Those readers with concerns about the effects of computer system complexity on our society have a must-read article by the Wall Street Journal staff on the legacy of the May 6, 2010 “Flash Crash”. Kudos to the WSJ for answering questions I’ve had about what really happened day, when the market gyrated, crashed, halted, and came back, albeit roughed up in the process.
“The whole system failed,” says John Bogle, founder of fund company Vanguard Group. “In an era of intense technology, bad things can happen so rapidly. Technology can accelerate things to the point that we lose control.” And that day, we did.
The solution implemented to date, on a pilot basis, is a five-minute trading halt on S&P 500 stocks that move ten percent within five minutes. These “collars” act as a circuit-breaker, allowing humans to intervene in otherwise automated trading systems. The question for the Securities and Exchange Commission, which will issue its own Flash Crash report this fall, is whether circuit-breaker trading halts are sufficient defenses to another crash.
My own conclusion is that circuit-breakers alone are insufficient to prevent a similar meltdown. Based on the WSJ article and my own discussions with market participants, there are four areas that require in-depth analysis and decision-making:
- All or None: The weak-sister exchange was the NYSE ARCA, whose quotations lagged other exchanges by two seconds, eternity when computer traders are working in millionths of a second. NASDAQ, followed by other exchanges, stopped routing orders to ARCA. But ARCA handles up to 30% of EFT trading, so cutting off ARCA from the rest of the market severed the widely traded S&P 500 ETF (SPY) from its option, individual stock component, and futures brethren. Bad things were bound to happen after that moment. Which raises the all-or-none question: if one exchange falters, is it better to cut it off and keep trading, or halt the market until the system of all markets can be synchronized again? The Flash Crash results suggest that treating all exchanges as part of a single integrated system that either runs or does not, all or none, deserves serious consideration.
- Capacity: The evidence is clear that inadequate computer and network capacity were major contributors to the Flash Crash rout. This resulted from an avoidable human planning error (or a negligent cost-avoidance decision). After all the heads-up we got in the 2008-2009 financial crisis about “black swan” events being more frequent than most expect, inadequate computer capacity would seem to be a no brainer to anticipate, and is certainly easy to solve. (Note to HP Sales: what are you waiting for?)
- Who’s Watching the Programmers? Two factors struck me as systems design and programming issues that should have probably been discovered and treated differently than they were, left to yield wrong or unexpected results. The first is stub quotes, which market makers use to set below a floor and above a ceiling on prices. They aren’t meant to be executed, yet numerous trades (i.e., Accenture for $.01 a share) at stub-quote were executed on Flash Crash day. The second anomaly involved an Apple trade with insufficient supply to meet the demand at a given price. The programmers, rather than declining the portion of the trade with no supply at the market price, substituted $99,999.99 a share for Apple stock. That fact should stop the heart of anyone who has ever placed a “market order” to buy at the market price. And I think an arbitrary $100,000 a share is a design flaw.
- ETF Disintermediation: Like the mortgage securities that brought our economy to its knees, stock indexes, Electronically Traded Funds (ETFs), futures, and options are all inextricably tied together. Ten or more years ago, if trading in a single stock was erratic, the market did not falter. Today, hundreds of millions of shares underlying the S&P 500 are continuously being bought and sold to keep the S&P 500 ETF, proxy for the U.S. equity markets, in line with its 500 component stocks. Now, layer on S&P 500 futures, arbitrage, and the short side of the market. This is one Gordian Knot of complexity. It’s unlikely we can prevent another Flash Crash until the pieces and the whole of ETF disintermediation and its effects on the market is fully understood.
The author’s resume includes design and implementation of a trust securities system for a money-center bank, financial industry marketing at two computer companies, and extensive experience in computer systems performance measurement and auditing.