The Smoke is Starting to Clear from the Aas/Nasdaq Blowup
Amir Khwaja of Clarus has a very informative post about the Nasdaq electricity blow-up.
The most important point: Nasdaq uses SPAN to calculate IM. SPAN was a major innovation back in the day, but it is VERY long in the tooth now (2018 is its 30th birthday!). Moreover, the most problematic part of SPAN is the ad hoc way it handles dependence risk:
- Intra-commodity spreading parameters – rates and rules for evaluating risk among portfolios of closely related products, for example products with particular patterns of calendar spreads
- Inter-commodity spreading parameters – rates and rules for evaluating risk offsets between related product
…..
CME SPAN Methodology Combined Commodity Evaluations
The CME SPAN methodology divides the instruments in each portfolio into groupings called combined commodities. Each combined commodity represents all instruments on the same ultimate underlying – for example, all futures and all options ultimately related to the S&P 500 index.
For each combined commodity in the portfolio, the CME SPAN methodology evaluates the risk factors described above, and then takes the sum of the scan risk, the intra-commodity spread charge, and the delivery risk, before subtracting the inter-commodity spread credit. The CME SPAN methodology next compares the resulting value with the short option minimum; whichever value is larger is called the CME SPAN methodology risk requirement. The resulting values across the portfolio are then converted to a common currency and summed to yield the total risk for the portfolio.
I would not be surprised if the handling of Nordic-German spread risk was woefully inadequate to capture the true risk exposure. Electricity spreads are strange beasts, and “rules for evaluating risk offsets” are unlikely to capture this strangeness correctly especially given the fact that electricity markets have idiosyncrasies that one-size-fits all rules are unlikely to capture. I also conjecture that Aas knew this, and loaded the boat with this spread trade because he knew that the risk was grossly underpriced.
There are reports that the Nasdaq margin breach at the time of default (based on mark-to-market prices) was not nearly as large as the €140 million hit to the default fund. In these accounts, the bulk of the hit was due to the fact that the price at which Aas’ portfolio was auctioned off included a substantial haircut to prevailing market prices.
Back in the day, I argued that one of the real advantages to central clearing was a more orderly handling of defaulted portfolios than the devil-take-the-hindmost process in OTC bilateral markets (cf., the outcome of the LTCM disaster almost exactly 20 years ago–with the Fed midwifed deal being completed on 23 September, 1998). (Ironically spread trades were the cause of LTCM’s demise too.)
But the devil is in the details of the auction, and in market conditions at the time of the default–which are almost certainly unsettled, hence the default. The CME was criticized for its auction of the defaulted Lehman positions: the bankruptcy trustee argued that the price CME obtained was too low, thereby harming the creditors. The sell-off of the Amaranth NG positions in September, 2006 (what is it about September?!?) to JP Morgan and Citadel (if memory serves) was also at a huge discount.
Nasdaq has been criticized for allowing only 4 firms to bid: narrow participation was also the criticism leveled at CME and NYMEX clearing in the Lehman and Amaranth episodes, respectively. Nasdaq argues that telling the world could have sparked panic.
But this episode, like Lehman and Amaranth before it, demonstrate the challenges to auctioning big positions. Only a small number of market participants are likely to have the capital, or the risk appetite, to take on a big defaulted position in its entirety. Thus, limited participation is almost inevitable, and even if Nasdaq had invited more bidders, there is room to doubt whether the fifth or sixth or seventh bidder would have been able to compete seriously with the four who actually participated. Those who have the capital and risk appetite to bid seriously for big positions will almost certainly demand a big discount to compensate for the risk of holding the position until they can work it off. Moreover, limited participation limits competition, which should exacerbate the underpricing problem.
Thus, even with a structured auction process, disposing of a big defaulted portfolio is almost inevitably something of a fire sale. This is a risk borne by the participants in the default fund. Although the exposure via the default fund is sometimes argued to be an incentive for the default fund participants to bid aggressively, this is unlikely because there are externalities: the aggressive bidder bears all the risks and costs, and provides benefits to the rest of the other members. Free riding is a big problem.
In theory, equitizing the risk might improve outcomes. By selling shares in the defaulted portfolio, no single or two bidders would have to absorb the entire position and risk could be spread more efficiently: this could reduce the risk discount in the price. But who would manage the portfolio? What are the mechanics of contributing to IM and VM? Would it be like a bad bank, existing as a zombie until the positions rolled off?
