Streetwise Professor

Today is the 25th anniversary of Black Monday, the 1987 Crash.  This event had a profound effect on my professional trajectory.  I worked at an FCM at the CME-GNP Commodities-during the Crash.  The firm was owned and run by Brian Monieson, a former chairman of the Merc, and a board member at the time.  I walked into his office on the morning of the 20th, and he told me about how the CME clearinghouse had almost failed.  I had only a dim appreciation for clearing prior to that, but that sparked my interest in the subject.

Little known fact: I Granger Caused the Crash.  Throughout the late summer and fall of  ’87 I had been having issues with my direct boss, and soon after I defended my thesis in August I had received a good offer from one of my advisors, Dennis Carlton, to work at Lexecon.  I tried to work things out at GNP, but by mid-October it appeared hopeless.  I told Brian, and we had several phone calls on the 17th and 18th where he tried to convince me to stay.  I thought it over, and on the evening of the 18th decided that it wasn’t going to work.  So I walked into my boss’s office at 7AM on the 19th and slapped down my resignation letter on his desk.  Apparently the news leaked, and the rest is history 😛

The anniversary has spurred retrospectives from many quarters.  Many of these focus on the role of computers in the Crash, and how this was a harbinger of the supposed horrors the machines are inflicting on the markets today.  One good-well, representative-example of this is “Rage Against the Machine” in the FT:

It has been 25 years since Black Monday, when stock markets crashed around the globe and Wall Street woke up to the risks of computerised trading.

Yeah.  The computers were blamed then.  Suspect number one was “program trading”, a good deal of which was garden variety index arbitrage.  This is not a destabilizing strategy, and numerous postmortems absolved program trading as a cause or aggravator of the Crash.

Indeed, if technology was a problem, it wasn’t that computers were trading too fast: it was that printers were printing to slow.  Seriously.

At GNP we did a little index arb trading.  In the morning, it appeared that the arb was extremely profitable.  The price of the cash index (we were doing MMI) was way, way above where the futures were trading.  It looked like a great opportunity.  My colleague Matt who was in charge of this called down to the CBT floor before doing a trade, to see where the floor was trading.  Our floor clerk said (I don’t remember the numbers exactly): “Somebody is offering X over here.  Somebody else is offering X+20 over there.  Somebody else is offering X-20 over there.  I have no fucking idea what the price is.”  Matt didn’t make the trade.

Which was good, because the apparent opportunity was a mirage.  That’s where the slow printers come into play. Specialist posts on the NYSE had old fashioned daisy wheel printers.  Very, very slow.  They were overwhelmed by the order flow.  At 10AM the printers were spitting out orders that had been entered at 8 or earlier.  So the orders being executed on the NYSE were actually very stale, and the prices didn’t reflect current market conditions.  The futures didn’t have that problem.  So at 10 or 11 NYSE prices were reflecting orders submitted hours before, and the futures were reflecting current conditions.  Hence the futures were way below the cash, making it appear that there was a huge arbitrage opportunity.

The problem was that this induced more people to submit sell orders to the NYSE before they caught onto the fact that NYSE prices were not reflective of current conditions.  This exacerbated the backlog on the printers.

So it wasn’t fast technology that caused problems on 19 October ’87.  It was slow technology.

Program trading wasn’t an issue, really, but portfolio insurance was.  This is sometimes called “computerized trading” but it’s not, really.  Yeah, computers were used to calculate deltas, and these deltas were used to determine trading strategies.  You could have used an HP-12C (complete with Reverse Polish Notation!-and no, that’s not an ethnic joke) and the normal cdf table from Handbook of Tables For Mathematics (pp. 922-929 of the Fourth Edition) to do the same thing.

The problem with portfolio insurance was that, unlike index arbitrage and most market making strategies employed today in HFT, it was a positive feedback strategy, rather than a negative feedback strategy.  As prices fell, portfolio insurers had to sell more, which drove down prices further, which led them to sell more . . . until they pulled the plug on the strategies.  That is, unlike market making strategies which are buy low-sell high trades, portfolio insurance was buy low, sell lower.

The FT article uses the ’87 Crash as a springboard to make the usual litany of complaints about current market structure and computerized trading.

Some of the criticisms are defensible.  The current fragmentation of equity structure in the US is indeed the product of SEC RegNMS, which I characterized at the time of its introduction as implementing an “information and linkages” approach to market structure, in lieu of mandating a CLOB.  This socialization of order flow has some problems, particularly during times of market stress.  Links are vulnerable during times of stress.  Moreover, the proliferation of order types that has complicated trading dramatically is directly the result of this approach: many of the new order types are related to order routing, which is a much more complex task with many execution venues as opposed to one (as in futures markets where order flows are not socialized).

Other criticisms are less defensible.  The authors of the FT piece are apparently scared of the dark, as they warn of the dangers of dark pools.  These things are not new.   They just used to be called “third markets” or “block markets”.

The FT piece also warns that in the computerized world, liquidity can evaporate quickly.  They acknowledge that liquidity evaporated quickly in the old days-as it did on Black Monday-but suggest that can happen more quickly today.  Not really.  At all.  Locals could shut up or keep their hands in their pockets very quickly back on the floor if it looked like order flow was getting toxic.  And they were pretty quick at picking up on that.  Moreover, during the Crash clearing firms pulled their locals from the pits.  You would have thought that the pits would have been in a state of pandemonium on the 19th, but after the initial frenzy, they were actually spookily empty, as most locals and many brokers had been yanked from the floor by their clearing firms.

In brief, the functions of markets haven’t changed.  The technology for performing these functions has.  But the fundamental economics of performing the functions hasn’t.  Face-to-face and machine-to-machine markets have the exact same basic vulnerabilities.  Those who focus on the technology miss the economics.

The FT article, and others like this one in Bloomberg,  blame computerization and HFT, and the consequent alleged “loss of confidence” among investors, for the decline in equity volumes recently.  Color me skeptical.  Futures volumes have dropped too, despite a very different market structure, and a different composition in market users.

Historically, trading volume has ebbed and flowed.  It ebbed and flowed in the pre-computer era. It shouldn’t be any surprise that it ebbs and flows now.

If you are looking for the big lesson from the  1987 Crash that we should heed today, ignore all the BS about technology.  Look at clearing.  Look at the Brady Report and other post-Crash studies of what happened in clearing.  That’s what almost brought the markets down.  The operation of variation margining-which is usually a source of stability-threatened to bring the markets to a halt, with even more devastating consequences.

Then think of what will happen during the next big market move, given that the scope of clearing has been expanded substantially due to Frankendodd, and parallel efforts around the world.  Don’t worry about the technology of execution: worry about what happens after the trade, during clearing and settlement.  That’s where the truly systemic vulnerability lies.

There’s an old adage in the military: amateurs talk tactics, professionals talk logistics.  In the markets, the parallel is that most commentary focuses on execution, when much of the real important action happens out of sight, in clearing and settlement-the logistics of the financial markets.  That happened in post-Crash commentary, and it is being repeated today.  That’s a big mistake.  When looking for object lessons from Black Monday, don’t look at the computers and the execution of trades.  Look at what almost happened with disastrous consequences in clearing.  And worry deeply about how the system is vulnerable to a repeat in the future.  Arguably more vulnerable, given the dramatic growth in the size of the markets and the mandated reliance on clearing, and margining of non-cleared derivatives.

That’s the lesson I first began to grasp in Brian’s office early on 20 October.  And that experience has decisively shaped my criticism of efforts, like Frankendodd, intended to fix the markets.