By Mary Campbell, Senior Reporter
Anyone who has lost a couple of hours surfing news Web sites while listening to CNN and eyeing the pile of print publications that has been accumulating, unread, for a week knows the problem: too much information.
And that’s just the average news junkie trying to keep abreast of current events; what about the traders and quantitative analysts and asset managers who are not only expected to keep up with current events, but to understand how those events will impact markets in general and their positions in particular?
Well, the news junkies are on their own, but Dow Jones is aiming to help the finance professional with a tool it recently unveiled – Lexicon.
Lexicon examines massive volumes of Dow Jones real-time news, identifying crucial words and phrases that signal trends or patterns. The words are analyzed for sentiment, frequency and other relevant analytics based upon proprietary dictionaries that determine the "tone" of the news. According to the firm, researchers and traders can then use this data to build customized indicators and deploy them as part of trading models.
The idea, says Rob Passarella, vice president and managing director of institutional markets at Dow Jones & Co., came from the company’s work with academics.
“We had been working with academics in the past,” Passarella told FINalternatives, “Many academics like to use what we call our text feed… to actually get our content and do analysis of it. So it pretty much started… with a few guys who wanted to do that, to do some textual content analysis and one of the gentlemen was a guy by the name of Paul Tetlock, who is now at Columbia University.”
What Tetlock wanted to do was to study the role of news in stock pricing. At a seminar at Princeton University’s Bendheim Center for Finance in April, Tetlock presented the results of a study using 29 years’ worth of data on all publicly traded U.S. firms in the Dow Jones news archive “to examine how firms’ information environments change during 2.2 million news events.” Tetlock called it “one of the largest quantitative records of financial news events ever constructed.”
What Tetlock did, says Passarella, was to run “simplistic sort of categorized words – positives, negatives – and what he found was that when people would talk about companies and they’d use more negative words it correlated very nicely with an earnings surprise.”
But it wasn’t just academics mining Dow Jones’ text feed for information:
“What we also found out,” says Passarella, “is that customers were taking our text feed and using it in a similar fashion. They were trying to figure out a way to come up with either sentiment scoring or some kind of quantitative way to look at the news… So we decided, ‘Okay, if our customers want to do that, how can we make it easier for them?’”
The answer was to make raw data available to customers looking to create their own indicators.
To do so, Dow Jones drew on experience gained in 2009, when the company launched its Economic Sentiment Indicator. To generate the ESI, Dow Jones analyzes 15 major print publications for “terms around recession and recovery,” says Passarella. “We have a certain list of things that we look for – word pairs and stuff like that – across all news and it pretty much gives a good look at sentiment around recession and recovery and that tracks very well against other economic indicators. So we noted that and we decided we would make basically that same type of raw analysis available to clients who wanted to do that on their own.”
Clients include asset managers who want to use the news as one factor among others – like market cap, EPS and momentum – when analyzing investments.
“They’ve also developed some measures to give them an idea of sentiment or at least tonality or in some cases velocity of news that’s coming out and use that as part of their screening process or positioning process for instruments that they either want to be in or get out of,” says Passarella.
One odd aspect of the effect of news on markets is that, in some ways, the accuracy of the news is not a factor. Passarella cites the “classic” example of Steve Jobs’ death. Although the story that Jobs had died of a sudden heart attack – posted by a “citizen journalist” on the CNN Web site in 2008 – was not true, it caused trading in Apple’s stock to skyrocket, with the share price falling about 10% before rebounding later that day.
“What I say,” says Passarella, “is we all know that information moves markets, so the question becomes how you want to look at it. There are those who… are fast trigger and there are those who are slow twitch… when it comes to reacting to what happens with news and I think that’s a beautiful thing that happens, because this product is not solely for those who are in the fast-twitch world. There’s a deeper analysis that can go on for those who are in the slow-twitch world and how they want to reflect and look at things over time.”
Lexicon can help the “slow twitchers,” says Passarella, by allowing them to filter information, uncovering historical patterns or sentiments in sectors that “may be a hunting ground” they should look at.
“We are inundated with information all the time and there are no really good strategies out there to quickly identify and do different things,” says Passarella. Clients often need to cull through content quickly, and it’s not just about finding “the negative news today and trading off that immediately,” he says. “Sometimes it’s an in-depth process and we think that’s really what we provide here… another set of meta-data for someone to use.”
Passarella adds that Lexicon goes beyond simply providing a news 'score'.
“A lot of times when I deal with quants they want to know the inner workings of how something was put together and a lot of times when they deal with someone who just provides them a score that can’t be done," he says. "This is a case of where, from the raw data itself and the count, they know exactly how it can be put together because they put it together.”