Public support for redistribution across borders: Who deserves what?

The concept of deservingness is central to the literature on public preferences for redistribution within and across countries. In welfare state research, scholars have consistently linked lower support for welfare spending to perceptions of beneficiaries as responsible for their own misfortune – most commonly, by being seen as lazy rather than unlucky (Petersen Reference Petersen2012). This association has been documented within different cultural settings, including African Americans in the United States (US) (Gilens Reference Gilens2009: 161–162), immigrants and asylum-seekers in Western Europe (van Oorschot Reference van Oorschot2006), and the unemployed and the working poor in former communist regimes (Gugushvili et al. Reference Gugushvili, Lukac and van Oorschot2021).

Typically, operationalisations of deservingness encompass multiple dimensions. For example, the so-called CARIN model outlines five criteria according to which people judge the deservingness of others: (1) control (are potential recipients to blame for their situation?), (2) attitude (do they show gratitude for receiving social benefits?), (3) reciprocity (have they contributed to the welfare state in the past or are they likely to contribute in the future?), (4) identity (are they part of the same cultural in-group as contributors?), and (5) need (are they genuinely in need of welfare support?) (Meuleman et al. Reference Meuleman, Roosma and Abts2020; van Oorschot Reference van Oorschot2000). More recently, scholars have proposed grouping these criteria into two broader dimensions: a situational dimension, capturing the level of need and responsibility for one’s adverse circumstances, and a relational dimension, comprising elements of (self-)identification and reciprocity within a shared political community (Heermann et al. Reference Heermann, Koos and Leuffen2023). Depending on their research aims, studies tend to emphasise one or several of these criteria in national settings or across borders.

In the context of international redistribution, much of the literature focuses on the EU, where scholars often use the term transnational solidarity to capture public support for redistribution among EU member states (Kriesi et al. Reference Kriesi, Moise and Oana2025; Reinl Reference Reinl2022). Regarding situational factors, research has repeatedly shown that perceptions of need and control are key drivers of public preferences for EU redistribution. First, citizens are more willing to redistribute when other countries are perceived as poorer relative to their own or as economically vulnerable at a certain moment in time (Afonso and Negash Reference Afonso and Negash2024; Katsanidou et al. Reference Katsanidou, Reinl and Eder2022; Reinl et al. Reference Reinl, Nicoli and Kuhn2023). Second, blame attribution strongly conditions perceptions of need, as citizens are more supportive of transnational solidarity when countries‘ economic problems are viewed as exogenous or linked to EU membership rather than as self-inflicted (Baute and Pellegata Reference Baute and Pellegata2023; Clasen Reference Clasen2024). This logic helps explain why support for EU solidarity was higher during the Covid-19 pandemic compared to the euro crisis (Ferrara et al. Reference Ferrara, Schelkle and Truchlewski2023; Katsanidou et al. Reference Katsanidou, Reinl and Eder2022; Reinl et al. Reference Reinl, Nicoli and Kuhn2023). Moreover, the design of solidarity instruments – for example, whether they are conditional or framed as mutual investments – can further influence public support for redistribution (Baute et al. Reference Baute, Nicoli and Vandenbroucke2022; Bechtel et al. Reference Bechtel, Hainmueller and Margalit2017; Bremer et al. Reference Bremer, Kuhn, Meijers and Nicoli2024).

