Twitter shakes up market: the impact of social media on algos
On 23 April 2013, the markets suffered a brief, sharp drop as algorithms reacted to “news” from the Associated Press’s Twitter handle that President Obama had been injured in a bombing attack at the White House. In a few minutes, the Dow Jones dropped 145 points, Standard & Poor’s 500 Index lost $136 billion in value, and blue chip names like Exxon Mobil, Apple, Johnson & Johnson and Microsoft lost about 1% of their value. By all appearances, the market looked as if it were headed for a crash at least as steep as the flash crash in 2010, writes Simon Jones, director of product marketing, IPC (right).
Just like the flash crash of 2010, the April 2013 drop was ephemeral. This time, the damage quickly vanished when it was revealed minutes later to be a hacker hoax, and the markets even ended the day higher than they had opened. However, the incident was a dramatic reminder of a very uncomfortable fact: “dumb” algorithms can only handle things as they are programmed and can do incredible damage in a short period of time due to their lack of human common sense.
What some call the Hash Crash, this second flash crash demonstrates that there are still many dangers in the automation of the market that humans cannot predict. Firms mine every potential data source for information to try to get a miniscule advantage and social media is increasingly used as one of the tools in a traders’ information arsenal.
As the use of social media networks becomes more and more ubiquitous, it’s clear that people are increasingly going to use social networks in their business, finance and trading activities. Social media has had a huge impact on our day to day lives and has changed the way information is disseminated. These changes are now flowing through into the trading environment with traders and algorithms alike using social media to gauge market.
This presents a real challenge for trading firms and for the regulators too. By its very nature social media is a two-way conversation and so firms are cautious about their staff using social media in the workplace, whether they are using it to influence trading decisions or to communicate information about the company. In April of this year the Securities and Exchange Commission gave permission for companies to use social media platforms such as Twitter and Facebook to make company announcements.
But despite this the rules around the use of social media in a trading environment are not entirely clear, although it’s obvious enough algorithms were tuned into the Associated Press Twitter feed for it to have a significant and almost instant effect on global markets.
Algorithms react to the news they detect far quicker than humans, which causes other algorithms react to the first algorithms’ movements, and the entire market is caught in a feedback loop of predetermined routines which are unable to react to the market reality. In fact, one could say it’s inevitable that something like this will happen again. So what can firms do to ride out next time?
As information moves through the markets at the speed of light 24/7, 365 days a year, more efficient and direct communication, particularly instant, trustworthy communication by voice, across the trading floor and between financial firms and their clients is imperative. Trading is a complex activity with a life cycle that moves from front office to back office. While algorithms blindly and near-instantaneously brought the market down, human traders used their voices and talked to each other. Through simple communication, they quickly examined the situation and discerned the likelihood of the news being a hoax due to the source of the news.
The markets recovered so quickly because the parties involved had the infrastructure in place to facilitate communication. An extensive array of tools, ranging from voice communication to integration with the PC applications traders use, such as CRM, OMS, and market data applications, enabled quick analysis and quicker communication. Systems that provide the middle- and back-office staff with access to the private lines used by traders, and line sharing, enabled traders to instantly access the appropriate team members. Secure lines to clients coupled with the role voice communication plays in creating, building, and enriching client relationships delivered critical information and enabled rapid decision making.
The 2013 flash crash was caused by rumours that humans would have approached with scepticism, but algorithms took at face value. Whilst you could argue that leaking rumours in an attempt to skew market trends is nothing new, the advent of social media means that the impact of one individual is dramatically amplified when compared to just a few years ago. The potential for criminals to exploit this in the future is huge, and trusting algorithms to discern what is real and what is fake is a risky strategy.
