1. Introduction
In this article, we investigate how Russia's invasion of Ukraine affected support for the EU among European citizens. The EU has arguably been the most successful example of regional integration that the world has ever seen. After two devastating World Wars, the French foreign minister Robert Schuman proposed a European Coal and Steel Community to tame and integrate its former arch-enemy Germany into a European community. The European Coal and Steel Community founded in 1952 covering six countries developed into the EU over the years comprising 27 countries today. However, while the EU has been a project led by political elites, popular support for the EU has gone down significantly across countries for which the Brexit is only the most extreme example (Hooghe and Marks, Reference Hooghe and Marks2009; Hobolt and Vries, Reference Hobolt and Vries2016). According to De Vries et al. (Reference De Vries, Hobolt and Walter2021b), public opinion plays a crucial role in sustaining international cooperation through organizations like the EU, especially in contexts of heightened public contestation.
We argue that Russia's invasion of Ukraine has fuelled popular support for the EU as the war in Ukraine led to a rallying around the European flag effect (Baker and Oneal, Reference Baker and Oneal2001; De Vries et al., Reference De Vries, Bakker, Hobolt and Arceneaux2021a; Steiner et al., Reference Steiner, Berlinschi, Farvaque, Fidrmuc, Harms, Mihailov, Neugart and Stanek2022; Orenstein, Reference Orenstein2023). The invasion of a neighboring country is a salient political event that can influence Europeans’ attitudes toward the EU. After decades of European integration, a war in Europe seemed virtually impossible to most citizens. The invasion in Ukraine however reminded Europeans that peace cannot be taken for granted and building on Gehring (Reference Gehring2022) and Steiner et al. (Reference Steiner, Berlinschi, Farvaque, Fidrmuc, Harms, Mihailov, Neugart and Stanek2022), we posit that the external threat has altered the view of citizens toward more support.
External threats can reinforce group identities, as social identity theory suggests (Tajfel and Turner, Reference Tajfel and Turner1979; Hogg, Reference Hogg2016). Research shows that war exposure increases cooperation within in-groups (Bauer et al., Reference Bauer, Blattman, Chytilová, Henrich, Miguel and Mitts2016), as seen when repressive policies strengthen immigrant identities (Fouka, Reference Fouka2019) or national threats bolster regional identities (Dehdari and Gehring, Reference Dehdari and Gehring2022). Even diverse groups can unite against common threats, such as the collective Arab identity taking shape in response to ISIS (Barrie, Reference Barrie2021). In Europe, a strong European identity correlates with higher EU support (Hooghe and Marks, Reference Hooghe and Marks2004; Marks and Steenbergen, Reference Marks and Steenbergen2004; Gehring, Reference Gehring2021). Notably, Gehring (Reference Gehring2022) found that the 2014 Crimean occupation increased EU identity and support for European integration in the Baltics. In this paper, we examine how Russia's 2022 invasion of Ukraine, a broader threat to all of Europe, impacted EU support among Europeans.
We evaluate the effect of the war in Ukraine on EU support using an unexpected event during survey design (Muñoz et al., Reference Muñoz, Falcó-Gimeno and Hernández2020). Using Eurobarometer data collected between February and April 2022 (N = 26,502), we find that the invasion of Russia led to a 4–5 percentage-point increase (11–15 percent of an SD) in EU support for the representative sample of the EU. The effect becomes more pronounced as more days pass. Unlike Gehring (Reference Gehring2022), we do not find a stronger effect in countries bordering Russia, indicating a general boost in EU support across Europe. Importantly, we observe significant individual-level variation, with left-leaning respondents becoming more critical of the EU after the invasion.
