Deus Ex Macchiato

Matt Levine has an excellent dealbreaker post which in turn references a Bloomberg story on Lehman’s derivatives. The facts first:

Almost five years after Lehman Brothers filed for bankruptcy and set off the global financial crisis, managers of the bank’s estate are demanding millions of dollars from retirement homes, colleges and hospitals… [For instance] The Buck Institute for Research on Aging in Novato, California, gave Lehman $2 million in October 2008 to cancel a swap contract used to manage fluctuating interest rates. Lehman [now] says it wants $12.1 million more and has assessed at least an additional $4.7 million in interest, the research center said in its most recent financial statement.

There are at least two important issues here. First, as Levine points out, when you closed your swap out with Lehman matters hugely for valuation. (I have cropped the market data he gives to show the USD 5y swap rate in the period after the default: look at that volatility in late 2008.) The CSA you had with Lehman matters too, as does whether you use market valuation or actual close-out (which in turn depends on the details of your master agreement with them), what CSA you had with the party you closed out with, and so on. Moreover, the naive idea that your claim against Lehman is the price you closed out isn’t necessarily true. Levine quotes from the Federal Home Loan Bank of Cincinnati’s 10-K:

We had 87 derivative transactions (interest rate swaps) outstanding with a subsidiary of Lehman Brothers, Lehman Brothers Special Financing, Inc. (“LBSF”), with a total notional principal amount of $5.7 billion. Under the provisions of our master agreement, all of these swaps automatically terminated immediately prior to the bankruptcy filing by Lehman Brothers. The terminations required us to pay LBSF a net fee of $189 million, which represented the swaps’ total estimated market value at the close of business on Friday, September 12… … On Tuesday, September 16, we replaced these swaps with new swaps transacted with other counterparties. The new swaps had the same terms and conditions as the terminated LBSF swaps. The counterparties to the new swaps paid us a net fee of $232 million to enter into these transactions based on the estimated market values at the time we replaced the swaps.

Now, that difference in value could have been a market risk gain based on a period of open exposure in very volatile markets. But it could also be partly a mismatch between the close-out amount the Home Loan Bank paid Lehman and the real market price. You can see how a lawyer might think that there is a case there, especially one paid to maximise the value of Lehman’s estate (for the benefit, let’s remember, of other creditors).

I am not sure how to react to this. The knee-jerk response is to demand that the close-out process is defined so as to lead to less disputable results, but doing that is not straightforward. What is applicable for the (unusual) Lehman-like events probably isn’t appropriate for much smaller (and more usual) close outs. Moreover, any claim on a bankrupt must, ultimately, be subject to scrutiny by the bankruptcy courts, and must adhere to underlying legal principles (like anti-deprivation). So the right policy here is not obvious. But certainly the risk of closing out then, years later, having that process challenged by the defaulter’s estate with the potential for large amounts of interest being assessed as well as the original claim, is material. Whether there is anything that can be done about it is less clear.

Posted in: Derivatives, Hedging and Convexity, Legal Risk and Trade Documentation by David /
1 Comment

It seems that JPMorgan’s travails have become a spectator sport, to be enjoyed with a snack of your choice. I am only half way through the senate report, let along the appendices, so I won’t add to the (already comprehensive) guides to the action‡.

Instead I want to focus on four key issues which emerge from this debacle.

1. CIO wasn’t hedging. Like Matt Levine, I had bought the firm’s line that the original portfolio was a macro hedge against the loan book. It is now clear that while that might, in the mists of time, have been the original motivation, the CIO’s office had turned into a prop trading center by 2012. This happened without, as far as I can tell, any authorisation, any redesign of the risk framework, or any changes in oversight. Mind you, given that the risk framework was not based on how well they were hedging anyway, that is hardly a surprise.
   The Machiavellian analysis of this is that they were trying to prop trade while avoiding Volcker. My gut feeling is that it wasn’t that: they were simply out of control.
2. As Lisa says, mis-marking is key here. The practice whereby, in complete violation of what the accounting standard actually says, US banks are permitted to mark derivatives anywhere between bid and offer must now receive attention. Supervisors must ensure that firms mark at where they can exit the position as it is absolutely clear that external auditors cannot be relied upon to police valuation practice.
3. My earlier conjecture that capital management was central to the whale losses is born out. But it is worse than I thought: capital optimisation was mostly about changing the model so that it generated lower numbers. This was wholly cynical, and is bound to increase the pressure to reduce the capital benefit available from the use of internal models§.
4. While JP undoubtedly kept things from the OCC, the OCC’s process allowed JP to make model changes without sufficient oversight, did not exercise control over valuation practices, and had little idea what was going on in the CIO. After all, the story was broken by journalists based on public information.
   While all the focus so far has been on JP’s mis-deeds, JP’s supervisors do not emerge from this covered in roses.

It will be interesting to see if the Senate can keep up the (encouragingly bipartisan) momentum here. One is uncomfortably aware that a confrontation may be brewing with politicians and public on one side, and the big banks, the OCC, and perhaps the FED on the other. If it really does pan out that way, the legitimacy of current regulatory arrangements may not survive the fall-out.

*The reference is to an extraordinary radio interview that Paul Roy, ex-head of equities at Merrill, gave about the old days on the London Stock Exchange, in which he claimed his equity traders used to enjoy a glass of madeira, or perhaps champagne, as a mid-afternoon pick-me-up. O Tempora, O Mores.

§See the recent speech from Stefan Ingves here. Ingves says that “Major [Basel Committee] projects currently under way include: … completing the review of the trading book capital requirements. This entails an evaluation of the design of the market risk regulatory regime as well as weaknesses in risk measurement under the framework’s internal models based and standardised approaches.”

Posted in: Credit, Derivatives, Hedging and Convexity, Securitisation and Tranching by David /
9 Comments

A timely reminder from Matt Levine:

nobody owns stock, they just own interests in their brokers’ interests in DTCC’s interest in stock. “Oh I own AAPL shares,” you say, but you don’t; you own like a second derivative on Apple shares. A delta-one derivative but still.

