Local Social Media Networks and the Emergence of COVID-19 Polarization

• In Event: COVID in the United States

Abstract

AIM: Political polarization of COVID-19-related beliefs, attitudes and behaviors has been extensively documented in Democratic- and Republican-predominant areas in the U.S. during the COVID-19 pandemic. To determine whether COVID-19 communication among the public is correlated with, and may drive, these polarization patterns, this study surveyed COVID-19 communication on Twitter by analyzing over three million geo-located and randomly sampled tweets, which were posted by members of the general public residing in 1,248 U.S. cities from January 2020 to October 2022.
METHODS: Using the American Ideological Project (AIP), we retrieved political ideology (mrsp) scores for 1,248 U.S. cities to which three million randomly collected tweets were geo-located. Each city’s ideology score had been previously derived by AIP, based on citizens’ voting records. City socioeconomic data was obtained from the U.S. Bureau of Economic Analysis. Data from the American Communities Project was used to define 15 major community types. To automatically classify the 14 major topics of COVID-19 communication, we filtered the tweets using a validated set of filters that were previously developed. Linear regression and negative binomial analyses were deployed to model the relationship between a city’s political ideology and Covid-19 communication patterns, and to determine which topics best predicted the engagement of audiences in U.S. cities.
RESULTS: We found very distinct patterns of Covid-19 communications among U.S. cities that were strongly correlated with the political ideology of cities. Thus, Twitter users from liberal cities tweeted more about race and ethnicity, and less about masking, religion, and law enforcement when communicating about Covid-19. By contrast, users from conservative cities tweeted more about religious issues and used significantly fewer foul words in their tweets. The tweets of users from liberal cities were significantly more positive and upbeat than those of conservatives and independents.
We next focused specifically on the communication about masking and vaccination – the major preventative approaches during the pandemic. We found that, starting in June 2020, users from liberal cities, unlike users in conservative and politically centric cities, tweeted significantly less about masking. Furthermore, users from liberal cities tweeted significantly more about vaccination, especially in spring 2021 when vaccination was widely introduced in American cities. Thus, the liberal ideology of a city predicted less tweeting about masking and more tweeting about vaccination.
We then asked whether people residing in different communities talked differently about Covid-19 throughout the pandemic. Users from college towns tweeted more about education, while members of African American and LDS (predominately white and religious) communities tweeted more about religion. Users from Hispanic centers employed more foul language in their tweets. Finally, users from the conservative “graying America” communities tweeted more about politics, extreme points of view and COVID-19-related conspiracy theories.
We next asked what Covid-19 topics were most engaging during the pandemic, and whether local political ideology predicted local audience engagement. We found that local audience reactions to Covid-19 communication by like-minded peers were strongly segregated along ideological lines. Followers of users in liberal and centric cities were very engaged by tweets about masking. Race and ethnicity were highly engaging for audiences in politically centric and liberal, but not conservative, cities. Political aspects of COVID-19 were very engaging for users in liberal cities but were de-engaging in politically centric and conservative cities. Tweets about education were engaging only for users in centric and conservative, but not liberal, cities. Finally, audiences across all cities, irrespective of political ideology, were de-engaged by anti-masking messages and tweets containing extremism and conspiratorial language.
CONCLUSIONS: These findings uncover robust geographical and temporal patterns of COVID-19 communication in the United States, which are driven by political ideology as well as socio-economic factors. We uncovered a polarized Covid-19 communication pattern across America along ideological lines. The results raise the possibility that local social media networks, which are assembled along ideological, geographical, and cultural lines, may restrict Covid-19 communication – making it more likely for people to encounter pro-attitudinal Covid-19 communication, which then facilitates the emergence, spread and persistence of Covid-19 political polarization. Finally, our findings can serve as a guide for devising effective health communication strategies and public health policies that could be employed to de-polarize the public and manage communication during future public health crises.

Authors

• Liviu Aron, Harvard Medical School

• Wenyou Ye, University of Maryland