Despite their name, CDSs are not swaps, but insurance policies. This is not merely a philosophical argument about names, as the following will make clear.
First, let's look at a real swap - for example, a foreign exchange forward swap.
This is how it works:
On Date X
Bank A sends to Bank B 100 million dollars
Bank B sends to Bank A 67.567 million euro
So far, this looks like a spot USD/EUR foreign exchange transaction taking place at a rate of 1.48 dollars per euro. Notice that actual cash changes hands on both sides: Bank A and Bank B have to send money to each other.
On Date Y, the transaction is reversed:
Bank A returns to Bank B QQ.qqq million euros
Bank B returns to Bank A 100 million dollars.
The exact amount of euros that is returned (QQ.qqq) and the duration of the swap (Date X to Date Y) are agreed and set at the inception of the deal. Again, actual cash payments are exchanged.
These swap transactions are extremely common, with hundreds of billions of dollars, euros, yen and other currencies swapped daily. The crucial point to keep in mind is that they are real swaps, i.e. the parties involved have to send money to one another at the same time.
There are other transactions also termed swaps, e.g. Interest Rate Swaps (IRS). They, too, involve the concurrent exchange of cash flows between the parties involved and are very common. More information on them can be found here.
Unlike the above, a Credit Default Swap transaction swaps nothing. One party buys an insurance policy against default and pays fixed premiums, typically for five years, while the other party sells insurance policies and receives premium income. Like all insurance operations, the seller is exposed to significant event risk and the buyer expects that his insurance company is well enough capitalized to meet its obligations, when it is called to do so.
At this point the difference between a credit insurance scheme and a real swap is obvious. Indeed, when they first started to become popular around 1999-2000, CDSs were not called swaps but, more accurately, Credit Insurance Contracts. Somewhere along the line, for reasons not entirely clear, the terminology was changed to the more familiar and innocuous-sounding "swap".
I leave it to the informed reader to deduce why a name which clearly denoted actuarial risk and the need for substantial reserves, was exchanged for one that falsely implies the concurrent exchange of cash flows.
Who is in the CDS market?
Credit defaults risk is certainly being spread around this way, but is it adequately reserved against? We won't know until the next credit cycle and, as Warren Buffet says, that's when we will find out who is swimming naked. Knowing how the above participants operate when in party mode, I wouldn't bet on many swimsuits being worn.
But don't take my word for it; look at what is happening right now with mortgage-backed CDOs, SIVs, etc. They, too, were constructed with the extremely optimistic assumptions that defaults would remain at record lows. Now that defaults are rising, even AAA paper is defaulting in one go. This toxic paper was manufactured at the exact same "plants" that also sold credit insurance, by the exact same workforce and under the same assumptions.
Notional CDS amounts matter and should not be immediately dismissed by comparing them to their market value.
The reason, once again, has to do with their insurance policy nature. For as long as defaults are low, anyone can make money writing enormous amounts of credit insurance. Property and casualty insurers are also very profitable, until three Category-5 storms hit in a row. Dismissing notional CDS amounts in favor of market values is similar to a P&C insurer disregarding how much insurance it has underwritten and focusing instead on the present value of its premiums.
The last time defaults rose sharply was in 2001-2002 (see charts below). At that time, the total amount of CDS outstanding was $918 billion, whereas today the amount has reached $45.5 trillion (ISDA). Very little debt was covered by credit insurance in 2001 and thus the effects on the financial system were minimal. We simply cannot compare that experience with today's massive risk exposure. Yes, we had the equivalent experience of three Cat-5's in 2001, but there was very little insurance outstanding back then.
Perhaps the CDS market is so spread out that it cancels itself out. Perhaps everything is just perfectly matched between credit assets and liabilities that nobody is holding a risk tail. But with $45 trillion around I wouldn't bet on it.
And you know what? Greenspan was worried about that, too. Last time he spoke on the subject he urged bank operations departments to rapidly clear their CDS paperwork so that open positions can be netted and the real risk profile of each may be assessed. Apparently there is some progress being made on this front.
But the truth be told, netting is not all that it's cracked up to be. All it takes is one nervous risk officer to hold up payment until he receives the other guy's money first and the whole structure goes poof. When push comes to shove everyone holds on to their cash much more tightly, everyone gets much more suspicious. No one wants to be on the no-show side of a "fail".
The full story will be told, like always, after the fat lady sings. For credit default insurance this means during the next high of the default cycle.