Did CECL and IFRS 9 Fix the Procyclicality Problem?

Now that we've experienced a pandemic-driven recession, it seems as good a time as any to assess whether the CECL and IFRS 9 reporting standards for expected credit losses (ECL) have had their desired impact. After all, these standards were expected to yield a more proactive loss-accounting system that would enable financial institutions to improve forecasting and build up loss reserves that could counteract the impact of any recession.

One of the problems with the incurred loss approach for credit losses, which was in place prior to CECL and IFRS 9, was that it was too procyclical. This resulted in firms not being able to develop proper loss reserves prior to the beginning of a recession, which, in turn, meant that they could not free up funds to support an earlier resumption of lending growth.

 Tony Hughes

Procyclicality and volatility were two of the key concerns of loss accounting critics prior to the launch of CECL and IFRS 9. But in the aftermath of post-implementation recessions (there have actually been two in the UK,) it is now possible to address whether ECL is truly fit for purpose.

Turning quickly to the issue of volatility, CECL and IFRS 9 replaced a reality-rooted incurred loss methodology – which reflected losses based on negative credit events that have already occurred – with forecast-driven approaches that are subject to a range of statistical errors associated with the specification and estimation of various models. Moreover, these new ECL methodologies also incorporate a healthy dose of management overlay, which is also subject to considerable bias and error.

Given all this, there was never much doubt that the new methods would be more volatile than those that they replaced. Subsequent studies have all but confirmed this reality.

Increased volatility may be tolerable for investors if the central aim of expected loss provisioning was actually achieved. Specifically, the overhaul was implemented to either reduce, eliminate or even reverse the procyclicality of bank loss reserves.

So, have CECL and IFRS 9 met this objective? Before answering this question, we first need to consider why the incurred loss methodology was replaced.

GFC Unveils a Loss Reserves Dilemma

In the latter stages of the housing boom that preceded the global financial crisis (GFC), banks held a large volume of bad loans on their books. But because the incurred loss method required lenders to wait for an adverse credit event before recognizing any expense, total loss reserves remained low.

Dodgy mortgages were made to look better than they were because house prices were still rising and the economy was still growing. When the recession finally came, a flood of delinquencies followed, and loss reserves quickly ballooned.

This has all been revealed with the benefit of hindsight – but before the Great Recession, there was far more uncertainty about the precise nature of the situation that confronted mortgage markets in real time.

Had the pre-GFC provisioning process been countercyclical, we would have seen reserves gradually building as the boom progressed, reaching their zenith around the first half of 2008. They then would have fallen rapidly, freeing up funds to support an earlier resumption of lending growth in the wake of the recession.

One of the ironies of this discussion is that countercyclicality is easy to achieve (with a small lag) as a mechanical process. If you tie provisions directly to observed industry loan growth or some coincident indicator of economic activity, you'll instantly and precisely induce the dynamics you desire.

CECL and IFRS 9: Not Recession-Proof

Expected loss accounting takes a more circuitous route. It is based on the principle that the shadow of recession will be apparent during the latter stages of an expansion period, causing loss forecasts to rise, and allowing reserves to build while the economy is still growing. Then, while the recession is raging, loss forecasters will imagine better times ahead and provisions will fall, freeing up funding for a rapid resumption of lending growth.

So, how did this work in practice? The 2020 COVID-driven recession has been analyzed by several authors and a picture is starting to emerge. The results are somewhat mixed – Chen, Dou, Ryan and Zou (2022) found that U.S. banks using CECL increased their provisions during the pandemic at a faster rate than banks that were yet to make the switch. In contrast, using regression analysis, Hansen, Charifzadeh and Herberger (2023) looked at the effect of IFRS 9 on EU banks during the recession, and found results consistent with an improvement in the overall extent of procyclicality.

People will read these papers and interpret the results in different ways. As evidenced by a handful of articles I wrote before the pandemic, I previously believed that CECL was likely to be worse than IFRS 9 and that neither approach would truly fix the procyclicality problem. The two papers therefore mostly align with my projections.

Personally, I find Chen's results more compelling, and tend to believe that CECL exacerbated the procyclicality problem. I'm skeptical of Hansen's finding that IFRS 9 was less procyclical than incurred loss, but I'm not surprised that IFRS 9 fared much better than CECL.

Pandemic Peculiarities and Recession Realities

The thing about the pandemic was that the recession hit very quickly, with little warning. Lenders had no time to anticipate the downturn and build reserves, so the surge in provisions happened after the recession had already started.

I think the designers of CECL and IFRS 9 imagined that recessions follow some sort of regular, predictable path. This is the notion that you may not know precisely when the economic worm will turn in the late stages of a boom, but you can reliably sense when that is about to happen. The idea is that things can only get better after a recession begins, but this ignores that it is possible to experience double-dip recessions and full-blown economic disasters.

The reality is that, by definition, recessions are abnormal events that are extremely heterogeneous and inherently difficult to predict, even when you're in the middle of one. Unlike the GFC, most recessions are caused by factors that are external to the banking system, meaning that bankers have no special insight into their likelihood or their effect. Likewise, the impact of recessions on borrowers is hard to fathom, with many downturns having little or no effect on vast swathes of borrowers.

Parting Thoughts

Research may show that CECL and IFRS 9 could have worked, retrospectively, during the GFC. This does not mean that they will be effective in other recessions or indeed the next one. They did not work very well, for example, during the pandemic.

By putting bank forecasts at the very center of loan loss provisioning, accounting officials have ensured that financial statements will become more volatile. If they also become more procyclical, it will be reasonable to question whether the whole shift to the ECL methodologies was worthwhile.

Indeed, a poor performance in the next recession could be the end of the line for CECL and IFRS 9.

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. He writes regularly on climate-related risk management issues at UnpackingClimateRisk.com.