Can Collective Sentiment Expressed on Twitter Predict Political Elections?

Abstract

Research examining the predictive power of social media (especially Twitter) displays conflicting results, particularly in the domain of political elections. This paper applies methods used in studies that have shown a direct correlation between volume/sentiment of Twitter chatter and future electoral results in a new dataset about political elections. We show that these methods display a series of shortcomings that make them inadequate for determining whether social media messages can predict the outcome of elections.