I’ve been beavering away at the clearing section of my next book, which will be titled Market Macrostructure. It’s been something of a struggle, because there are so many aspects to this issue that it is challenging to organize the material in a logical fashion. I analogize it to trying to write a history of a complicated battle, where many things are happening simultaneously over space, and these things interact. Inevitably, the narrative must jump around in either space or time or both, and the writer must summarize some material about one action at one time to relate it to the narrative of what is going on at another place at another time.
So it is with writing an analysis of clearing. There are many moving parts that interrelate and interact, so there is always the challenge of relating detailed analyses of important pieces to one another, and to clearing as a whole, and to the trading process (execution and clearing) as a whole.
One virtue of this struggle, however, is that it can lead the writer to new insights. Conceptualization at a fairly highly level facilitates organization, and the process of conceptualization can lead to new ways of understanding. So it is, I hope anyways, with the clearing analysis.
This post is my first attempt to crystalize those thoughts; indeed, one of the original purposes of the the blog was to provide a place where I could think out loud and commit to pixels some early thoughts to be refined going forward in more formal writing.
The issue that I have been grappling with is why are some trades cleared, and others not? Why do we see exchanges that trade cleared contracts operate side by side with bilateral markets trading similar contracts (and sometimes nearly identical ones) that aren’t cleared? What determines the division of trade between these alternatives?
This is an issue that I’ve been looking at since the ’90s, and I identified some factors, but I was never completely satisfied. But putting together some pieces that I have written about before, I think I’ve come up with a more complete explanation that captures some of the salient aspects of the economics of clearing and counterparty risk generally.
The first piece is that in theory-and indeed, in most theoretical treatments of clearing by academics (including yours truly)-clearing can operate as a classical risk pooling/insurance mechanism, in which market participants pool counterparty risk. To the extent that this risk is idiosyncratic, such pooling allocates risk more efficiently and makes risk-averse participants better off.
But as I pointed out in my earliest work, and emphasized more in some later papers, like any risk sharing arrangement, mutualization of counterparty risk creates the potential for adverse selection and moral hazard problems. Moral hazard problems are likely to be particularly acute. Clearing participants can affect the distribution of the default losses they impose on the mutualization pool by adjusting the riskiness of their trading positions in the cleared derivatives. They can also do so by adjusting the risks of their balance sheets, through, for instance, adjusting their trades in non-cleared derivatives (or in derivatives cleared at another CCP), changing their leverage, or adjusting the risk of other assets on their balance sheets
The prospect for moral hazard will inevitably lead to limits on the amount of insurance provided through the clearing mechanism. That is, not all default risk will be mutualized.
Clearinghouses use margins to limit the amount of risk that is mutualized. Only losses on defaulted positions in excess of margin posted by the defaulter are mutualized. The higher the margin cover, the lower the level of risk sharing.
In practice, CCPs utilize a “defaulter pays” model in which margin covers losses on defaulted positions with extremely high probability, e.g., 99.7 percent of the time. In a defaulter pays model, the amount of risk mutualization is very low. CCPs are not, therefore, primarily an insurance mechanism. They insure only tail risks (which has important implications for systemic risk and wrong way risk).
Note LCH.Clearnet’s boast that it had collected far more margin than necessary to cover the realized losses on the Lehman’s derivatives positions that it cleared. The CME also had more than enough Lehman margin to cover losses on its positions (although there were shortfalls on some product segments that were covered by excessive margins on others).
At the Paris conference I attended in September, the head of Eurex Clearing, Thomas Book, was adamant that the goal of his CCP, and of CCPs generally, was to avoid mutualizing risk if at all possible. His answer surprised Bruno Biais, whose model of clearing in the paper he presented focuses on the role of the CCP as a default risk insurer.
I confess that I have been inadequately appreciative of this point as well. Understanding its implications has important consequences, as I hope to show in a bit.
To summarize. CCPs generally operate on a defaulter pays basis: this is also sometimes referred to as a “no credit” system. That term will help illuminate the differences between cleared and uncleared markets. The defaulter pays system means that the amount of risk shared through a CCP is extremely limited. This limitation on risk sharing is best explained as the consequence of moral hazard.
In contrast to a cleared market operating on the no credit model, dealers in bilateral OTC markets historically extended credit to derivatives counterparties. Put differently, OTC deals often bundled a derivatives trade with credit provision. That is, dealers often willingly took on exposure to a default loss when entering into a derivatives deal with a customer. (This is not true for all types of customers. For instance, hedge funds typically had to post margin.) Taking credit exposure is equivalent to extending credit to the counterparty.
