Yesterday I explained the basics of a credit default swap. Now let's explain how they sunk AIG.
The obvious risk of the CDS is that the original bond issuer (Barb in the simple explanation) cannot repay her debt. In AIG's case, they did extensive computer modeling of this risk. They felt pretty good about the risk.
However, there were two additional risks not considered in the computer models. First, the CDS contract had provisions that could require the guarantor (like Sid in yesterday's example, or like AIG in real life) to put up collateral. That collateral would ensure that the guarantee would be solid. AIG received several calls for collateral from its counter-parties. That placed a significant burden on its balance sheet.
The second risk was marking the contract to market prices. For the guarantor, the CDS is a liability on the balance sheet. When other folks are trading credit default swaps, they establish a market price for this liability. AIG was forced to recognize larger liabilities on its CDS portfolio, even before it had sustained an actual loss.
Imagine this scenario: you've done a great job assessing risk, but the rest of the market is nervous. They start pricing CDS's as if they are very risky. You are right, they are wrong. What happens? We would like to live in a world in which the folks who are right profit, and the folks who are wrong lose money. But in this scenario, the guarantor gets in trouble even if its right about the credit risk, simply because it has to recognize losses on the market value of its portfolio, even though it has not had to actually pay off on any claims.
If the guarantor has a lot of staying power, it simply accepts a hit to its balance sheet while waiting to be proved right. As the CDS's expire over time, it can recover the reserves it booked when it marked the liabilities to market, showing huge profits. That is, if it has staying power.
However, it that guarantor is highly leveraged, the financial losses it has to report (before it has any genuine losses) cause its own credit rating to suffer. It has to post more collateral with its counter-parties. Its borrowing costs go up. Nobody wants to do business with it anymore, because its finances are shaky. Eventually, the company goes bust. Or gets bought by the U.S. Treasury, which may be even worse. And all of this bad stuff can happen even if the guarantor is ultimately right about the risks of the borrowers it was guaranteeing.
Listen to any of the radio personal finance shows, such as Clark Howard or Dave Ramsey. They'll tell you not to co-sign credit card applications or car loans or mortgages, not even for your own family members. If you're a corporation with a very strong balance sheet, you can co-sign a few notes and make a little easy money. But if you start doing this on a large scale, a large enough scale that it really helps your company, then you are piling up risk that ultimately could sink your company.
There's an old saw in investments: The market can stay irrational longer than you can stay liquid. Be careful.