MEDIA SLANT IS CONTAGIOUS: Evidence of high Fox News exposure in US counties influencing local newspaper content
12 Apr 2021
Does politically biased news messaging – the ‘media slant’ – of prominent media outlets distort other media organisations' news output? Recent work by researchers from ETH Zurich, the University of Bergamo and the University of St Gallen suggests that such media slant is indeed contagious. In US counties where exposure to Fox News is higher (relative to CNN and MSNBC), local newspapers' content resembles Fox News (rather than CNN and MSNBC).
The findings by Philine Widmer and colleagues imply that a higher local viewership of a cable news network shifts local newspapers' textual content toward that network's content. Televised media slant thus works not just through persuading viewers but also through influencing other media outlets. This effect seems to run mainly through the framing of news items rather than the choice of topics.
The researchers start with a corpus of 40,000 transcripts of cable news episodes from Fox News Channel (FNC), CNN, and MSNBC for 2005 to 2008. Using a supervised machine learning model, the researchers identify the most characteristic phrases for FNC on the one hand or CNN and MSNBC on the other hand. For example, the term ‘far left’ is more predictive of FNC, while the term ‘far right’ is often used by CNN and MSNBC.
Next, they collect over 24 million newspaper articles published in over 600 local newspapers. Local newspapers are essential drivers of citizen engagement and allow citizens to hold their political representatives accountable. They construct a measure of how similar its articles are with FNC (relative to CNN and MSNBC) for each newspaper. The similarity measure is based on the newspapers' use of the characteristic FNC phrases (versus CNN and MSNBC phrases).
The key research question is if local newspapers report in a more FNC-like fashion in counties where FNC is watched more. A challenge, thereby, is that both cable news viewership and newspaper content are influenced by variables such as the population's general political preferences. Such variables can confound a naive correlation of newspaper content similarity with FNC and FNC viewership.
Thus, this work relies on some variation in cable news viewership that is not associated with other county-specific characteristics: the relative channel numbering of FNC versus CNN/MSNBC. Various historical or geographical idiosyncrasies determine the channel position of a cable network. Notably, channel numbering is not only random but also predictive of viewership. Channels with a low position are, all else equal, consumed more than those with a higher position (they are more easily accessible when a viewer turns on their TV).
The research exploits this quasi-random variation in viewership from the channel numbering. The findings imply that a higher local viewership of a cable news network shifts local newspapers' textual content toward that network's content. Televised media slant thus works not just through persuading viewers but also through influencing other media outlets. This effect seems to run mainly through the framing of news items rather than the choice of topics.
Earlier contributions in economics and political sciences document that partisan media can affect voters' and politicians' behavior. Whether or not such persuasive effects can become a problem for the quality of democratic processes depends on many factors.
Importantly, the influences of specific kinds of politically biased news depend on whether they are embedded in a diverse news landscape where citizens can access information from various sources. If politically biased news by one outlet also distorts other organisations' output, different news sources are not independent in their reporting. Thus, policy-makers in charge of media regulation need to understand the contagion of content between outlets.