Social media is becoming a critical listening post for investors and traders who are looking for early signals on market direction. The volume of traffic on networks such as Facebook and Twitter is enormous and growing, and these larger data sets lend themselves to more meaningful analysis.
Facebook has an estimated 1.2 billion users. Twitter has grown to about 500 million users. At the same time, software developers have created more powerful tools for analyzing that data. Valerie Bogard, a research analyst at TABB Group, discussed how these new tools are playing a part in the investment and trading process. Here are edited highlights of those discussions.
It’s easy to be skeptical about the latest Internet trends. Can market participants really learn anything of value by listening to what people are saying on social media?
There is a really big potential here and people want to get ahead of the curve and familiarize themselves with it before it really takes off. It can give you a clue to pricing before other people find out about it. Studies have found significant correlation between social media sentiment and pricing. It you are able to see that before others, then you have an edge.
I think the skepticism is waning, to a degree. Sentiment analysis is increasingly able to weed out signs that are not important and focus on things that are. Sentiment analysis definitely has the potential to generate alpha, which is harder and harder to do.
Why are investors taking social media more seriously now than they were a year or two ago?
As more and more people use social media, the more important it becomes for investors to use social media. It is easier to mine for valid insights when you have a larger data set. And the tools have turned out to be effective in some situations. There was the case of the iPhone 4S. The media initially said that there was not a lot of positive public reaction to its launch. But sentiment analysis of what people were saying about the phone on social media showed that they were really excited about it. In the end, Apple’s earnings confirmed that social media sentiment was correct.
And social media was able to identify that the “hash crash” was a hoax before the media was able to do so. [the hash crash occurred on April 23, when hackers hijacked the AP Twitter account and erroneously reported that bombs had exploded at the White House, injuring President Obama. Stock prices fell in response to the erroneous report.]
How do these tools work?
Obviously, you have this huge data stream from Twitter and things like blogs and message boards, Facebook and other kinds of social media. Tools take this entire stream and, looking for keywords, analyze each individual post for sentiment, positive or negative, or a more complicated point in between. Then, the post is tagged for things like language or location. All of this is filtered and processed. Duplicates and spam are weeded out. Taken in the aggregate, we can see where sentiment is and how it changes over time.
If someone tweets “Apple is bad,” or “I hate my iPhone,” obviously the sentiment is negative, but if sarcasm is involved, it can be difficult to determine sentiment. Also, content can be weighted. Tools can determine whether one individual is making lots of posts on the same subject in a narrow window of time, or whether someone has lots of retweets and followers. The results of all this analysis generates charts which can be controlled in dashboards.
There are two very different kinds of sentiment analysis tools. Topsy Analytics and Gnip are applied to various industries, like marketing and media. And there are verticals that are applied to specifically to finance. All the market data suppliers offer some kind of sentiment analysis.
One vendor we talked to has developed tools for this more investor-focused approach to sentiment analysis. They filter both accounts and tweets exclusively for information about the intentions of investors or what they would call “indicative tweets.” They will also weigh accounts that tend to be more predictive. They argue that this approach leads to a more targeted analysis compared to general sentiment.
The tools have been around for about five years, but people are just beginning to figure out the best ways to get value out of the data. There is a certain amount of experimentation, but it is delivering some insights that you don’t get just by following the news. One example is an analysis that looks at what other investors are saying in the interest of isolating securities that are acting abnormally. This gives investors the opportunity to get ahead of the curve especially in short term trading strategies.
What is the potential of social-media in algorithmic trading?
We believe many firms are hesitant to integrate all types of social media into algorithmic trading methods because it’s still hard to understand the full context of short posts.
How are market participants interpreting the results of sentiment analysis and integrating them into the investment process?
Currently traders are primarily using the analysis as an early event detection system or to confirm their trading strategies. Obviously, after the hash crash, investors have been cautious to read too much into a single tweet or post. Once there is corroborating evidence, a factor that most platforms incorporate into their analysis, it becomes safer to trust. However, traders are not using social media analysis by itself. They are adding it to their overall interpretation which would include other pricing and internal data.
How sophisticated are these tools and how much more work needs to be done before they are sufficiently evolved for the mainstream?
Right now TABB thinks it’s in the first adapter phase, but is quickly moving to mainstream consumption. These tools have been developed on a horizontal front for a long time and now it is more about developing more vertical versions. Most firms are realizing that they need some sort of social analytics platform to stay competitive and to take advantage of all the information that is being posted on social media.