Thursday, December 11, 2008

CDS - To Be or Not To Be

John Dizard throws in the towel on credit default swaps (CDS). Felix Salmon is not ready to yet. Arnold Kling sees a vindication of a negative (prosecutorial) view.

Here's 2-cents. (This is a long post, but not complex):

Dizard weighs in on three points, "capital raising", "price discovery", and "risk mangement tool":

1. Capital Raising

The 'response' to Dizard's points is in three parts, complicated by the fact that CDS technology has been used in a variety of ways, so broad-brush is no good.

1(a) Balance-sheet management, credit intermediation, structured credits

'Capital raising' may mean the use of CDS in BISTRO-like products and synthetic products that transferred risk and may or may not have freed up risk-capital.

To the best I understand it, the writers of protection in this "structured market" are not the ones who are putting up collateral, now. The buyers put in upfront collateral, as best I know (and what I know about that is very limited).

There may have been additional trading of CDS outside of these structured products, yet related to them, but that is analytically separate - this is just about the products themselves.

It's not at all clear that the goal of credit inter mediation and build-to-suit credit is not worthy. They seem like fairly natural innovations, a step in the conceptual the evolution of risk-sharing and risk-management technologies. Before rejecting an innovation, I'd like to be certain that the risks were either practically or theoretically unmanageable. If there is evidence one way or the other in the current carnage, it needs to be sorted out thoroughly, not summarily.

1(b) Sharing a 'AAA' credit rating via CDS - a "naked" loan portfolio

If he means 'capital raising' to refer to something else, like AIG's (naked?) "regulatory capital CDS" contracts, then that is a different and interesting point. Should the financial strength behind a promise to pay be "tied up" in increasing (decreasing) amounts of capital, as dictated by an illiquid market? If a firm's exposure were just a loan gone bad, there would be no mark-to-market capital requirement. In that case, only a credit event would cause capital "to be posted", "violently" and all 'at once'. Is 'the system' more or less stable for having a seemless capital requirement all along the way, based on market-based assessments of the likelihood of loss?

To me, part of the answer to that lies in whether the CDS market is exaggerating moves in the cash market. So far, it doesn't look like it is. The two appear to be equally bad in tandem. Also, the illiquid, non-traded cash market does move in discreet intervals to require "collateral". As business conditions worsen, lenders' loan covenants get violated. You do not have to wait for "bankruptcy" for significant credit events (compare also ratings downgrades/upgrades). So, in a sense, there has always been a bit of mark-to-market of outstanding credit risk, in the form of demanding "collateral" or proof of ability to pay. The capital impact of credit loans gone bad also occurs explicitly in the loan-loss reserve. If a bank has bad loans, it would be required to estimate the likelihood of those loans being repaid. It takes a non-cash hit to capital, to the extent that the outlook worsens. Is that outlook better estimated by company managements than by CDS market prices?

Do market prices move ahead and faster than "business conditions"? Probably. Does that mean CDS are defunct? That's unclear. You'd have to posit that CDS market prices overshoot and have a (negative) business conditions feedback, more than cash bond/note prices do, right?

Is it as bad as Jesse Eissinger reports, with Bill Demchak agreed (see very end of article), that we have 24- year olds determining the price of things, introducing ridiculous levels of balance-sheet volatility? Maybe, maybe not. I'm not convinced yet which state of affairs is better/worse: illiquid unknowns (opaquely fragile balance sheets) or illiquid knowns (transparently fragile balance sheets - "frozen" in Dizard's terminology).

Last, in a broader view, you can argue that the capital requirements related to the credit risk inherent in a long CDS are the same or less than a loan. If I don't miss my guess, these credit exposures probably graph about the same, over time, as does a vanilla interest rate swap. The fact that the market for CDS might suddenly become illiquid means that VAR is not the only measure, but that notional limits are still a good idea, too.