Posted in: Broker/dealers, Derivatives, Hedging and Convexity, Markets by David /
3 Comments

It is perhaps to my discredit that I don’t have the interest, stamina, or bloody mindedness to go through Taleb’s technical papers (HT FT Alphaville*). Still, a glance at them and hefty dose of imagination quickly gave me what I think he might mean in the first few pages, at least from 50,000 feet.

The first thing to note is that options have vega: they are sensitive to changes in implied volatility. For our purposes vega can be thought of as a measure of sensitivity to being wrong about the asset return distribution. Naively it’s `I thought it was log normal with a 30% volatility, and it turned out to be log-normal with a 31% volatility – how much did I screw up?’

The second point is that vega is highly strike dependent. A 1% change in vol for a five year ATM option might make only a 2% difference in option value: it makes nearly an 8% difference for a 200% strike on the same underlying. This is (I think) what Taleb calls tail vega. It is simply the observation that being wrong about the distribution matters more for out of the money options. Pricing far OTM options is fragile: one knock of the distribution and you are screwed. Far OTM puts are particularly bad because vols rise as the markets crash, and you get crushed by both the gamma and the vega.

In contrast there are some products that are not very sensitive to being wrong about the distribution. What a savvy tail risk hedger tries to do is either buy cheap protection – i.e. find implied vols that are ‘too low’ – or find a product which is relatively vol insensitive yet is still long the downside.

The hard part is that ideally you want to do this analysis not with respect to a mis-parameterised distribution – it’s log normal but we don’t know the vol – but rather with respect to some class of distributions. What class do you pick, though? Too general, and you can’t prove anything, plus there are lots of distributions that really don’t occur in nature that you are trying to talk about; too narrow, and you risk missing the `real’** one. So the whole game in work like this is figuring out what class is general enough to capture some interesting fat tailed distributions, yet narrow enough you can prove something***. This is grandiosely termed meta-probability, but all it really is is hunting for a tractable class of distributions to do the meta-theory on.

*You would get a precise link except my session on the FT website has timed out and I can’t be bothered to login again. Honestly you would think the FT is trying to reduce readership…

***I am pretty sure I can come up with return `distributions’ (under a sufficiently loose construction of the term) that break all of Taleb’s theorems, but they would be insanely pathological – something Lebesgue integrable but not Borel measurable would be a good start. What, you wanted something continuous?

Posted in: Derivatives, Hedging and Convexity by David /
11 Comments

From the Wheatley review final report:

5.9 Therefore, the Review recommends that the number of currencies and tenors for which LIBOR is published be reduced. Specifically:

• publication of all LIBORs for Australian Dollars, Canadian Dollars, Danish Kroner, New Zealand Dollars and Swedish Kronor should be discontinued;
• for remaining currencies, publication of LIBOR for 4 months, 5 months, 7 months, 8 months, 10 months and 11 months tenors should be discontinued;
• continued publication of overnight, 1 week, 2 weeks, 2 months and 9 months should also be re-considered.

The maturity restrictions are probably not a big deal but the currency ones are. As IFR says, this could trigger years of lawsuits if the floating rate for trillions of dollars of swaps disappears.

Posted in: Derivatives, Hedging and Convexity, Legal Risk and Trade Documentation by David /
1 Comment

From the Wheatley review (HT Lisa Pollack at FT alphaville), three illustrations.

First, how many contracts use Libor?

Quite a lot then, especially swaps.

Second, which Libors do all those swaps use?

3 and 6 month. OK.

Third, which Libors actually trade?

Ah. So 3m kinda trades and 6m doesn’t trade. Still, what could do wrong with a multi-trillion-dollar business linked to a fictional interest rate?

Posted in: Derivatives, Hedging and Convexity, Markets by David /
Comments Off

There’s a problem in corporate swaps. It’s this.

• Basel 3 has a capital charge for CVA risk.
• This charge increases the price that some corporates will have to pay for their swaps, given their current credit support arrangement.
• The EU may or may not grant an exemption from this charge for many corporates. It’s a political issue, and impossible to call.
• So given you don’t know if you will have to pay the charge or not, do you price it in?

Risk magazine amusingly tells us:

Three banks that spoke to Risk for this article all claimed to be assuming there would be no exemption. They also said rival dealers are doing the opposite.

You might say ‘of course they would say that’. But there is a problem here, and it isn’t going to be resolved any time soon.

Posted in: Clearing and Collateral, Credit, Derivatives, Hedging and Convexity by David /
Comments Off

It is relatively simple to see the liquidity of a bond. You see how many times a day (or week, or year) it trades.

A commenter on a prior post suggested that the number and spread of these quotes is also a useful indicator of liquidity. I’d agree, with the caveat that a quote is not necessarily good in the size you have, nor is it necessarily a firm commitment to trade.

For OTC derivatives markets, though, things become murkier. This is because most derivatives have a maturity, and as time passes, that gets shorter. Your on-the-run five year swap today becomes a four year 51 week swap next week, and that isn’t liquid.

The problem is typically solved with a model. We build yeild curves, credit spread curves, vol surfaces and so on. The liquid instruments define points on these curves which the model is calibrated to; everything else follows by interpolation. (Or, if you want to take significantly more risk, extrapolation.) A model in this setting is just a fancy interpolator.

The advantage of this system is that it allows market participants – mostly – to price things that don’t trade. You can get a quote on your four year 51 week swap despite the fact that that particular instrument won’t trade this week precisely because its price is in a reasonably clear relationship to that of the five year swap that does trade.

Problem start to arise when the benchmarks themselves become illiquid or otherwise doubtful. The whole system relies on the benchmarks being liquid, so that the are known prices to interpolate between.

This brings us to Libor. If banks don’t lend to each other for three or six months, then 3m and 6m Libor are conjectural. That means that the benchmark swaps (and the Eurodollar futures) are themselves uncertain in value, and liquidity in them might start to decline. At that point you won’t be able to price any interest rate derivative.

Now, I don’t claim that this will happen. But without a floating rate that the market has confidence in, there has to be some risk of it. For me, this is a far more important financial stability issue than Mr. Diamond’s employment status.

