BY Alexandra Bertilsson

Advised by Professor Henry Shim and Assistant Instructor So Hye Yoon

Princeton University Economics Department 

1. Introduction

“It must, indeed, be one of these two events; either the nation must destroy public credit, or public credit will destroy the nation.” David Hume, “Of Public Credit”

Public debt is a critical instrument of fiscal policy: when used judiciously, it can stabilize output during downturns and fund essential investments. However, if it is mismanaged, it poses significant long-term risks. The Coronavirus Aid, Relief, and Economic Security (CARES) Act of 2020 – a $2.2 trillion stimulus package – illustrates this ambivalence. While it provided essential relief during the COVID-19 pandemic, it also increased the U.S. national debt from $23.2 trillion to over $27 trillion by the end of the year (U.S. Treasury, 2020). This shows one way fiscal expansions can support employment and output, while also eroding fiscal flexibility, crowding out investment, and triggering inflationary pressures.

However, the U.S. experience also highlights an important contrast: EU member states confront far more stringent fiscal and institutional constraints than the United States, including the Stability and Growth Pact, supranational oversight through the European Commission and Council, and the absence of a fully centralized fiscal authority capable of providing large-scale countercyclical support. As a result, debt accumulation in the EU carries distinct political and economic implications. Member states cannot rely on the same degree of federal backstopping, face external surveillance of national budgets, and must navigate electorates that increasingly attribute responsibility not only to national governments but also to EU institutions. These features make the study of public debt particularly salient in the European context, where fiscal pressures interact with questions of democratic legitimacy, electoral engagement, and multilevel governance.

While debt management falls primarily to national governments, economic governance in democratic systems is not exclusively the domain of technocrats – it is a collective task shaped by citizen engagement. In this view, elections are a critical mechanism through which voters influence fiscal and economic decision-making. Ideally, electoral outcomes reflect this dynamic, with voters acting as economic stakeholders who help shape the trajectory of national economic governance. This model presumes a unitary authority structure in which responsibility for economic conditions is clearly attributable to the national government. In reality, contemporary governance is increasingly multi-levelled. Institutions like the European Union now wield significant authority over economic policy, particularly in areas like monetary policy, fiscal rules, and budgetary oversight. This dispersal of authority complicates a core assumption of economic voting: that voters can accurately identify which political actors are responsible for economic outcomes. Attribution becomes blurred when national governments operate under supranational constraints, or when EU institutions shape the fiscal environment within which domestic policymakers must act. Moreover, despite extensive scholarship examining how macroeconomic indicators like inflation, unemployment, or growth influence vote choice, far less attention has been paid to how these conditions affect voter turnout – and even less to how public debt, a slow-moving but politically consequential fiscal variable, shapes citizens’ decisions to participate. By interrogating the link between public debt and turnout within this multi-level governance structure, this study speaks to an underexplored dimension of economic voting and democratic accountability.

This paper addresses that gap by examining how voters respond to government debt in both national and supranational elections in EU member states. It investigates whether debt levels influence voter turnout, and whether this relationship differs across institutional scales and degrees of integration within the European Union. Given the current context of rising populism, Brexit, and persistently low electoral participation, understanding how fiscal pressures shape democratic engagement is both timely and policy-relevant. By focusing on debt – a complex, long-term economic indicator – this research contributes to broader debates on political accountability, democratic legitimacy, and the evolving contours of economic citizenship in multilevel governance systems. In short, it asks: Do voters distinguish between national and supranational responsibility when evaluating economic performance, and how does this attribution affect their decision to participate in elections?

2. Literature Review

A central ambiguity in modern democracies – particularly those embedded in multi-level governance systems like the European Union – is how voters attribute responsibility for economic performance. When fiscal authority is divided between national and supranational institutions, the lines of accountability become blurred. This raises a critical question: in light of fiscal pressures, does this attributional ambiguity shape not only vote choice, but the more fundamental decision to participate in elections at all?

