Sunday, January 30, 2011

The Ratings Charade Continues: A CLO Investigation, Part I

The role of the ratings agencies in the financial crisis went largely unexamined by the powers that be. That's a crying shame. Because the game hasn't changed: Investment banks fork over big fees for ratings agencies to sign off on phony ratings for complicated products.

Today I'm going to prove it, step by step. I'm not going to show all my work (I don't want this expanding to the length of a Scribd academic paper), but I can separately (in the comments section or in a separate post) for anyone who's interested.

We start with one of Wall Street's darlings of complexity, called a collateralized loan obligation.

If you're going to hang with me here, you have to grasp the basics of how one works. It's like this: An investment bank bundles together say 30 leveraged loans (this is the risky debt that companies take on in leveraged buyouts). Now, recognizing that different investors have different risk appetites, the bank creates "tranches" of securities that receive payments in a "waterfall" structure, which is the complicated heart of the CLO.

Okay, that sounds confusing. But there's a simple way to look at it. Each of these 30 leveraged loans makes periodic payments (of interest, or interest and principal). Once you strap all the loans together, individually they still make the same payments on the same schedule. But how the money is distributed becomes a bit more complex.

That "pot of yield" generated by the 30 loans is divided as follows. The investors in the top, or safest tranche, get paid first. This tranche is generally ranked AAA. Investors in the next tranche down, which we'll say is AA rated, get paid after that. Then the A rated tranche holders receive their money, or "water." And so it goes, right down to the bottom layer of this structure, sometimes called equity (though it's not technically equity, for you finance nerds -- and there's often also something called an "overcollateralization" feature in a CLO, but we don't need to get into that here.)

When times are good, with all the leveraged loans paying on schedule, the waterfall is bountiful and everyone gets "wet" (i.e. paid). When some of the loans default, and the gushing waterfall of yield slows to something more akin to a trickle, there won't be enough money to go around. But you always start paying off investors at the top (AAA), then move down the structure. If the loans start to sour, the AAA guys are supposed to have an ample cushion before they feel any pain.

So, in a nutshell, an investment bank has taken 30 leveraged loans, tied to 30 companies that have 30 different stories, and roped them all together into a securitization that pours forth a stream of money that satisfies investors in the manner described above. If you're Standard & Poor's, it's a walk in the park to rate any one of those 30 loans compared with rating the slices of this Rube Goldberg-ian CLO. Which is probably why banks helpfully "suggest" the ratings to the ratings agencies and provide models to demonstrate their reasoning behind those "suggestions."

Now let's say you're Mr. Ratings Guy at Standard & Poors, in charge of signing off on CLO ratings. Your bull***t detector ought to be pinging pretty hard when something with these proposed ratings lands smack dab in the middle of your desk (I've condensed this from a Jan. 13 Bloomberg story):
Citigroup Inc. has revised the proposed interest rates on a collateralized loan obligation to be managed by WCAS Fraser Sullivan Investment Management LLC, according to people familiar with the terms.
A $15 million piece rated BBB by Standard & Poor's will pay lenders 400 basis points (note: there are 100 basis points in one percentage point) to 450 basis points more than the London interbank offered rate, while a $19 million slice, graded BB, will pay lenders 600 basis points more than the benchmark...
A $273 million piece rated AAA will pay lenders 160 basis points to 170 basis points more than Libor and a $13.5 million portion graded AA will pay lenders 250 basis points more than the benchmark, the people said. There is also a $31.1 million piece with an A rating and a $51.075 million slice of subordinated notes, the people said.
Why? Remember how our CLO was constructed: out of 30 leveraged loans. These loans pay a certain floating interest rate over Libor (the London interbank offered rate, or what banks charge when they lend to each other). And that's it. You can't wring out any more yield. So the size of our "waterfall" is constrained by what those underlying loans pay. Let's say it's 10% overall right now (not a bad assumption: a CCC and lower bond index right now is at 9.97%).

Now structuring isn't free. Citigroup isn't creating this CLO out of altruism. Here are some categories of CLO expenses: 1. The cost to structure the CLO and earn a profit. 2. The yearly costs to manage the CLO (for example, there's a reinvestment period, during which the manager must replace loans in the portfolio that pay down) 3. All other expenses, including paying the ratings agency.

Let's fill in some blanks. Let's say management fees average 51 basis points, or about half of 1% (source: 2009 report from PF2 Securities Evaluations). Let's say structuring fees run about 1.75 percent (this is according to a Bloomberg story). And, finally, let's say the life of the CLO will be six to eight years. Even though the management fee must be paid yearly, the structuring is a one-time expense, and can be averaged over the life of the securitization.

Do a little math and you get about 76 basis points as the yearly cost that has to be extracted from that 10% pot of yield you're getting every year. Now the size of that pot has been whittled down to 9.24% effectively.

So, as Mr. Ratings Guy at S&P, you should be getting a little suspicious at this point, even before you look at the proposed ratings: Citigroup claims to be able to strap these loans together and, through some bit of diversification/alchemy, just sort of poof! -- extract 76 basis points a year. If these loans, after being structured, were somehow "de-structured" but with all the fees still intact, you'd be left with the original 30 loans, but paying 76 basis points less apiece, which is a pretty significant gap in bond land.