Posted in: Derivatives, Hedging and Convexity, Liquidity and liquidity risk by David /
3 Comments

Good questions from Dealbreaker, apropos the ongoing Libor issues:

If Libor isn’t just a trimmed average of some numbers that some banks tell someone from Reuters every day, then it is … the risk-free rate? The unsecured borrowing rate for AA banks? The unsecured borrowing rate for an actual assortment of disparately rated, often barely investment grade, rather tarnished banks that mostly don’t actually lend to each other?

Don’t worry, though, there are only a few hundred trillion dollars worth of contracts linked to Libor. Bob is handing back his bonus, and that is an act that one can applaud. But there are more serious issues here than an attempt to move the rate – like what happens to all those swaps that reference 3 or 6 month Libor in a world where there is little or no 3 or 6 month unsecured borrowing by banks.

Posted in: Derivatives, Hedging and Convexity, Legal Risk and Trade Documentation by David /
2 Comments

From MacKenzie and Spears, apropos JPMorgan:

a ‘hedge’ is not a self-evident feature of the world, but a contestable cultural category.

Yes, Alphaville got there first, but honestly the whole paper really is worth reading. It very much backs up my suggestion a few days ago that it is not models-as-hedge-parameter-generators that were the problem, but rather models-as-indicators-of-absolute-value.

There is also an interesting discussion of what the authors call counter-performativity, and what I would call a crowded trade:

there are multiple mechanisms of counterperformativity, in other words multiple ways in which the practical use of a model can undermine its empirical adequacy… no-arbitrage models may be associated with a distinctive mechanism of counterperformativity, in which the hedging practices those models demand have effects on the market for the underlying assets that undermine the empirical adequacy of the views of asset-price dynamics embedded in those models.

In other words, if many people (or one very large player) is hedging the same position using a model, then the relationship between the derivative and the underlying can breakdown quite dramatically. My first exposure to this was the impact of the LTCM crisis’ on equity derivatives: many players were short long-dated equity vol into retail products, as were LTCM. A big bank, for reasons I won’t go in to, had to close out a big short vega position, and the market mostly knew it. Thus FTSE 5 year ATM vol went from the teens to the forties, and stayed there until the forced close out happened.

The model doesn’t say ‘this works if not everyone is the same way around’ but you would have to be a poor trader or risk manager not to know it…

Posted in: Derivatives, Hedging and Convexity by David /
1 Comment

So how would this pay for resolution by writing calls thing work?

Well, first you need to calibrate the system. Fortunately Joe Noss and Rhiannon Sowerbutts (HT FT alphaville) have recently analysed the implicit UK taxpayer subsidy to banks, and concluded that it is at least £30 billion per year across the cycle. As a starting point, let’s assume that we want half of that money back.

That means UK banks have to write £15 billion of physically settled one year at the money equity call options on their own stock to the Bank of England every year. You could determine individual bank amounts based on capital (or something else like balance sheet size); so for instance Bank H might have to write each year enough call options to be worth

£15 billion x Bank H RWAs / (sum of all Bank RWAs)

These would be priced using market implied volatilities, and hedged as usual by the central bank shorting bank H equity.

This process would lock in the value of those options, assuming that the hedge was successful – and frankly it isn’t too difficult to hedge one year plain vanilla call options in reasonable size: these are not funky credit derivatives. Big market moves would make the central bank money as it is long gamma and long vega; quiet markets, where all was well in the banking system, would result in the central bank failing to capture some of the calls’ value. That’s OK, though; you don’t mind having less in the pot when times are good providing the pot magically refills in bad times, as it will.

Now let’s roughly size the suggestion. HSBC has roughly 40% of UK bank RWAs, so it should pay 40% of the total. That is, each year, it should write £6 billion worth of calls to the Bank in this model, an amount corresponding to roughly 30% of its earnings. One at-the-money call on HSBC is worth very roughly 60p (I am rounding like crazy as this is all very approximate), so this year it would have had to have written calls on 10B shares. Ooops. That’s half the shares outstanding. There’s no way you can hedge a position that big efficiently, and market knowledge of dilution that large would kill the share price.

What do we conclude from this? Well clearly this idea can’t monetise all of the implicit taxpayer subsidy. But it certainly could monetise 10% of it. £3 billion a year in the pot to cover the costs of RBS/LBG/Northern Rock type events would be a good start.

Posted in: Derivatives, Hedging and Convexity, Regulation by David /
Comments Off

Acharya, Mehran, Schuermann and Thakor have an interesting but ultimately flawed idea. They suggest:

a special capital account in addition to a core capital requirement. The special account would accrue to a bank’s shareholders as long as the bank is solvent, but would pass to the bank’s regulators — rather than its creditors — if the bank fails.

This part is not so bad. It does rather imply that all banks will be rescued/resolved rather than giving supervisors the option to go through bankruptcy, but that is perhaps the new reality anyway. The problem comes in how they suggest that the level of the account is set:

the quantifcation of the capital requirement need not depend exclusively on the use of historical data for calibration of the bank’s risks; instead, it would rely on several different approaches, such as market-based signals of bank-level and systemic risk as well as regulatory intelligence gathered through periodic stress tests of the fnancial sector.

This is reasonable from a financial stability perspective, but less so from a capital planning one. Banks need some certainty about their capital requirements whether direct or in a new capital account. Moreover ‘market-based signals’ risk being hugely procyclical.

A better approach would be to force banks to issue a fixed fraction of their earnings to the central bank as equity call options. This would have several big advantages:

• It is anti-cyclical at both the systemic and the individual bank level;
• The supervisor could accrue a cash by hedging these calls. This hedging activity would also be anticyclical, shorting bank equity as equity prices go up, and buying back the short when prices fall.
• Those banks who profited most from the financial system would contribute the most.
• Banks would have a clear idea of the size of the buffer, facilitating capital planning.
• Supervisors would make more money hedging the calls just when it is most needed, that is in times of increased bank equity volatility.

Now of course the idea of a central bank having an equity derivatives hedging activity is new and perhaps radical. But radical ideas are not always wrong.

Posted in: Bank Run, Derivatives, Hedging and Convexity, Regulation by David /
2 Comments

Hedges, pace JP Morgan, are meant to reduce risk. There are, it seems to me, two important ways to hedge. Although real hedges overlap between these categories, the distinction is useful.

