Sunday, November 6, 2011

Known Unknowns

René Magritte, The Son of Man, 1964
[Edward Ferrars and Elinor Dashwood are baiting Margaret Dashwood, who is hiding]

Edward: “Oh... Miss Dashwood. Forgive me. Do you by any chance have such a thing as a reliable atlas?”
Elinor: “I believe so.”
Edward: “Excellent. I wish to check the position of the Nile. My sister tells me it is in South America.”
Margaret: [out of sight; laughs]
Elinor: “Oh. No. No, um... she’s quite wrong. Um... for I believe it is in Belgium.”
Edward: “Belgium? Surely not. I... I think you must be thinking of the Volga.”
Margaret: [still out of sight; appalled] “The Volga?!”
Elinor: “Of course, the Volga. Which, as you know, starts in...”
Edward: “Vladivostock, and ends in...”
Elinor: “Wimbledon.”
Edward: “Precisely. Where the coffee beans come from.”
Margaret: [revealing herself] “Ah! The source of the Nile is in Abyssinia!”
Edward: “Is it? How interesting.”

— Sense and Sensibility

It is a heartening feature of the internet that one can often determine the truth about a subject by asking for the input and advice of experts, who upon application will usually contribute their knowledge freely. In my experience, it is an even more effective method to adopt a strong and firmly argued position upon a topic you know very little about. This will flush out even more experts, who will dismantle your faulty reasoning and expose your flimsy command of the facts with fierce glee or kind patience, depending on how charitably they view your ignorance and presumption.

My recent post on the current state of counterparty credit risk in the global financial system has already elicited two excellent reponses, and I am reliably assured that more are coming. The first of these was submitted to me by email, by a mysterious personage (let us call him or her “X”) who appears to be even more skittish about his or her real identity than Yours Truly, which is saying something. X’s first messages to me assumed a higher level of knowledge on my part than I possess, and X declined to let me disseminate his or her thoughts for reasons of security. Fortunately, after X read my babbling here and discovered exactly how ignorant I am about the day-to-day finance and operations of large trading banks, he or she took pity on me and sent new material fit for publication.

I now quote M/Mme/Mlle X at length, for your education as well:

Let’s look at the example of Bank A hedging some exposure by trading with Bank B. Let’s say

(1) Bank A has bought $100mm of CDS from Bank B,
(2) The CDS is currently worth 65 points (i.e. the $100mm notional contract is worth $65mm),
(3) Bank B has posted $60mm of collateral to Bank A.

What is Bank A’s direct exposure to Bank B? I would argue that the correct number is $5mm. If Bank B were to default and have 0 recovery, Bank A would post an immediate loss of $5mm, since Bank A already has the $60mm in collateral.

The point is that direct counterparty risk only exists on the uncollateralized portion of any exposure. One term for this is “gap risk.” This is relevant because in your example, Bank A would not try to hedge out its exposure to Bank B by buying protection on Bank B from Bank C. Almost all of Bank A’s exposure to Bank B is already covered by collateral. As for the remaining part, generally the amount of uncollateralized exposure that Bank A has to Bank B is not correlated to Bank B’s credit rating, especially if there are a large number of trades in multiple asset classes between the two banks. Bank A can’t know a priori what the uncollateralized amount will be if Bank B defaults; it’s just as likely that the CDS in the above example has moved from 60 points to 55 points and Bank A actually owes Bank B collateral. Also note that since this is essentially portfolio risk, doubling the number of trades with Bank B doesn’t actually double the exposure, especially if (as is common) many of the new trades are offsetting in risk. There’s no gross buildup of residual risk; this just boils down to net risk against the counterparty.

Why was AIG different? The above is a fairly accurate stylized approximation of what happens for relatively liquid CDS (which do increasingly go through central clearinghouses anyway). Something like a corporate or sovereign CDS is a distinct product that trades and has an observable market price. In the AIG case, most of AIG’s CDS exposure came from much more bespoke deals on structured products. A typical AIG CDS contract might be on some particular complex mortgage product, for which the only CDS trade was the one in which AIG wrote the protection. It has no observable market price and has to be priced using model assumptions on the underlying. This contrasts with e.g. sovereign CDS, where a price can be observed in the market and multiple trades happen on the same CDS; i.e. where there does in fact exist an observable market price.

Why is this relevant? In the above example, we assume that banks A and B agree on the contract’s valuation. If instead Bank A believes the contract is worth $65mm but Bank B only believes the contract is worth $30mm and has only posted that much collateral, then Bank A has $35mm of exposure to Bank B, which it will need to hedge accordingly. But the point is that this is a valuation issue; if the two banks actually agreed on the value of the contract, but Bank B simply refused to post collateral, then Bank B would be defaulting outright on its obligations, and would have its positions closed out accordingly, rather than have the counterparty risk just continue to exist.

The above discusses direct counterparty exposure in the sense of “losing money if my counterparty defaults.” There is of course further risk; if Bank B defaults, Bank A is left with that $100mm of risk that it previously didn’t have. But the risk here is actually a function of Bank B’s net exposure, not Bank A’s gross exposure. If, for example, Bank B had an offsetting contract for $90mm notional with Bank C, then after a default by Bank B, you would expect that Bank A and Bank C would offset their newly acquired risk against each other, such that e.g. Bank A only ends up with a $10mm change in risk, and Bank C ends up with no change in risk. This is pretty much what happened after Lehman defaulted. In fact there was a special trading session arranged for just that purpose, though most of the risk rebalancing actually happened in normal trading after the default.

