Credit: Mimi Phan / iStock
Ideas Made to Matter
Ratings mismatch at play in collateralized loan obligations
As the U.S. economy continues to struggle, researchers have uncovered a troubling discrepancy between credit agency ratings and the risk profile for collateralized loan obligations.
Even more unsettling is the reason why this may sound familiar. “In the financial crisis of 2007 – 2008, ratings agencies were slow to downgrade structured finance products, including collateralized loan obligations,” said a visiting assistant professor at MIT Sloan. “And it looks like we may be repeating history during the current COVID-19 crisis.”
Collateralized loan obligations package risky corporate loans that are below investment grade into safer securities, known as tranches, which are then purchased by investors.
According to a new study co-authored by Nickerson, between March and August 2020, ratings agencies Standard & Poor’s and Moody’s downgraded approximately 25% of the collateral feeding into CLOs, but the total value of the CLO tranche downgrades was only around 2%.
That discrepancy suggests a conflict of interest may be at play — as it was during the financial crisis when ratings agencies came under fire.
As part of Standard & Poor’s settlement with the Department of Justice in 2015, S&P admitted that it didn’t want to downgrade a company’s underperforming assets because it was worried that doing so would hurt its own business. The reason? Companies pay ratings agencies to rate their loans; if they don’t like the rating they receive from one agency, they can go to another.
“This fear of losing business certainly is a potential reason why tranches haven’t been downgraded,” said Nickerson, who co-authored the research with John M. Griffin of the University of Texas at Austin. “I think what we’re finding is consistent with that potential conflict of interest.”
COVID-19 — a black swan event
CLOs are composed of corporate loans that are below investment grade, but agencies will give them a AAA rating because they are highly diversified, Nickerson said. “We assume that if something defaults [in a CLO], it’s relatively uncorrelated with the performance of other companies in the CLO at the same time.”
Or at least that was the assumption until COVID-19. As the effects of pandemic-related shutdowns took hold this year, bankruptcies began occurring across multiple industries, from department stores to car rental companies.
“If there’s ever going to be a time when those assumptions go out the window — that two businesses in unrelated areas wouldn’t default at the same time — COVID-19 would be it,” Nickerson said. “If there’s ever a time when these deals should falter, it would be during a 'black swan' event.”
And yet, ratings agencies weren’t downgrading CLO tranches at a rate that matched the financial devastation facing companies.
To investigate, Nickerson and Griffin gathered data on tranche rating actions, collateral pricing information, and performance metrics, among other factors. They then constructed a sample of CLOs’ collateral holdings data between January and June 2020 and developed algorithms to test their predictions.
Their results showed that as the probability of defaults increased, tranche ratings should have been downgraded accordingly to reflect that the underlying pool became riskier. Since the downgrades weren’t happening, investors were putting their money into investments that were riskier than they appeared.
“If the credit ratings are wrong, investors will not only lose trust in the ratings agencies, they also could lose a lot of money,” Nickerson said.
The researchers identified two possible explanations for the disconnect in ratings:
Qualitative factors play a role
Nickerson noted that some ratings agencies may be incorporating non-model factors, or qualitative factors, into their assessments — for example, factoring in the past performance of collateral managers. If an individual managing a CLO has more experience, ratings agencies might be more comfortable sticking with a higher rating.
Citing prior research he conducted, Nickerson said that ratings agencies shied away from their models leading up to and during the financial crisis and placed more weight on these non-model factors that would allow them to justify a AAA rating.
Normally, ratings agencies publish documentation that outlines their modeling approach, and how they’ve arrived at their ratings. Since non-model factors aren’t disclosed to the public ahead of time, they can easily contribute to opacity and add up to trouble fast, Nickerson said.
Kroll Bond Rating Agency, for example, recently said it will pay $2.01 million, mainly in fines, to settle U.S. Securities and Exchange Commission civil charges that its ratings practices lacked “analytical method.”
“I do think when you see for instance, Moody’s saying, ‘Here’s the expected effect on tranches,’ and the downgrade on tranches doesn’t match that, it does suggest that non-model factors are getting a little more weight right now,” Nickerson said.
When building a CLO, some portfolio managers pick loans that appear safer from the perspective of a ratings agency — and would therefore earn a more positive rating, Nickerson said. One example: picking a loan with a two-year maturity versus four. In this case, the risk hasn’t changed, but the pool of loans is window-dressed to appear safer, Nickerson said.
“When COVID-19 hit, you actually saw managers start trading into these shorter-lived loans that from a modeling perspective would be assigned a lower risk weight for a lower default probability,” Nickerson said. “They’re taking actions that are going to reduce the risk through the lens of this model while not necessarily reducing overall risk from an economic standpoint. It’s window-dressing in some sense.”
Too easy to bet the wrong way
CLOs are “really complex products,” Nickerson said, which means it’s important for ratings agencies to clearly outline their guidelines for how they assign their ratings.
Looking back to what happened during the financial crisis, “many people were betting the wrong way on these things,” Nickerson said. “They didn’t really understand the risks involved. From that standpoint, given the complexity of a CLO I think investors tend to lean a lot more on ratings from credit rating agencies than perhaps they would when assessing an individual loan.”
Ratings agencies typically have “a regulatory stamp of approval” from the government and the SEC, Nickerson said, noting serious consequences if ratings agencies stray from their methodologies without disclosing it. If credit ratings are incorrect, investors won’t just lose trust in the ratings agencies; they’ll likely lose their money, too.
Nickerson encourages investors to do more due diligence and look beyond the ratings before they invest. Likewise, he hopes that policymakers take note of the disconnect between risk and ratings because it impacts investors through structured investments and pension plans.
“We don’t have the access to data that a government organization or oversight committee could get access to, just to verify that they [the ratings agencies] are following the statements they laid out and disclosed to the public,” Nickerson said. “That’s something that policymakers could dig into for more details if they wanted to.”