Streetwise Professor

The commodity markets have been volatile of late.  The tumult started in silver, which sold off dramatically after reaching post-Hunt highs.  Last week the sell-off broadened to a good chunk of the commodities markets, with oil especially hard hit.  In response to the roiling markets, the CME raised margins sharply on silver, and today on oil and the rest of the energy complex.

This raises the question of the effect of margin changes on prices.  The CME’s head of clearing, Kim Taylor, vigorously defended the CME’s margin increases in silver (and would presumably make the same defense of today’s moves).  Kim’s justification characterizes the clearinghouse’s move as a response to changed market conditions:

Changes to margin requirements are a routine part of market surveillance at CME Clearing, the risk management and compliance unit at CME Group Inc. (CME), which owns Nymex. So far this year, CME has issued over 57 margin change notices.

“I don’t agree that the margin change was a trigger for changes in the market,” said Kim Taylor, president of CME Clearing. Taylor leads a team of sever hundred risk management and compliance professionals who monitor CME’s markets around the clock.

“The market is well tuned so that if there’s a market move that approaches or exceeds our volatility limits” participants know to expect an increase in margin requirements, she said.

The decision to alter trading margins on a contract typically involves risk management professionals involved in day-to-day monitoring of the specific market as well as senior management of the clearing house. The team looks at a variety of quantitative factors like rising volatility and qualitative factors like seasonality and relevant news events in making margin decisions.

“We try to make changes in a way that we can telegraph to the market, so that participants have notice. We try to be routine and predictable and provide no surprises,” Taylor said.

The trading margins are designed to cover around 95% of expected losses, and act as a pre-payment on the coming day’s market move.

“When market conditions become more volatile we would increase margins in anticipation of that, and when volatility decreases we don’t want to create unnecessary capital costs,” Taylor said.

In past instances, CME Clearing has often raised trading margins ahead of certain events in an attempt to damp their impact on trading. For example, the exchange-operator increased trading margins on several products, including crude oil, ahead of Hurricane Katrina to ensure traders were prepared for higher volatility .

“We try to be proactive with either something we can measure or something we can judge to likely effect our markets,”

This is sensible, in a way that I’ll explain in a moment, and the way that CCPs usually act.  Taylor tries to emphasize that the changes are predictable, and in some respects they are, but this is not sufficient to relieve CME of the charge that margin changes can cause market movements.

The changes that CME made, and which Kim explains, are sensible in a microprudential sense.  They help ensure that the CME has a sufficient buffer to absorb defaults by traders.  CCPs try to work on the “loser pays” model, and with more volatility, there are bigger losers–so margins have to be higher to ensure they can pay.

But the implicit assumption here is there isn’t feedback between margins and prices.*  Indeed, that’s the gravamen of Taylor’s argument (“I don’t agree that the margin change was a trigger for changes in the market”).  That is the essence of a microprudential approach.  That assumption, however, is quite tenuous, and almost certainly untrue.  This is particularly the case for big margin changes during unsettled market conditions.

That is, when setting margins, CCPs (and participants in bilateral markets too) typically act as if they are price (and volatility) takers, when in fact they are big enough and their decisions are material enough to be price and volatility makers.  Acting microprudentially, they typically fail to take into account the feedback between their decisions and market prices, or at least do not do so completely.

The feedback mechanism works because frequently market participants will respond to margin changes by liquidating positions.  Those who have already lost money as a result of big price moves are the most likely to liquidate, which tends to exacerbate the original moves.  That’s one reason why you can see bigger than expected moves (as occurred in oil last week) to fundamental shocks.

This is why CCP policies that are prudent in some sense can be macroprudentially dangerous.  Silver, and perhaps oil and other commodities, may be providing an object lesson of what is in store when clearing is extended to vast new markets.  When OTC clearing mandates kick in, the scope for these destabilizing feedbacks will expand dramatically.

Kim suggests that predictability mitigates these impacts, but CCPs are not that predictable, and margin methods are not necessarily that transparent.  So market participants cannot foresee the changes perfectly.  And even if they could, all this means is that they will likely react in anticipation of margin changes. The timing of the response may depend on predictability, but not the existence of the response.

This can create some really strange dynamics and non-linearities.  There’s a big selloff.  Longs anticipate that in response, CCPs will raise margins.  Those that have lost a lot of money know they can’t afford the higher margins, so they preemptively liquidate, exacerbating the selloff.  The bigger price move induces the CCPs to raise margins more than they would have without the preemptive liquidations.

This is exactly the kind of feedback mechanism that can induce chaos (in the technical and conversational senses of the word) in a tightly coupled system like a financial market. This mechanism is particularly prone to chaos if the CCP setting margin doesn’t take into account, or cannot predict accurately, the feedback between its margin changes and the trading of market participants.  And how could it know this effect with any accuracy?

This is a classic example of the tension between microprudential and macroprudential policies that I discussed in some earlier blog posts, and which is a major theme in my forthcoming white paper/discussion paper on clearing.  (This should be available within a week to ten days, if all goes well: final edits and approvals are in the works.)  The information and incentive structures tend to induce microprudentially sensible actions based on the price/volatility taking assumption.  But those actions can be destabilizing. CCPs acting in their own self-interest, and those of their clearing members, may not choose the systemically optimal actions.

It’s worth noting that the CFTC’s Proposed Risk Management Standards for Designated Clearing Organizations effectively codifies this sort of behavior.  Its proposed rule (a) requires that CCPs set margins “actual coverage of the initial margin requirements produced by such models, along with projected measures of the models’ performance, shall meet an established confidence level of at least 99%”, (b) determine the adequacy of margins on a daily basis, (c) backtest the adequacy of margins on a daily basis for products experiencing “significant market volatility,” and (d) backtest the adequacy of margins for all products at least monthly.  Similarly, both the Federal Reserve’s and the SEC’s proposed regulations require the use of risk-based models and at least monthly review of margin levels.

All of these requirements are microprudentially sensible, but will result in margin increases during periods of heightened market volatility.  As just noted, this can create destabilizing feedback effects, particularly during periods of extreme market volatility.

In contrast, the Committee on the Global Financial System recommended the use of a “through-the-cycle approach employing data from a long time series of market movements” when setting initial margins, precisely to reduce the procyclicality in margin that results from adjusting it frequently based on changes in market risks.

Which means don’t look at last week’s commodity mini-crash, or the earlier bigger crash in silver, as historical events.  Look at them as harbingers, as warnings.  Just the normal operation of a tight, mechanical variation margin process can be destabilizing.  Changes in initial and maintenance margins adopted in response to big price moves can destabilize things more.  The financial system can shake off such an occurrence in a market as small as silver, or even one the size of oil.  When it’s a bigger market–and especially when it’s a much bigger market, like some of the huge OTC markets–these changes will shake the markets, perhaps by more than they can withstand.

*Failures to account for feedback between one’s behavior and market prices is a common element in many market crises.  For instance, LTCM’s risk models–like most models in finance–presumed that the hedge fund was a price taker.  But once it became huge, it was no longer a price taker; its trades moved prices and affected liquidity.  Similarly, portfolio insurance strategies took price processes as exogenous, when in fact the strategies themselves affected prices.