BigSurv18 program
Wednesday 24th October Thursday 25th October Friday 26th October Saturday 27th October
Using Big Data for Electoral Research II: Likes, Tweets, and Votes? |
|
| Chair | Dr Josh Pasek (University of Michigan) |
| Time | Saturday 27th October, 16:00 - 17:30 |
| Room: | 40.006 |
Gauging the Horserace Buzz: How the Public Engages With Election Polls on Twitter
Ms Colleen McClain (University of Michigan) - Presenting Author
Mr Ozan Kuru (University of Michigan)
Dr Josh Pasek (University of Michigan)
Given the prominence of horserace news coverage, individuals are constantly bombarded with the latest poll results throughout presidential election campaigns. At the same time, today’s social media climate makes it easier than ever to engage with stories about polls online via liking, sharing, and commenting. While the effects of polling exposure on individuals’ political attitudes and behaviors are increasingly studied via surveys and experimental designs, data available via social media content provide us with new opportunities to investigate public engagement with poll coverage over the course of a campaign. Existing research suggests that polls that are close and polls that are different are more likely to be covered in traditional news media than their contemporaries, resulting in a biased set of results receiving disproportionate attention by the electorate. Whether these trends are amplified or mitigated by social media sharing, however, is an open question—and one with potential implications for perceptions of a given race, of the polling industry generally, and of the electoral process.
To this end, we provide a first look at several research questions in the current work. First, we ask how much social media attention and engagement polls elicit over the course of a campaign. Second, we examine whether polls receive differential levels of attention on social media based on their deviation from polling average results at a given time or the closeness of their result. To assess these questions, we analyze a large corpus of tweets with references to polls gathered over the course of the 2016 presidential election campaign. We define a set of conditions under which we can reasonably link large volumes of tweets mentioning polls to the specific poll results they refer to, using a lookup table referencing polling results from HuffPost Pollster. In doing so, we generate a novel metric to track the “result extremity” of a given tweet in comparison to the polling average over large volumes of data and over time. We analyze trends in tweet volume as a function of survey house, result extremity, and result closeness over time using a variety of computational techniques. Building on this work, we suggest directions for micro- and macro-level analyses of the dynamics of poll coverage on Twitter, keeping in mind the potential inferential problems with analyzing corpora of social media data from a total error/quality perspective. Importantly, our results contribute to an understanding of the public’s engagement with and perceptions of polls and public opinion in an increasingly volatile climate for public opinion research.
Social Media and Political Participation
Mr Sascha Göbel (University of Konstanz) - Presenting Author
Social media provides the opportunity to observe real-world behavior in real time without most of the limitations of, for instance, surveys. In principle, this lends social media data tremendous potential for researching general patterns and dynamics of individual behavior. Yet, political science research using social media data still largely focuses on social media as such and investigates its usage, structure, or impact. The impediments to using social media data for more general inferences are its non-probability nature, the shortage of user demographics, and selection based on a specific outcome or content. Thus, one data source’s weakness is another data source’s strength. I salvage social media’s untapped potential and bridge the limitations of traditional and social media data by linking publicly available US voter records of the state of Florida to the respective Twitter accounts. This linkage provides rich demographics of Twitter users, turnout histories, and knowledge of a clear target population (registered voters) which facilitates post-stratification for population-level inferences with known levels of confidence. The method further outperforms previous attempts linking voter records with social media data in terms of accuracy, replicability, and success rate. The social media activity for more than 80,000 registered voters in Florida will be collected in a daily rhythm until at least two months after the 2018 US midterm elections. Text analysis methods are continually used to distill (politically) relevant text features and quantities from social media activity. Substantively, the gathered data, including measures of actual offline and online political behavior, are used to study the relationship between online and offline political participation, to identify voter types based on general attitudes towards politics in a latent class model and relate them to turnout propensities, and to yield a proof of concept of social media-based individual-level election forecasts for House districts via multilevel regression and post-stratification. Hence, this project presents a case in point for how data linkage can improve on traditional survey and off-the-shelf social media data as well as enhance the study of general patterns and dynamics of political behavior.
Digital 'Cold Wars' in the U.S. Mid-Term Election
Mr Mark Polyak (Ipsos Public Affairs) - Presenting Author
Showcase near real time detection of social media “political bots” during the US mid-term elections. This pilot will identify which candidates and their policies are being targeted by political bots either for positive or negative amplification of political messaging.
The project correlates these attempts at digital strategic communication with daily estimate of candidates’ social media favorability and traditional polling favorability scores. The project is the first of its type to attempt to provide an integrated approach for combining polling, social media and “bot detection” for measuring impact of digital communication campaigns in a nationwide US election. The project relies on NLP and deep learning approaches for cross-correlating multi-stream data inputs and bot detection.
