Information Flow and Attitude Change: Structural and Individual Perspectives

Abstract

Attitude change is a necessary phenomenon in democratic societies; without it, there is not a functioning public sphere. This thesis explores plausible causes of opinion formation under existing theories, with a focus on information flow at the individ- ual and aggregate level. In The Nature and Origins of Mass Opinion, Zaller (1992) hypothesizes that individual propensity for attitude change is highest for cognitive moderates, or those who consume a medium amount of political information. This relationship is consistent with other literature on political cognition, but it is not clear that this relationship holds for all issues all the time; in particular, the degree to which attitudes about low-salience issues are stable and/or non-random among low awareness consumers is unclear from existing theory. Using longitudinal survey data from Pew Research Center from 2015 through 2018, I test three predictive models—– underlying preferences, elite cues, and partisan resistance—–of respondents’ proba- bilities of changing their mind on questions about international openness. I find that party and predispositions do a large amount of work in explaining variance in attitude change, and that political awareness has a mixed relationship with it. There is modest evidence that attitudes about low-salience issues are somewhat random among low awareness respondents, leading them to exhibit more variation in attitudes on surveys than they may truly possess. In contrast, attitudes on high-salience issues such as President Trump exhibit the non-monotonic pattern predicted by existing literature. I investigate this same relationship, between information flow and changes in attitudes, at the aggregate level, to understand the degree of influence the media have in moving public opinion. With data from FiveThirtyEight, a poll aggregator, and headlines from Fox News and The New York Times, I employ Granger causality models of order one, testing President Trump’s approval rating against both intensity and negativity of coverage of the President. I fail to find significant associations at the p = 0.1 level in both directions, suggesting that media coverage is not strongly linked to Presidential approval. Both the individual- and media-level data suggest that Americans generally take cues from elites on questions of low salience, and that an individual’s propensity to consume political information accounts for more variation in attitudes than do the media’s choices to put it out. In an era of hyper-selective consumers and increasing segregation of political media, these findings shed further light on the need for so- called “cognition moderates” in democratic politics.