Abstract
The majority of political communication research either studies a single media outlet in isolation of other outlets or focuses on the competing effects of multiple outlets. This study uses 2004 National Annenberg Election Survey data to go beyond these typical approaches to show the moderation-based complementary effects of two-sided messages (e.g., network TV news) relative to one-sided oriented outlets (e.g., FOX News) on attitudinal ambivalence. In addition, this study places ambivalence within a larger communicative process and shows that ambivalence mediates the relationship between consumption of political media and when citizens decide who to support during elections.
Notes
1For more details on the survey, see Romher, Kenski, Winneg, Adasiewicz, and Jamieson (2006).
2See Webster (Citation2007) and his analysis of Nielsen Media Research data.
3These same reliability and validity tests were run on all individuals who answered the trait-based questions, not just those individuals included in the regression models. Results show similar results with slight variation. There were similar alpha levels for both the Bush positive trait scale (Cronbach's α = .93) and Kerry positive trait scale (Cronbach's α = .93). Similar reliability scores were also found for both the Bush negative trait scale (Cronbach's α = .81) and the Kerry negative trait scale (Cronbach's α = .74). These scales also correlated with each other as expected. There were statistically significant negative relationships between the positive Bush and positive Kerry scales (zero-order r = −.74, p < .001), the positive Bush and negative Bush scales (zero-order r = −.68, p < .001), positive Kerry and negative Kerry scales (zero-order r = −.61, p < .001), and the negative Bush and negative Kerry scales (zero-order r = −.54, p < .001). There were also expected positive relationships between the positive Bush and negative Kerry scales (zero-order r = .60, p < .001) and the negative Bush and positive Kerry scales (zero-order r = .64, p < .001). Similar scores for the reliability and validity tests provide further support that these are reliable and valid scales that can be used create a measure of ambivalence.
4To assess the validity of the ambivalence scale, the measure was correlated with items with which it should have a relationship. First, previous research indicates that ambivalence should be correlated with strength of party identification (Rudolph & Popp, 2001). For this analysis, ambivalence was correlated with strength of party identification (with strong party identification coded as low). Results show a statistically significant relationship between ambivalence and strength of party identification (zero-order r = .29, p < .001, N = 4,534). That is, those with high ambivalence were more likely to have weak party identification. In addition, ambivalence should be negatively related people perceiving that they have enough information about the candidates running for president. Part of being ambivalent is that people feel they do not have enough information to resolve their attitudinal inconsistency. In other words, if a person holds both positive and negative attitudes toward Bush and Kerry, she or he will feel the need to seek out additional information to resolve their attitude inconsistency (Fournier, Citation2005). Results show a statistically significant negative relationship between ambivalence and having enough information about the candidates, (zero-order r = −.39, p < .001, N = 4,506). These correlations provide further evidence that this measure of ambivalence performs as it should.
Note. Table includes unstandardized regression coefficients with standard errors in parentheses.
*p < .05. **p < .01. ***p < .001.
Note. Table includes unstandardized regression coefficients with standard errors in parentheses.
*p < .05. **p < .01. ***p < .001.