First, there are P/L volatility hedges. Here the idea is that risk is earnings volatility, and that you have some position which generates earnings volatility, which you want to remove. At its extreme, you may be happy with the long run risk of the position, but the ride is too bumpy in the short term. Hence you put on a hedge which is likely to move from day-to-day in an equal and opposite way to the position. Buying protection on the CDX against owning a diverse portfolio of corporate bonds would probably count as a P/L volatility hedge.

Second, there are tail hedges. Here something bad, while perhaps unlikely, could happen, and you want something that will pay off to offset your losses in that bad situation. Thus for instance you might buy an out-of-the-money put against owning an equity to hedge against sudden bad news.

Now here’s the interesting thing. Often, the type-two hedges are more effective at protecting the firm when it matters, but type-one hedges can look better. Management scrutinise the P/L every day, but (in many firms, it seems) they don’t have detailed knowledge of the positions. The type-one hedges look good as the P/L doesn’t move. Moreover, risk reports that concentrate on small moves would show them as low risk. Hedges like this often fail in stress conditions though – exactly where type-two hedges work.

Hedging the second way is organizationally difficult. It involves educating management why the P/L is volatile from day-to-day, and why the risk reports show substantial risk from small moves (albeit proportionally much less from large ones). It also involves being right about what tail event(s) are possible, and covering those; there is nothing worse than paying for a tail risk hedge then discovering that you have hedged the tail of the wrong risk factor.

If the risk is large enough to get external attention, then the problem is even worse. Trying to explain tail risk hedging to journalists or equity analysts after you have had an earnings miss is difficult, and management don’t want to run the risk of that embarrassing call.

The net result of all of this, I suspect, is that a lot of type-one hedges are put on that are utterly superfluous to the bank, while too few type-two ones are executed.

Posted in: Derivatives, Hedging and Convexity by David /
7 Comments

When Magellan emerged from the strait that bears his name into the Pacific ocean, he thought that he was only a few days sailing from Portugal and home. Good try, but no cigar. A similar navigational issue seems to be plaguing folks over last night’s $2B JPMorgan loss. Here are some things we can, and cannot conclude from this ‘egregious’ loss.

• Buying protection on a CDS index sometimes doesn’t hedge a long bond portfolio very well. (JP is supposed to have invested excess deposits in short dated corporate paper, then hedged the credit risk with index CDS; or perhaps bought tips then constructed a synthetic corporate linker portfolio by selling protection – can someone please get to the bottom of this?)
• Hedging is difficult, but if you have a position that a lot of people have noticed, and written about, then you call it a storm in a teacup, don’t lose a $2B on it. You look silly.
• Defining what a hedge is precisely enough to allow hedging but prohibit proprietary trading, as the Volcker rule must try to do, is very hard.
• This is not evidence that JPM (or anyone else) is under-capitalized. Sorry, but it isn’t; they absorbed the $2B with little more than mild embarrassment. (Of course, they may be under-capitalized, but you can’t conclude that from this story, however much you might want to.)
• When you have fair value items hedging accrual accounted items, then you get earnings volatility. Your risk measure looks kinda screwy, too, if you include the former but not the latter in it. (See the 10-Q here, pg 74.). The accounting rules probably have not served JPM particularly well here.
• Social media manager for Goldman Sachs is, today at least, not the toughest job in finance.

Update. FT alphaville makes a similar point about the difficulty of identifying a ‘good’ hedge here.

Posted in: Credit, Derivatives, Hedging and Convexity, Reputational Risk, Risk Management, Writedown by David /
Comments Off

Yes, it (often) does. Credit Suisse offers us an example. First, what we know, from CS themselves.

In 1Q12, we entered into the 2011 Partner Asset Facility transaction to hedge the counterparty credit risk of a referenced portfolio of derivatives and their credit spread volatility. The hedge covers approximately USD 12 billion notional amount of expected positive exposure from our counterparties, and is addressed in three layers:

1. first loss (USD 0.5 billion),
2. mezzanine (USD 0.8 billion) and
3. senior (USD 11 billion).

The first loss element is retained by us and actively managed through normal credit procedures. The mezzanine layer was hedged by transferring the risk of default and counterparty credit spread movements to eligible employees in the form of PAF2 awards, as part of their deferred compensation granted in the annual compensation process.

We have purchased protection on the senior layer to hedge against the potential for future counterparty credit spread volatility. This was executed through a CDS, accounted for at fair value, with a third-party entity. We also have a credit support facility with this entity that requires us to provide funding to it in certain circumstances. Under the facility, we may be required to fund payments or costs related to amounts due by the entity under the CDS, and any funded amount may be settled by the assignment of the rights and obligations of the CDS to us. The credit support facility is accounted for on an accrual basis.

Basically, then, three parties own the credit exposure on CS’s OTC derivatives portfolio: the bank themselves, their employees, and a senior hedge provider. Selling the mezz to the employees (in lieu of bonus) is really smart as it gets around all sorts of disclosure and alignment of incentives issues associated with a third party hedge.

What’s left is presumably AAA risk or pretty close. But – and here’s the rub – the hedge provider has written a CDS on it. That’s an OTC derivative. So that generates CVA. Moreover like any senior tranche, while losses on it might be unlikely, there can be serious MTM volatility. So I bet, to keep the CVA down, CS has got its counterparty to agree to daily cash collateral but the counterparty, worried about the liquidity implications of this, has got CS to agree to lend it the money. It’s just a round trip. CS’s loan book lends money to the counterparty, who post it straight back as collateral under the CDS. Look, no CVA on the hedge, and all of our capital requirements on the CVA are gone. Magic, isn’t it?

(HT Dealbreaker.)

Posted in: Clearing and Collateral, Derivatives, Hedging and Convexity, Regulation by David /
6 Comments

When the RMMG (as it then was) issued the CVA capital rules in Basel 3, I said that they would lead to a number of capital arbitrage deals. Street talk was that the Swiss were first off the blocks; now we learn from Euroweek (HT FT Alphaville) of a deal by RBS:

Royal Bank of Scotland is in the market with a highly innovative capital relief trade, dubbed Score 2011-1, securitising a $2bn book of credit counterparty risk.

