Null effects of news exposure: a test of the (un)desirable effects of a 'news vacation' and 'news binging'

The study questions the widely held assumption that news exposure is inherently beneficial for democratic outcomes. It investigates whether increasing or decreasing overall news consumption affects desirable outcomes (political knowledge, participation, support for compromise) and potentially detrimental outcomes (attitude and affective polarization, negative system perceptions, and individual well-being). The authors hypothesize that because news tends to emphasize conflict and activate political identities, general news exposure—whether partisan or centrist—could exacerbate polarization, worsen perceptions of the political system, and harm well-being, even as it might aid knowledge and engagement. The research aims to provide a comprehensive test of both positive and negative individual-level effects of news exposure in realistic settings across two countries.

Prior research often highlights benefits of news use for knowledge and participation across media and countries. However, negative effects have been understudied and previous work frequently focuses on partisan media in the US, often using lab-like or forced-exposure designs and self-reports that may overestimate effects. Theoretical arguments suggest news content is conflict-focused, can activate partisan identity, and relay elite cues that may polarize citizens and increase out-group hostility. Existing encouragement and deactivation designs examine partisan outlets or social media use, with limited generalizability. The authors identify a need for ecologically valid tests of general news exposure, including potential adverse outcomes, using behavioral data in addition to surveys.

Design: Two preregistered encouragement experiments embedded in an international three-wave panel combining surveys with browser history via the open-source Web Historian tool. - US 'news vacation' (Wave 3): Participants (N=803 opted in; randomized 60% to treatment) were incentivized to avoid news entirely for 7 days across platforms (TV, radio, print, online, apps) and minimize conversations about current events. Control received no instructions. Post-test completion: treatment N=378, control N=288. - Poland 'news binging' (Wave 2): Participants (N=939 opted in; randomized 50% to treatment) were incentivized to consume more news than usual for 14 days via any medium. Control received no instructions. Post-test completion: treatment N=421, control N=402. Sampling: Quota samples approximating key census demographics (US via Lucid; Poland via Panel Ariadna). Ethical approvals obtained; preregistrations on OSF. Compliance measurement: Both self-reported and behavioral. - Self-report: Post-survey items (11 channels/platforms/devices) combined into a 0–1 index capturing change in the treatment direction. US: treatment reported lower news use than control (0.099 vs. 0.378). Poland: no difference (0.489 vs. 0.494). - Behavioral: Web Historian desktop trace data matched to comprehensive country-specific lists of news domains (US list ~5,400 sites; Poland 298), plus identified Facebook pages, Twitter handles, YouTube channels of these outlets. Compliance operationalized as change in average daily visits to news domains from pre-period to during-treatment. US: fewer daily news visits in treatment vs. control (3.86 vs. 5.22). Poland: more daily news visits in treatment vs. control (5.54 vs. 4.61). Alternative exploratory measures used a multilingual BERT classifier (accuracy 93%, Precision 0.92, Recall 0.91, F1 0.915) to categorize whether visited content was political, distinguishing political content within news domains (hard news) and political content anywhere online. Outcomes: Multiple-indicator constructs for (a) political knowledge (self-perceived and actual), (b) political participation, (c) support for compromise, (d) attitude polarization (importance and strength), (e) affective polarization (toward out-ideologues, out-partisans, and opposite-policy citizens via several measures), (f) attribution of malevolence to the out-party, (g) perceived societal polarization, and (h) well-being (mental and physical). Pre-wave measures included as controls where available. Analysis: OLS regressions estimating Intention-to-Treat (ITT). Complier Average Treatment Effects (CATE) via instrumental variable regressions using self-reported and behavioral compliance measures (including exploratory hard-news and any-political-content measures). False Discovery Rate (FDR) adjustment applied. Heterogeneous treatment effects tested by prior news exposure levels and ideological congeniality using both self-report and behavioral measures (domain ideology scores matched via validated models; US from Robertson et al., 2018; Poland open-source classification). Bayesian robustness checks with Bayes factors (BF01) comparing models including parameters of interest to baselines. Exposure baseline: In the month prior to experiments, only 3.10% of visits were to news domains (US 2.72%, Poland 3.72%), roughly one news domain per 33 sites visited.

