The gnomes at the Bank for International Settlements have produced a particularly gnomish paper called "Collateral requirements for mandatory central clearing of over-the-counter derivatives." Wait! It might be important! Hear them out. (There'll be charts ...)
Their goal is to measure how much more cash collateral the big dealer banks will need to encumber to make the transition to central clearing of derivatives (specifically interest rate swaps and CDS) under various forms of central counterparty regime. If you're interested in that sort of thing, and some people are, then this provides you with much fodder for chewing ruminatively. It sort of reads like an econometrics final exam that, speaking only for myself here, I would fail, so I can't really say how ruminatively you should chew it. You could clearly make different choices at many, many different points, and while each individual choice seems thoughtful enough you wonder whether the accumulation leaves you with a toy paper at the end.*
Now, if you're not particularly into the constraints on cash that would be driven by central clearing of derivatives, you can still get something from this paper. Specifically, this gives you a sensible look at how much damage the Financial Weapons Of Mass DestructionTM can do at any one time. Because "margin" is a proxy for something else, specifically likelihood of a big loss. Here's how they do their math:
We assume that prudent CCPs [central counterparties] would set variation margin calls equal to daily losses on the parts of these portfolios that they cleared and initial margin requirements equal to the 99.5th percentiles of possible five-day losses. A 99.5th percentile loss is the smallest possible ‘large’ loss if ‘large’ losses are defined as those with a chance of 1-in-200 of occurrence.
So, in the absence of central clearing, this is a rough rough proxy on how much damage derivatives can cause to the financial system in a big market move. Here's initial margin - that is, a 1-in-200 loss over a five-day period - for the G14 dealers (BofA, Barclays, BNP Paribas, Citi, CS, DB, GS, HSBC, JPM, MS, RBS, Soc Gen, UBS, Wells Fargo) for interest rate swaps:
These show initial margin requirements that the authors calculate - under low, medium, and high volatility environments - for all of the G14 derivatives dealers. In the right-hand panel, the bars show initial margin as a percentage of all assets, the diamonds initial margin as a percentage of unencumbered assets.
And here's CDS:
Those numbers are ... big, biggish anyway; you could read them and FREAK OUT, I suppose. The BIS authors don't particularly:
For both IRS [interest rate swap] and CDS portfolios, initial margins would only encumber a small proportion of G14 dealer’s assets, even when these margin requirements are set amidst high levels of market volatility. Although many of these assets may not be acceptable as collateral to CCPs in the first instance, dealers could swap them for eligible securities, either through outright sales and subsequent purchases or via asset swaps.
This is true, probably, and if you don't spend your time reading about rehypothecation the whole "dealers can post a few extra tens of billions of dollars of collateral without interrupting anything that they're doing" will maybe trouble you a bit but just let it go. That's for people interested in the whole cash-management-consequences-of-central-clearing thing. That's not your thing.
Your thing is: can interest rate swaps or CDS bring down the financial system? And, like, I dunno ... not really? Kind of? To me the most useful proxy is the high-vol CDS bars, reflecting how much of big dealer assets would poof in a five-day run of 1-in-200 credit-market badness. The answer is about $100bn total, or about 1% of assets of the various dealers. That's well within the buffer of bank capital requirements; in fact it's well within the G-SIFI surcharge, which seems sensible: the biggest interconnectedest derivatives-dealingest banks have are adding enough more than twice as much capital as they'd need to cover their worst-case derivatives losses. Except. Y'know. A one-in-200 loss comes, oh, about once every 200 days.
* Here's a quick check. "Dealer 9" is JPMorgan. JPMorgan's reported average daily 95% VaR in its fixed income business was $56mm in 4Q2011 and has been in that ballpark for a couple of years; call its 99.5% daily VaR like $80mm. Here's the paper's calculation of 99.5% one-day variation margin (defined as equal to the market value change, so should be apples-to-apples to VaR) for interest rate swaps (only - not CDS, not cash bonds, nothing) versus historical periods of low, medium and high volatility:
So ... in low-vol circumstances, one billion dollars then. That is more than $80mm. Obvs some of the risk on these swaps can be offset elsewhere in the portfolio so that JPM's overall fixed-income market risk could be less than the risk solely on its interest rate swap portfolio but ... I'm overall left with the impression that one of these numbers is wrong by, y'know, a lot. You can guess at which but remember that the BIS paper is a series of guesses layered on top of each other, while JPM's number is at least in theory developed by people who actually saw JPM's portfolio.