This study contributes to the literature in the following ways. First, our study complements and extends research by Gehring (Reference Gehring2022) and Steiner et al. (Reference Steiner, Berlinschi, Farvaque, Fidrmuc, Harms, Mihailov, Neugart and Stanek2022) on the effect of Russian invasion on attitudes toward the EU. In comparison to these earlier studies, our approach is able to demonstrate, using a representative sample of the EU, the overall impact of the Russian invasion of Ukraine on Europeans’ attitudes toward the EU. Second, because the event occurred early in the fieldwork collection period, we can examine the influence of the invasion in the very short (1 week) to medium term (4 weeks). This enables us to see how public opinion evolved in the month following the invasion. Third, we are able to show individual-level heterogeneity in treatment effects using a causal forest approach, with ideology standing out as the most important predictor of heterogeneous treatment effects. Finally, we contribute to the literature on “rally around the flag” effects by demonstrating how regional events can spark a rally effect around a supranational organization, and provide suggestive evidence on how the EU citizens increasingly perceive the EU as a security actor in the face of the conflict.
2. Research design
We identify the impact of Russia's invasion of Ukraine on Europeans’ attitudes toward the EU using an unexpected event during survey design (UESD) (Muñoz et al., Reference Muñoz, Falcó-Gimeno and Hernández2020). It is possible to causally identify the impact of significant events on attitudes when the shock is unexpected and significant, it does not prevent the survey from being conducted, and there are no imbalances owing to quota sampling or reachability and non-compliance issues.Footnote 1 In Appendix Section A we further explain how our design satisfies these conditions.
We use Special Eurobarometer 97.1 which was fielded between 21 February and 22 March 2022 (European Commission, Reference European Commission2022). The representative sample consists of 26,502 respondents aged 15 or older who are citizens (or residents) of each of the 27 Member States. Appendix Section A provides more information regarding sample procedures, survey representativeness, and sample sizes per country.
Our treatment variable, T i, takes the value 1 if respondent i was interviewed after February 23, 2022, and 0 otherwise. Since the Russian attack was announced early on February 24 (4 a.m. CET), we assume all respondents surveyed that day were treated with this information. We later relax this assumption in Tables 18 and 19. The number of participants within the treated and control groups for the main dependent variables are as follows: n Control = 1887 and $n_{\text {Treated}} = 22,\; 611$
.Footnote 2
We assess popular support for the EU based on respondents’ preference for remaining in the EU following the approach outlined by Hobolt and Vries (Reference Hobolt and Vries2016). In particular, Eurobarometer 97.1 asked, “Would your country face a better future outside the EU?” (1 to 4). This question holds particular relevance, especially during critical moments such as a war in a neighboring country, as it allows for an assessment of citizens’ stance on remaining within the EU.Footnote 3
We estimate the effect of the invasion on EU support using four specifications. The “Basic” model includes the treatment and country fixed effects, clustering standard errors at the country-day level. The “Extended” model adds time trends to control for unrelated temporal patterns. The “Full” model includes individual-level controls, and the “Balanced” model applies entropy balancing to correct any imbalance between treatment and control groups (Hainmueller and Xu, Reference Hainmueller and Xu2013). More details are provided in Appendix Section A.
3. Results
3.1. Main test
We begin by assembling a combined sample of respondents from all EU member states, which we refer to as the “EU Sample.” We estimate the effect of our treatment variable, Russia's invasion of Ukraine, on EU support. Figure 1 illustrates treatment effects. We find a statistically significant and sizeable effect of the invasion on EU support by a 4 percentage-point for the pooled EU sample (Table 6). This translates to roughly 11 percent of a standard deviation (Table 7). The effect is robust to including time trends, controls, and entropy weights.Footnote 4 We conduct further robustness tests in Appendix Section I. Importantly, when we remove the assumption that respondents were already treated on February 24, treatment effects go up to 15 percent of an SD (Table 18).
ITT in standard deviations at the EU level (N = 26,502). Notes: Appendix Table 7 illustrates the underlying regression table. The treatment coefficient is expressed in standard deviation.
We validate our findings by replicating the analysis on the eupinions (2020) dataset using the same analytical framework. Details are in Appendix Section H.2. The dependent variable measures respondents’ intention to vote for their country to remain in the EU. In the “Balanced” model (Table 14), we find a significant 4 percentage-point increase in EU support among the treatment group (p < 0.001), confirming that the effect holds across datasets.