Now consider whether a firm will prefer to trade a cleared derivative, requiring the posting of a high margin and which thus embeds no credit, or prefers instead to trade an otherwise identical OTC product that does embed credit. To fix ideas originally, let’s consider a firm that is cash constrained. Therefore, if it wants to trade the cleared product, it must borrow to fund the initial margin. The cleared derivative is a no credit transaction, but that doesn’t mean that moving to clearing necessarily reduces the amount of credit the firm obtains: it can borrow the money needed to post margin, and indeed, may have to borrow it.
One source of credit is dealer banks. So the firm could either enter into an uncleared bilateral trade with a dealer, or borrow money from the dealer bank to fund the IM. (The argument doesn’t really depend on dealing with the same bank on the derivatives deal and the borrowing: this just facilitates the exposition.)
Here’s were the loser pays aspect of margin comes in. Let’s say that the firm is selling a derivatives contract with payoff P. If the firm defaults, the OTC counterparty’s exposure is max[P,0]. But in a loser pays model, the margin M is almost always greater than max[P,0]. Thus, the borrowing from bank to fund margin almost always exceeds the default loss that the bank would incur if it entered into a bilateral deal with the firm.
I can show formally that if the firm already has debt outstanding, under standard pro rata/pari passu default loss allocation mechanisms, holding everything equal, the bank’s default losses if it extends credit to the firm to fund margin almost always exceed, and never are smaller than*, the default losses that it would incur if it had entered an uncleared bilateral trade with the firm. This, in turn, will make the cleared transaction more expensive for the cash-constrained firm, and it will prefer to trade the OTC product.
There are at least a couple of reasons why the cleared transaction can be more expensive. One is what is effectively a debt overhang problem. In order to induce the bank to lend the margin for posting at the CCP, the firm must promise it higher payments in non-default states than it has to promise the bank in these states when it trades OTC instead. Since the firm’s managers, acting in the interest of equity, only care about payoffs in non-default states, this means that returns to shareholders are lower when it borrows to fund margin than when it deals OTC. This can be seen another way. I can also show formally that the payoffs to the firm’s other creditors are almost always higher, and never lower, if it borrows to fund margin than if it trades OTC. Thus, the value of the the firm’s non-margin-related debt is higher if it trades a cleared product and funds the margin by borrowing, than if the firm uses the OTC product. Since the value of the firm’s assets doesn’t differ in the cleared vs. uncleared cases, and since value is conserved, this means that equity is less valuable when the firm trades cleared products than bilateral ones. Some of the benefit of borrowing to fund margin flows to other creditors; this is where the analogy to debt overhang comes in.
This is most easily seen in the following scenario. The firm can become insolvent when max[P,0]=0, i.e., the bilateral contract is out of the money to the bank, and the bank suffers no loss due to default, and all the losses of insolvency would fall on other creditors. However, if the bank had lent the firm money to fund margin, it would suffer a loss on the margin loan in this circumstance. This loss would reduce the loss suffered by the other creditors.
OTC is cheaper than cleared products in other models of capital structure. For instance, moral hazard (or adverse selection) mean that the firm will be credit constrained: the amount it can borrow is limited by its collateral, and/or the amount of cash flows that it can credibly pledge to lenders. The same formal analysis implies that more cash flows in non-default states must go to supporting a margin loan in a defaulter pays clearing model than in a bilateral transaction. This leaves less cash flows to support other borrowing, so by borrowing to fund margin loans the firm must borrow less to support other investments (which, in this sort of model, it has insufficient equity to fund itself). Thus, borrowing to fund margins on cleared transactions crowds out borrowing to fund positive NPV investments.
This analysis implies that this cash constrained firm will choose to trade OTC rather than cleared products with defaulter pays margins funded with loans. The bank is indifferent, because it will price the product or the loan to cover its costs, but the firm is always better off with the bilateral trade because it has to pay the bank less if it trades OTC than if it borrows to fund margin.
Moreover, the analysis implies that if the firm is forced to clear, it will either scale back its derivatives trading (because the cleared transaction is more expensive), and/or reduce its investments in positive NPV projects. Cutting back derivatives trading is costly if this trading reduces the deadweight costs of debt, for instance. Indeed, I can show that the cost of margin is especially high when the derivatives trade is a “right way” risk, as would occur when the firm is hedging.
Clearing mandates are therefore expensive for such firms, and there is a legitimate reason to exempt such firms from clearing requirements. (Note that even if these firms are exempted, other rules affecting dealer banks, e.g., punitive capital charges on OTC derivatives trades, can induce an inefficient use of cleared transactions.)
