Thursday, October 25, 2007

The Future of CDOs Episode I: Turn her around!

We've spent a lot of time talking about the CDO market, mainly because its one of my favorite subjects and its my damn blog. (Click here for a primer on how CDOs work.)

We stand at an important cross roads for CDOs. Banks, brokerages, and investors of various types have been hit hard by poor performance in the CDO product. I am on record as saying that CDOs were a primary source of helium for the housing market bubble. And yet I stand before you today to say that the CDO product is ultimately a positive for the economy. If we can figure out a way to eliminate some very deep and yet very solvable problems with how CDOs are structured and marketed, CDOs can be a source of liquidity and a very appropriate vehicle for various investors.

So this will be part one of a indeterminate series of solutions to the problems in the CDO market, and by extension, the ABS market. I don't want to spend a lot of time talking about what's gone wrong in the CDO market, because that's been pretty well covered. Also, I know that some of the solutions proposed below could not be implemented over night. Others would require investors to start demanding certain concessions from CDO arrangers. But hey, the beauty of the blogosphere is that an arrogant jerk like me gets a forum to preach his message. So here it goes.

The beauty of securitization is that risk becomes dispersed, as opposed to dangerously concentrated in a few places. The downside is that it creates a disconnect between the initial decision makers on a risk and the ultimate bearers of that risk. In subprime loans, mortgage originators had no apparent incentive to make good loans as long as they knew they could sell the loan into a pool. The same would hold true for commercial loans. As long as the bank assumes the risk can be passed on to someone else, what's the motivation to make good loans?

There is a relatively simple solution here. Force loan originators to retain more risk. This could be achieved in a variety of ways. It should be noted that MBS originators generally retain a certain amount of risk, called the servicing spread, which is usually on the order of 50bps of coupon. In essence that means that when the loan is sold into a pool, the originator gets the PV of the loan's cash flows less 50bps of interest. In total dollar terms, that usually amounts to something like 2-3% of the loan's overall risk being retained by the servicer.

But here is the problem with the servicing spread model. It still allows for woefully undercapitalized firms to remain involved in the mortgage market. An originator like New Century could just keep pumping out enough loans to make up for whatever losses it was taking in servicing, or else just sell the servicing rights to a bigger bank. Management just assumed they would personally make enough money to live comfortably regardless of what eventually happened to the firm.

What would have worked better is if originators had to take risk on both the top and bottom of the capital structure. First, you increase the servicing spread to at least 100bps on non-GSE pools. Then servicers would be required to write a letter of credit on the senior-most tranche of the pool. A letter of credit (LOC) is in essence a guarantee to pay principal and interest if the pool defaults.

The result of this is that investors in so-called low risk tranches would become more aware of the credit worthiness of the originator. Investors would invariably prefer higher-rated originators over lower-rated ones, because the LOC attached would be more valuable. This would either force the lower-rated originators to improve their capital situation or get out of the business. This would also force originators to put their own necks on the line when it came to the pools they originate. Investors would have more confidence in the whole system, since they know that the originator is standing behind the loans they make. The originator's incentives are aligned with the investor.

The same concept could be applied to the non-residential ABS, where the structure could be exactly the same. In the bank loan world, the loans aren't typically tranched prior to being put into a CLO. But banks could still retain more of the loan as a "servicing fee" which would improve the alignment of risk.

Here is a modest proposal: the ratings agencies simply shouldn't rate CDOs at all.

Look, its better to know what you don't know than to think you know something that you actually don't know. You know? Basically by rating the senior CDO tranches Aaa/AAA, Moody's and S&P were saying that they knew those securities were very low risk. Too many investors took it as a given that Moody's and S&P understood the risks embedded in these structures. Now its clear that basically no one fully understood how quickly subprime lending could all fall apart. Its time for a little humility in modeling. Yes, Monte Carlo simulation is the best way to analyze CDO pools, and the 2007 experience will undoubtedly allow the quants to build better simulations. That's all fine. But let's face it, a model is just a model, it can never incorporate all the complexities of real life markets.

So if the ratings agencies would simply admit this, then investors would go into CDO investing knowing what we don't know. That isn't to say that the ratings agencies would have no role in the CDO market. They could provide independent information on the deal's legal elements and opine on the deal manager's qualifications. This would be quite useful to investors who generally don't understand the obtuse reams of lawyereese populating offering memoranda, and who don't have the time to do site visits would CDO managers.

