Candidate Twitter Messaging in the 2022 U.S. Senate Election

• In Event: Analyzing Political Campaign Strategies across Platforms

Abstract

Evidence has suggested that many citizens use social media as a primary source of political information. This has led to fears among researchers that the selective algorithms of social media platforms may cause “echo chambers”, whereby citizens only receive information consistent with their existing beliefs. Much of this information comes from media outlets and/or candidate surrogates rather than the candidates themselves. This situation, coupled with the rise of auto-generating “bots” and the common presence of misinformation, calls into question whether political communication via social media is a positive or a negative force. Rather than enhancing civic engagement and voter education, social media has the potential to further polarize the electorate and negatively influence elections. As a semi-direct line of communication between themselves and the electorate, candidates can use their official social media accounts to try and cut through the noise and communicate their desired message.

Questions remain, however, about the role that the candidates’ social media communication plays in agenda setting and electoral outcomes and how candidate communications change over time. Questions also remain about whether candidates’ digital communications follow the more negative and toxic patterns often thought to be associated with social media-based political dialogue. The brevity of social media messaging mean that candidates must choose their words with an eye towards voter engagement. To determine patterns associated with this political communication, this study will explore the social media feeds of 2022 U.S. Senate candidates on one specific platform – Twitter (now known as X). This study focuses on the direct political communication of major party candidates to answer the following research questions:

(1) What were the primary topics of Twitter posts by U.S. Senate candidates, and how did that change over the course of election season (August – November)?
(2) What was the sentiment (positive, negative, or neutral) of Twitter posts by U.S. Senate candidates, and how did that change over the course of election season (August – November)?
(3) What was the toxicity of Twitter posts by U.S. Senate candidates, and how did that change over the course of election season (August – November)?
(4) What elements of Twitter posts (sentiment, toxicity, and linguistic choices) significantly predict the electoral outcomes of U.S. Senate candidates?

The study sample involves major party U.S. Senate candidates in 30 states, comprised of 32 Republican and 30 Democratic candidates in the 2022 midterm elections. Data from these candidates’ Twitter accounts was collected using public APIs via custom Python scripts. Data collection occurred for the 100-day period beginning August 1, 2022, and ending on November 8, 2022. This period was chosen due to it being the post-primary season, when final party candidates have been selected and voters focus on electing individual senators.

A combination of tools and techniques will be used to answer the research questions. Topic modeling, specifically Latent Dirichlet Allocation (LDA) techniques, will be used to identify the common topics discussed by candidates. This data will be examined for temporal patterns as well as for patterns by party affiliation and electoral outcome. Sentiment will be assessed using common tools (e.g. VADER), while toxicity will be assessed using the Google Perspective API. Linguistic choices, such as the use of emotional language, will be assessed using LIWC software. Binary logistic regression will be used to determine predictors for electoral outcomes.

By exploring the political communication of major party candidates during the height of the 2022 U.S. election season, this study will uncover patterns associated with topic selection, toxicity, sentiment, and language choices, as well as the predictive power of those elements towards electoral outcomes.

Author

• Jeffrey A. Stone, Penn State University