Economic interests

While most Europeans are not directly exposed to the war, many experience its economic consequences. The EU, US, and their allies adopted economic sanctions against Russia after it annexed Crimea in 2014, which they significantly expanded in the wake of the 2022 full-scale invasion (S&P Global 2024). These measures harmed Western firms tied to trade with Russia. Such companies represent a significant share of the European economy: a recent study found that 11 per cent of large European corporations mentioned anti-Russia sanctions in their annual reports between 2014 and 2017 (Davydov et al. Reference Davydov, Sihvonen and Solanko2022). The resulting economic strain – through reduced bonuses, salary cuts, and lay-offs – likely imposed costs on affected firms’ employees. Sanctions and the broader conflict also steeply increased retail energy prices (Ferriani and Gazzani Reference Ferriani and Gazzani2023). Although European governments implemented various mitigation measures, the rising energy prices contributed to a severe cost-of-living crisis (Coi Reference Coi2023), which particularly affected vulnerable groups, including pensioners and low-income households (Charalambakis et al. Reference Charalambakis, Fagandini, Henkel and Osbat2022; Ernst and Young 2024; Mandys Reference Mandys2024).

The economic costs of anti-Russia sanctions may explain some pro-Kremlin attitudes in Europe. As reviewed above, several studies find that foreign policy attitudes reflect cost-benefit calculations (Verba, Brody, Parker et al. Reference Verba, Brody, Parker, Nie, Polsby, Ekman and Black1967; Larson and Savych Reference Larson and Savych2005; Reifler, Clarke, Scotto et al. Reference Reifler, Clarke, Scotto, Sanders, Stewart and Whiteley2014), and that citizens are particularly sensitive to the costs of security and defence cooperation (Mader et al.Reference Mader, Neubert, Münchow, Hofmann, Schoen and Gavras2024) and policies towards Ukraine (Stolle Reference Stolle2024). In the current context, it is not Russia’s aggression per se that makes some citizens worse off, but rather the West’s response. If the aggression were seen as justified, sanctions – and their domestic costs – would appear illegitimate. Citizens bearing these costs may thus be more receptive to pro-Kremlin narratives. Pro-Kremlin disinformation campaigns have exploited this vulnerability by linking high living costs in the EU to sanctions (Serhan Reference Serhan2022; Canetta et al. Reference Canetta, Danko and Dukach2023). For instance, in Czechia, populist forces organised large anti-sanction, anti-government, and pro-Russian protests, under slogans like ‘Czechia against misery’ in 2022 and 2023 (Kostelka Reference Kostelka and Gladman2023). We thus hypothesise that, following the pocketbook logic, the more citizens depended on economic ties with Russia or struggled during the energy crisis, the less they support Ukraine.

Ideology

Citizens’ long-term political outlooks may shape their understanding of the current conflict. Indeed, some of the reviewed studies on public opinion on policies towards the war in Ukraine suggest that ideology trumps interests and utilitarian considerations (Eck and Michel Reference Eck and Michel2025; Moise and Wang Reference Moise and Wang2025a). More generally, earlier research found ideology to inform preferences on the use of military force, foreign policy, and transnational integration (Holsti and Rosenau Reference Holsti and Rosenau1996; Hooghe and Marks Reference Hooghe and Marks2004; Reifler, Scotto and Clarke Reference Reifler, Scotto and Clarke2011; Hobolt and de Vries Reference Hobolt and de Vries2016). Regarding the Russo-Ukrainian war, three ideological dimensions appear especially relevant: the general position on the left-right scale, conservative authoritarianism, and pacifism.

Prior studies show that extreme-left and extreme-right respondents tend to hold more pro-Russia attitudes than centrists (Onderco Reference Onderco2017). On the extreme left, this reflects historical ties to the Soviet Union, once seen as an ideological ally across Europe (March and Mudde Reference March and Mudde2005). After 1991, prominent extreme-left intellectuals opposed US foreign policy, denouncing it as imperialist while downplaying Russia’s own imperialism (Kostelka Reference Kostelka2022a, Reference Kostelka2022b). On the extreme right, pro-Kremlin leanings likely stem from affinities with Russia’s authoritarian political regime and some far-right parties’ proximity to the Kremlin, which we further discuss below.Footnote ⁷ We expect that when these other factors are controlled for, the extreme right will not be pro-Russian. However, without such controls, both extremes are likely to be positively associated with lower support for Ukraine.

Citizens’ ideological proximity to the belligerents’ political regimes may also inform public opinion on the war. Ukraine aspires to join the EU’s liberal democratic order, while Russia is an authoritarian state promoting strongman rule and ultra-conservative values (Diesen Reference Diesen2020; Frye Reference Frye2021). Citizens with authoritarian and culturally conservative views – who favour strong leaders and question minority rights – are more likely to sympathise with the Kremlin. Consequently, we hypothesise that authoritarianism and cultural conservatism correlate negatively with support for Ukraine.