I believe points (2) and (3) in your blog post boil down to concerns regarding net risk. I agree that large concentrations of net exposure would be a cause for concern, more so in illiquid positions but even to some extent in liquid ones. One way to get more comfortable with this in CDS space is just to look at the DTCC net notional numbers. By definition no entity’s net position can exceed the total net position. This ends up giving you a cap on how bad things can be; of course not ideal, but maybe less bad than you would initially think.

* *
One more thing—and you can share this too as long as it’s not attributed.

The “margin call contagion” scenario you propose is not representative of how banks operate. Just about everything in a bank’s portfolio will already be contributing to its funding. Bonds will be repoed out (i.e. for cash equal to the bond’s value, less a haircut), stock will be lent out, and collateral posted on derivative contracts will be rehypothecated.

It’s possible that e.g. repo haircuts will exceed the bid-offer on some instruments and selling a security might give me slightly more cash than repoing it, but the extra amount is small. In general the notion of “liquidating a valuable position for cash” doesn’t make sense for a bank. Of course this may be different for a buy-side firm, but it doesn’t make sense for a bank to sell a security for liquidity purposes when it’s already used to secure some cash. This is also less true for illiquid things that can’t be financed; it is however true for any collateralized derivative position due to rehypothecation.

Alles klar?

* * *

So, at the risk of having my mysterious interlocutor correct me once again, I will take the liberty of drawing a few conclusions.

First, I think X has substantially diminished my fears about investment banks being piles of counterparty credit kindling just one counterparty default away from causing massive systemic conflagration. The daily zero-limit, two-way settlement of collateral calls between large investment banks (now current practice among most large market participants, according to this BIS study) means that, except in very fast moving markets, one should expect that changes in net margin requirements triggered by changes in the value of underlying investment contracts should be reasonably well-reflected in the risk books of most major banks. Second, the fact that big trading banks settle margin exposure on a net portfolio basis—which, Harry Markowitz assures me, should net out to less than the simple addition of each individual exposure across large, multi-market and multi-instrument portfolios—gives me some comfort that whatever residual risks accumulate on bank balance sheets should not be extreme. Third, X’s assurance that investment banks prefer to use asset positions to fund their operations rather than sell them for cash in a market meltdown leads me to discount the risk of cross-market contagion and “death spirals” triggered by collapses in unrelated markets. All these points directly address the second concern I cited in my previous post.

However, if my concerns about the risk of market collapse inherent in the structure and operations of large trading banks have been partially assuaged, I remain less confident about systemic risk in general. In particular, I worry more about investment banks’ exposure to substantial net risks created by large hedge funds, other originating banks (e.g., Dexia), and non-bank participants (e.g., AIG Financial Products). If one is to believe X, this is where the major risks are created and packaged. If Bank A trades with Hedge Fund 1, which cannot meet its financial obligations and has no counterparty assets of its own to net against A, Bank A could still be seriously fucked if Hedge Fund 1 defaults. Especially if Hedge Fund 1’s default coincides, as it well might, with a substantial gapping out of risk exposure on the underlying trade with Bank A.

Assume, as I would certainly hope we can in today’s markets, that most investment banks aspire to pretty close to zero-net present value risk books across the firm. (This is the fundamental philosophical tenet of the Volcker Rule.) Then risks to the system will be created not by the banks at the center of the markets, but rather by the risk-takers (investors, hedge funds, etc.) at the edges. And, should you need reminding, risk-takers don’t hedge all their risks to zero. Duh.

I also worry that investment banks remain seriously exposed in illiquid, hard-to-value markets now and in the future. X him- or herself hints strongly that the nifty daisy chain of traditional bank risk mitigation can get dangerously frayed under such circumstances. Sovereign CDSs may be relatively transparent, given the monitoring and data publication of the DTCC, but this is not true in every market. In particular, I worry that the next dangerous net risk exposure will be created in one of those opaque, highly-illiquid, obscenely profitable new markets which Wall Street is so fond of creating. And if there is no central repositary of trade data in a particular security or derivative market, no standardization and reporting of net and gross positions, what is to prevent the rise of yet another bunch of idiots like AIGFP to create a huge net risk position of which their multiple, competing investment bank counterparties remain blissfully unaware?

Last, I retain a nagging worry about the sheer complexity of the balance sheets, risk books, and insanely complicated credit and financing plumbing upon which modern day investment banks rely. Long-time Readers will know I am no fan of complexity, because it introduces fragility and vulnerability into any system. X and his peers may have designed a beautifully functional risk transmission system for their employers, but what happens if one of the pipes clogs or breaks, due to human error or unforseen complications? (How likely are those, I ask you? Yeah.) I have every faith that the clever gnomes of Wall Street can figure almost anything out, if you give them enough time. The problem is, that when the shit hits the fan at an investment bank, your clients are sucking funds out at a blistering pace, and the ratings agencies and your shareholders are in a desperate race to write you off forever, you have almost no time at all.

© 2011 The Epicurean Dealmaker. All rights reserved.