The COVID-19 pandemic triggered a fundamental innovation in European Union (EU) economic governance through the adoption of NGEU and its core instrument, the recovery and resilience facility (RRF). The NGEU was a large-scale recovery plan (with over €800 billion in grants and loans) designed to promote resilience, social cohesion, and economic transformation across member states.
By allowing the European Commission to borrow jointly on capital markets and channel unprecedented volumes of funds to member states, the NGEU has been described as a turning point in EU fiscal integration (Bekker, 2021; Capati, 2023). Given that the RRF constitutes a temporary and exceptional instrument, created in response to an unprecedented crisis, sustainability concerns are less related to the long-term continuation of the RRF itself than to the political and democratic viability of similar large-scale EU interventions in the future.
While the NGEU framework is defined at the EU level through a common regulation, the allocation of funds is implemented via National Recovery and Resilience Plans designed and submitted by member state governments. These plans translate EU-level priorities into concrete national reforms and investments, including country-specific targets and milestones. They are subject to domestic political processes that may involve national parliaments and social partners. As a result, the governance of the RRF reflects a multi-level decision-making structure in which both EU and national executives play a central role, and where discretion in policy emphasis remains significant across countries.
From this perspective, incorporating citizens’ perceptions becomes particularly relevant, as public support and perceived legitimacy may condition whether comparable instruments can be replicated or institutionalised beyond the current framework. Examining sustainability and legitimacy through a citizen-centred lens, therefore, contributes not only to understanding the reception of the RRF but also to assessing the prospects for future EU-level crisis governance mechanisms.
While its primary goals include supporting green and digital transitions, economic cohesion, and social resilience, the plan’s legitimacy also hinges on the alignment between its allocation priorities and public expectations across member states (European Commission, 2021). Understanding how Europeans perceive the allocation of NGEU resources, and whether these perceptions align with their policy priorities, is therefore crucial for evaluating the legitimacy of this novel fiscal instrument.
This article examined the degree of alignment between citizen perceptions and preferences regarding the allocation of NGEU funds and examined cross-national variation in these priorities. Building on Eurobarometer survey data, individual- and country-level determinants of perceptions regarding whether the RRF contributes to recovery and resilience were assessed. In doing so, the article contributes to three debates: (1) the evolving literature on EU economic governance and fiscal capacity, (2) the role of mediating factors such as welfare regime legacies, labour-market structures, and national contexts in shaping public opinion, and (3) the emerging literature on the EU’s social dimension in the aftermath of COVID-19.
Findings showed that mismatches are systematic rather than incidental: in most member states, citizens placed a higher priority on social investment policies than they believed the EU is pursuing. In doing so, it contributes to the literature on public support for EU redistribution (Bechtel et al., 2014), preference formation in multilevel governance systems (Hooghe & Marks, 2005), and legitimacy of supranational crisis policymaking (Schmidt, 2020). The perception–preference gap could have relevant implications for how the EU sustains public trust in the aftermath of crisis governance, and for how future supranational initiatives can secure democratic legitimacy.
The remainder of the article proceeds as follows. Section 2 reviews the relevant literature on legitimacy and mediating factors. Section 3 summarises contributions on public opinion and EU economic governance. Section 4 describes the data and methods, explaining the Eurobarometer survey and the statistical techniques employed. Section 5 presents empirical results, first at the individual level and then through cross-country comparisons and cluster analysis. Section 6 concludes by highlighting contributions to ongoing debates in EU politics and outlining avenues for future research.
The formation of public preferences on fiscal and economic policy is a central topic in political science and public administration. The preferences citizens express regarding the allocation of supranational funds, such as those in the NGEU programme, are embedded in broader theoretical debates about political representation, opinion formation, and the perceived legitimacy of public policy. In the EU context, the distance between public preferences and policy outputs is a recurring concern, particularly when technocratic actors dominate decision-making processes (Majone, 1998). This raises questions about whether supranational agendas like the NGEU adequately reflect societal priorities or if they merely pursue elite-constructed goals in the name of European integration.
Traditional models of rational choice suggest that citizens weigh the costs and benefits of different policy options, aligning their preferences with their material self-interest (Downs, 1957). However, a growing body of literature emphasises the significant role of non-economic factors. Value-based theories argue that preferences are shaped by deeply held political, social, and moral values (Maldonado Hernández, 2008), while theories of political socialisation and social identity highlight the influence of national and European identities, as well as trust in political institutions (Hooghe & Marks, 2005). These frameworks suggest that a citizen’s preference for a specific policy area (such as green transition vs public health) is not solely a function of a rational calculation but is deeply intertwined with their worldview and group affiliations.
This dynamic is particularly salient within the context of multi-level governance, such as the EU. The EU’s perceived democratic deficit, often linked to a gap between elite decision-making and citizen priorities, makes the study of public opinion crucial for understanding levels of democratic legitimacy and institutional trust within the Union (Follesdal & Hix, 2006). A policy’s legitimacy (its acceptance by the public as right and proper) is not merely a given; it is a continuously negotiated outcome dependent on performance, procedural fairness, and public consent (Scharpf, 1999). The NGEU instrument, given its scale and scope, serves as a vital case study for examining this link. When citizens perceive that a policy’s objectives diverge from their own priorities, it can erode trust in institutions and undermine the political sustainability of the initiative itself.
Moreover, the NGEU could be theorised through Scharpf’s (1999) distinction between input and output legitimacy. The governance design of the NGEU raises questions regarding executive dominance (input legitimacy), as both the European Commission and national governments play a central role in agenda-setting and implementation. At the same time, this dominance is embedded in a multi-level structure in which national executives retain significant discretion over policy choices, and where opportunities for parliamentary and societal involvement vary across member states.
