Streetwise Professor

May 13, 2010

You Got a Lot of ‘Splainin’ to Do, SEC

Filed under: Derivatives,Economics,Exchanges,Financial crisis,Politics — The Professor @ 8:07 pm

CME Group Chairman Terry Duffy’s testimony before a subcommittee of the House Financial Services Committee is quite informative and fascinating.  Of course he is defending his firm and talking his book, but he makes a very persuasive case on many issues.  What he says does not demonstrate that the May 6 market boomerang did not begin at the CME (sorry about the double negative), but he does make it pretty plain that precautions built into the CME’s Globex system worked to arrest the decline, and that the rebound probably began at CME as a result.

The Duffy testimony provides concrete evidence that stop orders were accelerating the price decline.  An EMini trade at 1062 triggered a stop for a relatively modest 150 contracts.  Execution of this order led the market down to 1058.25: that big move in response to a small order indicates how illiquid the market was (i.e., how low the depth was).  (I will dig up an article by Hasbrouck that estimates depth coefficients using Bayesian methods to see how this compares to the usual market depth.)  This triggered another 150 car stop order which led the market down to 1056.

At this point Globex’s Stop Price Logic kicked in:

Stop Logic

The Globex trading platform is programmed to prevent the continuing execution of cascading stop orders when it detects certain conditions. The Match Engine monitors if the triggering of a stop order or series of stop orders will result in matched prices that exceed the contract’s No Bust Range from the price level that matching limit orders finished matching and caused the triggering of a stop order(s) . The contract is then placed in a reserved state, during which orders may be entered, modified or cancelled but not matched.

The market went into a stop state for 5 seconds.  The decline stopped, and the rebound began as value buyers–almost certainly including some dreaded HFTs–began to buy.

Duffy’s account confirms the main elements of the scenario I sketched out a couple of days back.  The market was relatively illiquid; something triggered a relatively large price move; that triggered stops and a further decline of liquidity, which accelerated the market decline; then value buyers moved in.   Duffy’s testimony also illustrates the essential element in keeping things from going completely non-linear: stopping stop orders.  I repeat: stopping stop orders.

Things were uglier on the cash equity markets.  The price declines in individual stocks, including heavily traded broad market gauges such as the SPY were more severe than on the index futures.  As is well known, some stocks traded at a penny before bungeeing back to reasonable levels.

This reflects, in part, the evolution in the market architecture in the years following the SEC’s approval of Regulation NMS.  As I wrote in this piece in Regulation Magazine, the SEC had a choice between two alternate ways of creating a national market system: a mandated central limit order book (“CLOB”) or what I referred to as an information-and-linkages approach.  The latter means that the mandated dissemination of quote information to permit direction of orders to the best-priced markets, and the requirement that different execution venues redirect orders to other venues displaying better prices.  In Reg NMS, the SEC choose information-and-linkages.

That approach is defensible, and can work well in normal market settings.  The Boomerang illustrates, however, that this architecture has an Achilles heel during periods of extreme stress and low liquidity.  The linkages don’t work properly, the markets fragment, and prices in some venues can diverge wildly from those in others.  During the crash, several venues declared “self-help” which meant that they would not direct quotes to other markets that were not responding within a second.  In other words, every man (market) for himself, and to hell with the women and children.  The linkages broke down.

As a result of NMS, the NYSE has gone from doing about 80+ percent of listed volume to less than 30.  That’s not a big deal in normal times, as information-and-linkages works OK then.  In the stressed time, though, it was a big deal.  Especially since one of the NYSE’s circuit breakers, its Liquidity Replacement Points stopped continuous trading on that market, but didn’t affect others.  So orders ricocheted off the NYSE, bypassing whatever liquidity was there, and overwhelmed liquidity on other markets.

I should note that even centralized markets can fragment.  Back when the pits were big (e.g., T-bonds, T-notes), when trading was active there would be different prices at different points in the pit.  This happened in the ’87 Crash too.  The firm I worked for was doing some simple index arb trades during that period.  Looking at our wondrous Telerate screens, it looked like there were huge arb opportunities.  (There weren’t, because the prices coming from the NYSE were absurdly stale.)  We called down to the floor to see where we could execute the futures leg.  The guy on our floor desk said: “Somebody’s bidding P1 over here, somebody else is bidding P2 over there, and there’s a guy offering P3.  So, I don’t know what the f*cking price is.”  [I don’t remember the exact numbers, except that they were index points apart.  I do remember that last sentence.  Go figure.]  But that can’t happen in a CLOB type electronic system.  (And Globex is effectively a CLOB.)

The Boomerang almost certainly means that the SEC will have to revisit the information-and-linkages approach.  It will have to coordinate the adoption of rules and technical interfaces that prevent the fragmentation, or it will have to consider jettisoning the concept altogether.  At a minimum, the rethink on the information-and-linkages approach will require coordinated circuit breakers.

And the CME experience suggests that the most essential thing to do is to implement a coordinated stop logic analogous to the CME’s.   It is imperative to derail that positive feedback mechanism.  The whole self-help thing (i.e., the every-man-for-himself-delinkage of the information-and-linkage system) also deserves close scrutiny.

The hyperventilating about HFT continues.  Duffy defends it, and defends it well.  Again, HFT is not a homogeneous thing, so it is essential that any evaluation of it take into account the diversity of these strategies.

Also, it bears repeating that the “where did it start?” question is secondary to “why did it spread the way it did?”  and “why were some markets impacted more severely than others?” questions.  And as for those questions, I think the SEC will have a lot of ‘splainin’ to do, Lucy.

Print Friendly, PDF & Email

1 Comment »

  1. I’m going to take a stab at this one.

    Order-driven markets are just that, driven by orders. The orders may be complex, contingent, etc, but, most importantly, they have been revealed to the marketplace or not. A “lack of depth” may mean that no one is willing to buy or sell at a particular set of prices or that market participants have not revealed their willingness to buy or sell at those prices. If no one is willing to buy or sell at a given set of prices and a contrary order comes in, the market will move and it should.

    However, if people are willing to trade at those prices and have just not revealed their orders, a good market structure will have mechanisms to allow the orders that have not been revealed to the marketplace an opportunity to be revealed. My preference is based upon my experience. An auction can be designed to provide the time and the incentive for people to reveal their orders. (Since the auction will lead to a single cleared price, I believe that I will be more willing to reveal my order to the marketplace if I feel that I am at an informational disadvantage than in a continuous market. I may be wrong about that.) An auction also provides the opportunity to cancel any aberrant order that may appear to have caused the market dislocation. (You can cancel any order over a certain size and force whoever sent it down to re-enter the order. This can prevent mistaken orders from disrupting the market though it could also be another source of abuse.)

    Perhaps an auction is not the answer, but I feel that an auction looks to achieve what continuous markets do not necessarily do well in times of turbulence, i.e., provide the time and the incentive for market participants to reveal their willingness to buy or sell as firm orders. Clearly, an auction mechanism would need to be coordinated between the exchanges. Do you agree?

    Comment by Michael Kelly — May 13, 2010 @ 10:01 pm

RSS feed for comments on this post. TrackBack URI

Leave a comment

Powered by WordPress