There are some challenges to getting both default risk and CVA charge capital relief in a securitization structure, but they aren’t insurmountable. I predict 2012 will see a goodly number more such deals.

Posted in: Basel, Credit, Derivatives, Hedging and Convexity by David /
Comments Off

The Libor probe is heating up. Bloomberg reports:

Global regulators have exposed flaws in banks’ internal controls that may have allowed traders to manipulate interest rates around the world, two people with knowledge of the probe said.

Investigators also have received e-mail evidence of potential collusion between firms setting the London interbank offered rate, said the people, who declined to be identified because they weren’t authorized to speak publicly.

Now I have no special knowledge of this situation, and I have no idea whether banks did indeed manipulate Libor. But I do think that the design of Libor is inherently flawed. When they are asked to submit the quotes that are averaged to produce Libor banks are asked

“At what rate could you borrow funds, were you to do so by asking for and then accepting inter-bank offers in a reasonable market size just prior to 11 am?”

Note the hypothetical: ‘could you… were you to…’. In other words, Libor isn’t necessarily the average of rates at which banks did borrow, but rather of rates at which they estimated they could probably borrow. That’s a big difference, and it makes it much harder for a bank to be sure that the numbers it submits to the Libor panel are right. [The issue is that banks don't borrow for longer tenors unsecured very much at all, so something like ten month Libor - or even one year Libor - may well be the average of guesses rather than of actual market rates.]

Obviously here there is a tension between having a lot of rates some of which are less well defined and having a rate that is really market price based. Personally though I would have thought that building a multi-hundred-trillion dollar industry on prices that may be the average of guesses might create some issues, such as the risk of manipulation…

Posted in: Derivatives, Hedging and Convexity, Legal Risk and Trade Documentation, Rules by David /
4 Comments

FT alphaville has a post on derivatives netting, which is mostly reasonable, although it links to a piece by (self-proclaimed?) expert Das, which isn’t.

To begin with, it is important to understand what a properly executed master agreement does. I think of it as glue: it binds up all the contracts between two parties, so that instead of many little contracts, there is one big complicated contract. As a result of this glueing, the parties owe each other whatever the net value of the big contract is. Thus we get two forms of netting: payment netting on everyday cash movements, reducing the number of cashflows between parties; and close out netting, which means that if one of the parties is in contractual default, then only a single net amount is payable.

In jurisdictions where this works (which is most of them – Russia and China being the most prominent examples where it may fail), this means that there is a single claim against the estate of a failed bank (or a single payment to it if the defaulter is in the money on the big contract).

Now, the really delicate thing is how this close out amount is determined. Unsurprisingly, a standard methodology is not imposed as part of the standard master agreement as this agreement has to deal with both bank-to-client relationships, where there are often a small number of derivatives which are easy to value, and bank-to-bank relationships, which may be much more complex. Of course, the vast majority of close-outs are of bank-to-client relationships – and you don’t hear anything about these proceeding without disputes (which they do, all the time).

A big bankruptcy like that of Lehman Brothers generates litigation on pretty much everything. The amounts of money at stake are large enough that it is worth sueing. So people do, on whatever can reasonably be disputed – and often on something things that can’t. Derivatives are part of this, but they are not especially problematic. Indeed, as Kimberly Summe points out, Lehman’s derivatives have received a lot of unnecessary and unwarranted stigma. The unpalateable truth is that it was real estate lending and bonds that broke Lehman, combined with liquidity risk, not swaps.

So far, we have noted that derivatives are not unusual in creating court cases, and that most close outs are simple and effective. But there is an issue that remains: how can it be that reasonable people differ on what the close out amount on a derivatives portfolio is? The answer is that while bankruptcy law usually has a simple idea of what you can claim, determining that amount is not straightforward. Thus for instance in UK law, broadly, if I suffer a loss of £10 because of your bankrupcty, I have a claim of £10 against the estate of the bankrupt. The obvious example is that I have lent the tenner to the bankrupt. But with a derivatives portfolio, what have I lost? Clearly it depends on how much it costs me to close out the risk. I can’t – especially if I want to look good in front of the judge – just use my own valuation: I have to actually go into the market and close out the risk, then add up the cost of doing that. And what I do has to be ‘commercially reasonable’. Thus for instance getting separate bids on the equity, credit and commodity derivatives sub-portfolios might well be commercially reasonable, but doing separate trades on every derivative rather than offering a portfolio of mostly offsetting instruments to the market probably isn’t. (This is a point which Das gets wrong and which lies at the heart of the Nomura vs. Lehman case.)

The problem at the heart of close out, then, is figuring out what value a bank has been deprived of when one of their derivatives counterparties fails. This is often simple, but for a large multi-asset portfolio, it can be both complicated and sufficiently uncertain that it is worth going to court about. The real story isn’t that there is a problem with netting: it is that the valuation of big portfolios of financial instruments is difficult, especially when you have to do it in a crisis.

Posted in: Derivatives, Hedging and Convexity, Fair Value, Legal Risk and Trade Documentation by David /
3 Comments

FT alphaville recently talked to Richard Comotto, author of the International Capital Market Association’s repo survey. Comotto points out, quite properly:

• Investors are increasingly trying to protect themselves by demanding higher haircuts from counterparties, with the focus on the party receiving the collateral and offering the cash.
• This fails to account for the fact that there is still an exposure on the other side of the trade.
• What if the financier (sitting on the collateral) goes bankrupt? The over-collateralisation leaves the counterparty exposed on what is effectively an unsecured basis. They have volunteered collateral which is worth more than the loan, but may now never receive the assets back.
• The trend towards overcollateralisation will put more pressure on the collateral market which is already short of quality collateral.

All true and all interesting. The point about overcollateralization being an unsecured exposure is particularly noteworthy; this is often also true of initial margin at a CCP of course.

