Predicting Stock Prices Using Topsy Social Sentiment

A number of studies have been done over the past year examining correlations between sentiment calculated from Twitter posts to stock price movement. Most of these analyses have been performed using sentiment calculations from generalized tweet content applied to broad based index fund movement rather than attempting to isolate results to a specific equity and related tweet content. We wanted to see if statistically significant correlations existed between a specific stock’s closing price movement and social sentiment calculated from posts containing references to that stock’s ticker symbol, company name or related terms.

Our hypothesis is that a stock’s price movement can be related to what people are saying within social media with statistical significance, so long as you have access to the data and tools to iteratively test and learn what terms and measurements relate to a stock’s price movement.

We tested this hypothesis using the NFLX stock symbol, isolating those tweets containing the terms “Netflix”, “NFLX” and “Qwikster” for an extended period of time, applying Topsy Social Sentiment™ to the content within these tweets and then comparing sentiment score movement to the company’s stock price movement. We ran a linear regression model and what we found was that the correlation between sentiment and NFLX closing stock price was statistically significant:

You can read more details about our analysis in our whitepaper by registering for it here.

From this analysis one can conclude that aggregating how people express themselves about a company, stock or related set of terms conversations, at scale, from public social media posts provides valid signal to anticipate an equity’s movement. Our technology enables access to multi-year, census-based Twitter data from which we extract a variety of measurements. These measurements coupled with our search technology allow us to perform similar analyses for any number of equities over multiple years.

Extracting valid social signal from this data to anticipate an equity’s price movement is certainly achievable. What’s needed to ensure proper risk mitigation is iterative testing using a multi-year data set where public social content can be precisely searched using terms relating to an equity, with different measurements calculated from these search results so they can be correlated to that equity’s price movement over time. We’re continuing to perform these type of analyses and will publish selected results as they become available. If your financial organization has interest accessing social data measurements over extended periods of time to perform similar types of analyses within your organization please contact us at bizdev@topsy.com.

One Response to “Predicting Stock Prices Using Topsy Social Sentiment”

  1. Global Commodity Trading House | Best Futures Broker Says:
    February 29th, 2012 at 12:47 am

    [...] Predicting Stock Prices Using Topsy Social Sentiment [...]