The pressure is on to ensure you are ready for the fast approaching IFRS 9 deadline, but many lenders have had challenges in both developing and validating models, which may delay them from entering the parallel run phase.
Development and approval of the models is by no means the end of the journey. Model validation techniques must be performed to understand how the models behave under different circumstances and what the overriding effects are on the final provision numbers.
Validation techniques, such as back testing, will test the performance of the models on historic data and enable you to understand the volatility of the outputs against historic periods prior to entering parallel run phase. In addition, devising ‘what-if’ scenarios helps to further validate the models on more extreme scenarios that haven’t necessarily been experienced within historic periods.
What is model validation?
Model validation is the critical point during the IFRS 9 project that enables you to check the accuracy of model estimation, which is key to achieving compliance. A number of larger firms, with sophisticated model frameworks, are now in parallel run phase and will be learning about the adequacy and stability of their provision estimates as a result of executing their models across a number of reporting points.
Given the complexities of these model frameworks, it may be challenging and time consuming to dissect the models to attribute any irregularities back to the component of the framework to fully understand the source of a problem.
What is involved in the validation phase?
The key is to ensure models produce provision estimates that are adequate, accurate and stable and model validation is an exercise to assess, with a degree of confidence, that the models fulfil this criteria.
Despite the time pressures, lenders don’t need to compromise on quality or increase the burden on already stretched resources. These challenges mean that lenders will be looking for solutions that reduce the complexity and investment at the same time as increasing automation and leverage work done with previous models. The validation of IFRS 9 models is quite different to the validation of other models, such as IRB models. As IFRS 9 uses different, unfamiliar kinds of models and methodologies it’s essential to validate against robust best practice IFRS 9 methodologies.
Model validation often focuses on the validation of each model component in isolation, although arguably the most important factor is how all the models are combined to calculate the final ECL numbers. Having that focus on the outputs and the movements in relation to the underlying portfolio trends and the macro-economy can be challenging, particularly within a complex model framework.
Jaywing has a number of offerings to help our clients meet the challenges of IFRS 9 validation, which includes Horizon, a tool that employs automated modelling techniques with inputs of your own data and economic scenarios to calculate IFRS 9 compliant ECLs, as part of our consulting toolkit. It offers a quick and cost effective method that can validate your model framework and the model estimates produced. The MI produced as part of Horizon will assist in the isolation of any potential problems with your existing models through a comparison of trends and outputs between your own model framework and Horizon.
Horizon can be provided to facilitate a one-off validation exercise; all required data is assembled into the prescribed format to enable it to be processed through to the tool, which will then produce the ECLs in a matter of minutes. These estimates and the resulting MI can then be compared to your own IFRS 9 models as part of a validation exercise.