- Primary result: Across both countries and experimental manipulations, changing overall news exposure (reducing for one week or increasing for two weeks) had no detectable effects on the tested outcomes. - Beneficial outcomes: No effects on self-perceived or actual political knowledge, political participation, or support for compromise. - Detrimental outcomes: No effects on attitude polarization (importance or strength), affective polarization (multiple measures), or attribution of malevolence to the out-party. - Small exceptions: (1) In Poland, ‘more news’ produced a small reduction in affective polarization as measured by feeling thermometer toward the out-party; (2) In the US, ‘no news’ modestly reduced perceived societal polarization. These effects were not robust across alternative measures and were small in magnitude (effects did not exceed about 2 percentage points on 0–100 scales). - Results held regardless of compliance (ITT and CATE estimates comparable), prior levels of news exposure, and ideological congeniality of prior news diets. - Bayesian analyses supported null effects: roughly 57% of BF01 > 10 (strong evidence for the null) and about 90% > 3 (moderate evidence for the null) across models. - Descriptive context: News constituted about 3% of overall browsing (US 2.72%, Poland 3.72%).

The findings directly address the hypothesis that general news exposure has meaningful short-term individual-level effects on political and well-being outcomes. Contrary to expectations of both positive and negative effects, neither decreasing nor increasing news use over 1–2 weeks altered knowledge, engagement, polarization, system perceptions (except a small perceived polarization decrease in the US), or well-being. The authors argue that these null results likely reflect the limited share of news in typical media diets, the entrenchment of political identities making short-term exposure shifts less consequential, and challenges in defining and capturing all forms of political information in today’s hybrid media environment. While news media are normatively central and can exert macro-level and long-term influences, short-term individual-level effects of modest changes in overall news exposure appear minimal in ecologically valid settings. The results suggest revisiting assumptions about the democratic benefits and harms of everyday news consumption and underscore the need for refined measurements and longer horizons to detect potential cumulative effects.

This study contributes an ecologically valid, preregistered, cross-national test of general news exposure effects using both surveys and behavioral browsing data. Incentivized ‘news vacation’ and ‘news binging’ interventions yielded largely null effects on a wide set of political and personal outcomes, with only minor, non-robust exceptions. The results suggest that short-term changes in overall news use, at typical exposure levels where news comprises a small fraction of media diets, have limited individual-level impact. Future research should: (1) identify exposure thresholds and durations necessary to move outcomes; (2) improve cross-device and offline tracking to better capture total news and political information exposure; (3) broaden conceptualizations of what counts as ‘news’ in hybrid media environments; (4) examine long-term, cumulative effects and macro-level consequences; and (5) test additional moderators (e.g., content type, medium, socio-demographics) that may condition effects.

- Sample generalizability: Non-probability online panels willing to share browsing data are not fully representative; estimates are not precise population ATEs. - Compliance: Although treatment groups showed shifts, behavior changes were modest; self-reports are biased, and some groups (Poland) showed no self-reported increase despite behavioral shifts. - Measurement coverage: Behavioral data capture desktop browsing; mobile and offline exposures (notably TV) were not behaviorally tracked. Social media incidental exposure to politics could occur outside measured news domains. - News definition: Domain-level classification may miss political content on non-news sites and non-political content on news sites; although exploratory political-content classification was used, other forms of content with ‘news value’ may be overlooked. - Short treatment duration: One to two weeks may be insufficient to detect changes; cumulative or long-term effects not assessed. - Unmodeled moderators: While prior exposure and congeniality (and exploratory ideology, education) were tested, other moderators (medium, platform, device, socio-demographics) could shape effects.