This analysis explains the preference of many firms, especially corporate end users, for uncleared OTC trades that embed credit, as opposed to cleared transactions that must be funded by increased borrowing. This, in turn, can explain the growth of OTC markets relative to exchange traded markets from the 1980s onwards.
It is useful to step back a bit here, and understand what is really going on. In essence, in this model clearing is expensive because it causes one agency problem to exacerbate another. CCPs adopt defaulter pays because of an agency problem: the moral hazard associated with risk sharing. Debt is expensive or constrained for firms because of agency problems (e.g., debt overhang problems, or constraints on borrowing due to moral hazard). Effectively, the firm must obtain more credit to support a cleared position than an uncleared one, and the cost of debt arising from agency problems makes this higher level of credit more expensive.
This analysis raises the question of why firms would ever choose to clear. There are a couple of answers to that.
First, there may be other (private) benefits. For instance, netting economies may be greater with clearing. Of course, the efficiency effects of netting are equivocal (because netting primarily has the effect of redistributing losses among creditors), but as a positive matter is is pretty evident that netting offers private benefits, and thus the mulitilateral netting that can occur in clearing, but not to the same degree in bilateral trades, could induce some traders to prefer clearing.
Second, the analysis started from the assumption that the firm at issue is cash constrained and hence has to borrow to fund margin on the cleared trade. Some firms are not. One example would be an ETF like USO or USNG, which collect the entire notional value of derivatives in cash from their investors. These firms do not require any credit to meet margins. Large real money funds, like a Pimco, that collect cash from investors and use derivatives to gain exposure to price risks would be another.
Even many hedgers may not suffer from cash constraints that limit their ability to trade cleared contracts. Consider commodity trading firms. They typically use derivatives to hedge inventories of commodities. Banks, in turn, are willing to lend against these inventories as collateral. Thus, the commodity trader can fund margin using borrowings secured by commodity inventories, and the lender does not share in default losses pro rata with other creditors. This type of borrowing is fundamentally different than the borrowing considered above, in which the firm borrows against its balance sheet and default losses are shared with other creditors.
Thus, the simple models would predict that whereas cleared derivatives used to hedge liquid inventories are as cheap or cheaper than uncleared derivatives, it is much more expensive to use cleared derivatives to hedge cash flows on illiquid assets, or to hedge broad balance sheet risks. This is largely consistent with my understanding of the pattern of usage of cleared and uncleared derivatives.
This model, which combines a model of the cost of risk sharing at a CCP with a model of the capital structure of firms, has both positive and policy implications. In particular, it can explain the adoption of defaulter pays by CCPs. It can also explain the disparities between OTC and cleared markets when CCPs utilize defaulter pays. Moreover, it demonstrates that clearing mandates can be inefficient if they are applied too broadly. One source of this inefficiency is that the mandate leads to a perverse interaction between agency problems.
Of course, a rationale for clearing mandates is that clearing reduces systemic risk. Anyone who has read my work will know that I am dubious of that rationale on many grounds, but it is worthwhile to consider the implications of the foregoing analysis for systemic risk.
The model is too sparse to make very strong conclusions, but one consideration does stand out. If firms borrow from OTC dealer banks to fund margins, holding the rest of the firm’s liabilities and assets constant, these banks suffer larger losses when the firm goes bankrupt if they lend to them to fund margins on derivatives trades than if they enter into identical uncleared OTC derivatives trades. As noted before, there is a distributive effect: other creditors suffer smaller losses. The systemic implications of this redistribution depend on the relative systemic importance of the banks and the other creditors. Dealer banks are definitely systemically important, but other creditors may be too. Therefore, it is not evident how this cuts, but if one believes that large financial institutions that serve as OTC dealers are especially crucial for systemic stability, moving to cleared trades would tend to increase systemic risk because clearing actually increases their credit exposures to customers.
But again, caution is warranted here. The redistribution result holds capital structure and derivatives trades constant, but of course these will be different if firms have to clear than if they don’t. These changes are difficult to predict, and the systemic riskiness of different configurations is even more difficult to compare given how little we really know about the sources of systemic risk. But the fact that clearing can lead to adverse interactions between agency problems should raise concerns about the systemic implications of forcing clearing.
*To be more precise. When P>M, the CCP and the lender suffer default losses, and the sum of these default losses is the same as the bilateral counterparty would incur on an uncleared trade. Relatedly, other creditors of the bankrupt firm suffer the same default losses in the cleared and uncleared cases when this condition holds. They suffer smaller losses whenever this condition does not hold and the firm goes bankrupt.