Synthetic CDO ‘First Loss’ tranche safer than ‘Second Loss’ Tranche
In the past few weeks there has been some more noise from regulators and the rumor mill related to synthetic CDO’s. It hasn’t hit the frenzy surrounding the Abacus deal last year, but the questions are starting to pop up again. Who was long what? Who was short what? Who picked the portfolio? Etc.
While all that is important, I figured now is a good time to trot out one of my favorite pieces. In many cases the ‘first loss’ or ‘equity’ or ‘unrated’ tranche was actually safer than the ‘second loss’ or ‘mezz’ or ‘rated’ tranche above it. Does this make any sense? On the surface, NO! An investor would expect the more senior tranche to underperform the ‘equity’ in good scenarios and outperform in bad scenarios. That makes sense, yet many of the synthetic CDO’s were created in such a way that there was almost no scenario where the second loss tranche outperformed the equity tranche. Bizarre, but true.
It’s important to understand this phenomenon as it helps explain a lot of the strategies that were employed and also why the defense of ‘I owned the equity’ is not a very strong defense. I will now try and walk you through how this paradox is possible.
Cash CDO vs Synthetic CDO
The best way to demonstrate ‘safe’ or ‘free’ equity is to show how we got there and why its unique to the synthetic CDO’s. Free equity does not exist in the cash CDO world (maybe that’s why that market was a fraction of the synthetic one?).
I will walk through 2 hypothetical examples that illustrate the point. The example is more similar to a corporate CDO than a mortgage backed one, but the principals remain the same, and in reality, the trade first appeared in the credit referenced world and then moved to the mortgage world as that market developed.
For a cash CDO, let’s assume we have 100 bonds each paying T+150, each with a maturity of 5 years. Let’s assume that the 5 year treasury yields 4.5% (reasonable at the time most of these deals were created). So we have a portfolio of 100 bonds paying 6% each.
For a comparable synthetic CDO, let’s assume we have 5 year CDS on the same 100 bonds (issuers) and each CDS pays 100 bps. That’s a reasonable ‘basis’ where the 5 year bond pays T+150 and the CDS to the same maturity is 100 bps. Its simplified for our purposes, but realistic.
If each position is $10 million, then the $1 billion bond portfolio generates annual income of $60 million. The $1 billion CDS portfolio generates $10 million of annual income.
Now let’s assume a simple capital structure. A 3% ‘first loss’ tranche, a 10% ‘mezz’ tranche and a 87% ‘super senior’ tranche. Again, this is simplified, but not out of line on a 5 year corporate structure to have the mezz as BBB and the Super Senior as AAA.
Let’s assume that the senior tranche of the cash deal gets T+60 or 5.1%. They get less than the average since there is so much subordination. Assuming a ‘basis’ of 50 bps, the synthetic deal would pay 10 bps on this tranche. Lets assume the mezz tranche earns T+300 or 7.5%. They are receiving a premium to the average because although they have second loss, they do have leverage once losses start hitting that tranche. That mezz tranche for the synthetic would pay 2.5% keeping the 50 bps basis.
So here is what we have:
Cash Deal Synthetic Deal
Number of assets 100 100
Size per name 10 million 10 million
Asset pool size 1 billion 1 billion
Asset yield 6% 1%
Total Income 60 million 10 million
Size 870 million 870 million
Cost 5.1% 0.1%
Cost 44.4 million 0.9 million
Income Remaining 15.6 million 9.1 million
Size 100 million 100 million
Cost 7.5% 2.5%
Cost 7.5 million 2.5 million
Income remaining 8.1 million 6.6 million
Size 30 million 30 million
0 default Return 27% 22%
So in a no default case, the returns look reasonable for the risk. The funded returns are higher, but that does reflect the use of cash.
The key element here, the driver of the free equity, is the income versus cost of the senior tranches. In a cash deal, 86% of the income is used pay the interest due on senior tranches. For a synthetic deal, only 34% of the income is needed to pay the senior tranches. This excess income, and how little of an impact any individual default has on the income stream of a synthetic deal, is what creates free equity.
Now let’s take what might be an extreme case, but is great at illustrating the difference between the cash deals and synthetic deals. Let’s assume 20 defaults occur, all with 0 recovery. This would generate $200 million of losses. It should wipe out the equity, the mezz, and eat into the super senior. Where it gets interesting, is when you look at the residual income.
With $200 million of assets gone, the cash deal would only generate $48 million of income. This would be enough to pay the full interest due on the senior tranche, but not enough to cover the ‘mezz’ tranche and nothing left for the equity. That makes sense to me. It’s on the synthetic side that you get a very interesting dynamic. The residual income would be $8 million. Not only is that enough to continue to paying the interest on the senior and mezz tranches, but it leaves $4.6 million for the ‘first loss’ tranche which is 15% per annum on the notional!
It all comes down to the ‘waterfall’ of how the interest is distributed. On the cash deal, its easy to see that with 20% default, on day 1, with 0 recovery, that the equity would be wiped out, the mezz would be wiped out, and even if there were no waterfall/cashflow restrictions, there would be no distributions to the first loss holder.
On the synthetic side, the story is very different. The equity would be wiped out, but receive $23 million over the 5 year deal in residual interest flows if there were no more defaults. The mezz tranche would be wiped out but receive $7.5 million in interest over 5 years. So the mezz loss would be 92.5/100, or 92.5%. The first loss piece would lose 6.9/30 or only 23%. It seems amazing, but the first loss loses less than the second loss in a synthetic cdo with a straight waterfall.