An ideological orientation particularly salient to war is pacifism. At its core, pacifism is the belief that wars and violence are fundamentally unjustifiable (Howes Reference Howes2013). In the current context, pacifists should oppose Russia as the aggressor. However, pro-Kremlin disinformation has instrumentalised calls for peace, claiming that supporting Ukraine means prolonging the war (Disinformation Analysis Center 2023). Therefore, pacifism may be confounded by factors such as exposure to pro-Russian narratives and disinformation. Only when these factors are accounted for may the positive relationship between pacifism and pro-Ukrainian attitudes become apparent.

Partisan alignment

The next set of hypotheses relates to elite cues and partisan alignment. Research has repeatedly shown that citizens possess shallow knowledge of foreign policy issues and, instead, rely on heuristics to form opinions (Zaller Reference Zaller1992; Berinsky Reference Berinsky2009; Guisinger and Saunders Reference Guisinger and Saunders2017). In the context of the war in Ukraine, elite positions offer the most accessible heuristic.

As with other foreign and complex policy issues, citizens are likely to follow their government or favourite political parties when assessing Russia’s aggression. With a few significant exceptions, such as Viktor Orban’s cabinet in Hungary and Robert Fico’s recent cabinet in Slovakia, European governments have clearly supported Ukraine (Trebesch et al. Reference Trebesch, Antezza, Bushnell, Frank, Frank, Franz, Kharitonov, Kumar, Rebinskaya and Schramm2023). Party positions vary more, with populist and eurosceptic parties generally less supportive (Hooghe, Marks, Bakker et al. Reference Hooghe, Marks, Bakker, Jolly, Polk, Rovny, Steenbergen and Vachudova2024). Earlier research found that the long-term proximity of far-right and far-left parties to the Kremlin guided their supporters’ views on Russia (Fisher Reference Fisher2021). At the same time, there is variation even within these families: only about half of far-right parties oppose sanctions and blame the West (Wondreys Reference Wondreys2025). It may thus be essential to distinguish between individual parties and not to rely on party families when studying attitudes to the conflict. We hypothesise that party preferences and support for the incumbent cabinet account for individual-level variance in attitudes towards the war. The closer citizens are to parties favourable to Russia, the less they support Ukraine. Similarly, the more citizens support their government, the more likely they are to align with its position towards the war.

Our theory assumes that the potential congruence between party or government positions and those of their supporters, as articulated in Hypotheses 3.1 and 3.2, reflects processes of elite cueing. This implies, among other things, that such congruence should be stronger among individuals with more pronounced party identification, since they are more likely to be exposed to, attend to, and internalise their party’s positions.

Two caveats are in order. First, our hypotheses are agnostic about the timing of the cues, which may have occurred before and after the 2022 invasion. Partisan alignment may therefore arise not only from elite communication about the invasion itself, but also from longer-standing foreign policy orientations and pre-invasion messaging about the belligerents. We are unable to disentangle these temporal dynamics in the following empirical analysis and leave this task to future research that can measure elite cues directly.

Second, observed congruence could in principle reflect reverse causality: citizens may support parties or governments because of those actors’ positions on the conflict. For instance, individuals holding pro-Ukrainian preferences might shift their support toward elites expressing similar stances following Russia’s invasion. This alternative explanation appears substantially less plausible, given that prior research shows that voters care far more about domestic issues than foreign policy (Page and Shapiro Reference Page and Shapiro1992; Lavine et al. Reference Lavine, Sullivan, Borgida and Thomsen1996; Guisinger Reference Guisinger2009; Saunders Reference Saunders2022: 227). Moreover, we present suggestive empirical evidence below (see Analysis 3) that further reduces the likelihood of this reverse causal pathway. Nevertheless, that caveat – together with the previous one – still necessarily qualifies any conclusions that can be drawn from any positive test of Hypotheses 3.1 and 3.2.

Disinformation

Besides elite cues, usually transmitted through mainstream media, citizens may also rely on political information from less conventional sources such as alternative news websites, social media, chat applications, and emails with political content. The content coming from these channels often tends to question official narratives by spreading disinformation and conspiracy theories (Terracino and Matasick Reference Terracino and Matasick2022; van Ham et al. Reference van Ham, Sipma and Fiselier2025). Such messages frequently resonate in polities with low levels of political trust, such as many countries in CEE (Marinov and Popova Reference Marinov and Popova2022).

Exposure to disinformation and beliefs in conspiracy theories likely correlate with pro-Russian attitudes. In the past decade, the Kremlin has extensively used these tools to sway global opinion and destabilise democracies (Yablokov Reference Yablokov2015; Kostelka and Krejcova Reference Kostelka and Krejcova2017; Zilinsky et al. Reference Zilinsky, Theocharis, Pradel, Tulin, De Vreese, Aalberg, Cardenal, Corbu, Esser, Gehle, Halagiera, Hameleers, Hopmann, Koc-Michalska, Matthes, Schemer, Šteˇtka, Strömbäck, Terren and Zoizner2024). Regarding the conflict in Ukraine, disinformation and conspiracy theories have been widespread, promoting pro-Russian narratives and casting doubt on major events such as the downing of MH17 by pro-Russian separatists (Golovchenko et al. Reference Golovchenko, Hartmann and Adler-Nissen2018). We thus hypothesise that the more citizens are exposed to disinformation and the more they tend to believe in conspiracy theories, the less they support Ukraine.