Furthermore, the ratings agencies could still model CDO deals in their Monte Carlo simulators for a fee. Investors could then run the Monte Carlo themselves, inputting default and recovery rates, default patterns, and correlation as they see fit. Rather than getting one or two perspectives on what the default/recovery/correlation patterns should be, investors could impose their own stresses. Some of this capability is available on sites like ABSNet, but obtaining access to a deal can be difficult and expensive. It should be where investors can have access to deal modeling in an open and relatively inexpensive manner. The days of information on CDOs being closely guarded needs to end.

Eliminating formal ratings from the CDO world would also foster more competition among those offering credit analysis on a deal. Currently it costs between $500,000 and $1 million to get a CDO rated by both Moody's and S&P. If a CDO arranger only needed to hire one of the two, and the ratings agency merely commented on the perfection of the legal structure, and assured investors that the collateral manager is reputable, I'd say the cost of would drop to less than $100,000. The CDO arranger would then need to find multiple firms to model the deal's cash flow into a Monte Carlo simulator. But this may or may not be the big two ratings agencies. If investors were allowed to personally compare various simulations of a deal's performance, the barriers to entry in the ratings business would plummet. Investors could easily pick out a modeler who was too generous to the CDO arranger, because that model would stick out compared with others. Currently Moody's and S&P basically tells investors to "trust them." If the models were more open and the inputs were modifiable, then we wouldn't need to "trust" anyone. Trust creates a giant barrier to entry. Openness tears that barrier down.

Besides, if a deal really needs a rating, then obtaining either a LOC or insurance policy would be more appropriate. If someone like Bank of America wrote a LOC on a CDO tranche, then the tranche would be legitimately be Aa-rated, rather than having a phantom Aaa rating.

So hopefully this is a start. Look forward to your comments.


Anonymous said...

Nicely written.

However as for the investor's concerns in deal modelling, you said,

"Investors could then run the Monte Carlo themselves, inputting default and recovery rates, default patterns, and correlation as they see fit. Rather than getting one or two perspectives on what the default/recovery/correlation patterns should be, investors could impose their own stresses"

Aren't they imposing their own stresses when they give in an investor request? Or what about the reverse-engineering of deals that some investors carry out to make the deal model at their end?

Although I agree to RA point you mentioned, they simply make too much money without having ANY accountability whatsoever, but I wouldn't like to say that the investors are foolish and don't understand the risks involved. They simply turned a blind eye to it!

As they say, why fix it when it ain't broke. Fortunately or unfortunately it indeed has broken down now.

Anonymous said...

Were the rating agencies to stop rating structured credit then the market for the structured products would, of necessity, diminish greatly. Why? Basel II lays out the capital requirements based on the rating agencies ratings. If there is no rating then you are going to take a whopping capital charge and purchasing the unrated note becomes uneconomic.

Basel II is just starting to kick in in Europe this January (though with a floor of 90% of B I for the first year) so the B II requirements will become increasingly important.

My thoughts, anyway.


Accrued Interest said...

I admit that the no-ratings idea is a tough one. Currently what a lot of banks and SIV programs did was buy the senior CDO tranche and buy an insurance policy against it. So even if there hadn't been a rating on the CDO itself, there would have been a rating on the security purchased by the bank.

Notice that if MBIA sells XYZ bank an insurance policy on a CDO tranche, then MBIA is in effect rating the CDO, but MBIA also has actual skin in the game. That creates more accountability.

Anonymous said...


Good article. I want to address a couple points you made as well as make my own suggestions.

Accountability: I don’t think this is realistic because 1) it defeats the whole purpose of securitization and 2) it will make residual piece unmarketable. First and foremost, you’re assuming the originator is also acting as the servicer: for non-agency RMBS, this is not always true. Americans enjoy high level of home ownership among industrialized nations because securitization allows banks and originators to syndicate risks out to investors, allowing them to recapitalize and consequently make more loans to sponsor growth in home ownership. Your LOC idea would have to be reflected on the balance sheet of originators some how, thus putting a cap on origination capacity and securitization. In fact, if the LOC had to be reflected on the balance sheet somehow for securitization, there would be no point to securitizing because you’re effectively retaining the risk of the deal (~-70-80% of total collateral since AAA typically have 20-30% credit enhancement these days). If you were to charge 100bps for servicing, deal will not be economical for the residual piece – making it difficult, if not impossible to place. Your LOC idea may make more sense if the originator and servicer are one and the same: this, however, still puts a cap on origination capacity, thus defeating the point to the securitization process.

