What is a synthetic subprime CDO?
Six months ago I was asked to explain what a synthetic CDO was. Well pal, the sky is blue because bankers in Wall Street…. I sort of did my best and drew a few diagrams and we got about halfway there and came to a halt. I understand roughly how a synthetic CDO works, I think, but the problems implied in such a thing are not easily imparted to someone who does not follow finance. Another one of my friends asked me the same question and I thought maybe I could gin up a couple of diagrams that other people might find useful as well.
The trouble with descriptions on paper is that it is hard to build a mental model in your head. I have had a go at doing so below.
I have built these based on what I’ve read in books like The Big Short (Michael Lewis) and The End of Wall Street (Roger Lowenstein). I am not directly familiar with CDOs and have never read a CDO prospectus. I have taken some liberties with the structure of mortgage bonds, simple interest, etc to simplify everything as much as possible. This is a super simplified diagram that I hope will prove useful.
An illustration of the steps that lead to the creation of synthetic CDOs:
Step 1: Get mortgages
Banks lend money to people to buy houses. Some loans are more risky than others. Not all loans get paid back. ‘Sub-prime’ loans, the primary cause of the crisis, are riskier loans to less reliable borrowers.
When you repay a loan, you must repay principal (the loan sum, e.g. $500,000) and interest (the interest, e.g. at 3% per annum, that accrues on the principal. I.e., you pay 3% of $500,000 per year in interest).
If a loan is not repaid, the bank takes the house and sells it to get payment for the loan. Sometimes the house does not sell for a high enough price for the bank to get its money back, and the bank takesx a loss.
Step 2: Make mortgage bonds
Once a mortgage has been written, if you are a bank and no longer want to take the risk of that loan not being repaid, it is very hard to ‘sell’ a mortgage to another bank or investor. If you package up (say) 1000 mortgages into a mortgage bond, however, they can now be sold and a bank can rid itself of the risk by selling the mortgage bond to someone else. The bank gets some money, and those other investor/s who bought the mortgage bond, earn interest on the mortgages and take the risk of not being repaid their principal.
Mortgages in a bond are ranked from highest (AAA) to lowest (BBB-). Investors at the top get the lowest rate of interest, but also take the lowest risk and have the highest likelihood of being repaid, for reasons I’ll explain in a moment. A bond could be worth $100m or more, so one investor does not own the whole thing. Little pieces of the bond are owned by dozens of investors. Some investors will hold only AAA, while others will hold BBB-, for example. This is a sample mortgage bond, with just 5 ‘tranches’, or levels.
Note the 7% figure in the BBB- level. This is the level of losses required before the BBB- holders will be wiped out. Now, imagine that this bond contains 1000 mortgages. 10 of those mortgages (1% of 1000) fail and the mortgages are so bad that once the house is sold the investors in the bond recover no money. Investors in the BBB- tranche of the mortgage bond take 1% losses. However, because the BBB- tranche becomes worthless with just 7% losses, BBB- investors have actually lost 1/7th, or ~14% of their investment.
Many finance books use a ‘waterfall’ analogy, with the ‘waterfall’ (cash; mortgage + interest repayments) falling on AAA holders first, before excess spills down to AA, and then A holders, and so on.
Ratings agencies decided which parts of a bond were how risky, e.g. did a tranche have a AAA or BBB- level of risk?
Step 3: Package multiple mortgage bonds into a Collateralised Debt Obligation, or CDO
In order to sell more mortgage bonds of crappy mortgages, Wall Street packaged parts of numerous mortgage bonds (100 or more) into a kind of ‘super bond’ called a CDO.
Due to flaws in the rating agencies (S&P, Moody’s, Fitch), Wall Street was able to ‘game’ the ratings agencies models and get many BBB- securities re-rated as AAA. See the below diagram:
In this diagram, the BBB- tranches from 5 mortgage bonds (top) are piled into a CDO (bottom left). The ratings agencies would then re-rate as much as 80% of this BBB- debt to AAA (bottom right).
CDO creators were able to convince ratings agencies that BBB, or lower rated debts, were somehow not BBB at all when they were piled into a CDO. Quoting Michael Lewis in the Big Short, up to 80% of low-grade debt could be subsequently re-rated as AAA, or highest quality.
