“All this is not a product of nature.”
No I’m not referring to the display, as it were, by Miley Cyrus at the Video Music Awards. If the shortest distance between two points is a straight line, what do you call a stock market with 13 exchanges, forty broker-operated alternative systems, 4,300 FINRA-regulated brokers and dealers, 2,000 order types, a complex routing scheme for moving millions of quotes per second and all the associated messaging traffic at light speed (oh yeah, and hopefully some investing too)?
Well, not a straight line. The Wall Street Journal’s Holman Jenkins wrote a compelling opinion last weekend on the Nasdaq data outage Aug 22, which he titled “How to Think About the Nasdaq Freeze,” and from whence I borrowed today’s opening salvo. You should read it.
As you’ve heard – at length on CNBC – the Nasdaq halted trading for three hours last Thursday due to a connectivity issue that led to failure propagating the marketwide data stream providing consolidated quotes in Nasdaq securities.
The WSJ’s Jenkins argues that whatever the root cause of this latest in a long line of troubling market mishaps, “complexity breeds snafus.” The market where your investors compete to demonstrate belief in the story you deliver is a tangled web.
It wasn’t a web to begin. The SEC wanted no monopolies in stock-trading, so it mandated that trades go wherever the best price is. The straight line turned into a winding road and then a ball of yarn.
That happened only after electronic markets like Island and Archipelago (now the automated matching engines for the Nasdaq and the NYSE) had already gobbled market-share by audaciously matching buyers and sellers beyond the purview of regulators. So regulators imposed the Order Handling Rules, conjoining square pegs and round holes and ultimately melding them as exchanges bought electronic markets and became themselves electronic.
Which brings us back to The Freeze – the trading halt last week. Trading occurs simultaneously in many asset classes now. On Aug 19, options and futures on Treasuries, stocks, indexes, ETFs and currencies for September expiration began trading. That day, a wave of new Limit-Up/Limit-Down (LULD) circuit-breakers was also rolling out in equity markets. Suddenly, volatility halts exploded in exchange-traded products (ETPs) at Arca, the NYSE’s electronic market and a top derivatives exchange. We saw 60 or so in the first half-hour. Most were commodity, currency, bond and interest-rate instruments.
On Tuesday, Aug 20, Goldman Sachs’s automated algorithmic trading in options markets short-circuited, mispricing swaths of high-speed options trades. We continued to see waves of volatility pauses in Arca-traded derivatives. On Wednesday, Aug 21, volatility futures and options expired and more halts in derivatives ensued.
Then on Aug 22, around 11am ET the Nasdaq invoked Self Help against Arca, which it then revoked about ten minutes later. “Self Help” is regulatory language that permits one exchange to skip another when routing orders because that exchange’s systems aren’t responding normally. About one hour later at 12:14pm ET, the Nasdaq abruptly halted trading.
It could be coincidental, of course. But in markets rife with complex trades in multiple asset classes, are these coincidences or outcomes? My firm writes algorithms. What vexes us is bad data. So take a vast array of simultaneous multi-asset-class trades.
And pause one.
What might that do to the data that will later be consumed by algorithms, or data used to calculate index products or risk-exposure or to link markets for regulatory routing? The bane of all models, not just ours, is bad data.
We’re not trying to prove our hypothesis – that this problem traces back to SEC-mandated LULD volatility pauses. But it’s the straightest line we can find.