![Sample our Politics & International Relations journals, sign in here to start your FREE access for 14 days]({"type":"image" , "src" : "/sda/5947/TOC-Subject-Sample-2021-Politics-International-Relations.jpg"})
Abstract
To what extent are groups selectively exposed to news that affects their self-interest? We theorize that having an interest at stake in an issue increases the importance of related information, promoting selective exposure to politics. The framework we develop generates hypotheses we test with data from 29 U.S. public opinion surveys conducted between 1997 and 2007 with a combined sample size of over 40,000 respondents. We find that though the propensity to follow news is a general predisposition, people's exposure to specific issues varies based on whether or not information is relevant to their interests, as defined by social group (or issue public) membership. While general awareness of political matters is normatively desirable, exposure to issues relevant to one's self-interest is an equally important and necessary precondition for political engagement and issue public formation. Evidence from this study suggests that citizens with low levels of general news exposure may, nonetheless, attend to personally relevant information, enabling democratic accountability across dynamic information environments.
[Supplementary material is available for this article. Go to the publisher's online edition of Political Communication for the following free supplemental resource(s): iPoll survey identifiers for the data used in this article and full regression results for all statistical models reported in the article.]
Notes
1. We use the term “selective exposure” broadly to refer to a preference for any particular sort of information (see, e.g., CitationHolbrook et al., 2005; CitationIyengar et al., 2008).
2. As we discuss at length below, we equate group identity with an individual's self-interest. This draws not only from a large literature in social psychology on the antecedents of attitude importance (CitationBoninger, Krosnick, & Berent, 1995), but also from literature on “linked fate” in political science (CitationDawson, 1994; CitationGay & Tate, 1998; CitationHerring, Jankowski, & Brown, 1999; CitationSimien, 2005; CitationTate, 1994). CitationSimien (2005, p. 529) defines linked fate as “an acute sense of awareness (or recognition) that what happens to the group will also affect the individual member.” For instance, among African Americans, perceptions of linked fate arise from shared experiences of oppression, and this “stage of identification, where individuals come to see themselves as sharing a linked fate with other African Americans, leads to collective action as a necessary form of resistance” (CitationSimien, 2005, p. 530; CitationDawson, 1994; CitationTate, 1994).
3. Individual differences in awareness affect the connection between self-interest and policy choices because exposure to cueing communications alters the priority given to self-interest in decision making (CitationChong, Citrin, & Conley, 2001; CitationMiller, 1999).
4. An additional, growing literature examines partisan selective exposure, which attempts to link party identifiers or ideologues to particular news media or to particular news stories (see, for example, CitationDilliplane, 2011; CitationFeldman, 2011; CitationGarrett, 2009a, Citation2009b; CitationIyengar & Hahn, 2009; CitationStroud, 2007, Citation2008, Citation2010). This literature is obviously relevant, but we focus here on studies about non-partisan, non-ideological forms of self-interest. We also focus on exposure or attention to group-relevant stories, not specifically on stories that are congenial or uncongenial to one's particular group, party affiliation, or predispositions.
5. Note that “researchers typically make inferences about respondents’ self-interest by making their own analysis of the consequences of [a] policy [or issue] for different [demographic] groups in society” (CitationChong et al., 2001, p. 542). Objective measures of self-interest have been shown to differ significantly from subjective assessments and to be more closely connected to an individual's opinions (CitationChong et al., 2001, p. 545). Differentiating issue public members from nonmembers can be “a controversial decision because it can be done in a number of different ways” (CitationKrosnick & Telhami, 1995, p. 538). One limitation of our approach is that it may misidentify or fail to fully identify all issue public members. Another method of identifying issue publics is to ask survey respondents directly how important an issue is to them personally (CitationAnand & Krosnick, 2003, p. 37). Nonetheless, our approach is widely used throughout this literature.
6. Self-relevant information also leads to higher levels of encoding and recall (CitationDutta & Kanungo, 1975; CitationKuiper & Rogers, 1979; CitationRogers, Kuiper, & Kirker, 1977).
7. Knowledge accumulation is the result of selective exposure to self-relevant information and selective elaboration of related information. The implication of this is that “most citizens are likely to be knowledgeable about the few issues that they care deeply about and are therefore equipped to be ‘responsible voters’ if and only if they are first attentive to those issues” (CitationHolbrook et al., 2005, p. 767).
8. This battery of questions was typically at the beginning of each survey, with some surveys including the battery as the first question and others as late as the 27th question. The median starting point for the battery within each survey instrument was as the third question.
