Knowledge Versus Beliefs: How Knowledge and Beliefs Mediate The Influence of Likeminded Media use on Political Polarization and Participation

Abstract

Using cross-sectional data from the 2008 National Annenberg Election Survey, this study tests 2 models that explicate the relationship between politically likeminded media use and political polarization and participation. The knowledge model suggests that the effects of exposure to likeminded media on individuals’ attitudinal polarization and political participation are mediated by knowledge of candidate issue stances. The belief model proposes that likeminded media use indirectly influences political polarization and participation via political beliefs. The results provide evidence that individuals’ beliefs mediate the influence of likeminded media consumption on attitudinal polarization and participation, but there was no support for the knowledge model. These findings indicate that individuals who consume politically likeminded news tend to develop polarized attitudes and are motivated to participate in political activities by forming biased beliefs associated with candidates rather than by gaining factual issue knowledge.

Acknowledgments

The author thanks Dr. Maxwell McCombs, Dr. Natalie Jomini Stroud, and Dr. Homero Gil de Zúñiga for their comments on previous versions of this manuscript.

Notes

1. Although selective exposure technically refers to individuals’ seeking out information that agrees with their perspectives while avoiding challenging viewpoints, this study focuses specifically on the use of likeminded media outlets. Recent literature, for example, argues that exposure to confirmatory information needs to be distinguished from avoidance of attitude-discrepant information because the consequences of these two forms of exposure—for instance, affirming one’s sense of correctness and producing negative emotional reaction, respectively—are not equivalent (see Garrett, Carnahan, & Lynch, 2013). Indeed, many studies have focused on the selection of likeminded information, particularly the consequences of selective exposure on individuals’ attitudes and behaviors, and the impact of cross-cutting exposure has been separately investigated in the literature (Dilliplane, 2011; Johnson, Bichard, & Zhang, 2009; Mutz, 2002a; Price, Cappella, & Nir, 2002). In addition, it is important to note that it is very difficult to know, in survey research, whether media selectivity is conscious (i.e., driven by individuals’ choice or motivation) or is not necessarily motivated by media users per se, but externally imposed (i.e., de facto selective exposure; see, e.g., Garrett, 2009). Despite this distinction, researchers have conceptualized selective exposure without clarifying those different types of selectivity (e.g., Gil De Zúñiga, Correa, & Valenzuela, 2012; Stroud, 2010). Some researchers, however, tend to use the term “likeminded media use” instead of “selective exposure” (Dilliplane, 2011), which may ease concerns about the clear definition of selective exposure regarding the issue of selectivity. Thus, this study will focus on individuals’ likeminded media use throughout to be clear that the focus of the current study is about the consequences of likeminded media consumption rather than on what causes media selectivity.

2. Although this study focuses on attitudinal polarization, it is worth noting that there is a literature in the interpersonal political discussion domain dealing with ambivalence using a similar but not identical measure. Ambivalence is defined as the “individual’s endorsement of competing considerations relevant to evaluating an attitude object” (Lavine, 2001, p. 915). The concept of ambivalence captures individuals’ simultaneously holding both positive and negative attitudes toward a single attitude object including political candidates. Individuals’ ambivalence toward candidates—McCain (M) and Obama (O)—can be measured with the following equation: Ambivalence = (M + O) / 2 - | M - O |. Ambivalence score was measured by taking into account: (1) the average attitudinal intensity of the two ratings, and (2) the disparity between them (see Lavine, 2001; Nir, 2005). With the rating scores ranging from 0 to 10, with 0 indicating the least favorable rating and 10 the most favorable, the theoretical range of the ambivalence score is from -5 (= least ambivalence) to 10 (= most ambivalence). For example, a respondent who rated McCain = 0 and Obama = 10 shows the lowest ambivalence score of -5. As such, the measures of polarization and ambivalence are not identical, but very similar. To see how polarization and ambivalence are related, ambivalence was additionally calculated using the above equation. It turned out that polarization and ambivalence are highly correlated (r = -.88, p < .001), suggesting that ambivalence can be considered just a different operationalization of polarization. Indeed, analysis of ambivalence in the models showed very similar results to models with polarization.

3. Literature suggests that general political knowledge should also be controlled for because it is associated with individuals’ learning during a campaign (Brians & Wattenberg, 1996; Price & Zaller, 1993). One may wonder if general political knowledge and factual issue knowledge would be highly correlated and thus influence the relationships found. It turned out that they are moderately correlated (r = .42, p < .01), but dropping general political knowledge did not change the results. In any case, general political knowledge was included as a control variable in both the knowledge model and the belief model to have the same sets of control variables across both models.

4. Given the nature of this study, which focuses on testing the potential mediating role of political knowledge in the relationships between likeminded media use and political polarization and participation, the influence of control variables is not discussed in the text. The effects of demographic variables (gender, age, education, income, and race), political orientation (political efficacy, campaign interest, and general political knowledge), general media consumption (TV, radio, newspapers, and online), the time of interview, and dissonant media use were included in the model in which such variables, as control variables, predict all outcome variables. Furthermore, the path model for knowledge does not provide goodness of model fit since it is a saturated model, which leaves no degrees of freedom for testing the goodness of fit of the model. This saturated model, however, is appropriate for the goal of this study, which is to investigate the potential mediating role of issue knowledge in the relationships between likeminded media use and political polarization and participation, rather than building a model that is as parsimonious as possible and satisfactorily fits a set of data.

Additional information

Notes on contributors

Yonghwan Kim

Yonghwan Kim (Ph.D., University of Texas at Austin) is an assistant professor in the Department of Media and Communication at Dongguk University, Seoul, Republic of Korea. His research interests include emerging information communication technologies such as digital/social media, media convergence, journalism, and media effects.