Following the News? Reception of Uncivil Partisan Media and the Use of Incivility in Political Expression

AbstractPartisan, pundit-based media gets blamed for making political discourse more uncivil, and studies on incivility in mediated discourse have found that uncivil political media can induce negative reactions in audiences. However, how use of uncivil media affects the way individuals express their political views has yet to receive substantial scholarly attention. I hypothesize that tuning in to uncivil political media leads to an increased propensity to use incivility in textual political expression. I develop an index to identify incivility in political expressions, and test my hypothesis using panel data analysis and an open-ended survey item in the 2008 National Annenberg Election Survey. I find that, consistent with my hypothesis, use of uncivil media—specifically pundit cable news and political talk radio—leads to an increased use of incivility when expressing text-based political opinions. Furthermore, this only occurs with reception of like-minded uncivil political media. I note the implications this has for online political discourse and effective deliberation.Keywords: incivilitypartisan mediapolitical discourseonline discoursesocial media Notes1. Scholars have found that the transparency of debate among elites affects policy decision-making; because representatives feel pressure to placate their partisan base when speaking in public, openness in elite deliberation has been shown to increase partisan polarization (Stasavage, Citation2007).2. Whether uncivil political programs can generate negative emotions might also depend on whether the program consistently takes a negative tone. The inclusion of lighthearted segments and jocularity, as one might expect to see on “soft news” programming, might neutralize the effect of incivility by inducing positive emotions.3. While studies have shown selective use of partisan media to be limited (Garrett et al., Citation2013; Gentzkow Prior, Citation2013), differences in hostility have not been considered. That is, certain media might be equally partisan but not equally provocative. Partisan audiences may watch more traditional newscasts from both conservative and liberal viewpoints, but selectively avoid strident shows professing disagreeable views.4. If selective exposure theories adequately explain patterns in media use, then it is unlikely that someone who identifies as a liberal will choose to watch Bill O’Reilly, unless they do not become upset by O’Reilly when watching him—in which case they would not feel the need to retaliate.5. The NAES online panel started in October 2007 with a random probability sample of about 30,000 people. Respondents were reinterviewed another four times throughout 2008, during the primary, election, and post-election periods. Retention rates between the waves averaged about 82%.6. The exact wordings of the questions were: “Is there anything in particular that you like or dislike about Barack Obama/John McCain? If so, please fill in the boxes below.” All respondents who gave a thermometer ranking score for the candidates (on a scale of 0–100) in response to a previous question were asked this question.7. The 2008 NAES was designed to avoid the problems with self-reports that Prior (Citation2007, Citation2013) presents. The analysis of self-report measures in the 2008 NAES by Dilliplane et al. (Citation2013), which finds the measures to be reliable predictors of increases in political knowledge, should provide some confidence that overreporting is not a big factor. Dilliplane et al. (Citation2013) also compare the most popular programs according to 2008 NAES self-report measures to the most popular as indicated by Nielson “people meter” data, finding the two to be strongly correlated—however, they caution that researchers’ ignorance of Nielson’s methods makes it difficult to verify self-reports with their people meter data. Generally, that self-reports of media use may change due to factors such as interest in the election, but actual viewership does not, is always a possibility. Should this be the case, connections between reported media use and use of incivility will be spurious.8. As the subsample was random, the panel is unbalanced; as discussed below, this is not problematic.9. The research assistants were told that the goal was to have a conservative estimate and were thus instructed to give the benefit of the doubt to civility—comments they felt were on the borderline should not be marked as uncivil. To ensure that there were three independent analyses of the responses, the research assistants were instructed not to discuss specific responses or types of responses among each other or with me during the coding process.10. See online Appendix 2 for a theoretical discussion as to why satirical news should be separate from other partisan media.11. Specifically, listening to the program All Things Considered. See online Appendix 2 for more information.12. Breaking these media categories into smaller groups has little effect on the size and significance of coefficients. Therefore, I use these larger groupings for the sake of parsimony.13. Because party identification was measured over time, these measures can take into account how individuals’ attachment to a party may have shifted over the course of the campaign and panel study.14. Individuals who consistently used uncivil media had the highest rate of incivility use, and those consistently unexposed had the lowest incivility use, with “changers” falling in the middle.15. For the wave 2 period, a familiarity factor with the candidates and campaign may influence the propensity to offer uncivil comments. A Pew Research Center (2007) study in the fall of 2007, taken 2 months before wave 2 began, found that only 62% could name Barack Obama as a Democratic presidential candidate, and a mere 24% named John McCain as a Republican candidate. The visibility of both candidates most certainly increased thereon.16. Additionally, including these variables interacted with the waves has little impact on the size and direction of the other coefficients in the model. None of the demographic interactions were close to significance.17. Dummy variables for the two waves allow me to gauge the impact of the particular time periods. Wave 4 is left as the baseline wave to avoid collinearity. Including dummies for any two of the waves has no impact on the model.18. The political interest measure was part of a public affairs profile in which nearly all NAES participants partook, usually before their wave 1 interview. It ranges from 0 (no interest) to 3 (very interested). Essentially, it gauges how interest, uninfluenced by the campaign, affected the propensity to use incivility as election season wore on.19. I calculate the predicted probabilities using the observed value approach (Hanmer & Kalkan, Citation2013), which involves holding each of the other independent variables at the observed values for each case in the sample, calculating the relevant predicted probabilities for each case, and then averaging over all of the cases.20. The descriptive statistics included in Table A7 in online Appendix 5 show that uncivil political media users tend to be more partisan and ideological than non-users, and are overall more politically active.21. Note, however, that the election does not seem to have influenced viewership of uncivil media. Virtually the same percentage of Republicans (just under 60%) and Democrats (just under 40%) tuned in to uncivil political media in each wave. This is not to say that feelings about the candidates, the election, and politics did not change with the events of the election—but the election did not appear to systematically affect mass viewing habits.Additional informationNotes on contributorsBryan T. GervaisBryan T. Gervais is Assistant Professor, Political Science and Geography, University of Texas at San Antonio.