Can #Twitter_Trends Predict Election Results? Evidence from 2014 Indian General Election

Aparup Khatua, Apalak Khatua, Kuntal Ghosh, Nabendu Chaki

Abstract: Extant literature finds that twitter trends can capture electoral sentiment. However, empirical evidences are ambiguous in nature. Thus, this study uses the context of 2014 Indian General Election to test the predictive power of Twitter in a large and politically diversified country. We have analyzed roughly 0.4 million tweets during the period March 15, 2014 to May 12, 2014. We observe that tweet volume as well as sentiment analysis can predict election results. We also find that sentiment scores can predict changes in vote share. We note that in a multi-party system the nationality of a party can be an important factor. However, these results should be interpreted with caution. We emphasize the relevance of contextual understanding for efficient data collection and analysis.