Central counterparties work on the principle of mutualization. CCP default funds are a mechanism for mutualizing–sharing–counterparty loss risk, where the CCP-insurer has some protection in the form of a “deductible” that consists of the defaulter’s margin plus default fund contribution.
Insurance works when risks are diversifiable. So, to the extent that the risks of default exhibit relatively little correlation across clearing members (CMs), mutualization through a CCP can improve the allocation of risk: through diversification, the variability of the average default loss across all CCP members is smaller than the variability of any CM’s default loss profile.
It should be noted that these diversification effects work in OTC markets too: market participants who deal with multiple counterparties diversify exposure to counterparty losses. Moreover, in OTC markets, it is often possible to hedge counterparty risk through the CDS market. [Which raises a question: how do you hedge CCP credit risk? It is highly unlikely that CDS will be traded on CCPs. Perhaps using a portfolio of CDS on CMs? But who clears those CDS? Sounds like the wrong way risk from hell.]
CCPs have been sold as a means of reducing systemic risk. The basic story is that in bilateral OTC markets, the failure of one entity can cause the failure of other entities in falling-domino-like fashion. In this telling, a CCP reduces interconnectivity of the financial market, and thereby reduces the likelihood of that the failure of one firm–even a big one–will cause the failure of others. Put differently, in this view default/counterparty risk is insurable because it is diversifiable, and therefore mutualization can improve the allocation of risk.
But is the domino-model really a good description of systemic risk? Is the system really threatened by the isolated, idiosyncratic failure of one firm, even a big one?
Here’s an alternative story, in which a big, system-wide shock hitting many institutions simultaneously is the source of systemic crisis. Some examples: Black Monday in 1987, the Asian and Russian crises in 1998, and the subprime meltdown of 2007-2008.
In such episodes, there are large price moves across asset classes. These price moves often blow through margins–and by a lot. During the subprime crisis, for instance, even well-before the Lehman episode, some prices moved by huge amounts. In August, 2007–more than a year before Fannie-Freddie, Lehman, and AIG–David Viniar of Goldman lamented that there had been several 25 standard deviation moves in multiple markets. In a “normal” (Gaussian) world, 25 sd moves don’t occur. Period. People don’t set margins based on the possibility of 25 sigma moves. So when such moves occur–and they occur in the real, non-normal world–margins get blown through. Meaning that if someone defaults, they’re well into the default fund, and the protection of the “deductible” is gone.
Moreover, such earth-shaking episodes are typically associated with adverse shocks to bank (CM) balance sheets. What’s more, such events raise serious doubts about the solvency of banks/CMs: adverse selection problems can run rampant, causing markets to freeze. This tends to cause funding and liquidity to dry up. Due to actual insolvency, and illiquidity driven by widespread fears of potential insolvency, those facing big margin calls may face substantial difficulty in meeting them. Thus, they are more likely to default. This is what happened to AIG and Lehman.
Since the shock hits all financial institutions simultaneously, and since the shock is big, this risk of default is not idiosyncratic; it is highly correlated across firms. Given such a shock, if one firm defaults, others are at elevated risk of default too.
If defaults that breach margins occur, the surviving CMs who contribute to the default fund must absorb the loss, and take a hit to their capitals. But this hit occurs exactly when capital is very valuable/costly for them. That is, hits to the default fund are likely to come at the absolute worst times–during adverse moves in the overall market.
It is possible that a CCP’s default exposure in such a circumstance could crater the clearinghouse. But even if the CCP survives, its surviving members suffer a loss (via to their exposure to the default fund) when they can afford it least.
This has some interesting implications. In particular, it means that the systematic risk of default fund capital is very high, which in turn means that default fund capital is extremely expensive. The default fund is like a deep out of the money short put on the market. Such puts are essentially bets on systematic risk, and hence are very expensive.
The expense of such capital means that CMs are likely to want to minimize the amount they post. They have every incentive to try to construct risk measures that make CCPs look safe, with very low risk of default, in order to reduce their exposure to the default fund.
Think that this can’t be done? Think again: that was essentially the entire game in CDOs. As Coval-Jurek-Stafford show in an AER paper, “super senior” CDOs were claims that had low probabilities of default, but that defaults would occur precisely during periods of crisis–analogous to the foregoing characterization of the risk profile of default funds. C-J-S present empirical evidence that these CDOs–which were AAA rated–were grossly underpriced, meaning that those who were buying these securities didn’t fully appreciate how heavily they loaded on systematic risk because analysis–and ratings–were framed almost exclusively in terms of unconditional default probabilities, and the rating agencies and many investors and regulators ignored the fact that defaults would occur when the value of capital was very high. I can see the same thing happening with CCPs–who will reflect the interests of their members–focusing on default probabilities to justify their capitalization to regulators.
Another interesting implication is for what instruments should be cleared. Instruments that have big losses during big systematic shocks–those that load heavily on systematic risk–are exceptionally dangerous for CCPs. This means that clearing things like senior or supersenior tranches (or derivatives on them) is exceptionally dangerous.
This has a corollary: anybody that uses AIG to justify clearing mandates doesn’t know what the hell they’re talking about. AIG’s positions in derivatives on supersenior tranches were, as C-J-S show, systematic risk time bombs. Clearing those things would dramatically increase the systematic risk of default funds, because they would expose default funds to losses precisely during systemic crises.
Some people know this. Jon Gregory, for instance. When discussing “market coverage of a CCP”, he states:
Finally, CDO tranches (especially supersenior) would be expected to have serious correlation [i.e., systematic risk] impact since in a scenario where there are losses on these tranches it is likely that many CCP counterparties [and CMs] are in a distressed state.
