Mass Media and Electoral Preferences during the 2016 US Presidential Race

Christopher Wlezien and Stuart Soroka

Research typically concentrates on how mass media influences voters’ perceptions and preferences.   This is not surprising – we know that election campaigns spend a good deal of time and effort trying to influence the public, and that most campaign information reaches voters through mass media.  But there also are good reasons for news coverage to reflect public opinion rather than affect it, some sociological and others economic.  This expectation may surprise some readers, but we do not regard it as original or especially contentious, just under-studied and so not well understood.  Our recent work explores this aspect of the relationship between media and citizens, focusing on the 2016 US presidential election.  Over roughly 200 days leading up to the November election, we examine the bi-directional relationship between vote intentions and news content.

Our analyses rely on two time series: (1) vote intentions and (2) media “tone.”  The first is based on the many trial-heat polls conducted during the 2016 election year, drawing on data archived by the Huffington Post.  Specifically, we rely on 308 separate national polls during the election year, 100 of which from after the unofficial Labor Day kickoff of the general election campaign.  We aggregate these data to create time series of macro electoral preferences.  The second time series captures the balance of sentiment in news coverage of the two main candidates.  We capture the tone of media coverage for each candidate from roughly 30,000 artices from nine major US newspapers during the election year.  The tone of coverage is then identified using the Lexicoder Sentiment Dictionary (LSD), a simple dictionary-based approach to sentiment analysis.  Using these data, we produce net Clinton-Trump tone in daily coverage.  The analysis relates these two measures, with a focus on whether and when media coverage leads and/or follows voter intentions.

We begin with a basic first-order autoregressive distributed lag (ADL) models relating the variables, where each variable is regressed on lagged values of both itself and the other variable.  This allows us to see whether one variable influences the other, independent of the latter’s own history.  Results show that media coverage does predict vote intentions, but also that those preferences predict the tone of coverage.  That is, the media appeared to both lead and follow the public during the 2016 campaign.

We then incorporate the effects of various campaign events into the analysis.  Doing so reveals that both media coverage and voter preferences reflected certain events, primarily the conventions and the initial Comey intrusion. Taking account of these events, media tone appears to have little impact on voters.  Indeed, the relationship between media and public opinion appears to have changed over the course of the election, and the final weeks of the campaign show no effect of media coverage on electoral preferences.  There similarly is no effect of electoral preferences on the tone of media coverage, at least over short time horizons.  There is some evidence of followership over longer time horizons (beyond several days), however. In sum, media coverage in 2016 appears to have followed electoral preferences at least as much as it led them.

Our analysis has focused on a single election in a single year.  What about presidential elections in other years?  What does the research tell us about the effects of election campaigns more generally?  Clearly, one cannot generalize based on an analysis of 2016 alone.  We nevertheless expect that similar analyses of news coverage will find instances in which media powerfully lead opinion and others where they do not.  Indeed, it may be that the relationship between media coverage and vote intentions varies over the course of campaigns, as seemingly was the case in the 2016 US presidential election.  It is for subsequent work to contemplate the broader role of news coverage – as both leader and follower – during election campaigns in the US and other countries.

Christopher Wlezien is Hogg Professor of Government at the University of Texas at Austin.  Further information about him and his work is available on his website.

Stuart Soroka is Michael Traugott Professor of Communication Studies and Political Science at the University of Michigan.  Further information is available on his website.



Leave a Reply

Fill in your details below or click an icon to log in: Logo

You are commenting using your account. Log Out /  Change )

Google+ photo

You are commenting using your Google+ account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s