Investors are constantly seeking new signals that provide leading indicators of market direction – scouring news feeds, market data, economic data, and even the weather as signals that indicate where the market is headed. It seems obvious, then to also tap into social media to provide an indicator of where the Dow will head, and clearly many firms have already been doing this for some time.
One firm, however, has gone very public with their approach of relying heavily on Twitter posts as their basis for investment decisions. Derwent Capital based in London launched this year and is making its investment decisions primarily on data culled from Twitter – they analyze about 10% of the posts on Twitter to assess market mood and sentiment, and make investment decisions based on what they find. Derwent, through July of 2011, has beat the S&P and outperformed the average hedge fund slightly. Not a bad showing, but we’ll see how their performance trends over time. Given that its not very difficult to do what they are doing – likely other funds can replicate their approach and any outsize returns will quickly be arbitraged to zero.
How to identify market sentiment using Twitter posts?
A research paper published at the University of Manchester and Indiana University by Johan Bollen, Huina Mao, Xiao-Jun Zeng details the results of a project they ran which did enable them to accurately predict market direction based on Twitter posts.
The researchers took a sample of all public tweets on Twitter, and filtered out all posts with URL links as well as filtered out for stop words. The researchers then focused on tweets indicating mood or emotion, specifically those containing phrases like “I feel,” “I am,” “makes me,” “I don’t feel,” and analyzed the results to categorize them into 6 basic emotional categories: Calm, Alert, Sure, Vital, Kind, and Happy. They did their analysis in late 2008 during which the major events were the presidential election and the Thanksgiving holidays:
Based on their assessment of the public’s mood – they then ran analyses against market performance and were able to determine correlations between market movements and these public emotional states, and found that one emotional indicator, Calm, does have a predictive ability for market direction, specifically when indicators of Calm went down (e.g. anxiety increased), 3 to 4 days later the market had a downward trend. From the paper:
In this paper, we investigate whether public mood as
measured from large-scale collection of tweets posted on
twitter.com is correlated or even predictive of DJIA values.
Our results show that changes in the public mood state can
indeed be tracked from the content of large-scale Twitter feeds
by means of rather simple text processing techniques and that
such changes respond to a variety of socio-cultural drivers in
a highly differentiated manner. Among the 7 observed mood
dimensions only some are Granger causative of the DJIA;
changes of the public mood along these mood dimensions
match shifts in the DJIA values that occur 3 to 4 days later.
Surprisingly we do not observe this effect for OpinionFinder’s
assessment of public mood states in terms of positive vs. negative mood but rather for the GPOMS dimension labeled
“Calm”. The calmness of the public (measured by GPOMS)
is thus predictive of the DJIA rather than general levels of
positive sentiment as measured by OpinionFinder.
There are other factors to consider, such as the fact that the time frame in which this research took place (late 2008) was very unique due to the financial crisis and bank bailouts. One would need to run these analysis over a longer time frame to have a better understanding of predictive value.