Solution: Ultimately, accountability rests with the investors. It is there responsibility to do good credit work and thorough due diligence on servicers and originators. Investors cannot simply rely on rating agencies to evaluate servicer and originators, as corporate investors do not solely rely on sell-side research. Investors need to get off their asses, do some work, instead of blaming their own laziness and incompetence on originators, servicers, underwriters and rating agencies. And the end of the day, if I am an originator and I originate crap and can’t sell it – I obviously won’t be in business very long. Doing due diligence and staying away from bad originators and servicers will ultimately impose a sort of Darwanism on originators and servicvers and thus eliminate most of the poor performing loans we currently have outstanding today.

Ratings: I believe most people would agree there was probably some conflict of interest between rating and structuring of ABS/CDO. However, it should be noted that rating agency are receiving the brunt of the blame from so-called “sophisticated” investors, who were not very sophisticated, did no research and basically relied on the rating agency to do all the credit work. In corporate land, everyone knows the credit rating is a LAGGING indicator of the risk of a bond. That’s why most investment house have corporate credit analyst to stay ahead of the curve in evaluating risk in their corporate bond holdings. I digress.

Rating methodologies are flawed, but its only because CMO/CDO technologies are so complicated. There are so many moving parts in a CMO/CDO and so many factors driving ultimate cashflows on a deal. In a way, CMO/CDO are like a microcosm of an economy and in some ways, rating agency are being asked to project how these miniature economies will perform for the next 5-7. There lies the problem and possibly one with no clear solutions: models and humans cannot predict the future, especially when they are few years out.

Solution: Most people already acknowledged that the rating agency underestimated losses because they were extrapolating on limited and, at times, unrelated data. Rating agencies are an INTEGRAL part of ABS/CDO and should continue to be an INTEGRAL part to this business. In some ways, they act as private regulators in this industry and have done a fair job to bring and continue to bring consistency and transparency into a very opaque sector. Moody’s, for instance, is making headway in trying to get originators to 1) provide monthly loan tape so that investors can better model their deal’s performance 2) require third party oversight of accuracy of those monthly loan level data and 3) stronger and more uniform reps and warrants for originated loans. Rating agency have already incorporated changes to how modified loans will affect performance triggers, which greatly affect how principal cash flows will be diverted.

Overall, rating agencies have made fair amount of changes to the way they rate ABS/CDO. One of the biggest changes they have brought is demanding higher levels of credit enhancement on current and future transactions. This makes originators and underwriters ACCOUNTABLE because it reduces leverage on the transaction and makes marketing of a larger residual piece more difficult (no one will be buying equity in a 15-20x levered transaction if originator or deal shelf has spotty records)

Anonymous said...

M-LEC banks stand to make billions Renée Schultes
29 Oct 2007

The banks participating in the Master Liquidity Enhancement Conduit, the super-fund planned by four US banks to ease liquidity concerns in the market, could make more than $1.3bn (€900m) in management fees and even more in other charges, according to people who have seen an early prospectus.

Bank of America, Citigroup, JP Morgan and Wachovia are backing M-LEC, which is expected to raise assets of $75bn to $100bn and which is supported by the US Treasury. Three people who have seen a prospectus dated October 8 – a week before the fund became public – said the banks will earn 1% on structured investment vehicles of less than $5bn, and 1.5% for SIVs over $15bn.

The banks will also earn fees for the provision of backstop liquidity facilities. Structured investment vehicles are off-balance sheet funds that seek to profit from the difference in the cost of short-term borrowings and higher returns on structured credit.

Christian Stracke, an analyst at CreditSights, said: “Even at 1% it’s still very high. These are vehicles that have little wriggle room in what they can afford to give up, especially SIVs that have taken a beating in their asset values.”

The prospectus also details what SIVs will receive for selling their assets to M-LEC. Qualifying SIV holders will be eligible for up to 94% of the value of the assets they sell in cash, or 89% cash and 5% in senior capital notes, in the form of medium- term notes, that will participate in part of the upside when the assets mature.

SIVs can also buy junior capital notes based on a formula set by M-LEC, where SIVs would get 3% plus half the discount at which they sell their assets. For example, if the SIV sells its assets at 98, which is a discount of 2%, it could take 94% of that in cash and 4% in junior capital notes.

The banks, which met again last Friday evening, may have refined the initial terms in the past three weeks.

The banks declined to comment.

Anonymous said...

Amazed that you believe in "dispersion of risk". There is no such thing in reality. Why can't you grasp this and why do act like any of this is new? It's all been tried before and has broken.

