Streetwise Professor

My most recent clearing post explored the possibility that the formation of a clearinghouse for OTC derivatives is efficient, but that distributive considerations and collective action problems have heretofore prevented its formation. In this post, I consider the possibility that the costs of formation of an OTC clearinghouse (for CDSs, for instance) exceed the benefits.

A central counterparty (CCP) is a default risk sharing mechanism. In a typical CCP setup, a group of intermediaries who trade on their own account and as agents for customers share default risk among themselves. If a customer defaults, the intermediary through whom that customer clears bears the loss. The risk sharing mechanism kicks in only if the one of the CCP member firms defaults.

Two factors drive the default risk that the CCP members share. The first is the riskiness of the instruments that are cleared. The second is the riskiness of the balance sheets of the member firms. A default occurs only if (a) a member firm incurs a loss on its position(s), and (b) this loss exceeds the value of the other assets (net of liabilities) that the firm can draw upon (e.g., sell) to cover this loss. Thus, the members of a CCP take on an option-like exposure, where the option is a complex one with a payoff that depends on multiple factors, namely, the prices of the instruments cleared and the values of the other assets and liabilities in each member’s portfolio.

Recent events (though not in the context of a CCP) indicate that both sources of risk are important in determining the likelihood of default. Lehman and Bear Stearns defaulted on their derivative obligations not because of losses incurred on these derivatives, but because of losses incurred on other investments (primarily mortgage securities). In contrast, AIG collapsed because the huge losses on derivative positions overwhelmed the capital on its otherwise healthy balance sheet.

To operate effectively, any risk sharing mechanism must price risks appropriately. Mispricing of risks provides perverse incentives. If the default risk posed by a particular member of a CCP is underpriced, that member will trade excessively, thereby imposing excessively large risks of loss on the other members.

It should be noted that in practice CCPs typically do not charge different members different fees to reflect differential risks. That is, there is no “experience rating” of clearing fees. Instead, CCPs price risk indirectly by choosing collateral (margin) levels and capital requirements. Since the CCP is effectively a risk sharing mechanism, where the risks are not priced directly and “premiums” do not flow from one member to another, a CCP ideally sets collateral and capital levels so that the expected default cost is the same across all members. By doing so, there are no wealth transfers across members. Failure to do so leads to a transfer of wealth from one set of members to another.

Even though CCPs do not price default risk through insurance premia, as a convenient shorthand I will refer to the pricing of default risk by the CCP, with the understanding that this pricing is indirect through the setting of collateral.

Information asymmetries challenge any risk pricing/insurance mechanism. If the insured (i.e., a clearing member) has better information about his risks than the insurer (i.e., the CCP), he will trade too much and hold positions that are too big when his information indicates that the insurance is underpriced, and trade too little and hold positions that are too small when his information indicates that the insurance is overpriced. Asymmetric information problems usually result in incomplete insurance of risks. In the extreme, information asymmetries and the associated adverse selection and moral hazard problems can make risk sharing inefficient altogether.

In the context of a CCP, asymmetric information about the instruments traded and the balance sheets of the dealer firms that trade them–and who would be the members of a CCP–impedes the sharing of default risk. The severity of these information problems can vary by instrument, and by the nature of the firms that intermediate the trading of these instruments. An examination of CDSs in particular, and more exotic derivatives in general, strongly suggests that asymmetric information about both risks is likely to be acute. This tends to reduce the net benefit of introducing a CCP for these products, and may make this net “benefit” negative, especially when the costs of establishing and operating a CCP are considered.

First consider the issues relating to the risks of a particular type of instrument. To price risk, a CCP uses information on (a) the risk-return characteristics of this instrument, and (b) the current price/value of the instrument. Given the current value of the instrument, and holding the (true economic) capital of a member firm constant, the likelihood of default depends on the probability distribution of returns on the instrument and the size of the position. Therefore, margin setting/risk pricing depends on information regarding the price behavior of the instrument. Moreover, information about the current price of the instrument is important. A CCP uses an estimate of market price to adjust collateral. Using an incorrect price leads to an incorrect estimate of the gain or loss on a position, and therefore to an incorrect determination of the risk exposure, and relatedly, the collateral level.

