To What Extent Sentiment Analysis of Social Media is able to Forecast Electoral Results? Evidence from France, the United States, and Italy

The growing usage of social media by a wider audience of citizens sharply increases the possibility to investigate the web as a device to explore and track policy and political preferences. Still relying on a proper methodology for sentiment analysis remains a crucial issue. In this presentation we apply the recent supervised method proposed by Hopkins and King (2010) to analyze the voting intention of Twitter-users in three different contexts: France (for the 2012 Presidential ballot and the subsequent Legislative election), US (for the 2012 Presidential election), and Italy (for the two-rounds of the centre-left 2012 primaries). Despite internet users are not necessarily representative of the whole population of country’s citizens, our analysis shows a remarkable ability of Twitter to forecast electoral results, in some cases improving on mass surveys results. On the contrary, unsupervised sentiment analyses provide results far less satisfactory than ours. We also illustrate that the predictive ability of the method we employ strengthens as the number of citizens’ expressing on-line their opinion grows, provided they act consistently on that.