Soner Tunay will be joining us for the CECL 2017 Congress. Soner is the leading Quantitative Analytics in Accenture’s Finance& Risk practice. Until recently, he was the Head of Risk Analytics at Citizensbank where we developed CCAR models, Economic Capital and Basel-type risk rating methodologies. Previously he held similar leadership roles in model development and validation functions at a global systemically important bank a super-regional bank.
Soner, can you tell the Risk Insights’ readers about yourself and your professional experiences?
As your readers would know, I have been active in modelling circles. I have been holding full-day workshops and reaching out to risk professionals through various speaking engagements and webinars. Recently, we held a workshop on “Integrated Credit Modelling: From CCAR to CECL” in New York City which took place ahead of the Risk Americas Convention. And I plan to do a similar event the next year as well.
[The Risk Americas Convention is the leading premier risk and regulation gathering where over 500 risk professionals share insight and knowledge on current challenges, and will be returning in New York on May 2018.]
Can you provide our audience an insight into the life of loan concept under CECL and how this differs from current practices?
There are number of new concepts being introduced in the CECL framework for loan loss reserving. Arguably, the most impactful of all is the ‘life-of-loan’ concept. It is going to be a fundamental change in many activities in Banking because almost none of the current models or processes are designed to deal with the life of a loan estimates.
The current reserving methodology is based on Incurred Loss Period (ILP) concept. The premise of ILP is that an event that has happened in the past that will lead to losses in the future, which is inherently backward looking. The new ECL (Expected Credit Loss) framework is designed to be forward-looking throughout the life of a loan.
Because of the P&L impact of the changing reserves due to the life of loan concept, we expect an alignment in the rest of the systems and models in a bank. As you could imagine, many such systems use reserve estimates as an input. We predict that pricing models, loan origination systems, profitability analysis and even Economic Capital calculations need to be all in alignment with the life of loan estimates.
Many will be looking to leverage CCAR under CECL, to what extent do you believe this is feasible given the horizon differences?
We believe that CCAR models are a good starting point for CECL purposes. However, some adjustments are needed. The most critical adjustments will be better calibration of baseline estimates, aligning with business forecasts, being able to deal with a longer time horizon because of the life of loan forecast and better integration of drivers of expected life estimates (e.g. prepayment models) and credit loss forecasts.
Most CCAR models were calibrated to do well in the stress scenarios and many had conservative assumptions built into them. It would take a careful study to re-calibrate CCAR models to be more accurate in benign economic environments. Similarly, we need to ensure that the baseline estimates line up with the business forecasts.
CCAR models are designed and tested at relatively short horizons, at 9 quarters. CECL will change this horizon. Most of the loans will have expected lives longer than 9 quarters. During the life of a loan, CECL models should be driven by a macroeconomic scenario within the reasonable and supportable period as in the CCAR practice but should be able revert to longer term experience after that. This aspect of CECL alone is a major challenge that practitioners have been struggling with.
And finally, the integration of credit loss models and other models (i.e. prepayment) that help estimate the remaining life of a loan is more important now.
How can institutions ensure the right mix of short and long term horizons into their model?
As mentioned above this is a new type of challenge for practitioners. In the past, we either built Point in Time (PIT) or Through the Cycle (TTC) models. With CECL the lifetime forecast would include components of both PIT and TTC elements. And we know from our experience of developing both PIT and TTC models, that each one takes a different type of driver. For example, macro factors might be the most influential drivers in the CCAR models which are closer to PIT. On the contrary, macro factors have very limited room in a typical TTC model, as such models are driven by obligor-level attributes.
Banking professionals are developing some ideas to adopt to this new kind of framework. One group plans to switch from PIT models to TTC models at the end of the reasonable and supportable period. We believe such an approach would have certain challenges. CCAR style models are driven by the macro scenarios and therefore more PIT. TTC models, being calibrated to historical default experience, typically have been backward-looking, which is not an ideal application for CECL. While the debate continues on this topic, we have proposed a new approach which we call Forward-Looking TTC. Forward-Looking TTC should combine the scenario driven nature of PIT, but also reflect the current and future portfolio composition in its default rate calibration, thereby combining the best of both worlds.
What challenges do you foresee on the horizon for CECL that institutions should be aware of as the industry moves forward?
We mentioned some of the methodology challenges above. In addition to analytical changes, we anticipate a number of enhancements in data management and technology infrastructure. Particularly, we believe that data lineage verification, governance around model production, infrastructure and reporting are going to be more intense than what we are used to with CCAR.
We also predict a considerable amount of effort required to align origination systems, loss estimates, product strategies and product pricing to the new reserve calculations. End to end system integration would also be in place.