of internal models should act as a catalyst to bring change to internal models for the financial health of the enterprise.

It is enough to mention IRB, IFRS9, and Stress Testing to paint a picture of how diverse and complex a credit risk modelling landscape is in today’s bank. In order to fulfill their purpose, these models must be correct, accurate and work in unison. What happens if they are not? Then they produce excessive estimation errors. Because major business decisions rely on the numbers produced by those models, a bank is likely to misjudge its risks, misprice its products, misestimate the expected loss and required capital. If so, such mistakes lead to a loss of the bank’s competitive power, take a toll on its profits and increase a chance of bankruptcy. While it is commonly perceived that regulatory scrutiny is guaranteed to keep bankruptcy as distant as it is supposed to be, bank’s profits and market position are its own concern. Validation is there, in particular, to make sure that model-induced errors are minimized, so that the bank’s profitability and business performance do not suffer. It is still a large-scope mandatory exercise, but it is important to remember that it is in the bank’s own interest to have solid and reliable internal models.

Why is validation difficult?

Building a model is not so difficult. Building a good one is hard. Among many challenges is to give assurance that the final model is indeed good, i.e. correct and accurate, and that the bank can trust the resulting numbers. This is done by validation.

Because model errors can come from multiple sources, validation requires a comprehensive approach combining subject knowledge with technical skills, semantic precision with methodological rigor, relevant regulation with common sense, as well as a clear understanding of the nature of model risk and the roles of the parties involved in its management. Regulatory rules endorse this approach, but unfortunately do not provide all the answers to the many questions arising in the process, often limiting themselves to articulating only the end-goal rather than a particular way of reaching it. Despite significant improvements made by the regulators in the past years, important definitions are sometimes still missing or imprecise, allowing for multiple subjective interpretations of concrete requirements. Uniform industry standards are still in the making too. In addition, numerous academic articles and working papers borrow statistical tests from other scientific fields, though their assumptions are unrealistic in credit risk applications. The use of such tests leads to indeterminate validation conclusions, causing a situation in which model deficiencies either remain unidentified or the collected evidence is too weak for a decisive action.

How can Credit Analytics help?

Being a prerequisite to the use of internal models, strong validation procedures are the key to success in today’s banking business. At Credit Analytics, we help our clients by offering truly independent and professional quality assurance of their credit risk models. For us, this work does not equate to ensuring regulatory compliance. It is more. We use our experience as model developers, validators and former regulators to help those enterprises that believe in the importance of data-driven decisions to generate more value for their customers and shareholders by making their internal models more accurate and reliable. This is why we view those models as a strategic asset rather than an imposed cost, and uncovered issues not as problems but rather as opportunities.

As external consultants, we bring an impartial expert judgement rooted in the broader industry best-practices. We follow the latest ideas and results published in professional journals and rely on innovative solutions, some of which are based on our own research. Our effort is aimed, in particular, at assisting the host validation team in selecting most appropriate statistical tests and confidently interpreting multiple interconnected findings.

With these considerations in mind, we have built our validation platform. It can serve as a convenient decision-making tool in the matters of model recalibration, redevelopment or retirement by giving on-demand interactive overview of model risk and enabling the bank to do fast, regular and automated model check-ups. It relieves the bank of the manual ad-hoc burden, characteristic of past validation practices, and leaves more time for understanding and explaining the results to stakeholders.