Another follow-up from my previous post relates to the issue of self-clearing. On Twitter and elsewhere, some have suggested that clearing through a 3d party would have been an additional check. Surely an FCM would be less likely to fall in love with a position than the trader who puts it on, but the effectiveness of the FCM as a check depends on its evaluation of risk, and it may be no smarter than the CCP that sets margins. Furthermore, there are examples of FCMs having the same trade in their house account as one of their big customers–perhaps because they think the client is really smart and they want to free ride off his genius. As a historical example, Griffin Trading had a big trade in the same instrument and direction as its biggest client. The trade went pear-shaped, the client defaulted, and Griffin did too.
I also need to look to see whether Nasdaq Commodities uses the US futures clearing model, which does not segregate positions. If it does, and if Aas had cleared through an FCM, it is possible that the FCM’s clients could have lost money as a result of his default. This model has fellow-customer risk: by clearing for himself, Aas did not create such a risk.
I also note that the desire to expand clearing post-Crisis has made it difficult and more costly for firms to find FCMs. This problem has been exacerbated by the Supplementary Leverage Ratio. Perhaps the cost of clearing through an FCM appeared excessive to Aas, relative to the alternative of self-clearing. Thus, if regulators blanch at the thought of self-clearing (not saying that they should), they should get serious about addressing the FCM cost issue, and regulations that inflate these costs but generate little offsetting benefit.
Again, this episode should spark (no pun intended!) a more thorough reconsideration of clearing generally. The inherent limitations of margin models, especially for more complex products or markets. The adverse selection problems that crude risk models can create. The challenges of auctioning defaulted portfolios, and the likelihood that the auctions will become fire sales. The FCM capacity issue.
The supersizing of clearing in the post-Crisis world has also supersized all of these concerns. The Aas blowup demonstrates all of them. Will CCPs and regulators take heed? Or will some future September bring us the mother of all blowups?
Lest we forget, it was also a spread trade that took down Barings.
Comment by charles — September 20, 2018 @ 12:13 pm
Great analysis and very informative as always.
Comment by Miltie — September 22, 2018 @ 1:16 pm
Carig,
Who would have guessed that Nasdaq is still stuck in the 1980s with SPAN! I haven’t looked at how CCPs manage their margin positions but elsewhere risk management has moved on a peg or two and I — innocently — thought that CCPs would have adopted some of the newer technologies.
As I understand SPAN it is essentially a scenario technique which looks at the absolutely worse outcome under multiple scenarios and computes the worst case. Perhaps not the best way to think of spread risk as this story shows.
A key question, one wonders, is whether CCPs and regulators will wake up to the issues you discuss here. I wonder.
Peter
Comment by Peter Moles — September 24, 2018 @ 3:22 am
Not only are Nasdaq using SPAN, the CME still use it too. Which is why I agree the self clearing issue is a bit of a red herring. The clearing members just upload positions into SPAN to calculate risk anyway.
If you’re relying on a simplistic position limits to prevent blow ups, what is all your fancy ‘risk modelling’ for?
If people who work in banking (which is far from perfect on the risk modelling side of things) knew how the CHs calculate their risks, there would be a major reassessment of their exposures to these jokers.
The guys doing the trading I think know this very well. The PBs and banks don’t really want to know.
Comment by Bob — September 24, 2018 @ 4:28 am
Not all CCPs still rely on SPAN.
German Eurex margining methodology is based on real-time intra-day risk evaluation for positions using current underlying prices/volatilities, which is like a Porsche vs. Volkswagen Beetle. So, prudent risk management is addressed at least at some CCPs.
But nice article anyway.
Lars
Comment by LarsB — September 24, 2018 @ 5:31 am
CME also uses SPAN, if you spend any time looking into how clearing houses ‘manage’ risk, you’ll be dumbfounded. It’s so basic they should be embarrassed.
The clearing members probably just upload their positions into SPAN and add something on, so using an FCM may not have made much difference unless they had position limits.
But what’s the point of all the ‘risk modelling’ and oversight if you’re relying on blunt position limits?
The large traders I expect understand the limitations of the CH models very well, the CFTC not so much.
Comment by Bob — September 24, 2018 @ 9:18 am
@Lars–Thanks. But to clarify: I never said all CCPs rely on SPAN. Further, yes, using current underlying prices/volatilities is likely better than SPAN, but if these are used in a VaR-type model, they can give highly misleading results. Furthermore, correlations also matter a lot, and estimation of conditional correlations is far more difficult, and they tend to vary substantially more than vols.
