Not so long ago, pioneers of social media as market data had to prove the correlation between tweets and market movements using academic studies. Now, the relationship is plain for all to see, after a fake tweet from the Associated Press’ Twitter account about an attack on the White House sent markets plunging.
The hoax—call it it cyber-terrorism or just a new twist on old-fashioned market manipulation (remember to short Treasuries before hacking Twitter feeds of news agencies)—has spurred fresh debate over the use of social media platforms in the financial markets, where everyone agrees there is value to be gained, but everyone disagrees on how great that value is, and how to harness these new sources of information to achieve it.
“Social media data needs to be used with great caution… because as things stand today, we can’t trust it,” says Hugh Cumberland, solution manager for payment and settlement services at Colt, which recently published a survey on the use of social media in financial markets.
But those concerned with the accuracy of tweets may be barking up the wrong tree. Of course you should double-check before basing trades on a 140-character update whose top trending topics include (at time of writing) #infowarspickuplines. But anyone making money from trading Twitter probably doesn’t care whether tweets are accurate: what they care about is how the Twitterverse reacts to those tweets, and how they can profit from riding the wave of that reaction.
Meanwhile, vendors like New York-based startup Estimize, are applying social media principles to create more accurate market data. Estimize uses crowd-sourcing to create consensus earnings estimates—which it says are more accurate than traditional estimates—and is looking to apply this model to other data types, having signed a deal for social media aggregator Gnip to carry its data.
However, most social media content is in unstructured free text, and requires analysis and processing to turn it into a usable numerical value or signal, so key to getting value out of unstructured social media data is Natural Language Processing and search. The market data industry is no stranger to high volumes of data, but—other aspects of Big Data aside—the desire to harness social media creates a new set of challenges and demands new technical solutions.
Natural Language Processing serves dual purposes: it allows computers to understand unstructured data—such as news, research, blogs or social media posts—rather than merely numerical values, thus providing context and nuances to figures. But it also allows traders to query today’s vast amounts of data using plain English, rather than having to remember combinations of codes or shortcuts. For example, Thomson Reuters designed its Eikon terminal mindful that tomorrow’s traders would be more familiar with web searches than shortcuts and clunky function keys. Its latest iteration incorporates natural-language search for querying the terminal’s data and analytics.
Thomson Reuters isn’t alone. Austin, Texas-based financial search startup 9W Search—which enables financial advisers, analysts and researchers to search and compare company financial data from public sources and company filings, as well as other data sourced from Edgar Online—has just released the first production version of its search platform, and is now developing a new screener tool to help users filter its datasets more quickly.
Now imagine if these services were all made available in the app store of a web-based data platform and tied together with some unified search mechanism that could link and cross-reference their data. Because that’s where I believe the financial data desktops of the future are headed. Sure, social media in some form will be an important component, but I believe search will be even more important.
Anthony and James delve into how the systematic internalizer regime is shaping up, and then examine the regtech sector.Subscribe to Weekly Wrap emails