No support was found for the hypothesis that social media use contributed to the level of affective polarization; instead, it was the level of affective polarization that affected subsequent use of social media

Affective polarization in the digital age: Testing the direction of the relationship between social media and users’ feelings for out-group parties. Maria Nordbrandt. New Media & Society, September 19, 2021. https://doi.org/10.1177/14614448211044393

Abstract: There is considerable disagreement among scholars as to whether social media fuels polarization in society. However, a few have considered the possibility that polarization may instead affect social media usage. To address this gap, the study uses Dutch panel data to test directionality in the relationship between social media use and affective polarization. No support was found for the hypothesis that social media use contributed to the level of affective polarization. Instead, the results lend support to the hypothesis that it was the level of affective polarization that affected subsequent use of social media. The results furthermore reveal heterogeneous patterns among individuals, depending on their previous level of social media usage, and across different social media platforms. The study gives reason to call into question the predominating assumption in previous research that social media is a major driver of polarization in society.

Keywords: Affective polarization, echo chambers, Facebook, reversed causality, social media, Twitter

Check also other literature with references: Politically partisan left-right online news echo chambers are real, but only a minority of approximately 5% of internet news users inhabit them; the continued popularity of mainstream outlets often preclude the formation of large partisan echo chambers

Conclusion and discussion

The study shows that starting using social media or elevating usage did not impact an individual’s level of affective polarization over time—contrary to H1 and to common assumptions. Instead, the results suggest that affective polarization affects social media usage, in line with H2, depending on the history of previous usage, as suggested in H3. These results should essentially be good news from a democratic point of view and should alleviate the widespread worry that social media is a major driver of polarization in society.

As with any study, some remaining questions and limitations need to be discussed. To start with, as others have previously noted (Prior, 2013), in strict terms, causal inferences require exogenous variation. Nonetheless, there is widespread agreement in the literature that panel data are the best non-experimental data for approaching making causal inferences (Allison, 2005: 1). Moreover, I have taken several important measures to reduce the risk of temporal and non-temporal confounders and the patterns remain the same regardless of model specification. It furthermore deserves to be highlighted that panel data often have the upper hand vis-à-vis experimental treatments in providing a picture of processes as they naturally unfold rather than in a manipulated setting.

Turning to a discussion about the measures of affective polarization and social media use; I only had access to a measure of polarization that tapped into the respondents’ sympathies for political parties, but polarization can take on many other, and perhaps more troubling, expressions. Nevertheless, to the extent that dwindling sympathies for out-group parties appear in concert with increasing inter-group hostility and distrust that follow political lines, it is arguably something we need to be wary of.

Because of data limitations, the scope of the study was furthermore limited to inferences about the average effect of/on usage of various social media platforms. This measure had the important property of providing a picture of whether social media use as such seems to be a driver of affective polarization in society from an aggregate point of view. Nonetheless, qualitative aspects, such as purpose of usage and the content users are exposed to, are likely crucial for a deeper understanding of the mechanisms through which affective polarization and social media are connected.

The study was furthermore limited to the study of one particular country. Recent evidence suggests that The Netherlands exhibits lower levels of affective polarization than most other western countries (Reiljan, 2020) possibly making it a least likely case for detecting a relationship between these. Theorizing from a comparative perspective is beyond the scope of this study, but cross-country differences should certainly be of interest to future research.

The most striking, and complex, finding of the study calls for some reflection; namely that there was heterogeneity in how polarization affected the amount of social media usage. First of all, those who were non-users to moderate users in the previous wave increased their level of Facebook use and decreased their level of Twitter use significantly as they gained more polarized attitudes. One possibility is that individuals who developed an expressive need turned to platforms such as Facebook because they judged their chances of impacting their strong-tie and weak-tie offline relationships to be greater than their chances of impacting their more distant Twitter followers. Others may increasingly have shied away from Twitter because of unease or anger as a result of exposure to polemical Twitter debate and instead resorted to platforms that allow users to take part of a larger share of apolitical content. A third possibility can also be raised. Some strongly polarized and frustrated social media users may have abandoned conventional social media platforms in favor of other more alternative and niched platforms or even abandoned them altogether (Purhonen et al., 2021). In fact, this happened among many Trump supporters following the suspension of the account @realDonaldTrump from Twitter in the aftermath of the Capitol riots on 6 January 2021. All these explanations may have some element of truth to them, but the former should be less troubling than the latter ones. For instance, people can have legitimate reasons for feeling upset. Social media doubtlessly fills an important deliberative function as an arena for people with an expressive interest “to call something to the attention of a wide audience” (Moles, 2007: 54). However, to the extent that social media increasingly becomes a megaphone for aggressive or divisive rhetoric, this is more troublesome. There is a risk that some people with an expressive need engage in a “spiral of silence” and refrain from voicing their concerns because of a worry of being subjected to resentful comments or social exclusion (Kruse et al., 2018; Noelle-Neumann, 1974).

Second, according to the results, rising levels of polarization exerted a much stronger effect on previous non-users or moderate users than more regular users, in line with H3. This gives some substance to the notion that a variety of rationales besides affective polarization drives frequent usage and that everyday users are more resilient to reducing their level of usage, even if they acquire more polarized attitudes.

A final thought on perhaps the most consequential finding of the study vis-à-vis previous research; namely that increased social media use did not seem to amplify affective polarization. It is probably warranted to be skeptical of unlimited usage of social media for several reasons. For instance, it may gradually make us more tolerant of uncivil behavior, invoke unhealthy social comparison, and reduce occasions for in-depth in person contacts. Still, the results give reason to doubt the notion that elevated affective polarization is among the suggested sinister consequences. Insofar as the findings translate to other contexts as well, they suggest that social media may not be a reliable barometer for assessing affective polarization in society and that explanations for any surge of polarization should primarily be found elsewhere. This discursive correction is important because the stronger the impression that we are deeply polarized, the higher the risk that we eventually judge it as pointless to engage in cooperation and dialogue across ideological lines. This would surely be detrimental for society and for democracy. My hope is that these results spur an interest in future evaluations of the causal direction of the relationship to avoid unnecessary consequences of what may potentially be an erroneous perception.