The last point leads me nicely to Australia, specifically to the problems they are having as there isn’t enough government debt to act as collateral. In response, the RBA is creating a committed liquidity facility or CLF. The basic idea is to meet banks’ need by allowing them to pledge non-qualifying collateral and get in exchange assets which will work for Basel III liquidity purposes. Eligible instruments include

..domestic issues by supranationals and other foreign governments, ADI [Authorized Deposit-taking Institution] issued debt securities and asset-backed securities, including residential mortgage-backed securities (RMBS).

However for the purposes of the CLF, the RBA will also allow banks to present certain related-party assets such as self-securitised RMBS.

The RBA, then, is meeting banks’ needs for quality collateral, albeit for a fee. Perhaps a more interesting question though is why there is this collateral squeeze. Basel III and clearing together, as we have said several times, are creating a huge demand for high quality collateral, and the systemic implications of this have not yet played out.

Postscript. Are you impressed that I resisted the urge to post a picture of a haircut?

Posted in: Clearing and Collateral, Derivatives, Hedging and Convexity by David /
1 Comment

I am out of town so this will be brief, but there are a lot of good things around today:

• Zoltan Pozsar has a nice paper on VoxEU, ‘Can shadow banking be addressed without the balance sheet of the sovereign?’. One of his main points, which we have made before, is that the demand for safe assets was an important contributor towards the crisis. As he says, Seeing the shadow banking system from the perspective of the safe asset demands of institutional cash investors is crucial for drafting the right policies for shadow banking.
• A lovely piece from FT alphaville on the interaction between CVA hedging and sovereign CDS. Amusingly, they use data from the EBA stress tests to show how big this problem is for European banks.
• Another good piece from Alphaville (these guys are on fire this week) about the impact of the IRC on European sovereign bond prices. They kindly refer back to an old point of mine about the IRC creating pressure to move bonds to the banking book, as ask
  

  All the other factors now driving the crisis speak against holding sovereign debt in banking books as well, increasingly. No?

  They are of course right about that. But if the TB/BB boundary is flexible, in other words you can just move sovereign bonds that you had in the TB into the BB with no impact on strategy and a big saving on capital, why wouldn’t you. A great question in this context is where is sovereign bond repo booked: TB or BB.

• Finally, a Dealbook piece by Jesse Eisinger that quotes me as saying that ‘I’m pretty certain that clearing is being imposed without anyone actually knowing whether it actually reduces counterparty risk or not.’ Just to be clear, what I mean by that specifically is that the problem of whether the multilateral netting benefits of clearing (clearable OTC trades with A and B are now both with the CCP, resulting in risk reduction) outweigh the disadvantage of splitting netting sets (trades with A are split into clearable ones, now with the CCP, and unclearable ones, now still with A) is still open. So far as I know, there is no study of this that makes remotely plausible assumptions about the multiplicity of CCPs for multi-currency, multi-asset portfolios.

Posted in: Basel, Credit, Derivatives, Hedging and Convexity, Liquidity and liquidity risk, Markets, Sovereign Default by David /
Comments Off

FT alphaville gets it:

right now, CVA desks are driving something like a quarter of the demand for sovereign CDS.

My guess is more like a third, but yeah, CVA desks are major players in the sovereign CDS market.

Doing this is in part a regulatory arbitrage for them such that more hedging means a lower capital requirement.

Not quite. It will, once Basel III is implemented, mean a lower capital requirement. And it isn’t regulatory arbitrage in the sense that this CDS does indeed somewhat hedge the position: it just doesn’t completely hedge it (because for instance the CDS might not trigger when you want it to).

To the extent that other market participants think that such CDS are now worth less because of politicians’ attempts at financially engineering around a credit event for Greece, that will cause spreads to move tighter, making the arb cheaper. Go Team Basel III!!

Yeah, the less the market believes that CDS hedge, the cheaper they are, and so the cheaper the CVA hedge is to put on.

We haven’t even mentioned the death spiral whereby sovereign CDS widen out and CVA desks are obliged to buy more, driving spreads higher… That’s one crowded trade all pointing in the same direction. Keep this in mind the next time spreads are widening out and politicians howl about the evil speculators that are exacerbating moves in the market (psst they may just be CVA desks following Basel regulations).

Exactly. CVA convexity is a big deal.

It’s also worth reiterating that nearly all sovereigns do not post collateral, but do demand it themselves — the dreaded one-way CSA. Banks have been trying to encourage sovereigns to start posting collateral so that they wouldn’t have to hedge against with CDS quite so much.

With the exception of a few minor countries (Portugal, for instance) they have not had much success. If sovereigns would agree to full two ways CSAs (zero threshold, daily collateral, cash only), then the CVA problem goes away. But, for now at least, they won’t, which is a shame as that is the only sensible way out of this mess.

Posted in: Basel, CDS and Negative Basis, Clearing and Collateral, Derivatives, Hedging and Convexity by David /
Comments Off

Goodness me there are some fishy things being written about BofA moving their derivatives to the retail bank (which of course has FDIC insured deposits). Some of the things that are not true include ‘If a retail bank is a derivatives counterparty, then it doesn’t need to post nearly as much collateral’ and ‘The derivatives aren’t themselves insured by the FDIC, but they have extremely senior status, which means that the bank can use its deposit base to pay off derivatives counter parties.’ (These actually consecutive sentences too – clearly Reuters does not bother to fact check blogs.)

What is true is that a retail bank often has rather better liquidity than a broker/dealer. This alone makes it safer, and hence (assuming that liquidity does not get spread around the rest of the group – something that should not happen too much) the retail bank is more attractive as a derivatives counterparty. Note though that these days everyone uses the same interdealer CSA (zero threshold, daily margin, cash only), so there is no collateral advantage to being in the retail bank; moreover being in a retail bank doesn’t suddenly make derivatives super senior; they are just good only fashioned pari passu with senior debt, the way they always were – and not FDIC-protected.

Posted in: Derivatives, Hedging and Convexity, Legal Risk and Trade Documentation by David /
5 Comments

With a title like that, you know this is going to be technical. So I am going to assume you know what CVA is. The key point in hedging the credit spread sensitivity of the CVA is to note that you can calculate the sensitivity of the CVA to a small (1 basis point say) move in credit spreads and buy CDS with equal an opposite sensitivity. That way if the credit spread moves, the P/L in the CVA is offset by that in the CDS (ignoring gamma effects anyway) and hence you are hedged.

