Lately a bunch of us were discussing the notorious topic of subprime mortgage debacle, that investment banks created instruments so complicated that even CEOs and board of such banks did not understand how large the risk was. To facilitate discussion, I put together the below as a recipe to construct such monstrous innovation which are so popular and high in demand in 2007 that they went out of subprime mortgages to be packaged.
Construction of a Mortgage Backed Securities Collateralized Debt Obligation (MBS CDO)
1. Bundle up mortgages, can be residential or commercial (or both), into a group of asset that generates revenues using mortgage payouts.
2. These mortgages can come from a variety of sources. When the good ones are taken, worse mortgages have to be considered. Most mortgages do not have good rating, and especially after the good ones are packaged away, subsequent bundles contained a lot of sub-prime mortgages, with a BBB rating (worst).
3. With these bundled up asset, securitize (i.e. issuing shares of) these bundles with the mortgages as backing. Let say there are 10 mortgages, and we can now issue 100 shares representing these mortgages. So each share contains the asset of 1/10 of an average mortgage of these ten. These shares are now called Mortgage Backed Securities. For detail you can go here in Wikipedia for a nice explanation.
4. The MBS can be traded in the market, but since there are so many different type of mortgages, trading these MBS is not very efficient. To solve this problem, a new product called Collateralized Debt Obligation is created by bundling up various MBS together, with the mortgage payment as revenue stream for these vehicles.
5. There are different seniority level (called tranches) in a CDO. The most junior ones are called "first loss" and would take all the loss if some of the mortgages go default. The most senior tranche are protected from such loss, until all the tranches beneath are defaulted. Of course different tranches are priced differently based on the risk carried.
6. For the most senior tranche, usually investment bank would add Credit Default Swap (CDS) with a small payment per year to the CDS issuer. CDS is essentially an insurance policy which the issuer would pay the purchaser a large amount if the underlying product goes default. In return, the purchaser would pay a small premium periodically (like 2.5% for good debt) depending on the risk and likelihood of default of such protected asset. CDS is liquid and can be traded in Over-The-Counter OTC market which is not regulated or monitored by the SEC. Using their statistics model, the CDS issuer (AIG being one of the biggest) considered it to be a very rare event (black swan?) for the most senior tranche to go default, as all mortgages would have to go default in the entire CDO for the most senior tranche to go busted, the risk of such tranche is low, and thus lower CDS premium.
7. Combine the protection of CDS into the CDO and now these structured product would appear very safe and have a new credit rating. A lot of them got rated AAA (best) even though ultimately the asset are mostly subprime mortgages (BBB or lowest rating).
Risk is characterized by the likelihood of such event. If the company is very unlikely to go default, then it is said to have very low risk. It was assumed that it is very unlikely for all subprime mortgages to go default, and in addition also very unlikely for the CDS issuer (AIG) failing to pay back the policy purchaser. Thus presumably these CDO products have very low risk and thus have very high credit rating (AAA, as high as US Treasury Bonds). The problem was that these rare random events are NOT orthogonal (i.e. not uncorrelated). So if one goes, the others go with it in a chain reaction fashion, and trigger further snow ball effect in a downward spiral. Thus the probability of the most senior tranche going bust is no longer a multiple 6-sigma event. A 6-sigma event is loosely borrowing the low risk meaning of six sigma quality where it guarantees that there is a chance of 99.9999999% for the product to pass. Noted that such faulty assumption of orthogonality had already brought down Long Term Capital Management (LTCM, a hedge fund created by Nobel Laureates including the creator of the industry standard Black-Scholes model for option pricing) back in the 90s. Thus it is not entirely honest for the credit agencies to just blame on wrong assumption of their models. They would not, and should not have missed this.
Of course there were lots of profit taking involved (for instance the credit rating agency will get a handsome “registration transaction” fee if the product get a good rating, thus creating incentive to bless the product; in turn if the products have good rating the banks can sell them at a better price).
But that’s just CDO. The demand of these vehicles was so high (high yield and low risk, who doesn’t want it?), they actually went out of mortgages to be packaged as CDOs because there were only a limited number of physical American home buyers. Actual MBS and their CDOs were soon sold out to foreign investors. I am not joking. To satisfy demand, the investment banks created Synthetic CDOs which had no physical backing (i.e. does not have any attachment to anything in the real world) but were just bets against each other with the MBSs described above as reference (**), without actually owning any MBS. In contrast a CDO has real MBS asset backing.
The now famous Goldman Sachs Abascus 2007 (where they created a product so that John Paulson could bet against Goldman's own customers) was a special company (SPV) registered in Cayman Island constructed by these synthetic CDOs "referencing the performance" of a pool of MBS. In turn this vehicle was sliced up into tranches which investors can buy at different price and yield. You can find the details of this company in this Goldman Sachs presentation. Can you imagine being the accountant who filed tax for such “company”?
** An example is the weather future, where farmers can buy to hedge against the risk of their crop yield in case bad weather happens. Obviously you cannot claim or ask for delivery of good/bad weather in such a derivative. The future is just tracking weather as its reference. Weather or mother nature does not care or owe anything to either side of the future trade. It is like gambling which team will win in a soccer match. Actually a lot of ETFs are created this way, without owning any real stocks but just “tracking the performance of the underlying”. Following the same logic if panic sets in again, ETFs may go busted before the real stocks do as the counterparty can no longer fulfill their payment duty, regardless of what may happen to the real companies/market.
For a time CNBC hailed these sophisticated products as the latest great invention being “manufactured in Wall Street”. Anyway the investment communities as a whole were and still are drinking the same kool-aid. If there was no demand, there would be no supply. There were simply too much money out there seeking a place to go. If we are to do a witch hunt, the target would probably be regulator (SEC and Fed) and the credit rating agencies.
