Believe those who are seeking the truth. Doubt those who find it. Andre Gide

Friday, March 20, 2009

CDO Squared, Anyone?

Along with Martin Hellwig, I think that the other bright light in the field of finance is Gary Gorton of Yale. He has published several very interesting papers on historical banking panic episodes; see here. He gives a detailed account of the subprime mortgage market and the financial innovations associated with it in his paper entitled "The Panic of 2007."

In this paper, he describes the nuances of subprime mortgages; in particular, how their particular design made them very sensitive to the underlying asset price (unlike conventional mortgages). Evidently, this was by design (there was no other way for creditors to make money servicing this particular demographic).

He goes on to describe how these subprime mortgages were packaged into mortgage backed securities (MBS). This is a common form of securitization (although, the design of these also differed in a subtle, but important manner, from standard securitizations). As with other securitizations, mortgages were pooled and then tranches were formed; e.g., a senior tranche, a mezzanine tranche, and an equity tranche.

This type of securitization has an economic rationale. Higher rated tranches can be sold to insurance companies and pension funds (whose liabilities are longer term in nature). In principle, the originator of the MBS should could then hold on to the junior (equity) tranche. This gives the originator the incentive to construct a sound MBS; as the originator is the first in line to potential losses.

What I do not understand is what followed. These MBS were then used as backing for new securities, called Collateralized Debt Obligations (CDOs). For example, the mezzanine tranche of the MBS (rated BBB) would then be divided into senior, mezzanine, and junior tranches. The mezzanine tranche of this CDO would then be divided again into senior, mezzanine, and junior tranches (a CDO squared). The senior tranches of the CDO squareds would be assigned AAA ratings!

I do not understand the economic rationale for this further subdivision of the original MBS. I presume that there must be one (perhaps to get around some government regulations?). Is there an expert in finance out there that can help me out? I have had little luck in finding anything that explains the motivation for why CDOs exist.


  1. I think your real question is why would anyone buy these CDO^2 instruments. Why they were created is obviously to make money from selling them which many people (initially) did.

  2. Well the question should also be how did these tranches get AAA in order to attract investors?

  3. Alexander: I am disappointed in you. This is like me answering your question by saying that "people bought them because they obviously thought they were a bargain."

    Pani Pani: Yes, good question. I have a not entirely satisfactory answer to the question. Assume, for example, that in the worse case scenario, a BBB tranche can lose at most 50% of its value. Assume that the senior tranche of the CDO backed by this BBB asset has first claim to 50% of its value. Then this tranche of the CDO is risk-free (i.e., it should be rated AAA).

    But my real question is why go through the bother of creating this new AAA CDO? Why doesn't the purchaser (e.g., a pension fund or an insurance company) simply buy the initial AAA MBS? Why is there an economic need to create a new derivative security?

  4. It is my understanding that CDO^2 and synthetic CDOs arose because they didn't require any actual mortgage or loan orgination. By using credit default swaps, or by re-carving up existing CDOs bankers could increase the volume of structured products without any real constraint. This allowed them to meet heightened demand for high yielding fixed income products and to generate a lot of fees. Everyone is happy until home prices start falling and the whole thing unravels.

  5. Shock Minus Control: I'm still not sure I understand.

    Imagine that I dig up a bar of gold (originate a mortgage). I issue a note that is backed by the gold (a mortgage-backed security).

    Now, imagine that the noteholder wants to create a new note (a CDO) that is backed by his existing note (that is backed by the gold bar).

    Creating his CDO and disposing of it is tantamount to disposing his original note (the MBS). Why not just dispose of the original note, instead of going through the bother of creating a new one that has no apparent purpose?

  6. But it does have a purpose - demand for high yield fixed income was higher than supply and it takes time to originate and securitize new loans. Rather than wait for Joe's Hot Subprime Loans in Orange County to originate a new mortgage, you can take the highest risk tranche of an MBS (the tranche that pension funds are not allowed to buy) repackage it as a CDO, pay S&P/Moody's to slap an investment grade rating on the top tranches and viola - you have a product you can sell to funds that are restricted to investing in only investment grade assets.

