On June 21, 2019, S&P Global Ratings published its revised criteria for rating collateralized loan obligations (CLOs) and collateralized debt obligations (CDOs) backed by corporate assets, including loans, bonds, and synthetic assets (see "Global Methodology And Assumptions For CLOs And Corporate CDOs"). We are publishing this article to provide answers to questions we believe market participants may have regarding the updated criteria, and to help them better understand the updated framework.

We believe the changes we're making to our CLO and Corporate CDO criteria are necessary to better align our ratings on these transactions with our ratings definitions (see "S&P Global Ratings Definitions," Oct. 31, 2018) and the economic stress scenarios we associate with them (see "Understanding S&P Global Ratings' Rating Definitions," June 3, 2009). Based on data we have reviewed and research we have done, some of which is outlined in the responses to the questions below, we expect that the updated criteria will enhance ratings comparability and produce rating outcomes that are more consistent with our ratings definitions, associated economic stress scenarios, and expectations.

Frequently Asked Questions

Can you summarize your current rating framework for analyzing credit and cash flow risk in a CLO portfolio?

In 2009, we updated the fundamental assumptions we use in our analysis of CLO and corporate CDO transactions (for ease of reference, we'll refer to these as "CLOs" in the remainder of this article). In our credit analysis of a CLO portfolio, we seek to assess the proportion of assets that we expect will default when subjected to different levels of economic stress we associate with our rating levels. For example, obligations with a 'AAA' rating from S&P Global Ratings (across all asset types, not just CLOs) are intended to survive an "extreme" level of economic stress without defaulting. We associate an extreme level of economic stress with downturns such as the Great Depression, which afflicted the U.S. economy in 1929 and the 1930s. Thus when analyzing whether or not a CLO tranche can be assigned a 'AAA' rating, we begin by using our CDO Evaluator credit model, which incorporates our criteria assumptions, to estimate the gross level of defaults the CLO portfolio might experience under a hypothetical extreme economic stress that we view as commensurate with a Great Depression scenario. This linkage of economic stress levels with ratings is why we refer to the CDO Evaluator model output at each rating level as a "scenario default rate" (SDR).

In our CLO rating process, the SDR at the rating level for a tranche represents the gross level of cumulative asset defaults the tranche will have to withstand in our modeling while still paying all of its required interest and principal. Our cash flow modeling then estimates the level of asset defaults a CLO tranche can withstand under various default timing and interest rate scenarios without missing any of its required interest or principal payments--a number we refer to as the tranche's "breakeven default rate" (BDR). In order for a tranche to be assigned a given rating, the BDR for the tranche generally must equal or exceed the SDR at the tranche rating level, in addition to meeting other rating considerations. Chart 1 provides a high-level overview of our CLO rating process.

What was the process through which you calibrated the criteria and updated the assumptions in CDO Evaluator in connection with the criteria update?

To calibrate our criteria and CDO Evaluator, we started by determining target portfolio default rates that we believe a hypothetical CLO portfolio (which we refer to as the "archetypal portfolio") would suffer under two different rating stress levels: 'AAA', which is intended to survive an extreme level of economic stress commensurate with the Great Depression, and 'BBB', which we associate with a moderate level of economic stress. For purposes of our CLO criteria, we associate a 'BBB' level of economic stress with the worst corporate default rates seen since 1981. The target portfolio default rates we set in connection with the CLO criteria calibration process are the level of gross cumulative asset defaults we would expect to see for our archetypal pool--the hypothetical pool of CLO assets we referenced above--under the economic scenarios above.

The 'AAA' target default rates in the updated criteria are unchanged from the targets in the September 2009 criteria. The 'BBB' target default rates have been slightly adjusted from the September 2009 numbers to reflect additional global historical data, but this change did not substantially affect the results.

Tables 1 and 2 in the CLO criteria provide the target portfolio default rates we developed for the 'BBB' and 'AAA' stress levels, respectively. By way of example, for an archetypal CLO pool comprising five-year assets from 'B' rated companies, we estimate 48.5% of them would default under a moderate ('BBB') stress and 68% would default under an extreme ('AAA') stress.

With our 'AAA' and 'BBB' target portfolio default rates in hand, we then used our hypothetical CLO archetypal pools to calibrate the model assumptions in our CDO Evaluator credit model to come close to producing SDRs commensurate with them. In the revised criteria, we redefined the archetypal pools with which the target portfolio default rates are associated, to make their level of diversification more consistent with the average diversification we have seen in securitized pools in the past decade and closer to the real diversification in the economy.

