10th Annual
Credit Risk Management, Modelling, and Validation
Develop, test and utilise credit risk frameworks and models incorporating Basel IV, IFRS9, IRB, climate risk and stress testing standards
ENTER YOUR DETAILS FOR THE CONFERENCE AGENDA
How are you dealing with credit data challenges to ensure relevance and quality?
We are big fans of data validation pipelines and schema libraries for consistent data types and automated data quality checks. A simple example, if there is a parameter called PD in any of our codes or tables it will be linked to the centralised schema library which defines it as a floating point number between 0 and 1. Any input data that needs PD is checked dynamically against this schema and exceptions are handled accordingly. Furthermore, the default definition can be modified and extended in a different context, for example if we talk about PD in a downturn scenario which would have different properties and requirements.
What are the areas you are particularly focusing on at the moment and in the next few months?
Automatic data validation is something that we already have and will keep developing further into the future. Currently, a major focus is automated backtesting and dynamic feedback loop for updating the models and their parameters. As I mentioned, here a big challenge is the combination of data-driven and expert-based information, however, a key to solving this, is treating the expert estimates and the ones from models the same. For example, if a model estimates a 5% probability of default but the expert says 10%, from a backtesting point of view it doesn't matter how exactly the estimate was obtained, all that matters is how good it is relative to the actual observed data and how well do you update it as a continuous process.
What are the main challenges credit risk model developers and validators are facing at the moment?
I see three main challenges which are also very related. The first one, is the design of integrated modelling methodologies, this is needed in order to have an efficient process that connects the dots between different regulations and the credit risk theory, in order to produce models that are both regulatory compliant and statistically optimised. The second one, is the dynamic combination of statistical evidence and expert-based information. The rare events accounted for in risk are often lacking large amounts of data unlike other fields like image recognition, this requires integration of human intelligence in the form of expert-based information, however, the latter also needs to be updated dynamically as data availability grows. Finally, constructive validation is something very powerful but hard to achieve. Model validators often act as administrators which simply need to fill a report template. This can make the job boring and also brings little value added to the modellers and ultimately the quality of the models. While the administration can be automated, there are a lot of challenges in model risk that can be approached objectively and scientifically, like what is the quantitative impact of a given modelling decision on the final estimates and the business decisions they drive? A good validator would aim to quantify this objectively as the nature of model risk.
How can the effective validation of IFRS9 models be ensured given the specific challenges involved?
IFRS9 has a lot of generic concepts which need to be translated to statistical properties that can be tested objectively against real world data and all of this needs to be set-up as a continuous process. This sounds simple, but if you imagine a daily provisioning process over the lifetime of a 1-year loan, this means that each of those 365 days the provision will incorporate forecasts for each of the remaining days until the end of the lifetime. This gives a total of ~66,430 predictions and each of these represents a distinct promise which can be tested objectively after the actual values are observed. Validating the accuracy of these predictions as a continuous process is a huge challenge.
Ahead of the 10th Annual Credit Risk Management, Modelling, and Validation conference we spoke with Dr Konstantin Vasilev, Group Head of Advanced Risk Analytics at Revolut. Dr Konstantin Vasilev is originally from Bulgaria. He moved to the UK to study Economics at the University of Essex and subsequently completed his Msc and PhD there. After a PhD traineeship at the ECB in Frankfurt, he worked at several of the largest European banks in the fields of stress testing, model validation and model development for credit risk. His journey at Revolut started in 2020 as the first member of the Model Risk team, where he helped to establish and grow the validation function. He then founded and headed the Credit Risk Modelling team with main objectives in the areas of ECL modelling and stress testing. Following the success of the team, as well as the needs for analytical capabilities for all financial risk types, the team was expanded to Advanced Risk Analytics.
18-19 September, 2023
Amsterdam, Netherlands
Dr. Konstantin Vasilev will be presenting a session during day two of the 10th Annual Credit Risk Management, Modelling, and Validation conference!
Session: Confront IFRS9 model validation challenges to ensure regulatory compliance
- Identify the best procedure to develop IFRS9 models meeting regulatory compliance to effectively forecast ECL.
- Understand the importance of back testing and performing ongoing monitoring of PD and LGD models through the entire model lifecycle to identify their performance in different scenarios.
- Assess the relevance of data and its function through the model lifecycle to predict forward looking results.
For registration pricing and multiple attendee discounts, please contact:
Ria Kiayia
All Rights Reserved. marcus evans ® 2023
Interested? Do you feel you will benefit?
An interview with Dr Konstantin Vasilev, Group Head of Advanced Risk Analytics at Revolut
Why would it be beneficial for credit risk experts to attend the next edition of the credit risk modelling and validation event?
This event is a great opportunity to meet industry leaders and expand your network. I encourage active participation, as it broadens the horizons and allows you to sync your clock with the best in the credit risk modelling and validation areas.
