CECL and IFRS 9: Superior Risk Management or Psychological Ploy?

By now, we all know the drill with respect to forecasting expected credit losses (ECL). When a recession is predicted, for example, banks must increase their loan-loss reserves in compliance with CECL and IFRS 9 financial accounting requirements for ECL. But is this truly logical and necessary?

As we saw during the pandemic, a huge increase in credit losses does not always follow an economic downturn. Indeed, for residential mortgages, this has not been the case in three of the last four U.S. recessions.

Model failures have taught us, moreover, that the subjectivity and inexactness of scenario analysis makes CECL and IFRS 9, in part, psychological exercises.

Connecting the Dots: Recessions and Credit Losses

To better understand the link between economic crises and credit losses, let’s consider an ECL scenario. Suppose we want to determine expected losses for a residential mortgage portfolio, with the aim of calculating loan-loss reserves under CECL and/or IFRS 9.

Assume that the central bank raises rates by about 300 basis points as the core rate of CPI inflation surges to 5.7% from a baseline within the target range. The central bank’s actions cause a sharp recession – unemployment rises to 7.8% while median house prices fall by 5% peak-to-trough.

For risk analysts in 2023, this exercise feels like the uncanny valley: familiar, but slightly off. We have experienced inflation and rate rises, but house prices, in many countries around the world, have surged and unemployment has been surprisingly quiescent. The scenario is like our current situation – plus the recession that everyone was predicting last year.

Tony Hughes

Sharp readers may have spotted the game I’m playing – the scenario is precisely what happened during the 1990/91 recession in the U.S. In this case, we know exactly what happened to mortgage delinquencies: they barely budged, peaking at an industry average of 3.5% in 1992, roughly in line with the long-run average.

This example, combined with our recent experience during COVID, should be enough to convince you that recessions don't necessarily cause elevated credit losses. Sometimes they do, but often they don’t.

Mortgage losses certainly surged during the 2008/09 subprime crisis. Why was that recession so unusual?

The big difference between the 2008/09 and 1990/91 crises was the nature and quality of the borrower cohort on the eve of each recession. Prior to 1990, banks maintained a traditional approach to mortgage underwriting that applied the 5 Cs of credit. In 2006, by way of contrast, you could get a no-doc liar loan while working as a volunteer video game tester.

Creditworthy borrowers can usually withstand generic recessions – unless they are bowled over by a wave of losses stemming from loans to less worthy obligors.

IFRS 9 and CECL 

Mortgage losses remained low in three of the past four U.S. recessions them. We have a pretty good understanding of the preconditions that generate high mortgage losses, and they don't seem to be present currently. Keeping this in mind, would it be reasonable for a mortgage analyst to predict something close to baseline losses under a severe recession scenario?

History suggests that it would.

However, given CECL and IFRS 9 current ECL requirements, it’s pretty obvious that regulators and auditors would never accept such a finding. If a recession is ever viewed as likely or probable by the broad consensus, lenders will be obliged to build very conservative loss reserves, regardless of the underlying quality of the loans on book.

We’ve talked in the past about excessive loss reserve volatility and the difficulty in successfully forecasting economic turning points. Here’s the way things often work: a recession is predicted, reserves build, the recession fails to materialize, and reserves are drawn down. Reported bank earnings at first fall and then rise, but nothing actually happens. 

Under CECL and IFRS 9, the analysis presented here suggests that reserves will be unnecessarily volatile, even if all the recessions are correctly predicted. In most of the downturns that occur, reserves on mortgage loans will increase but elevated credit losses will not follow.

A Psychological Crutch?

I suspect that many bankers are, from a personal perspective, quite comfortable with the new loan-loss accounting methods. Increasing loss reserves because the consensus is forecasting a recession is a very easy story to tell both shareholders and boards of directors. 

However, there are certainly circumstances under which such a discussion might ultimately prove quite awkward. Just think back, for example, to the fall of Washington Mutual (WaMu).

For many banks, in the years prior to 2008, the narrative they were using to justify strong growth in subprime mortgages involved the democratization of home ownership. Kerry Killinger, the CEO of Washington Mutual, said all the way back in 2003 that he wanted his bank to be the “category killer,” the “Walmart of consumer finance.” He likened WaMu to companies like Starbucks and Home Depot, aimed at lower- and middle-class people underserviced by regular banks.

Killinger, of course, didn’t see a collapse coming, and WaMu remains the biggest bank failure in American history. But would it have made a difference if WaMu had been subject to current financial accounting requirements?

Imagine a world where WaMu was subject to CECL and IFRS 9 expected loss accounting in its final few years. At the time, the models in use at the bank had been calibrated on earlier recessions that were relatively safe for mortgages. As economists started to predict the onset of recession in 2007, WaMu’s models would not have moved very much, suggesting that CECL loan-loss reserves would have been relatively stable during the first half of 2007.

As defaults on subprime mortgages started to build, WaMu’s risk analysts would have begun to see a divergence between actual results and the predictions provided by the CECL and IFRS 9 models. It is only at this point that loan-loss reserves could have started to truly expand.

For reserves to rise any earlier than this, key insiders would have had to realize the flaws in the company’s business model and senior executives, including Killinger, would have had to publicly admit them. In other words, they would have had to inform shareholders and investors that WaMu needed to double loan-loss reserves, after describing the bank as the “Walmart of consumer finance” just a quarter earlier. Bear in mind, also, that this would have had to happen before the company had any evidence of actual defaults. 

Parting Thoughts

CECL and IFRS 9 are, at least in part, psychological or strategic exercises, because of the subjectivity and uncertainty of scenario analysis. This rings particularly true amid recession prognostications.

If the industry consensus is to downgrade pre-recession macroeconomic forecast, there is no shame in, say, an individual bank declaring weaker earnings, because it is likely that peer institutions will soon be following suit. 

For many portfolios, these loan-loss reserve building exercises will continue to be overly conservative. There seems little doubt, moreover, that earnings numbers will become far more volatile under the new protocols.

Where things are likely to get dicey is when banks get caught with their hands in the cookie jar. If a risk management department discovers an idiosyncratic implosion in credit quality, scenario loss projections should, in principle, rise disproportionately more than under generic recession conditions. Indeed, credit quality is arguably a more important factor in determining credit risk than underlying macroeconomic conditions.

For complete ECL transparency to happen in public disclosures, it will take an exceptional team of ethical senior managers. Looking at it the other way, the prospect of publishing the bank’s mea culpa may act as a deterrent to crazier profit-making schemes.

Tony Hughes is an expert risk modeler. He has more than 20 years of experience as a senior risk professional in North America, Europe and Australia, specializing in model risk management, model build/validation and quantitative climate risk solutions.