The New York Fed and the Wall Street Journal have both been studying how liquid the CDS market today and have released their conclusions today. Short answer: not that liquid. From the WSJ:

In recent years, credit-default swaps—contracts that give the buyer the right to collect a payment from the seller if a borrower defaults on its obligations—have risen from obscurity to an avidly tracked barometer of the financial health of everything from Bank of America Corp. to Greece. … Yet a Wall Street Journal analysis shows that actual trades in these widely cited derivatives are few and far between—and the quotes that market observers bandy about often aren’t based on actual trades at all.

What I liked most about the FRBNY study is that it not only looks at overall liquidity but – sort of – gives you a window into the breakdown between what you could call “initiation trades” and “closeout trades.” And this in turn tells you something about not just “liquidity” in the abstract but about how market makers go about providing that liquidity.

Now the starting point is that, unlike in cash markets, there is no obvious way to measure CDS close-outs. If I buy 100 shares of Bank of America (the WSJ’s CDS example) and then get bored with it in a year, I go sell 100 shares of BAC, and I am left with zero shares of BAC and some profit or loss. Straightforward enough.

If on the other hand I buy $10 million of 5-year credit protection on BAC (*note “buy” = pay some up-front premium and agree to make quarterly coupon payments for the next five years), and in a year I get bored with it, then I could in theory do one of about three things:

1. Sell back my remaining 4 years of protection to the dealer I bought it from, i.e. tear up the contract at some profit or loss. The FRBNY calls this, sensibly, a “termination.”

2. Sell my remaining 4 years of protection to someone else, and convince the dealer I bought it from to deal directly with that person and leave me out of it – so that I have no position, just some profit or loss. The FRBNY calls this an “assignment.”

3. Sell offsetting 4 years of protection to someone else and, for the next four years, have offsetting payments until both contracts terminate. This is like #1 and #2 but leaves me with CVA and administrative pain.

In a rational world you would do #1 or #2, whichever gave you the best price – if a dealer would pay you more than a third party, you’d terminate; if a third party would pay more you’d assign. Simple enough.

Here’s what happens:

As FT Alphaville says in their smart look at the FRBNY study:

This shows that 90 per cent of activity is new trades, when measured by notional. This doesn’t directly translate to an increase in the overall economic, i.e. net, position as some of these trades will offset other trades that a given trader already has.

I think this tells us a couple of things that are probably obvious to CDS traders but interestingly confirmed by this data. First of all, it is very hard to get out of a CDS trade. The FRBNY study doesn’t just measure “liquidity,” it measures turnover. BAC stock trades about 3% of its market cap every day; in other words its market cap turns over about 8 times in a year – and every trade is turnover in the sense that someone is getting out of the stock and someone else is getting in. In contrast, only 10% of CDS notional trades are true turnover – that is, termination/assignment trades, my categories 1 and 2, where one party actually gets out of a position.

Why would that be? One obvious answer is that CDS are bilateral contracts that don’t generally have free assignment rights. So dealers can quote tight markets for initiation trades, but make a lot of their money on restructuring/termination/assignment. A client who is, say, winding down a fund and has to terminate a CDS contract may not have much negotiating leverage – and more importantly, he can *only* deal with the dealer who sold him the contract in the first place, who can therefore quote him more or less whatever price he wants. The only way to keep him honest is to get a competitive price for a new trade from someone else – that is, if you have the ability to take on more gross exposure. So gross notional keeps swelling. Some people think that’s a bad idea, since gross CVA and systemic risk keep growing right alongside.

The nerdier question is, how does approach #3 work: do you actually do an offsetting trade in the same tenor, or do you sell offsetting protection that is not 4 years such that you have a roughly credit-neutral position by some metric. Thus for example I could sell $5 million of 5-year credit protection on IBM and $5 million of 3-year protection to be approximately notional and duration neutral. Or I could sell $8 million of 5-year protection to be approximately duration neutral (but have a non-zero jump to default risk).

That sounds sort of silly, but because CDS is so concentrated in the 5-year tenor there’s reason to think that the market has a big component of exactly that sort of rough-offset trade. That sort of sucks because (1) if you try to offset 4-year risks with 5-year protection you will have imperfectly matched risks (zero DV01 in some range but different jump-to-default risks) and (2) it creates a Swiss guinea pig problem if the only thing that ever trades is the 5-year.

The single-name data, though, suggests otherwise:

So almost half of single-name trades are in the standard 5-year maturity. But although this shows that the 5-year is by far the most liquid, it also suggests that customers can still get a decent price for their stub single-name maturities. There’s about as much activity in 0-to-4 (combined) as there is in the 5-year, which you could imagine means that every $47 of brand new 5-years are created while $12 of last year’s 5-years, $10 of two years ago’s 5-years, … etc. are offset by new 0 to 4-year trades which add up to something like $40 – meaning that you can expect a large majority of each year’s “new” 5-year CDS to eventually be offset by shorter-dated trades when the initial holder wants to exit the position. Add in the trades that are exited by actual terminations and you’re left with a pretty small stub of trades that are held to maturity.

Which suggests that, unlike in the termination/assignment trades, dealers make pretty competitive markets in “off-the-run” trades, and that CDS market liquidity – though it may be light in total, and may not always be the best “avidly tracked barometer” of everything you might want to know – actually fulfills its main function of allowing customers to trade in and out of credit exposure.