  7. SMC: what you say makes sense to me at an intuitive level. But I have learned to distrust intuition!

    Consider the following MBS. It has three layers; a senior, mezzanine, and equity tranche.

    Assume that the senior tranche is completely safe (rated AAA). Assume that the mezz tranche is divided equally between safe and risky mortgages (rated BBB). Assume that the junior tranches consists solely of risky mortgages (unrated).

    What you appear to be saying is that there was a huge demand for AAA securities. Moreover, it takes time to originate mortgages. So what is the solution? Why not take the mezz tranche above and create a CDO with two tranches; the senior tranche taking the safe component (so it is rated AAA and hence available to pension funds) and the other half left unrated. Hence, the CDO constructed in this way creates a new AAA security that fulfills the demand for AAA assets.

    This argument makes sense. Except for one thing. It takes the original design of the MBS as unexplained. The original MBS could have met this demand for AAA assets by unbundling its BBB tranche into AAA and NR components to begin with. Why did it not do so? Given that it did not, then I can see the need for the CDO to unbundle and inefficiently packaged MBS. But this does not explain why the original MBS was inefficiently designed.

    Or am I missing something obvious?

  8. SMC: After thinking about the argument I constructed above, together with your own contribution, I think that I have arrived at a quasi-satisfactory answer to my original question.

    Let us take the structure of demand for AAA, BBB, NR assets as given for the moment. Assume that this structure is more or less stable, but perhaps subject to an occasional exogenous shock (a regime change, so to speak; possibly the product of regulatory change).

    The initial supply of MBS are tailored to the prevailing structure of demand. It is difficult to embed contingency clauses in these debt instruments that might lead them to become unbundled in the event of a regime change.

    If this is the case, then should a regime change happen (say, a sudden proliferation of new pension funds around the world) that increases the relative demand for AAA assets, the natural market response is for the emergence of CDOs that effectively repackage the original MBS to meet the new demand structure.

    Something like this. If this is true, then it is an argument in favor of CDOs. This is the way the market reconfigures the structure of existing assets so that it is better alligned with the structure of demand.

    What do you think?

  9. I think you're on the right track - with the regime change being very low yields on traditional fixed income products leading to heightened demand for structured products like MBS. It will be interesting to see if CDOs remain some part of the fixed income market now that they have effectively imploded.

  10. I should add, other than the cynical reasons I've given, there is, or was, a purpose for CDOs - they helped to mitigate prepayment risk for MBS and therefore allowed certain types of funds to better match assets and liabilities. Also, they allowed for a high degree of customization according to risk profile.

  11. What do you mean by "prepayment risk?" The risk that a significant number of mortgages would NOT incur the prepayment penalty? Gorton (2007) claims that during the housing price boom, most of the high return on subprime MBS came from the prepayment penalties incurred by households.

    And yes, it will be interesting to see whether CDOs will remain a part of the fixed income market. If the line of reasoning I pursued above is correct, the answer will undoubtedly be yes. Although, the market for CDOs backed by subprime mortgage assets will no doubt contract.

  12. I'm probably misunderstanding this (since I'm not a professional economist nor a finance expert, but a layperson), so:

    Wouldn't the low yields on traditional fixed income products tend to reflect the high demand for AAA-type products? And wouldn't a shock in demand drive yields down further for both products? Why did the CDOs (have to) offer a higher yield?

  13. Prepayment risk is the risk to the holder of the MBS that more borrowers will pre-pay their mortgage before maturity - contracting the horizon over which the investor was counting on receiving cash flows.

