Exploring Correlations Between TV Viewership and Twitter Conversations in Italian Political Talk Shows

Abstract

The practice of using a ‘second screen’ while following a television program is quickly becoming a widespread phenomenon. When the secondary device is used to read or contribute to online comments about a watched program, most of the discussion takes place on popular social media such as Facebook and Twitter. Previous research has shown that the analysis of these contents could lead to a better understanding of the behavior of 'networked publics' and of the structure of the show itself. Leveraging on this background, the paper presents what is, to our knowledge, the first attempt of creating a statistical model with the goal of predicting the audience of a TV show from Twitter activity. During the last TV season, we collected all the Tweets (2,497,885) containing at least one of the official hash-tags of the eleven political talk shows (1,077 episodes) aired by the Italian free-to-air broadcasters from August 2012 to June 2013. We found a significant correlation between the Tweet rate and Twitter contributors per minute during airtime and the audience of the show’s episode. On this premise, we trained a multiple regression model on a random sample of 538 episodes (R^2 = 0.96) and successfully tested the model on remaining observations.