To make matters worse, the mezz tranches were often structured so that they stopped receiving their interest payments on any portion of their tranche that had been used to cover a default. In the above example, that $2.5 million per annum would have then gone straight to the first loss holder. The second loss would have received 0 in income and had to pay for the $100 they lost. The so-called first loss piece would actually receive 7.1 millon per annum (24%) and over the life receive 35.7, so even after paying away $30 in losses, they would have received a positive return, while the tranche above them, the more senior tranche, the IG rated tranche, would have lost 100% and even the super senior would have had a negative return!
Scary but true. And yes you could argue over the time value of money, and you can argue over capital structure, and you can argue that individual spreads were be different and in all likelihood higher spread names would default first, all of which make the ‘free’ equity less free, but its almost impossible to argue that the equity is as risky as you would think. The risk/reward that exists does not match what you would expect.
If there are no defaults, it’s clear that the first loss portion will outperform the mezz tranche. Its now also clear that under extremely adverse conditions, the first loss outperforms the tranche above it, which is truly bizarre. At least, maybe, there is something in the middle ground, that salvages the mezz tranche?
How about there are 10 defaults on day 1, each with a 70% recovery. This is fairly unrealistic, but should be an example skewed against the first loss tranche. The 10 defaults each losing 30% means a total loss on the portfolio of 3% so the first loss would have to pay out that amount. The mezz tranche would have no losses. The mezz tranche would receive the $2.5 million a year its entitled to. Even with 10 defaults, the residual income to the first loss tranche would be 5.63 million. A total of 28.2 million over 5 years. So a slight net loss of $1.8 million or an annualized loss of just over 1%, compared to a gain of 2.5% for the mezz tranche. Yes, in this case the mezz moderately outperformed, but any additional defaults would primarily impact the mezz so the outperformance would disappear quickly. And this is the most harmful case I can think of where mezz receives all that it is entitled to and the equity receives the bare minimum.
There were deals where the equity was getting large ‘guaranteed’ interest payments, sometimes in excess of 20%, so more than the amount at risk over the life of the trade. Guaranteed coupons this high were a sign of how unrisky the risky tranche was.
One European dealer created a structure, where, by using a reserve account to build up excess cash flow, they got the rating agencies to rate the ‘first loss’ tranche HIHGER than the tranche above it! Boggles the mind that the rating agencies did it, but more proof that the concept is real.
How could this happen?
One of Wall Street’s biggest flaws seems to be complacency when something is working. The CDO structure had worked with bonds and loans and other cash instruments as the underlying. The flows made sense. The first loss did very well in good times, but underperformed the more senior tranches in times of high default rates. There really was no reason to suspect that the dynamics of a synthetic CDO changed all that. On the surface it seemed the same. Heck, it even got a similar ratings profile to the cash deals. The relative value paid to senior investors in synthetics seemed in line with what investors demanded on the cash side. People had Guassian Copula models attributing the spread to various tranches ‘fairly’. Though I remain convinced that the quants just liked the word copula because it reminded them of copulation.
Another reason it happened, is because have you ever tried to explain to risk management that the first loss tranche is safer than the second loss tranche? How the unrated tranche is less risky than the BBB rated tranche? It’s so counterintuitive it’s not an easy argument. The first time someone told me it existed, I shook my heading thinking they were missing something. It just doesn’t seem right, but it is.
This happened primarily for the early deals. As people became more aware of the issue, bells and whistles were added to protect the senior investors (at their request). After defaults, cash flows to the first loss would stop, or at least some portion would stop, to build up a reserve or cushion against future defaults. It helped and seemed fair. Over time the product evolved and in the later credit deals, the risk/rewards had gone back to being more in line with expectations.
As a market developed for single name CDS on the mortgage side, synthetic CDO’s backed by them were also created. For better or worse, mainly for worse, most Wall Street firms seemed to have a mortgage department that was in direct competition with their credit department. In chasing P&L and accolades, the amount of communication between the desks was often minimal. Investors, who probably understood the concept of ‘free’ equity less than the street, also tended to run their investments in credit cdo’s separately from their mortgage backed cdo investments. I believe that this allowed many of the early mortgage deals to create the same sort of free or low risk equity that had existed in the early days. There is nothing wrong with it, it just skews the risk reward and means that you have to be careful when making the claim that the equity holder was taking the real risk.
Have you ever wondered why so many bears were either long the ‘equity’ tranche or seemed willing to be long?
Well, wonder no more. You now know that depending on structure, the first loss, or equity, was actually very well protected. Deals could be structured in such a way that by owning the first loss and shorting more senior tranches, the base case was high teens returns, the best case were massive gains as big losses hit the super senior, and the scenarios that caused a loss were minimal (if not non-existent). These are complex deals. I have simplified the analysis to make a point, but the math works, and it does help explain why there are few stories about how hard it was to place the equity of synthetic cdo’s and why many of the people most bearish the underlying markets were long them and were short higher up the ‘capital’ structure.
So, if we get another round of discussion about the CDO market, like we did surrounding Abacus, at the very least I hope this makes you question what the parties are saying and spend more time figuring who had what risk, based on the documentation and the math, rather than just the name of the investment.