    For example - say I own one asset, your mortgage, and I have one liability (my own mortgage). I am using your monthly payments to fund my own liabilities. If you suddenly pay off or refinance your mortgage with a lump sum payment , possibly because interest rates have fallen, I now have to go find some other source of financing to match my own liabilities. If interest rates are lower, I may not be able to completely offset my liabilities. This is known as asset/liability matching and its important for pension funds, insurance companies, etc.

    Now, i'm am far from being an expert in structured finance, but I believe that CDOs can be useful in spreading this prepayment risk across investors.

  14. Pani Pani: Well, I am a professional economist and believe me, I'm even more confused!

    But to answer your questions: [1] Yes, low yields would reflect both high demand AND high quality of AAA assets; [2] Yes; indeed, I think this is what happened; and [3] I am not sure whether AAA rated CDOs did offer a signficantly higher yield; if they did, it would have reflected the market's assessment that MBS CDOs were not the same grade as (say) Treasuries.

    SMS: Sure, I understand. I was just trying to be cute. That is, if what Gorton (2008) says is true (that prepayment penalties were a major source of return on MBS), then the "risk" in question is that people might NOT refinance and pay the hefty prepayment penalty (common on subprime mortgages).

  15. May I offer a visual suggestion? The cashflow in all these complicated CDO models is similar to a waterfall (hence why there usually called waterfall tranches).

    So when cash first starts to flow in, the top tray fills (Senior Tranche, AAA) followed by the middle (Mezzanine, BBB) and then the last tray (Equity / Junior, NR).

    You can tell that there is progressively more risk with each tray. If there isn't enough water (cash) flow, the bottom tray doesn't get filled - but at the same time it gets EVERYTHING that's left over - higher risk and return.

    SMC hit the nail on the head with the previous comment. There is essentially no market for BBB investment vehicles (when compared to AAA and NR). Therefore they need to be repackaged into something people can buy.

    Analogy: Looking at publicly traded companies - Most people will buy equity or fixed income (junior and senior respectively). Mezzanine financing is usually a special game played by venture capitalists and pre-IPO companies etc.

    This is where the recursion comes in. Because there's a pretty good chance you that if you take that middle tray and cut it again, you can get a similar recursive waterfall effect (with credit enhancements). That is to say, there's still a pretty good chance that the top tray of a BBB slice will still get filled and the bottom tray has similar characteristics to the NR slice.

    The marginal difference between a BBB - AAA slice is *close* to AAA (especially with credit enhancements like a Credit Default Swap) and a BBB - NR slice is *close* to an NR slice (technically marginally safer). But you can take your unpopular 1/3 and turn it into an unpopular 1/9 (and sell the other 2/9 as regular products)

    Did that make things clearer? Muddier? Or did I just repeat what everyone was just saying?

    I really like this "visual aid" I found

    It really helps visualize the basics.

  16. Hi Joshua: Thanks for your post.

    I did check out your visual aid. As with most of these I've seen, it contains an element of truth (as I see it). But any virtue it has is greatly outweighed by its faults (not the least of which is the "creepy" voice that speaks to the proletariat; helping them understand how they were screwed once again by the capitalist system).


    [1] financial panics have occurred even prior to the emergence of a subprime mortgage market
    [2] most of those "not so responsible" subprime mortgage holders continue to pay their mortgages
    [3] nothing was said about the government pressure to get banks to lend to subprime (low income, high risk, disproportionately minority households)
    [4] and nothing (substantive) was said about how a crash in real estate manifested itself in a worldwide systemic financial crisis. The same sequence of events described also occurred in 1990-91, for example (and yet no systemic crisis followed).

  17. Would you be able to recommend (if any exist) a book at an advanced undergraduate level on financial crises in general?

  18. Pani Pani:

    None exist (to the best of my knowledge). However, I plan to address the issue in the next version of my textbook (to be published by Blackwell-Wiley). Not sure how successful I'll be, but I'll give it my best shot.

  19. No, people bought them to sell them to other people, it's a ponzi scheme, I am pretty sure many buyers/sellers knew that, the ones that didn't are whining now.