For homogeneous, linear, traditional instruments traded in liquid, transparent markets, a CCP is likely to have information on these variables that is nearly as good as, and perhaps better than, that in possession of its members. For an actively traded instrument (e.g., S&P 500 futures), transactions are numerous and observed, so positions can be marked to current value with no difficulty. Moreover, extensive historical data is readily available to calibrate risk models, and “rocket science” modeling is not necessary to estimate these risks. Thus, for such instruments, centralized clearing is unlikely to face severe adverse selection problems, and risks can be priced correctly. One would expect to observe central clearing for such instruments–and one does.

Things are quite different for exotic instruments. These instruments are traded less frequently, and so current market price information is harder to come by. Indeed, at times, there may be no transactions, so it is necessary to “mark-to-model” rather than “mark-to-market.” Moreover, sophisticated modeling is necessary to quantify and understand the risks of these instruments.

This pricing complexity is likely to be an important issue for credit derivatives, especially on individual names. The products are relatively new. Often they are not traded in high volumes in liquid transparent markets. Marking to market is not a trivial exercise as a result. Moreover, understanding the risks of these things is difficult. This is particularly true for portfolios of these instruments, due to the potential for default dependence between different instruments (i.e., the unconditional risk of default of a single instrument is not sufficient to determine the risk posed by a CCP member with positions in multiple instruments; you need to know about the correlations across instruments, and these dependencies are especially tricky to model and calibrate with credit products).

Big dealer firms specialize in developing models designed to quantify and characterize these risks. Moreover, these dealers expend resources to develop the data to calibrate and test these models. They have strong incentives to develop good models, as with better models they can manage their own risks better, and perhaps generate higher trading profits. Therefore, it is highly, highly likely that for a product like CDSs, dealer firms that engage in proprietary trading of these instruments will have better models and better information than a CCP.

To reiterate something I’ve said in previous posts, this is not to say that these models are great in some absolute sense. The key issue in risk sharing is asymmetry, which depends on the relative quality of information. If a CCP member has a better model than the CCP, there is an information asymmetry problem, even if the former’s model is, well, pretty bad. As I’ve said before, the one eyed man is king in the land of the blind.

(For an illustration at the problems associated with third party’s attempts to evaluate the risks of heterogeneous, complicated, non-linear derivatives, consider the rating agencies’ disastrous experience with CDOs, and monoline insurers.)

Thus, in my view, it is almost certainly the case that for exotic derivatives, and for CDSs in particular, dealer firms that make markets in these products have much better information about their values and risks than would a CCP. This creates the potential for adverse selection, which reduces the benefit of default risk sharing through a CCP.

It should also be noted that big dealer firms that (a) are the primary intermediaries in exotic derivatives and CDS products, and (b) would be the members of any OTC derivatives clearinghouse, are very complex financial firms with relatively opaque balance sheets. They are in the business of providing information intensive intermediation, through their loans, investments, and trading. As a result of this information intensity, it is challenging for third parties to appraise their balance sheets, and quantify the risks associated with those balance sheets. It is certainly the case that the intermediaries (banks and investment banks and in some cases insurance companies) have better information than third parties about the value of the assets and liabilities on their balance sheets, and the risks of those assets and liabilities–including the correlations between those asset and liability values and the values of derivatives positions on its books.

This means that if a CCP’s members are complex dealer firms engaged in information intensive intermediation, the potential for asymmetric information about balance sheet values is acute.

Put these two things together, and it is evident that the formation of a CCP to clear exotic products, including credit derivatives, and which has big financial institutions with complex balance sheets who supply information intensive intermediation, faces far more daunting challenges than a CCP clearing “vanilla” products and which has simpler firms has members. The pricing of risk is almost certainly far more difficult, and more importantly, the potential for information asymmetries is far greater. Thus, adverse selection problems would almost certainly be far more acute if the big dealer firms formed a CCP to share default risks for credit derivatives and other exotic products.

This suggests caution is in order in moving towards a CCP structure in CDS products in particular. Myriad regulators, market participants, commentators, and legislators have seized on central clearing as the panacea for these products. There has been little–if any–recognition in the public debate of the special challenges that these kinds of products and the institutions that trade them would impose on a central clearing mechanism. Thus, it is quite possible that a CCP has not been adopted for these products is not due to the selfishness of dealers who fear that clearing would increase competition for them, but instead reflects a sober appraisal of the greater costs of sharing default risks for these products.

