"The path to default is beset on all sides by the inequities of economic downturn and the tyranny of underperforming credits. Blessed is the manager who, in the name of his O/C ratio, shepherds the trust through periods of turmoil. For he is truly the investor's keeper and the finder of lost value. And I will strike down upon thee with great vengeance and furious anger those who attempt to par build in the wrong industries. And you will know I am the Market when I lay my vengeance upon you." (With apologies to "Pulp Fiction.") 

A collateralized loan obligation (CLO) default has historically been a rare event. From 1994 to 2017, S&P Global Ratings issued just under 10,000 U.S. CLO ratings, of which fewer than 0.4% have defaulted--the majority were subordinate and mezzanine tranches from CLO 1.0 transactions (those issued pre-financial crisis) strained by the financial crisis and defaulting in the years following. However, in November 2017, the rare default became reality for Airlie CLO 2006-II Ltd. (Airlie), and for the first time in three years, we lowered our rating on the class D note to 'D (sf)' from 'CCC+ (sf)' (see "One Airlie CLO 2006-II Ltd. Rating Lowered To 'D (sf)' And One Discontinued," published Nov. 7, 2017), about 11 years after it originally closed on Dec. 21, 2006. The downgrade of the class D note, which was originally rated 'BB (sf)', reflected the unpaid principal balance (13.59% of the original) after the optional redemption on Oct. 20, 2017. It raised the cumulative U.S. CLO default count among originally 'BB (sf)' rated tranches to 20.

What went wrong? We took a closer look at the deal's structure and its overcollateralization (O/C) timeline, as well as what impact the manager's trades had on the par loss the CLO was experiencing. Airlie will not likely be the last CLO to default, but looking at this example could help in examining future CLO defaults.

The Deal's Structure: Nothing Too Out Of The Ordinary

This transaction's structure was not materially different from other 2006 vintage transactions, or the CLO 1.0 transactions that came after it. Airlie did, however, have a somewhat longer reinvestment period of seven years. Although the average reinvestment periods of CLOs steadily increased for the 2003 through 2007 vintages, as of 2006, the average reinvestment period was around six years.

Airlie reinvested through the financial crisis in 2009-2010, concluded its seven-year reinvestment period in early 2014, and amortized through the energy/retail downturns starting in 2015. After sustaining significant stress during the financial crisis--when the subordinate O/C test briefly dipped below 100%--the test reached a post-crisis high of 103.74% in January 2012. Airlie didn't seem like an obvious candidate for default--with two years remaining in its reinvestment period, it seemed like it had a chance to continue building up credit and par after rebounding from the rough patch.

The timing of the oil and gas and retail downturns in 2015-2017 ended up being somewhat unlucky for this transaction. Amortizing through a downturn (even a sector-specific one) can be challenging for a CLO as the manager's flexibility to manage the portfolio and recovery becomes limited. Of the U.S. CLO tranches that have defaulted (see "Twenty Years Strong: A Look Back At U.S. CLO Ratings Performance From 1994 Through 2013," published Jan. 31, 2014), several were from CLOs that were amortizing through the financial crisis, highlighting the importance of positioning the portfolio properly to enable its smooth run-off during the amortization period.

Airlie CLO 2006-II's class D closed just a few years before the height of the financial crisis. The impact of the crisis on this portfolio was similar to that of other CLO 1.0 transactions: downgrades of the corporate ratings and a negative swing in the loan prices contributed to O/C test declines and failures. The class D O/C ratio of Airlie failed for most of 2009 and early 2010, causing about $13 million of interest proceeds to be diverted to pay down the senior class A-1 notes' principal instead of the junior CLO notes' interest and equity. But this was the CLO behaving as it should: as economic conditions caused a loss of par and credit, excess spread was diverted away from the lower part of the CLO capital structure and used to restore credit enhancement and pay down the CLO's senior notes. This helped to somewhat improve the class D O/C ratio after the second half of 2009. By the start of 2012, it had reached a post-crisis high of 103.74%; however, defaults started to accumulate in 2012 beginning with Quiznos' default in late 2011. This caused the first in a series of noticeable blips in the CLO's O/C ratios starting in 2012. When the reinvestment period ended in January 2014, the class D O/C had declined to 102%, leaving the CLO with less cushion to endure the impending energy and retail slowdown.

Chart 1 shows the full timeline of the class D O/C ratio.

While these events shed some light on the matter, the questions still remain. Why did this CLO default while most of the other CLO 1.0s did not? Why did it take this long to default? The answer may lie in the manager's trades.