  20. this is a nice article about some of those guys that were buying these pieces of paper:

  21. Hi David: nice blog you have here. I found it while looking for commentary on the de Long/Boldrin face off.

    >This argument makes sense. Except for one

    >thing. It takes the original design of the MBS

    >as unexplained. The original MBS could have met

    >this demand for AAA assets by unbundling its

    >BBB tranche into AAA and NR components to begin

    >with. Why did it not do so? Given that it did

    >not, then I can see the need for the CDO to

    >unbundle and inefficiently packaged MBS. But

    >this does not explain why the original MBS was

    >inefficiently designed.

    >Or am I missing something obvious?

    Two possible explanations:

    1. Fannie and Freddie. They started buying large quantities of subprime and alt-A paper starting in 2003. They were mindful of taking too much risk (obviously not mindful enough) so generally stuck to AAA tranches. They most likely stipulated the tranche structure of paper that they were willing to buy, hence the 80/20 split. Other regulated entities that purchased sub-prime paper might have followed a similar patterns: Federal Home Loan Banks, wholesale Credit Unions. That relative conservatism created opportunities in the lower tranches for CDO packagers. A potential problem with this story is CDOs contained all kinds of crap not just subprime/alt-A loans.

    2. Deals were done quickly and in bulk with standard terms. The left scope for the "crows" to peck over the corpse looking for opportunities. (This is essentially a catch-all explanation)

    Both stories rely on compliant fee taking ratings agencies willing to endorse the product with AAA ratings.

  22. Couldn't read every comment, so I apologize if I'm restating something already discussed. You have mortgage pool A and mortgage pool B; let's say that the loss distribution of these pools are completely indipendent. Let's suppose that with the BBB tranche you have a 5% default probability; a SIV buys the two mezzanine and issues a note that pays a yield with the proceeds of the mezzanines. The note has a rating > BBB because in order to have a complete default an event with 5% prob. should occurr on pool A and another indipendent event with 5% prob. should occurr on pool B. This way you've built a better rated note, and if everything's fine the SIV's shareholders can also gain a good dividend. This is one of the main points of the game. Hope it helps.

  23. very interested read your post and your post's comments :)
    finance solution

  24. you got a really useful blog, i have been here reading for about an hour. i am newbie and you success is very much an inspiration for me


    Information Technology

  25. Different angle on the discussion.

    Say for instance I am a foreclosure defense attorney. A borrower comes to me requesting that I look into the facts and circumstances of his mortgage.

    Looking at the correspondence from
    his "lender"(who is merely a mortgage servicer), I notice that the borrowers loan is being foreclosed upon by a trustee representing a securitized trust.

    Hypothetically, I am smarter than the average bear attorney, and I understand structured finance.

    Should I then challenge the "lender" to show evidence of what tranche my particular loan is in (if any). Was the borrowers loan in a mezzanine tranche that was sliced into a CDO, then resliced into a CDO squared??

    Was the borrowers loan already indemnified to the "investor",if one could even prove a chain of ownership of the borrowers note and mortgage at this point?

    Therefore, under the waterfall structure discussed, if the borrowers loan was in a super senior (AAA) tranche, or in the alternative in the senior tranch of the resliced mezz tranche in the new CDO, the question becomes where is the loss to the investor?

    Hasn't this loss (borrowers default) already been indemnified? But the trust is still allowed to foreclose to collect again?

    What of credit default swaps, has the borrowers obligation already been indemnified, and even more insidiously because the originator purposely set the loan up to fail because it too was in on the swap bet as well.

    What of Mortgage Guarantee Insurance, or pool insurance, did this already indemnify the investor for the default?

    What if the servicer gets to keep the residual "clean up" funds, is it right that the servicer forecloses on an already indemnified default in the name of the trust which has already been paid for this loss?

    You are the attorney, what would you advise?

  26. This comment has been removed by the author.