Wednesday, March 21, 2007

CDOs By The Numbers
...and Where Does The Yen "Carry" Fit In?

I have been discussing Collateralized Debt Obligations (CDOs) and associated credit derivatives recently (March 15 post: "Well, what do you want it to be?"). Today I present some numbers to go with the story. Data are from SIFMA, the Securities Industry and Financial Markets Association. The link page also provides a concise explanation of the collateral, structure, purpose and other features of CDOs.

First, global CDO issuance has expanded at a furious pace - up six times in just three years. The chart below tracks the new issues of CDOs every quarter since 2004.

Of the total $489 billion issued in 2006, almost 80% were issued in US dollars, 18% in euro and the remaining 2% in yen, sterling and other currencies. Given the size of US debt markets, such a concentration is expected.

The next chart breaks out the CDOs issued in 2006 by collateral type. Structured finance collateral (60% of all issues) includes residential and commercial mortgages, asset-backed securities, credit default swaps, other CDOs (resulting in CDO-squared or -cubed) and other securitized and structured products. Another 34% is backed by high yield loans made to borrowers rated Baa3/BBB- or lower. Investment grade bonds make up a minor 5% and all other account for the rest (rounding error).


It is apparent that CDOs are used mostly to pool and securitize low-rated obligations. Using the cascade structure, CDOs can transform a very high percentage of risky debt into securities rated AAA to A because they currently use optimistic (in my opinion) default and recovery assumptions in structuring the tranches. The vast majority of CDOs issued during 2006 were designed to capture exactly that difference between the high yield of the underlying collateral versus the lower yield of the CDO tranches. According to SIFMA, such "arbitrage" issues amounted to 86% of all CDOs issued in 2006 (the rest were "balance sheet" transactions done to remove assets from balance sheets, used to maintain regulatory capital ratios).

Another way to categorize CDOs is by issuance type. Cash flow CDOs are designed to pay investors by passing through to them interest and principal payment cash flows from the underlying debt collateral. Synthetic CDOs sell credit default swaps (CDS) to create a replicate cash flow. By doing so they also assume the liability of paying out on the CDSs in case of defaults, so buyers have to deposit funds in a special purpose vehicle entity (SPV) as security, or margin. Naturally, they do not deposit the entire or even a large portion of the potential liability - that would defeat the whole purpose of the CDS and the related CDOs - but only a very small portion. How small depends on default and recovery assumptions and, once again, I believe these assumptions are too optimistic. Hybrid CDOs combine those two structures: they collect and pass through both "real" cash flows from debt collateral and "replicated" cash flows from a synthetic structure. Finally, market value CDOs are designed to provide payments from both cash flow plus selling off part of the collateral, so the market value of the collateral greatly impacts the payout and performance of the CDO. The break down by type is shown below.

Unfortunately, SIFMA does not break down the dominant category of "cash flow plus hybrid" into its two components, so we do not have a clearer picture of the systemic risks involved in the CDO market. It is quite obvious that synthetic and hybrid issues contain significantly higher risk, based on the leverage imbedded in their structure (CDS and margin).

Finally, a reader asked how the yen carry is involved in the CDO market - apart from the simple strategy of borrowing cheap yen to buy higher-yielding CDOs on margin, of course.

The more "creative" use of yen loans comes in the initial funding of synthetic and hybrid CDO's. As we saw above, cash must be deposited into an SPV to collateralize the potential CDS liabilities in case of defaults. A speculator may simply provide hard cash for this purpose, but in this age of leverage and maximum return competition, it is much more likely that he simply gives his prime bank a security from his portfolio as collateral to obtain margin money through as cheap a source as possible. Enter the yen (or the swiss franc)...

I close by stressing, yet again, that all this talk of "ample liquidity" can be very distracting and even deceiving. Despite what many believe, liquidity is not some absolute, measurable quantity of cash sitting somewhere as a stagnant pool, just waiting for the nimble speculator to suck it up and put it to use.

Liquidity is the accessibility to and the pricing of credit, i.e. lending standards and interest rates. The former are cyclical, running the gamut from "fog the mirror" to "indenture your children as collateral". Until very recently we were as close to the former as we have ever been (CDO issuance is certainly a clue), but now the pendulum is clearly starting to swing in the other direction. As for interest rates...what can I say, but the obvious: they are going up everywhere, including Japan and China.

Like the rest of finance these days, liquidity is best described as a derivative: the mathematical first derivative of debt. Which begs the question, of course, first derivative against what? Time, GDP, income, gold, crude oil...? That's a question to ponder for another posting..

7 comments:

  1. Hellasious,

    Thanks again for a great explanation of a difficult subject.

    It boggles the mind that CDS investors are willing to except only a small deposit as collateral against the possible default by the "insurer" in the event of a default by issuers of the underlying debt instruments. If my understanding is correct there is little if any check on the creditworthiness of the "insurer", i.e. seller of the CDS, in the event there is widespread default by the underlying issuers.

    I guess the possibility of such a default on the underlying debt instruments has already been determined to be mathematically impossible.


    Unfortunately, it appears that the buyers and sellers of CDS instruements are masquerading around pretending as if these CDS are rock solid insurance policies while a quick look will tell you that the bond ratings of many CDS sellers(i.e. investment banks)are just above junk status.