The Portfolio And Its Evolution: Manager Trades To Improve Declining Portfolio Credit Quality At The Cost Of Par Loss

As of April 2007, during the beginning stages of Airlie, the initial portfolio had about 106 issuers with an S&P Global Ratings weighted average rating factor (WARF, or weighted average five-year probability of default; see Appendix for more details) of 2,752, a weighted average life (WAL) of 6.4 years, a weighted average spread (WAS) of 2.8%, and initial par balance of $436 million (including cash). There was $26.5 million of subordination (the portfolio par balance minus the sum of rated liabilities), or 6.08% of credit enhancement (subordination divided by the initial par balance).

Jump forward to January 2014, after the transaction ended its seven-year reinvestment period, the portfolio had about 127 issuers with an S&P Global Ratings WARF of 2,547, a WAL of 4.0 years, a WAS of 3.5%, and par balance of $406 million (including cash). By the end of the reinvestment period, the portfolio par balance had declined by about $30 million, or about 7% of the initial April 2007 portfolio par balance, greater than the 6.08% worth of initial credit enhancement! Luckily, the interest diversion mechanism of the CLO structure caused $13 million worth of interest proceeds to be diverted to partially pay down the class A-1 notes, retaining about $10 million of subordination, or 2.53% of credit enhancement at the end of the reinvestment period (relative to the values as of April 2007, a decline in subordination of $16.5 million, or credit enhancement of 3.6%). With just over 6% enhancement at close and a par loss of about 7%, without the structure's protective mechanisms, the recovery of the class D notes would have arguably been much worse (if any) rather than getting the 86% recovery it ended up with.

The initial portfolio (as of April 2007) was built during a benign economic period. Given the stresses this CLO faced during the financial crisis and afterwards, however, nearly half of the loans in the initial portfolio were downgraded before their maturity dates and about 13% of the portfolio had ratings lowered to non-performing. Some of the par loss incurred was due to the manager's efforts to remove these problem assets during the downturn. By the end of the reinvestment period, over 90% of the names in the initial portfolio had been removed, and the manager had traded into and out of 385 issuers over the course of the seven-year reinvestment period. The WARF of the new purchases during the reinvestment period (at the time of purchase) was 2,693, slightly better than the WARF of the initial portfolio assets (2,752). However, given the economic stress Airlie suffered during its 11-year life, 45% of the assets acquired ended up being downgraded, and 15% saw their credit ratings lowered to non-performing before the loan maturity date. Many of the loans that ultimately defaulted were purchased during the first half of the reinvestment period.

The portfolio ended the CLO reinvestment period more diversified and with a better WARF, partly due to its greater exposure to a handful of investment-grade assets. However, the portfolio's par balance had declined by 7%. Defaults continued to afflict it during the amortization period, many of which came from purchases made during the latter half of the reinvestment period. In chart 2 below, we analyze the credit history from the month of purchase through the maturity dates for each of the new purchases made per year (see Appendix for the methodology used for the chart). We can analyze these trades into new assets by looking at each loan issuer's credit rating history.

We found that during the latter half of 2007, the manager completed a moderate volume of trades: they purchased relatively riskier assets (compared to the initial portfolio), several of which eventually defaulted. Leading into the crisis in 2008, the manager made fewer new purchases. The WARF of the new purchases were in line with that of the initial portfolio. But during the crisis in 2009, the manager made the largest volume of new purchases, of which, most were in line with the WARF of the initial portfolio. The average WARF of these purchases was higher though (10,000 averaged for non-performing ratings), due to one loan from an issuer that had a non-performing credit estimate at the time of purchase.

As the dust settled by the second half of 2010, the volume of trades declined. The credit quality of the new purchases this year was higher, of which, just one issuer defaulted. Like other later-vintage CLO 1.0s, Airlie had a low cost of funding due to tighter spreads on pre-crisis CLO liabilities; as underlying loan spreads increased during and after the crisis, the portfolio WAS increased well above the target spread. With portfolio WAS values well above their triggers, some managers were able to purchase lower-yielding credits with higher credit ratings. The lower (i.e., better quality) WARFs of the new trades done in 2010-2012 were partially due to purchases of lower-yielding investment-grade assets (some of which were purchases of other CLO tranches).

In 2013, the manager made several purchases into issuers that were then rated 'CCC', making the WARF of that year's trades higher than others. Some of these 'CCC' purchases were from energy and metals issuers that eventually defaulted during the looming energy slowdown. The manager did not focus heavily on energy-related issuers, but by diversifying into the energy sector, especially with the 2013 purchases, the portfolio saw an increase in defaults in the following years.

