03 Jun The Shortfalls of Loan-Level CRE Stress Testing in a COVID-19 World
The Shortfalls of Loan-Level CRE Stress Testing in a COVID-19 World
By Adam Mustafa Invictus Group Chief Executive Officer
The coronavirus has presented the first threat to community banks since the 2008 financial crisis. For the first time, stress testing is a real exercise. What community banks across the country are discovering with dread right now is that their regulator-approved models are useless for the pandemic environment. Community banks need to quickly recognize that stress testing is no longer about satisfying regulators and immediately arm themselves with the right tools to prepare for a severe downturn.
The “Incumbent” Stress Test – An Overview
The most common form of stress testing in the community banking industry is a loan-level approach that is entirely focused on the commercial real estate (CRE) portfolio, mainly because regulators required this in 2006 guidance. This type of stress test, which we will refer to as “CRE stress testing,” is laser-focused on two metrics: the debt-service coverage ratio (DSCR) and the loan-to-value (LTV).
Here is how it generally works, although several variations exist:
1. The bank needs to gather the latest reported net operating income (NOI) for each borrower.
This task often proves to be difficult, time consuming, and prone to human error as it’s not naturally stored in the data core. Some ambitious banks try to go further and capture the components of NOI, such as rental rates, occupancy rates, and operating expenses, but this is often quickly abandoned following the trauma associated with going down this data gathering rabbit hole.
2. NOIs are then shocked with the same level of severity across the portfolio, say 20 percent or 40 percent.
The DSCR is re-calculated under these shocked conditions and benchmarked against underwriting standards. Loan balances that are no longer in compliance with these standards are summed and compared to the overall portfolio to provide management with a sense of the bank’s exposure under such a scenario.
3. Collateral values are also shocked, say with a 20 percent to 40 percent severity.
Some banks, albeit very few, may also try to incorporate a “cap rate” driven approach into this as well. The LTVs are also recalculated and compared to underwriting standards to give management a sense of exposure. The deficits of loans whose LTVs exceed 100 percent when shocked are summed, which is assumed to be a meaningful proxy of the bank’s risk.
The above exercise is simple, seems logical, and is so prevalent within the industry that many regulators often assume you have a shortfall in your risk management practices if you are not doing this type of analysis, especially if you are a bank with a CRE concentration as a percentage of capital approaching or exceeding 300 percent.
CRE stress testing can also serve as a prudent analysis to inform underwriting decisions for individual credits at the time of origination, review, and renewal. If a borrower does not perform well under such an analysis, the credit memo should articulate mitigating factors, such as the loan having recourse and being backed by the borrower’s liquidity and other assets. But a simple NOI shock of say 40 percent ignores some important context. If the borrower is not dependent on a small concentration of tenants, and has a high occupancy rate (think multifamily), the likelihood of a NOI shock along these lines is lower than say another borrower (think strip mall) with one anchor tenant and a handful of smaller tenants. The bank should also move quickly to work with the borrower if there is a material decline in NOI upon receipt of the borrower’s annual financial statements long after origination.
The Shortfalls in the Current Environment
The stark reality is that the incumbent CRE stress testing models do not work for the unprecedented COVID-19 economy. Several reasons:
1. The CRE Portfolio is not necessarily where most of therisk lies.
If a bank is only stressing the CRE portfolio, it has severe blind spots to the COVID-19 fallout. Yes, hospitality and retail reside in the CRE portfolio, but community banks likely have as much or more risk in their C&I and CRE-owner-occupied loan books with exposures to high risk industries such as manufacturing, oil & gas, and restaurants. Also, when we look across our client base, many of the loan modifications that have been granted during the COVID-19 crisis are associated with residential mortgages to borrowers employed in those industries.
