Can electoral popularity be predicted using socially generated big data?

Abstract: Today, our more-than-ever digital lives leave significant footprints in cyberspace. Large scale collections of these socially generated footprints, often known as big data, could help us to re-investigate different aspects of our social collective behaviour in a quantitative framework. In this contribution we discuss one such possibility: the monitoring and predicting of popularity dynamics of candidates and parties through the analysis of socially generated data on the web during electoral campaigns. Such data offer considerable possibility for improving our awareness of popularity dynamics. However they also suffer from significant drawbacks in terms of representativeness and generalisability. In this paper we discuss potential ways around such problems, suggesting the nature of different political systems and contexts might lend differing levels of predictive power to certain types of data source. We offer an initial exploratory test of these ideas, focussing on two data streams, Wikipedia page views and Google search queries. On the basis of this data, we present popularity dynamics from real case examples of recent elections in three different countries.