The key model assumptions within CDO Evaluator include the following:

• Asset default rate assumptions, informed by historical data, for assets rated 'AAA' through 'CCC-' and with years to maturity ranging from one to 30 years;
• Pairwise asset correlation assumptions that are also informed by historical data and which vary depending whether two assets come from companies that operate within the same industry, country, and region; and
• Rating quantile assumptions (or "cut points") the model uses to convert the output of its default simulation into SDRs for specific rating levels, which are designed to achieve as close as possible the target SDRs for the archetypal pools.

Once the calibration of CDO Evaluator was established in this way, each CLO portfolio, with its specific portfolio composition and characteristics, can be analyzed by the model and yield an SDR for a given CLO rating level that we believe reflects the economic scenarios we associate with that rating level, consistent with our ratings definitions criteria.

What is the effect of the changes you are making to your credit analysis in CDOs?

We are making two types of changes to our credit analysis that have an impact on the SDRs and BDRs:

• Updating default rates and rating quantiles for our CDO Evaluator model, and
• Analyzing corporate exposure using the pool's actual maturities (as opposed to maturities extended to mimic the effect of reinvestment) for transactions in which the manager uses a CDO Monitor test as a reinvestment condition.

Results may vary from transaction to transaction and by rating level, but our testing on CLO pools we rate indicates that the aggregate effect of these changes may be a decrease of approximately 3%-5% on average to the SDRs produced by CDO Evaluator. The update to default rates and quantiles is informed by historical data, as well as by an additional analysis of the sensitivity of the default rate of 'B' and 'BB' rated obligors to macroeconomic variables, which supports that the updated calibration would produce ratings that remain consistent with our ratings definitions. The change to maturity assumptions increases the consistency between our rating analysis and the conditions governing reinvestment in transaction documents, where these include the use of a CDO Monitor test.

We are also making changes to our cash flow analysis, primarily for simplification purposes, which are described in more detail later in this article. These changes had no impact on the SDRs generated by CDO Evaluator.

Why are you updating your CLO criteria now, when there appears to be consensus in the market that we are close to a turn in the credit cycle?

We are making these changes to reflect observations from the past decade. We believe the changes we're making to our CLO and corporate CDO criteria are necessary to better align our ratings on these transactions with our ratings definitions and the economic stress scenarios we associate with them.

The changes to our criteria incorporate recent performance data. We now have 10 years of additional data on corporate ratings performance since we introduced our prior CLO rating framework in 2009, which was created using data up to 2008. Furthermore, with CLO 1.0 transactions (CLOs issued before the global financial crisis) almost all fully redeemed, we now have a more representative picture of global CLO performance over the past 20-plus years--a period that spanned multiple economic downturns. The changes in the criteria also aim to provide an update and simplification of our rating approach and make it more transparent.

We note that our CLO criteria are not primarily calibrated to the recession in 2008-2009, and are not specific to any one point in the economic cycle. Our ratings are calibrated to economic stress scenarios outlined in our "Understanding S&P Global Ratings' Rating Definitions" criteria, which we are not changing. These ratings definitions do not vary as a function of where we are in the credit cycle; as such, our CLO criteria are intended to produce ratings which are able to survive their relevant level of economic stress regardless of where we may be in the cycle.

Why do you feel comfortable making these changes?

In connection with criteria recalibration process, we conducted an analysis to assess the consistency of the CLO ratings that would result from the revised approach with our ratings definitions.

First, we determined that the criteria changes are supported by a macroeconomic analysis that links corporate defaults in an extreme stress scenario to economic variables. As there wasn't a leveraged loan market in the 1930s, we analyzed the dependency of corporate default rates on key macroeconomic variables in order to build a relationship between the rates and the variables and make inferences about the projected levels of default in an extreme stress scenario commensurate with the Great Depression.

These macroeconomic variables included the U.S. Treasury yield curve, return volatility on the S&P 500 stock market index, real GDP growth for the U.S. economy, and credit spreads between corporate bonds with different ratings. We looked at historical data and used a regression analysis of these economic parameters that we have observed during more recent downturns. We then used actual macroeconomic variables from the Great Depression to estimate the level of gross cumulative defaults that corporate assets might experience under an extreme stress scenario. The analysis showed that there is a range of projected corporate default rates, depending on the specific scenario used. We used this analysis to assess how well the calibrated model reaches the set targets.

We concluded that the projections for tenors of five years and more support our 'AAA' targets and exhibited a range of potential outcomes as we had expected, depending on the specific scenario used. This also supports the fact that the modestly lower SDRs triggered by the change to the criteria calibration remain consistent with what we might expect in an extreme level of stress. The CLO ratings resulting from the revised approach are in line with our ratings definitions.

