Does News Exposure Increase Perceived Polarization? Evidence from Trace Data

• In Event: Media and Polarization

Abstract

There is a widespread agreement that Americans have become more divided across party lines and report increasingly negative attitudes toward their political opponents. A related, but less studied trend is that citizens perceive society and politics as increasingly polarized and – in fact – this perceived polarization has grown at higher rates than actual polarization. These perceptions can have tangible influence on various political outcomes such as turnout or evaluation of public institutions. Despite the growth in perceived polarization and its consequences, we know surprisingly little about its predictors. This project focuses on media exposure as the key reason for why citizens see the polity as sharply divided. After all, media are the only avenue from which most citizens learn about elite behavior and American news media increasingly focus on political conflict. Some experimental evidence suggests that exposure to media coverage of polarization makes people perceive the country as more polarized. Our pre-registered project extends the limited past evidence in several key ways. First, unlike experimental designs, which expose people to content they may not consume in real life, we measure actual exposure to news in naturalistic settings over time. We rely on web browsing history records from large samples in the US, Poland, and the Netherlands paired with a three-wave panel survey on participants attitudes and perceptions. This unique data set that allows us to examine over time changes in perceived polarization due to actual online news exposure in between the waves. Second, we disentangle two kinds of content that might have distinct effects on perceived polarization: (a) the coverage of polarization (i.e., the fact that politics or society is polarized) and (b) polarizing content itself (e.g., using extreme or derogatory language when discussing politics). For each of the three languages, we develop machine-learning classifiers to measure exposure to coverage of polarization as well as to polarizing content on the news sites people consumed. We use cross-lagged panel models to estimate the causal effect of exposure to these types of polarization content on perceived polarization over time and also account for the fact that the coverage of polarization might have different effects depending on whether it comes from partisan or centrist news sites, as determined on validated classification of news websites’ ideological leaning. As a complementary piece of evidence, we experimentally test whether a reminder that the media talk a lot about political conflict increases perceived polarization. Lastly, given the likely negative consequences of perceived polarization, we explore potential mitigation strategies in two related survey experiments: We test whether priming people with a norm of good citizenship, and with a norm of viewpoint diversity, affect perceptions of polarization. The data are being analyzed.

Authors

• Magdalena Wojcieszak, UC Davis, U of Amsterdam

• Bernhard Clemm von Hohenberg