Recent research on the EU’s crisis management suggests that the legitimacy of policy responses depends not only on their effectiveness but also on their perceived alignment with public concerns (Schmidt, 2020). A growing body of work introduces the notion of a ‘perceived policy mismatch’, a gap between what citizens view as important, and the areas prioritised by policymakers (Braun & Tausendpfund, 2016; Ferry, 2021; Peters, 2018). This mismatch can erode institutional trust, especially in contexts where the EU is seen as a technocratic authority detached from popular will.
The NGEU plan, with its strong focus on digitalisation and climate neutrality, exemplifies this potential misalignment. While these are long-term strategic goals promoted by EU institutions, citizens, especially in times of crisis, may prioritise more immediate concerns, such as employment, health care, or inflation control (Wlezien, 2017). If these preferences are not reflected in policy design or communication, perceptions of disconnect may emerge.
The NGEU represented an institutional innovation in EU economic governance. It combined large-scale EU borrowing with conditional grants and loans to member states, thereby introducing elements of fiscal capacity at the EU level (Zgaga, 2023). Authors have emphasised its dual novelty: on the one hand, the unprecedented volume of EU-level borrowing, and on the other, the conditionality linking national recovery plans to EU oversight (Bekker, 2021; Capati, 2023). This design has raised debates on whether the NGEU constitutes a temporary emergency response or the foundation of a more permanent fiscal capacity at the EU level (Zgaga, 2023).
The governance of NGEU has also raised questions of centralisation and legitimacy. The RRF is implemented primarily through national recovery plans negotiated with the European Commission, giving executives a dominant role. Parliamentary institutions, at both the EU and national levels, are less central, leading to debates on executive centralisation (Borghetto, et al., 2025; Munta, et al., 2023). This shift heightened concerns regarding the democratic legitimacy of the instrument, as public acceptance becomes critical (Capati & Christiansen, 2025).
Additionally, under the RRF Regulation, member states are required to allocate at least 37% of their plans to climate-related objectives and 20% to digital transformation. However, beyond these minimum thresholds, governments retain discretion in allocating resources to other domains (health systems, labour-market policies, and social services).
While there is growing research on the governance design of NGEU, studies of public opinion remain scarce. Fernández-Pasarín and Lanaia (2025) hold that while fiscal solidarity had been explored in relation to the Eurozone crisis, perceptions of NGEU effectiveness were under-researched. This article addresses this gap by analysing determinants of citizens’ evaluations of the NGEU.
Therefore, understanding citizens’ priorities for the NGEU requires situating individual attitudes within broader institutional and structural contexts. Following the literature on European integration, fiscal capacity, and welfare regimes, three mediating mechanisms could be highlighted: (1) welfare regime legacies, (2) labour-market structures, and (3) national political contexts. They may shape how individuals translate preferences into expectations about EU-level action.
Research on welfare state regimes has long emphasised that institutional legacies shape citizens’ expectations of public provision (Esping-Andersen, 1990; Ferrera, 1996). In Nordic and social-democratic regimes, comprehensive national health and education systems may reduce the perceived need for supranational intervention, leading to greater support for social investment measures such as youth education and training. By contrast, in Southern and Eastern European regimes with weaker or fragmented welfare arrangements, citizens may assign higher priority to EU investment in health or social protection (Bekker, 2021; De la Porte & Jensen, 2021).
The organisation of labour markets also conditions preferences. In countries with high dualisation, precarious employment, and weaker protection for non-standard workers, citizens are more likely to perceive EU funding for improving working conditions as salient (Emmenegger et al., 2012; Rueda, 2014). Conversely, in coordinated labour markets with stronger protections, citizens may prioritise long-term investment over immediate labour-market regulation.
Finally, attitudes towards EU spending priorities are mediated by national experiences of crisis and political framing. Countries more severely hit by the COVID-19 pandemic, or with recent histories of austerity, may exhibit stronger demand for EU-level intervention in social areas (Munta et al., 2023; Zgaga, 2023). Levels of political trust and perceptions of legitimacy also influence whether citizens view the EU as an appropriate actor to deliver such policies (Borghetto et al., 2025; Capati & Christiansen, 2025).
The data source comes from the Eurobarometer 99.4 conducted in 2023 (European Commission, Brussels, 2024), a multi-topic, cross-national public opinion survey conducted on behalf of the European Commission. It includes a set of questions specifically designed to measure citizens’ preferences regarding the NGEU plan. The survey covers EU member states and is based on face-to-face interviews with nationally representative samples, typically around 1,000 respondents per country (European Commission, 2021). However, Croatia and Cyprus were not included in the sample. In the first case, it was dropped from the model due to collinearity, and in the second case, the following questions on the NGEU were not included in the questionnaire.
The core of the analysis is two key survey questions that directly address citizens’ perceptions and preferences regarding the NGEU recovery programme. The key outcome variable is based on the questions:
‘On which of the following areas do you think the recovery plan NGEU is currently spent? Firstly?’
‘On which of the following areas would you like the recovery plan NGEU to be spent in priority? Firstly?’
The response categories were: (1) scientific research and development, (2) youth education and training, (3) culture and media, (4) energy, (5) transport, (6) climate change and environmental protection, (7) agriculture and rural development, (8) health, (9) improvement of working conditions of EU citizens, (10) support to SMEs, (11) defence and security, (12) public administration, (13) industry, (14) digitalisation of economy and society, (15) other, (16) none, and (17) do not know.
The methodology comprises three main parts. First, a descriptive analysis considering a frequency analysis of the responses to both the perception and preference questions. This descriptive step is crucial for identifying the most significant discrepancies between perceived and preferred fund allocation, which serves as the foundation for our subsequent inferential analysis.
