Have you ever wondered how CLO AAA/AA/A/BBB/BB tranche ratings are created from a portfolio of non-investment grade loans?
This article will try to address this question simplistically.
The key to the creation of a solid CLO rating is the waterfall concept or credit subordination. Before we go into the credit subordination concept, let us take a closer look at the portfolio credit risk profile.
Three key factors drive the shape of the portfolio credit loss profile:
- Probability of default (PD) of each loan in the portfolio (driven by its credit rating and tenor)
- Recovery rate (or loss given default (LGD))
- Default correlation between loans (driven by sub-industry, industry, regions and country)
Let’s assume that the weighted average PD of the portfolio (say 50 credits spread across many industries) is 20% (over 8 years), and its weighted average recovery rate is 50%. Does this mean that if one were to run 10,000 scenarios (based on the Gaussian copula function), most scenarios would show an average portfolio credit loss of 20%*50% (LGD)= 10.0%? The answer is no.
Why? This is due to default clusters (as defined by default correlation).
Companies would have a higher default correlation with other companies operating in the same industry than those operating in different sectors. Not all companies in the same declining industry would suffer the same fate, as some might still emerge as winners even when their peers went out of business. Therefore, this could explain why default correlation within the same industry is typically lower than most people assume.
As companies are ‘real businesses’, they could cut costs, restructure and transform themselves when they run into trouble. Maybe this is why the track record of CLO rated tranches (backed by corporate loans) over several credit cycles have been impressive. More importantly, a typical CLO portfolio is well-diversified across regions and industries. What is the chance of seeing default clusters across several industries?
Going back to the shape of the portfolio credit loss curve, one would expect to see a tail risk (low chance of high losses). If we were to assume a 30% intra-industry default correlation and a 5% inter-industry default correlation, we would see the sort of credit risk profile as shown in the table below:
| Table 1 | Portfolio credit losses |
| Expected Loss (%) | 10.0% |
| Max | 30.0% |
| Value @ Risk (%) | |
| 5.0% | 18.8% |
| 4.0% | 19.3% |
| 3.0% | 20.0% |
| 2.0% | 21.0% |
| 1.0% | 22.8% |
| 0.5% | 24.3% |
| 0.1% | 27.0% |
As shown in table 1, the simplistic modelling result (from only 10,000 scenarios) shows a 0.5% chance (or less) that this portfolio might suffer 24.3% or more credit losses.
Therefore, if a CLO tranche can withstand that huge amount of credit loss, it deserves a solid credit rating. How can this be done?
This is where the waterfall or credit subordination concept comes into play. The AA rated tranche would be supported by lower-rated tranches (single A to BB tranches and excess interest cash flows). Hence, a AA rated tranche is protected and can withstand a considerable amount of portfolio credit loss.
For example, from the waterfall perspective, the Aa2 tranche is well covered by the portfolio notional of X. The over collateralisation ratio at the Aa2 level would be X/Y. Rated tranches could also be protected by additional interest cash flows that have been diverted (away from the equity tranche or lower-rated tranches) to pay down the AAA tranche upon the breach of any OC tests.
If a tranche could absorb up to 27% of credit losses under different default timing and interest rate scenarios without defaulting, this tranche could be rated at the AA level. Table 1 shows a 0.1% chance (or less) that the underlying portfolio could suffer 27% or more credit losses. When comparing 0.1% to the rating agencies’ PD benchmark table, we can determine the rating associated with this 0.1% PD – AA rating based on its weighted average life.
Of course, the above illustration is simple. The purpose of this article is to show how the CLO technology works – which allows various solid CLO tranche ratings to be created from a portfolio of non-investment grade loans.
Rating agencies have a highly comprehensive approach to rating CLOs. While Moody’s and S&P have their methodologies, the fundamental concept is the same even though Moody’s uses the expected loss (EL) approach. Nonetheless, there are two components involved – portfolio credit risk profile and cash flow modelling.
For example, Moody’s uses the expected loss (EL) calculation to rate a CLO tranche. They typically use the binomial expansion technique to associate asset default scenarios with the likelihood of each scenario (a default distribution). They then use cash flow modelling that relates each assets’ default scenario to the cash flows that the rated tranche receives in that scenario. After applying the default distribution to the cash flow model, they calculate the EL for each tranche. Finally, they compare the tranche’s EL to the relevant EL benchmark, based on its weighted average life, to determine the rating associated with such an EL.
On the other hand, instead of using the EL approach, S&P first assesses the credit portfolio scenario default rate (SDR), which corresponds to the level of defaults that is likely to affect the portfolio in a given rating stress scenario. As a second step, S&P analyses the transaction’s cash flows and payment profile. S&P will test the various scenarios, based on key rating drivers to determine the maximum level of defaults that a transaction may sustain while still repaying the noteholders in full and on time. This is the breakeven default rate (BDR). To assign a rating at a given level, S&P looks for the SDR commensurate with that rating to be at or lower than the BDR.
While Moody’s uses the EL approach and S&P uses SDR vs BDR approach, there are two components involved – portfolio credit risk profile and cash flow modelling.
Disclaimers
The information, research, data, research related opinions, observations and estimates contained in this document have been compiled or arrived at by CLO Research Group, based upon sources believed to be reliable and accurate, and in good faith, but in each case without further investigation. None of CLO Research Group or its service providers; authorised personnel, or their directors make any expressed or implied presentation or warranty, nor do any of such persons accept any responsibility or liability as to the accuracy, timeliness, completeness or correctness of such sources and the information, research, data, research related opinions, observations and estimates contained in this document. All information, research, data, research related opinions, observations, and estimates in this document are in draft form as of the date of this document and remain subject to change and amendment without notice. Neither CLO Research Group nor any third-party providers shall be subject to any damages or liability for any errors, omissions, incompleteness or incorrectness of this document. This article is not and should not be construed as an offer, or a solicitation of an offer, to buy or sell securities and shall not be relied upon as a promise or representation regarding the historical or current position or performance of any of the deals or issues mentioned in it.