Scaling the Civility Wall: Examining Social Media Discourse on Immigration During the 2016 U.S. Presidential Election
Dr Thomas Johnson (University of Texas at Austin) - Presenting Author
Mrs Heloisa Aruth Sturm (University of Texas at Austin)
Dr Patricia Rossini (Information Studies, Syracuse University)
Public discussion about pressing issues is essential to a healthy democracy, and most recent scholarship on deliberative politics has addressed how social media platforms have reshaped political communication online. While some researchers have argued uncivil discourse may increase political polarization (Mutz, 2006; Anderson et al, 2016), others contend heated online discussions are responsible for a revived public sphere, promoting democratic emancipation through disagreement (Papacharissi, 2004).
This study intends to examine a key topic of the 2016 elections – immigration – through the lens of civility research, examining effects of online public talk on incivility. Recent studies have shown that deliberation and incivility can co-exist in digital spaces, and less extreme instances of incivility may actually boost political participation (Chen, 2017).
More specifically, this study examines the campaign messages concerning immigration delivered on Twitter and Facebook by Republican candidates Donald Trump and Ted Cruz, as well as Democratic candidates Hillary Clinton and Bernie Sanders. Trump’s rhetoric when addressing immigration (“Build that wall”) and Cruz’s voter base in the border state of Texas made them the most prominent Republican candidates addressing this topic while leading Democrats Clinton and Sanders presented a positive message on the immigration issue. In addition, the comments sparked by those candidates’ posts will also be analyzed.
The data for this study is from Illuminating 2016, a computational journalism project led by Columbia University and Syracuse University that focuses on how presidential candidates communicate through social media and what types of campaign messages resonate with the public (Stromer-Galley et al, 2016). The project uses a lexicon-based approach to create clusters of salient issues and to identify political topics in social media messages (Jackson et al, 2017). The dataset includes 4,261 Facebook and 19,283 Twitter messages sent between January 2015 and November 2016 from all the Republican, Democratic, and third-party presidential candidates, as well as the public discussions elicited by these messages.
The focus of analysis will be on messages delivered by the four candidates that either attack or advocate for immigration policies as well as on comments posted by the public concerning immigration. How do the four candidates’ posts differ in degree of civility? How is civility of the candidates’ posts related to civility of the commenters? This analysis will also address who and what are the different targets of negative posts on immigration. Have candidates become the target of negative comments even on their own posts? Finally, we will examine differences between Twitter and Facebook in civility of posts and comments.
Although incivility is nothing new in U.S. politics, the speed and range of online communication have normalized and weaponized that practice (Chen, 2017), spurring a need for studying uncivil discussion within the 2016 election context. While most research has focused on overall commenting on newspapers' websites or their Facebook pages, this study contributes to the nascent literature by focusing on a specific topic and examining how uncivil discourse emerges through the interactions between political actors and their social media base.
Analyzing Right-Wing Discourse on Twitter: A Case Study of the 2017 German Federal Election
Ms Aubrey O'Neal (University of Texas at Austin) - Presenting Author
Download presentation 2
The global rise of populism is made increasingly apparent in this year's series of elections and nationalist protests. The case of Germany's federal election, in which a fledgling far-right party, Alternativ für Deutschland, made its way into parliament, provides an important case study given Germany's history and placement as a world power. This study explores textual and visual discourse on Twitter surrounding the 2017 German federal election, exploring themes in conversation used by and against AfD constituents. Given that social media activity is linked to the events of a specific country, this study makes a distinct separation between Twitter users within specific German voter regions and users in the greater global community. Within these geographically bound areas, tweets are analyzed for sentiment towards AfD as well as common themes in conversation. Additionally, Tweet images are organized into clusters using the Google Vision API, and qualitatively analyzed for themes using a grounded-theory approach. This study finds that tweet volume about a party is not directly indicative of sentiment or success of the party on election day. However, the less the volume of tweets about AfD in a region, the greater support for AfD on election day, suggesting that the Spiral of Silence applies both online and offline on this topic or that the AfD demographic is largely not on Twitter. While more than half (57.2%) of all election tweets pertained to AfD, 39.2% of these tweets were supportive of the party, while 43.5% were in fact critical. Thematically, this study finds that the AfD supporters relied heavily upon criticizing Chancellor Angela Merkel and the established party, and de-emphasized the party platform. Meanwhile, the opposition tended to discuss ideologies and history (Nazism, right-extremism, populism, immigration) and to plea for others to vote.