Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Political attitudes differ but share a common low-dimensional structure across social media and survey data

Antoine Vendeville, Hiroki Yamashita, Pedro Ramaciotti

Abstract

Does polarization online reflect the state of polarization in society? We study ideological positions and attitudes on several issues in France, a country with documented issue nonalignment. We compare distributions on X/Twitter with a nationally representative sample, focusing on two key properties: ideological polarization and issue alignment. Despite significant issue-wise divergences, positions of both the X population and the nationally representative sample present a similar bi-dimensional structure along two dominant bundles of aligned issues: a Left-Right divide, and a Global-Local divide. We then study how our results vary when accounting for key structural parameters of the online public sphere: activity, popularity, and visibility. We find that the dimensionality of attitude distributions shrinks as ideological polarization increases when selecting more active users. The divergence between political attitudes on social media and in survey data is greatly mediated by the combination of activity and popularity of social media users: users benefiting from the most exposure are also the most representative of the general public. Together, our results shed light on the structural similarities and differences between political attitudes from social media users and the general public.

Political attitudes differ but share a common low-dimensional structure across social media and survey data

Abstract

Does polarization online reflect the state of polarization in society? We study ideological positions and attitudes on several issues in France, a country with documented issue nonalignment. We compare distributions on X/Twitter with a nationally representative sample, focusing on two key properties: ideological polarization and issue alignment. Despite significant issue-wise divergences, positions of both the X population and the nationally representative sample present a similar bi-dimensional structure along two dominant bundles of aligned issues: a Left-Right divide, and a Global-Local divide. We then study how our results vary when accounting for key structural parameters of the online public sphere: activity, popularity, and visibility. We find that the dimensionality of attitude distributions shrinks as ideological polarization increases when selecting more active users. The divergence between political attitudes on social media and in survey data is greatly mediated by the combination of activity and popularity of social media users: users benefiting from the most exposure are also the most representative of the general public. Together, our results shed light on the structural similarities and differences between political attitudes from social media users and the general public.
Paper Structure (20 sections, 5 equations, 7 figures, 1 table)

This paper contains 20 sections, 5 equations, 7 figures, 1 table.

Table of Contents

  1. Introduction
  2. Results
  3. Dimension-wise comparison of online and offline positions
  4. Alignment between political dimensions
  5. Low-dimensional structure
  6. Variations of political positions with activity and popularity
  7. Visible opinions in the online sphere
  8. Discussion
  9. Methods
  10. X dataset
  11. European Social Survey
  12. Taking into account survey design for ESS data
  13. Binning X panelists
  14. Measuring ideological polarization
  15. Wasserstein distance
  16. ...and 5 more sections

Figures (7)

  • Figure 1: Distributions of positions inactive the X panel and the ESS panel.(a) Histograms of attitude and ideology position distributions. Dotted vertical lines indicate the mean. Distributions for the X panel are sorted into bins to facilitate comparison with the ESS panel. (b) Polarization values across dimensions. Brackets indicate groups of dimensions that share the same support, and within which polarization values can be compared. (c) Normalized Wasserstein distance between X and ESS distributions, sorted from lowest (top) to highest (bottom).
  • Figure 2: Pearson correlations between dimensions. Significance levels *: $p<.05$, **: $p<.01$, ***: $p<.001$. Immig: immigration, SocLib: Social liberalism, Env: environmental policies, Redist: redistribution, DirDem: Direct democracy. Percent of significant correlations: 100% (X), 71% (ESS). Average value of significant off-diagonal correlations: 0.56 (X), 0.18 (ESS). Strongest off-diagonal correlation: $0.97$ (X, Anti-elitism and EU integration), $0.41$ (ESS, Anti-elitism and Direct democracy). Each matrix differs from the identity (Bartlett's sphericity test: $\chi_2>300, p<0.001$). Non-significant correlations and diagonal cells are colored in white.
  • Figure 3: Variance explained by the principal components. We show explained (a) and cumulated explained variance (b) of populations on the European Social Survey (ESS) and X (previously Twitter). ED stands for Effective Dimensionality.
  • Figure 4: Loadings of political dimensions along principal components. Each arrow shows the loading of a political dimension along PC1 (x-axis) and PC2 (y-axis). Arrow length is proportional to the absolute value of the loading. Red-blue arrows correspond to dimensions related to the Left-Right attitudes, and green-brown arrows correspond to dimensions related to Global-Local attitudes Plus (+) signs are short for "pro", and minus (-) signs for "anti".
  • Figure 5: Distribution of panelists in the low-dimensional space. The (PC1,PC2) plane is divided into a grid. Vertical bars indicate the number of individuals in each cell. Bar colors map the average position of individuals on the Left-Right and Anti-elitism dimensions. (a,b) X panel, grid size $30\times 30$. (c,d) ESS panel, grid size $20\times 20$.
  • ...and 2 more figures