Data & methods

Our main analyses rely on original survey data from five countries (FC dataset). Four (Czechia, Poland, Slovakia, and Romania) lie in CEE, where geographic proximity makes the war particularly salient. Each of these four countries experienced one (Romania) or several (Czechia, Poland, and Slovakia) Soviet occupations in the 20th century. In addition, Russia and its Austrian and Prussian allies deprived Poland of independent statehood between 1795 and 1918. Today, three of these countries (Poland, Slovakia, and Romania) share a border with Ukraine, and the fourth (Czechia) had the largest per-capita intake of Ukrainian refugees as of August 2023 (Plevak Reference Plevák2023). France serves as a comparison case. Its general level of support for Ukraine in earlier surveys was similar to that of other large countries in Western Europe.Footnote ⁸

The survey was programmed in Qualtrics and fielded via computer-assisted online interviews by the company Cint between November 21 and December 11, 2023. For each country, we obtained a quota-based sample defined in terms of sex, age, region, size of municipality, and education.Footnote ⁹ As noted below (see Footnote 26), the distribution of our dependent variables is consistent with probabilistically sampled data from Eurobarometers. The final sample size is 12,221 respondents (approximately 2,450 respondents per country). This research received ethical approval from the European University Institute.

We study two dependent variables measuring support for the war’s belligerents. In Analysis 1.1, the dependent variable operationalises adherence to the Kremlin’s narrative via blame attribution for the war in Ukraine. Respondents rated the responsibility of Ukraine, Russia, and NATO for the ‘open conflict between Ukraine and Russia’ on a 0–10 scale where 0 meant ‘not at all responsible’ and 10 meant ‘totally responsible’.Footnote ¹⁰ We expect that the more respondents adhere to the Kremlin’s narrative, the more they will blame Ukraine (or NATO) and the less they will blame Russia. To facilitate interpretation, we use as the main dependent variable in Analysis 1.1 a simple difference between the scores given to Ukraine and Russia, respectively (ie negative scores indicate a pro-Russian position blaming Ukraine). Alternative operationalisations, intended for robustness checks, employ a variable that compares Russia’s score to NATO’s score, and a Likert scale based on the three scores (for Russia, Ukraine, NATO).Footnote ¹¹ In Analysis 1.2, the dependent variable measures preferred war outcomes based on a closed question with three possible answers besides ‘don’t know’: ‘Ukraine’s victory’, ‘neither Russia’s nor Ukraine’s victory’, and ‘Russia’s victory.’ The two dependent variables strongly correlate, even though the outcome variable can take only three values (Pearson’s r= 0.56).

We analyse the dependent variables through OLS (blame attribution) and multinomial logit (outcome preference) regressions. To focus on within-country variation, all our analyses employ country-fixed effects. Besides substantive independent variables described below, full model specifications include standard socio-demographic controls used in public opinion research on foreign policy (eg Onderco Reference Onderco2017): gender (dummy variables for women and non-binary respondents), age (a linear variable), and education (dummies for vocational, A-level and university education).

To operationalise Hypothesis 1.1, respondents rated on a 0–10 scale how important selected European economic powers (Germany, Spain, Russia, Great Britain, and Italy) were to the operations of their business or employer over the past five years. The variable Business Importance uses the score for Russia (the larger the score, the more significant Russia was for one’s business or employer).Footnote ¹² The operationalisation of Hypothesis 1.2 relies on how difficult it was for respondents’ households to cope with the rise in energy prices since autumn 2021. The variable Energy Crisis ranges from 0 (‘with no difficulty at all’) to 10 (‘with extreme difficulty’).Footnote ¹³

Regarding the ideological explanations, Hypothesis 2.1’s operationalisation uses respondents’ self-positioning on an eleven-point left-right scale (0 = left, 10 = right). To allow for non-linear relationships, we recoded the scale into five dummy variables (Extreme Left, Moderate Left, Centre, Moderate Right, Extreme Right)Footnote ¹⁴ using the centre category as a reference. Hypothesis 2.2 is operationalised through the variable Conservative Authoritarianism,Footnote ¹⁵ which is an additive index that combines a measure of cultural conservatism and a measure of authoritarianism.Footnote ¹⁶ Hypothesis 2.3 corresponds to a variable Pacifism, which is a Likert scale based on two questions about the use of military force and war (see Appendix J for details).Footnote ¹⁷

Concerning partisan alignment, Hypothesis 3.1 on party proximity corresponds to a scale combining respondents’ self-reported probabilities to vote (PTVs) for political parties in a future election (0–10)Footnote ¹⁸ and parties’ positions towards the government of Russia extracted from the Chapel Hill expert survey fielded in Spring 2023 (Hooghe, Marks, Bakker et al. Reference Hooghe, Marks, Bakker, Jolly, Polk, Rovny, Steenbergen and Vachudova2024).Footnote ¹⁹ For each respondent ${_i}$ in a party system with J parties, the variable Pro-Kremlin Party Proximity calculates the average of the parties’ positions weighted by their respective PTVs increased by one (see Equation 1).Footnote ²⁰ The generated variable remains on a 0–10 scale, but, in practice, it varies from 1.43 (the lowest proximity) to 9.22 (the highest proximity). Hypothesis 3.2 is operationalised through the variable Government Support, which is measured on a 0–10 scale where 0 means ‘no support at all’ and 10 ‘total support’ for pro-Ukrainian cabinets.Footnote ²¹ For Slovakia, which is the only country in the FC dataset where the cabinet is more pro-Russian, the variable is reverse-coded (‘total support’ = 0).