Comment by cpirrong — September 24, 2018 @ 2:47 pm
@Bob–Methinks that one reason why banks are so keen on requiring CCPs to have more skin the game is precisely driven by their reservations/concerns about CCP risk modeling. Model risk is a big deal, and the first principle of risk bearing is that the party that can control the risk at lowest cost should bear it. CCPs control models, and hence the model risk.
This observation relating model risk was at the center of a paper I wrote years ago trying to explain why some things were cleared, and others weren’t.
Comment by cpirrong — September 24, 2018 @ 2:50 pm
@Craig, I agree that is a reason why banks are so keen on increasing SITG for the CCPs, but I think the group of people within the banks who are sufficiently informed about how the CCP risk modelling operates is surprisingly small.
On the back of having worked in one I get asked in interviews about what happens, this would be with senior risk people, and they can’t believe what I tell them.
That’s just the margin modelling side of things, if they looked into the default management processes they’d have many sleepless nights.
Weirdly, they’re also unaware of what the LCH did during the day of Brexit for instance. The general level of education of how the CCPs operate is too low in my opinion.
The risk model at Eurex is far better I agree, but I doubt it would have correctly risked this guy’s positions. At the end of the day, it’s still a 3 year look back historical VaR.
Comment by Bob — September 25, 2018 @ 3:48 am
I’m still sniggering immaturely at the guy’s name being “Aas”.
Nomine facta concordant, and no mistake.
Comment by Green as Grass — September 25, 2018 @ 10:28 am
@Bob–Thanks for the insights. Very informative, though not particularly surprising.
LCH during Brexit was a nightmare and illustrated–in spades–why it is foolish to expect CCPs to mitigate systemic risk. LCH’s self-interested/self-preserving/sauve-qui-peut moves are exactly the kind of thing that can create a systemic event.
I was at FIA Chicago last year, and several FCM people griped about what had happened. Shocking that people that weren’t directly involved aren’t more aware.
I’d appreciate any observations re the default management process.
Comment by cpirrong — September 25, 2018 @ 4:20 pm
@Green–sometimes you just can’t make up this stuff.
Comment by cpirrong — September 25, 2018 @ 4:20 pm
Well the DM process is a large subject, I expect you’ve see this (since you’re quoted often) https://www.mercatus.org/system/files/Clearinghouse-PDF.pdf.
It very useful because it links to a lot of CH documents, which are mostly about how great their processes are.
But in https://www.cftc.gov/sites/default/files/idc/groups/public/@aboutcftc/documents/file/mrac_040215_transcript.pdf
Scott Flood from Citi hits the nail on the head. A big bank defaulting will default in all CHs and all asset classes simultaneously, the DM process relies on seconding traders from the clearing members, just when they have their own survival issues. There won’t be enough expertise to go around. Try calling in the head of rates at Goldman if MS have gone bust.
The thing to realise is that the auction process is irrelevant, it’s the hedging process that matters.
In the event of default, the staff of a clearing house will go into the market to hedge the portfolio, at least in the initial stages before their seconded traders show up.
LCH has an enormous swap book, what will the impact be of them trying to hedge a bank’s portfolio into a stressed market? When the market realises which way around they are, how much do you think it will move?
IRS IM in 10 year swaps is around 30-40 bps depending on the currency. If another Lehmans happened over a weekend, how many bps would the market move by on Monday morning?
What would the market impact be of trying to hedge a massive swap book?
Remember Lehmans day? Now imagine a re-run with LCH trying to hedge their swaps, ICE trying to hedge a CDS book etc.
Comment by Bob — September 27, 2018 @ 7:44 am
I see Clarus are writing blog posts of the ‘nothing to see here’ variety.
If there is a domain more populated by charlatans than the ‘market risk management’ one, I’ve yet to see it.
Comment by Bob — October 8, 2018 @ 5:05 pm
@Bob. Yeah, I saw that. SPAN is just great, right?
Comment by cpirrong — October 8, 2018 @ 11:52 pm
It wouldn’t do to rock the boat when your pitching for business in the clearing ecosystem.
That their argument would apply equally well before the Nasdaq default doesn’t seem to occur to them.
Comment by Bob — October 9, 2018 @ 3:14 am
you’re dammit
Comment by Bob — October 9, 2018 @ 5:48 am