In addition, we back-tested the revised CLO criteria against corporate defaults from the 2008-2009 downturn. We tested Standard & Poor's CDO Evaluator credit model (inclusive of the criteria changes) on about 500 CLO 1.0 portfolios to see how many defaults the model would predict for each portfolio under a modest ('BB') economic stress and a moderate ('BBB') economic stress. We chose these scenarios because we believe they define the boundaries of stress levels by which the pools we tested were affected through their life, including the 2008-2009 recession. To complete our back-testing analysis, we compared the projected portfolio default rates from CDO Evaluator (i.e., the SDRs) at these 'BB' and 'BBB' rating levels with empirical default data for each of the portfolios during the downturn and the years that followed.

The results of the back-testing showed that our updated CDO Evaluator model remains consistent with our ratings definitions. Charts 2 and 3 below show this graphically, with the amount a marker (each representing a CLO portfolio) is above the diagonal line indicating the extent to which CDO Evaluator's estimate of portfolio defaults at each stress level exceeded the actual observed defaults during and after the downturn.

Finally, we also take note of the strong performance of the CLO 1.0 transactions rated before the 2008-2009 downturn and the CLO 2.0 transactions rated since. Ten years on from the global financial crisis, we now have more visibility on the performance of the CLO 1.0 transactions through the most recent stress period. Even after withstanding an economic downturn, these CLOs saw few tranches default (see tables 1 and 2 below). Because the post-crisis CLO 2.0 transactions generally have more par subordination than their pre-crisis CLO 1.0 siblings, we would expect these transactions to see even fewer tranches defaulting under a global financial crisis-type downturn, all else being equal.

Table 1

Table 2

With the increase in leverage and weaker covenants in leveraged loans, the market generally expects recovery rates to be lower than they have been in the past. Do your recovery rate assumptions reflect this risk?

We did not change our recovery rate assumptions from our previous criteria. These can now be found in "Guidance: Global Methodology and Assumptions for CLOs and Corporate CDOs," June 21, 2019. Our CLO rating methodology captures shifting risk conditions in the loan market through its use of recovery ratings ('1+' through '6'), and these are currently available for more than 95% of the assets found in U.S. and European broadly syndicated loan (BSL) CLOs. Recovery ratings reflect our forward-looking view of recovery prospects for a specific loan or bond of a rated company and are intended to take into account debt characteristics such as higher leverage, smaller debt cushions, covenant-lite status, and other factors we believe will influence the level of recoveries in the case of a default.

For example, in recent years, the proportion of loans with lower recovery ratings has increased substantially (see "When The Cycle Turns, Assessing How Weak Loan Terms Threaten Recoveries," Feb. 19, 2019; "Lenders Blinded By Cov-Lite? Highlighting Data On Loan Covenants And Ultimate Recovery Rates," April 12, 2018; and "Lean Senior Debt Cushion Threatens Recovery Prospects For U.S. Leveraged Loans," Nov. 30, 2017). In our rating approach for CLOs, weaker recovery ratings directly lead to lower recovery assumptions in our cash flow modeling and, hence, higher CLO tranche enhancement requirements, all else being equal, for a given CLO rating level.

The shift in recovery prospects can be seen in the recovery ratings assigned to newly issued loans in the U.S. from 2007 through 2018. In the years following the 2008-2009 downturn, many corporate loans were structured with lower leverage and greater debt cushions, and this was reflected in the recovery ratings we assigned to the loans in those years. In 2010 and 2011, we assigned recovery ratings of '1,' '2' and '3'--representing expected recovery outcomes of 90%-100%, 70%-90%, and 50%-70%, respectively--in roughly equal proportion. Starting in 2012, loans with a recovery rating of '3' began to represent an increasing proportion of loan issuance, to a point where, in 2018, they represented 60% of all new-issue loan recovery ratings--three times the proportion of the next highest category. This decrease in recovery ratings since 2012 has led to an increase in CLO par subordination associated with a given CLO rating level over the period.

Chart 4

We also reviewed the calibration of recovery rate assumptions in the CLO criteria for assets that don't have a recovery rating against the most recent historical observations, considering the level of stress experienced at the time, and concluded that those assumptions remain valid.