Unfortunately, you are not hedged on a jump to default, at least with standard single name CDS. This is because a portfolio of derivatives are not deliverable into the CDS, so you can’t just hand over the thing that generated the CVA and get its par value. (This is true even if the portfolio of derivatives has close out value exactly equal to the CDS notional you have bought – the point isn’t what it is worth; it is that it is not deliverable.)

Now you might argue that ISDA receiveables are pari passu with senior debt and hence a cash settled CDS should be fine. There are several problems here. First, the cash settlement amount depends on an auction, and that has its own dynamics. The amount fixed here may turn out to be rather different from the eventual senior debt recovery. Second, events of default on derivatives are often wider than those on senior debt, so you may be able to close out the derivatives before there is a senior default. This again gives rise to a basis risk between the derivatives recovery and the CDS recovery. Finally, because of the deliverability issue discussed above, some (many?) supervisors do not give credit for CDS bought against CVA in the default part of the counterparty credit risk calculation. (The Basel 2 piece not the Basel 3 CVA charge.) This means that CVA hedging does not reduce regulatory capital right now. Messy, that, isn’t it?

Posted in: CDS and Negative Basis, Derivatives, Hedging and Convexity, Legal Risk and Trade Documentation by David /
1 Comment

Ever since synthetic CDO became for many a long way of saying trash, investors have been nervous about synthetic structures. The recent kerfuffle about ETFs is one example of this. Let me do a little fact from fiction separation here.

In a synthetic structure someone wants to take exposure to some asset class and someone is providing it. For instance a fund may wish to track the FTSE 100 without the tedious and expensive business of owning the component shares. Therefore it enters into a swap where it receives the total return on the index and pays some financing rate. So far, so straightforward.

An major division in synthetic structures is between those where the exposure provider owns the underlying asset and those where they don’t. In the simplest and safest version of the structure, the exposure provider owns exactly the asset they are providing exposure to, and they pledge this asset as collateral against the swap. (Typically this happens when the exposure provider has advantages that the exposure taker does not, such as better market access or an advantageous tax position with regard to returns on the asset.) Here the worst that can happen in the event of failure of the exposure provider is that the exposure taker seizes the asset. Providing that they can legally own the asset, this is highly likely to involve little or no loss: counterparty risk is minimal.

In any other version of the structure, there is are more risks. These can include the exposure provider having difficulty in providing the promised return because they don’t own the asset, and counterparty risk, especially if the structure is leveraged and/or the collateral posted against the swap does not cover the exposure. Morningstar has a good discussion of the issues for ETFs here. (Another risk arises if the return the swap provides is not precisely the one the exposure taker wants, but we won’t discuss that.)

Why were synthetic CDOs so dangerous? Well part of the answer at least is that the exposure provider often did not own the asset. Indeed, sometimes the structure was created because the exposure provider wanted to be short. This was (probably) the case for Goldman’s famous Abacus deal, for instance.

One particular issue here is that if the exposure provider owns the asset, then the derivative has not changed the market for the underlying. Instead of the exposure taker buying the asset (and hence presumably increasing its price), the exposure provider buys it instead. The market in the underlying is unaffected. However if the exposure provider is short synthetically – through providing exposure but not hedging that obligation – then the existence of the derivative has changed the market in the underlying.

An extreme example of this is where there is more exposure written than there is underlying available and this fact is not evident to market participants. For instance, suppose I sell physically settled CDS protection on more notional of a corporate bond than has been issued. Clearly if there is a credit event the CDS buyers will not be able to find enough bonds to deliver to me, and the market in the bonds will be disrupted. The Amherst trade is a great example of this.

Personally I think that one of the great advantages of trade repositaries is that they could be used to ensure that a false market in the underlying does not develop. If market participants knew what was going on synthetically just as they know (thanks to exchange reporting) what is going on in the cash market, then prices will better reflect the real supply and demand in the market. A real systemic risk of synthetic structures is that they can be used by people to hide what they are doing: now that the technology exists to eliminate this loophole, it should be used.

Posted in: Credit, Derivatives, Hedging and Convexity, Securitisation and Tranching by David /
1 Comment

FT alphaville askes How much is this plain vanilla derivative in the window?, noting

Banks aren’t marking a [-n uncollateralized] swap to market anymore, but to the model which dictates their internal cost of funds.

Why do we care?

Because it means pricing even the most basic (uncollateralised) swaps is now very complex.

Well, yes and no. A few points:

• A vanilla derivative is a collateralized one under the standard CSA these days (cash collateral in the same currency, daily MTM, daily margin). Anything else is exotic, because it involves an exotic collateral option.
• Accounting standards require firms to take account of their own cost of funds in calculating fair value. So using your own cost of funds to discount uncollateralized flows from you to a counterparty is not just standard, it is necessary.
• It is true that, as the IFR article Alphaville references says, Unsecured trades now present a serious valuation headache. But, um, that’s because they are really hard things to value. ‘Mark to market’ is a chimera here: fair value is the answer, and that is an institution-specific thing because it depends on funding cost.

Posted in: Derivatives, Hedging and Convexity, Fair Value by David /
3 Comments

FT alphaville suggests that a consequence of the CFTC proposed rule whereby sovereigns (except the US, naturally) will have to post collateral to US swap dealers on OTC derivatives is that US swap dealers won’t do many trades with sovereigns. They are of course right. The large European, Asian and Latin American debt management offices will not post collateral at the whim of the Americans – or anyone else.

Should they (and indeed the US government) be obliged to post? Well, it depends. If they had to, they would have to find the money from somewhere, and that would increase their levels of debt. On the other hand, one way CSAs, whereby the dealers have to post collateral to the sovereign but not vice versa, are a liquidity drain on the banking system.

