Segregation vs. Omnibus Accounts: Discretionary Credit vs. Committed Credit
Content warning!: this is a post about clearing minutia. But believe it or not, there are a select few who actually like that kind of thing–so this one’s for you!
I was thinking more about the segregation issue, and particularly the credit and systemic risk issue. I think I got the essence of the conclusion right, but was not completely happy with the supporting analysis, so I gave it some more thought and came up with something that I think is a more robust.
The issue is whether segregation or futures-style omnibus accounts lead to more market fragility during times of stress. My original intuition, and my newer analysis says that segregation is more fragile in times of stress; that is, a CCP is more likely to fail with segregation than with omnibus accounts. The basic reason is that segregation is more dependent on discretionary credit during times of stress; this dependence is a source of fragility.
The potential problem has its origins in the timing of cash flows in a CCP setup. CCPs make margin calls on brokers (FCMs) that the FCMs have to meet in a short time period, on both house and customer accounts. Often the FCM has to make payment on the customer account before the customers have made their variation margin payments; to meet the CCP deadline, therefore, credit has to be extended to these customers. The FCM’s bank may make the payment to the CCP even though there are insufficient funds in the FCM’s account at the bank to cover the payment. That is, banks for FCMs often extend credit to the FCMs to permit timely payment of margin calls to the CCP.
With omnibus accounts, the payment that the FCM owes on its customer account is the net payment owed to the CCP by all its customers on all their positions cleared through that CCP. This means that the maximum amount of credit that is needed to ensure that the bank makes the margin payment to the CCP on behalf of the FCM is this net payment owed.
With segregation, it won’t be possible to net customer payments against one another. This increases the amount of credit banks have to extend to ensure prompt payment of variation margin to CCPs. And that is where the vulnerability arises.
A very stylized example demonstrates the point. It is unrealistic in many ways but the elements of unrealism don’t overturn the basic result. Assume an FCM has zero capital. It has 10 customers, 5 who have to pay variation margin of 1000 apiece, and 5 who should receive 900 in variation margins apiece. Thus, the net amount owed on this FCM’s customer account is 500=5x(1000-900).
Assume that the margin payment is due immediately, but none of the customers who owe have posted the necessary cash. With some probability all will stump up the necessary cash in the future, but with some probability none will. In the latter event, there is a total loss of 5000 that is borne by others in the market: just who bears it depends on the type of account (0mnibus or segregated).
With omnibus accounts, the FCM owes the CCP 500. If the FCM’s bank lends the FCM the 500, it can meet the margin call and there is no default to the CCP. If all the customers who owe pay, the FCM can repay the bank, and the bank and the FCMs other customers get paid all they are owed. However, if the customers who owe don’t pay, the FCM defaults on its loan to the bank for 500, and also cannot pay its customers who are owed money the 5×900=4500 that is due them.
Note that in this example, the risk that the customers who owe variation margin payments will default is borne by the bank and the other FCM customers. That’s the “fellow-customer” risk in this setup.
Now consider the situation with segregation. Here, it is not possible to net customer pays and receives. Thus, it is necessary to provide 5×1000=5000 in credit to the FCM, or the FCM’s customers, to make it possible to meet the CCP deadline. If the customers end up paying what they owe, everything is the same as with co-mingling. If they don’t, however, the bank loses 5000–if it decides to extend the credit.
The total credit loss in the two alternatives is the same–5000–but the distribution is different. The bank is on the hook for 100 percent with segregation, but only 10 percent under omnibus accounts–again assuming that it extends credit in each case.
That is a crucial difference, because the bank has the discretion to extend credit: that is, it is not a given that the bank extends credit. It can say no. If it says no, the FCM goes into default on its obligation to the CCP, and the CCP is on the hook for the loss in the event the customers who owe default. This loss falls on the CCP’s capital, or on its default fund. This could conceivably threaten the CCP.
Clearly, the bank is more likely to say “no” to a request to borrow 5000 than a request to borrow 500. Note particularly this situation is likely to arise when the market is under stress. That’s when big price moves occur that can raise doubts about the ability of customers to meet their margin calls, and when excess margins held at the FCM are going to be blown through. It’s also when banks are likely to be more averse to taking on credit risk.
With omnibus accounts, there is credit extended, but much of it is not discretionary. In the example, the FCM’s in-the-money customers provide 4500 of credit. They have committed to do so by being customers of the FCM, and being on the winning side of their trades. In effect, customers provide irrevocable letters of credit to other customers–and in return, receive irrevocable LCs from those customers. These are not traditional LCs because their payoffs are conditional on market prices (and in a more realistic example, on the financial condition of the FCM). But in essence, if you are a customer of an FCM under an omnibus system, you commit to provide credit to other customers under some circumstances, and in return receive a commitment from other customers to grant credit to you under other circumstances.
