A recent Office of Financial Research paper by Samim Ghamami and Paul Glasserman asks “Does OTC Derivatives Reform Incentivize Central Clearing?” Their answer is, probably not.
My overarching comment is that the paper is a very precise and detailed answer to maybe not the wrong question, exactly, but very much a subsidiary one. The more pressing questions include: (i) Do we want to favor clearing vs. bilateral? Why? What metric tells us that is the right choice? (The paper takes the answer to this question as given, and given as “yes.”) (ii) How do the different mechanisms affect the allocation of risk, including the allocation of risk outside the K banks that are the sole concern in the paper? (iii) How will the rules affect the scale of derivatives trading (the paper takes positions as given) and the allocation across cleared and bilateral instruments? (iv) Following on (ii) and (iii) will the rules affect risk management by end-users and what is the implication of that for the allocation of risk in the economy?
Item (iv) has received too little attention in the debates over clearing and collateral mandates. To the extent that clearing and collateral mandates make it more expensive for end-users to manage risk, how will the end users respond? Will they adjust capital structures? Investment? The scale of their operations? How will this affect the allocation of risk in the broader economy? How will this affect output and growth?
The paper also largely ignores one of the biggest impediments to central clearing–the leverage ratio. (This regulation receives on mention in passing.) The requirement that even segregated client margins be treated as assets for the purpose of calculating this ratio (even though the bank does not have a claim on these margins) greatly increases the capital costs associated with clearing, and is leading some banks to exit the clearing business or to charge fees that make it too expensive for some firms to trade cleared derivatives. This brings all the issues in (iv) to the fore, and demonstrates that certain aspects of the massive post-crisis regulatory scheme are not well thought out, and inconsistent.
Of course, the paper also focuses on credit risk, and does not address liquidity risk issues at all. Perhaps this is a push between bilateral vs. cleared in a world where variation margin is required for all derivatives transactions, but still. The main concern about clearing and collateral mandates (including variation margin) is that they can cause huge increases in the demand for liquidity precisely at times when liquidity dries up. Another concern is that collateral supply mechanisms that develop in response to the mandates create new interconnections and new sources of instability in the financial system.
The most disappointing part of the paper is that it focuses on netting economies as the driver of cost differences between bilateral and cleared trading, without recognizing that the effects of netting are distributive. To oversimplify only a little, the implication of the paper is that the choice between cleared and bilateral trading is driven by which alternative redistributes the most risk to those not included in the model.
Viewed from that perspective, things look quite different, don’t they? It doesn’t matter whether the answer to that question is “cleared” or “bilateral”–the result will be that if netting drives the answer, the answer will result in the biggest risk transfer to those not considered in the model (who can include, e.g., unsecured creditors and the taxpayers). This brings home hard the point that these types of analyses (including the predecessor of Ghamami-Glasserman, Zhu-Duffie) are profoundly non-systemic because they don’t identify where in the financial system the risk goes. If anything, they distract attention away from the questions about the systemic risks of clearing and collateral mandates. Recognizing that the choice between cleared and bilateral trading is driven by netting, and that netting redistributes risk, the question should be whether that redistribution is desirable or not. But that question is almost never asked, let alone answered.
One narrower, more technical aspect of the paper bothered me. G-G introduce the concept of a concentration ratio, which they define as the ratio of a firm’s contribution to the default fund to the firm’s value at risk used to determine the sizing of the default fund. They argue that the default fund under a cover two standard (in which the default fund can absorb the loss arising from the simultaneous defaults of the two members with the largest exposures) is undersized if the concentration ratio is less than one.
I can see their point, but its main effect is to show that the cover two standard is not joined up closely with the true determinants of the risk exposure of the default fund. Consider a CCP with N identical members, where N is large: in this case, the concentration ratio is small. Further, assume that member defaults are independent, and occur with probability p. The loss to the default fund conditional on the default of a given member is X. Then, the expected loss of the default fund is pNX, and under cover two, the size of the fund is 2X. There will be some value of N such that for a larger number of members, the default fund will be inadequate. Since the concentration ratio varies inversely with N, this is consistent with the G-G argument.
But this is a straw man argument, as these assumptions are obviously extreme and unrealistic. The default fund’s exposure is driven by the extreme tail of the joint distribution of member losses. What really matters here is tail dependence, which is devilish hard to measure. Cover two essentially assumes a particular form of tail dependence: if the 1st (2nd) largest exposure defaults, so will the 2nd (1st) largest, but it ignores what happens to the remaining members. The assumption of perfect tail dependence between risks 1 and 2 is conservative: ignoring risks 3 through N is not. Where things come out on balance is impossible to determine. Pace G-G, when N is large ignoring 3-to-N is likely very problematic, but whether this results in an undersized default fund depends on whether this effect is more than offset by the extreme assumption of perfect tail dependence between risks 1 and 2.
Without knowing more about the tail dependence structure, it is impossible to play Goldilocks and say that this default fund is too large, this default fund is too small, and this one is just right by looking at N (or the concentration ratio) alone. But if we could confidently model the tail dependence, we wouldn’t have to use cover two–and we could also determine individual members’ appropriate contributions more exactly than relying on a pro-rata rule (because we could calculate each member’s marginal contribution to the default fund’s risk).
So cover two is really a confession of our ignorance. A case of sizing the default fund based on what we can measure, rather than what we would like to measure, a la the drunk looking for his keys under the lamppost, because the light is better there. Similarly, the concentration ratio is something that can be measured, and does tell us something about whether the default fund is sized correctly, but it doesn’t tell us very much. It is not a sufficient statistic, and may not even be a very revealing one. And how revealing it is may differ substantially between CCPs, because the tail dependence structures of members may vary across them.
In sum, the G-G paper is very careful, and precisely identifies crucial factors that determine the relative private costs of cleared vs. bilateral trading, and how regulations (e.g., capital requirements) affect these costs. But this is only remotely related to the question that we would like to answer, which is what are the social costs of alternative arrangements? The implicit assumption is that the social costs of clearing are lower, and therefore a regulatory structure which favors bilateral trading is problematic. But this assumes facts not in evidence, and ones that are highly questionable. Further, the paper (inadvertently) points out a troubling reality that should have been more widely recognized long ago (as Mark Roe and I have been arguing for years now): the private benefits of cleared vs. bilateral trading are driven by which offers the greatest netting benefit, which also just so happens to generate the biggest risk transfer to those outside the model. This is a truly systemic effect, but is almost always ignored.
In these models that focus on a subset of the financial system, netting is always a feature. In the financial system at large, it can be a bug. Would that the OFR started to investigate that issue.