2. It ignores other critical risk characteristics besides DSCR and LTV.
One of the fatal flaws of CRE stress testing method is that it treats the DSCR as the gospel. As previously mentioned, DSCRs are not all created equal. That is because NOI cannot just be measured in quantity, but it also needs to be measured in quality. Also, the CRE Stress Test completely ignores the risk rating assigned to the loan, which is far more important than the DSCR. The risk rating will be highly correlated to DSCR but will also capture other mitigating risk factors not easily seen in loan-level information, such as the ability of the obligor to subsidize shortfalls in the cash flows of the underlying property. (Note that in a pandemic stress test, the “global” strength of the underlying borrower is more important than the DSCR of the specific business or property purposed for the loan.) Other risk factors are missed, including the structure of the loan, maturity date, origination date, unfunded commitment, and cross-collateralization. Most importantly as it relates to COVID-19, CRE stress testing does not distinguish between loans that have been modified and ones that have not. Even though modified loans for COVID-19 do not have to be treated as troubled debt restructures, let alone be downgraded, they have a higher risk of default than loans not modified, accounting gimmicks aside.
3. It assumes a single outcome for each loan.
In other words, this type of analysis assumes that every loan has either a 0 percent or 100 percent chance of default, if subject to the same amount of stress. The probability of default for every loan is actually neither, but somewhere in between. While it’s impossible to estimate with perfect precision, the best models will take the weighted average of multiple possible outcomes for each loan. For example a 4-rated loan has a certain chance of absorbing a downturn and remaining a 4, or it could slip to 5, become a special-mention credit, become a classified, wind up in non-accrual, or flat out default. Before you shrug and think that complex models are required for such an analysis, remember that this is what happens in the real world. Some loans get downgraded. Some of those loans that were downgraded default. But some do not. This type of analysis can be performed without needing a PhD. It can be estimated with a little bit of data (far less than you expect), some effort and qualitative support. And it’s worth the effort because otherwise you will leave out loans that are more vulnerable to stress and end up with more false positives.
4. There is no connection to the loan loss reserve (either ALLL or CECL).
Provision expenses have an impact on capital, not charge-offs. This is especially critical in a CECL-world, where provisions are driven by expected, not probable losses. The methods typically used by banks to estimate their loan loss reserves are entirely different than the CRE stress test, even for the CRE loans. This makes no sense, and it is no wonder that banks struggle to use their CRE stress test for strategic and capital planning. In fact, the biggest mistake banks are making with CECL is their failure to integrate it with stress testing. In both the incurred loss method and CECL, most of the reserve today is driven by q-factors on pass-rated credits. Many banks struggled to estimate their reserves when the coronavirus emerged as a crisis in March. If stress testing and loan loss reserving are not speaking the same language, you have no chance at making sense of what the CRE stress test means to capital and strategy.
5. Contingency and strategic plans cannot be properly evaluated.
CRE stress testing only tells part of the story. Besides the need to stress test the rest of the portfolio, banks need to be able to put the stress test results into context by evaluating the impact on earnings and capital. It is vital in this crisis environment for banks to know their capital adequacy and the margin of safety for remaining sufficiently capitalized. Community banks also need to be able to evaluate strategies such as de-leveraging, dividend revisions, and expense reductions on stressed capital. Some community banks fool themselves into thinking their top-down ALM model is sufficient for this exercise. Unfortunately, those models are estimating loan losses using historical losses, often from other banks (usually weak ones). These models do not even consider the underwriting characteristics of the actual loan portfolio and are therefore useless and irrelevant for COVID-19. This is just another variant of a check-the-box exercise, which may lead to misleading and dangerous results.
Why This Matters
Yes, CRE stress testing has a role in the risk management playbook. Frankly, it should be required when underwriting, reviewing, and renewing any CRE loan. It’s simple and can easily be understood by management and directors. Regulators have influenced banks to adopt this type of stress testing for good reason.
However, it just doesn’t work for strategic and capital planning, and it certainly won’t help guide banks through the pandemic. Regulators had good intentions when they recommended CRE stress testing. They did not want community banks to become victims of the cottage industry of stress testing consultants and software pushers that make millions of dollars from the largest banks in the country. They also did not want banks relying on models they do not understand. Coming out of the 2008 Financial Crisis, CRE stress testing was a big step for community banks in the right direction. The regulators deserve a lot of credit as the industry is far better off because of it.
But the coronavirus crisis has unmasked the limitations of the incumbent CRE stress testing models for practical purposes. To get in front of the pandemic economy, banks need to take the next step in their stress testing journey by finding ways to quickly overcome these limitations. If you wait for your regulator to tell you what to do, it may be too late.