At the same time, relational factors are also important determinants of transnational solidarity. Regarding identity, studies have found that individuals are more willing to support EU redistribution when they feel geographically or linguistically closer to other countries (Afonso and Negash Reference Afonso and Negash2024), when they show strong attachment to Europe and support for European integration (Kuhn and Kamm Reference Kuhn and Kamm2019; Verhaegen Reference Verhaegen2018), or when they endorse cosmopolitan values more broadly (Kleider and Stoeckel Reference Kleider and Stoeckel2019). Other studies highlighted the role of perceiving the EU as a bounded community that differentiates insiders from outsiders (Oana and Truchlewski Reference Oana and Truchlewski2024) or as a political community united by shared norms (Heermann et al. Reference Heermann, Koos and Leuffen2023). Similarly, regarding reciprocity, scholars have found that redistribution is more acceptable when recipients are depicted as ‘ready to do their bit, if other people – when they are capable – are also ready to do their bit’ (Baute et al. Reference Baute, Nicoli and Vandenbroucke2022: 722). Elsewhere, reciprocity has been operationalised through close economic ties – such as trade – which strengthen the sense of mutual dependence (Afonso and Negash Reference Afonso and Negash2024) or in terms of past contributions to EU solidarity schemes (Heermann et al. Reference Heermann, Koos and Leuffen2023). Depending on the study, relational factors function either as moderators or as primary drivers of public support for EU solidarity.

While stereotypical perceptions are not an explicit focus of this literature, they are implicitly present in portrayals of [un-]deserving individuals. In the US, for example, women of colour stereotyped as ‘welfare queens’ are depicted as exploiting honest taxpayers through their dependence on state support (Gilman Reference Gilman2014). Because the concept of deservingness emphasises a group’s responsibility for its own economic hardship, such portrayals often centre on policy outcomes or specific behaviours. However, stereotypes are generalised perceptions about groups that can exist independently of knowledge about particular policies or behaviours. Returning to the example above, stereotypes of African Americans as lazy are rooted in the legacy of slavery (Gilens Reference Gilens2009: 154) and thus predate political debates about the welfare state. From this perspective, the process of stereotyping precedes – and often informs – perceptions of deservingness. To understand this process more fully, we turn to the social psychology literature and highlight the long-standing presence of national stereotypes in Europe.

The role of stereotypes in public support for redistribution

In social psychology, the study of stereotypes has a long tradition of over a century (Nelson Reference Nelson2016). Cognitively, stereotypes have been linked to automaticity and the unconscious mind (Dijksterhuis Reference Dijksterhuis, Fiske, Gilbert and Lindzey2010): by allowing the rapid categorisation of people and events, stereotypes function as ‘energy-saving devices’ that help systematise and simplify information (Macrae et al. Reference Macrae, Milne and Bodenhausen1994). While stereotypes are often associated with prejudice, the concept is broader and more neutral; for instance, one may be aware of a negative stereotype without endorsing it (Devine Reference Devine1989). Beyond facilitating information processing about other groups, stereotypes also serve a justificatory function, enabling individuals to rationalise why they treat some groups in specific ways (Crandall et al. Reference Crandall, Bahns, Warner and Schaller2011). Once the element of justification appears, stereotypes have shifted from the unconscious to the conscious mind – as explanations of existing prejudice.

The difference between the unconscious and the conscious use of stereotypes is crucial for understanding their connection to deservingness and public preferences for redistribution. By definition, the notion of deservingness assumes a conscious narrative of blame attribution: ‘People in group X are at fault for their adverse situation, and therefore are not entitled to welfare benefits’. Transposed to debates about redistribution in the EU, this narrative was particularly visible during the euro crisis – even at the highest level of EU institutions. A striking example comes from the Eurogroup, a Council body responsible for making key decisions on financial assistance. In 2017, its then-president Jeroen Dijsselbloem – the Dutch Finance Minister – remarked that Southern Europeans ‘spend all the[-ir] money on drinks and women and then ask for help’ from the EU (Khan and McClean Reference Khan and McClean2017). This narrative of undeservingness alluded to accusations of financial mismanagement by the Greek government prior to the crisis (Vasilaki Reference Vasilaki2018) and illustrated the conscious use of stereotypes about fiscal irresponsibility to justify strict conditionality in financial assistance.

