An online presentation by Elliot Ash (ETH Zurich). This event is part of the Political Economy Lunch Seminar.
(with Philine Widmer and Sergio Galletta)
This paper provides causal evidence on how partisan news messaging from cable television influences the content published by newspapers in U.S. localities. We introduce a new parallel corpus of newspaper articles (24M articles in 600+ local newspapers) and transcribed television news shows (40K cable news episodes from Fox News Channel, CNN, and MSNBC) for the years 2005-2008. We measure media influence using a supervised learning model that predicts, for a given piece of text, the probability that it comes from a Fox News transcript, rather than from CNN or MSNBC. After validating the measure, we apply it to the local newspaper article texts. Exogenous variation in news viewership across localities comes from relative channel numbering, which we use as instruments. We find that an exogenous increase in local viewership of a cable news network shifts the textual content of local newspapers toward that network's content. Televised media slant works not just through persuading viewers, but through influencing other media outlets.