That lack of discretion is crucial in preventing a market breakdown. “Freezing” of credit markets during times of market stress essentially means that lenders choose not to lend. The more that you rely on the discretion of lenders, the more vulnerable you are to a market freeze.
During the Crash of 1987, the reluctance of banks to lend to FCMs was what threatened the major CCPs. Consider that in the context of the example. In the example, under segregation, the bank is less likely to extend credit, making an FCM default more likely. Moreover, since there is no sharing of risk among customers, with segregation the CCP is on the hook for a bigger loss: 5000. With omnibus accounts, it loses only 500. CCP failure is obviously more likely when it has to bear a 5000 loss than one a tenth as large. Thus, with segregation, there is a larger probability that the CCP will absorb a loss, and conditional on a loss occurring, that loss is bigger. This means a greater likelihood of CCP failure–or the need for greater CCP capital.
In brief, the clearing system is more likely to break down during periods of market stress, the greater the reliance on discretionary credit to meet variation margin calls. This reliance is greater with segregation, than with futures-style omnibus accounts. The latter reduces reliance on discretionary credit by the use of (implicit) commitments among FCM customers to lend to each other. That makes the system more robust.
There is considerable symmetry in the commitments that FCM customers make, but that symmetry is not complete. Some customers are more creditworthy than others. They essentially subsidize the less creditworthy customers (unless, as I discuss below, commissions or other terms offset this effect). Moreover, tail risks that create default risks are not symmetric. Customers who are long stock index futures, for instance, are likely to experience big margin calls than those who are short because crashes are more likely than spikes. Customers who are short CDS are more likely to experience big margin calls than those who are long because of jump to default risk.
Even though the exchange of commitments is not necessarily of equal value (i.e., a given customer may provide a commitment that is more valuable than the commitment he receives from the other customers due to these sources of asymmetry), there is an exchange of value. Moreover, more creditworthy customers should be able to negotiate better commissions to reflect the benefits they are providing to other customers–which helps the FCM attract the business of these other traders. The ability to negotiate commissions and other terms of agreements between customers and FCMs (e.g., margins) reduces the scope for cross subsidies between customers. This limits the distributive consequences of omnibus accounts.
These credit issues are likely to arise only in rare circumstances, but those are exactly the circumstances in which a market breakdown and CCP failure would occur. A liquidity crunch occurs when due to heightened risk and asymmetric information, lenders choose not to lend. This is most likely to occur during periods of market stress–which is exactly when big margin calls increase the demand for liquidity/credit. That is, stressed conditions reduce the supply of credit and increase its demand. It is desirable to reduce reliance on discretionary credit in these circumstances. Omnibus accounts do just that. They are a liquidity pooling mechanism by which customers implicitly commit to extend credit to one another. This commitment reduces the need to rely on discretionary extensions of credit by banks in order to raise the cash necessary to make payments to CCPs. This, in turn, reduces the likelihood of a CCP collapse.
This analysis strongly suggests that the systemic risks of segregation are far greater than the CFTC acknowledges. Indeed, in its NOPR on OTC derivatives segregation, the agency argues that segregation reduces systemic risks. Its analysis fails completely, however, to consider the implications of segregation on the demand for credit during times of market stress. The foregoing analysis implies that segregation leads to higher demands for liquidity and credit precisely under the conditions in which liquidity and credit are likely to be in short supply. That, in turn, means that the possibility of market breakdown or CCP failure is greater with segregation.
Market participants will adjust on other margins, of course. CCPs will presumably require greater margins and greater default fund contributions with segregation: these are also essentially precommitments of resources that substitute for the precommited resources lost under segregation. They are also costly, and it is nigh on to impossible to know whether these costs are larger or smaller than the costs inherent in a system based on omnibus accounts.
But apropos my earlier post, CCPs, exchanges, FCMs, and their customers have strong incentives to adopt the system that minimizes these costs. They have skin in the game, and a bigger pie to split when the choose the lowest-cost mechanism. They also have better information to evaluate these costs than any regulator. No, their incentives aren’t perfect. Their information isn’t perfect either. But perfect isn’t an option, so those aren’t meaningful objections to relying on competition and contract to determine the choice of institutions, rather than regulatory fiat.
@streetwiseprof Unfortunately, as your piece highlights, the terminology linked to some of the details has become ambiguous and confusing, thereby harming the current debate.
