Telecracy: Testing for Channels of Persuasion

Can biased information persuade in the long run? Political information on Italian TV has been biased towards Berlusconi's party since 1994. We exploit a shock to viewers' exposure to Berlusconi bias: idiosyncratic deadlines to switch to digital TV from 2008 to 2012. Digital TV increased the number of freeview channels tenfold. As a consequence, viewership of Berlusconi-controlled channels by digital users dropped by 19% from 2008 to 2010 elections. Although the control of most pre-digital outlets by Berlusconi was widely known, the switch caused a drop in his coalition vote share by 5.5 to 7.5 percentage points. The e ect was stronger in towns with older and less educated voters. At least 30% of digital users had not ltered out the bias from 1994 to 2010. Moving to digital TV a ected voting via turnout: previous Berlusconi supporters went to vote less than others, hence his vote share dropped. We discuss several Bayesian interpretations, and argue that they cannot fully explain these results. Coarse thinking, selective attention, and persuasion bias are broadly consistent with the evidence.