Echo chambers, filter bubbles? The degree of heterogeneity of one’s political news consumption, as well as voting intentions for a specific party are complex psychological constructs with many different variables, each single variable only having a small effect

The degree of heterogeneity of news consumption in Germany—Descriptive statistics and relations with individual differences in personality, ideological attitudes, and voting intentions. Cornelia Sindermann, Christopher Kannen, Christian Montag. New Media & Society, December 30, 2021. https://doi.org/10.1177/14614448211061729

Abstract: This study aimed to examine the degree of homogeneity versus heterogeneity of individuals’ political information environments across offline and online media types and relations with sociodemographic variables, personality, and political attitudes. In two online surveys, German participants (sample 1: N = 686; sample 2: N = 702) provided information on sociodemographic variables, consumption of political news, and voting intentions, and completed the Big Five Inventory and Right-Wing Authoritarianism (RWA) and Social Dominance Orientation (SDO) scales. Results revealed that absolutely homogeneous political news consumption was evident for a small proportion of individuals (2.04% and 0.43%). Openness (positively) and Agreeableness (negatively) exhibited significant associations with the degree of heterogeneity of political information environments across samples. No consistent patterns of relations with either the ideological attitudes of RWA and SDO or voting intentions were observed. The findings shed light on the existence of absolutely homogeneous political information environments and “who” might be prone to a more homogeneous versus more heterogeneous information environment.

Keywords: Echo chamber, filter bubble, homogeneous information environment, ideological attitudes, news consumption, personality, political news, voting

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

Discussion

The current study sought to contribute empirical data to the ongoing discussions around homogeneous political information environments. Three specific aims were addressed in this study: (1) An examination of the degree of heterogeneity of information environments of individuals in relation to political news, (2) analyses of the extent to which the degree of heterogeneity of an individual’s political information environment is associated with individual differences in sociodemographics and personality, and (3) an exploration of the associations of the degree of heterogeneity of individuals’ political information environments with ideological attitudes and political voting intentions. Analyses focused on political news consumption across various offline and online media types and were implemented in two independent population-based samples to derive robust, replicable, and reliable findings.

Across the two independent samples, the media type where individuals received on average the most homogeneous and attitude-consistent political news was podcasts; see Figure 2. It should be highlighted, however, that relatively few individuals in sample 1 (8.02%; but sample 2: 37.61%) actually used podcasts to consume political news. The generally low scores for consuming counter-attitudinal news via podcasts in both samples might be due to the time which needs to be invested into listening to a podcast. Given this, individuals might need to be more selective with regard to which podcast they listen to. It is also possible that individuals tend to choose to listen to podcasts delivered by specific individuals who then advertise podcasts by other individuals with similar attitudes, further reducing the heterogeneity of news consumption of listeners. The media types where individuals on average received the most heterogeneous political news were social media (sample 1) and online news websites (sample 2); see Figure 2. Thus, these results do not support the notion that social media platforms pose the greatest risk for homogeneous information environments (see arguments highlighted in the Introduction of the present study). In addition, the results do not show that political news consumed via online media types, where algorithmic filtering is possible, are generally more homogeneous than those consumed via offline media types. Unfortunately, we cannot draw any conclusion as to why heterogeneity scores differ between media types. Heterogeneity could be due to incidental, algorithmically based, or self-initiated confrontation with counter-attitudinal news. Investigating the reasons for differences in heterogeneity scores across media types will be an important research approach for forthcoming studies. Relatedly, mechanisms to increase the heterogeneity of news consumed within and across media types will need to be investigated. Based upon these investigations, citizens should be informed about the degree of heterogeneity of news presented via different media types and ways to increase heterogeneity. Further development of add-ons (e.g. for internet browsers) informing users about their personal heterogeneity of news consumption could be the first step in this direction—an add-on is, however, always limited to the online context (see examples mentioned in Bozdag and van den Hoven (2015)).

Across all media types used, only a few individuals responded “never” to seeing news contradicting their existing opinions when consuming political news (n = 14, 2.04% in sample 1; n = 3, 0.43% in sample 2). These results indicate that few individuals are exposed to absolutely homogeneous information environments across media types. This is in line with previous research on incidental and counter-attitudinal news exposure on different media types (Newman et al., 2017; Vaccari et al., 2016).

Older age and being male (versus female) were associated with greater heterogeneity of one’s political information environment across samples. These results fit with findings from previous studies (Benesch, 2012; Sindermann et al., 2020) although, in one of the cited studies significant associations with education were also found, which was only observed in one of our samples (sample 1).

With respect to individual differences in personality the following results were obtained: Significant positive associations of Openness (in correlational and regression analyses) and negative associations of Agreeableness (in regression analyses) with the degree of heterogeneity of political information environments were found across samples.