That should make you go "hmmm." But once you look at the actual numbers for the proposed ratings, your reaction should be something like, "no way."

Just look at the generous yields on the tranches of the CLO! The AAA slice is 160 to 170 basis points over Libor. That's a super-juiced AAA yield. A AAA corporate bond -- once you make a few tweaks (for the fixed-to-floating swap rate, the difference in Libor vs. Treasury -- I won't show my work now but can later for anyone interested) has a comparable yield of about 54 basis points. How can this be, at a time when the credit markets are relatively calm, when even junk debt is selling like hotcakes? This isn't a period of high market stress and irrationality.

(Brief aside: Some readers may object: "Well, a AAA bond doesn't imply the same risk profile as a AAA slice from a securitization." If you think that, you may want to look at S&P's own writing on the issue from January 2010: "In developing our updated corporate CDO criteria (note: a CLO is a type of CDO), we collaborated with Standard & Poor's corporate and government ratings group to promote comparability of CDO ratings with ratings in corporate, municipal and sovereign, as well as other areas of structured finance. When we assign the same rating level to debt instruments in varying sectors, we are expressing the opinion that they have comparable credit risk.")

Back to our unfolding narrative! This is what Citigroup is essentially saying to you, Mr. Ratings Guy: "Hey, ya know, we just structured it and all, and found this great big pot of yield left over! Son of a bitch, funny huh? I mean, there was so much yield we shook out of this thing, thanks to our genius in structuring, that we could pay the structuring fees, pay the annual manager fees, pay all other fees, PLUS hand out extra yield like candy canes right up and down the waterfall structure!"

Because here's what you have: AAA is getting 111 basis points (1.11 percentage points) more than comparable corporate bonds. AA is raking in an extra 149.5 basis points, BBB an extra 229.5, BB an extra 218.5 ... (the spread for the A rated isn't given, but it's got to be consistent with the others because this grade lies between AA and BBB, so I extrapolated that one.)

That gives us another 116 basis points of yield a year, over the size of the entire CLO, that the structuring genies claim to have conjured from somewhere, for a total of 1.92%.

[Update: Reader “Anonymous” makes a good point below about the need to adjust the numbers to account for a call option premium. A fuller explanation of what that means appears at the end of this post. So the structuring genies are actually conjuring up closer to 152 to 180 basis points of yield out of thin air -- not 192 -- but that’s still a whole heck of a lot.)]

Think about that. These individual loans pay 10% overall. That was presumably their fair value. Somehow Citigroup is claiming, through the miracle of structuring, that it has been able to shave almost 2 percentage points -- one whole fifth -- of that risk away.

This structure makes no sense, right on the ground floor. You can't extract a bunch of fees, pay a bunch of rich yields, and have the math work out, considering there's finite money being paid out by the underlying loans. Structuring, in and of itself, can't produce such enormous savings. If it could, everything in the corporate debt universe would be immediately structured for huge and immediate gains.

Now, Mr. Ratings Guy, you should be saying, "Something smells really fishy here." And if you were thinking like this, you would reach an obvious conclusion:

These ratings have to be bull***t. The AAA tranche of this CLO, for example, deserves a grade closer to junk than to AAA.

Yet Mr. Ratings Guy still signs off on the ratings. Why? Hmmmm...

Stay tuned for Part II in which we answer: Why does Mr. Ratings Guy sign off on ratings he knows can't be correct and why does this farce exist at all? Is what's going on a benign "nobody gets hurt" kind of transgression? And who gets burned if these ratings blow up down the line?

Update: One objection that's been raised: S&P doesn't actually see the pricing when these ratings are proposed. It doesn't see that the AAA tranche, for example, would pay 160 basis points over Libor. Okay, that's somewhat exonerating for S&P, but it doesn't change the math. And what's more, Mr. Ratings Guy isn't that stupid. He can figure out what's going on.

He can easily find out what AAA rated debt pays for other asset classes. Even if he doesn't have the CLO pricing in front of him, he'll discover the same problem I outlined above. This structure supports a tremendous amount of what should be low-yield AAA, and even after you add in yields for the other stuff, there's an awful lot of leftover yield to go around (some of which is used to pay structuring and management fees). All of which leads back to the same questions: How does that act of structuring manage to create so much extra yield? How can these CLO ratings be accurate?

Update, Part II: I wanted to sneak in a second update for those readers who will say -- rightly -- wait a moment, don't loans amortize? So aren't you really receiving the interest rate on the loan plus a certain chunk of principal each year? That's typically correct, and I reference that up high in the post. But just to make clear: I am keeping this example simple with a focus on the interest rate portion only, because the yield is the sexy part. That's what you make over and above your initial investment -- the return of principal is just making you whole. If anyone has further questions/comments, put 'em below and I'll tackle them. The bottom line is the math doesn't really change.