It may be easiest to consider the matter under three headings
– Gross v Net margining
– collateral segregation
– “customer loss mutualisation”
“Gross v Net Margining”
When talking about “Omnibus”, today we can be referring to a)Net omnibus where the obligations and positions of all clients are offset, which was commonly the model in European Futures CCPs; or b) Gross omnibus, the model adopted by OTC CCPs which entails collecting gross margins by the CCP on a “net portfolio” per customer.
You’ve correctly highlighted that the Net Omnibus model provides funding benefits to the Clearing Member by allowing the offset of margin requirements between members, whilst still providing for the clearing member to collect the gross margin from each customer. Traditionally this has been the source of additional revenue for clearing members [interest earnings] as well as additional funding to the extent that the margin was provided by customers who didn’t demand segregation from their clearing members [permitted outside the US]. However, in general, Net Omnibus is not a model being operated in the OTC arena and nor is it being anticipated by the CFTC.
As such, liquidity is going to be a key issue for Clearing Members and one that I believe regulators have largely ignored in their reshaping of the OTC landscape, especially coming on top of the liquidity/solvency demands of other new regulations.
The model currently favoured by the CFTC of “Legally segregated, operationally co-mingled” or “LSOC” model, first developed at Swapclear, is based on gross margining.
“Collateral Segregation”
Separately, when discussing segregation, we can be discussing loss mutualisation between customers which is a present feature of Futures markets under certain circumstances you describe above. Yet it can also be describing how collateral is held eg legally separate from the CCP & clearing members and also from under customers.
In general, it is agreed that segregation of customer assets from clearing members is important in order to avoid customers having a credit exposure to their clearing member. A separate consideration is whether one needs to hold customer assets on a basis that is legally separate both from the CCP and other customers. In the case of separation from the CCP, the question to consider is to whom one wants a credit exposure and the related capital implications of that. Notably, the capital charges of an exposure to a CCP are far less than to a credit institution, which is whom one would have a credit exposure if placing cash collateral on a segregated basis. Only cash collateral placed with a central bank would attract a lower charge than a CCP, a proposal that has been made in some European quarters.
A further twist on “segregation” has been whether collateral placed by end users has to be fully traceable to their CCP account or whether clearing members may perform collateral substitutions eg customer provides a US Treasury bill, but clearing member posts cash to the CCP. The operational challenges of providing such traceability are considerable, with the solutions for customer typically being unpalatable eg advance posting of collateral.
“Loss mutualisation”
The main debate in this area surrounds the participation by customers in loss mutualisation. To the extent that customers funds will be included in the default waterfall in the event of their own clearing member defaulting as a consequence of a customer default, so this lessens the burden on other parties. Removing this contribution, by definition, increases the burden elsewhere either in the form of pushing up margins or increasing the requirements on the default fund. Notably, this aspect of loss mutualisation is independent of how collateral is held.
Unsurprisingly, customers are in favour of no loss mutualisation, whilst those who would pick up the burden of such an arrangement are arguing against. Aside from cost implications, moral hazard is cited in that customers will have no incentive to perform due diligence on their clearing members. What such an argument ignores is that client confidentiality typically prevents any customer having information on who else uses the same clearing member, thereby impeding adequate scrutiny.
“A Better Model”
Personally I believe all end users of CCPs should participate in loss mutualisation but of the market as a whole by making default fund contributions, for many reasons including
– clearing is a market utility and as such all beneficiaries should participate in supporting its costs, which includes loss mutualisation, as well as enjoying its’ benefits
– by only requiring customers to post initial margin and not default fund, they fail to pay the full costs of the risk they introduce to the CCP and thus enjoy leverage over direct clearing members. Moreover, this creates a subsidy from clearing members to end users. Although funding costs of default fund contributions can be passed on, any losses normally cannot.
– Not requiring end users to contribute to the default fund incentivises them to argue for lower initial margin and higher default fund. It is for this reason that the argument about governance has been so strained ie why should those with no skin in the game get a say. If end users do contribute to the default fund, so their participation in governance would be entirely reasonable and uncontentious. Moreover, it would remove their incentive to favour the form of risk cover.
– Portability is frustrated by default fund contributions, placing a huge burden on incoming clearing members to post additional liquidity to the CCP when accepting new clients from a defaulting member at a time of great market stress. If customers funded default fund contributions on their own positions, so their own porting would not create any burden on an incoming member.
– By placing the burden of default fund contributions on clearing members, so the costs of default is concentrated into a small number of institutions rather than being more widely dispersed as it was under a bilateral model. This creates greater systemic risk.
