Prevalence, Dynamics, and Effects of Behavioral Polarization on Facebook

• In Event: Political Parties and American Democracy Mini-Conference III: How the Public Used Facebook and Instagram in the 2020 U.S. Election

Abstract

Both the public and researchers worry about the social consequences of affective polarization, specifically that Americans are unwilling to interact with people who do not share their partisan or ideological identities. This is a key issue surrounding the potential retrenchment of American democracy. Specifically, many worry that social media facilitates polarization by allowing people to curate social networks that are politically congenial. This motivates our main research question: How are political leanings associated with people’s propensities to create and sever ties on social media, and what moderates these relationships?

In this study, we analyze two unique datasets -- aggregate data on the behavior of over 150 million Facebook users and a targeted survey of over 63,500 of these users collected during the 2020 U.S. presidential campaign and its aftermath---to assess a behavioral manifestation of affective polarization: political (un)friending on social media platforms. We assess the dynamics of friendship creation and destruction over a tumultuous period in American history, which included the Covid-19 pandemic, the 2020 election, and the January 6th insurrection.

We make several novel contributions stemming from the unique availability of data on user behavior on Facebook. First, we ascertain the prevalence of political (un)friending during the 2020 U.S. general election and its aftermath and offer the first ever evidence of the differential preference for congenial friendships and aversion to cross-cutting friendships. To better contextualize this prevalence, we map the fluctuations in political (un)friending and its two components over the course of the 2020 presidential campaign and we additionally benchmark political (un)friending around Election Day and the January 6th Capitol storming against users' unfriending patterns during New Year’s Day, 2020-2021. This allows us to assess how people's behavioral response to changes and conflicts in the political environment compares to other events that may prompt reconsideration of the composition of one's social networks.

Second, we analyze the prevalence of political (un)friending—”in-group affinity” and “out-group aversion”---among various sub-groups of Facebook users who, theoretically, should be more or less likely to be affectively polarized (e.g., liberals vs. conservatives; those with strong versus weak partisan identities; those with high vs. low on platform political engagement and exposure to political content). Examining the variation in network curation patterns among these different social groups can shed light on which kinds of people are most likely to inject political considerations into their social lives.

Third, in conjunction with one another, these unique datasets---on-platform behavioral data and survey self-reports from the same individuals--- make it possible to address an important
methodological problem that has complicated the study of both affective polarization and political (un)friending as its manifestation and social consequence. We compare the rates of political (un)friending in behavioral data to self-reported feeling thermometers of partisans and also to survey self-reports of political (un)friending.

Authors

• Fang Guo

• Neil Malhotra, Stanford University

• Jaime E. Settle, College of William & Mary

• Magdalena Wojcieszak, UC Davis, U of Amsterdam

• Annie Franco, Facebook