PREDICTIVE POWER OF TWITTER? CASE: THE NEXT POPE
On February 11, 2013, Pope Benedict XVI announced that he resigns and immediately speculations about who will be the next pope started. Newspapers started to list most likely candidates and bet offices started to take bets on possible candidates (e.g. http://www.paddypower.com/bet/novelty-betting/current-affairs/pope-betting).
The wisdom of crowds on Twitter have earlier been mined to make predictions about stock markets (Bollen et al., 2011) to box-office revenues (Asur & Huberman, 2010), and from flu epidemics (Lampos & Cristianini, 2010) to outcomes of elections (Tumasjan et al., 2010). However, it has also been argued that the predictive power of Twitter has been exaggerated and too little research has focused on developing reproducible methods to validate the results obtained from social media data (Gayo-Avello, 2012).
With this in mind, would it be possible to use Twitter to predict who will be the next Pope? Or at least say something about who would be the popular choice of Twitter users.
On February 12, 2013, data collection for tweets containing ”next pope” or ”new pope” was started from Twitter. By March 11, 2013, (the day before the conclave is planned to meet) a total of 291,018 tweets had been collected. These were queried for mentions of the 115 cardinal electors in the papal conclave and for the cardinals in papabili (which overlapped without one exception). The cardinals that were mentioned more than 100 times are listed below.
Next time series analysis were done with the most frequently mentioned names. The time series show that some cardinals from the papabili gained immediate mentions that didn’t last (e.g. Scola), while others only gained mentions a bit later (e.g. Bertone, Wuerl, Mahony and O’Malley). These peaks are probably due to some newspapers publishing articles listing these cardinals as strong candidates.
Ouellet has been mentioned during the whole month of data collection, but clearly less frequently than the most frequently mentioned cardinals.
The three most frequently mentioned cardinals Turkson, Dolan and Tagle have some clear peaks but they also get mentioned during almost the whole month.
Is it now, based on this data, possible to say who is the most likely winner? No, at least not without doing some content analysis on the tweets that made these graphs. The graphs alone do not tell use whether the mentions are positive, negative, neutral or perhaps just spam. This is also the next step of this research. In the meanwhile, you can make your own interpretations from the graphs above, and if you want, place your bets and wait for white smoke to rise from the chimney of Sistine chapel.
Asur, S. & Huberman, B.A. (2010). Predicting the Future with Social Media. arXiv preprint archive, arXiv:1003.5699 [cs.CY]. [link].
Bollen, J., Mao, H. & Zeng, X. (2011). Twitter mood predicts the stock market. Journal of Computational Science, vol. 2, no. 1, pp. 1–8.
Gayo-Avello, D. (2012). “I Wanted to Predict Elections with Twitter and all I got was this Lousy Paper” — A Balanced Survey on Election Prediction using Twitter Data. arXiv preprint archive, arXiv:1204.6441 [cs.CY]. [link]
Lampos, V. & Cristianini, N. (2010) Tracking the flu pandemic by monitoring the Social Web. In Proceedings of the 2nd International Workshop on Cognitive Information Processing (CIP). [link]
Tumasjan, A., Sprenger, T.O., Sandner, P.G. & Welpe, I.M. (2010). Predicting Elections with Twitter: What 140 Characters Reveal about Political Sentiment. In Proceedings of the Fourth International AAAI Conference on Weblogs and Social Media. [link]