Social media, education, and the rise of populist Euroscepticism

Public sentiment toward EU institutions has shifted markedly over recent decades, with support for Eurosceptic parties more than doubling since the early 1990s. Concurrently, the media landscape has transformed, with online platforms and social media becoming central to political information flows, often amplifying divisive content. Economic shocks (e.g., the global financial and European debt crises) and rising inequality also contributed to anxiety and political disaffection, especially among vulnerable groups. Education emerges as a crucial factor: it proxies cultural capital and social trust, both associated with resilience to divisive narratives and higher support for integration. The authors hypothesize that exposure to online political content is associated with stronger Eurosceptic attitudes predominantly among lower-educated individuals, whereas higher education mitigates or reverses these associations. They test this using European Social Survey (ESS) data and Italy’s Multipurpose Household Survey (MHS), focusing on both trust in the European Parliament and preferences for exiting the EU, and differentiating between general online political use and social-media-mediated political information.

The paper defines Euroscepticism as skepticism or opposition toward the EU and its integration process, encompassing both soft (low trust in EU institutions) and hard (support for leaving) forms. While not identical, Euroscepticism and populism are closely associated in Europe; many populist parties (right and left) are also Eurosceptic. Social media’s design—engagement-maximizing, popularity-based algorithms—facilitates the spread of emotionally charged, polarizing content, potentially heightening partisan animosity and mistrust in institutions. Prior studies suggest online political exposure can foster polarization and sectarianism, with experiments showing social media deactivation reduces polarization and increases well-being. Literature consistently links lower formal education to higher Euroscepticism, via economic self-interest (reduced competition for the highly educated in integrated markets), cognitive skills fostering flexibility in internationalized environments, and sociocultural capabilities to relate to abstract political communities like the EU. This paper builds on these strands by examining how education interacts with online political exposure—especially via social media—to shape Eurosceptic attitudes.

Data: (1) European Social Survey (ESS) rounds 8 (2016) and 9 (2018) across 17 EU countries. Outcomes: trust in European Parliament (0–10) and an EU-exit voting intention dummy. Key regressors: years of education; exposure to online politics (dummy = posted/shared about politics online in last 12 months, including blogs/email/social media). Controls: sex, age and age squared, marital status, foreign-born status, urban/rural residence dummies, household income deciles; country fixed effects; 2018 round fixed effect; survey weights. Estimation: ordered probit for trust, probit for EU-exit; models include an interaction between online politics and education. (2) ISTAT Multipurpose Household Survey (Italy) 2013–2016. Outcome: trust in European Parliament (0–10). Education: three levels (compulsory, high-school diploma, tertiary). Online political information: (a) via social media (Facebook/Twitter), (b) online without social media (websites linked to traditional media or blogs). Controls: sex, age group, marital status, household type, urban level; Italian region fixed effects; year fixed effects (in pooled models). Estimation: ordered probit with robust SEs clustered at the individual level, including interactions between online source type and education. Robustness: address omitted-variable concerns (computer skills) by adding self-assessed computer/software skill proxies (available 2015–2016). Address endogeneity of online exposure using IV ordered probit (2014–2016) with first-stage probits for (a) online politics without social media and (b) via social media, instrumented by access to connection types: DSL, smartphone, SIM/USB, ISDN (yes/no). IV models split samples into low educated (compulsory schooling) versus higher educated (high-school, tertiary). All models include region fixed effects and standard controls.

