The “reasonable and supportable” clause of the new Current Expected Credit Loss (CECL) standard “is the most hotly debated part” of the regulation, according to Cristian deRitis, Senior Director, Consumer Credit Analytics at Moody’s Analytics. He was part of a round-table discussion about the transition to CECL that was webcast by the Global Association of Risk Professionals (GARP) on January 10, 2018. CECL is the new credit impairment standard under Financial Accounting Standards Board (FASB).
“Auditors are grappling with what ‘reasonable and supportable’ means, too,” he added. For best practice in terms of modelling credit risk, the phrase boils down to whether other models come up with similar results to a bank’s chosen model.
DeRitis broke CECL into two components: the credit loss model used, and the economic forecast that feeds the model. He cautioned that if only one economic scenario is used “then you are tying yourself to that.” It’s preferable to run the analysis over multiple scenarios, to create a range.
Of the three panellists, deRitis had the most recent on-the-ground perspective of the challenges faced by European banks as they struggled to implement IFRS 9, part of the International Financial Reporting Standards. IFRS 9 is a close cousin of the CECL standard. “Many [European] banks underestimated the timeline,” he said, predicting that some will “still be working out the kinks in 2018.” The lesson for CECL adoption? “It’s never too early to start,” he said.
When it comes to modelling, CECL is not prescriptive. Banks should not feel overwhelmed by modelling credit risk. “A model is just a structured, systematic way of organizing the information,” deRitis said, and it’s part of having a more holistic view of banking, starting from origination to the final stages of a loan.
“To best fit different asset classes, we will have to use different methods, because we have different data,” he said. For example, residential mortgages come with a lot of information gleaned from the lengthy application form, whereas “for an unsecured personal loan, we might have only a credit score.”
During the Q&A session, deRitis spoke about vintage cohorts methodology. Some organizations cannot handle an overly complex model. The benefits of vintage cohorts over the loan level approach are three-fold: “it compacts the data” therefore there is less to have to handle; the compaction reduces noise in the data; and there may be better correlation.
When asked which model to choose, he said it’s something to be done on a class by class basis. He said a bank “might want to leverage an existing model” or “build out the system.”
In the short term, much work remains to be done on CECL. Ultimately, the trouble will be worth it because “CECL will move us to a better banking system overall,” he said. ª
Related Links on Moody’s CECL
Click here to visit the Moody’s Analytics CECL website.
Related Links for this GARP webcast
Click here to read a summary of comments by the panel member Masha Muzyka, from Moody’s Analytics.