(1) $Pro {\hbox{-}} Kremlin\,Party\,Proximit{y_i} = {{\mathop \sum \nolimits_{j = 1}^J {\rm{Party's\;position\;on\;Russi}}{{\rm{a}}_j}\,{\rm{*}}\,\left( {{\rm{PT}}{{\rm{V}}_{ij}} + 1} \right)} \over {\mathop \sum \nolimits_{j = 1}^J \left( {{\rm{PT}}{{\rm{V}}_{ij}} + 1} \right)}}$

We use two variables to test the disinformation-related hypotheses. Altern. News Consumption captures exposure to disinformation (Hypothesis 4.1) by measuring respondents’ reliance on alternative versus conventional sources of political information. Respondents selected their primary political news channels from eight options. We assume that individuals differ in their propensity to consume different news sources. We conducted a confirmatory factor analysis on the resulting binary indicators, which corroborated the presence of a factor positively correlated with traditional media (eg television or mainstream news websites) and negatively correlated with alternative sources (eg alternative media websites, social media, chat applications, and political emails).Footnote ²² We reverse-coded and standardised the extracted factor to vary from 0 (conventional media) to 1 (alternative sources). Finally, the variable Conspiracy Beliefs measures vulnerability to disinformation (Hypothesis 4.2). It is a Likert scale (0–10) based on respondents’ agreement with two common conspiracy claims.Footnote ²³

To compare the explanatory potential of the different factors and assess individual hypotheses, we use model fit statistics and regression coefficients. To make all regression coefficients comparable, we rescaled all quantitative independent variables onto a 0–1 scale.Footnote ²⁴ The regression coefficients below can thus be read as the predictors’ maximal potential effects.

Our analyses aim to identify the correlates of support for Russia’s narrative and determine which of the competing accounts (economic interests, ideology, partisan alignment, and disinformation) is the most strongly associated with respondents’ views on the war. Of course, they cannot establish causality as the identified relationships might be confounded by omitted variables, and causality could flow in the opposite direction. Nonetheless, in many cases, such risks are relatively low, and we did our best to reduce them. Many of the predictors, such as ideological orientations, are long-term and stable traits unlikely to shift because of the war (Miller and Shanks Reference Miller and Shanks1996; Prior Reference Prior2010; Reeskens, Muis, Sieben et al. Reference Reeskens, Muis, Sieben, Vandecasteele, Luijkx and Halman2021). To further attenuate the risk of reverse causality, survey items measuring independent variables preceded any reference to the war,Footnote ²⁵ and respondents could not return to previous screens to revise their answers. In the only question mentioning Russia prior to the war section (the variable Business Importance), Russia appeared among other countries (presented in a randomised order), to avoid disclosing the focus of our study. Furthermore, the survey fieldwork coincided with Israel’s early offensive in Gaza, which likely diverted attention from Ukraine and reduced the risk of priming effects on the items used to operationalise our independent variables.

Results

Figure 1 presents the distribution of blame attribution variables in the FC dataset. In all countries, a plurality blames Russia most, but support for the mainstream narrative is far from universal. The share of respondents fully blaming Russia ranges from 55 per cent (in Poland) to 32 per cent (in Slovakia). Between 17 per cent (in Slovakia) and 9 per cent (in Poland) deem Russia only partially responsible, while up to 14 per cent consider Russia not at all responsible (the value of 0). Blame attribution to Ukraine and NATO shows greater variation, with the mode values (ie ‘none’ or ‘partial’ responsibility) generally selected by 20 per cent of respondents. Worryingly, more Slovaks think Ukraine is fully responsible (19 per cent) than not at all responsible (13 per cent).

Figure 1.

Responsibility for the war: overview.

Note: 0 means ‘not at all responsible’ and 10 means ‘totally responsible.’ Total N for each question is 12,221. Respondents were not offered a ‘don’t know’ option.

Figure 1. Long description

The image contains three sets of bar graphs, each representing different European countries’ perspectives on the responsibility for the war in Ukraine. The first set of bar graphs at the top shows data for Czechia, France, Poland, Romania, and Slovakia regarding their views on Russia’s responsibility. The second set in the middle shows data for the same countries regarding their views on Ukraine’s responsibility. The third set at the bottom shows data for the same countries regarding their views on NATO’s responsibility. Each bar graphs displays the distribution of percentage values, with the x-axis representing the percentage and the y-axis representing the percentage of respondents. The data suggests varying levels of agreement and disagreement among the countries regarding the responsibility for the war.