9. We also found stories related to other groups (Catholics, Latinos, union members, etc.) but excluded them prior to the analysis due to the limited number of stories our theory would expect these groups to attend to. Our mapping of the three main groups in our analysis is consistent with the “group interest” literature, which maintains that “category membership and identification with a group and a sense of shared fate lead to group-based assessments of self-interest” (CitationBobo & Kluegel, 1993, p. 445; see also CitationBobo, 1983, 1988; CitationBobo & Hutchings, 1996; CitationJackman & Muha, 1984; CitationMutz & Mondak, 1997).
10. When two or more group-relevant stories were found in a single survey for the same group (e.g., women), each was excluded from the global follows calculation. When group-relevant stories were found for multiple groups (e.g., women and older Americans) in the same survey, separate calculations of global follows were made for each group (excluding only the story relevant to their group).
11. The number of observed levels (and cutpoints on the latent exposure variable) for the global and specific measures is identical. Averaging responses to all questions maintains the observed range of this outcome and its comparability to the specific measure, though our results are generally robust to alternative coding of the outcome measures.
12. The choice of a dummy variable in these models aids substantive comparison across the three identified groups (females, Blacks, and older Americans). While responses to the age question are continuous, we use age as an indicator, with those 60 and older set to 1 and those 59 and under set to 0. Models that include alternative specifications of age (including a binary split at 65 or a continuous predictor) provide substantively similar results, and models containing a binary split at 60 or 65 produce statistically similar results. The choice of 60 attempts to include those at (early) retirement age in addition to nearly retirement age.
13. Education and income are both ordinal variables ranging from 1 to 8. Republican and Democrat are dummy variables. This specification reflects the nonlinear relationship between partisanship and news attention (i.e., Democrats and Republicans are generally more attentive than nonpartisans, but neither group is more attentive than the other). On a subset of surveys where a follow-up “leans”-type party identification question was asked, those who leaned toward one party or the other were coded .5 in the respective variables.
14. Modeling with OLS or using the full interval scale of the global measure yields substantively and statistically comparable results. In the analysis presented here, we use the four-category global measure to preserve comparability with the specific measure and to ease presentation of results.
15. It is also possible that respondents reacted to race, age, or sex-related cues in the questions, leading them to provide socially desirable responses (i.e., reporting more attention to stories that they “should” be attentive to). We thank an anonymous reviewer for this observation.
16. Due to space constraints and for clarity of presentation, we do not report the full results for all global and specific models. Also, note that we include partisanship variables in all models. Democrats are typically slightly more attentive than Republicans and Independents, though the latter groups do not generally differ from one another and Republicans are sometimes more attentive than Independents (one or both partisanship variables were significant for 34 general attention models). Democrats are, similarly, more attentive to 27 specific stories than Independents, as are Republicans in 10 instances.
17. The results of the analysis of older Americans’ attention to news in the 2006 HNI060 survey are potential outliers due to the large oversampling of adults age 65 and older. The core sample size is 1,217, with an oversample of 1,300. Excluding results from this survey, the average values of coefficients and their difference are similar: β global = 0.15, β specific = 0.49, and β specific – β global = 0.34.
18. Given our directional hypotheses, our statistical significance tests are one-tailed t tests for differences between the coefficients for each group's indicator in the specific and general models, under the assumption that general follows and specific follows are uncorrelated. These two measures are sometimes positively correlated, however, although these correlations never exceed .50 when present.
19. In fact, for only four stories in the entire sample is specific attention lower than global attention on stories regarding RU486 (female, 2000), Social Security (age, 2000), and Medicare (age, 2003 and 2004). In each of these cases, however, the specific coefficient is not significantly different from the global coefficient.
20. These stores are the debate over Ebonics (1997), news that African Americans are disproportionately more likely to suffer heart failure (2004), Rosa Parks's death (2005), and news of Hurricane Katrina recovery efforts (2006).
21. While this is the smallest difference between general and specific attention, it reflects statistically significant differences in most cases. For 11 of 18 surveys, this difference is statistically significant at p < .01, and in one additional survey the difference is marginally significant at p < .10. In three of the remaining surveys, attention to stories we identified as group-relevant was lower than general attention: one about Social Security (2000) and two about Medicare (2003 and 2004).
22. These results come from an analysis of iPOLL data sets USPEW2002-06NII for Catholics (question 5g) and USPSRA2006-HNI059 for Hispanics (question 2d). The regression models are identical to the other analyses but include additional indicator variables for these two groups.