Update, Part III: Explanation of the accounting for the call option: the equity investors (the ones who hold the junkiest tranche) have a call option on the CLO fund, which typically can be exercised after 3 to 5 years. The existence of that call option is undesirable for the other investors, so they’ll demand a premium to compensate for it. So in other words, if investors wanted to be paid 30 basis points above Treasuries for a AAA CLO tranche that can’t be redeemed early, once you add a call option, they’ll want even more.

How much more? That’s the key question. See more details in my reply in the comment section, but basically I give the example of a Wells Fargo note that’s effectively 6-year debt with a call option in two years, where the option appears to be worth 18.5 basis points. In a longer-dated note, the option is worth more: a Bank of America note that matures in 13 years has a call option that kicks in 3 years from now that’s worth 39.9 basis points.

Of course those examples aren’t CLO tranches. Still, for CLOs, the call seems even less valuable. Babson Capital Management looked at spreads for CLO bonds and found in the first quarter of 2007, they were 22 - 26 basis points for the triple A tranche. So even if you assume the investor is assigning negligible credit risk (say 10 basis points, which is paltry) to the asset itself, that leaves only 12 to 16 bps for the call option.

So, relating this to the example above, you can subtract somewhere from 12 to 40 basis points from the 192 estimate to account for the call option. Still, you get a good 152 to 180 bps of mis-rating -- which is quite a lot.


  1. So tell me, Finance Guy - how much credit losses have been substained to date on the senior-most, originally 'AAA' rated CLO bonds? Closer to zero than the train wreck that you are preaching, no?

  2. Yes, I've heard that argument many times before. But it's not my point. Is your tinderbox house flame retardant because the fire trucks extinguish a blaze before it hits your front porch? Bernanke's unprecedented central bank activism saved a lot of assets that were probably heading for the crapper.

    My larger point is that the ratings have to be wrong. You can't extract as many fees as the securitization machine does and have enough yield left over to properly compensate investors for the risks they are taking (assuming you believe in a reasonably efficient market). Leveraged loans in particular will be heavily exposed and highly cross-correlated in a sharp market downturn, giving them a decidedly non-AAA vulnerability.

    But look: the proof is in the pudding. If the AAA CLOs are really AAA, they should trade close to Treasuries, reflecting their near-risklessness. But they don't. (See above.) And for the heck of it, I checked out a recent CLO; the AAA slice was being priced at 150 basis points over Libor, which is roughly comparable (after making a Libor-Treasury adjustment) to an A or A+ corporate credit, some five levels below AAA.

    The market clearly doesn't see this stuff as triple A. Rather it's "triple A" (wink, wink).

    1. 1: AAA spreads are much tighter than 150, currently at 110-115.

      2: AAA Government Notes (bonds) generally 10 year pieces of paper, that do not have call options. AAA CLO paper is being priced with a call option, b/c the AAA investor assumes their paper (debt) will be called by the deal earlier (through a refinance or another measure). This call option adds X basis points to the deal, which is why AAA debt trades rather high currently.

    2. Never bothered to check the date this was published, but point 2 stands.

    3. Your first point: If AAA spreads for CLOs are currently 110-115, that would argue for less mis-rating now than there was when I originally wrote this post. Still, regular AAA debt is closer to 30, so the gap is still significant.

      Your second point: This is a good one, and I will revise the post above to reflect the need to adjust for the call option. But the big question is: What is that option worth?

      Measuring the value of embedded call options isn’t my specialty. But it appears that such an option isn’t all that valuable. Take this Wells Fargo note: CUSIP 94986RJD4. At the moment, it’s effectively 6-year debt with a simple call option in 2 years. The option-adjusted spread suggests the option is worth 18.5 basis points. Or, if you prefer a longer-dated note: CUSIP EI7197557. The issuer is Bank of America, and it matures in 13 years, with a call in 3 years. The option-adjusted spread suggests that call is worth 39.9 basis points.

      Of course those examples aren’t CLO tranches, but regular debt. For CLOs though, the call seems to be worth even less. Babson Capital Management studied credit spreads of CLO bonds and found in the first quarter of 2007, they were 22 - 26 bps for the triple A tranche. So even if you assume the investor is assigning negligible credit risk (say 10 basis points, which is paltry) to the asset itself, that leaves only 12 to 16 bps for the call option. (Caveats: I’m not 100% certain Babson didn’t already remove the option value from its spread numbers, though it appears unlikely. Second caveat: the value of a call will vary over time, of course, depending on volatility of interest rates, along with expectations of whether they’ll fall or rise.)

      So, relating this to the example above -- if you subtract somewhere from 12 to 40 basis points for the call option, you get a good 152 to 180 bps of mis-rating -- which is still a lot. But look, even if you ignore the “over-rating” argument I’m making above, the fact that structuring fees and other costs are shaving 7% off the revenue stream that the loan is kicking out is going to automatically suggest that there will have to be some mis-rating (through clever use of models, and misapplied theories of diversification, etc.) to make this investment palatable to buyers.

  3. It's been a long time since this article was released but I just wanted to make something clear. CLOs are not my specialties but I think you should consider the illiquidity risk which is clearly different than the credit risk. We all know CLO's assets are far less liquid than any leveraged loan traded in the market...