To examine the determinants of citizens’ preferences over the allocation of NGEU funds, a series of probit models was estimated exclusively for those policy areas where the absolute difference between the share of respondents who selected the area as a preferred spending priority and the share who perceived it as an actual priority of the NGEU exceeded 5% points. By restricting the dependent variables to these high-gap areas, the modelling strategy targets the policy domains where understanding the determinants of preferences is most relevant for identifying potential misalignments between citizen priorities and perceived governmental action.
The probit models are appropriate given that the dependent variables in each case are binary, coded as 1 if the respondent selected the corresponding policy area among their top spending priorities and 0 otherwise (Wooldridge, 2010). This approach models the probability that a given respondent prioritises a specific domain as a function of their sociodemographic characteristics, economic situation, and political attitudes.
The estimated coefficients cannot be interpreted directly as changes in probability; instead, the average marginal effects were computed to express the expected change in the probability of prioritising a given area for a one-unit change in each predictor, holding other variables constant. Country fixed effects are also included in extended models to account for unobserved heterogeneity across member states. The models were estimated using survey weights provided in the Eurobarometer dataset to ensure representativeness of the EU population. Robust standard errors clustered by country are employed to correct for within-country correlation in the error terms.
Country fixed effects are included to absorb unobserved national characteristics that are constant across individuals, such as welfare-state institutions, labour-market regimes, or the health-system capacity. In the country fixed-effects specifications, the estimated coefficients for the country dummies capture the difference in the probability of prioritising a given policy area relative to the reference category (Germany), holding all other individual-level characteristics constant. For the computation of average marginal effects by country, the dummy for the country of interest is set to 1, and all other country dummies are set to 0, rather than centring them at their sample means.
Germany was chosen as the reference category because it represents a central and substantively meaningful benchmark in EU fiscal governance: it is the largest member state, a key agenda-setter in the negotiation of the NGEU, and a country whose welfare and labour-market institutions combine social investment with fiscal restraint. Using Germany as the baseline allows country effects to be interpreted as deviations from a central case rather than from a peripheral or extreme one.
This procedure ensures that the reported marginal effect reflects the discrete change in predicted probability when moving from the reference country to the country of interest, controlling for the same baseline in all other cases. Such an approach facilitates direct and meaningful comparisons across countries, as it isolates the effect of being from a specific country. Failing to account for these factors would risk attributing structural national differences to individual-level characteristics. As a robustness check, post-estimation country-level marginal effects were computed for other member states by setting each country dummy to one while holding all others at zero. The resulting country rankings and the cluster structure remain stable irrespective of the omitted category, indicating that the substantive cross-national patterns are not driven by the choice of Germany as the reference country.
As an additional robustness check, the analysis was extended using multilevel probit models with random intercepts at the country level, combining individual-level Eurobarometer data with country-level indicators capturing welfare-state legacies, labour-market arrangements, and economic performance. The results from these multilevel models closely mirror the baseline estimates: the direction, magnitude, and statistical significance of the individual-level predictors remain largely unchanged. Moreover, the estimated between-country variance is statistically significant, confirming the presence of meaningful cross-national heterogeneity in citizens’ policy priorities beyond individual characteristics (see Appendix).
To assess whether the clustering results are sensitive to the underlying statistical model, the country clustering was replicated using country-level random intercepts obtained from multilevel probit models with random intercepts at the country level. For each policy domain, best linear unbiased predictions (BLUPs) of the country-specific random effects were extracted, standardised, and used as inputs in a hierarchical clustering procedure identical to the baseline specification. The resulting country groupings are highly consistent with those obtained using fixed-effects marginal effects, indicating that the identified clusters do not depend on the choice between fixed-effects and multilevel modelling. This stability across specifications strengthens confidence in the substantive interpretation of the clusters.
Table 1 describes the set of independent variables that include a range of sociodemographic characteristics (age, gender, level of education, among others) and other key attitudinal indicators from the Eurobarometer, such as a respondent’s self-reported political ideology, among others. The selection of individual-level covariates is guided by three complementary theoretical mechanisms emphasised in the literature on welfare preferences and EU legitimacy: life-cycle risk exposure, socio-economic position, and ideological orientation. Variables capturing institutional trust or EU support are not included in the main specifications because they are likely to be endogenous to perceptions of NGEU priorities, potentially biasing estimates.
Independent variables
| Label | Values | Mean | Std. Dev. | Min. | Max. |
|---|---|---|---|---|---|
| AGE | Respondent’s age. | 51.13 | 18.49 | 15 | 98 |
| AGE_SQ | AGE*AGE | 2,956.20 | 1,894.09 | 225 | 9,604 |
| BACHELOR | 1 if the respondent’s highest level of education attained is a bachelor’s degree and 0 otherwise. | 0.12 | 0.33 | 0 | 1 |
| CENTRE | 1 if respondent places themselves at 5 or 6 on a 1–10 political scale (left-wing) and 0 otherwise. | 0.36 | 0.48 | 0 | 1 |
| DOCTORAL | 1 if the respondent’s highest level of education attained is a doctoral degree and 0 otherwise. | 0.01 | 0.10 | 0 | 1 |
| LEFT | 1 if respondent places themselves between 1 and 4 on a 1–10 political scale (left-wing) and 0 otherwise. | 0.27 | 0.44 | 0 | 1 |
| MASTER | 1 if the respondent’s highest level of education attained is a master’s degree and 0 otherwise. | 0.12 | 0.32 | 0 | 1 |
| MEN | 1 if male and 0 otherwise. | 0.47 | 0.50 | 0 | 1 |
| MIDDLE_CLASS | 1 if the respondent self-identifies as belonging to the middle class and 0 otherwise. | 0.73 | 0.44 | 0 | 1 |
| RURAL | 1 if living in rural areas and 0 otherwise | 0.34 | 0.47 | 0 | 1 |
| UNEMPLOYED | 1 if unemployed and 0 otherwise | 0.04 | 0.20 | 0 | 1 |
| WORKING_CLASS | 1 if the respondent self-identifies as belonging to the working class and 0 otherwise. | 0.23 | 0.42 | 0 | 1 |
Source: author’s calculations based on European Commission, Brussels (2024).