A longstanding strand of research in comparative political behavior examines how economic conditions shape citizens’ willingness to participate in elections. Within this literature, Radcliff offers one of the earliest and most influential arguments linking macroeconomic performance to voter turnout. He frames economic downturns as a challenge to democratic responsiveness, contending that adverse conditions suppress turnout by discouraging “potential punishers” from participating (Radcliff, 1992, 444). This highlights a core assumption in economic voting theory – that political dissatisfaction naturally translates into higher electoral engagement – but overlooks the possibility that economic distress itself constrains political participation. This paper examines whether government debt affects not just political sentiment, but actual political participation.

The economic voting literature traditionally focuses on vote choice, where elections serve as mechanisms of retrospective accountability (Merrifield, 1993, 657). It builds on the reward-and-retribution model, which defines elections as the tool by which voters endorse or penalize incumbents based on economic performance. A smaller but growing body of work examines the decision to vote at all. In his seminal work “Economic Adversity and Voter Turnout,” Rosenstone offers a dynamic model of turnout, arguing that economic hardship depresses participation by raising opportunity costs, fostering alienation, and shifting voter priorities toward material survival. Rosenstone’s approach echoes the pocketbook voter hypothesis, which assumes that individuals base political decisions on their immediate financial well-being. Other authors argue that economic adversity can serve as a mobilizing force, driving voter participation by heightening political engagement. Burden and Wichowski, for example, find that economic downturns increase turnout as citizens seek to hold leaders accountable and demand corrective policies (Burden & Wichowsky, 2014, 887). However, their study, like Rosenstone’s, focuses primarily on unemployment rates, reinforcing the literature’s tendency to privilege short-term, tangible economic indicators over more abstract macroeconomic conditions such as public debt. This paper adds to that tradition by introducing government debt as a longer-term, lower-salience economic factor, and asking whether it operates differently depending on the level of government responsibility. This allows for the possibility of heterogeneous effects across levels of governance.

Understanding how macroeconomic precarity influences turnout requires understanding how voters assign responsibility for those conditions. In the EU context, this is particularly complex due to the intertwined national and supranational governance structures. At the national level, member states retain formal sovereignty and maintain their own governments, constitutions, legal systems, and many public policies. However, in key areas—especially monetary and fiscal policy—competence is partially delegated upward to EU institutions such as the European Commission, the European Central Bank, and the Council of the EU. This pooling of authority constrains the fiscal autonomy of member states, meaning that national policymakers cannot fully control economic conditions within their borders. As a result, citizens may struggle to disentangle which level of government is responsible for macroeconomic performance. When decisions are made at the supranational level, democratic accountability can become more diffuse. Citizens often feel further removed from EU decision-makers, and this institutional distance can foster perceptions of alienation from policy outcomes enacted in Brussels. Hobolt & Tilley show that voters struggle to differentiate between national and supranational authority, often directing their anger towards the EU rather than local politicians. Specifically, they write that in the multilevel system of the EU, citizens face the additional challenge of distinguishing between the powers of multiple levels of government. They show that this effect is particularly strong for fiscal policy. This has direct implications for turnout: if voters perceive no clear locus of accountability, they may disengage altogether. Alternatively, they may enact rewards and retribution towards the ‘wrong’ political actors. This paper uses cross-level electoral variation to examine how attributional ambiguity shapes the relationship between debt and voter turnout.

Relatedly, second-order election theory suggests that European Parliament elections are often viewed as lower-stakes contests, where voters express discontent or abstain altogether (Reif & Schmitt, 1980). This logic assumes that voters apply similar evaluative criteria to both arenas. However, if economic conditions, such as government debt, are interpreted differently depending on perceived responsibility, this challenges the second-order model’s assumption of apathy alone. When citizens perceive high public debt as stemming from EU-imposed fiscal constraints or supranational economic governance, abstention or protest voting may be directed toward the European level. Conversely, when debt is attributed to domestic fiscal mismanagement, national elections may become venues for mobilization. This is especially important in the context of the mobilization hypothesis: Garrick et al. (2014) suggest that voter engagement increases when elections are perceived as having greater political significance. Their findings imply that economic distress, amplified by rising debt burdens, can heighten the perceived stakes of elections—but not uniformly across levels of governance. As a result, public debt may either depress turnout (where responsibility seems diffuse and agency limited) or mobilize participation (where accountability appears clear). Recognizing this variation can help us understand how debt-related economic precarity shapes electoral behavior in multilevel systems such as the EU.

