Panem Et Circenses: The Persuasive Effect of Soft News in Berlusconi's Italy

Abstract

How do politicians influence news content? And are voters affected as a result? This paper contributes to the long-standing debate on media capture and bias by innovating in four directions. First, we develop a new measurement strategy to achieve content scores applying unsupervised text scaling on an original dataset of verbatim transcriptions of almost 20,000 hours of broadcasting from all main Italian televisions’ newscasts between 2010 and 2014. Crucially, transcriptions exclude all politicians’ airtime and thus allows us to identify purely content-induced media bias. Second, by exploiting a natural experiment, we document how during his tenure as a prime minister Mr. Berlusconi significantly altered public television coverage to mimic his private companies by increasing the soft news concentration. Third, we investigate the mechanism of news manipulation, by identifying a filtering, or agenda-setting, and a framing component. Finally, we couple the estimated content scores with a representative panel dataset of Italian voters documenting that content variation increased Berlusconi’s electoral support. Results are discussed in their implications for voting behavior and voters’ democratic competence.