By contrast, unconscious stereotypes tap into essentialist and identity-based logics that are often independent of any knowledge of past economic or policy-relevant behaviour. In Europe, national stereotypes have circulated for centuries, long before the euro crisis. On a continent comprising a mosaic of nation states, stereotypes developed as people started to associate ethnic or entire national groups with certain attributes, which could be either positive or negative (Leerssen Reference Leerssen2018). Present in art, literature, and the media, national stereotypes were often satirical, focusing on character traits that made other groups appear dysfunctional in some way (Mamadouh Reference Mamadouh2017). Common examples include the frivolous French, the humourless Germans, the tax-dodging Italians, the heavy-drinking Poles, and so forth (Fenn, Reference Fenn2012; Hidasi, Reference Hidasi1999). While some stereotypes have evolved over time, many of them have proven remarkably resilient (Barkhoff and Leerssen Reference Barkhoff and Leerssen2021; Connelly Reference Connelly2014). In the context of public preference for redistribution, long-standing unconscious stereotypes are in theory much more problematic owing to their resistance to change, disconnected from perceptions of specific behaviours at a given moment in time.

Theoretically, the role of unconscious stereotypes can be understood with the help of Social Identity Theory (SIT) (Brown Reference Brown2000; Harwood Reference Harwood2020) – which has become an umbrella term for a set of theories on intergroup behaviour. The original proponent, Henri Tajfel, defined social identity as ‘that part of an individual’s self-concept which derives from his knowledge of membership in a social group (or groups) together with the value and emotional significance attached to that group membership’ (Tajfel Reference Tajfel1974: 255). According to SIT, the tendency to categorise people into groups is a natural cognitive process based on our inclination to group things together (Oakes Reference Oakes2003). In doing so, we also categorise ourselves as belonging to some of those groups – a process known as self-identification (Harwood Reference Harwood2020: 11). Once we identify with a group (the so-called ‘in-group’), we look for ways to derive positive emotions from that membership. In fact, the need for a positive self-concept or self-esteem is a basic assumption of SIT (Abrams and Hogg Reference Abrams and Hogg1988). One way to do this is by perceiving our in-group more favourably than other groups (which become ‘out-groups’). This desire for positive distinctiveness within our in-group helps explain why people develop negative stereotypes toward out-groups, which can lead to prejudice and, eventually, discrimination (Turner and Reynolds Reference Turner and Reynolds2003).

Moreover, SIT is concerned with the psychological consequences of a group’s status on its members and in relation to other groups. In this context, status is not an objective resource at one’s disposal but a subjective measure of how groups evaluate themselves in comparison to others (Tajfel and Turner Reference Tajfel, Turner, Hatch and Schultz2004: 286). Two dimensions are key to understanding status: (1) how people perceive the location of their own group within a social structure (high vs low status) and (2) the perceived nature of that social structure (legitimate vs illegitimate, stable vs unstable) (Turner and Reynolds Reference Turner and Reynolds2003: 134). While one might intuitively expect high-status members to derive positive feelings from in-group favouritism, empirical studies have shown that in some instances, low-status members display out-group favouritism because they have internalised the existing social structure as legitimate and stable (Brown Reference Brown2000; Tajfel Reference Tajfel1974). In the context of public debates on redistribution, the literature on class stereotypes is particularly useful because it defines status as the possession of relevant resources such as wealth, income, educational level, titles, or job prestige (Fiske Reference Fiske2017: 6).

Transposed to the EU, SIT helps us identify potential dynamics of national stereotypes at play against the background of increasing demands for redistribution in European integration. We describe the mechanism of this logic in the next section.

Research design

To test our hypotheses, we conducted an observational survey in four countries capturing regional variation in the EU: Germany (Western Europe), Italy (Southern Europe), Romania (Eastern Europe), and Sweden (Northern Europe). The element of geography was seen as relevant for EU policy positions more generally: for instance, research on the Council has shown that national governments often form coalitions along regional lines due to similar interests across issues (Naurin and Lindahl Reference Naurin, Lindahl, Naurin and Wallace2008). Furthermore, the debate about redistribution was expected to be more salient in euro-area countries (Germany and Italy) as opposed to non-euro-area countries (Sweden and Romania). An additional factor referred to the status of a Member State as a beneficiary of the EU budget (Romania) as opposed to a net contributor (Germany, Italy, and Sweden). Among net contributors, Italy is a special case owing to its consistent high levels of public debt and reliance on quantitative easing by the European Central Bank (ECB 2021), which makes it perceived as a ‘debtor’ country despite never having received financial assistance formally (European Stability Mechanism 2019). As a result, we expected Italians and Romanians to be, on average, more in favour of EU redistribution than Germans and Swedes. Below we describe our survey design and measurement.