3. Hypothesis

This paper investigates whether voters attribute fiscal pressures – specifically, public debt – to national governments or supranational institutions such as the European Union, and whether this attribution influences their decision to participate in elections. The existing literature suggests two competing theoretical mechanisms:

• Withdrawal Hypothesis: High public debt signals economic mismanagement, undermining political trust and fostering alienation. Under this framework, elevated debt levels are expected to depress voter turnout.

• Mobilization Hypothesis: High debt signals economic crisis or instability, heightening public concern and prompting voters to engage politically in hopes of influencing corrective policy. Here, debt is expected to increase turnout.

Importantly, the applicability of these mechanisms may vary across levels of governance. If voters primarily hold national governments responsible for fiscal management, the relationship between debt and turnout should be more pronounced in national elections. Conversely, if voters perceive fiscal constraints as the product of supranational oversight – particularly EU fiscal rules and austerity measures – then this relationship may be stronger in European Parliament elections. Alternatively, if voters struggle to assign responsibility clearly in a multi-level governance system, the effect of debt on participation may be weakened or inconsistent. These considerations motivate the following hypotheses:

• Hypothesis 1: High levels of debt have no significant effect on turnout in national elections, as voters may not associate long-term fiscal indicators with national-level policy responsibility. The diffuse, abstract nature of debt renders it less salient in elections typically dominated by more tangible concerns, like inflation or unemployment.

• Hypothesis 2: High levels of public debt are associated with higher turnout in European Parliament elections, as voters turn to supranational institutions they perceive as more capable of managing broad, long-term macroeconomic challenges.

• Hypothesis 3: The relationship between public debt and turnout varies across EU member states - likely stronger in post-crisis economies like Greece or Spain, where debt is linked to supranational governance, and weaker in fiscally stable countries like France or Germany.

4. Data

1. Data Sources

To test these hypotheses and examine the relationship between public debt and voter turnout, I used four data sources to construct a panel dataset containing data on 27 EU member states from 2000 to 2023. The following sections describe each source and how it was used.

Dependent Variable: Voter Turnout Data

Data on voter turnout was obtained from the Constituency-Level Elections Archive (CLEA), which provides consistent turnout figures for both National Parliamentary Elections and European Parliamentary elections across all 27 EU member states. To ensure comparability, I only used first-round turnout figures, as second-round data were inconsistently reported and may reflect strategic participation instead of baseline voter engagement. Turnout is measured as a proportion between 0 and 1 (e.g., 0.90 indicates 90% turnout).

Independent Variable: Government Debt

Public debt data, expressed as a percentage of GDP, was sourced from Eurostat, the European Union’s official statistical office. The variable reflects the general government gross debt-to-GDP ratio at current market prices, in accordance with the Treaty on the Functioning of the European Union. This measure includes all consolidated debt – currency and deposits, debt securities, and loans – across central, state, and local governments, as well as social security funds. Because excessive debt may raise concerns about repayment capacity, crowd out private investment, or trigger inflationary pressures, it serves as a meaningful indicator of fiscal stress. In this context, the key question is whether voters perceive these macroeconomic conditions as politically salient—and whether they link them to national or supranational institutions.

Control Variables: Economic and Demographic Indicators

To account for other structural factors that influence turnout, I include a range of economic and demographic controls: GDP growth (rate at which the monetary values of all goods and services produced within a country's borders grows from year to year), consumer price index (CPI), unemployment, education, net earnings, and the Electoral Democracy Index (EDI).

• Economic Indicators: GDP growth, CPI, and unemployment data were drawn from the World Bank. GDP growth is measured as the annual percentage change in real GDP. The CPI is indexed to a 2010 base year, and unemployment is expressed as a percentage of the total labor force. Including these controls ensures that the estimated effect of public debt is not conflated with broader macroeconomic fluctuations that may also influence political behavior.

• Socioeconomic Controls: Education and net earnings data come from Eurostat. Educational attainment is measured as the percentage of individuals aged 25–64 with tertiary education, capturing a key determinant of civic engagement. Net earnings refer to the average annual income of a single, childless individual earning 100% of the national average, measured in Purchasing Power Standards (PPS) to ensure cross-country comparability. Both variables help account for socioeconomic status and life conditions that may shape an individual’s likelihood of voting.

