Why Political Messages Spread

• In Event: Non-Stop: Platform Affordances, Diffusion, and the News

Abstract

Social media has become a more prominent feature in modern American politics in recent years. President Trump’s tweets are often covered in the national news, as were social media messages by various candidates during the 2016 election. Social media platforms are being used widely in Congress today, as every U.S. Senator has at least one account (see Straus et al. 2016), and many use them to reach wide audiences on a variety of subjects. For example, during the Senate’s consideration of a “skinny” repeal of the Affordable Care Act in 2017, Senator Chris Murphy (D-CT) issued a series of tweets, hosted a Facebook Live session, and participated in a podcast recording in advance of the vote. Democratic members even used social media to show live coverage of a protest on the floor of the House of Representatives after Speaker Ryan ordered the chamber’s cameras to be turned off (Fandos and Isaac 2016).

As social media has gained use and attention in political life, scholars have examined these platforms with increasing frequency (Gainous and Wagner 2014; Stromer-Galley 2014). Political scientists have explored various dimensions of social media use in Congress including the characteristics of members who adopt it (e.g., Gainous and Wagner 2014; Straus et al. 2013; Glassman et al. 2009), the frequency with which legislators use it (e.g., Lassen and Brown 2011; Straus et al. 2016; Lawless 2012), and the content of the messages they generate (e.g., Auter and Fine 2016; Gainous and Wagner 2014; Golbeck et al. 2010; Evans et al. 2014; Stromer-Galley et al. 2018). There also is work that examines social media’s potential to impact votes and outcomes (Gainous and Wagner 2014) and polarization in the political system (Settle 2018). Collectively, these elected officials generate thousands of social media messages daily; some of these messages ‘go viral’ while others are ignored altogether. Although recent scholarship sheds light on who uses social media and what content they generate, we know comparatively less about what happens with these messages after they are created (for a notable exception, see Stromer-Galley et al. 2018). To fully understand the impact of social media in American politics, we need insight into why individual social media messages spread, especially those generated by political elites. The current paper fills this gap in the literature by examining the factors that affect elite message diffusion on social media.
Specifically, we examine the impact of attacks and negative sentiment on the number of times a particular message generated by a members of the U.S. House and Senate is shared (retweeted) on Twitter. We assess attacks and negative sentiment in a sample of members'/senators’ tweets through two methods: (1) content analysis with multiple human coders and (2) coding using IBM Watson’s Natural Language Processing tool. We find that political messages attacking others are more likely to spread, as are those with more negative sentiment. We also find that messages are most likely to spread if they contain language that exhibits negative emotions like disgust. Each of these effects persist even when controlling for whether the tweet addresses a policy issue, the party and ideological extremity of those who generated the tweet, and the number of followers they have on Twitter. These findings fit with the broader literature on emotionality of messages and their diffusion in person and online.

In addition to providing insight into why political messages spread, our results also have implications for polarization in American politics (Settle 2018). As tweets that attack and those with negative sentiment spread much more than tweets with information or other types of messages, senators who covet social media attention may have an incentive to dial up the negativity and further polarize the landscape in Washington.

Author

• Jeffrey A. Fine, Clemson University