Selecting stereotypes

The first goal of the survey was to identify which stereotypes were prevalent among respondents from different countries at a given moment in time (July–August 2024). Based on the logic above, we used the SeeGULL dataset (Jha et al. Reference Jha, Davani, Reddy, Dave, Prabhakaran and Dev2023) to compile a list of stereotypes relevant for EU policy-making (see Table 1). The higher number of stereotypes about Southern and Western Europe reflects their proportion in the SeeGULL dataset, where – as can be expected – there are also more stereotypes about larger member states. For instance, Germans alone are associated with 43 of the 237 EU-specific stereotypes identified in the dataset. By testing only 25 stereotypes in our survey, we wanted to confirm the patterns found in the SeeGULL dataset, and more generally, we aimed to explore who had stereotypes about whom and how much variation was present in the four samples (Sweden, Romania, Italy, and Germany).

Table 1. Stereotypes retrieved from the SeeGULL dataset

Stereotypes retrieved from the SeeGULL dataset (Jha et al. Reference Jha, Davani, Reddy, Dave, Prabhakaran and Dev2023).

a) recognisability = number of human annotators from Europe who recognised the stereotype.

b) offensiveness = mean of scores given by human annotators who categorised the attribute as offensive.

The categories ‘recognisability’ and ‘offensiveness score’ were taken from the SeeGULL dataset, whereas the column ‘sentiment’ was added for our purposes to capture whether stereotypes were positive or negative (in inverse correlation with the offensiveness score).

In general, when it comes to stereotypes in circulation among citizens across the four countries in our sample, we expected to corroborate the results of the SeeGULL dataset – finding predominantly positive stereotypes about Northern Europeans, predominantly negative stereotypes about Eastern and Southern Europeans, and mixed stereotypes about Western Europeans.

Who holds which stereotypes?

We first present descriptive results regarding the distribution of stereotypes in our sample. Figure 1 shows how often respondents named their own and other countries in connection with all, positive, and negative stereotypes. On average, each country is named for about one-third of the stereotypes presented. One pattern that emerges is that German and Swedish respondents are more likely to identify their own country in connection with positive stereotypes and less likely to identify their own country in connection with negative stereotypes. Italians and Romanians, by contrast, name their own countries slightly more often for both positive and negative stereotypes. This does not necessarily mean that they think less of their countries, but it reflects a recognition that Italy and Romania are more often linked to negative stereotypes.

Figure 1. Histograms of the frequency with which respondents name their own vs. other countries in relation to different stereotypes.

Figure 2 illustrates the distribution of some important economic stereotypes. It shows which countries were named the most and least often as trustworthy, hardworking, corrupt, and lazy. What we find generally aligns with the SeeGULL attribution of stereotypes to countries. Swedes are most often named as trustworthy, and Romanians least often – except in Romania, where Germany is named trustworthy most and Italy least. On hardworking, respondents generally name their own country most often – except Italians, who point to Germans. Greeks receive the fewest mentions as hardworking in the pooled sample, consistent with dynamics observed during the Euro crisis (Pew, 2013).

Figure 2. Most and least named countries for key stereotypes.

For corruption, Romanians are named most frequently, followed by Italians. Both name their own countries as corrupt, while Nordic countries are generally least associated with corruption. Romanians are also most frequently associated with laziness. Again reflecting the stereotypes circulating during the Eurozone crisis, Greeks are also often labelled as lazy, while Italians name themselves most frequently in connection with laziness. Nordic countries and Germany are named ’lazy’ the least frequently.

