Predicting elections with Twitter: What 140 characters reveal about political sentiment

From AcaWiki

Citation: A. Tumasjan, T. O Sprenger, P. G Sandner, I. M Welpe (2010) Predicting elections with Twitter: What 140 characters reveal about political sentiment. Proceedings of the Fourth International AAAI Conference on Weblogs and Social Media (RSS)

Tagged: Computer Science (RSS) elections (RSS), politics (RSS), political science (RSS), Twitter (RSS), sentiment analysis (RSS)

---

Summary:

Research questions

• Does Twitter provide a platform for political deliberation online?  

• How accurately can Twitter inform us about the electorate's political sentiment? 

• Can Twitter serve as a predictor of the election result?

Corpus

104,003 tweets posted August 13th and September 19th, 2009, prior to the German national election which took place on September 27th, 2009. They collected all tweets that contained either:

• names of one of the 6 major parties represented in the German parliament (CDU/CSU, SPD, FDP, B90/Die Grünen, and Die Linke) [~70,000 tweets]
• names of selected prominent politicians of these parties. [~35,000 tweets]

Results

Political Deliberation

They find serious discussions in Twitter, and give an example for each party, such as "CDU wants strict rules for internet" and "Whoever wants civil rights must choose FDP". However, 50% of users only post once in their corpus; the remaining 50% of users make 90% of the posts.

Political Sentiment

They use radar maps to plot multi-dimensional profiles of the politicians discussed, using the relative frequency of 12 LIDC aspects. Then they computed a distance measure, finding relatively small differences between politicians, and that, on the whole, there is more difference between politicians than between political parties.

LIDC aspects used

Orientation

• future orientation
• past orientation

Polarity

• positive emotions
• negative emotions

Emotions

• sadness
• anxiety
• anger

Certainty

• tentativeness
• certainty

Hot topics

• work
• achievement
• money.

(categorization added)

Predicting Election Results

The relative volume predicts the election result, with mean absolute error (MAE) 1.65%. "This is in line with the findings reported by Véronis (2007) who has shown that, in the case of the 2007 French presidential election, the simple count of candidate mentions in the press was a better predictor of electoral success than many election polls."

Selected References

• Adamic, L. A., and Glance, N. 2005. The political blogosphere and the 2004 US election: Divided they blog. In Proceedings of the 3rd International Workshop on Link Discovery, 36-43. Chicago, IL

• Koop, R., and Jansen, H. J. 2009. Political blogs and blogrolls in Canada: Forums for democratic deliberation? Social Science Computer Review, 27(2): 155-173

• Pennebaker, J.; Chung, C.; and Ireland, M. 2007. The development and psychometric properties of LIWC2007. Austin, TX.

• Schneider, S. M. 1996. Creating a democratic public sphere through political discussion. Social Science Computer Review, 14(3): 373-392

• Sunstein, Cass. 2007. Neither Hayek nor Habermas. Public Choice, 134(1-2): 87-95

• Véronis, J. 2007. Citations dans la presse et résultats du premier tour de la présidentielle 2007. http://aixtal.blogspot.com/ 2007/04/2007-la-presse-fait-mieux-que-les.html

Theoretical and practical relevance:

Twitter volume is used to predict the election. Some deliberation and persuasion is occurring in political conversations in Twitter, but a minority of Twitter users are generating the volume of tweets.