- Descriptive context: In ESS 2016–2018, mean trust in the European Parliament ≈ 4.26 (full sample) and 4.33 (employed); about 19% favor leaving the EU, with substantial cross-country variation (e.g., Ireland ~8% leave; UK ~40%). In Italy (ISTAT 2013–2016), mean trust ≈ 3.75 among employed; it declined from 2013 to 2015 and stabilized in 2016. - ESS regression results (Table 1): • Education is negatively associated with Euroscepticism: higher years of education lead to higher trust and lower support for exiting (e.g., EU-exit probit: years of education coefficients around −0.043 to −0.058, p<0.01). • Without interaction, online politics has mixed/weak effects on trust but is positively associated with EU-exit propensity (e.g., full sample coef 0.063–0.079, p≤0.05–0.01). • Interaction: Online politics × education significantly moderates effects: — Trust model (ordered probit, Panel A): interaction positive and significant (e.g., 0.009**; 0.018***), implying that online exposure is associated with higher trust among more educated individuals, and conversely with lower trust among the less educated. — EU-exit model (probit, Panel B): interaction negative and significant (e.g., −0.026***, −0.030***, −0.037***), indicating that online exposure’s association with supporting EU exit is concentrated among lower-educated individuals. • Marginal effects (Fig. 2): Online political exposure increases the probability of “no trust” and of being pro-exit only at low education levels; at high education, effects are neutral to positive for trust and negative for exit. - ISTAT Italy results (Table 2): • Mode of online information matters. Online politics via social media is negatively associated with trust (e.g., full sample −0.046***; employed −0.083***), while online politics without social media is positively associated with trust (e.g., full sample 0.048***; employed 0.045***). • Lower education (compulsory or high-school) is associated with lower trust vs tertiary. • Interactions: The trust-decreasing association of social media use is significantly stronger for the least educated (e.g., online via social media × compulsory school: −0.098*** full sample; −0.184*** employed). - Robustness with computer-skill controls (Table 3, 2015–2016): Including multiple proxies for self-assessed computer/software skills (jointly significant by F-tests) does not overturn results; the interaction of social media use with compulsory education remains negative and significant. - IV analysis (Table 4, 2014–2016): Instrumenting online exposure with connection-type availability: • Among low educated, online politics via social media remains negatively and significantly associated with trust (IV coef ≈ −0.216***). • Online politics without social media is not significant in IV models. • Among higher educated, neither online exposure mode has a robust negative association with trust (e.g., social media IV coef ≈ 0.057, ns). • First-stage diagnostics suggest strong instruments (F-tests > 10). - Overall: The association between online political exposure and Euroscepticism is concentrated among lower-educated individuals, and the negative association is driven by social-media-mediated political information, not generic internet use.

The findings support the hypothesis that education conditions the relationship between online political exposure and Euroscepticism. Social-media platforms, designed to maximize engagement, algorithmically amplify divisive content, disproportionately affecting less-educated users who are both more skeptical of EU integration and less likely to filter misinformation. This helps explain the rise and electoral resonance of Eurosceptic and populist narratives that capitalize on distrust of EU institutions. The results underscore social media’s systemic impact on democratic attitudes, showing that platform-mediated political communication can empower fringe or extreme narratives and marginalize moderates, thereby eroding institutional trust. By distinguishing between social-media and non-social-media online information, the study clarifies that the medium itself—social networking platforms—plays a key role in shaping Eurosceptic sentiments among vulnerable groups, beyond mere internet access.

The study documents that lower education is associated with higher Euroscepticism and that exposure to online political content reinforces Eurosceptic attitudes primarily among the less educated. Importantly, it is not general internet use but specifically social-media-mediated political information that is negatively associated with trust in EU institutions for lower-educated individuals. Using cross-country (ESS) and national (ISTAT) surveys, interaction models, robustness checks with computer-skill controls, and IV strategies for online exposure, the results are consistent and robust. Future research directions include linking social media diffusion directly to voting behavior in elections and referenda; differentiating effects across far-left versus far-right populisms; and exploring interactions between social media exposure and public health behaviors (e.g., vaccination attitudes). Policy-oriented work on platform governance and institutional reforms could help mitigate polarization and bolster democratic resilience.

- Causal identification is challenging: online exposure variables are potentially endogenous (politically motivated or Eurosceptic individuals may self-select into social-media political content). - IV strategy relies on connection-type availability (DSL, smartphone, SIM/USB, ISDN) available only for 2014–2016 in Italy; instrument validity, while supported by strong first-stage F-statistics, cannot address all concerns (e.g., exclusion restrictions). - Omitted-variable risks (e.g., digital literacy): addressed using self-assessed computer-skill proxies only in 2015–2016; such proxies may be imperfect. - Generalizability: ESS covers a subset of EU countries and time points; ISTAT results pertain to Italy. - Outcomes are attitudinal self-reports, not observed voting behavior. - Data constraints: ISTAT microdata access is restricted; some analyses cannot be replicated publicly.