Stereotypes and support for economic redistribution

Next, we examine how stereotypes are associated with preferences for redistribution, taking each measure of such preferences in turn. We start with the most general dependent variable, which measures participants’ agreement with the statement that ‘The EU should promote solidarity between its member states’ on a four-point Likert scale. Figure 3 shows the regression coefficients for our aggregate stereotype measures based on regressions with control variables; the full regression table can be found in the online Appendix. We find that positive economic stereotypes have strong associations with support for solidarity. On average, naming one additional country (other than one’s own) per positive economic stereotype is associated with a 0.069-point decrease in support for solidarity on a four-point scale. While this may seem small, the numbers imply that holding two more positive stereotypes about other EU countries has roughly the same effect as moving from low to high trust in EU institutions – which is equivalent to increasing support for solidarity by 9% of a standard deviation. Another way to contextualise these associations is to compare them to baseline differences in country samples. Country-fixed effects reveal that Romanian respondents show the highest support for EU solidarity, followed by Italy, with Swedes and Germans showing lower support.Footnote ⁷ Our regressions predict that a German respondent who names an average of three positive economic stereotypes about other countries will, on average, support EU solidarity at the same level as an Italian who names no positive economic stereotypes about others.

Figure 3. Coefficients of aggregate stereotype measures, by country, DV = Solidarity between EU countries.

Negative economic stereotypes are, as expected, significantly and negatively associated with preferences for solidarity. Naming one additional country (excluding one’s own) for a negative economic stereotype is associated, on average, with a 0.052-point increase in solidarity support, or 7% of a standard deviation. Again, the predicted difference between those who hold no stereotypes and those who hold many negative stereotypes is larger than the baseline difference between country samples. Our regressions predict that an Italian respondent who names an average of three negative economic stereotypes about other countries will, on average, support EU solidarity less than a German who names no negative economic stereotypes. The measure that combines all stereotypes – positive and negative – also has a positive and statistically significant association with support for solidarity, suggesting that positive stereotypes seem to matter more than negative stereotypes in explaining support for solidarity. Note that, as mentioned above, these estimates are likely lower bounds of the magnitude of associations between stereotypes and attitudes, because measurement error using our proxy should bias our results toward zero.

Results for control variables, reported in the online Appendix, are all in the expected directions. We find that having a stronger European identity is strongly positively associated with support for EU solidarity, while low trust in EU institutions is strongly negatively associated with support for solidarity. We also find that, compared to the baseline of not voting, voting for parties on the left or centre in the European Parliament election is strongly positively associated with support for more solidarity, while voting for far-right parties is strongly negatively associated with support for more solidarity.

As shown in the country-specific panels of Figure 3, there is some heterogeneity between country samples. For Sweden, both negative and positive stereotypes are significantly associated with support for solidarity in the expected directions: among Swedes, negative stereotypes are significantly associated with less support, while positive stereotypes are associated with more support for general EU solidarity. For Germans, positive stereotypes are significantly associated with support for solidarity, while the coefficient for negative stereotypes is negative and almost, albeit not quite, statistically significant. In contrast, in our Italian and Romanian samples, only positive economic stereotypes are statistically significant, with coefficients for negative stereotype measures close to zero. One possible explanation for these differences is that German and Swedish respondents expect their countries to be net contributors to EU solidarity. They might be worried about both the positive and negative traits of recipient countries that could influence perceived deservingness and how well the provision of help, financial or otherwise, is used by recipients. Conversely, Romanian and Italian participants might expect to receive help as part of solidarity, be it in the form of economic help or – especially in the case of Italy – other types of help, such as the redistribution of refugees. Consequently, they might be more concerned with how generous, well-meaning and honest the providers of the help are.

