Physical partisan proximity outweighs online ties in predicting US voting outcomes

Marco Tonin, Bruno Lepri, Michele Tizzoni

Abstract

Affective polarization and increasing social divisions affect social mixing and the spread of information across online and physical spaces, reinforcing social and electoral cleavages and influencing political outcomes. Using aggregated and de-identified co-location and online network data, we investigate the relationship between partisan exposure and voting patterns in the USA by comparing three dimensions of partisan exposure: physical proximity and exposure to the same social contexts, online social ties, and residential sorting. By leveraging various statistical modeling approaches, we consistently find that partisan exposure in the physical space, as captured by co-location patterns, more accurately predicts electoral outcomes in US counties, outperforming online and residential exposures across metropolitan and non-metro areas. Moreover, our results show that physical partisan proximity is the best predictor of voting patterns in swing counties, where the election results are most uncertain. We also estimate county-level experienced partisan segregation and examine its relationship with individuals' demographic and socioeconomic characteristics. Focusing on metropolitan areas, our results confirm the presence of extensive partisan segregation in the US and show that offline partisan isolation, both considering physical encounters or residential sorting, is higher than online segregation and is primarily associated with educational attainment. Our findings emphasize the importance of physical space in understanding the relationship between social networks and political behavior, in contrast to the intense scrutiny focused on online social networks and elections.

Introduction

The "Great Sort" of American politics is frequently attributed to two primary drivers: residential segregation—living in politically homogeneous "red" or "blue" neighborhoods—and online echo chambers, where social media algorithms and user choice curate ideologically narrow information environments. However, these perspectives often overlook a critical middle ground: experienced exposure in daily activity spaces.

Individuals do not spend their entire lives at home or on digital platforms; they move through physical environments such as workplaces, transit hubs, and public parks. This study addresses whether these daily routines, which facilitate casual physical encounters, are more or less significant than residential or online ties in driving partisan segregation and predicting electoral outcomes.

Data and Methodology

To compare the relative influence of different dimensions of partisan exposure, the researchers integrated four major datasets:

  1. Colocation Maps (Meta/Facebook): An aggregated, de-identified dataset measuring the probability that two individuals from different counties were in the same 600m x 600m area for at least five minutes.
  2. Social Connectedness Index (SCI): Also from Meta, this measures the relative probability of Facebook friendship between individuals in different counties.
  3. Residential Exposure: Calculated from voter registration records, measuring exposure based on the nearest 1,000 neighbors.
  4. ANES 2020-2022 Social Media Study: Individual-level survey data from 2,420 respondents regarding their vote choice and the perceived partisanship of their offline and online social circles.
Conceptual framework and spatial data overview. Panel (a) illustrates the definitions of co-location, online friendship, and residential proximity. Panels (b) and (c) show the spatial distribution of co-location and friendship probabilities across the US. Panel (d) presents partisan exposure distributions.

The analysis employed Spatial Autoregressive Lag Models to account for regional influences, Dominance Analysis to rank predictor importance, and Machine Learning (Gradient Boosting and Random Forest) with SHAP explainability to identify key socioeconomic drivers of segregation.

Key Results

The study's findings challenge the common narrative that online platforms are the primary drivers of political division.

Predictive Power of Physical Space

Physical partisan exposure (co-location) was found to be the most accurate predictor of county-level voting outcomes, achieving an $R^2$ of up to 0.97. This significantly outperformed online ties ($R^2 \approx 0.85$) and residential proximity ($R^2 \approx 0.80$). This dominance was particularly pronounced in metropolitan areas. In individual-level logit models, the effect of offline exposure on vote choice was roughly 2 to 3 times stronger than that of online exposure.

Partisan Segregation and Education

Partisan segregation is notably higher in physical spaces than in online networks. While social media is often criticized for creating echo chambers, users are actually more likely to encounter opposing views on Facebook than during their daily physical routines.

Partisan exposure distributions for Republicans and Democrats across offline, online, and residential dimensions, categorized by U.S. counties, Metropolitan Areas, and Non-Metro Areas.

The strongest predictor of this offline partisan isolation is educational attainment. Highly educated areas exhibit greater Democrat-leaning segregation in physical movements, whereas less educated areas show higher Republican-leaning segregation. Furthermore, the most intense partisan reinforcement occurs in immediate, local physical environments; removing interactions within one's own county significantly reduced the predictive power of the co-location data.

Discussion and Limitations

These results emphasize that passive, casual encounters in physical space—often overlooked in favor of curated online feeds—are more strongly associated with voting behavior. However, several limitations remain. The data is derived from Facebook users, who may not represent the entire US population. Additionally, co-location (being in the same 600m tile) does not guarantee direct interaction or persuasion; it measures exposure rather than active influence. Finally, individual-level data relies on self-reported perceptions of friends' politics, which are subject to bias.

Future research should seek unified datasets that track both online and offline networks for the same individuals to better disentangle the flow of political influence between digital and physical realms.

Summary updated

Abstract

Affective polarization and increasing social divisions affect social mixing and the spread of information across online and physical spaces, reinforcing social and electoral cleavages and influencing political outcomes. Here, using aggregated and de-identified co-location and online network data, we investigate the relationship between partisan exposure and voting patterns in the USA by comparing three dimensions of partisan exposure: physical proximity and exposure to the same social contexts, online social ties, and residential sorting. By leveraging various statistical modeling approaches, we consistently find that partisan exposure in the physical space, as captured by co-location patterns, more accurately predicts electoral outcomes in US counties, outperforming online and residential exposures across metropolitan and non-metro areas. Moreover, our results show that physical partisan proximity is the best predictor of voting patterns in swing counties, where the election results are most uncertain. We also estimate county-level experienced partisan segregation and examine its relationship with individuals' demographic and socioeconomic characteristics. Focusing on metropolitan areas, our results confirm the presence of extensive partisan segregation in the US and show that offline partisan isolation, both considering physical encounters or residential sorting, is higher than online segregation and is primarily associated with educational attainment. Our findings emphasize the importance of physical space in understanding the relationship between social networks and political behavior, in contrast to the intense scrutiny focused on online social networks and elections.
BibTeX
@misc{tonin2024physical,
  author = {Marco Tonin and Bruno Lepri and Michele Tizzoni},
  title = {{Physical partisan proximity outweighs online ties in predicting US voting outcomes}},
  year = {2024},
  eprint = {2407.12146},
  archivePrefix = {arXiv},
  url = {https://arxiv.org/abs/2407.12146},
}
Figure