ABSTRACT
Prior research suggests that Twitter users in the United States are more politically engaged and more partisan than the American citizenry, who are generally characterized by low levels of political knowledge and disinterest in political affairs. This study seeks to understand this disconnect by conducting an observational analysis of the most popular accounts on American Twitter. We identify opinion leaders by drawing random samples of ordinary American Twitter users and observing whom they follow. We estimate the ideological leaning and political relevance of these opinion leaders and crowdsource estimates of perceived ideology. We find little evidence that American Twitter is as politicized as it is made out to be, with politics and hard news outlets constituting a small subset of these opinion leaders. Ordinary Americans are significantly more likely to follow nonpolitical opinion leaders on Twitter than political opinion leaders. We find no evidence of polarization among these opinion leaders either. While a few political professional categories are more polarized than others, the overall polarization dissipates when we factor in the rate at which the opinion leaders tweet: a large number of vocal nonpartisan opinion leaders drowns out the partisan voices on the platform. Our results suggest that the degree to which Twitter is political has likely been overstated in the past. Our findings have implications about how we use Twitter and social media, in general, to represent public opinion in the United States.
Supplementary Material
Supplemental data for this article can be accessed on the publisher’s website at https://doi.org/10.1080/10584609.2022.2075061.
Disclosure Statement
No potential conflict of interest was reported by the author(s).
Data Availability Statement
The data described in this article are openly available in the Open Science Framework at 10.17605/OSF.IO/QDVT7.
Open Scholarship
This article has earned the Center for Open Science badge for Open Data. The data are openly accessible at 10.17605/OSF.IO/QDVT7.
Notes
1. Respondents to the Pew Survey were, for instance, 2–3 more likely to be registered voters than the population at large.
2. We validate this assumption in the Appendix using crowdsourced ideological scores for those “neutral” opinion leaders.
3. Only the MTurk survey was pre-registered. All other data used in this study were collected before the survey was conducted. Pre-registration details are publicly available at https://osf.io/ytkj7 and https://osf.io/uj4ed respectively. We needed two pre-registered surveys because the first did not yield enough responses for all opinion leaders.
4. Donald Trump has since been banned from the platform.
5. Including these two accounts in the ideal points distribution did not substantively change our results.
6. We also estimated the ideologies of the ordinary American users as manifested in their following decisions using the ideal point algorithm. In line with our expectations, most ordinary users do not follow enough political elites to estimate an ideal point. If we impute these respondents’ ideal point as 0, the distribution of ideology on Twitter is decidedly unimodal (see, Figure A2 in the Appendix.)
7. In our classification of opinion leaders, each opinion leader may belong to more than one professional category, as described in the Data section. In the Robustness section in the Appendix, we show that the results presented here are very similar, even if we only use the primary category of each opinion leader (Figure A4).
Additional information
Notes on contributors
Subhayan Mukerjee
Subhayan Mukerjee is an Assistant Professor in Computational Communication at the National University of Singapore.
Kokil Jaidka
Kokil Jaidka is an Assistant Professor in Computational Communication at the National University of Singapore.
Yphtach Lelkes
Yphtach Lelkes is an Associate Professor in the Annenberg School for Communication at the University of Pennsylvania.