To identify distinct groups of EU member states based on their preferences for NGEU fund allocation, a two-step clustering approach was employed. First, all marginal effects for health (MFX_HEALTH), youth education and training (MFX_YET), and improvement in working conditions (MFX_IMP_WC) were standardised to z-scores. This step ensured that each variable contributed equally to the distance calculations, preventing any single variable from disproportionately influencing the clustering process due to differences in scale.
Subsequently, a K-means clustering algorithm was applied to the standardised marginal effects. This iterative partitioning method aims to divide n observations into k clusters, where each observation belongs to the cluster with the nearest means. To determine the optimal number of clusters (k), the Calinski–Harabasz index was utilised. This criterion, which is a ratio of the sum of between-clusters dispersion and the sum of within-cluster dispersion, was evaluated for solutions ranging from 2 to 9 clusters using the cluster stop command in Stata (e.g. Everitt et al., 2011). The analysis indicated that k = 2 yielded the highest Calinski–Harabasz score while also providing the most parsimonious and interpretable solution, leading to the selection of two distinct country groups.
Following the determination of k = 2 as the optimal number of clusters, a hierarchical clustering analysis using Ward’s linkage method was performed. Ward’s method minimises the total within-cluster variance, aiming to create clusters of relatively uniform size. The clustering structure was generated, which was then visualised through a dendrogram. This visual representation confirmed the appropriateness of the two-cluster solution by revealing a significant cut point in dissimilarity, solidifying the division of EU countries into two primary groups based on their fund allocation preferences. Country assignments to their respective clusters were then extracted for further interpretation and mapping.
This descriptive analysis focuses on two key aspects of public opinion regarding the NGEU programme: citizen perception of the current allocation of funds and their preferences for where these funds should be invested. The findings presented below are based on a weighted frequency and percentage analysis of the Eurobarometer 99.4 survey (Table 2).
Citizen perceptions and preferences on the allocation of NGEU funds
| Category | Share of respondents (%) | Gap in answers (percentage points) | |
|---|---|---|---|
| On which of the following areas do you think the recovery plan NGEU is currently spent? Firstly? | On which of the following areas would you like the recovery plan NGEU to be spent in priority? Firstly? | ||
| Health | 6.5% | 17.2% | −10.7 |
| Youth education and training | 6.0% | 13.4% | −7.4 |
| Improvement of working conditions of EU citizens | 4.4% | 9.4% | −5.0 |
| Agriculture and rural development | 4.9% | 7.1% | −2.2 |
| Scientific research and development | 4.7% | 5.8% | −1.1 |
| Support for SMEs | 4.0% | 5.1% | −1.1 |
| Culture and media | 1.4% | 1.2% | 0.2 |
| Transport | 1.8% | 1.5% | 0.3 |
| Climate change and environmental protection | 14.3% | 13.8% | 0.5 |
| Industry | 4.3% | 3.3% | 1.0 |
| Energy | 9.2% | 6.7% | 2.5 |
| Public administration | 4.0% | 1.4% | 2.6 |
| Digitalisation of economy and society | 6.2% | 2.9% | 3.3 |
| Defence and security | 11.4% | 7.1% | 4.3 |
| Other | 0.8% | 0.3% | 0.5 |
| None | 1.4% | 0.4% | 1.0 |
| Do not know | 14.7% | 3.4% | 11.3 |
| Total | 100% | 100% | – |
Source: author’s calculations based on European Commission, Brussels (2024).
The findings indicate that public perception regarding the use of NGEU funds is heterogeneous. The area most frequently perceived by citizens as receiving investment was climate change and environmental protection, with 14.3% of respondents mentioning it. This was followed closely by funds allocated to defence and security and energy, at 11.4% and 9.2%, respectively. In contrast, areas such as transport, culture, and media were perceived as the least prioritised, with only 1.8% and 1.4% of respondents believing that funds are being invested there.
When citizens were asked about their preferences, the response pattern differed significantly. Health emerged as the most prominent priority, with a notable 17.2% of respondents preferring that funds be allocated to this sector. This represents a marked contrast with the current perception. The second most mentioned priority was climate change and environmental protection (with 13.8% suggesting this area), followed by youth education and training (with 13.4%).
A comparative analysis reveals a substantial perception-preference gap in several areas.
Health: a significant gap exists, with only 6.5% of citizens believing health is the NGEU priority, while an overwhelming 17.2% would prefer it to be.
Youth education and training: this area shows one of the largest discrepancies, with the 13.4% preference surpassing the 6.0% perception.
Improvement of working conditions of EU citizens: the preference also exceeds the perception, as only 4.4% of citizens believe funds are invested in it, but 9.4% would prioritise it.
The analysis reveals that the largest perception-preference gaps, in absolute value, are consistently found in areas where preferred rates exceed perceived rates. An additional key finding is that while 14.7% of respondents reported being unaware of how the NGEU funds were being spent, this figure dropped significantly to only 3.4% when asked about their preferences. This substantial reduction in ‘do not know’ responses from the perception question to the preference question suggests a potential information problem, where a lack of awareness about the programme’s priorities may be a factor driving the public’s desire for a different allocation of resources.