One way to further probe the origin of these aggregate effects is to look at stereotypes regarding specific countries. In the online Appendix (Figures B.7 through B.21), we present replicas of Figure 3, where we calculate the association between positive and negative stereotypes about specific countries and attitudes toward solidarity. Aggregated across all survey participants, we find little clear-cut differentiation between countries by economic wealth: negative and positive stereotypes about Germany and France have similar coefficient sizes as negative and positive stereotypes about Romania and Poland. However, we observe some interesting heterogeneity when focusing on respondents from specific countries. While most results are consistent with our hypotheses, one unusual finding stands out. In the Italian sample, negative stereotypes about countries affected by the Eurozone crisis – Greece and Italy itself – are positively associated with support for solidarity, whereas negative stereotypes about the non-Eurozone country Poland are negatively associated with support for solidarity. This suggests a potentially complex relationship between stereotypes and attitudes towards EU solidarity among Italian respondents. Periods of heightened political salience, such as the euro crisis, may have affected both attitudes toward intra-EU solidarity and stereotypes in a way that our control variables cannot easily capture. Specifically, those Italians who had a particularly negative experience with the euro crisis might have both increased their demand for within-EU solidarity and updated their knowledge about stereotypes. After their experience of the euro crisis, these participants might now be more likely to recognise the negative stereotypes about Italy (and Greece) that circulated during the crisis. Beyond country-fixed effects, we do not have control variables that allow us to capture the potential omitted variable of negative euro crisis experiences.

Table 2 lists the individual stereotypes by sample whose association with our solidarity outcome variable is statistically significant (at the 10% level or beyond). Appendix Figure B.2 shows the underlying regression coefficients. Most of our findings on individual stereotypes go in the expected direction. When pooled across participants from all countries, the attributes trustworthy, rational, passionate, orderly, honest, hardworking, generous, and efficient are all positively associated with support for solidarity in the EU. In contrast, the stereotypes thief, lazy, greedy, dishonest, and corrupt are associated with less support for solidarity in the EU. In line with the heterogeneity found above, we also find that both positive and negative stereotypes are statistically significant in the German and Swedish samples, while we find fewer negative associations between negative stereotypes and support for solidarity in the Italian and Romanian samples. While the bulk of these findings align with our hypotheses, there are a few surprising results. For example, arrogant and argumentative are positively associated with support for solidarity across all respondents.

Table 2. Negatively and positively associated stereotypes by sample, DV solidarity

Type of stereotype: [black color] both directions, [red] negative, [blue] positive.

Another puzzling finding is that in the Italian sample, naming more countries as vindictive is associated with higher support for solidarity. One explanation for this may be linguistic: the Italian translation of vindictive (‘vendicativo’) is uncommon in everyday use. Another possible explanation is substantive, pointing to an important omitted variable of ‘euro crisis experience’ – as discussed above. Italians who felt particularly disadvantaged by the crisis may both call for more solidarity and ascribe more negative stereotypes to countries perceived as having treated Italy poorly during the crisis.

Of course, solidarity could be interpreted as economic solidarity or as solidarity in other areas, such as defence or migration. We therefore turn next to two outcome variables that measure preferences for economic redistribution specifically. First, Figure 4 presents the regression coefficients of our three aggregate stereotype measures, where the dependent variable captures respondents’ support for lowering economic inequality between EU member states. The figure is based on weighted OLS regressions with covariates; the corresponding regression table is in the online Appendix.

Figure 4. Coefficients of aggregate stereotype measures, by country, DV = Reduce inequality between countries.

Across our pooled sample of respondents, both negative and positive stereotypes are significantly associated with support for lowering economic inequality between EU member states. As expected, the number of countries (excluding one’s own) linked to negative economic stereotypes is negatively associated with support for lowering economic inequality. Naming, on average, one additional country when asked about negative economic stereotypes is associated with a 0.070-point decrease in support on a four-point Likert scale. This coefficient is statistically significant. Conversely, positive economic stereotypes are positively associated with support for lowering income inequality, with one additional other country named corresponding to a 0.05-point increase, also statistically significant. We find no significant effect for the overall number of countries named across all stereotypes, likely because positive and negative stereotypes cancel each other out in the aggregate measure of all stereotypes.

