Effects of Online News Use on Voting: Evidence from Surveys and Tracking

• In Event: The Effects of News and Ads on Participation

Abstract

Whether news media influence who people vote for is one of the oldest questions in political communication research (Lazarsfeld et al., 1948). The rise of more partisan outlets and the growth of social media, as well as a decline in party loyalties (Denver & Garnett, 2014), have all increased the relevance of studies on the persuasive potential of news media in elections. Against this backdrop it is perhaps surprising that there is “little in terms of systematic research […] that addresses the impact of media on voting behaviour” (Boomgaarden & Schmitt-Beck, 2016, p. 145).

In this study we use the context of the 2019 UK General Election to explore the effects of online media use on people’s voting behaviour. We focus in particular on the six-week campaign period, examining the effect of both news media use and social media use (Facebook, Twitter, YouTube) on (i) ‘activation’, where media use mobilises non-voters to vote, (ii) ‘conversion’, where voting intention is shifted from one party to another, and (iii) ‘de-activation’, where news media use causes people to dismiss early preferences and not vote (Cappella & Jamieson, 1997; Boomgaarden & Schmitt-Beck, 2016; Dilliplane, 2014).

Our data comes from a combination of passive internet tracking and online panel surveys (Stier et al., 2020). A representative sample of 2,161 UK internet users were tracked on desktop/laptop and mobile for six weeks before the campaign, and for the six-week campaign itself. Online news media use was coded as Conservative-endorsing, Labour-endorsing, or Non-endorsing because during UK General Elections all national newspapers explicitly endorse a political party (or make an explicit non-endorsement) and all broadcast outlets must adhere to strict guidelines ensuring that they do not favour one party or candidate over another. A subset of 838 panellists were surveyed at the beginning of the campaign (t1), of which 752 completed a second survey immediately after polling day (t2). Passive tracking provides more accurate and fine-grained measures of online media use than recall-based approaches (Guess, 2015), and by linking this to panel surveys, which were used to measure voting intention at the start of the campaign and immediately after polling day, we can make stronger causal claims about the effect of online media use than would be possible with cross-sectional data.

We analyze our data using multinomial logistic regression because our dependent variable, vote choice at t2, is categorical with more than two categories. We first analyse the data using an autoregressive model by controlling for vote choice at t1, and second using random-effects within-between model to explore within-person effects and between-person effects simultaneously–controlling for other variables such as political attitudes and offline media use where appropriate.

In line with previous research, we find some limited evidence of conversion effects where a small proportion of the sample switched their vote choice to a party endorsed by online news media they used. We also found some limited evidence of de-activation effects from social media use, perhaps because this shifted people’s attention away from politics and towards entertainment. However, we did not find any evidence of activation effects whereby people voted after previously thinking they would not.

Taken together, these findings offer some small support to the idea that partisan news media can shift people’s preferences during an election campaign. However, it is also clear from the analysis that the overwhelming majority stuck with their pre-campaign preferences and intentions, and among those that did change, other factors such as political interest and attitudes towards the European Union, were considerably more influential (Ford et al., 2021).

Authors

• Richard Fletcher, University of Oxford

• Anne Schulz, University of Zurich