How Leaders Use Social Media to Estimate the Costs of Subverting Democracy

• In Event: Democracy, Autocracy and the Internet

Abstract

A distinguishing feature of the recent trend toward democratic backsliding in advanced economies is the gradual, methodical undermining of institutions and basic freedoms by elected leaders. This piecemeal approach stands in stark contrast to previous episodes of democratic breakdown, which were often the result of military coups. Democratic reversals in Latin America, Africa and East Asia between 1960 and 2000 were generally abrupt, while recent episodes of erosion in Turkey and Western democracies like the United States fit the gradual pattern.

We argue that new technologies, especially social media platforms and the data they generate, play a key role in gradual democratic erosion. A majority of extant work on this issue has focused on the negative effects of new media on its consumers, i.e. the voters. In contrast, we make a supply-side argument: social media sites help political leaders calculate with greater certainty the political costs of challenging institutions and curtailing freedoms. Leaders can send out information on a regular basis and obtain feedback from relevant groups: their loyal base, the party as a whole and the opposition. This detailed feedback allows politicians to estimate the costs of their actions more accurately and estimate which types of attacks on democracy are likely to face lower resistance. Indeed, subverting or gradually eroding democratic rules and institutions requires more specific information on the preferences and reactions of voters, groups, and institutions. Elections only generate a limited and imperfect amount of information on social support, and social media becomes key in determining how to attack democratic institutions.

Our point of departure is that some political leaders have an explicit agenda to subvert democratic practices and concentrate power in their hands. This has occurred, arguably, in Turkey, Hungary, Poland, and the United States. Yet the sequence of attack on institutions used by political leaders is still unclear, and we think social media can help us understand it better. We focus on three aspects of supply-side social media: (1) the nature of the reaction from the politician's most loyal base, (2) the reaction by the leader's broader coalition (the party and its members), and (3) the reaction of institutions outside the leader's coalition --the opposition and the courts, usually. An excited base, combined with a favorable outlook by the party and a lukewarm response by the opposition, usually leads to a greater push on a given issue. Lack of enthusiasm with the loyal base and strong opposition by outside institutions generally leads to postponing an attack as it is deemed too costly to undertake at the moment. We see then supply-side media as an instrument facilitating a social-cost discovery process beyond election results.

We use the case of the United States under Donald Trump to test the argument. We have collected all of President Trump's tweets since his election in November 2016, a total of 14,179 tweets, and analyzed them in multiple ways. First, we have plotted how often certain topics appear in his tweets and clear patterns emerge. An attack on the media was launched first and Trump's loyal base received it with enthusiasm, with some of the highest number of retweets and likes coming for tweets containing the words `fake news'. The term `fake news' disappeared completely afterward for a full year and a half. We contend it was no longer necessary to attack the media after the first attack had been so successful with the loyal base and the party. References to `media' confirm this pattern: they were by far most popular within Trump's first six months in office. Trump then moved to immigration in the second half of 2017 and early 2018. A sentiment analysis of these tweets reveals that they are highly negative in tone, almost two standard deviations below Trump's non-immigration related tweets. Attacks on the court system and judges have become increasingly common and more negative in tone, peaking in late 2019 amidst the impeachment proceedings. Conversely, terms such as `constitution' are practically absent from Trump's tweets --references only appear also in late 2019 and are related to impeachment as well. 42 tweets contain the word `constitution' and only 5 of these were published before 2019. There are also no attacks on the military.

This initial analysis gives us some early evidence that leaders use social media strategically on topics that are central to democratic erosion. With this big data approach, we will further explore the content of these tweets and investigate more precisely who interacted with them (retweet, like, or comment). opposition fought back. An analysis of Trump's tweets at this level of detail is unprecedented and can give us great insight to determine the extent to which leaders use social media as a tool to obtain information as they build their strategy to undermine democracy.

Authors

• Joan C. Timoneda, Purdue University

• Abel Escriba-Folch, Institut Barcelona d'Estudis Internacionals