The subsequent two figures (2 and 3) graphically present our main dependent variables, allowing for a direct comparison of respondents’ sympathies towards Russia and Ukraine. Both confirm respondents’ overall support for Ukraine. Nevertheless, they also reveal a significant presence of ambivalent and even pro-Russian attitudes. Figure 2 shows that although all five publics attribute more responsibility for the war to the Russian side, there are significant cross-country differences. Poland (4.5-point difference on average) and France (3.1) are the most pro-Ukrainian countries. In the other three cases, Russia is blamed for only between 1.8 (Czechia) and 0.9 (Slovakia) points more than Ukraine. Regarding the preferred outcome of the war (Figure 3), respondents in all countries prefer Ukraine’s victory to Russia’s victory. However, up to 20 per cent of Slovaks prefer a Russian victory. Even more surprisingly, in Czechia, Romania, and Slovakia, the most preferred option is neither side’s victory, favoured by nearly half of the respondents. This comes as a surprise in light of these countries’ experience with Soviet occupations and their vulnerability to the war’s potential escalation. The contrast with France, which neighbours neither Russia nor Ukraine and where Russian soldiers have not set foot since 1815, indicates that other factors, such as those studied below, may outweigh historical and geographic influences in shaping attitudes.Footnote ²⁶

Figure 2.

Responsibility for the war: Ukraine vs Russia.

Note: The figure shows average differences in perceptions of Russia’s and Ukraine’s responsibility for the war. It draws on the first two rows from Figure 1. The resulting variable ranges from −10 to +10, but the country averages are all in the 0–5 interval. Positive values indicate greater blame on Russia. 95% confidence intervals.

Figure 2. Long description

A horizontal dot plot compares the perceived responsibility for the war between Ukraine and Russia across five countries: Czechia, France, Poland, Romania, and Slovakia. The x-axis represents the difference in responsibility, ranging from negative values indicating Ukraine is considered more responsible to positive values indicating Russia is considered more responsible. The y-axis lists the countries. Each dot represents the average perceived responsibility difference for that country and is accompanied with a confidence interval. Czechia shows a value of 1.78, France 3.10, Poland 4.54, Romania 1.62, and Slovakia 0.91. All these estimates are statistically significant. The red dashed line at zero marks the neutral point where both countries are seen as equally responsible. All values are approximated.

Figure 3.

Preferred outcome of the war.

Note: 95% confidence intervals. DK means ‘Don’t Know’.

Figure 3. Long description

The bar graph compares the preferred outcomes of the war in five countries: Czechia, France, Poland, Romania, and Slovakia. The x-axis represents the preferred outcome of the war, with categories for Ukraine, Neither, and Russia. The y-axis represents the percentage of respondents. Each country has three vertical bars representing the percentage of respondents who prefer victory for Ukraine, neither, or Russia. Czechia has approximately 40 percentage for Ukraine, 50 percentage for Neither, and 10 percentage for Russia. France has approximately 47 percentage for Ukraine, 43 percentage for Neither, and 10 percentage for Russia. Poland has approximately 70 percentage for Ukraine, 28 percentage for Neither, and 2 percentage for Russia. Romania has approximately 43 percentage for Ukraine, 50 percentage for Neither, and 7 percentage for Russia. Slovakia has approximately 30 percentage for Ukraine, 50 percentage for Neither, and 20 percentage for Russia. The color scheme uses blue for Ukraine, gray for Neither, and red for Russia. All values are approximated.

Altogether, the descriptive evidence underscores the need to better understand attitudes towards the Russo-Ukrainian war. For both dependent variables, public opinion in all five countries favours Ukraine. However, more nuanced positions are surprisingly common and, in three countries, even predominant regarding the preferred outcome of the war.

Analysis 1.1

The first analysis investigates blame attribution for the war. Table 1 displays seven models ranging from a baseline with only country-fixed effects (1), through a series of models testing the different types of factors separately (2–5), to comprehensive models with all independent variables (6) and controls (7).

Table 1.

Responsibility for the war (Ukraine vs. Russia) regressed on predictors

Table 1. Long description

The table presents seven models analyzing blame attribution for the war. The models range from a baseline with only country-fixed effects to comprehensive models with all independent variables and controls. The table includes economic factors, ideology, partisan alignment, disinformation, and controls. Each model shows coefficients and standard errors for variables such as business importance, energy crisis, political ideology, party proximity, news consumption, conspiracy beliefs, and demographic controls. The table aims to investigate the responsibility for the war between Ukraine and Russia.

Note: The DV is coded from −10 (Ukraine is fully responsible for the war) to 10 (Russia is fully responsible for the war). Positive coefficients indicate that respondents attribute more responsibility to Russia. OLS regression. Standard errors in parentheses. Significance levels: ${{\rm{\;}}^{\rm{*}}}{\rm{\!}}p \lt 0.05{,\,^{{\rm{**}}}}{\rm{\!}}p \lt 0.01{,\,^{{\rm{***}}}}{\rm{\!}}p \lt 0.001.$

The separate tests (Models 2–5) support most of our hypotheses. The more respondents deem that Russia was important for their business (H1.1) and the more they suffered during the energy crisis (H1.2), the more they blame Ukraine. The Kremlin’s narrative also finds support among extreme leftists (H2.1), authoritarians (H2.2), respondents close to pro-Kremlin parties (H3.1), consumers of alternative news (H4.1), and believers in conspiracy theories (H4.2). Pacifists (H2.3) are also more likely to blame Russia, although the picture is, as expected, more complex. In the separate test in Model 3, Pacifism has the opposite (ie negative) sign, suggesting that pacifists could be pro-Russian. However, the coefficient becomes positive and statistically significant in the comprehensive model specifications (0.46, $p \lt 0.05$ in Model 6), which indicates the presence of confounding. Additional analyses in Appendix Table A16 confirm that the key confounder is disinformation: when consumption of alternative news and conspiracy beliefs are controlled for, Pacifism takes the hypothesised positive sign. For some respondents, pacifist attitudes reflect a penchant for alternative sources of political news.