As the reinvestment period wound down, the volume of new purchases declined in 2014, and several purchases were again made into investment-grade credits. None of the trades done in the final year of reinvestment had defaulted by fourth-quarter 2017.

Overall, the manager did remove some assets before they deteriorated in credit quality, and actively removed non-performing assets, keeping the proportion of defaulted loans from getting too high. They also sold some assets before they defaulted: of the 62 purchased assets that were lowered to non-performing ratings before their maturity dates, only 42 of these exposures were still held within the portfolio after the rating downgrades; meaning, 20 names were cleared out before their ratings were lowered to non-performing.

Manager Trades Were Ultimately Ineffective In Maintaining Or Improving The Portfolio

In the end, Airlie CLO 2006-II found itself exposed to two sets of economic stresses: one broad (the financial crisis) and the other sector specific (stresses from the energy, commodities, and retail sectors in 2015 and beyond). The trades collateral manager made were not able to preserve the portfolio's value given the stress this transaction endured during its 11-year life. A significant amount of par was lost during the financial crisis and after, cash was reinvested into energy-related issuers in the ensuing years when valuations were high. Several of the purchases made during this period ultimately soured, and when the energy slowdown was in full swing in late 2015 and 2016, depressed loan prices made it impossible to liquidate the portfolio. Market prices did recover by 2017, but not after a few more names experienced deterioration. The transaction started off with about $26.5 million worth of subordination underneath the class D notes. After 11 years, the class D notes experienced a loss of $2.6 million after the transaction captured about $13 million in excess spread, meaning the portfolio lost, in total, about $42.1 million worth of par, or about 9.7% of the par balance of the initial portfolio.

The events that led this CLO note's default may not be unique to Airlie. At the start of 2018, there were a small number of S&P Global Ratings-rated U.S. CLOs with subordinate O/C values in the sub-104% range. Some of these transactions are from earlier vintage CLO 2.0s that have lost par due to the energy and retail slowdown. Their low O/C ratios expose the subordinate notes of the CLO to the risk of downgrades should the portfolio's credit quality continue to deteriorate. These transactions may also be less prepared to withstand further par losses during their amortization periods. We will continue to monitor these transactions, and we will take appropriate rating actions as we deem necessary.

Appendix

S&P Global Ratings' weighted average rating factor (WARF)

The S&P Global Ratings' WARF of a portfolio is the five-year probability of default given the S&P Global Ratings' credit rating of the CLO portfolio assets and the S&P Global Ratings' default table used in CDO Evaluator E7.2, weighted by each assets par balance and then multiplied by 10,000. WARF does not factor in the tenor of the asset. The five-year default scale was chosen for the WARF since the average weighted average life of reinvesting CLOs has been closer to five years.

Methodology for chart 2

In chart 3, we aggregate the balances of new purchases that occurred during each year within the bars, as reported by the monthly trustee reports. We define new purchases as loan exposures that appear in the trustee report that was not present in the prior month's trustee report. We focused on the new purchases because it represents a clear decision by the manager to add exposure to a particular issuer. A loan that dropped out of the portfolio may have been sold at the manager's direction, or if could have been redeemed, which is not at the direction of the manager. Thus, we focused on purchases into new issuers to reflect the manager trades. The sum of the par of these new trades per year can give an indication to the level of trading activity the manager engaged in during the year. We also calculated the WARF of the new annual additions based on the credit rating as of the month the issue first appeared in the portfolio, reflecting the credit quality at the time of the new purchases for each year. We aggregate the lowest rating of each of these trades between the month of purchase and the maturity date of the loan (the expected exposure period) within the bars, splitting each bar into three cohorts: affirm or upgrades, downgrade, and default. We focus on the lowest rating of the expected exposure period because by purchasing an asset of a particular tenor, the manager exposes the portfolio to the asset's potential risk, even if the deterioration occurs before the asset's maturity. (For example, the manager purchased a five-year loan from an issuer rated 'B+' in 2007. After two years, the issuer experiences credit deterioration and is downgraded to 'CCC'. Even if the issuer emerges from this event the year after and is raised to 'B+', and then maintains this rating at the loan's maturity, we will still capture this trade as a downgrade within the 2007 bar even though the loan matured with the same rating as when initially purchased. By purchasing this loan, the manager exposed the portfolio to a distressed issuer in 2009. If the manager intended to do a credit risk sale in 2009 when the issuer was rated 'CCC', the deal would have likely taken a par loss from this potential trade). The framework used does not consider adding additional exposure to an issuer that already exists within the portfolio as a new purchase.