How IFRS 9 and CECL work better when using a panel data model
This week I was going to talk about how IFRS 9 and CECL work better when using a panel data model, but just days ago, Alan Forrest beat me to it with a discussion of how Simpson's Paradox arises when separate models are used for Stage 1 (12-month) and Stage 2 (lifetime) forecasts. https://lnkd.in/gAczpbc2
The perspective in my book was about simplicity, integrating with cash flow models, and long-term accuracy. It had not occurred to me that those using separate models are likely creating biased results. My own studies when CECL was being rolled out simply showed that vintage-based methods (Age-Period-Cohort) or account-level survival models are more accurate. See my lengthy comparison study in Living with CECL: Mortgage Modeling Alternatives.
Honestly, few people care about lifetime accuracy of a calculation that is purely for regulatory compliance, but there is a broader point here. If you adopt an IFRS 9 or CECL method that has other high-value purposes, such as being the loss and prepayment components of a cash flow model, then it is no longer a regulatory expense -- it's an investment into running the business better. If you're going to go through all the effort and expense of gathering the data, building the models, validating the models, implementing the models and defending them to examiners, why not build models you can actually use?
Joseph Breeden
Posted on LinkedIn
Recent View Blog