Table 3 reports the average marginal effects from three probit models predicting the probability of prioritising health spending (HEALTH), youth education, and training (YET), and improving working conditions (IMP_WC) in the allocation of the NGEU funds. These baseline models include individual-level sociodemographic, political, and economic variables, but no country fixed effects.
Marginal effects from the probit models with individual-level predictors
| Change in the P(HEALTH = 1) | Change in the P(YET = 1) | Change in the P(IMP_WC = 1) | ||||
|---|---|---|---|---|---|---|
| dy/dx | Std. Err. | dy/dx | Std. Err. | dy/dx | Std. Err. | |
| AGE | 0.00226*** | (0.0006) | −0.00727*** | (0.0013) | 0.00270*** | (0.0008) |
| AGE_SQ | −7.20e-06 | (6.86e-06) | 0.00006*** | (0.0000) | −0.00003*** | (7.00e-06) |
| MEN | −0.04601*** | (0.0117) | −0.02125** | (0.0103) | −0.00187 | (0.0044) |
| BACHELOR | 0.00874 | (0.0224) | 0.00771 | (0.0103) | −0.00946 | (0.0094) |
| MASTER | −0.06391*** | (0.0220) | 0.01723 | (0.0122) | −0.04439*** | (0.00710) |
| DOCTORAL | −0.08722*** | (0.0307) | 0.01077 | (0.0351) | −0.03209 | (0.0270) |
| UNEMPLOYED | 0.01121 | (0.0138) | 0.02413* | (0.0142) | 0.04216*** | (0.0129) |
| RURAL | 0.02078 | (0.0175) | −0.02102** | (0.0072) | −0.02111** | (0.0107) |
| WORKING_CLASS | 0.04181* | (0.0216) | −0.02422 | (0.0249) | −0.00400 | (0.0265) |
| MIDDLE_CLASS | −0.01159 | (0.0244) | −0.01741 | (0.0347) | −0.03300 | (0.0235) |
| LEFT | −0.03817** | (0.0152) | 0.01738** | (0.0081) | −0.00186 | (0.0118) |
| CENTRE | −0.01814 | (0.0164) | 0.02166** | (0.0107) | −0.01297 | (0.0148) |
| Observations | 24,846 |
Source: author’s calculations based on European Commission, Brussels (2024).
Note: ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels respectively.
For HEALTH, older respondents are significantly more likely to favour this allocation, which is consistent with life-cycle arguments suggesting that individuals’ policy preferences shift towards areas directly affecting their own age cohort (Busemeyer et al., 2021). Men are less likely to prioritise health, reflecting well-established gender gaps in support for social spending (Edlund & Pande, 2002). Higher education levels, particularly master’s and doctoral degrees, are associated with a lower probability of selecting health, potentially reflecting greater awareness of competing budgetary priorities or reflecting greater opportunity costs perceived by higher-educated individuals (Rehm, 2016). Working-class respondents are more likely to prioritise health, aligning with evidence that lower socioeconomic status increases demand for public welfare provision (Powell et al., 2020). Finally, respondents identifying as left-leaning are less likely than right-leaning individuals to select health. This finding runs counter to the general expectation that the political left favours expansive welfare spending (Busemeyer, 2014), and may indicate that among left-leaning respondents, EU-level funds are seen as better directed towards other areas rather than health, which is often regarded as a national-level responsibility in many welfare states (Pereirinha & Pereira, 2021).
For YET, the results show a negative age effect (p < 0.001) and a positive coefficient for age squared (p < 0.001) that reveals a U-shaped relationship: support is highest among the youngest and oldest respondents, and lower in middle age, consistent with generational interest theories (Busemeyer, 2014). Men are less supportive (p < 0.05), and rural residents are also less likely to prioritise this area (p < 0.01), potentially reflecting lower perceived direct benefits from education and training investments or initiatives in rural contexts. Unemployed respondents show higher support, consistent with self-interest models where those facing labour market vulnerability prefer investment in human capital (Rehm, 2016). Political ideology is significant for both LEFT and CENTRE categories; left-leaning respondents and centrists are more likely to support investments in this area. For left-leaning individuals, this aligns with progressive commitments to equality of opportunity through education (Busemeyer, 2014). For centrists, the finding is consistent with prior evidence that education and training can serve as ‘valence issues’ attracting broad ideological support when framed in terms of competitiveness and growth (Hooghe & Marks, 2005).
For IMP_WC, age is positively associated (p < 0.01), while the negative age squared term (p < 0.001) suggests diminishing returns of age. Master’s degree holders are less supportive (p < 0.001), perhaps due to higher labour market security. Unemployed respondents are significantly more likely to support this allocation (p < 0.01), which aligns with self-interest theory predictions that those with greater exposure to labour market risks are more supportive of policies enhancing job quality (Rehm, 2016). Rural residents show lower support (p < 0.05), and attachment to the EU is negatively associated (p < 0.05), possibly indicating a preference for other forms of EU intervention among more pro-European respondents. It is worth noting that in this case, political ideology does not have significant effects, suggesting that preferences regarding labour market improvements are not shaped by ideological orientation. This finding is consistent with comparative welfare research showing that labour market policy preferences are often cross-cutting in their political appeal (Häusermann, 2010).
Table 4 reports results for models including country fixed effects (Germany omitted), capturing unobserved national-level factors such as institutional arrangements, welfare state structures, and pandemic impacts. When country fixed effects are included, the individual-level patterns persist, but country dummies reveal substantial cross-national heterogeneity. These marginal effects represent the change in the predicted probability relative to Germany, holding all individual-level characteristics constant.
