Is News to Blame? A Causal Test of (Un)Desirable News Exposure Effects

• In Event: News Engagement

Abstract

Democracies function best when citizens are up-to-date on political events and knowledgeable about the issues of the day. For this reason, news exposure is seen as normatively desirable. In this study, we problematize this normative view. We argue that tuning in to news can generate a wide range of adverse outcomes, such as attitude and affective polarization, negative perceptions of the political system, and a decreased sense of personal well-being. We study these in concert with two democratically beneficial outcomes, i.e., political engagement and political knowledge. We do so in two complementary preregistered experiments: one examining the effects of taking a 'news vacation' (N = 858) in the US, the other of 'news binging' (N = 804) in Poland.

Both experiments are part of a larger project involving panel surveys combined with browsing data in two different sociopolitical contexts. In both countries, we conducted a 3-wave panel study in which the same participants completed a 20-minute survey about their political views and news consumption every three months. As part of Wave 1, we asked participants to install Web Historian. This open-source plug-in tracks their browsing history across the survey's full duration on their local computers. Before each subsequent Wave, we asked participants to review their browsing data via the visualizations (e.g., network graphs, word clouds, search tables) provided in Web Historian. We subsequently asked participants to fill out an informed consent form and upload their browsing data before participating in the survey. We use these data in conjunction with a comprehensive list of news websites and machine learning algorithms to construct sophisticated behavioral compliance and (like-minded) news exposure measures.

We embedded the first experiment in Wave 3 of the US panel survey. We recruited participants via Lucid, an organization that aggregates survey participants from many different sources. We used quota sampling based on demographic information to acquire a sample composition that approximates the US population. Before collecting Wave 3, we randomly assigned participants to an experimental (~60%) or control (~40%) condition. Out of the 858 participants in Wave 3 (i.e., pre-test), 801 (93%) agreed to participate in the experiment. Ex-ante power analyses reveal that this sample size suffices to observe small treatment effects and that only very small effects would return an insignificant result. Participants in the experimental condition (N = 366) were incentivized to go on news vacation (i.e., stop following the news for one week). Participants in the control (N = 273) received no instruction. During this time, we recorded compliance in two ways. Every two days, we asked participants to fill out a short survey containing questions about their media diet. Besides this self-reported measure, we use their browsing data to check compliance for participants' online news consumption. Both checks confirmed that participants in the experimental group decreased their news consumption more than those in the control group. After seven days, both groups completed the post-test.

We embedded the second experiment in Wave 2 of the panel survey in Poland. For the data collection, we relied on Ariadna, an organization with over 286,000 registered survey participants. Ariadna ensures high-quality samples by sending rewards to its panelists by courier to assure that there are no bots in its panel and that an individual does not register multiple times. We drew a sample from Ariadna's database in such a way that its composition mirrored the population's key demographic characteristics. The treatment group assignment took place before Wave 2 (i.e., the pre-test; ~60% in the experimental and ~40% in the control condition). At the end of Wave 2, we asked the 982 participants to participate in an experiment, of which 801 (82%) opted in. Here, too, ex-ante power analyses reveal that our sample size suffices to observe small treatment effects if they exist. Participants in the news binging treatment (N = 398) were incentivized to consume more news for two weeks, while the control (N = 406) received no instructions to alter their news consumption. As in the first experiment, we checked compliance by asking participants to fill out a short survey every two days (6 in total) and analyzing their browsing history. After 12 days, participants completed the post-test.

We found no evidence of treatment effects: Neither the incentivized increase nor decrease in news consumption had a discernible impact on positive or negative outcomes. We observed that this null pattern exists among all participants, irrespective of their prior levels of self-reported or online (like-minded) news consumption. Together, our findings challenge scholarly and popular wisdom on the effects of news exposure.

Authors

• S.E. Edith Leeuw, University of Amsterdam

• Magdalena Wojcieszak, UC Davis, U of Amsterdam

• Andreu Casas, Vrije Universiteit Amsterdam