Messing, S., & Westwood, S. J. (2014). Selective Exposure in the Age of Social Media: Endorsements Trump Partisan Source Affiliation When Selecting News Online. Communic. Res., 41, 1042–1063.
Summary
This paper challenges the widespread pessimism that internet-driven media fragmentation inevitably deepens partisan selective exposure. Messing and Westwood argue that social media reconfigures the fundamental act of news consumption: instead of choosing a habitual partisan source, users choose individual stories recommended by a politically heterogeneous social network. Through two randomized experiments simulating social-media and news-aggregator interfaces, they demonstrate that social endorsements (aggregated “recommends,” Facebook likes, “most emailed” rankings) act as powerful decision heuristics that boost selection of endorsed content and, crucially, dilute the partisan source cues that normally drive selective exposure to levels indistinguishable from chance.
Key Contributions
- Recasts the internet’s role in polarization, providing experimental evidence that social media can reduce rather than amplify partisan selective exposure.
- Delivers what the authors describe as the first direct social-science comparison of the relative effects of social cues versus partisan source cues on news selection.
- Extends information-utility models of news selectivity by conceptualizing social endorsements as heuristic cues that trigger a bandwagon effect and signal social utility.
- Draws implications for agenda-setting theory, suggesting a diffusion of agenda-setting power from newsrooms toward social networks (a modern two-step flow).
- Flags design and policy concerns: filtering algorithms and “filter bubbles” could offset the diversifying potential of social endorsements.
Methods
Two randomized experiments used custom web applications mimicking social-media and aggregator interfaces, extending Iyengar and Hahn (2009). Study 1 (N=739, Amazon Mechanical Turk) used a three-condition between-subjects design (partisan-label-only, social-endorsement-only, combined), with live RSS-fed articles across four categories; endorsements were manipulated as low (0–1,000 recommends) versus high (>10,000). Source labels spanned partisan brands (Fox, MSNBC) and non-partisan outlets (USA Today, Reuters). Study 2 (N=141 undergraduates) used a richer within-subjects interactive interface with 80 daily-harvested stories, a countdown timer, and real Facebook like counts and “most emailed” rankings. Analysis relied on mixed multinomial (conditional) logit models (Study 1) and mixed logistic regression with random participant intercepts (Study 2), alongside comparisons against chance. A manipulation check confirmed source ideology was perceived as intended.
Findings
- In the partisan-label-only condition, selectivity was evident: Republicans chose Fox more than Democrats (0.38 vs 0.21), and Democrats chose MSNBC more than Republicans (0.30 vs 0.20).
- When social endorsements were present, Republican and Democratic selection rates became nearly identical, reducing partisan selectivity to chance levels.
- Partisans were more than twice as likely as chance to select strongly endorsed articles from ideologically dissonant sources — endorsements drew people toward counterattitudinal content.
- Strong endorsement effects were robust and, if anything, slightly stronger when partisan cues were also present.
- Republicans relied on high social cues more than Democrats, suggesting partisan differences in selection processes (possibly conformity-linked).
- Study 2 replicated the endorsement effect in a realistic browsing environment (0.141 vs 0.129 selection rate) with scant evidence of partisan source selectivity.
Connections
This paper is a foundational, optimistic counterpoint in the debate over whether digital platforms fuel political polarization, and it pairs naturally with Bakshy2015-rn, which similarly examines exposure to cross-cutting content via algorithmic and social curation on Facebook. Its argument about socially heterogeneous networks broadening exposure speaks to broader work on partisan sorting and echo chambers such as Iyengar2019-jj, Barbera2015-je, and Del-Vicario2016-uj, and its endorsement-as-heuristic framing connects to research on partisan sharing and misinformation dynamics like Osmundsen2021-et and Guess2023-ur.