A couple of other points are worth noting. First, traditional CCPs typically do not vary risk pricing (i.e., collateral levels) to reflect the balance sheet risks specific to each member firm. CCPs typically set collateral levels based on the risks of the instruments held in each member’s portfolio of cleared products, and two members with the same portfolio would post the same collateral even if their balance sheet risks are quite different. CCPs do impose some constraints on balance sheets, but normally through capital requirements that are simply a function of collateral levels. For instance, the CME clearinghouse sets minimum member capital equal to a multiple of the member’s margin level. Importantly, it does not vary capital requirements or collateral levels based on assessments of the balance sheet risks of member firms. Thus, in most CCPs balance sheet risks are not priced.

This may not be a big problem when CCP members are relatively simple–and relatively homogeneous–firms with relatively simple and transparent balance sheets. Many futures commission merchants and securities firms that are clearing members fit this characterization. Some are subsidiaries of complex, broad-based intermediaries, but from the perspective of the clearinghouse, the relative simple balance sheets of the actual member firms are what matters.

If big banks and investment banks continue to dominate intermediation of CDS and exotic derivatives, the members of any clearinghouse for these products will inevitably be firms with complex, opaque balance sheets, for which asymmetric information is likely to be quite acute. Simply adopting futures and options CCP practices will not be adequate for this business. Moreover, I have serious concerns that the difficulties in assessing and pricing balance sheet risk will seriously impair the effectiveness of a CCP for credit derivatives.

The second point is that it is incorrect to say that default risks are not shared among dealers in a bilateral OTC structure. Dealers trade with dealers. I daresay that big dealer firms had the biggest exposures to Lehman, Bear, and AIG. Inter-dealer trade means that default risks are shared among the dealers who would be the members of a clearinghouse even in a formally bilateral market.

This has an important implication. Specifically, the choice between a “bilateral” and “centrally cleared” market structure is a choice between different ways of sharing default risks among dealers (and between dealers and customers); it is NOT a choice between interdealer sharing and no interdealer sharing.

Thus, it is imperative to analyze the comparative costs and benefits of these alternative sharing mechanisms. This is a multi-faceted problem, but I will focus on one issue, one that I think is of primary importance.

In a CCP structure, the CCP is responsible for evaluating the balance sheet and product specific risks of each member firm. That is, there is a centralized risk evaluation and risk pricing mechanism. In contrast, in a “bilateral” market, each dealer firm makes individualized assessments of the product specific and balance sheet risks posed by each counterparty. Moreover, since dealers trade with one another, each dealer assesses and monitors the counterparty risks of each of the other dealers.

Centralized monitoring has one clear benefit–there is no duplicative monitoring. Each dealer is evaluated once. In contrast, in a bilateral market, each dealer assesses and monitors every other dealer, meaning that for N dealers there are 2 to the power N monitoring costs incurred.

However, decentralized monitoring likely has a substantial benefit when complex financial institutions trading complex products are involved. These firms interact with one another in many markets, and have various sources of hard and soft information about the balance sheet risks of its counterparties. This information is almost certainly superior to what a clearinghouse is likely to collect. Moreover, since due to the complexity of the balance sheets of the big dealer firms, any individual firm’s assessment of the balance sheet risks of another is likely quite noisy. Multiple signals should be more precise than a single one. Therefore, monitoring by multiple counterparties generates information that is likely to be collectively far more precise than the information collected by any one. If this information can be aggregated in some way, multiple counterparty risk assessments in a bilateral market can be far more efficient than a single risk assessment by a CCP. (The mechanism for aggregation is a complicated subject in itself. Dealers will make inferences from the actions of others. If dealer A cuts back on trading with B, or decides not to trade with B altogether, dealers C, D, etc., will make inferences about A’s information regarding B’s performance risk. One issue that deserves study is whether this aggregation method encourages information cascades that can lead to inefficient outcomes.)

Furthermore, dealers in bilateral markets can–and do–charge counterparty-specific prices for balance sheet risk. Specifically, they require different counterparties to post different amounts of collateral, where variations in collateral levels across counterparties reflect both product-specific risks and the particular balance sheet risks of each counterparty.

(Member-specific collateral levels in a CCP setting would also put tremendous stress on the governance of a clearinghouse. Each member would have an incentive to influence the CCP to impose higher collateral requirements on competitors. Heterogeneity puts stress on collective/cooperative bodies such as clearinghouses. Even though institutional mechanisms such as the non-profit form can mitigate heterogeneity-driven rent seeking in such organizations, it cannot eliminate it, and these rent seeking costs are typically greater when members are heterogeneous.)

