Algorithmic trading hits yet another bump
The Twitter feed of the Associated Press was hacked last week and a false post about Barack Obama being injured pulled down the Dow,momentarily shaving $136 billion off the market. The power of social media,it was reported,had extended beyond politics to influence markets. This appeared to be a reasonable conclusion because the incident occurred after the US Securities and Exchange Commission (SEC) had allowed companies to share market-relevant announcements over Twitter and Facebook on April 2. But consider the time-frame of the incident: four minutes from slide to recovery. No human can evaluate a breaking story that fast and initiate or cancel trades accordingly. What the incident actually revealed was the rise of the machines on the trading floor.
Algorithm-based high-frequency trading is not the same as playing old-world stock exchanges. Exclusively electronic and focused on speed,it gains leverage by cutting out the human factor from decision-making. It is driven by algorithms which use the vast volumes of data that modern trading operations generate and evaluate it in ways and at speeds that humans cannot. Computers may relinquish and regain positions several times per second,creating hot potato volumes in trading. The mechanics are poorly understood since algo traders are generally tight-lipped about technicals,but it produces glitches that unambiguously suggest that the markets are no longer for mere mortals,only for specialists who can access heavy-duty computing and bulk data.
The Twitter post was the second big glitch. The first warning was the flash crash of May 6,2010,when aggressive selling by algo traders,in response to a big sale by a mutual fund,had tipped over a market worried about the Greek debt crisis. The day before,there had been protests in the Acropolis in Athens and gold had hit a record high. Trading machines had reacted to uncertainty conservatively,dropping their positions and exiting the market. They are programmed to have better sense than some human traders,but the exit of machines in a flock had exacerbated the negative human sentiment prevailing in markets that day. But again,the crisis had developed and passed in minutes,while regulators spent almost five months pondering the forensics.
Regulators have been concerned about microsecond trading since last August,when a glitch caused $440 million losses to Knight Capital. In February,the US SEC had prescribed regular checkups for exchanges and trading platforms to see that they maintained technical standards and redundancy mechanisms. The agency will also trawl market data to anticipate flash crashes,or at least produce better forensic reports.
But this is merely symptomatic treatment that avoids fundamental questions. This week,a parallel forum is raising them the currency markets. EBS (Electronic Broking Services),one of the worlds biggest currency trading platforms,is telling its clients that speed doesnt matter. The preferred platform for trades involving the dollar,the Euro,the yen and the Swiss franc,it had moved early to facilitate high-frequency trading. But now,it has observed that first-come-first-served trading disproportionately advantages firms that invest in faster computers and thicker data pipes,encouraging a digital arms race. In fact,firms that are physically located closer to markets also have a slight edge in the transmission time of orders. Since competition in trading should be generated by strategy rather than speed,EBS wants to process orders not serially but in batches that are shuffled like a deck of cards. A lottery every millisecond is not the most elegant solution,but at least it would end the monomaniacal pursuit of speed.
Flash crashes are made possible by high-frequency trading,but social media do have an impact. However,the use of Twitter in taking market decisions is not new. Markets have always been mindful of the news,which is why pink papers have political correspondents. Now,when even journalists are reading Twitter feeds instead of the wires,it is not exactly radical for market players to follow suit. It is the fastest news channel ever. It is self-correcting like regular news,offering all versions of a story,but it is prey to the common internet flaw of bubbling you live in a sensory bubble where you are more likely to see posts that confirm your beliefs rather than contrary opinion.
The coverage of the manhunt for the Boston marathon bombers,which was crowdsourced using social media,suggests that reporters understand the phenomenon. Reddit was quick to blame a missing Indian who,inconveniently enough,was later found to be dead; formal media was careful to air only attributable opinions. But perhaps traders arent as hard-nosed as reporters. Last August,a fake Twitter post attributed to Russian Interior Minister Vladimir Kolokoltsev had alleged that Syrian president Bashar al-Assad had been killed or injured,and the prices of crude were immediately affected. Perhaps traders need to teach their algorithms to be wary about Twitter truths.