Conference Agenda

Overview and details of the sessions of this conference. Please select a date or location to show only sessions at that day or location. Please select a single session for detailed view (with abstracts and downloads if available).

Session Overview
C 3: Campaigning and Social Media
Thursday, 10/Sep/2020:
3:30 - 4:30

Session Chair: Pirmin Stöckle, University of Mannheim, Germany

Show help for 'Increase or decrease the abstract text size'

Cross-Platform Social Media Campaigning: Comparing Strategic Political Messaging across Facebook and Twitter in the 2016 US Election

Michael Bossetta1, Jennifer Stromer-Galley2, Jeff Hemsley2

1Lund University, Sweden; 2Syracuse University, United States of America

Relevance & Research Question: This study is the first in the American context to compare political candidates’ social media communication across multiple social media platforms. This topic is relevant, as the large majority of digital political communication studies only focus on one social media platform. We therefore ask two research questions:

RQ1: Do political campaigns broadcast the same messages across multiple social media accounts, or does campaign messaging differ depending on the platform?

RQ2: What explains the similarity or difference in political campaigning across social media platforms?

Methods & Data: We combine three types of computational analysis – fuzzy string matching, automated content analysis, and machine learning classification – to compare the Facebook and Twitter posts of Hillary Clinton and Donald Trump during the 2016 U.S. Election.

Results: Our results show a relatively high degree of content recycling across platforms. At the highest level, over 60% of Clinton’s Facebook posts were also present on Twitter, whereas approximately 1/4 of the Trump campaigns posts were recycled across the two platforms. We do, however, find key strategic differences relating to how this content was conveyed to electorate. Our machine learning algorithm categorized posts by topic issues and message type, and we found the latter to be a significant predictor of platform differentiation through chi-squared tests. That is, candidates promoted the same policy issues across platforms, but the strategic intent behind their messages differed. Most notably, the Clinton campaign messaged Hispanic audiences in Spanish solely on Facebook. The Trump campaign promoted livestreams predominantly on Facebook, while reserving Twitter for broadcasting information relating to mass media interviews.

Added Value: The added value of the study is two-fold. First, while the state-of-the-art suggests candidates use different platforms for different messaging, we find a relatively high degree of content recycling across platforms. Moreover, we go beyond the existing literature and uncover what explains differences in cross-platforms posts. It is not the policy content of messages, but rather the strategic motivations that campaigns perceive in light of the audiences on each social media platform.

No need to constantly innovate: Interesting lessons from two election campaigns within a year

Yaron Ariel, Dana Weimann-Saks, Vered Elishar Malka

Academic College of Emek Yezreel, Israel

Relevance and research question: During more then a decade now, political candidates have been using central social networks as their leading platforms in election campaigns, constantly trying to improve their performance and enhance their influence over potential voters. In 2019 only, Israel has seen two general elections. Comparing patterns of online platforms' political usage between these two campaigns reveals some surprising changes: within a few months, innovative components that were used in the first election campaign were abandoned in the second (e.i. using live television broadcasts on candidates' Facebook accounts). Could these changes indicate a broader phenomenon? Can we detect evidence of a decline in exposure to innovative platforms already in the first campaign?

Methods and data: Four consecutive surveys were passed during the last month before the first 2019 Israeli general elections, with 520-542 respondents participated in each. The samples represented the Israeli voters' population and were transmitted via an online panel to match the actual distribution of the population. The questionnaire included 50 questions on voting trends, news exposure patterns, and political content exposure in traditional and new media, most of them on a Likert scale.

Results: A One-Way Analysis of Variance was conducted to examine overall exposure to online political content. No significant difference was found at the four-time points examined [F (3, 2153) =0.271, p> 0.05]. More specifically, a Kruskal-Wallis test was performed to examine whether there was a statistically significant difference in the use of various social media applications. There were no significant differences in the consumption of Facebook, Twitter, and Telegram apps; however, a change in exposure to political content was detected on Instagram [Kruskal-Wallis H = 9.42, df=3, p < 0.05] with decreased of the mean rank score.

Added value: The findings of the current study suggest that politicians' attempts to be innovative in online media are not necessarily effective. Despite politicians' efforts, there is no detectable increase in exposure to new platforms as Election Day approaches.

The Sequencing Method: Analyzing Election Campaigns with Prediction Markets

Oliver Strijbis

University of Zurich, Switzerland

Relevance & Research Question:

What are the effects of campaigns on voting behavior? Despite a long tradition in research on the question we know surprisingly little about it. A main reason is that the analysis of polls of polls—the most important method for the comparative analysis of campaign effects—is confronted with formidable methodological problems. In this paper, I propose and apply an alternative method for the analysis of campaign effects on voting behavior, which is based on prediction markets.

Methods & Data:

The proposed method to analyze campaign effects in elections and direct democratic votes is based on real money online prediction markets with automatic market scoring rules. In contrast to the current use of prediction markets, I propose to let traders bet on the probabilities according to which sequences of vote shares will be the outcome of a vote. I use original data that I have collected in the context of 16 direct democratic votes in Switzerland.


I demonstrate that on average direct democratic campaigns in Switzerland have substantial mobilizing effects and make up for about 5% of Yes-vote shares. Furthermore, campaign effects vary in predictable ways between ballots.

Added Value:

I illustrate this "sequencing method" with the first time–series cross–section analysis of direct democratic campaigns. Since this is the first paper to quantify the total effect of campaigns for direct democratic decisions it has a major impact on our understanding of direct democracy.