Twitter's Unheavenly Chorus? Assessing Inequalities of Voice Online

• In Event: Online Incivility: Outrage and Hate

Abstract

Political activity often tilts to a narrow corner of society, towards the affluent, the well resourced, men, and the highly educated (Schlozman, Verba, and Brady 2012). This is particularly true, of course, for the expenditure of dollars, but it is also true of a wide range of political action. In particular, most forums for political communication tend to replicate inequalities and diminish the voice of traditionally underrepresented groups (Mansbridge, 1983; Karpowitz and Mendelberg, 2014). In the early years, techno-optimists argued that the Internet would create new channels of political communication that could offer multifold pathways of connections to the political process, offering a democratizing platform for expression. Subsequent research suggested, instead, that emergent forms of expression, such as blogs, tilted to the same categories of people as the pre-Internet world (Hindman 2009).

Twitter potentially challenges this pattern because of its relatively low barrier to entry and relatively frictionless communication. In general, communication patterns are heavily influenced by design features of the forum in which the communication occurs. Recent research has shown (Kennedy et al., 2019) that online platforms can have designs that limit the advantages of those who traditionally hold communication advantages, for example, by limiting the ability to interrupt others, by removing non-verbal cues that shape interpersonal interactions, and by stripping out some (but not all) of the race, class, age or gender identifiers from the message itself. Understanding voice and attention on Twitter is extremely important because of its critical role in the current collective expression in politics. Twitter is the public town square that enables participation at scale in a sense that has never truly existed before.

We evaluate the question of voice on Twitter by examining a large panel of Twitter accounts linked to voter data, focusing on political expression during the 2016 election. This dataset was previously reported in Grinberg et al.’s (2019) study of the distribution of fake news on Twitter during the 2016 presidential election. The dataset matches 16,442 Twitter users who are registered U.S. voters with data provided by TargetSmart. The protocol to identify matches between Twitter accounts and the TargetSmart voter data required the name and location records to uniquely match within a U.S. state. The Grinberg study demonstrates the representativeness of the panel along the normal battery of demographic indicators.

Our overarching question is: Does Twitter reinforce or undermine existing biases in political expression? Our more specific empirical analyses will focus on the socio-cultural determinants of who chooses to express themselves on Twitter. We will explore, among those who tweet, and then specifically who chooses to tweet about politics?

The Twitter panel data records each panelist’s tweets, followers and followees. The Grinberg dataset includes validated classification of whether tweets are on political or non-political topics. To assess the relationship between race, age, education, gender and geographic region on voice on Twitter, the analysis will use count regression methods to test for the covariates that are associated with the rates that panel members tweet on any topic (all tweets) and compare those when the analysis is limited to political tweets. For comparison, we will model the propensity for the panelists to participate in traditional ways, such as voting and making contributions.

The Grinberg dataset is a unique opportunity to examine voice and representation on Twitter for the first time. Modeling the observed associations between our demographic covariates and these participation measures using regression is the optimal analysis for this project because the goal is to determine the presence or absence of disparities. As Twitter and other forms of digitally-mediated communication emerge, it is critical for our democracy to understand the properties of the platforms and how they shape political communication.


Grinberg et al. 2019. “Fake News on Twitter during the 2016 Presidential Election.” Science 25 Jan 2019: 374-378

Hindman. 2009. The Myth of Digital Democracy. Princeton UP.

Karpowitz. and Mendelberg. 2014. The Silent Sex. Princeton UP.

Kennedy et al. 2019. ``Demographics and (Equal?) Voice: Assessing Participation in Online Deliberative Sessions.'' Political Studies.

Mansbridge. 1983. Beyond Adversary Democracy. U Chicago Press.

Schlozman et al. 2012. The Unheavenly Chorus. Princeton UP.

Authors

• Stefan McCabe, Northeastern University

• Kevin M. Esterling, University of California, Riverside

• David Lazer, Northeastern University