We want to specifically highlight the positive associations between Openness and the degree of heterogeneity of political information environments. These positive associations are in line with our initial hypothesis and the findings of previous studies (Matz, 2021; Sindermann et al., 2020). Moreover, it supports the intellectual stimulation hypothesis proposed by Kim and Kim (2018). Individuals scoring high in Openness seem to enjoy engagement in news reports dealing with a range of different political opinions. This is also in line with one study where a positive association between Openness and engagement in political discussions was reported (Mondak and Halperin, 2008); but we should also acknowledge the results of other studies that do not support this link (Gerber et al., 2012; Hibbing et al., 2011). As can be seen in Supplementary Table 5, Openness might be more strongly related to the degree of heterogeneity of one’s online versus offline political information environment (see results from sample 1 reported in the Supplementary Material). A greater potential for the individual to create his or her online news environment versus the offline news environment might contribute to these differential associations (see Supplementary Material).

Contrary to our hypotheses, we did not find any significant association of ideological attitudes with the degree of heterogeneity of one’s political information environment across samples. Despite RWA and SDO having been found to positively relate to closed-mindedness (Berggren et al., 2019), they do not seem to be related to the degree of heterogeneity of one’s political information environment. Interestingly, however, additional analyses (Supplementary Material) revealed that RWA was negatively related to the heterogeneity of one’s online, but not offline, political information environment across both samples. On the one hand, it is possible that high-choice online environments in particular offer the possibility for homogeneous political news consumption for individuals high in RWA. On the other hand, more information filtering online (e.g. via algorithms) compared to offline sources might influence RWA. These results and the causal direction of associations will need to be (re-)investigated in future studies.

It is difficult to draw any firm conclusions regarding potential associations between the degree of heterogeneity of one’s political information environment and voting intentions. Lower heterogeneity was not associated with intentions of not voting, or to vote for a more extreme right-wing party, across samples. Models derived from linear discriminant analysis on associations of sociodemographic variables and the degree of heterogeneity of one’s political information environment with voting intentions revealed quite low classification accuracies. Therefore, we should not overinterpret these particular findings. Future studies might want to examine the associations between the degree of heterogeneity of political information environments and political attitudes in German samples in a slightly different way. For example, one could investigate the strength of party support, topic polarization, affective/partisan polarization via the “feeling thermometer,” or overall left–right ideological self-placement (e.g. Heatherly et al., 2017; Lee et al., 2014). Based on this idea, on an exploratory basis, we calculated the correlation between a 10-point left–right ideological self-placement item and the HoHe score in sample 2 (the left–right self-placement was only assessed in sample 2). The association was nonsignificant (ρ = –.02, p = .660). However, the association between an extremity score (absolute difference between self-placement on the left-right dimension and the mean scores 5 and 6) and the HoHe score was small but significant (ρ = .08, p = .037; these analyses were not preregistered). This indicates that these associations should be investigated in more depth in further studies to understand these complex relations.

Some limitations of the present study must be acknowledged. First of all, neither of the two samples is completely representative of the general German population. Moreover, it is important that the generalizability of the findings will be tested in future studies in other countries. Nevertheless, we are confident that the results that were replicated across the two samples of the present study will also be found in other samples. Next, it should be emphasized that data collection for both surveys was conducted in times of an exceptional crisis, namely, during the COVID-19 pandemic. Therefore, many news stories across different platforms are likely to have included information about the virus and legal/policy regulations to combat it (e.g. lockdowns, stay-at-home orders, etc.; GöfaK Medienforschung, 2020, as cited in Statista, 2021; tagesschau.de, 2020). The focus on these topics across news platforms might adversely affect the generalizability of findings and might explain findings that were inconsistent with our hypotheses. For instance, according to a recent survey, voters of the AfD were mostly against the measures to fight the COVID-19 pandemic (Forschungsgruppe Wahlen, 2020), which might (in part) explain high HoHe scores in this group. The present study should therefore be replicated when topics related to the pandemic are not dominating the news. Another limitation is that the present surveys were cross-sectional. Any conclusions about causality are limited, accordingly. One might, however, cautiously infer causal influences on the degree of heterogeneity of one’s political information environment specifically for associations with sociodemographic variables and personality traits, which are generally seen as relatively stable (Costa and McCrae, 1992a; Edmonds et al., 2008; Roberts and Mroczek, 2008). Another limitation is the application of self-report measures. Assessment of personality, ideological attitudes, and voting intentions via self-reports might be influenced by response biases or lack of ability for introspection (McDonald, 2008). Also, the statements on how many news sources one consumed within the past six months prior to participation, and on how often one was confronted with counter-attitudinal news, might be (un-)consciously biased. Nevertheless, the application of self-report measures allowed us to assess the degree of heterogeneity across various offline and online media types, which is very difficult to assess via objective methods. The HoHe score and its constituent items might be improved in future work, for example, by adding variables on the frequency of use, or the time spent on each media type. Another potential limitation is that across all results effect sizes were mostly small to medium, according to Cohen’s (1992) rules of thumb. This indicates that the degree of heterogeneity of one’s political news consumption, as well as voting intentions for a specific party, are complex psychological constructs and many different variables and their interactions contribute to explaining variance in them, with each single variable only having a small effect (Götz et al., 2021).