Abstract
The vote share of the 2010 UK general election is forecasted here by applying twice the concept behind Galton's predictive "wisdom of the crowds," first, by aggregating at the media level (Facebook, Twitter, Twitter Sentiment, YouTube, Google) the political opinion of the audience and second, by averaging at the media level each prediction. The auto regressive integrated moving average (ARIMA) model performs predictions for the Labour Party, Liberal-Democrats, and Conservatives that are respectively only 0.48, 0.83, and 0.19 percentage points off the real vote share, thus well exceeding the predictive power of more traditional and expensive polls. © 2013 Copyright Taylor & Francis Group, LLC.