There are two chief deviations from our expectations. First, Hypothesis 2.1 receives only partial support: extreme right-wingers blame Russia as much as centrist respondents in Model 3.Footnote ²⁷ In the comprehensive models (6 and 7), the regression coefficient of Extreme Right remains positive. It even becomes statistically significant, indicating that right-wingers are more likely to blame Russia once we control for other factors, including proximity to pro-Kremlin parties and vulnerability to disinformation. This reveals that the relationship between left-right ideology and support for the Kremlin’s narrative is more linear than hypothesised – presumably reflecting the historical affinity of the left with the USSR.

Hypothesis 3.2 also receives little support as the regression coefficient of Incumbent Support is substantively small and statistically insignificant across all models. Appendix analyses (Table A14) demonstrate that the tested relationship is fully nested in the general association between Party Proximity and the dependent variable. When Incumbent Support is tested separately without Pro-Kremlin Party Proximity, its coefficient becomes positive and substantively and statistically significant (2.42, $p \lt 0.001$ ). The null finding in Table 1 thus does not imply that citizens do not align their views with those of political elites, but rather that they pay greater attention to party positions rather than government positions.

The models reveal marked differences in the substantive significance of the observed effects, especially when comparing economic factors to the rest. The regression coefficient of Business Importance (−0.3 in Model 2), which is twice as large as that of Energy Crisis (−0.16), means that the estimated difference between a respondent who considered Russia had been extremely important to their business activity and a respondent who thought the opposite is 0.3 on the 20-point blame-attribution scale. The magnitude of the regression coefficients of Conservative Authoritarianism (Model 3), Pro-Kremlin Party Proximity (Model 4), Conspiracy Beliefs, and Altern. News Consumption (Model 5) are, respectively, 21.3, 45.2, 18.1, and 11.8 times larger (ie −6.41, −13.56, −5.42, −3.53).

Including all four factors jointly (Model 6) and adding controls (Model 7) leaves most substantive results unchanged. Apart from Pacifism, all regression coefficients retain their sign. The most noteworthy difference, besides those commented upon above, is Energy Crisis’s loss of substantive and statistical significance. It suggests that the relationship between one’s economic situation and attitudes towards the war essentially translates political factors such as respondents’ party preference, ideology, and vulnerability and exposure to disinformation. Concerning the controls, pro-Ukrainian attitudes are positively associated with age and level of education, which presumably reflects more experienced and more educated respondents’ geopolitical awareness.

In terms of explained variance, partisan alignment performs the best ( ${R^2} = 19.6{\rm{\% }}$ ), followed by disinformation ( ${R^2} = 19.4{\rm{\% }}$ ) and ideology ( ${R^2} = 15.8{\rm{\% }}$ ). Economic factors explain significantly less ( ${R^2} = 9.9{\rm{\% }}$ ), adding only 3.7 percentage points to the model fit of the baseline fixed-effects specification ( ${R^2} = 6.2{\rm{\% }}$ ). The four types of factors jointly account for a solid 30.2 per cent of variance in the dependent variable (Model 6) and 31.2 per cent when the controls are included (Model 7). These proportions demonstrate that individual attitudes towards the war are structured and vary significantly within countries as a function of respondents’ characteristics. The differences in ${R^2}$ between Models 1 and 7 mean that respondents with similar characteristics who come from different countries are closer to each other in their attitudes towards the war rather than to their fellow compatriots with different attributes.

Analysing each country separately (Appendix Table A8), Pro-Kremlin Party Proximity, Conspiracy Beliefs, Altern. News Consumption and Conservative Authoritarianism consistently exhibit the hypothesised (negative) sign and are usually statistically and substantively significant. Pro-Kremlin Party Proximity is the strongest predictor in all countries except Romania, likely reflecting Romanians’ low trust in parties (Tatar Reference Tatar2022). The most interesting variation concerns left-right ideology. Its full, unmediated effects (see Appendix Table A9) show that the left is more pro-Russian and the right more pro-Ukrainian in Czechia, Slovakia, and Romania, presumably because of the historical ties between these countries’ left and the USSR (especially before 1968). In Poland, the moderate left is pro-Ukrainian, and no ideological group is distinctly pro-Russian. This probably reflects Poland’s painful experience with the Tsarist and later Soviet domination and externally imposed communism (Kitschelt, Mansfeldova, Markowski et al. Reference Kitschelt, Mansfeldova, Markowski and Toka1999). In France, the right is more pro-Russian than the left. Overall, these ideological patterns appear related to cross-country differences in the association between the economic and cultural attitudinal dimensions (Rovny Reference Rovny2014; Kostelka and Rovny Reference Kostelka and Rovny2019).