Country marginal effects from the probit models
| Change in the P(HEALTH = 1) | Change in the P(YET = 1) | Change in the P(IMP_WC = 1) | ||||
|---|---|---|---|---|---|---|
| dy/dx | Std. Err. | dy/dx | Std. Err. | dy/dx | Std. Err. | |
| AUSTRIA | 0.02868 | (0.0633) | −0.03421*** | (0.0088) | 0.02988*** | (0.0069) |
| BELGIUM | 0.13884*** | (0.0150) | −0.06261*** | (0.0066) | 0.02753*** | (0.0075) |
| BULGARIA | 0.07920*** | (0.0124) | −0.08321*** | (0.0063) | 0.13710*** | (0.0135) |
| CZECH REPUBLIC | 0.00816 | (0.00845) | −0.08609*** | (0.0053) | 0.03910*** | (0.0083) |
| DENMARK | 0.06652*** | (0.0123) | −0.09310*** | (0.0045) | −0.01902*** | (0.0041) |
| ESTONIA | 0.05333*** | (0.0108) | −0.06540*** | (0.0081) | 0.01531** | (0.0074) |
| FINLAND | 0.06270*** | (0.0114) | −0.02627*** | (0.0096) | −0.01544*** | (0.0046) |
| FRANCE | 0.24900*** | (0.0184) | −0.05514*** | (0.0075) | −0.00904* | (0.0048) |
| GREECE | 0.38746*** | (0.0204) | −0.06940*** | (0.0065) | 0.05467*** | (0.0097) |
| HUNGARY | 0.13421*** | (0.0153) | −0.08885*** | (0.0050) | 0.08052*** | (0.0091) |
| IRELAND | 0.30684*** | (0.0186) | −0.06393*** | (0.00784) | 0.00336 | (0.0059) |
| ITALY | 0.05406*** | (0.0132) | −0.05231*** | (0.00646) | 0.11155*** | (0.0111) |
| LATVIA | 0.25199*** | (0.0170) | −0.09149*** | (0.0050) | 0.00512 | (0.0065) |
| LITHUANIA | 0.11851*** | (0.0133) | −0.05528*** | (0.0081) | 0.02451*** | (0.0094) |
| LUXEMBOURG | 0.12815*** | (0.0143) | −0.06297*** | (0.0072) | −0.01088** | (0.0050) |
| MALTA | 0.33234*** | (0.0199) | −0.04030*** | (0.0089) | 0.00141 | (0.0061) |
| NETHERLANDS | 0.04900*** | (0.0110) | −0.03940*** | (0.0099) | 0.00015 | (0.0074) |
| POLAND | 0.20085*** | (0.0173) | −0.09289*** | (0.0047) | 0.01627*** | (0.0063) |
| PORTUGAL | 0.50143*** | (0.0229) | −0.09348*** | (0.0044) | 0.13199*** | (0.0118) |
| ROMANIA | 0.07977*** | (0.0126) | −0.06532*** | (0.0066) | 0.04979*** | (0.0077) |
| SLOVAKIA | 0.14772*** | (0.0147) | −0.08909*** | (0.0054) | 0.04506*** | (0.0080) |
| SLOVENIA | 0.21341*** | (0.0172) | −0.07518*** | (0.0068) | 0.06226*** | (0.0100) |
| SPAIN | 0.26496*** | (0.0187) | −0.05951*** | (0.0073) | 0.12367*** | (0.0106) |
| SWEDEN | −0.01552** | (0.0062) | −0.06333*** | (0.0068) | −0.01304** | (0.0051) |
| Observations | 24,846 |
Source: author’s calculations based on European Commission, Brussels (2024).
Note: ***, ** and * indicate statistical significance at the 1%, 5% and 10% levels respectively.
For HEALTH, the largest positive differences are observed in Portugal (+0.501, p < 0.001), Greece (+0.387, p < 0.001), and Malta (+0.332, p < 0.001), indicating substantially higher support for channelling NGEU funds into health care compared to the German baseline. Other countries with significantly higher probabilities include Ireland (+0.307), Spain (+0.265), and Latvia (+0.252). The magnitude of these effects suggests that in Southern and certain smaller member states, public opinion is more strongly oriented towards immediate social spending, possibly reflecting legacies of comparatively lower per capita health expenditure and the heightened salience of health systems during the COVID-19 crisis (Pereirinha & Pereira, 2021). By contrast, Sweden (− 0.016, p < 0.05) is the only member state showing a significant negative difference, consistent with evidence that strong pre-existing health infrastructures can reduce perceived urgency for additional supranational investment (Gallie, 2007).
In the case of YET, all country effects relative to Germany are negative and statistically significant, with the largest gaps in Denmark (−0.093, p < 0.001), Poland (−0.093, p < 0.001), and Portugal (−0.093, p < 0.001). These findings suggest that, after controlling for sociodemographic and attitudinal factors, respondents in many member states are less inclined than Germans to view ‘youth education and training’ as the primary allocation target for EU recovery funds. Potential explanations include satisfaction with domestic education systems in Nordic countries (Busemeyer, 2014) and more urgent competing policy priorities in Southern and Eastern Europe, such as job quality and wage growth.
For IMP_WK, substantial positive marginal effects are found in Bulgaria (0.137, p < 0.001), Portugal (0.132, p < 0.001), and Spain (0.124, p < 0.001), with Italy (0.112) and Hungary (0.081) also showing elevated support. These countries have historically exhibited higher levels of labour market segmentation and lower job security (Gallie, 2007), which may amplify public demand for EU investment in working conditions. Conversely, negative effects are observed in Sweden (−0.013, p < 0.05), Finland (−0.015, p < 0.001), and Luxembourg (−0.011, p < 0.05), reflecting lower perceived need in contexts with stronger labour protections.
Given these results, a hierarchical cluster analysis was made, which identified two distinct groups based on their standardised marginal effects of preferences regarding the allocation of the NGEU funds. The identification of two optimal clusters was supported by the Calinski–Harabasz criterion (e.g. Everitt et al., 2011).
