How a fake tweet sent stockmarket into free fall

When share trading is dominated by ultra-fast computers programmed to glean news feeds for data, chaos is only 140 characters away, writes Glenda Kwek.

Within two frantic minutes on Tuesday, $US134 billion was wiped off the New York Stock Exchange. The flash crash stemmed from a tweet by the Associated Press on Tuesday reporting that explosions had rocked the White House and US President Barack Obama was injured.

The tweet was quickly verified as a fake, and the Dow Jones Industrial Average rebounded within 10 minutes, reversing the 1 per cent plunge.

While the White House and its occupants were unscathed by an attack that never happened, investors and traders were shaken by how some trading algorithms appeared to have been triggered by keywords in the tweet, such as "explosions" and "White House", and sold stocks, without any human intervention.

Where was the oversight, and could a similar crash happen here in Australia?

"It's fine having a plane on auto-pilot, but you actually would like to know that there's a pilot watching what's happening," said Don Hamson of Plato Investment Management, who has managed Australian equities for two decades.

Algorithmic trading, also called algo-trading, has become a central feature of global stockmarkets. Computers dominate our day-to-day activities, and as financial markets become larger and more complex, computerised trading has also become commonplace.

One of Australia's largest investment banks said more than 95 per cent of its trades on the ASX were carried out by algorithms, which sift through data and make buy and sell decisions in milliseconds.

After the Australian Securities and Investments Commission's investigation into financial markets last year, the regulator concluded at least 99.6 per cent of all trading messages were from an automated order processing program.

"Most of the trading messages ... originated from the algorithmic programs used by market participants and buy-side clients," the ASIC report said.

These algorithms are meant to operate under the watchful eye of a trader. "It's not like they are running and nobody is watching them," said one of the investment bank's senior traders.

"Traders are monitoring them and adjusting the parameters, adjusting the aggression, adjusting the prices. And it's not like there's just one algorithm, our traders would have access to 20 different algorithms that do different things."

The difference in the Wall Street crash earlier this week was that stocks were traded based on a tweet.

While algorithms have traditionally digested and acted on market data, there are now programs that glean information from news feeds from traditional news sources such as Reuters and Bloomberg. What's not clear is how many trading firms take the next step and use algorithms that also scrape news from Twitter and other social media channels.

Armando Gonzalez, the chief executive of RavenPack, a US specialist data provider that counts Australian banks and hedge funds among its clients, said although his firm had developed algorithms that read and analysed news stories from major news wires and newspapers, they had so far shunned social media as a source as being "too noisy" and unreliable.

"In this case, there were some pretty basic elementary algorithms that some high-frequency trading shops used to trade immediately off a single tweet without verifying its validity," Mr Gonzalez said.

"Arguably, this particular tweet came from a well-known and credible source, but what is very well known, at least in the institutional space, is that Twitter is not the most secure to convene news."

Traders Fairfax Media spoke to believe a similar micro-crash could happen in Australia, given the growing use of social media by companies to share official announcements.

The US Securities and Exchange Commission said earlier this month companies could use social media to make key announcements - as long as they told investors which sites they used.

In contrast, the Australian Securities Exchange said its listed companies were bound by a requirement to make all key announcements through its market platform. It also issued a guidance to companies on how to deal with rumours on social media and other news outlets, including using the platform to respond to the news. Three Australian companies - Whitehaven, Macmahon Holdings and David Jones - have been the target of hoaxes in recent times.

For portfolio manager Vasilios Piperoglou of Leyland Private Asset Management, regulation is key. Although the federal government last year approved new policies to regulate high-frequency trading such as mandatory "kill switches" to avoid flash crashes, he believes more can be done.

"I think regulation has yet to catch up. These programs do move markets," Mr Piperoglou said. "If you were unfortunate enough to follow the herd and sell down, like that fake bomb scare, sometimes there are other people on the other side profiting from that."

The FBI and the Securities and Exchange Commission were investigating the incident, which a group of pro-Syria hackers calling themselves the Syrian Electronic Army has taken responsibility for.

At the same time, Twitter was reporting tests of a two-step verification security method that would make hacking its accounts more difficult.

For David Scutt, a treasury dealer at Arab Bank Australia, the growing role of algorithms in every decision-making step on financial markets was changing the philosophy of why trading exists - which is to test the strength of a company's fundamentals.

"I think it is going too far. Where the marketplace used to be about looking at company fundamentals, monetary policy and fiscal policy, if you look at advertised jobs now for people working in finance in that sector, they are for mathematical degrees and [quantitative analysts]," Mr Scutt said. "There's not even a skerrick of anything about economic fundamentals."