Streetwise Professor

Apparently not having enough on its plate, the CFTC has seen fit to vote out a NOPR on collateral segregation for cleared swaps.  The Commission decided to jettison the omnibus customer account model used successfully for donkey years in futures markets, and to require OTC CCPs to segregate collateral in their books and records, while permitting comingling of customer moneys in one account.  This legal segregation/operational comingling of customer collateral is intended to protect customers against “fellow-customer risk,” i.e., the risk that  the margins of non-defaulting customers of a futures commission merchant (FCM) that (a) has defaulted, and (b) has customers in default will be used to meet shortfalls in the defaulting customers’ accounts.

I wrote about this last year in a post titled “Alfred E. Newman Chooses an FCM: the Moral Hazard of Segregation.”  I argued that it is necessary to understand why futures exchanges would choose–and stick with–the omnibus model that exposes customers to the risk of each others’ default.  The most likely answer is that it provides customers with an incentive to monitor their FCMs.  FCMs that are undercapitalized, or which do not manage risk properly, pose a greater risk customers.  By giving customers skin in the game, they have an incentive to monitor the FCM’s capitalization and the adequacy of its risk management.  With segregation, customers can play Alfred E. Newman: “What?  Me worry?”  They have zero, zip, nada incentive to monitor their FCM.  Their money is safe even with an undercapitalized, or even crooked, broker.  This lack of customer monitoring leads to the survival of riskier FCMs, thereby increasing the risk in the system.

The CME and ICE Trust made a similar argument in letters to the CFTC.  Their letters apparently, uhm, cut no ice with a majority of the commissioners.  The CFTC evaluated the argument and rejected it.  In a nutshell, the Commission bought into the argument advanced by big money managers and pension funds that they don’t have the information to monitor FCMs.

This assertion has problems on many levels.  Yes, customer information about the safety and soundness of FCMs is imperfect.  They can’t evaluate that risk perfectly.  But they can see things like whether the FCM is overcapitalized, and by how much.  They can get some kind of idea as to what kind of customers the FCM serves.  They can observe first hand how the FCM manages the risk of that particular customer: does the FCM insist on prompt payment of margins?  Does it collect and evaluate information on the customer’s creditworthiness?  To the extent that a customer perceives that an FCM is slacking off with him, he can infer that’s a more general problem.  So they can–and do–observe signals on FCM safety and soundness.

And note that there is monitoring by multiple customers.  Each one might observe a noisy signal of FCM creditworthiness, but collectively they may get a relatively precise signal.  The literatures on global games and bank runs suggests that there are mechanisms whereby customers who decide whether to patronize an FCM based on a noisy signal of its riskiness can provide effective monitoring.

A sketch of a model would go something like this.  Customers observe a noisy signal on FCM riskiness.  They know other customers are observing noisy signals.  Each customer adopts a cut-off rule that says “I leave if I get a signal that the riskiness of the FCM rises above a certain level.”  In equilibrium, when enough customers get a bad signal, they run.  Sometimes that’s inefficient because they are fooled by bad draws of the signal.  But with enough customers, that’s relatively unlikely.  Instead, runs tend to occur when the FCM is in fact risky.  The prospect of being destroyed in a run gives an FCM an incentive to control its risk.

Things appear to work this way in practice.  Runs on FCMs often precede revelations that there is something seriously remiss.  There was a run on Refco before it imploded.  Ditto Lehman:

Some argue that the futility of relying on risk mutualisation in the default waterfall was illustrated by the fact that Lehman’s omnibus customer account dropped by roughly 75 percent during the week prior to its default. However, others say this proves the current system works.

“It shows market participants have the incentive to monitor the financial health of their FCM. Moving to an LSOC regime may well undermine this incentive,” said Taylor at the CME.

Exactly, Kim.

There is also empirical evidence that monitoring of relatively opaque financial institutions does occur when there is an incentive to do so.  In the 19th century, notes of different banks sold at varying discounts to reflect creditworthiness, and market participants would often run on banks that were perceived to be unduly risky.  Evidence presented by Calomiris and Gorton and others shows that these runs were not Diamond-Dybvig “sunspot” runs, but reflected bad fundamentals.  Rates in the Fed Fund market depend on the creditworthiness of the borrower.  Similarly, prices on CDS depend on creditworthiness of the counterparties.

As I noted earlier, big money managers and pension funds were pushing for segregation–indeed, greater segregation than the CFTC actually settled on.  Their protestations of their inability to monitor should be taken with a grain of salt–and would be quite alarming if true.  No, I think that the main motive is that they do in fact incur costs to monitor FCMs, but that other customers effectively free ride of their monitoring.  Segregation allows them to play Alfred E. Newman, and not worry about their FCMs risk–and thus not have to incur the costs of monitoring.