Control Variable: Institutional Quality

Finally, I include the Electoral Democracy Index (EDI) from the Varieties of Democracy (V-Dem) project. The EDI captures the integrity, fairness, and inclusiveness of electoral processes, serving as a proxy for institutional quality. Given the diversity of democratic institutions across the EU – from consolidated systems like France and Germany to newer democracies like Hungary and Poland – this index adjusts for cross-country variation in the strength and legitimacy of electoral systems. It also helps account for broader political and institutional factors that could mediate the relationship between debt and voter turnout.

________________

2. Exploratory Data Analysis

Table 1: Descriptive Statistics

Table 1 presents descriptive statistics for the key variables included in the analysis. The data reveal substantial variation across economic indicators and voter turnout, reflecting the heterogeneity of the EU’s political and economic landscape. The average government debt-to-GDP ratio stands at 60.25%, with a wide range spanning from 3.9% to 209.4%, indicating substantial fiscal disparities among EU countries. Voter turnout also varies considerably, with national elections exhibiting higher participation (mean = 66.72%) compared to European Parliament elections (mean = 46.45%). The Electoral Development Index (EDI), which measures the quality of electoral systems, has a high average of 0.84, though substantial variation across countries reflects differing electoral system maturities within the EU. Economic indicators display expected variation, with GDP growth ranging from a severe contraction of -14.8% to a rapid expansion of 24.5%, further underscoring economic disparities. The Consumer Price Index (CPI), indexed to the base year 2010, reveals substantial inflationary pressures across the period from 2000 to 2023. These descriptive patterns point to the need for fixed effects in the analysis to account for country- and year-level differences.

A correlation matrix was generated to examine the pairwise relationships between all variables (Table 3). It inherently accounts for differences in scale between variables by dividing the covariance by the product of the standard deviations of the two variables. As a result, there's no need to standardize the variables beforehand.

Table 3: Pairwise Correlation Matrix 

The analysis reveals several factors influencing voter turnout in both national and EU parliamentary elections. A strong positive relationship exists between net earnings and voter turnout, with higher economic prosperity linked to increased political engagement, as reflected in the significant correlations with both EU (0.720) and National Turnout (0.661). Education also positively correlates with turnout, with moderate associations to both EU (0.344) and National Turnout (0.265), suggesting that more educated populations are more likely to participate in elections. Unemployment shows weak negative correlations with turnout (-0.152 for EU, -0.258 for national), aligning with the economic voting literature that suggests economic hardship may dampen voter engagement. While CPI and debt exhibit subtle correlations with voter turnout, their effects are relatively weak, indicating that inflation and government debt do not significantly influence participation in this dataset. Interestingly, debt is moderately correlated with unemployment (0.545), implying that higher debt burdens may contribute to economic strain and rising unemployment, potentially reducing political engagement. Ultimately, these results underscore the complex interplay of economic factors – especially earnings, education, and unemployment – in shaping voter turnout.

One concern is that Debt and GDP are highly correlated, potentially introducing multicollinearity into the model. Indeed, the correlation matrix reveals a coefficient of -0.133 between debt and GDP. which is fairly weak, but still deserves our attention. While multicollinearity does not bias the estimates, it can inflate standard errors, making the true effects appear statistically insignificant. Estimates are not biased because the regression model remains correctly specified, with an error term mean of zero and regressors unrelated to the error term. Standard errors however may be inflated because when regressors are correlated, they contain overlapping information about the outcome and the model struggles to attribute this variation to one regressor or the other. This supports the interpretation that initial observed insignificance may be due to imprecision rather than the absence of an effect.

Despite these relationships, there initially seems to be little systematic association between debt and turnout, as evidenced by Figures 1 and 2.

