Social media and high-speed algorithmic FX trading a dangerous mix
As use of social media becomes more widespread as a tool to disseminate and garner FX market information, concern is mounting about the threat it poses when combined with high-frequency algorithmic trading.
US regulators have announced they will explore how social media is used in derivatives trading after April’s hack crash, where a fake news tweet from the Associated Press’s (AP) account wreaked havoc in the markets.
Dollar-yen plunged 0.7% instantly and in the 180 seconds before order was restored, $136 billion was wiped off the Dow’s value after algorithms allegedly reacted violently to keywords in a tweet sent from a hacked AP account saying the White House had been attacked.
Markets have always been vulnerable to the spreading of false information, whether unintentionally or for more sinister motives. But social media massively increases the reach of such information and can add a gloss of credibility, especially when it is re-tweeted 4,000 times – as was AP’s tweet – by journalists and other trusted sources.
And while it remains unclear whether the algorithms of high-frequency trading (HFT) firms triggered the sell-off, it is clear that algorithms programmed to spot rapid and/or large price moves exacerbated and accelerated the moves that crashed markets.
While most of the focus has been on HFT incidents involving equities markets, the much more liquid FX market might be more exposed and not just because of currencies’ utility as a proxy for any trade over the short-term.
According to the Aite Group, HFT accounts for more than 40% of spot FX trading volume, coming from almost nothing a decade ago. Since 2010, spot market trading has grown by 38% to $2 trillion per day, contributing about 40% to the surge in global FX market activity to $5.3 trillion daily, according to the Bank for International Settlements (BIS).
BIS attributes the surge directly to the increased use of algorithmic trading and execution strategies, and the rising participation of specialized HFT firms in the FX market.
FX derivatives, both on-exchange and off-exchange OTC, might be even more heavily automated.
April’s crash underscores just how influential Twitter now is in markets despite concerns over tweets’ reliability compared with other data, the potential for abuse and how it might ratchet up systemic risks posed by high-speed algorithmic trading.
“An algorithm isn’t going to stop to double-check the source or accuracy of a piece of information – it’s just going to react, but then so do human beings,” says Michael Hewson, senior market analyst at FX provider CMC Markets. “I’ve often had a headline flash up on one of my screens, only for a correction to be issued 30 seconds later.
“Hacking and verification of high-profile Twitter accounts could be addressed with an additional layer of encryption in respect of the log-in.
“In terms of information dissemination in FX markets, social media has been fairly good, but you are going to get the exceptions where people will tweet for their own ends and unfortunately I don’t think you can do anything about that.”
Hewson, an award-winning currency analyst with 12,000 Twitter followers, adds: “HFT provides a lot more depth to the market. Unfortunately, sometimes it will react to a news story that’s incorrect. Even if you didn’t have HFT systems, you’d still get that from ordinary traders reacting immediately to the possibility the news is true.
“Whether it would be as violent is really the key question.”
Hewson notes exaggerated moves produced by HFT can also provide a buying opportunity or a selling opportunity if the algorithms get it wrong, so it can work both ways.
“If it comes down to effective risk-management on the part of the trader and the investor, it can give them a cheap trading opportunity, or on the flip side it will give you the opportunity to get out that they wouldn’t ordinarily have had,” he says.
Regulators are behind the social media curve, responding to developments reactively rather than pro-actively.
With commerce’s zealous embrace of the phenomenon already seeing markets being moved by tweets and Facebook postings, the SEC was forced in April to concede the new reality and authorize companies to use social media to broadcast earnings news.
Bloomberg has incorporated some Twitter accounts into its data service while DCM Capital, an online trading service, has launched a trading platform with a social-media sentiment feed enabling traders to incorporate information from the social-media channel into trading decisions.
The service tracks positive or negative postings on FX, commodities, stocks and indices on social media to create a sentiment score.
With social media going mainstream in capital markets, the Commodity Futures Trading Commission wants to address what it sees as a gaping hole in Dodd-Frank which addresses neither HFT nor social networking.
While FX strategists, such as Société Générale’s Sebastien Galy, argue hack crash makes a strong case for only paying attention to newswires though the conventional channels, providers of trading infrastructure and data say losses mean lessons will have been learned.
“Those algos which traded on this false news most likely made a loss and those people who made those losses will have adjusted their algos, so it’s quite unlikely it will happen again,” says a product manager at one of the leading providers of trading technology to banks and brokerages.
While social media, use of which by market participants has exploded in the past three years, looks as if it’s here to stay, scepticism remains over not only the automated trading risks but whether it can ever translate into actionable market insight.
“It’s a case of user beware,” says Jason Rolf, manager of the Amati Systematic Trend Fund, which trades FX futures. “Putting a lot of reliance on what comes out of social media is asking for trouble because it allows critical information to be announced, effectively, directly into the world with no gatekeepers, no corroboration.
“You’re using a machine to read something that has been typed in by a human so the potential for errors and abuse is enormous.
“Adjusting algorithms may stop them reacting to suspect tweets but not the impact of that information on prices. A price that’s moving much faster than over a recent period of time is going to pull in algorithms to sell it.”