Next, we examine treatment effects over time (see Table 5 for weekly observation counts). Figure 2 illustrates EU support in the very short run (first week), short run (2 weeks), and medium run (3 weeks), compared to the full sample (4 weeks). We find that the effect increases over time, likely due to increased exposure to news about the invasion.
ITT in standard deviations at the EU level—treatment effect over time (N = 26,502). Notes: Appendix Table 8 illustrates the underlying regression tables. The treatment coefficient is expressed in standard deviation.
We also examine the impact of the invasion on respondents facing varying degrees of threat from Russia. The results do not show heterogeneous treatment effects by residing in a country bordering Russia, in contrast to the findings of Gehring (Reference Gehring2022) (Table 9). We conduct several robustness checks which confirms this finding (Table 10). Finally, we explore additional outcome variables from the Special Eurobarometer and discuss mechanisms extensively in Appendix Section H.
3.2. Individual-level heterogeneity
In this section, we explore how the war's impact on EU attitudes varies across individuals using the causal forest approach, a machine learning method based on the Generalized Random Forests (GRF) algorithm by Athey and Imbens (Reference Athey and Imbens2016), which estimates conditional average treatment effects (CATEs) across covariates. This method detects significant effect size variations by dividing the data with regression trees. Additional details are in Section A4.
Figure 3 shows the variable importance measure, highlighting ideology as the top predictor of heterogeneous treatment effects, followed by age. We also estimate interaction effects (Tables 12 and 13) and calculate marginal effects by ideology and age (Appendix Figures 10 and 11). The results reveal that left-leaning individuals, initially most supportive of the EU (pred. 3.16/4), decreased their support (pred. 3.06/4), while center- and right-leaning individuals increased theirs, thereby narrowing the ideological gap. All age groups showed increased EU support, with the youngest group (<30) demonstrating the largest increase in support.
Heterogeneity based on causal forest (N = 26,502). Notes: Variable importance measures variable importance as the sum of the absolute values of the standardized total causal effect estimates (CATEs) across all trees in the forest, for each covariate.
4. Conclusion
This paper examined the impact of Russia's invasion of Ukraine on Europeans’ attitudes toward the EU. Using data from Special Eurobarometer 97.1, which included 26,502 respondents from all 27 EU member states, we found a significant increase in EU support following the invasion. These findings were further confirmed with the eupinions dataset, where we observed identical effect sizes on a question directly asking whether respondents would vote to remain in the EU.
Compared to earlier studies, our findings are more moderate. Steiner et al. (Reference Steiner, Berlinschi, Farvaque, Fidrmuc, Harms, Mihailov, Neugart and Stanek2022) observed a larger increase in EU support among Erasmus students, while Gehring (Reference Gehring2022) found a stronger effect in the Baltics after Crimea's annexation. Our smaller effect size is likely due to analyzing the average effect across a representative sample of all EU citizens.
Our analysis also revealed that the invasion's impact on EU support differed by ideology, with left- and right-leaning respondents reacting in opposite ways. This finding aligns with Truchlewski et al. (Reference Truchlewski, Oana and Moise2023), who highlighted the interaction between geopolitical events and ideology. However, while they focused on polarization over time, we demonstrated that ideological orientation influenced responses to the invasion itself.
In conclusion, our study demonstrates that external events like the invasion can significantly shape attitudes toward the EU, with potential long-term implications for its trajectory. However, we acknowledge the limitation imposed by the control group size, preventing us from conducting a precise analysis for each country individually. Future research could examine cross-country differences, explore the mechanisms through which major events shape attitudes toward the EU, and identify the influence of other pivotal events on these attitudes, further enriching our understanding of European public opinion in response to transboundary crises.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/psrm.2024.62. To obtain data and replication material for this article, https://doi.org/10.7910/DVN/WRLXWI.
Financial support
We thank the VolkswagenStiftung for generously funding the COVIDEU project on which this study is based (grant: 9B051).
Open access