The clinch argument for me, though, is CVA. Uncollateralised derivatives create a big CVA. (See here or here for a description of CVA and how it arises.) Basel 3 will require banks to hedge their CVA well or take capital against it. Critically this includes CVA from sovereign transactions. The way you hedge CVA is to buy CDS. Thus when you see a picture like these two from the Alphaville post:

What should ask yourself is ‘how much of this increase in CDS trading is caused by CVA hedging?‘

The prior post suggests that there is only $4-30 Billion net notional of CDS on even the biggest countries. My guess is that the major swap dealers need a good deal more liquidity than that to hedge their collective CVA position without moving the market.

The Basel 3 CVA rules, then, are a huge shot in the foot for governments. (A 50mm shell in the foot, maybe.) By forcing banks to hedge their sovereign CVA, they have increased the demand for sovereign CDS. That drives out spreads, which in turn increases borrowing costs for governments. The problem could be largely solved if sovereigns would agree to post collateral – but they won’t.

Posted in: CDS and Negative Basis, Clearing and Collateral, Derivatives, Hedging and Convexity, Sovereign Default by David /
2 Comments

Suppose that you wanted to do QE synthetically. What a government would do is buy the long end of the government curve and fund at the short end. So wouldn’t receive 10y govy rates, pay 3m bills do? Well no, it wouldn’t, and the reason is interesting.

It’s counterparty risk. If, say, the US government did that trade with a bank uncollateralised, then it would bear the credit risk of the bank. That changes the economics of the trade substantially. So what if the deal is collateralised? Well, what with? The only logical thing is USD cash, but that has to be funded too. For the government it is easy, it can just print it; but if the collateral flows from the bank to the government, then the bank has to fund that, and it does pay a spread for money.

What this means is that there is no counterparty that a government could go to to repeat the economics of QE synthetically. Moreover this argument may or may not extend to other counterparties: it would depend on their cost of funding.

Posted in: Clearing and Collateral, Derivatives, Hedging and Convexity by David /
Comments Off

Doug, in a truly excellent comment on my previous post, says this:

The reason speculative markets tend to price assets and risks correctly is because over time bad speculators make losing trades and have shrinking capital bases, and good speculators grow their capital base.

With the nonlinear assets (credit protection, vol, skew, correlation, carry trades, many quantitative mean reversion strategies, a lot of the volatile energy products) this mechanism breaks down. At any given time some players who are making losing trades with ex-ante negative expectation will have very large capital bases, because they’ve been consistently collecting the small payout and haven’t hit a large asymmetrical loss.

In other words, if you have only unleveraged longs and shorts in the market, and no derivatives assets, price discovery works. But as soon as you have convexity, so for instance people can write puts and pick up premium, then the mechanism starts to break down. Doug goes on:

This is especially true in a world where managers market to raise capital and people base their investment decisions off of short(ish) term track records. In a world where people only managed their own money and grew or shrunk their capital base from the returns on the initial investment, many of these asymmetrical payoff strategies wouldn’t have time to grow that large before collapsing. However the nature of hedge funds creates a world where capital grows super-linearly with returns (growing from the returns on the initial capital, as well as the influx/outflux of funds that a good/bad track record brings).

Thank you Doug; that is a most insightful comment.

Posted in: Derivatives, Hedging and Convexity, Fair Value by David /
2 Comments

Andy Haldane thinks that we should consider Cocos with market triggers as capital instruments for banks. Um. Dunno.

For me the issues are complex. First, whatever the Coco trigger is, it should be objective. There market will react badly to a trigger which is, in effect, whatever the supervisors want. That’s partly because as we have a longer period of calm the wisdom will develop that Cocos are never triggered; then, if they are, the shock will be enormous. So having an objective trigger – and one that is not too far from the money – is important.

Second, partial triggering is important. You do not want to be in a situation where the bank needs a bit more capital but not so much that the whole Coco needs to be triggered with the accompanying dilution for shareholders. IThe way to do that is to allow partial conversion.

Third, there is the hedging issue. Some folks will hedge Cocos by shorting equity against them. (And/or trading CDS.) The way to ensure that this hedging is not too disruptive, and is less likely to push the bank down, is to spread the gamma of the Coco around a range of strikes. This is a particular form of partial conversion where as the share price slides, more and more of the Coco converts. (Of course you need to adjust the current trigger for the dilution effect of earlier conversions.)

Fourth, false positives. Haldane is reasonably convincing that market based triggers do detect troubled firms. But do they detect trouble where none exists? My gut feeling is that part of the answer here is improved disclosure, but I’d want to see a thorough historical analysis of market based Coco conversions – and a behavioural analysis of how they would have been hedged – before truly embracing these instruments.

Update. See here for an interesting further discussion from VoxEU.

Posted in: Capital and Contingent Instruments, Derivatives, Hedging and Convexity by David /
Comments Off

Q: What’s a complex derivative?

A: One the respondent does not understand.

It seems like that is sometimes the answer anyway. Just because something is new, or new to the writer, does not necessarily mean it is complex, nor are old familiar things necessarily simple. As Merton demonstrated, for instance, common stock is far from common, being a derivative on the net asset value of a firm. That’s a complicated idea but the market has no problem with pricing equity.

Thus when FT Alphaville – normally it must be said the most reliable of sources – starts using terms like ‘OTC-squared’, one wonders if the thrill of the new has distorted judgment a little. Stripping away the hyperbole, there is a CDS options market. Primarily, this is a market on options on CDS on the CDX and iTraxx indices (and their tranches). This is very much like an OTC equity index market: you buy calls and puts on the index. You hedge with delta weighted amounts of the index. Um, that’s it.

FT alphaville discusses whether these CDX options should be cleared. Well, given that the CCPs are having problems offering clearing for any OTC options – including something as simple as plain vanilla index options on the S&P 500 – you would not have thought that it was a priority. Indeed, there are very good reasons that this is complex, including price discovery when the cleared option goes far away from the money (which it very well might).

So why pick on CDX options rather than, say, FX double barriers or capped callable floaters or globally floored cliquets on a custom stock basket? Because CDX options made a serious error for a derivative, an error few derivatives recover from. They are credit derivatives and thus, de facto, in some eyes at least, the spawn of satan. Exorcizo te, immundíssime spíritus, omnis derivativus credidi…

Posted in: CDS and Negative Basis, Clearing and Collateral, Derivatives, Hedging and Convexity by David /
4 Comments