Synthetic CDO Tsunami, not so much

I mentioned yesterday that synthetic CDOs might prove to be a massive windfall for certain financial companies. I wrote this based on an article I'd seen quoted several times a few weeks ago. After briefly puzzling with an explanation while trying to write yesterday's post, I realized that I hadn't dug into the details of the matter. So today I did.

The upshot is that the writer of the original article Alan Kohler seems to have misunderstood the underlying financial products, and that the large transfers hinted at are unlikely. Among the usual chaff comments at MetaFilter's reference to the original, there are several salient critiques, and what follows is a long and dorky explanation of the whole shebang. If you're not up for that, the take-away is that synthetic CDOs aren't going to be a cataclysm for pension funds, nor will they sprinkle pixie dust on the financial industry.

Ready for some TLAs?

Good. The first one is CDO, for Collateralized Debt Obligation. These are financial structures to distribute risk and return disproportionately among investors. Not all investors wish to take the same amount of risk, nor do they hope for the same level of return on their investments; an insurance company typically wants modest return and little risk, while a hedge fund manager is willing to accept significant risk in exchange for higher returns. Furthermore, the universe of fixed income securities in aggregate is unlikely to match the risk versus return profile of the universe of fixed income security investors in aggregate. (In fact, we should be surprised if this were the case, as it would imply a rather shocking set of coincidences.) CDOs provide a means to match the needs of the buyers with the offerings of the sellers, and Wall Street has made a nice business of setting them up. It's nothing evil.

Let's suppose we have $200M of 30-year bonds in XYZ Corp that produce a certain income stream, say $10M per year. If we form a legal entity, say a limited partnership, and sell 1000 shares according to the following plan: 70 shares are 'senior' and share equally the first $7M of yearly return from the XYZ bonds. The next 20 shares are 'mezzanine' and share equally the next $2M of yearly return, provided that there is any after the senior shares have had their take. The remaining 10 shares are 'equity' and take whatever is left. (In financial engineering terminology, each class of shares is referred to collectively as a 'tranche'.) Obviously each senior share is worth more than each mezzanine share which is in turn worth more than each equity share, and since the aggregate is worth around $200M, each senior share will be worth more than $2M and each equity share less than $2M. Suppose the price breakdown is $2.4M/senior share, $1.4M/mezzanine share, and $0.4M/equity share. (In $M that works out to 70*2.4 + 20*1.4 + 10*0.4 = 168 + 28 + 4 = 200.) What is the percentage yield for each tranche? For the senior tranche, it's $0.1M/$2.4M = 4.1733%; for the mezzanine, $0.1M/$1.4M = 7.14%; and for the equity, $0.1M/$0.4M = 25%. The senior shares pay a modest return, but in the event of a default have a much higher chance of full recovery than the equity shares, which would in all likelihood be worthless. Furthermore, if XYZ Corp should meet adverse or advantageous business conditions, the risk of default rises or lowers, and in the secondary bond market, the price of the underlying $200M may fall or rise, respectively. Likewise the share prices of the various CDO tranches would fall or rise, but the change would disproportionately affect the equity shares.

Next up is CDS, for Credit Default Swap. It's a kind of insurance in which the protection seller guarantees the notional amount be given to the protection buyer in the case of a default by the reference entity. (It's actually more complicated, but that's the gist of it.) Here's a simple example. Suppose you have $10K in your savings account at your local bank. If your bank becomes insolvent the FDIC guarantees that your savings will be returned to you; your bank buys this insurance for you, paying FDIC a few pennies per $1K per year. In this case you are the protection buyer, the FDIC is the protection seller, the reference entity is your bank, and the notional amount is your savings balance, $10K. Now suppose the protection buyer is a pension fund, the protection seller is a large investment bank, the reference entity is XYZ Corp, and the notional amount is $200M. If the pension fund holds $200M in XYZ bonds, then it might well want to reduce its exposure to the risk of XYZ defaulting by paying the investment bank a yearly premium. The pension fund is less at risk should XYZ meet adverse conditions, and if all goes well, then the investment bank gets a small profit. All is well, provided that the investment bank stays solvent and remains able to meet its obligations.

(Ah, hindsight.)

If you stand on your head, you can see that a CDS contract can behave much like a bond on the reference entity. Huh? Suppose the investment bank determines the risk of XYZ defaulting is 10%, and so buys $20M of US Treasury bills and charges the pension fund $1M for the insurance. If the bank has a few billion insured, then so long as they've properly estimated the risk of default, their reserves will be adequate to cover any losses. In this case the bank gets a return on their $20M very much like the return they would see if they held XYZ Corp bonds. If XYZ Corp defaults, they will lose their investment, if not, they see a 5% yearly return.

Now if we take a set of CDS contracts and package it as a CDO we have a synthetic CDO, so named because we have synthesized the underlying bonds with CDS contracts. That's really all there is to it. But note that synthetic CDOs still have tranches, still have cash flows, and function much like CDOs. And that's the point. A few defaults among the reference entities will not cause the entirety of the notional value of these instruments to be transferred from one party to the next, but rather, just like a CDO, each tranche will be impaired and then wiped out in turn.

Now go and (re)read Kohler's article.

Cui bono?