1(c) Short-term trading of a "view" on credit

If capital raising means just the collateral posted by hedge-funds and dealers swapping views on the direction of this or that credit, then consider this: those risk-transfers are a zero-sum game. If collateral is not also a zero-sum game, then perhaps that needs to be explored, as Felix says. Also, if this is truly short-term trading, then one suspects that these trades unwind/terminate, even at abusive, illiquid prices, rather than being the cause of a sustained, significant, non-economic build-up in collateral.

2. Price Discovery

Dizard says:

But CDS are derivative instruments, whose price is "discovered" these days as a function of equity volatility, since buying equity puts is one way to dynamically hedge the illiquid legacy books.

So CDS dealer sales of Citigroup equity through derivatives means higher equity volatility, then higher CDS spreads, leading to more margin calls, leading to more sales of bank stocks . . . This has become a system-wide tail-swallowing exercise in lunacy
The second part first. This seems to be a variation of the age-old theme that derivatives increase the volatility of the spot/underlying markets. In this case, the author suggests more than that, that there is also a destructive feedback loop, an inherent instability created.

First, I may be completely wrong, but I don't think you dynamically hedge (long) CDS using equity puts, in the normal sense of a dynamic, parametric hedge. Buying a put on the equity is a (costly) and sufficient hedge. As the probability of default goes to one, the put value goes up. You are covered. If someone is buying and selling partial hedges as part of an overall strategy, that seems to require another name.

The market maker who bought/sold the equity put might need to delta-hedge in the equity market. To the extent that they are using a market-maker exception to short naked, that's a potential problem. To the extent that non-market makers are restricted to legal shorts, it's not an obvious problem.

Apart from an interesting and important consideration, the sum of this problem seems to be that "price discovery" at or around high default likelihood is problematic. D'oh! The notion that equity volatility is high near times of default is ... not particularly alarming, either. It doesn't have to be high default likelihood, either. It could just be conditions in which no one wants to bear the risk of credit or too few the financing/capacity to take advantage of the price-opportunities created by peak risk-aversion or market illiquidity.

3. Risk Management Tool

Dizard notes:
A credit default swap is a very different creature than the traded equity of a "reference entity". CDS cover only one on-off risk, that is, default on reference obligations. So in the airless world of ideal models, CDS values are better considered as binomial probability distributions, rather than as the continuous probability distributions that are closer approximations of equity values.
One could also conjecture that what we are seeing is a dual-pricing regime for CDS, not a bi-modal distribution. One type of pricing for high probability risk of default, when "replication protection" is paramount (for hedging), and another for moderate to low default risk, when replication/no-arbitrage is less important.

Were this a wild ride, you'd expect the CDS spreads to be really, really wide compared to the cash bond spreads, right? At least I would, as all the "weird" theoretical assumptions got violated and knocked risk-models around, etc. Well, so far as I can tell, the CDS are trading at or near the cash bond prices (sometimes even inside the cash spreads).

The reg-cap arbitrage problem at AIG, in my view, could be solved with notional limits, at a minimum, rather than relying on VaR, or sophisticated peak credit-risk calculations. This reflects the fact that a seller of protection (long the credit risk) could end up with a de facto loan. You need capital to cover that "worst case" scenario. Of course, the caliber of the top leadership at AIG at the time appears to be such that that was unknowable...

I confess I do not understand this paragraph, so I cannot comment, although I feel I should be able to, eventually. First pass, it looks like this is a way of estimating some of the parameters of the pricing models in use, and I don't understand its applicability:

When implied probability of default, and equity volatility, are relatively low, you can do some seemingly plausible regression analyses to fit the series. But at high levels of default risk and equity volatility, if you hedge the one with the other you get frantic, self-defeating activity.

The point about whether an on-off contract is the best way to trade continuous views about credit spreads is one worth thinking through. Why doesn't the market just trade put/call options, with various strikes, on a variety of credit spreads, for instance? What's the particular allure of the CDS construct?


For those like Arnold Kling looking for a fuller theoretical understanding of CDS, here is Hull & White, available online (not to difficult for those who already have an understanding of options pricing theory):



Here is the math for backing implied default probabilities for simple contracts.

No comments: