6th Edition: 
Credit Risk Modeling and Management 

18-20 Nov 2019   |  Chicago

James Raborn

SVP, Chief Administrative Officer
BCB Community Bank

"Overcoming obstacles on the road to compliance."

David Gumpert-Hersh

VP, Business Analytics 
Wescom Credit Union

"Resolving data challenges: Internal and third party
data sources."

"Model validation in credit card portfolios under CECL"

SVP, Head of Credit Risk
& Capital Validation

U.S. Bank

Wei Zheng

 For registration details and multiple attendee discounts, please contact:

Alexia Mavronicola 
AlexiaMav@global-gfmi.com 

Peter Glassman

SVP, Head of Credit Risk Analytics
& Reporting
Northern Trust 

"Combine resources for CECL and CCAR compliance."

We brought together key industry leaders  from financial institutions that have devoted their time and energy into Credit Risk Modelling and Management.

Interested? Do you feel you will benefit?

THE SPEAKERS

© Copyright 2019 GFMI

Prime city centre locations and venues ensure your event experience is as convenient as possible

LOCATION

CHICAGO

ABOUT THE INTERVIEW

Could you give us a brief overview of your topic/presentation at the upcoming conference and what you hope delegates will take away from it?

January 2020 is approaching fast. Banking and finance industry is concerned about stepping into the lifetime allowance calculations, and uncertainty about credit card is close to the top of the list. For term loans, with their given payment schedule, the concept of lifetime loss is reasonably well developed. Credit card is on the opposite side of the spectrum: it has no defined term, and exposure at default can considerably exceed the current balance. Variability and complexity of some card programs and limited historical data may aggravate the problem.

While there is no “one size fits all” recipe, my expectation is that the industry participants, who have found solutions to a few key tasks, can deliver simple and robust expected credit loss models for their card portfolios. First, address regulatory requirements about unconditionally cancellable debt and lifetime estimation method, and build a calculation framework around your choices. Second, make sure that you have dependable data and can compensate for data shortages. Finally, understand how your lifetime loss estimates behave against the main drivers and how the estimated losses correspond with realized losses in the historical perspective.

In your opinion, why is there no “easy” way to approach CECL in credit cards?

Transition from the incurred loss allowance to the lifetime loss without a defined contractual maturity is not easy. CECL regulation is principle-based and gives the lenders considerable flexibility in the modelling approach – but there are key requirements in place.

The credit card debt obligations are unconditionally cancellable by most lenders, and the regulation stipulates that the cancellable off-balance sheet commitments do not contribute to the allowance. This requirement differs from the stress testing and capital planning approach which lies in the foundation of loss forecasting models in multiple institutions. Although some lenders may not have segmented their card loss forecasts by payment behavior for capital planning, they would have to do so for CECL: most, if not all, expected loss comes from revolvers.

The regulation also allows for two lifetime estimation methods: First in, First out (FIFO) and Proportional. FIFO, the most common method, unwinds the current balance by full strength of estimated future payments. The proportional method is more conservative and usually chosen by larger banks.

The institution’s choices in response to regulatory requirements lead to the lifetime loss estimate. This estimate may end up being quite volatile – which runs contrary to the assumed precision of the accounting world. To offer guidance to the decision makers, the institution should build an imprecision framework and produce qualitative adjustments to complement the expected credit loss. The imprecision framework reviews known model errors, uncertainty of economic forecasts, early warnings, and business strategy to come up with qualitative adjustments that reflect the expected economic situation in the short and longer term.


Can you briefly talk us through the process of finding reliable data sources for card loss modelling?

The data reliability questions are common: whether the right information is available, whether it has a reasonably long history, and what can be done to redress for the gaps.

The risk management time series data are often the first choice for credit risk modelers. Such time series may be readily available with origination, ongoing balance, and charge off data.Yet, the CECL analysis does not stop there. The modelers need to separate revolver accounts from transactors and to segment the portfolio by payment speeds for an adequate lifetime estimate. This involves other data dimensions: purchases, payments, financials statements, and possibly individual transactions with the associated interest rates. If these dimensions are outside the first choice data source, going to the production system time series may be the right step.

To understand historical payment and loss patterns with internal data, the data source should cover a full business cycle, i.e. go back 12 years or more. Not all regional institutions have credit card data of this length; smaller lenders are likely to be at a higher disadvantage. If that is the case, external data may supplement the internal data. The supplementary data exist at national or regional scale, with granularity ranging from aggregate down to instrument level. The aggregate data could improve macroeconomic dependence in loss models while the granular data could enhance the behavior and payment detail.

How do you go about estimating the lifetime loss, under unconditionally cancellable credit requirement, within your session?

Each institution aligns the depth of its loss model design with strategic significance and size of the lending product portfolio. Ultimately, the model estimates exposure at default, probability of default, and loss given default at the chosen granularity. In the session, I will show sample results of how EAD and PD are affected by creditworthiness, along with cumulative loss at the portfolio level. Another point is the “apples-to-apples” backtest which allows to measure the realized loss consistently with the requirement of unconditionally cancellable credit.

What would you like to achieve by attending the 6th Edition Credit Risk Modeling?

For the last few years, transition to CECL is on the minds of multiple groups of stakeholders across the banking and finance industry. The modelling and implementation questions will stay with us long past 2020. This conference allows to see the big picture of CECL preparation through the key dimensions of regulatory interest, data preparation, credit risk modeling, governance and compliance. I am looking forward to learn how financial institutions of different size and portfolio structure creatively handle the transition issues, especially in data analysis and quantitative modeling. Plenty of ideas shared will motivate to reflect on what we can still do better in the remaining half a year. And, last but not least, there must be a real treat in store in the masterclass about machine learning in credit risk modelling. See you there!

Ahead of the 6th Edition: Credit Risk Modeling and Management, we spoke with Andy Blokhin, Senior Vice President and Consumer Quantitative Analytics Lead at Regions Bank about approaching credit card modeling under CECL.

ANDY BLOKHIN PRESENTATION TOPIC

Andy Blokhin,  will be presenting during the second day, 13th of June at 9:15 am. 
  
Presentation topic: Approaching credit card modeling under CECL


• Simple and robust approach to CECL in credit cards:
Why there is no “easy” button? 

• Finding reliable data sources for behaviour segments and migrations

• Estimating and backtesting lifetime loss under unconditionally
cancellable credit requirement

REGISTER NOW!