Table 5 reports descriptive statistics of the standardized marginal effects for three outcome variables (HEALTH, YET, and IMP_WC), across two country clusters identified in the analysis. The distribution of member states in each cluster reveals a meaningful geographic and institutional divide. Cluster 1 includes Austria, Belgium, Estonia, Finland, France, Ireland, Lithuania, Luxembourg, Malta, the Netherlands, Romania, and Sweden, while Cluster 2 comprises Bulgaria, the Czech Republic, Denmark, Greece, Hungary, Italy, Latvia, Poland, Portugal, Slovakia, Slovenia, and Spain. This configuration reflects the presence of both long-standing Western and Northern European welfare states and newer or Southern members, highlighting how structural and historical factors may shape policy preferences.
Cluster-level descriptive statistics of the standardised marginal effects
| Label | Description | Cluster | Mean | Std. Dev. | Min. | Max. |
|---|---|---|---|---|---|---|
| MFX_SHEALTH | Standardised marginal effect of the HEALTH variable | 1 | −0.25 | 0.86 | −1.34 | 1.32 |
| 2 | 0.25 | 1.11 | −1.16 | 2.61 | ||
| MFX_SYET | Standardised marginal effect of the YET variable | 1 | 0.71 | 0.70 | 0.08 | 2.04 |
| 2 | −0.71 | 0.70 | −1.33 | 0.74 | ||
| MFX_SIMP_WC | Standardized marginal effect of the IMP_WC variable | 1 | −0.59 | 0.43 | −1.09 | 0.26 |
| 2 | 0.59 | 1.07 | −1.16 | 2.06 |
Source: author’s calculations based on European Commission, Brussels (2024).
The results in the case of HEALTH showed that Cluster 1 had a negative mean standardised marginal effect (−0.25), while Cluster 2 recorded a positive mean (0.25). This suggests that, on average, respondents in Cluster 2 are more likely to prioritise health spending under the NGEU plan. Moreover, the larger dispersion within Cluster 2 (standard deviation = 1.11; range up to 2.61) compared to Cluster 1 (standard deviation = 0.86; max = 1.32) indicates greater heterogeneity in the salience of health priorities. This pattern resonates with earlier findings on the differentiated legacies of welfare regimes across Europe. Southern and Eastern European states often face weaker healthcare infrastructures and higher vulnerability to shocks, which may heighten public concern with health investment (Brooks, et al., 2021; Ferrera, 1996). By contrast, Northern and Western states, characterised by relatively robust healthcare systems, may exhibit less urgency to prioritise health in recovery spending, although they still view it as an important policy domain.
Regarding YET, the opposite trend is found. Cluster 1 countries display a positive average marginal effect (0.71), while Cluster 2 shows a negative effect (−0.71). This divergence underscores the extent to which education and training are valued differently across regions. In Northern and Western Europe, education is strongly associated with social investment strategies, where policy is oriented towards building human capital and promoting long-term economic competitiveness (Busemeyer, et al., 2020; Hemerijck, 2013). Conversely, in many Southern and Eastern European member states, more immediate socioeconomic challenges such as unemployment and labour-market precariousness may overshadow investment-oriented policies, reducing the relative prioritisation of education and training (Bohle & Greskovits, 2019). The clustering, therefore, highlights the extent to which structural economic differences condition citizens’ attitudes towards the allocation of EU recovery funds.
Concerning IMP_WC, findings showed another contrasting pattern. Cluster 1 has a negative mean marginal effect (−0.59), while Cluster 2 records a positive mean (0.59). This implies that individuals in Cluster 2 countries are significantly more likely to prioritise improvements in working conditions compared to their Cluster 1 counterparts. This finding can be linked to the persistence of dualised labour markets and weaker employment protection regimes in several Southern and Central-Eastern European countries, where precarious contracts, informality, and youth unemployment remain pressing issues (Bohle & Greskovits, 2019; Emmenegger, 2012). In contrast, Cluster 1 countries, many of which are high-income Western and Nordic welfare states, have more established labour protections and inclusive social systems, potentially lowering the salience of employment-related spending in the context of EU recovery policy.
Taken together, these results point to a systematic divide in citizen preferences regarding EU-level recovery policies. Cluster 1 countries demonstrate a profile aligned with a social investment orientation, emphasising education and training, while assigning comparatively less weight to health and labour-market conditions. Cluster 2, by contrast, highlights health and employment-related concerns, consistent with structural vulnerabilities in their welfare and labour-market regimes. These differences echo long-standing debates on the diversity of welfare states in Europe and the challenges of forging consensus in EU social policy (Ferrera, 1996; Hemerijck, 2013). The findings therefore suggest that support for NGEU initiatives is filtered through national institutional contexts and historical welfare trajectories, underscoring the difficulty of articulating a uniform European social investment strategy (Brooks, et al., 2021; Busemeyer et al., 2020).
Theoretically, the results underscore the relevance of mediating mechanisms. Welfare regime legacies and labour-market structures condition which domains gain salience in public opinion. In countries with fragmented welfare systems and dualised labour markets, citizens are more likely to demand EU investment in health and working conditions. By contrast, in consolidated universalist regimes, social investment measures, such as education and training, attract greater support.
In terms of legitimacy, these dynamics echo Scharpf’s (1999) distinction between input and output legitimacy: while NGEU may be justified by its effectiveness, its sustainability depends on whether citizens perceive that it addresses their own priorities. A disconnect between supranational objectives and social expectations risks undermining trust in the EU (Follesdal & Hix, 2006; Schmidt, 2020).