This analysis raises the question of whether it is efficient to bundle the trading of these instruments with the supply of information intensive intermediation. Big banks and investment banks with opaque balance sheets (due to the information intensity of their various financing and trading activities) pose greater balance sheet risks than would simple, “vanilla” intermediaries that specialize in trading these derivatives and eschew other activities.

But we don’t see such simple, vanilla, boutique intermediaries. I think that for a variety of reasons this makes sense, as there are likely scope economies between trading things like CDSs and offering other forms of financial intermediation. After all, banks specialize in evaluating credit risk, and the resources and skills used to do this in lending and underwriting can be put to use in valuing and assessing the risk of derivatives, and most particularly, credit derivatives. Moreover, there are scope economies in modeling and risk evaluation. Thus, I think a strong case can be made that bundling of intermediation of complex derivatives (and especially credit derivatives) with the supply of other information intensive financial intermediation activities is efficient.

In sum, before charging ahead with a CCP for credit derivatives and other exotic products, it is worthwhile to step back and ask why this hasn’t happened before now. Sure, it may reflect inefficient strategic behavior by dealers and collective action problems. But it may be the result of high costs of centralized risk sharing for complex products traded by firms who supply information intensive intermediation, and therefore have somewhat inscrutable balance sheets and balance sheet risks.

I have serious reservations about the efficiency of a CDS clearinghouse consisting of big banks/investment banks. Balance sheet risks are a big deal when such firms are involved, and traditional CCPs have not priced these risks (or have priced them in a very, very crude way.) Moreover, the risks of the instruments themselves will be difficult for a CCP to quantify. Indeed, sometimes it will be difficult for the CCP just to determine the market value of these instruments. (Which is why firms like Markit are included in CDS CCP proposals.) Given the instruments, and the firms trading them that would be in a clearinghouse, risk pricing will be extremely difficult, and beset by asymmetric information problems. To me, it is by no means clear that centralized assessment, pricing and sharing of position and balance sheet risks is more efficient than decentralized, “bilateral” assessment, pricing and sharing of these risks.

This argument may seem somewhat Panglossian–to some it may appear that I am suggesting that we live in the best possible world, and intervention to change performance risk sharing mechanisms is therefore unwarranted. I would disagree with that characterization. Suffice it to say that starting from the basics of risk sharing economics, and analyzing the specifics of credit derivatives and other exotic products, a strong argument can be made that a bilateral OTC market structure is efficient relative to alternative structures, including a CCP-based system. To put things in legal terms, I should think that the burden of proof should lie on those who implicitly or explicitly assert that the existing market structure is inefficient compared to a realistic alternative.

And I will close by noting that some practitioners with skin in the game share reservations similar to those I have expressed here. This Bloomberg article by Matt Leising quotes “electronic trading pioneer” Thomas Peterffy’s objections to the CME’s proposal to clear CDS products. Long time Chicago broker Chris Hehmeyer (CEO of FCM Penson GHCO) is also “balking at” this initiative. Peterffy “doubts that the exchange will be able to determine CDS pricing because they trade infrequently.” Although neither Peterffy nor Hehmeyer (or David Rutter, also quoted in the article) mention the very different balance sheet risk posed by big CDS dealers, I wouldn’t be surprised if that’s a concern as well. Though they aren’t quoted as saying so, these gentlemen have to realize that when the big dealers clear CDS products, the other members of the clearinghouse–firms like theirs–not only bear the risks associated with the CDSs, but also the risks associated with subprime, or CDOs, or whatever happens to be floating the banks’ boats at the moment. This cannot be a comforting thought.

So, before we rush in to create a CDS clearinghouse, it behooves us to examine carefully the risks of doing so. Not all instruments and not all intermediaries are created equal. Some are well-suited for the sharing of default risk through a central clearing mechanism. Others are more problematic. Credit derivatives are among the most problematic of all. Creating a central counterparty without a full assessment of the unique challenges that these products pose is itself a risk–a risk of a train wreck somewhere down the line when the mispricing of default insurance encourages excessive risk taking, and a member (or members) of the clearinghouse defaults. Conditional on the probability of default, a central counterparty is probably the efficient way to bear risk. But if the CCP prices risk wrong, it could increase the probability of default and actually create more problems than it solves.

Do I know this will happen for certain? No, I don’t. But I do think there is considerable basis to conclude that it is a serious possibility. So before venturing down the CDS CCP road, it would be advisable to (a) give this issue serious attention, and (b) design the CCP to mitigate these problems.