[Reference: Figure 1]

Figure 1: Scatterplot of National Turnout and Public Debt

[Reference: Figure 2]

Figure 2: Scatterplot of EU Parliamentary Turnout and Public Debt

While the exploratory data analysis above – including the correlation matrix and the scatterplots – relies on raw, unstandardized variables for clarity and interpretability, all subsequent regression models use standardized versions of continuous variables. This ensures comparability across variables with differing units and scales and reduces the risk that high-variance variables dominate the analysis. Standardization also improves interpretability of regression coefficients and facilitates valid cross-variable comparisons. The transformation was done via Stata’s standardize command, applying the formula:

Standardized Value = (Value - Mean of Variable) / (Standard Deviation of Variable)

Table 2: Descriptive Statistics  for Standardised Variables

3. Clustering the Data

Given the diversity among EU member states, I employed k-means clustering to group countries based on their economic, demographic, and electoral characteristics. This unsupervised learning method was chosen to identify natural clusters of countries with similar profiles. The optimal number of clusters was determined using the Calinski-Harabasz (CH) index, which evaluates the quality of clustering by comparing the variance within clusters to the variance between clusters. This method produced two clusters (CH-index = 8.83):

• Group 0 (group0 == 0): Austria, Belgium, Denmark, Finland, France, Germany, Italy, Luxembourg, Netherlands, Sweden.

• Group 1 (group0 == 1): Bulgaria, Croatia, Cyprus, Czechia, Estonia, Greece, Hungary, Ireland, Latvia, Lithuania, Malta, Poland, Portugal, Romania, Slovakia, Slovenia, Spain.

  
  

These clusters were then analyzed to assess whether different groups of countries exhibited significant differences in their political and economic profiles (Table 4).

________________

Table 4: Summary Statistics by Group (non-standardized variables)

Group 0

Group 1

Group 0 exhibits a higher Electoral Development Index (EDI) with a mean of 0.884 (SD = 0.021) compared to Group 1’s mean of 0.815 (SD = 0.089), indicating stronger electoral institutions and less variability. Voter turnout also differs significantly; Group 0 demonstrates greater participation in both EU parliamentary elections (mean = 0.57 vs. 0.4) and national elections (mean = .788 vs. .612), reflecting a more robust democratic culture. Economic indicators further underscore the contrast: Group 0 shows a slow but stable GDP growth (mean = 1.55, SD = 2.555), while Group 1 experiences faster growth but more extreme fluctuations (mean = 2.993, SD = 4.369), and a lower debt-to-GDP ratio (mean0 = 66.499 vs. mean1 56.57). CPI inflation is higher in Group 0 (mean = 102.978) compared to the more stable mean in Group 1 (100.677), but this difference is not statistically significant (p =0.0537). Unemployment is notably lower in Group 0 (mean = 6.735, SD = 2.177), while Group 1 experiences higher unemployment (mean = 9.281, SD = 4.879), with more pronounced volatility. Additionally, Group 0 benefits from better socioeconomic conditions, with higher education (mean = 31.581 vs. 25.933) and net earnings (mean = 24073.77 vs. 13937.42). Overall, Group 0 exhibits stronger democratic institutions, more stable economies, and higher participation rates, while Group 1 faces greater volatility, weaker institutions, and lower levels of engagement and economic performance. These findings suggest that democratic strength, economic stability, and voter participation are interlinked, with Group 0 showing more favorable conditions for political and economic stability. Other factors that could have influenced this turnout are electoral system design, ease of registration and voting procedures, and level of trust in political institutions.

4. Handling Missing Data

My data set contained two types of missing data: missing data in voter turnout in EU and National elections – which is normal due to the cyclical nature of election cycles – and missing data for Net Earnings (51 missing observations) and Education (6 missing observations). I selected turnout data as the focus of my missing data analysis because turnout is not randomly assigned. Certain individuals and groups are systematically more or less likely to vote. If these differences correlate with key explanatory variables (e.g., economic conditions, political engagement), then analyses of turnout may suffer from selection bias. By creating inverse probability weights, we adjust for differences in the probability of voting, ensuring that our results are not driven by non-random turnout patterns. This allows us to estimate relationships between variables (such as debt and turnout) in a way that better represents the full electorate rather than just those who are more likely to vote. Additionally, we know that countries that joined the EU after 2000 - like Bulgaria which joined in 2007, or Croatia which joined in 2013 - are underrepresented in EU turnout data because of structural differences in voting eligibility. A report from the Center of European Reform (CER) shows that structural adaptation (electoral registers, voting information, etc.) often lag in new members, creating a structurally lower turnout rate in early elections (Grant & Csaky, 2025). To make our sample more representative of the actual electorate, we use inverse probability weighting and give more weight to underrepresented observations. Finally, if education, income, or political attitudes disproportionately affect national turnout more than EU turnout, weighting helps balance these characteristics across groups, making the comparison more meaningful.