If the foregoing analysis is correct–and the CFTC’s arguments certainly don’t convince me otherwise–the proposal will increase total risk in the system, and shift costs from one kind of customer (the money managers and FCMs) to other customers.  This will come in the form of higher initial margins and higher default fund contributions because the inability to mutualize risk among customers requires shifting that risk elsewhere.  It will also come in the form of greater–and perhaps much greater–operational costs.  But note that these will all hit the big money managers and pension funds less than smaller entities.  They typically have lower funding costs (and hence are less impacted by the margin increases) and since many of the increased operational expenses are fixed costs, they have a bigger impact on the smaller firms.

Note too there are other possible regulatory responses that can reduce monitoring costs without interfering with incentives to monitor by fiddling with segregation.  In particular, improving FCM disclosures would help.  These could include providing customers with statistics on the fellow-customer risk they face, based, for instance, on stress tests.  CCPs perform such evaluations to estimate the risk they take on from FCMs and their customers, and properly protected and presented, such information would allow customers to make discriminating choices among FCMs based on risks and rewards.

There are other problems with the proposal.  In particular, given the concern about systemic risk in clearing, the proposal is particularly troubling.  CCPs mark-to-market on a daily or twice-daily or greater basis.  Thus, once or twice a day it is necessary for FCMs to meet variatio margin calls on both customer and house accounts.  Frequently the FCM needs to meet the call before the customer has provided the funds.  So in meeting the call, the FCM extends credit to the customer.  Moreover, settlement banks typically extend credit to the FCM to meet the variation margin payment.  In the omnibus system, banks view the credit as being collateralized by the omnibus account.  This is well diversified, usually, across customers and across products in the customer accounts.

Moving to segregation basically makes this credit mechanism unworkable, or at least far more costly than under the omnibus system.

This credit mechanism is extremely important.  Indeed, the possibility of its breakdown during times of market stress is one of the things that poses threat to the viability of a clearing system.  The near failure of this credit mechanism on 20 October, 1987 is what threatened the collapse of the major Chicago CCPs.

So what replaces it? It’s not immediately obvious how credit could be extended to meet variation margin payments in a segregated system in the way it is in an omnibus system.  With segregation, a bank extending credit to an FCM doesn’t have recourse to the entire portfolio of customer funds.  Individual accounts pose more risk, and are more difficult for the bank to evaluate.  It is also more operationally difficult to evaluate creditworthiness and to extend credit on hundreds of single accounts rather than a single omnibus account. So it will be much more costly to use credit to meet the metronomic variation margin calls.

But margin calls will still need to be met, and fast.  With credit being far more costly, FCMs will have to hold more liquid assets, and will likely require customers to post higher margins in order to have the cash necessary to meet the calls: in effect, segregation increases–perhaps dramatically–the precautionary demand for liquid assets. Moreover, during periods of market stress, even these additional holdings of liquid assets may not be enough to meet margin calls.  Riddle me this, Gary: What happens then, when they can’t be financed by credit?

This represents a major change in the operation of the market.  Given that it was possible for CCPs to operate in the past, but chose not to, it is reasonable to conclude they did not do so because it is more expensive than the omnibus system.

There’s also a monitoring angle here too.  Banks that extend credit to FCMs monitor them.  They are not going to extend credit to FCMs that are undercapitalized, or which do not manage properly the risk in the customer accounts.  So the omnibus system gives settlement banks an incentive to monitor FCMs.  Go to segregation, and that bank incentive to monitor vanishes.

There is a broader issue here.  Why does the CFTC perceive that it is necessary to impose a particular mechanism on market participants?  What is the market failure here?

The choice of segregation mechanism, and the specification of the default waterfall more generally, involves complicated trade-offs and incentive effects.  Yes, fellow-customer risk is a bummer, but we don’t live in Nirvana: we can’t make it disappear.  To reduce fellow-customer risk you either have to push the risk elsewhere or pay real costs to reduce it.

Exchanges and CCPs internalize many of the costs and benefits of those trade-offs and incentives.  If they choose an costlier mechanism, or one that allocates risk inefficiently, or one that provides bad incentives to control risk, the demand for their services declines.  They see lower volumes and lower prices as a result.

Now this may not work perfectly to achieve a first best outcome in the omniscient social planner sense.  But exchanges and CCPs (and integrated exchanges that operate CCPs) tend to have better information and stronger incentives than any regulator.  Absent some convincing demonstration that there is some major externality or other problem that leads them to choose badly, there is no basis to prescribe just how they should operate their businesses.

But no.  The CFTC just can’t seem to leave well enough alone.  It has again arrogated to itself the authority to impose a system on the marketplace, without making any showing whatsoever that competition and private contract lead to inefficient outcomes that can be improved by regulatory fiat.  This is especially perverse in this particular instance given that the likely outcome of the dictat is to increase risk–and indeed systemic risk.  All in the name of implementing an act ostensibly implemented to reduce risk.

How much irony can we take?