– BY having “skin in the game”, end users would be incentivised to scrutinise CCPs more closely and not solely consider which CCP is charging the lowest initial margin, which is what is happening today
Inevitably, all CCPs would have to operate on this basis in order to avoid arbitrage between CCPs. Yet at a stroke it solves some many of the current arguments in the industry, creates a level playing field and properly distributes to costs to all beneficiaries of clearing.
Comment by John Wilson — June 18, 2011 @ 6:04 am
@John. Thanks for the extended and thoughtful comment. Apologies if my ambiguities re terminology are sowing confusion, rather than reducing it.
A lot to chew on, but a couple of quick comments.
First, I agree completely that “liquidity is going to be a key issue for Clearing Members and one that I believe regulators have largely ignored in their reshaping of the OTC landscape, especially coming on top of the liquidity/solvency demands of other new regulations.” Indeed, what I am trying to work through–and given the complexities, it is a lot of work and I can’t say I have it right with any confidence–are the implications of clearing mandates for liquidity demand during times of market stress. That’s the goal of the post (not a complete analysis, but a provisional one of just one piece of the puzzle, there being many other pieces). I think that is a major, major issue–in fact, the most important issue relating to the systemic consequences of clearing mandates–but one which you correctly note regulators have studiously ignored.
As I see it, the crucial issue is that during periods of market stress, variation margin calls lead to big spikes in the demand for liquidity. The question is how those are going to be accommodated/supplied. What I’m grappling with in the post is how treatment of collateral and risk sharing rules will affect that. Moreover, in a cleared environment, there is little flexibility in terms of timing to respond to those spikes: they have to be accommodated on a fixed time schedule or there is a risk that a CCP will go into default. My view (of long standing–I wrote about this in the late-90s) is that precommitted liquidity is essential. It seems to me that the loss mutualization mechanism in which customers are part of the waterfall provides a mechanism for precommitting liquidity: customers effectively precommit to lend to one another.
Second, re loss mutualization. I like your model at first glance–I’ll kick it around some more. But here’s what strikes me. It brings us close to full circle. After all, the primary reason for clearing is to shift default risk from customers to financial institutions who are the residual risk bearers in the CCP. But as I think we’ve both realized, that has some adverse effects in terms of incentives (skin in the game). So your proposal shifts more risk back to the customers. The distribution differs from that in a bilateral market, but the direction is clear. One wonders about the prudence of setting out on a journey when one soon recognizes that considerable backtracking is necessary.
Third, I’ll take some issue with your pushback on moral hazard. The perfect is the enemy of the good. Yes, confidentiality limits the amount of information FCMs can disclose about their customers to other customers. But customers can still monitor–hell, even seeing who else is on the golf outing can be informative–if they have the incentive to. They can monitor process to some degree by evaluating how the FCM deals with them. There is all sorts of soft information people can pick up. No, the monitoring is noisy, but as I note in the post, the collective informativeness of multiple noisy signals can be pretty high, and the possibility of customer runs can impose a strong discipline on FCMs: the wisdom of crowds, if you will. The incentive to lift a finger to collect any signal is zero when collateral is perfectly secure, and similarly there is no incentive to run. There is an extensive literature in banking that shows that customer monitoring with even noisier information can impose costs on high risk banks. The same is true with FCMs too, IMO.
Thanks again for the comment. I’m sure we’ll be continuing the discussion.
I had assumed that the clearing members’ Default Fund would not simply come from their (as it were) start up capital, but would be a charge on their clients (so, the buy side participant will have to both put up some set of assets to act as a collateral fund and would also have an additional charge (membership of the DCM?)). That additional charge would go to funding the Default Fund …
But I’m also conscious that I haven’t read or heard anything about this and it is very much my personal assumption without much real eveidence.
Comment by Hugh Griffiths — July 26, 2011 @ 2:06 am
@Hugh. Whilst a clearing member may seek to pass on the funding costs and capital cost, a member can neither use client assets to fund its contribution, nor share losses it incurs from use of the default fund by the CCP.
Comment by John Wilson — July 31, 2011 @ 11:14 am
[…] Craig Pirrong, argues for co-mingling: https://streetwiseprofessor.com/?p=5227 […]
Pingback by Segregation models and Eurex | The OTC Space — February 7, 2012 @ 10:56 am
[…] in the summer of 2011, I wrote that segregation could make the markets more fragile, because it would tend to reduce credit (mainly intraday credit) used to finance variation margin. […]
Pingback by Streetwise Professor » Pick Your Poison — October 16, 2012 @ 3:40 pm