    No wonder CDS premiums are so low because the chances of the seller actually being able to make good on them in the case of a debt crisis are so slim.

    ReplyDelete
  2. Dear anonymous,

    It's the old question of who insures the insurers..but at least regular insurers are by law required to maintain adequate loss reserves, based on actuarial data going back centuries - insurance is very old business.

    The CDS business is so new (notional outstanding $0.9 trillion in 2001, today around $30+ trillion) that it has never been stress tested in through even one complete credit cycle.

    It gets worse: there is no limit on how much CDS is written against just one "name": for example, ABC Corp. may have $10 billion outstanding in bonds. It is perfectly legal to issue $20 billion in CDS against the $10 billion. Of course in a bona-fide, complete default situation the maximum loss for all CDS combined will be limited to $10 billion, because the CDS owner has to provide the physical bonds in order to be paid by the CDS seller.

    Sounds almost like a good thing, a sort of "over-insurance" but it isn't really, because if ABC does not default but just scrapes by, there is no payment required, but the value of its CDS's will certainly decline (risk premium up). In the above example 2 times more CDS's will be affected (20/10), in turn affecting the value of synthetic and hybrid CDOs "manufactured using those CDS's). Vertical risk infection, instead of risk mitigation as some credit derivative boosters proclaim.

    It can get even worse...CDS's are now being written on indexes of CDS's and there you have cash settlement, i.e. no limit on the potential loss since the owner of the index CDS does not have to produce any physical bonds...thus a single default can theoretically create unlimited losses.

    The mind boggles, but it all comes down to this: the next down credit cycle is going to be extremely vicious and have domino effects nobody can now imagine.

    Warren Buffet calls the stuff Weapons of Mass Financial Destruction and I believe him 100%.

    I like your allusion in the end...CDS insurance is cheap because it is worthless, hahahaha good one!

    (I doubt that Goldie, Morgan, et al will be amused tho)

    Regards

    ReplyDelete
  3. Again so plain, so clear. And the more I understand, the more discouraged I get. Storing wealth is so unbelievably tricky.

    ReplyDelete
  4. hellasious,

    Any clue as to the actual value of the underlying bonds for the $30 trillion in CDSes are?

    ReplyDelete
  5. hellasious,

    If it is correct that the notional value of CDSes insuring a particular set of bonds can actually exceed the total value of the underlying bonds themselves, is it possible then that the total losses experienced by sellers of CDSes would then exceed the actual default losses experienced by the underlying bond holders themselves in the event of a total default by the bond issuer?

    Perhaps, I am misunderstanding something but lets says a group of sellers have sold $20 billion of CDS on $10 billion of underlying bonds (CDS total potential liability of $20 billion), what is to prevent that group of sellers if a default situation is arising from buying up the $10 billion of bonds or intervening to make the debt payments in order to prevent the default situation? Better to pony up the $10 billion than possibly lose $20 billion in a total default situation.

    I remember reading something recently about settlement problems because a group of CDS buyers could not get a hold of enough bonds of a bankrupt issuer.

    Maybe there is something I am missing because this can’t possibly be the situation, right?

    ReplyDelete
  6. hellasious,

    the links below discuss the very situation I am alluding to with physical settlement problems e.g. the Delphi bankruptcy where $20 billion of CDS were issued on $2 billion worth of bonds

    http://www.tavakolistructuredfinance.com/ft20.html
    http://www.creditflux.com/resources/problems+with+physical+settlement.htm

    If such "technical squeeze" problems arose in a situation involving default by just one issuer, what type of squeeze issues will arise in the case of widespread CDO defaults by a number of different issuers?

    Is physical settlement required for a CDS covering CDOs that packages subprime mortgages?

    If so, in the event of widespread default by subprime borrowers, will that create a technical squeeze for the underlying subprime mortgages?

    I am really confused. I can't figure out whether CDS premiums are low because the CDS sellers are junk rated or the CDS sellers know they will never have to make good on a default because of this technical squeeze dynamic.

    ReplyDelete
  7. In a default situation, the owner of the CDS must present the actual "physical" bonds to the CDS issuer in order to be paid on his "insurance policy". Therefore the total liability can never exceed the amount of debt outstanding, no matter how many times more CDS are written against the debt - even if it is 10x or 20x.

    This raises some very interesting possibilities of what will actually happen in case defaults start rising out there. I'm thinking through some scenaria and will be posting on them if I come up with anything interesting.

    This is a brand new market and has not been tested in a real down cycle with more that minimum defaults. There are too many unknown variables that could come into play.

    In the case of mortgage CDO's, the mortgages are owned and held by the SPV that created the CDO in the first place, so it's the same as with any other corporate bond: physical delivery must be made.

    Lastly, I haven't seen hard data on the amount of bonds underlying the total issued CDS, but I have seen this: The major interdealer brokers for this market report that they commonly see only about 600 "names" being used to write CDS against their debt. There is no question in my mind that the notional amount of CDS's exceeds the amount of debt they insure. I just don't know by how much.

    As if all this weren't enough, there is a rising issuance of CDS's against indexes, i.e. multi-party CDS that DO NOT require physical settlement. This is a real curve-ball and can totally alter what happens in default situations.

    I'll discuss more in an upcoming post.

    Thank to all for your interest.

    ReplyDelete