The findings demonstrate that the legitimacy of NGEU cannot be assessed solely through its strategic objectives (such as the green and digital transitions) but must also be evaluated in terms of its perceived responsiveness to citizens’ immediate social concerns, such as health, education, and working conditions. The perception–preference gap identified here suggests that, for many Europeans, the priorities associated with the EU recovery framework, as they are perceived and communicated, are not fully aligned with their most pressing needs. Importantly, this gap emerges within a multi-level governance context in which national governments retain significant discretion over the translation of EU-level objectives into concrete spending choices. As such, perceived misalignment reflects not only institutional design at the EU level but also the complex interaction between EU frameworks, national policy choices, and public communication.
The analysis provides a multi-layered perspective on citizens’ preferences for allocating resources within the NGEU framework. The descriptive evidence indicated that health, youth education and training, and improving working conditions of EU citizens emerged as the three domains in which the largest discrepancies between citizens’ preferences and their perceptions of NGEU spending were found. This heterogeneity suggests that while the programme was conceived as a common response to the COVID-19 crisis, public expectations for its use remain highly context-dependent. Moreover, while respondents consistently identify climate and environmental protection as a visible area of NGEU investment, they would prefer a much stronger emphasis on the social pillars of health, education, and labour market quality. This perception–preference misalignment suggests that, despite the programme’s unprecedented scale, many citizens feel that their most immediate social concerns are not adequately addressed at the EU level.
The probit models clarify who supported which domains. Regarding health, older citizens are more likely to prioritise it; men are less likely; higher degrees (Master’s/PhD) reduce the probability; and working-class respondents are more likely to select health. Ideologically, left-leaning respondents are less likely to prioritise health within the NGEU framework. Rather than reflecting explicit considerations about EU competences (which are not directly observable in the survey), this pattern likely captures differences in how policy priorities are framed across ideological groups, as well as varying expectations about the appropriate scope of EU-level intervention.
In the case of youth education and training, support follows a U-shape over the life cycle, is lower among men and rural residents, and higher among the unemployed; both left and centre identifiers are more supportive than the right, consistent with ‘social investment’ frames that attract broad coalitions. For the improvement in working conditions of the EU citizens, age is positive with diminishing returns, a master’s degree lowers support, unemployment raises it, rural residence lowers it, and (unlike the other domains) ideology is not significant, consistent with the cross-cutting appeal of job-quality policies. These individual-level patterns align with scholarship linking life-cycle risks, labour-market position, and ideology to social policy preferences.
Introducing country fixed effects (Germany omitted) reveals systematic cross-national differences. For example, in the case of health, most countries registered a significant and positive impact on the probability (relative to Germany), the largest marginal effects were found in the cases of Portugal, Greece and Malta (0.501, 0.387, and 0.332, respectively) while only in the case of Sweden, a significant and negative impact was found (−0.016). Regarding youth education and training, the country effects are uniformly significant and negative relative to Germany, with the most negative reported for Portugal, Denmark, and Poland (−0.093 each). In the case of improving working conditions for EU citizens, the largest positive impact appeared in Bulgaria (+0.137) while the largest negative impact was registered in the case of Denmark (−0.019). Taken together, these signs indicate that, holding individual traits constant, Southern and parts of Central-Eastern Europe exhibit stronger demand for EU-level investment in health and job quality, whereas Germany is relatively more favourable to YET than most peers. This pattern is consistent with well-documented regime and labour-market differences across Europe.
The clustering of standardised country-level marginal effects consolidated these gradients into two coherent groups, with two clusters selected via the Calinski–Harabasz criterion and confirmed by Ward’s hierarchical structure. The first cluster exhibited a negative mean for health and improving working conditions, while the mean for youth education and training was negative. The second cluster showed the mirror image. These signs (and the larger dispersion in Cluster 2 for health and improving working conditions reinforce a substantive cleavage: a social-investment profile in Cluster 1 (education/training salience) versus a health-and-job-quality profile in Cluster 2, which maps closely onto classic welfare-regime and labour-market divides.
Including country-level variables (social-protection spending, youth unemployment, and GDP growth in 2020) as second-level covariates produced generally non-significant coefficients, except for GDP growth in 2020, which is marginally significant and negative for preferences regarding improvements in working conditions. This suggests that where the pandemic produced a stronger economic contraction, citizens were relatively more likely to prioritise job quality (consistent with heightened salience of labour-market insecurity in economically affected contexts).
The results extend the comparative welfare state literature by situating EU-level fiscal instruments within long-standing regime differences. The clustering analysis reveals two distinct country groups: one emphasising youth education and training and another prioritising health and working conditions. This typology resonates with established regime distinctions in Europe while also highlighting how EU-level instruments are filtered through domestic legacies. As such, the study advances the understanding of how supranational fiscal solidarity interacts with national welfare regimes and diverse citizen expectations. Moreover, the EU can enhance the legitimacy of joint fiscal tools by tailoring communication and programming: for example, emphasising education and training where a social-investment frame resonates (Cluster 1), and foregrounding health and working-conditions improvements where those concerns are salient (Cluster 2). Rather than a one-size-fits-all narrative, differentiated emphasis—within common EU objectives—may better sustain broad support.
These findings contribute to the growing literature on EU policy legitimacy by showing that citizen support is not only shaped by distributive outcomes but also by the perceived alignment between collective spending and individual priorities. By uncovering the specific domains where EU citizens detect the strongest gaps, this study provides concrete evidence that legitimacy at the supranational level hinges on bridging these preference–perception divides. The results reinforced that welfare-regime legacies and labour-market structures shape supranational preferences, mediating the translation of EU-level frames into citizen demand. Overall, the study showed that the success of NGEU as a vehicle of European solidarity depends on recognising and accommodating preference diversity, not merely on macroeconomic efficacy.