The Appendix shows that our main substantive results hold under a variety of alternative model specifications.Footnote ²⁸ When we compare the blame attributed to NATO (instead of Ukraine) and Russia (Table A10), the results remain very similar. Conversely, when we compare NATO and Ukraine (Table A11) in a sort of a placebo test, none of the independent variables is statistically significant in any model specification, and the full model explains only 1.3 per cent of variance. This suggests that, regarding the start of the conflict, public opinion does not exhibit differences in perceptions of NATO’s and Ukraine’s roles. Indeed, using the combined (Ukraine-Russia-NATO) scale as the dependent variable increases the share of explained variance by Model 7 to 36.2 per cent (Table A12).

Analysis 1.2

The multinomial logit regressions of preferred war outcomes closely mirror the results from Analysis 1.1.Footnote ²⁹ As the raw regression coefficients may be misleading (Long and Freese Reference Long and Freese2014; Wulff Reference Wulff2015), they are presented only in the Appendix (Tables A20 and A21), and we turn to marginal effects for interpretation. Figure 4 presents average marginal effects based on the full multinomial logit model that includes exactly the same independent variables as Analysis 1.1.Footnote ³⁰ They indicate the change in the probability of preferring Ukraine’s victory over any other outcome (Russia’s victory or neither’s victory). The signs of the predictors and the relative differences in substantive and statistical significance are essentially the same as in Table 1. Respondents who want Russia to win are close to pro-Kremlin parties, believe in conspiracy theories, are exposed to alternative media, and hold authoritarian attitudes. These four variables are associated with decreases in the probability of supporting Ukraine by up to 0.74, 0.25, 0.21, and 0.17, respectively. These four variables are by far the strongest correlates of the dependent variable, as the absolute values of the other coefficients never exceed 0.11. Like in Analysis 1.1, economic factors are less related to the dependent variable than partisan alignment, exposure to disinformation, and ideology.

Figure 4.

Correlates of support for Ukraine’s victory.

Note: Average Marginal effects from the multinomial logit analysis (Model 6 in Tables 2 and 3). The coefficients indicate changes in the probability of preferring Ukraine’s victory (over any other outcome) resulting from one-unit changes in the independent variables (holding other variables at their observed values). 95% confidence intervals.

Figure 4. Long description

A vertical dot plot displays the correlates of support for Ukraine’s victory. The x-axis ranges from -1 to 1, representing the correlation values, while the y-axis lists various factors such as Business Importance (Russia), Energy Crisis (difficult), Extreme Left, Moderate Left, Moderate Right, Extreme Right, Conservative Authoritarianism, Pacifism, Pro-Kremlin Party Proximity, Incumbent Support, Altern. News Consumption, and Conspiracy Beliefs. Each dot represents the average marginal effect from the multinomial logit analysis. Notable patterns include high positive values for Business Importance (Russia) and Energy Crisis (difficult), and high negative values for Pro-Kremlin Party Proximity and Conspiracy Beliefs. All values are approximated.

Figure 5 uses predicted probabilities to show that individuals scoring lowest on the four main correlates (0) are highly likely to express a desire for Ukraine’s victory (probability of 0.94) as opposed to Russia’s victory (0). Conversely, those scoring highest (1) show the reverse pattern – strong support for Russia’s victory (0.83) and none for Ukraine’s (0).

Figure 5.

Predicted probabilities of the preferred outcomes.

Note: Predicted probabilities from the multinomial logit analyses (Model 6 in Tables 2 and 3) for individuals scoring respectively the maximum and minimum values on the four main correlates of the preferred outcomes (Pro-Kremlin Party Proximity, Conspiracy Beliefs, Altern. News Consumption, Conservative Authoritarianism). Other variables are left at observed values. 95% confidence intervals.

Figure 5. Long description

A horizontal dot plot compares predicted probabilities of preferred outcomes for three categories: Ukraine, neither, and Russia. The x-axis represents probabilities ranging from 0 to 1, while the y-axis lists the categories. Each category contains two dots: one at the minimum value (0) and one at the maximum value (1) on the four main correlates. The Ukraine category shows a dot near 1 at the maximum value and a dot near 0 at the minimum value. The neither category has dots near 0.2 at the minimum value and near 0.2 at the maximum value. The Russia category displays a dot near 0.8 at the maximum value and a dot near 0 at the minimum value. The plot highlights the distribution of predicted probabilities across the three categories, indicating varying levels of support.

