Streetwise Professor

I’m on my way back from my annual teaching sojourn in Geneva, plus a day in the Netherlands for a speaking engagement.  While I was taking that European non-quite-vacation, a Norwegian power trader, Einar Aas, suffered a massive loss in cleared spread trades between Nordic and German electricity.  The loss was so large that it blew through Aas’ initial margin and default fund contribution to the clearinghouse (Nasdaq), consumed Nasdaq’s €7 million capital contribution to the default fund, and €107 million of the rest of the default fund–a mere 66 percent of the fund.  The members have been ordered to contribute €100 million to top up the fund.

This was bound to happen. In a way, it was good that it happened in a relatively small market.  But it provides a sobering demonstration of what I’ve said for years: clearing doesn’t eliminate losses, but affects the distribution of losses.  Further, financial institutions that back CCPs–the members–are the ultimate backstops.  Thus, clearing does not eliminate contagion or interconnections in the financial network: it just changes the topology of the network, and the channels by which losses can hit the balance sheets of big players.

Happening in the Nordic/European power markets, this is an interesting curiosity.  If it happens in the interest rate or equity markets, it could be a disaster.

We actually know very little about what happened, beyond the broad details.  We know Aas was long Nordic power and short German power, and that the spread widened due to wet weather in Norway (which depresses the price of hydro and reduces demand) and an increase in European prices due to increases in CO2 prices.  But Nasdaq trades daily, weekly, monthly, quarterly, and annual power products: we don’t know which blew up Aas.  Daily spreads are more volatile, and exhibit more extremes (kurtosis), but since margins are scaled to risk (at least theoretically–more on this below) what matters is the market move relative to the estimated risk.  Reports indicate that the spread moved 17x the typical move, but we don’t know what measure of “typical” is used here.  Standard deviation?  Not a very good measure when there is a lot of kurtosis (or skewness).

I also haven’t seen how big Aas’ initial margins were.  The total loss he suffered was bigger than the hit taken by the default fund, because under the loser-pays model, the initial margins would have been in the first loss position.

The big question in my mind relates to Nasdaq’s margin model.  Power price distributions deviate substantially from the Gaussian, and estimating those distributions is challenging in part because they are also conditional on day of the year and hour of the day, and on fundamental supply-demand conditions: one model doesn’t fit every day, every hour, every season, or every weather enviornment.  Moreover, a spread trade has correlation risk–dependence risk would be a better word, given that correlation is a linear measure of dependence and dependencies in power prices are not linear.  How did Nasdaq model this dependence and how did that impact margins?

One possibility is that Nasdaq’s risk/margin model was good, but this was just one of those things.  Margins are set on the basis of the tails, and tail events occur with some probability.

Given the nature of the tails in power prices (and spreads) reliance on a VaR-type model would be especially dangerous here.  Setting margin based on something like expected shortfall would likely be superior here.  Which model does Nasdaq use?

I can also see the possibility that Nasdaq’s margin model was faulty, and that Aas had figured this out.  He then put on trades that he knew were undermargined because Nasdaq’s model was defective, which allowed him to take on more risk than Nasdaq intended.

In my early work on clearing I indicted that this adverse selection problem was a concern in clearing, and would lead CCPs–and those who believe that CCPs make the financial system safer–to underestimate risk and be falsely complacent.  Indeed, I argued that one reason clearing could be a bad idea is that it was more vulnerable to adverse selection problems because the need to model the distribution of gains/losses on cleared positions requires detailed knowledge, especially for more exotic products.  Traders who specialize in these products are likely to have MUCH better understanding about risks than a non-specialist CCP.

Aas cleared for himself, and this has caused some to get the vapors and conclude that Nasdaq was negligent in allowing him to do so.  Self-clearing is just an FCM with a house account, but with no client business: in some respects that’s less risky than a traditional FCM with client business as well as its own trading book.

Nasdaq required Aas to have €70 million in capital to self-clear.  Presumably Nasdaq will get some of that capital in an insolvency proceeding, and use it to repay default fund members–meaning that the €114 million loss is likely an overestimate of the ultimate cost borne by Nasdaq and the clearing members.

Further, that’s probably similar to the amount of capital that an FCM would have had to have to carry a client position as big as Aas’.   That’s not inherently more risky (to the clearinghouse and its default fund) than if Aas had cleared through another firm (or firms).  Again, the issue is whether Nasdaq is assessing risks accurately so as to allow it to set clearing member capital appropriately.

But the point is that Aas had to have skin in the game to self-clear, just as an FCM would have had to clear for him.

Holding Aas’ positions constant, whether he cleared himself or through an FCM really only affected the distribution of losses, but not the magnitude.  If Aas had cleared through someone else, that someone else’s capital would have taken the hit, and the default fund would have been at risk only if that FCM had defaulted.  But the total loss suffered by FCMs would have been exactly the same, just distributed more unevenly.

Indeed, the more even distribution that occurred due to mutualization which spread the default loss among multiple FCMs might actually be preferable to having one FCM bear the brunt.

The real issue here is incentives.  My statement was that holding Aas’ positions constant, who he cleared through or whether he cleared at all affected only the distribution of losses.  Perhaps under different structures Aas might not have been able to take on this much risk.  But that’s an open question.

If he had cleared through another FCM, that FCM would have had an incentive to limit its positions because its capital was at risk.  But Aas’ capital was at risk–he had skin in the game too, and this was necessary for him to self-clear.  It’s by no means obvious that an FCM would have arrived at a different conclusion than Aas, and decided that his position represented a reasonable risk to its capital.

Here again a key issue is information asymmetry: would the FCM know more about the risk of Aas’ position, or less?  Given Aas’ allegedly obsessive behavior, and his long-time success as a trader, I’m pretty sure that Aas knew more about the risk than any FCM would have, and that requiring him to clear through another firm would not have necessarily constrained his position.  He would have also had an incentive to put his business at the dumbest FCM.

Another incentive issue is Nasdaq’s skin in the game–an issue that has exercised FCMs generally, not just on Nasdaq.  The exchange’s/CCP’s relatively thin contribution to the default fund arguably reduces its incentive to get its margin model right.  Evaluating whether Nasdaq’s relatively minor exposure to default risk led it to undermargin requires a more thorough analysis of its margin model, which is a very complex exercise which is impossible to do given what we know about the model.

But this all brings me back to themes I flogged to the collective shrug of many–indeed almost all–of the regulatory and legislative community back in the aftermath of the Crisis, when clearing was the silver bullet for future crises.   Clearing is all about the allocation and pricing of counterparty credit risk.  Evaluation of counterparty credit risk in a derivatives context requires a detailed understanding of the price risks of the cleared products, and dependencies between these price risks and the balance sheet risks of participants in cleared markets.  Classic information problems–adverse selection and moral hazard (too little skin in the game)–make risk sharing costly, and can lead to the mispricing of risk.

The forensics about Aas blowing up real good, and the lessons learned from that experience, should focus on those issues.  Alas, I see little recognition of that in the media coverage of the episode, and betting on form, I would wager that the same is true of regulators as well.

The Aas blow up should be a salutary lesson in how clearing really works, what it can do, and what it can’t.   Cynic that I am, I’m guessing that it won’t be.  And if I’m right, the next time could be far, far worse.