Credit Unions Must Prep Now for CECL
Experts provide three steps credit unions should take now and when you know it's too late to implement CECL.
The rules, which require credit unions and other financial institutions to record expected losses when they make new loans, don’t kick in for credit union Call Reports until the beginning of 2022. But CECL preparations can easily take years and procrastination can be a costly mistake, industry pros cautioned. Here are three things they said credit unions should start doing now if they want to be ready for CECL on time – and when it’s too late to start.
1. Figure out what data to collect and how to collect it. “That’s the first priority right now,” Thomas Griswold, director of advisory services at financial institution advisory firm ALM First in Dallas, Texas, said.
“The big pieces are going to be credit score, LTD ratio, debt-to-income ratios – those are going to be the big components,” he said. “The good thing is with three years effectively to adopt, I would think that gives plenty of time for an institution to look at their data warehouse and how data are flowing through the institution, and be able to pull reports on a frequent basis to inform that credit profile.”
Of all the things on the CECL preparation list, that frontline data capture is probably the hardest thing to do, Griswold added.
“I think that’s why it should be the first and foremost step, as it may be costly to reformat your data warehouse and take some man-hours to harvest that data,” he explained.
That’s a challenge that is likely setting some credit unions back, added Jared Mills, who is a consultant at the Raleigh, N.C-based financial information company Sageworks.
“The requirements for a lot of these models and types of analysis requires a lot of the data that credit unions haven’t been actively storing and saving,” he said. “To put it simply, we need lifetime’s worth of data.”
Credit unions should set aside several months to do the work, Mills noted.
“In a perfect world we’d want to allow, first of all, enough time to actually do a thorough investigation of our data: How much we have, how much we don’t have, potential issues we have with it. I would at least allot half a year to kind of doing that before even starting running some models,” Mills warned.
2. Settle on a CECL model. Processing all that data is a critical component of developing new models, and that isn’t a last-minute endeavor either, Mills warned.
“That can take some time. That isn’t something that usually you can just get figured out in one run. That’s likely something we’d want to do a few times ahead of the point of making, maybe, some decisions,” he said.
Choosing from a multitude of models often comes down to the complexity of the credit union and what it wants to get out of CECL, Griswold noted. More expensive, complex models might be good for credit unions with unique loans or strategic planning needs, for example, but they could be less necessary for credit unions with straightforward operations or simple needs, he explained. Credit unions need to make time to decide what they want and then find something that works for them.
“Is it just a check-the-box item where I just want to know my allowance and move on? Or do I want to take this opportunity to capture all of this credit data and truly inform my entire institution of any nuanced risks that sit on the balance sheet?” Griswold asked. “If the former is the answer, and it’s just a simple check-the-box item, then the answer should be: Choose a simple model that’s inexpensive and gets you the number that you want. If it’s the latter and you really want to get a significant amount of the decision-making power from this process, then you need to look at a more complex model.”
Credit unions also need to make time to understand the assumptions that go into various models, Griswold added.
“Making sure you understand how data are used, the assumptions informing the loss projections – that’s going to be critical,” he said. “If an institution is looking at vendors, I’d say make sure that every model report you receive includes assumptions – and detailed assumptions – as well as data limitations. That should make it a completely transparent process, to where it eases any heartburn for the institution and makes it easily auditable, which I know auditors will definitely appreciate.”
3. Test and make adjustments. Credit unions can’t procrastinate on running data through their models – those models may spark some time-consuming discussions, Griswold said.
“Once you get the model results from a financial side, make sure that you agree with those expectations. Is it wildly different than what you’re actually considering your risk profile to be? If it is wildly different, that means maybe it’s some time to do some soul searching as an institution. Are we marking our loans appropriately? And I think that component right there, you know, ‘What do we do after we get our reports?’ – that’s why it’s important to get started on this process,” he said.
“You want to have that detail as soon as you can to make moves and strategic decisions,” he said.
The Point of No Return
“If we’re sitting at January 1, 2021, and you haven’t really taken that initiative to start digging through your data, that’s a bit too late,” Griswold warned.
For NAFCU Chief Economist and Vice President of Research Curt Long, most credit unions today still have some work to do.
“We try to ask our members about this issue – it is a big one for them. There are some who are well into the process of not only gathering data but looking at some of the different modeling options, or even further down the road than that, having selected their models. There are still a fair number who are in the very early stages of that process or even some who have not yet begun,” he explained.
He added, “I’m not sure any of it’s going to be painless. I’m noticing a lot of angst in the industry, and probably more so as we get closer to the implementation deadline. I know I’ve said it before but I’ll go back to it again: Something that everyone should be doing right now is gathering that data, because that is something that they’re probably going to need.”