In an era where digital interactions increasingly dominate the social landscape, new research underscores the enduring and powerful influence of offline social networks in shaping political behavior. Groundbreaking findings by Michele Tizzoni, Marco Tonin, and their colleagues at the University of Trento reveal that physical proximity between individuals with varying political affiliations plays a more critical role in predicting voting patterns in the United States than online social connections or residential sorting. Their comprehensive study, published in PNAS Nexus, leverages large-scale co-location data from Meta’s Data for Good program alongside additional datasets to offer an unprecedented look at how real-world interactions influence political partisanship.
The study harnesses anonymized data gathered from Facebook users who enabled location tracking on their smartphone applications to track patterns of physical colocation—defined as two or more people occupying the same spatial area no larger than approximately 600 meters by 600 meters at the same time. By analyzing these interactions, the researchers were able to create a robust measure of “offline social networks” that reflect real-world encounters and proximity between individuals. This innovative approach moves beyond conventional methods that often focus exclusively on digital interactions or residential demographic data.
Political affiliation in this research was inferred from the county of residence for each individual, allowing the team to correlate spatial mixing patterns with known local voting behaviors. The counties where the researchers observed minimal mixing between self-identified Republicans and Democrats tended to exhibit strong voting polarization, observable both in physical spaces and in voting outcomes. The intensity of partisan segregation in tangible spaces far exceeded that found in online social networks, suggesting that physical barriers and social behaviors profoundly sculpt political groupings in the United States.
When juxtaposed with online data—Facebook friendship connections—and residential proximity based on voter registrations of the nearest thousand neighbors, physical co-location demonstrated superior predictive power for county-level voting outcomes. The statistical models revealed that offline exposure to partisanship explained a remarkable 97% of the variance in voting patterns across counties. By contrast, online social networks accounted for 85-87% of variance, while residential proximity explained an even lower range of 75-80%. These findings indicate that while digital connectivity captures some aspects of political alignment, it cannot fully substitute the nuanced influence wielded by real-life interactions and shared physical environments.
To deepen their understanding, the researchers analyzed survey data from 2,420 Americans conducted between 2020 and 2022 as part of the Social Media Study administered by the American National Election Studies. This individual-level data confirmed that offline social ties exerted a more potent influence on voter choice than online interactions. The respondents reported greater exposure to politically like-minded individuals in their physical social networks compared to their online worlds, reinforcing the concept that political attitudes are strongly shaped by face-to-face encounters and localized social environments.
The study sheds new light on the relationship between political partisanship and social context, illustrating that physical spaces remain a vital arena for political influence. Educational attainment emerged as a significant factor structuring partisan segregation, with higher education levels correlating with different patterns of spatial mixing and political alignment. This highlights the complex interplay between socioeconomic factors and political geography and suggests that educational status may act as an important moderator in how political attitudes and affiliations manifest in daily life.
Importantly, these results challenge the common narrative that internet technologies and social media platforms are the primary drivers of increasing political polarization in the United States. Instead, this work elucidates that political polarization is not confined to cyberspace but is deeply rooted in the tangible social and physical landscape people inhabit. The real-world social segregation based on political identity, as opposed to solely online echo chambers, plays a foundational role in shaping voter behavior.
The methodology of this research is pioneering in its use of colocation data as a proxy for offline interaction networks. By harnessing location services data collected passively from millions of smartphone users, the study achieves an unparalleled spatial and temporal resolution of human movement and interaction patterns. This approach allows for accurate capture of the diversity and intensity of real-world social exposures—dimensions that are impossible to replicate through self-reported surveys or digital friendship graphs alone.
Furthermore, the research incorporates a multi-level analytical framework combining large-scale, county-level aggregation with granular, individual-level social network data. By bridging these scales, the authors provide a comprehensive and multidimensional portrait of the social environments in which political beliefs and voting decisions are formed and reinforced. This integrative perspective is critical for understanding the mechanisms underpinning political partisanship and voter behavior.
One of the most profound implications of this work is the renewed appreciation for the social ecology of political polarization. Unlike isolated online interactions, real-world social networks are embedded in complex physical and social infrastructures—workplaces, neighborhoods, public spaces—that dictate when and how individuals encounter political diversity or homogeneity. These environments facilitate stronger affective and social ties, which in turn potentiate the alignment of political attitudes and behaviors.
Moreover, the high predictive validity of offline social exposure for election outcomes suggests potential avenues for political campaign strategies and interventions aimed at bridging partisan divides. Targeting real-world social settings and leveraging physical interactions may prove more effective at altering political beliefs or reducing polarization than solely focusing on online digital platforms. The work urges policymakers, political scientists, and social technologists to reorient efforts towards acknowledging and integrating the power of offline social dynamics.
In summary, this compelling research provides robust empirical evidence that physical partisan proximity overwhelmingly outweighs online social connections as a predictor of voting outcomes in the United States. It invites a more nuanced understanding of political polarization as a phenomenon deeply embedded in spatial and social realities rather than merely a product of internet-driven digital fragmentation. As societies continue to navigate the evolving digital age, appreciating the foundational role of real-world social networks remains essential to grasping the true dynamics of political behavior.
The study opens new frontiers for interdisciplinary research in political science, social network analysis, geography, and data science. Future inquiries can build on these findings to examine how different types of physical social environments contribute to political identity formation, how mobility patterns influence exposure to political diversity, or how interventions designed around offline social spaces might mitigate partisan polarization. This work stands as a pivotal contribution to the ongoing quest to decode the complex social fabric underpinning electoral outcomes in modern democracies.
Subject of Research: Political partisanship and social networks in the United States.
Article Title: Physical partisan proximity outweighs online ties in predicting US voting outcomes.
News Publication Date: 21-Oct-2025.
Image Credits: Tonin et al.
Keywords: Political science, social networks, partisanship, offline social networks, online social connections, voting behavior, political polarization, co-location data, Meta Data for Good, electoral outcomes.