As before, results for control variables show expected patterns. A stronger European identity, more left-wing voting, and higher trust in the EU all correlate with greater support for lowering inequality between EU countries. Among country samples, Romanians are most supportive of reducing inequality, on average, followed by Italians, Germans, and lastly the Swedes (see Figure C.1 in the online Appendix).

Country-specific results again reveal substantial heterogeneity. As Figure 4 illustrates, the aggregate effects are driven largely by the German sample, where both positive and negative stereotypes have statistically significant associations in the expected directions. For Romania, only negative economic stereotypes are significantly associated with less support for redistribution at the 10% level of significance, but positive economic stereotypes have no significant effect. For Italy, only positive stereotypes are significantly associated with support for lowering inequality. For Sweden, we find no strong associations, despite the fact that Swedish respondents report a similar number of stereotypes as others. This likely reflects Sweden’s non-Eurozone status, as Swedes may simply not think that the financial burden of lowering inequality is likely to fall on them, so their stereotypes about other EU nationalities are less relevant. By contrast, for the general measure of solidarity discussed above, Swedish respondents likely thought that this affects their country directly, as Sweden is contributing to defence (having just joined the North Atlantic Treaty Organization (NATO) and has taken many refugees in the last decade.

As for our solidarity measure, we present equivalent results regarding stereotypes about specific countries in the online Appendix (Figures C.7 through C.21). In line with our expectations, the aggregate and German sample findings are driven by positive and negative stereotypes about presumed net recipient countries: Italy, Greece, Romania, and Poland. By contrast, in most cases, stereotypes about Germany, Finland, or Sweden show no significant association with attitudes toward lowering economic inequality in the EU.

Finally, we examine how individual stereotypes relate to support for lowering inequality between EU countries, as shown in Table 3. In the pooled sample, the stereotypes that are positively and statistically significantly associated with support for lowering inequality are unsurprising, including trustworthy, rational, orderly, honest, hardworking, generous, and efficient. Similarly, the list of stereotypes that are negatively and statistically significantly associated with support for lowering inequality also aligns with expectations, including the traits unreliable, thief, lazy, dishonest, corrupt, and beggar.

Looking at individual countries, we again find strong heterogeneity. For Germany, both positive and negative economic stereotypes are statistically significantly associated with support for lowering inequality, generally in the expected directions. For Romania, only negative stereotypes are significant, while for Sweden, almost no individual stereotypes are statistically significant. Italy again presents an interesting deviation, as the largest statistically significant and positive coefficient is for the stereotype ‘beggar’. The disaggregation of effects by target countries in the online Appendix shows that Italy’s unexpected result for ‘beggar’ is driven by respondents who associate it with Germany and, to a lesser extent, Finland. As before, we interpret this as a backlash against Italy’s treatment during the Eurozone crisis: Italians who experienced the crisis more negatively might now be more supportive of EU redistribution and may, at the same time, perceive other countries as vindictive, greedy, or beggars based on their treatment of debtor countries during the crisis.

Our results for the third dependent variable – support for an EU-wide social welfare system – echo our results for reducing economic inequality. To avoid repetition, we present these results in the online Appendix.

Throughout this discussion, we have been careful to speak of ’associations’ rather than making causal claims. This is because our survey design cannot rule out potential reverse causality, where preferences for EU solidarity or economic redistribution influence stereotypes. While this is possible, we believe that the hypothesised direction – stereotypes shape attitudes – is more plausible in most cases. As explained in The role of stereotypes in public support for redistribution section, stereotypes tend to be long-standing, often unconscious beliefs that are more likely to start the causal chain that ultimately shapes attitudes. Now that we have identified important stereotypes in circulation that correlate with attitudes, future experimental work could verify whether the relationships we identify are truly causal.