So you have a CDO composed of junk (BBB-) being sold as gold (AAA). AAA debt was more highly priced so banks made a lot of money via this practice.
Now in the original mortgage bond, the BBB- tranche takes losses first. Once the first, say, 7% of losses (people not repaying their mortgages) are taken, the bottom BBB- tranche is wiped out. However the re-rating of the debts in the CDOs created an illusion.
In this image, a regular mortgage bond takes 4% losses, wiping out just over half of the BBB- layer:In the above diagram. We’re assuming that the BBB- tranche is wiped out with 7% losses. So 4% losses (40 mortgages out of 1000) wipe out most – 4/7ths – of the triple B tranche, as intended. However, since so much of the higher-rated tranches in CDOs were actually disguised triple B tranches from mortgage bonds, 4% losses in a CDO might look more like this:
Instead of losing just 4% of the bottom tranche, a CDO might lose 4% from all or most tranches, because they were all disguised BBB-. If there were 13 tranches and all were BBB-, 4% losses in the original mortgage bonds could be a 52% loss to the CDO. Quoting Michael Lewis again, 4% losses in mortgage bonds happen in good times.
Due to the way they were structured, very small losses could hurt higher tier investors also.
This is just for illustration. In reality CDOs were far more complex. Next up, is the ‘synthetic’ CDO.
Step 4: You can’t get enough mortgages to feed the demand for CDOs. Enter the ‘Synthetic’ CDO.
There weren’t enough mortgages to go around to feed the creation of CDOs. So Wall Street started creating ‘synthetic CDOs’ from the payments from investors betting against CDOs. To keep this short I assume you know what a credit default swap is, but I am going to use an ‘insurance policy’ analogy to describe them.
Situation 1: This is how ‘The Big Shorters’ (Mike Burry et al) and the CDO market originally worked. Investors bought ‘insurance’ (credit default swaps) against a regular CDO defaulting. Companies like AIG Insurance sold these types of ‘insurance’ products:
The typical interest payment to CDO owners (orange) might have been something like ‘2% above LIBOR’ (London interbank offer rate) but I have just called it 3% for simplicity.
At one point, AIG FP stopped writing insurance. Now Wall Street banks and other investors started writing the ‘insurance’ risk, and Wall Street used the ‘insurance’ payments to create synthetic CDOs to feed demand:
Read the diagram. Unlike a normal CDO, there is no property or mortgages underlying these ‘synthetic’ CDOs. This has a few implications. If the investors buying the insurance policy (credit default swaps) go bankrupt, CDO investors lose everything. Additionally, in this situation there is no property that can be sold to recover the losses, because there are no mortgages.
I couldn’t tell you if that was fraud, as I am not a lawyer. But it looks pretty outrageous.
The greatest concern is actually a bit more abstract – with synthetic CDOs and credit default swaps, there is no limit to the risk (defined as total maximum amount of loss) that can be placed in the financial system. For example, no matter how bad the loans in the USA got, there is always a certain maximum amount of risk. For example say the population is 300 million people. There is a vague, but fairly firm, limit to the number of mortgages you can write for this many people. Just for illustration, there couldn’t be more than, just say, 400 million homes and, say, 800 million home loans. If each loan is $300,000 on average, the total maximum amount of money that can be lost on the loans is something like $240 trillion, plus interest. Admittedly this is an outrageous sum, and far more than was lost in the GFC. But at least with a mortgage you know the upper limit on your risk, and there is a property that can be sold to reduce some of the losses.
However, with synthetic CDOs you can create as much risk as you like because the number of homes and borrowers is not a constraint. This is why many crisis commentators and post-crisis books take an apocalyptic tone – because the world really could have been looking at financial Armageddon.
I hope you found the diagrams useful.
I have no financial interest whatsoever in any of the above mentioned products or companies, directly mentioned or implied. This is a disclosure and not a recommendation.
This is a very rough description of how they worked and I know that it glosses over several important aspects. This post is more about writing something that would explain how a CDO works (to someone with no financial background), than it is trying to be a definitive treatise on CDOs.
If you know more about CDOs and think I have misconstrued something or missed an important point, I would be glad to hear from you.