The above analysis is to some degree subjective but one general point–probably beyond argument–is that the most important trades products to be traded through CCPs are arguably the most risky from the point of view of the stability of the CCP. There has been much recent interest to trade all CDS index products and single-name products through CCPs. . . . The general point is that the products that market participants will most want and need to trade through a CCP will be the precise products that are most difficult to handle in this way (pp. 381-382).
In sum, the belief that CCPs can reduce systemic risk is predicated on a particular narrative of the likely source of a future crisis. This view is that the failure of a single institution, arising from some shock idiosyncratic to it (a bad trading decision, an operational risk such as a rogue trader) can lead to the failures of other institutions via a contagion mechanism.
But, in my view, that isn’t an accurate characterization of the actual crises that have occurred over the years. These have been associated with broad-based shocks that hit large numbers of institutions simultaneously. Again: Black Monday, Asia/Russia, subprime.
There have been multiple examples of large, one-off collapses of highly interconnected and leveraged institutions that have not resulted in a cascade of failures that led to a systemic crisis. To name several off the top of my head: MG (1992), Barings (1995), Amaranth (2006). For a while, it looked like the Hunt collapse in 1980 could have broader, systemic consequences, but that fear evaporated rather quickly.
The contrast between the effects of the collapse of Amaranth in 2006 and the crisis surrounding the problems of LTCM in 1998 is particularly instructive. Amaranth was actually bigger than LTCM (though it should be noted that the markets were bigger too), and had large OTC positions. The difference was that Amaranth’s collapse was not caused by a systematic shock, whereas LTCM’s was. LTCM engaged in strategies that blew up because of major shocks arising in Asia and Russia. It was thus on the verge of failure precisely when liquidity had dried up, and its (big bank) counterparties were in bad financial shape because of the crisis.
The Amaranth collapse was effectively a diversifiable risk, and potentially clearing improved the allocation of that risk (many of Amaranth’s positions were cleared). But the bilateral markets could have likely handled it quite well too. In contrast, LTCM’s risk was not diversifiable. Clearing would have still passed the risk of its failure onto stressed financial institutions, and likely resulted in some similar emergency measure to mitigate the impact of its huge losses. (I’ve long argued that the same would have been the case with AIG.)
Along these lines, FTAlphaville has a very interesting post describing some forthcoming research which shows that incorporating “common jumps” into models of index option prices go a long way towards explaining the volatility skew, and reconciling the skews of index components and the skew of the index:
According to a technical paper published on Risk.net by Alex Langnau, global head of analytics at Allianz Investment Management and Daniel Cangemi, head of FICC trading at EFG Financial Products, Wall Street’s next top (risk) model is actually a copula that attempts to explicitly link correlation skew to systemic risk so as to improve tail risk management of large portfolios.
The basic premise lies in the idea that the downside risk of a portfolio of assets is generally substantially higher than the downside risk of its components — because assets tend to correlate more strongly in times of crisis.
. . . .
As the authors conclude this could become extremely useful for trying to figure out the additional risk attached to asset portfolios:
We have presented a generalisation of the Merton jump option formula to multi-assets as well as diffusive skews and found that common jumps are an appropriate and intuitive way to define copulas that describe the systemic risk of a portfolio.
Financial institutions should start treating systemic risk as market data across asset classes and start to mark, risk-manage and price these effects in a systematic way.
This suggests that these common jumps are of first order importance. And these common jumps aren’t diversifiable. What I argue above is that the major risk that CCPs face arises from such system-wide, common shocks. And in the event, mutualization is all but irrelevant. Except, perhaps, to the extent that CMs attempt to limit their exposure to this risk via the default fund by undercapitalizing CCPs, thereby threatening the ability of these vital–and soon to be even more vital–pieces of market infrastructure to survive a systematic shock. In which case, mutualization makes things worse because if some CCPs don’t survive, the knock-on effects from the original shock will be even more catastrophic.
In some respects, the issues I’ve discussed here reflect an older literature on bank runs. One view of bank runs is that they are idiosyncratic in origin (i.e., a bad shock–perhaps even a sunspot–hits a single bank, leading to a run), but have systemic effects due the interconnectedness of the financial system. In this view, widespread panics spread like a wildfire from a single, isolated spark.
An alternative view, associated most strongly with the work of Calomiris and Gorton, is that runs occur when depositors receive adverse information about the aggregate economy, which raises concerns about the viability of all banks. In this view, panics involving runs at many financial institutions reflects the fact that all banks are exposed to a common risk.
The Calomiris-Gorton view, and the evidence they amassed to support it is persuasive. In more recent times, major periods of crisis have been associated with a common shock hitting many financial institutions, rather than from the contagion of an idiosyncratic shock to one big institution laying low otherwise healthy institutions. In these types of events, mutualization of risk isn’t of much help, and can actually make things worse. Meaning that clearing mandates were sold largely on the basis of a misleading portrayal of, or misunderstanding of, the real dynamics of financial crises.
One takeaway from this is that there should be a very hard look at CCP capitalization, and at measuring the cost of default fund capital. The exposure that default funds face to systematic risk affects the cost of capital–and hence the cost of clearing. That cost should be taken into consideration when deciding the scope of clearing mandates–that is, when deciding what should be cleared. For the analysis suggests that clearing can be very expensive, and that a major source of cost has been largely overlooked. Forcing too much to be cleared can create costs with no corresponding benefit.