Data and methods

To test the generalizability of our main finding – the role of party proximity – we replicate Analysis 1.1 using data from the Solidarity in Europe (SiE) dataset (Kriesi, Stolle and Cicchi Reference Kriesi, Stolle and Cicchi2023). The survey was fielded by the company YouGov in seventeen European countries,Footnote ³¹ four of which also appear in the FC dataset, between March and April 2023. It includes a question similar to that used for the dependent variable in Analysis 1.1: ‘From what you’ve read and heard, who do you think is responsible for the current situation in Ukraine?’. The possible answers were from 1 (‘Entirely NATO’), 2 (‘More NATO than Russia’), 3 (‘NATO and Russia equally’), 4 (‘More Russia than NATO’), and 5 (‘Entirely Russia). To make the regression coefficients comparable to those in Analysis 1, we rescaled the resulting dependent variable to range from –10 (‘Entirely NATO’) to 10 (‘Entirely Russia’).Footnote ³²

Similarly to Analysis 1.1, we conduct OLS regressions with country-fixed effects. The main predictor, operationalising Hypothesis 3.1, gives the position towards Russia’s government – extracted from CHES – of the political party to which the respondent feels close.Footnote ³³ Unfortunately, the survey does not allow us to operationalise most other hypotheses. The only exception is Hypothesis 2.1 on ideology, whose operationalisation consists of the same five dummy variables as in the previous analyses (Extreme Left, Moderate Left, Centre, Moderate Right, Extreme Right).Footnote ³⁴ Control variables exactly match Analysis 1, except that age information is only available in categorical form.

Results

Figure 6 plots average blame attribution by country. The picture closely mirrors Figure 2 from the FC dataset. In all countries but Bulgaria, respondents are more likely to blame Russia than NATO. Nonetheless, cross-country differences are substantial. For the four countries included in both datasets, the relative ranking is exactly the same as in Figure 2: Poland is the most pro-Western, followed by France, Romania, and Slovakia. On the whole, however, Eastern countries (in the lower part of the graph) appear more pro-Russian than Western countries. Again, this suggests that painful historical experiences can be overshadowed in citizens’ attitudes by other factors, which may include partisan alignment.

Figure 6.

Responsibility for the current situation in Ukraine: NATO versus Russia.

Note: SiE dataset. The figure shows the country averages on the dependent variable which ranges from −10 to +10. Positive values indicate that respondents attribute stronger responsibility to Russia. 95% confidence intervals. Total N = 24,261.

Figure 6. Long description

A horizontal dot plot displays the perceived responsibility for the current situation in Ukraine, comparing NATO and Russia. The y-axis lists 20 countries, while the x-axis ranges from -10 to 10, with negative values indicating NATO responsibility and positive values indicating Russia responsibility. Each country has a dot representing its perceived responsibility, with some countries having multiple dots indicating different data points. Notable values include Denmark at 7.62, Finland at 6.72, and Germany at 0.22. The plot shows a clear division, with many countries leaning towards Russia being more responsible. All values are approximated.

The regression results presented in Table 2 remarkably resemble those from Analysis 1, fully supporting Hypothesis 3.1. The regression coefficients of Pro-Kremlin Party Proximity are consistently negative, substantive, and statistically significant ( $p \lt 0.001$ ). Interestingly, the magnitude of the regression coefficient in the full Model 4 (−6.56) aligns closely with that observed in Model 6 in Table 1 (−8.50) despite using entirely different data. When the analyses are conducted separately by subregion (Models 5 and 6), the regression coefficients remain similar, although the magnitude is slightly stronger in CEE (Model 6), likely reflecting the conflict’s heightened salience because of geographic proximity. The results for ideology also echo Table 1. While extreme leftists are more pro-Russian and moderate-rights are more pro-Western, extreme-rights do not significantly differ from centrists (the reference category).Footnote ³⁵ Overall, ideology seems to matter much less than party proximity. Finally, the controls exhibit similar patterns to those in Table 1: female, younger, and less educated respondents report more pro-Russian attitudes.

Table 2.

Responsibility for the current situation in Ukraine

Table 2. Long description

The table presents regression models analyzing responsibility for the current situation in Ukraine. The table has six columns labeled (1) to (6), representing different regression models. The rows include variables such as partisan alignment, pro-Kremlin party proximity, ideology, age groups, gender, education levels, and a constant. Each cell contains regression coefficients with standard errors in parentheses and significance levels indicated by asterisks. Notable trends include consistently negative and significant coefficients for pro-Kremlin party proximity across all models, with varying magnitudes. Ideological alignment shows that extreme leftists are more pro-Russian, moderate rights are more pro-Western, and extreme rights do not significantly differ from centrists. Controls indicate that female, younger, and less educated respondents report more pro-Russian attitudes.

SiE Dataset. The DV is coded from −10 (NATO is entirely responsible) to 10 (Russia is entirely responsible). Models 5 and 6 include only Western European (Denmark, Finland, France, Germany, Greece, Italy, Netherlands, Spain, Sweden, UK) and CEE (Bulgaria, Croatia, Hungary, Lithuania, Poland, Romania, and Slovakia) countries respectively. Positive coefficients indicate greater blame attribution to Russia. OLS regression. Standard errors in parentheses. N (16,572) is smaller than in Figure 6 due to missing values on the independent variables. Significance levels: ${{\rm{\;}}^{\rm{*}}}{\rm{\!}}p \lt 0.05{,\,^{{\rm{**}}}}{\rm{\!}}p \lt 0.01{,\,^{{\rm{***}}}}\!\!\!\!{\rm{\;\;}}p \lt 0.001$ .