Anonymous said...


You have written today another excellent piece.

I also have some attachment to CDO’s. I like to believe that I saw their birth. I guess it was the summer of 1988 when William E. Simon’s Western Federal Savings & Loan issued its first $100 million 7-year bond backed by high yielding bonds. I was involved in a similar issue a few days after that, and am still proud that our asset portfolio went unscathed despite the problems with Fruit of the Loom and the Trump Casino in Atlantic City.

Its legal name was a “CDO” but it was a very different animal. These days there’s a great effort in placing the more senior “tranches”. In those days there was only debt and equity, and there was a greater effort in placing the (large) equity layer. I remember numbers like LIBOR + 14%, using a 7x leverage using only bank finance. I remember calling it a “synthetic bank”.

Forbid my boring you with this memorabilia. It’s the sign of the age before I get back to your clear and excellent piece.

I am also convinced that the structuring of a CDO is an ingenious and efficient mechanism to gather and parcel risk. And agree with the potential disconnect between the risk source and the risk taker. But I feel that I also have to agree with the previous comments.

My understanding is that the source of so many problems with CDO’s today is that they showed up in money market funds, or in bond funds carrying some perceived capital guarantee.

For this reason, I would wish that all instruments carrying some actual or perceived capital guarantee should not be allowed to invest in any non-primitive security. And if they are allowed to invest in a non-primitive security they should carry a “NO SMOKING” label with similar relative appearance in the prospectus.

Let’s be clear: CDO’s like any derivative security will always be described as opaque in terms of its risk. So, the mathematical complexity of derivative securities can only be the source of social progress as long as it is handled by legal entities described as “professionals”. By the same token the so-called “toxicity” of CDO’s does not arise from its risk. It comes from the potential contamination of what I would describe as “popular” savings.

I guess that what I am trying to say is that what one describes today as non-transparent has no reason to become more transparent or more regulated. Its degree of transparency should be what is required by its buyers. But similarly it should not be allowed to contaminate the domain of the markets where transparency is the rule.

You will probably say that I envision a world where the higher returns can only be accessed by the private/rich banking clients. I guess you may be right. As I think of it, this is as far as I can get now.

Thanks. F (Lisbon, Portugal)

Accrued Interest said...


Risk dispersion does work. There's a difference between me saying that risk is dispersed vs. someone claiming that finanancial contagion is dead because of risk dispersion.

The former point merely says that distributing financial risk allows for greater stability of the entire system. A repeat of the Great Depression is lesser because of financial dispersion. On the other hand, the risk that bad actors in a single corner of the financial world will cause many investors to lose money may indeed be greater.

Anyway, in the future, please state why you disagree with me. Its much more interesting.

Accrued Interest said...

On the subject of ratings...

It might be unrealistic to expect the CDO/ABS market to exist without ratings. But as Sigma pointed out, it would be possible to have the ratings agencies share more of their data with us. I think the spirit of Sigma's comments is similar to my idea of allowing investors to use their Monte Carlo simulators.

Regardless, I still think it'd be a good idea. Consider F's comment about the very early days of CDOs. Like he said, they involved only a couple tranches, and the rating was all about the subordination, not about trigger tests and the like. What if the top tranche was supported by an insurance policy and the subordinate tranches. Its not impossible.

Accrued Interest said...

Also to Sigma: Your point on servicers and originators being two different people is a good one.

I think its safe to say that going forward, investors will pay more attention to originators. I mean, 5 years from now when the subprime loan market is viable again, I still think people will be looking hard at who the originator is.

But as Anon points out, the next financial meltdown will come from someplace else. And people will act like its all new, when its really all the same. We go from fear to greed back to fear then back to greed. I'm just trying to think of some more creative solutions that can persist through cycles.

If the ratings agencies agree to modify their CDO criterea, but don't consider their fundamental flaw (lack of model humility) then we really haven't learned anything have we?

Accrued Interest said...

BTW there is some good commentary from Hymas here:

Anonymous said...

You left out the biggest thing needed: standardization. Cookie-cutter templates for a few standard-model CDOs, with strict rules about the debt/securities/etc that are used for construction, so once built, investors no longer have to worry about particular contents, ratings, etc.; only about the behavior of that class of CDO.

Makes modeling a lot more useful, as a meta-model of each of the classes would soon be available; these models would be more reliable to boot, as they would be without the noise caused by the esoteric particulars of their hand-crafted forbears.