The second of three parts of Dennis and O’Harrow’s series on the downfall of AIG introduces two new features—credit default saps and Joseph Cassano. Together the two managed to bring down AIG.
Thus far we have only read about how AIG got into the business of writing CDSs — it seemed too good to be true — money for nothing. My guess is that the first ones they wrote did amount to money for almost nothing. However such opportunities are few and if one considers one year’s money for nothing as the minimum acceptable profits for next year you will end up getting money for a while for destroying the world’s largest insurance company.
To me the interesting part is how they got into the CDS market
in early 1998: …
… a new kind of contract known as a credit-default swap. For a fee, the firm essentially would insure a company’s corporate debt in case of default. The model showed that these swaps could be a moneymaker for the decade-old firm and its parent, insurance giant AIG, with a 99.85 percent chance of never having to pay out.
The computer model was based on years of historical data about the ups and downs of corporate debt, essentially the bonds that corporations sell to finance their operations.
My guess is that the model was right and that returns on writing (less than 10 year) CDSs on corporate debt were huge while the risk was nothing that AIG couldn’t handle. This profit opportunity could exist because of binding prudential regulations — banks worried about capital requirements and managers of endowments and pension funds with rules requiring much of their assets to be AAA corporate debt or safer would be willing to pay much more than the actuarially fair premium for a CDS on, say, AA debt. Given the correlation of default on different AA bonds, writing CDS on a broad array of them is essentially money for nothing. The most likely value for payouts is zero and the expected value plus, say, 10 standard deviations would be less than the fees.
I have bolded the two factors which convince me that the model might have been reasonable when it was written. First this was corporate debt not new financial products so the ratings were based on decades of experience and not on, say, the assumption that the probability of a nationwide average house price decline is zero (an actual assumption made by Standard and Poor’s really). Second, I sure hope the “years” also amounted to decades including recessions and such. The use of sample frequencies to estimate probabilities is always risky, but if the sample size is a few years during an unprecedented house price boom, it is insane.
The problem is that once one has gotten money for nothing it is very hard to convince oneself that there is no more to be had.
The aim of this article seems to be to describe Hubris* — the pride that came before a fall. This is an old theme in drama — the oldest. I always read about it when I read stories of financial collapses. I suspect that it might be added, because it makes a morality play out of math. I suppose it might often be relevant because of human nature. However, I also think it is particularly relevant to finance.
The efficient markets hypothesis (adjusted for inflation) says you don’t find hundred dollar bills on the sidewalk. It is false. However, you don’t find a hundred dollar bill on the same concrete square day after day. It is possible for financial operators to find an anomaly and make a huge profit with moderate risk. At best they can find two or three. There don’t seem to be many of those who don’t conclude that they are geniuses who can find such profit opportunities year after year. A shining exception is Andrew Lahde who made a killing and cashed in. However, he isn’t in the business anymore.
I don’t see a solution to this problem. You have to put up either with people who think they are geniuses, because they have been very successful or who have something to prove. In particular, I think the replacement of Tom Savage by Joseph Cassano may have been very costly for AIG (Savage really retired according to all accounts). The huge profits made by Savage and Sosin (his predecessor) set a standard which Cassano was determined to surpass. It is very unwise to employ someone who considers finding hundred dollar bills on the sidewalk to be barely adequate.
It is also interesting that Cassano is not a quant:
A Brooklyn College graduate, the 42-year-old Cassano was not one of the “quants” who had mastered the quantitative analysis and risk assessment on which the firm had been built. He had no expertise in the art of hedging. But he had excelled in the world of accounting and credit — the “back office,” as it is known on Wall Street.
Now quants are clearly dangerous, but at least they know the silly assumptions they made when writing their models. In particular they must understand the fact that quantitative models require parameter estimates which are definitely not reliable unless based on large amounts of data. I don’t think someone who is good in the back office understands that estimates based on small samples are not just less precise but also have distributions which can’t be determined at all based on available data.
Anyway, the third article in the series should be bloody.
* I’m pretentious enough to use the word but I won’t spell it correctly as Brad does.