Validation of CECL models: What needs to be validated?

Validation of CECL models: What needs to be validated?

By Jennifer Matney, SVP, Director of Operational Risk Management, UMB Financial

CECL 2019 is taking place in New York City on 27-28 March, 2019 – find out more here

Can you please tell the Risk Insights readers a little bit about yourself, your experiences and what your current professional focus is?

I am a native of Kansas City and have been fortunate to work with many of our largest companies Including the Federal Reserve Bank, Cerner, DST Output, and now UMB. All my roles prior to UMB were in finance such as a Financial Analyst to Director of Finance all with primary responsibilities in budgeting and forecasting. In 2015 I shifted my focus and developed the Model Risk Management program at UMB. It started with a shell policy and inventory that has been refined over the past 4 years and now includes a team of validators to allow UMB the ability to conduct all validations internally.
In late 2017 I acquired responsibilities of third-party risk management and corporate insurance, so all combined we became Operational Risk Management. Then in late 2018 the contracts team was restructured and now is part of Operational Risk Management. Going into 2019 all things operational risk is my focus from development of new models such as CECL, ongoing model validations, third party concentration risk and how to mitigate or at a minimum inform and be aware, as well as fraud losses and other non-credit operational losses and what can UMB do to minimize these inherent risks.

What, for you, are the benefits of attending a conference like CECL and can you provide an overview of what you hope to discuss at the event and why the topic is so important?

CECL conferences such as this, have numerous benefits for me and so many others for different reasons. For me, from a model risk perspective, it will be incredibly helpful and valuable to learn from peers on their validation assumptions and approaches, timing of parallel runs, impact of CECL with other models, different perspectives on sensitivity of macro-economic variables, and the ability to network and learn from peers for future outside conversations. Validating CECL models is new territory for all of us whether it is an internal validation team or an external consultant. Having a strong model risk foundation that includes sound procedures, techniques, and standards to validations is key. I feel strongly UMB has that. Where we will all learn from each other is the differences in our approach to the complexities of CECL – it is unique in that it is an accounting standard that will now require an economic forecast. There are accounting rules that must be governed and applied, and are very black and white, versus a forecast that will be applied to account for the grey that exists. CECL is a high risk and highly complex model given the different model methodologies that can be used for different portfolios, possible financial impact for the accounting change, ongoing impact to the financial statements, and the regulatory impact. This is new for all of us!

Why is it so important to monitor performance on incurred but unrealised losses? How can this benefit an institution?

It is very important for a financial institution to monitor its potential losses to prepare for the future. The Great Recession taught us that the previous method for accounting for losses was “too little, too late”. To ensure proper liquidity and capital ratios we must look to the future to determine the quality of our loan portfolios. By monitoring the performance of incurred but unrealized losses we can account for those losses on financial statements before they occur. This allows us to be prepared when the losses do hit. As risk managers it is our job to inform and ensure there are no surprises (to the best of our abilities). The forecasting element to CECL is what makes this accounting standard so different and will hopefully allow financial institutions to be more prepared in the event of an economic downturn.

Can you provide some examples of when CECL forecasts have had differences to reality and how these can be overcome?

We are still in the development phase of building our CECL models, so I do not have any actual examples of CECL forecasts differing from reality. However, some similar methodologies and practices are in play with CECL that were for DFAST. No forecast is ever exact, there will always be some variances to reality. To overcome significant differences from forecast models to reality, back-testing and sensitivity testing are key. Back-testing allows the user to test the effectiveness and accuracy of the model to reality. If there are significant variances the model assumptions or variables used may need to be altered. By testing the sensitivity of the variables used in the model, the user can determine if those variables are important to the model and how sensitive they are to economic factors.

How can operational capabilities be optimized whilst aligning CECL, Stress Testing and Capital Planning? Are there any challenges that will need to be overcome to do so?

There are many benefits to aligning CECL, Stress Testing, and Capital Planning, however, there are some differences to all three that make it challenging to do so. For example, CECL has life of loan requirements that requires significantly more data and more detailed data that stress testing and capital planning models likely do not have. Over the next 12-18 months as CECL models are more final and ultimately implemented, it is imperative for the various teams that monitor CECL versus capital planning, versus stress testing communicate to ensure assumptions that are used in all three models align. If CECL is predicting increased unemployment over the next 12 months but stress testing models maintain a flat unemployment it will add confusion and undermine the validity of the models. To me, that is one of the most difficult challenge because we have different model owners for each of those models so ensuring assumptions and economic forecasts align will be very important.

Finally, in your opinion, how do you foresee CECL impacting the industry once all institutions have moved to implementation?

Similar to DFAST, there is an immediate impact to financial institutions right now as we are all developing models, testing, validating, finally implementing. All of this requires resources that financial institutions may not have had before, it requires different skill sets, different way of thinking. In addition, there is still an unknown of how CECL will impact financial statements. A recent CECL Preparedness letter to bank CEO’s from the OCC dated December 14, 2018 showed 66% of banks expect the allowance to increase over the current method.  That letter also stated most banks expect operational changes as a result of CECL from reducing the average maturity of loans, to pricing adjustments, to changing repayment schedules, and even changing credit management practices. These changes are still unknown until we have models developed and tested and compared to reality. This will take some time, but once CECL is more mature, maybe by 2021, financial institutions will have a better idea of just how CECL has impacted us.

You may also be interested in…

Sign up for your free Risk Insights account: