I started university in the aftermath of the Great Financial Crisis, which I was wholly oblivious to, because this was long before economics and finance appeared on my interest-radar. Given the uncertainty of what’s currently going on in today’s financial markets (and especially the uncertainty regarding what’s likely to happen), I therefore thought this would be a good time to make up for previous failings and to learn more about previous market crises.

One of the things that have struck me about market- and business-cycles is how similar their anatomy is, with the embers of growth springing from the ashes of what came before, and how irrational exuberance dominates the narrative at the peak (with references to how ‘it’s different this time’ becoming more and more common, as if it is a protective spell potent enough to ward off what’s coming).

Another similarity between then and now is the structural mis-estimation of risk that seems endemic in the lead-up to debt crises. For example, as I was reading about the Great Financial Crisis, I was intrigued by the many similarities in the narrative around the structural mis-estimation of risk in subprime mortgage lending (then) and the conversations around leveraged corporate loans (today). This intrigued me because mis-estimating the level of risk could lead to hyper-normal levels of debt defaults with a negative impact on high-yielding debt instruments in times of crisis.

The bulk of this blog post aims to provide a bit of background, but, to summarise: During the Great Financial Crisis, debt instruments like collateralised debt obligations (“CDOs”) were implicated in the financial instability that followed the weakening of the real-estate market. Altogether, the structural weaknesses that condensed out of the system in the form of CDOs were (partially) the result of banks hoping to make money by selling loans and investors ‘hunting for yield’ in a low-interest-rate environment, which was further exacerbated by lax regulatory standards that had caused lending standards to drop. Eventually, these structural weaknesses exposed the financial system to unaccounted-for vulnerabilities when the economy started slowing.

Today, CDOs have subsequently declined in popularity as part of the fallout from the Financial Crisis. However, many of the underlying drivers that caused the weaknesses to appear in the system have not been rectified since, and instead the last 10 years have seen increases in the popularity of a similar instrument, the collateralised loan obligation (“CLO”).

As I was learning more about CLOs, I grew a little bit concerned about what impact they might have on the financial system if growing numbers of companies start defaulting on their debts because of the economic fallout from COVID-19. This post is an attempt to outline some of my concerns. Do note, however, that I am not an expert on CLOs or other similarly-structured financial instruments. As a result, important nuance might be lacking and there might be unintended errors of omission. I apologise for these in advance,

What is a CLO?

First, it might be helpful to understand a bit better what a CLO actually is: CLOs are a type of CDO (a type of securitised loan) that pools together multiple loans made out to businesses (rather than individuals), with the aim of the pooling being to increase the available supply of lenders. Like CDOs more generally, this ‘securitisation’ of loans helps lower the cost of debt for the borrower but also makes it easier for the lenders to sell the loans to investors (and thus pass along the risk). However, while CDOs were growing increasingly complex in the lead-up to the Great Financial Crisis, today’s CLOs are less complex, and they don’t typically include credit-default swaps or securitisations. They also include more diversified collateral. This means that while CDOs and CLOs are similar in theory, they can be quite different in practise. See the table, below, for a side-by-side comparison of CDOs and CLOs.

A side-by-side comparison of CDOs and CLOs from the Bank of International Settlement’s (“BIS”) September 2019 Quarterly Review [link].

A characteristic of CLOs is that they are composed of ‘leveraged loans’, that is, loans made out to indebted companies. (CDOs, on the other hand, are typically composed of mortgages.) Because of the inherent risk of leverage, loans of this kind are normally not considered ‘investment grade’’. However, because of the pooled structure of a CLO (typically involving the pooling of 150 – 300 different loans), the constituent loans can collectively be considered more diversified than the individual loans themselves (more companies, exposed to different sectors). This is thought to lower the overall risk of the resultant security (the CLO), as long as the risks have been properly accounted for. In addition, while each individual loan comes with a set interest yield for the investor, the loans comprising the CLOs are collectively divided into ‘tranches’. This ‘tranching’ allows investors to choose what level of risk they are willing to accept for a particular interest yield.

To help visualise this, the tranches in a CDO/CLO can be thought of as similar to the floors of a building; being composed of lower, middle, and higher levels/tranches. The interest payments will flow from the higher tranches to the lower tranches, which make the higher tranches the most
likely to receive the payments. This makes the higher tranches the least risky. Conversely, the lower tranches will be last to receive the interest payments, which make them the most risky.

The lower tranches in a CDO/CLO are the riskiest because if a borrower should be unable to repay the loan, their default will be ‘captured’ by the lower tranches to insulate the higher tranches, which would not be affected by the default. An investor in this lower part of the CDO/CLO structure is compensated for taking on this higher level of risk because of the higher interest yield that is offered by that tranche.

Rating agencies like Standard and Poor’s and Moody’s help rate the tranches in the CDO/CLO according to the perceived level of risk; the diversified nature of the loans and their aggregate sensitivity. The higher tranches in a CDO/CLO will be rated more highly than the lowest tranches, and the ratings will span the spectrum from investment-grade for the higher levels all the way down to high-yield debt (junk) for the lower tranches.

Over the past 10 years, investors (banks, insurance companies, and asset managers) have been incentivised to take on more risk as they hunt for yield in the current low-interest-rate environment, and some of this demand has been satisfied by CLOs. (Some CLOs can yield higher returns than either high-yield bonds or private equity.) As a result, CLOs have exploded in popularity since 2012, and in December 2019 the amount of outstanding leveraged loan debt was estimated at around $1.2 trillion (in $US-denominated debt—which is similar to the size of the 2007 subprime mortgage market, of which CLOs are estimated to account for 50 %). Global estimates further peg the total amount of leveraged loans currently outstanding at closer to $2.2. trillion.

Much like the CDOs before them, the growing demand for CLOs has however resulted in declining debt covenant standards, and the last couple of years have seen a higher proportion of ‘covenant-lite’ loan issues and to more-heavily indebted companies. The infographic below is from BIS, again, which helps give an idea of how CLOs have grown in popularity and evolved since the Financial Crisis:

How CLOs could be potentially destabilising

One of the fundamental assumptions underlying the perceived risk of a CLO is that the economic environment will not deviate much from ‘normal’. Indeed, many financial models are (still) based on the assumption that events will follow a normal distribution, ignoring the growing body of work showing that the distribution of many natural (and economic and social) phenomena are better described using fatter-tailed, power-law distributions. Partly, this is because of the difficulties inherent to modelling such scenarios because of their non-linear nature.

In orthodox financial theory (which operates under the assumption of normality), unlikely events are extremely unlikely (often misunderstood as something like Taleb’s ‘black swans’). Conversely, under the assumption of a fat-tailed power-law-like distribution, unlikely events are much more common, and perhaps best accounted for as outliers. (To understand this, consider how the concept of outliers applies to the stock market, where extremely well-performing companies like Amazon or Constellation Software are not black swans but outliers that occur with a much-higher frequency than the normal distribution would suggest.)

To revert back to loans, the flawed assumptions around the shape of the distribution (normal vs fat-tailed) and the distribution and nature of the risk were the downfall of the CDO industry in the lead-up to the Great Financial Crisis. This is because the models used to assess the risk of the loans (mortgages) did not take the much-higher-than-assumed probability of mass-defaults of borrowers into account. Indeed, most models assumed no more than 5 % of borrowers defaulting at any given time, which would have been be tolerated by the CDOs. However, as a result, the financial system was under-prepared for dealing with the fallout of an outlier event occurring (where peak default-rates reached 40 %). The complexity of the CDO structure further exacerbated the impact.

Reading around CLOs, I am intrigued that some aspects of CDOs are true also of CLOs today. While CDOs have become more heavily regulated since the Financial Crisis, most of the new regulations have focussed on specific lending standards rather than the underlying assumptions regarding the distribution of risk. As a result, CLO marketing materials are free to focus on the ‘diversified nature of the loans’. Indeed, as one promotional flyer (from AXA [link]; emphasis mine) put it:

Of course, investors could lose a substantial portion of their investment in the case where multiple loans of the underlying pool default at a rate significantly exceeding levels anticipated at CLO inception. In the worst-case scenario, i.e. the event of default of the CLO, the holders off the AAA tranche can decide to liquidate the loan pool to repay note holders in order of seniority. While possible, the scenario where a CLO structure has been unwound due to an event of default, with no cash flows left to allocate to the equity tranche holders, has never occurred in the more than 20-year history of CLO issuance, even in the worst years of 2008-2009.

The overall message of the CLO industry has been that companies across industries are extremely unlikely to default on their debt en masse, with justifications being that ‘nothing of this sort has ever happened before’. Indeed, CLOs emerged from the Financial Crisis relatively unscathed, and now, risk assessments are (still) being based on (incomplete) historical data and poorly-fitted models (including the assumption that default rates are unlikely to top 5 % and that recovery rates will be around 70 %, when they could be half that in a worst-case scenario).

These assumptions have been put to the test since SARS-CoV-2 hit in the form a ‘perfect storm’ of cross-sector and -industry economic impact, putting unprecedented pressure on leveraged companies.

As it happens, leveraged loans are collectively sensitive to similar economic conditions, and when these conditions start to deteriorate, leveraged loans will start to be downgraded (signalling an increased perceived risk of default) and some of the loans will start to actually default. Loan downgrades will also have a direct impact on the leveraged companies that benefitted from the loans, and, for example, the Financial Times reports that some 20 % of CLOs (representing $600 billion of leveraged loans) have already been downgraded, while Moody’s has placed 20 % of their rated CLOs (40 % of which had been rated ‘AAA’ previously!) on a watchlist for potential downgrades.

Growing defaults will also make the leveraged loans increasingly illiquid, which will be exacerbated as some investors are forced to sell because of their investment criteria. This could have an indirect impact in affecting investor returns, and many could suffer losses.

Further impacts of CLO downgrades will be to reduce the availability of loans for struggling companies (because investors have gotten spooked) and the impact on borrowers and lenders alike could potentially exacerbate the impact of the economic slowdown regardless the size of potential stimulus packages. For example, the need for banks and insurance companies to liquidate assets to raise capital to satisfy capital requirement could lead to fire-sales, while the growing proportion of people nearing retirement age will be hit by declining asset values as their pension pots are impaired. Altogether, this could hurt the people and companies who are the most in need of help.

Fortunately, more than half of $US-denominated leveraged loans are due by 2024 and 2025 (with only ~$150 billion being due in 2021) but shrinking investor appetite for risky debt could have effects earlier than that. For example, MSCI estimates that if the current economic conditions persist, that leveraged loan defaults could reach “unprecedented levels”, with the rating agency Fitch estimating that leveraged loan defaults could reach as high as 10 – 15 %. (This is in excess of the 5 % risk assumption that is built into the CLOs.)

The pace of the CLO-rating downgrades could exacerbate issues further, by increasing the proportion of high-yield tranches in CLO investor portfolios. Again, much like during the Great Financial Crisis, this shows that the real, underlying, ‘fat-tail’ risk of the securities has not been well-understood.

So, why should we care?

I think developments in the CLO space will be important to monitor for people hoping to better understand the potential economic fallout from COVID-19 since CLO downgrades are correlated with economic performance as companies are cut off from access to funding. Therefore, one reason for my concern is the comparatively high yield of the riskiest CLOs, despite this still being a relatively early point for fully understanding the full economic impact of COVID-19. For example, CCC-rated yields today are equivalent to those seen during the more acute stages of the Financial Crisis in 2008/2009 (a full 2 years after the first signs of trouble had emerged and the fallout was becoming critical). 

Focussing on that part of history, the similarities between what was happening to the market in 2007 and what’s happening today are quite intriguing. To wit, the first signs of something being amiss in the lead-up to the Financial Crisis occurred in autumn 2006 as banks started to lose money on CDOs. February 2007 saw HBSC’s announcement that it had taken a surprising hit on subprime mortgage investments. After this, the market stabilised and the US Federal Reserve said in March that the impact was likely to have been “contained”. In the US, credit quality also improves in March/April because of tax returns. Meanwhile, the stock market seemed relatively unaffected—until more bad news arrived in June 2007. This was when the market reacted, and the fallout lasted for the next 18 months. 

I am intrigued by the similarities between the reaction of the 2007 and 2020 stock markets and the timeline of the developments. For example, while the market saw a strong impact from the growing spread of SARS-CoV-2 in February and March, it has since stabilised. Many Q1 numbers also did not take the economic impact of COVID-19 into account, and the Q2 numbers might therefore have a larger impact. (For example, the most recent earnings call of one of the companies most heavily exposed to CLO deterioration, Oxford Lane Capital, articulated that “We know the economy is going to have a horrible second quarter”.) Low underlying consumer demand will also exacerbate the current situation, since easing lockdowns will do little to reinvigorate suffering economies unless people feel safe enough to return to normal. This will require more effective policies around the economic reboot than we have seen to-date, and leveraged loan defaults are expected to increase from current levels through to the second half of 2020. 

In addition, analysts following the CLO market are expecting continued downgrades as more and more CLOs are failing their stability checks (i.e. the underlying debt pools are seeing higher-than-expected defaults), and some suspect that this is only the first act of something that will take longer to play out. (Indeed, Oxford Lane Capital, again, said in their most-recent earnings call that they are “concerned about a likely increase in defaults”.) Conversely to the last crisis, where banks were the worst hit from the CDO fallout (causing a structural financial crisis), the organisations that are exposed to the most risk from a potential CLO fallout today are the leveraged companies themselves and their investors (making this more of a corporate debt crisis).

Of course, government stimulus packages might be able to stave off some of the financial fallout from growing default rates, but they will also be far from perfect, since they won’t be addressing the underlying cause of the fallout: The virus itself and our primitive understanding of risk and the financial markets themselves.