Whether we use recovery ratings or not, our recovery assumptions decrease for higher CDO note ratings, with the severity of the haircut being the greatest when doing analysis of 'AAA' rated CLO tranches. For example, when analyzing a 'AAA' CLO tranche, the recovery ratings for the loans in the collateral pool provide for an average recovery assumption in the low 40s for a U.S. BSL CLO, and high 30s for a European BSL CLO. This reflects empirical evidence that recovery rates are inversely correlated to default levels. The lower recoveries are in line with the expectations for the credit cycle, where higher defaults and a lack of liquidity will likely increase the number of businesses that liquidate rather than restructure, thus putting a stress on recoveries. It also reflects the general conservatism of the assumptions we use when analyzing a 'AAA' rated CLO tranche.

What are the other changes you're making to the criteria?

In addition to the changes to the criteria and CDO Evaluator described above, we have made other changes to specific areas of our analysis, including the following:

• Adjusted our assumptions regarding the tenor of the exposure to the assets for transactions in which the manager commits to maintain or improve the consistency of the portfolio's credit quality with the notes' original rating as a condition to reinvestment, for example by using S&P Global Ratings' CDO Monitor tool. In this case, in both our credit and the cash flow analysis, we generally propose to base our analysis on the assets' actual maturities, as opposed to simulating potentially longer tenors to reflect the expected effect of reinvestment. This is because such tests incorporate the effect on the transaction's credit risk of extending asset exposure through reinvestment.
• Updated the asset amortization profile we use in our cash flow analysis of CLO transactions to reflect typical CLO portfolios we have seen. This is because portfolios are typically not fully invested at closing and a portfolio's maturity profile could change over time due to the reinvestment process.
• Updated our assumptions for default timing and patterns consistently with our global cash flow criteria, taking into account the assets' characteristics in our cash flow analysis. For pools of leveraged loans from issuers typically rated in the 'BB' to 'CCC' range, for example, we assume defaults are more likely to occur in the early years of the transaction, while scenarios in which no defaults occur for the first few years of the transaction are less relevant given the assets' risk profile.
• Differentiated our analysis of foreign exchange exposure in CDOs based on the transaction's potential sensitivity to this risk. This is to reflect the various characteristics we have observed across transactions historically.
• Removed the percentile approach to BDRs in our cash flow analysis, given our focus on a smaller number of the most relevant cash flow runs, and consider the minimum BDR for comparison with the portfolio's SDR.

We did not make material changes to the determination of the rating inputs, industry classifications, supplemental tests, or the recovery assumptions.

What is the expected impact of these changes on CLO tranche par subordination?

Our testing indicates that the effect of the criteria update on our existing CLO ratings is generally positive, with an average increase in rating cushion of approximately 4%. The cushion refers to the difference between the SDR from CDO Evaluator at a given rating level and the level of defaults that a tranche can withstand in our modeling (the BDR from our cash flow model, Cash Flow Evaluator). Because we don't generally upgrade ratings on actively managed CLOs during the reinvestment period, we expect the rating impact of these changes to be limited, with rating changes mostly affecting subordinate tranches of CLOs in their amortization phase, and a large majority of these rating changes being upgrades.

Under the updated criteria, new-issue CLO transactions may see different rating cushions, CDO Monitor Test cushions, and subordination levels compared to what may have been seen under the prior criteria. Subordination varies depending on different variables, such as the levels of asset and liability spreads, overcollateralization test thresholds, and other structural factors. The level of rating cushion for a CLO tranche can affect the likelihood that a rating may be lowered if the collateral were to experience deterioration, and the CDO Monitor Test cushion for a given CLO transaction can affect the likelihood that the CDO Monitor test may fail. All other things being equal and assuming no increase in CDO Monitor buffer or rating cushion, we think the increased cushion due to the criteria recalibration could convert into a roughly 3% decline in subordination for a typical CLO tranche. For context, prior to the criteria update, the typical par subordination for a CLO tranche rated 'AAA' by S&P Global Ratings from a transaction with a four-year reinvestment period ranged from 37% to more than 40%.

Related Criteria

• Global Methodology And Assumptions For CLOs And Corporate CDOs, June 21, 2019
• Understanding S&P Global Ratings' Rating Definitions, June 3, 2009

Related Research

• Updated EMEA Corporate Recovery Calibration: Q1 2019, Feb. 8, 2019
• A 10-Year Lookback At Actual Recoveries And Recovery Ratings, Feb. 4, 2019
• 2017 Annual Global Leveraged Loan CLO Default Study And Rating Transitions, Oct. 9, 2018
• S&P Global Ratings Definitions, Oct. 31, 2018
• U.S. CLO 1.0 Chapter To Close With Minimal Defaults, Aug. 2, 2018