5. Methodology

1. Regression Model Specification

To analyze the impact of government debt on voter turnout, I will begin by treating the dataset as a single cross-section, performing two pooled OLS regressions with standardized variables. Following this, I will run models that account for both identity and time fixed effects. This approach allows for a more detailed understanding of the relationship between government debt and voter turnout. Since the regressions will be conducted separately for Group 0 and Group 1, and with both fixed effects included, this will result in a total of eight regression tables.

(3) National Parliamentary Turnoutit= α+ β⋅Government Debtit+ γ⋅Xit+i +t+  it

(4) EU Parliamentary Turnoutit  = α+ β⋅Government Debtit+ γ⋅Xit +i +t+  it

• Xit : Control Variables (GDP, CPI, Unemployment, EDI, Net Earnings, Education) 

• i: Identity Fixed-Effects 

• t: Time Fixed-Effects

The fixed-effects model operates under the assumption that entity-specific fixed effects are correlated with the independent variables. This aligns with the economic intuition that a country's time-invariant characteristics are likely to influence the level of government debt it accrues. This assumption makes the fixed-effects model more appropriate for this analysis than the random-effects model, as the random-effects model relies on the assumption that the entity-specific effects are uncorrelated with the independent variables, which could make it produce biased estimates in the debt/turnout context. Finally, the fixed-effects approach is particularly valuable because it addresses omitted variable bias by effectively "differencing out" these unobserved, time-invariant effects.

6. Results

Table 6 presents the results of the pooled OLS regression examining the relationship between public debt and voter turnout across a sample of 27 countries, combining Group 0 and Group 1. All variables are standardized to ensure consistent magnitudes and comparability of the coefficients. The table reports two specifications: Model 1, which is a pooled OLS regression with controls, and Model 2, which incorporates both time and identity fixed effects to control for unobserved heterogeneity. Model 1 uses the weights “weight_national” and “weight_eu,” while Model 2 uses “avg_weight_national” and “avg_weight_eu” to ensure inter-country consistency.

Table 6: Pooled OLS Regression (Group 0 & Group 1 Combined)

The results suggest that debt does not have a statistically significant impact on voter turnout at either the national or EU level in any specification. In Model 1 (without fixed effects), debt appears to be positively associated with both national and EU turnout, but these relationships are not significant. After incorporating fixed effects in Model 2, the coefficient for debt on national turnout turns negative (-0.167), while it remains positive for EU turnout (0.126), but neither reaches statistical significance.

This lack of significance indicates that debt alone is not a decisive factor influencing voter turnout when controlling for other economic and institutional factors. However, while debt may not directly mobilize voters, it still contributes to broader perceptions of economic performance, institutional legitimacy, and policy pressure – all of which shape turnout. The inclusion of fixed effects also suggests that unobserved heterogeneity – such as country-specific political or economic conditions – may be masking any potential direct effect of debt. Furthermore, the relatively small sample size (133 observations for national elections, 101 for EU elections) may limit the statistical power needed to detect more subtle relationships.

Model 2 provides stronger evidence by accounting for unobserved heterogeneity. GDP growth significantly reduces national turnout (-0.146, SD = 0.0490, p < 0.01), suggesting that economic stability pacifies voters, weakening the mobilizing effect of economic downturns. Inflation also negatively impacts national turnout (-0.335, SD = 0.188, p < 0.1), reinforcing the "pocketbook voter" hypothesis, but is insignificant at the EU level. This reflects a broader feature of the European political landscape: economic grievances do not translate uniformly across governance levels. EU elections typically follow a second-order structure in which voters attribute primary economic responsibility to national governments rather than EU institutions.

Consequently, macroeconomic indicators that influence participation domestically often have weaker or indeterminate effects in EU contexts. Similarly, higher electoral development (EDI: -0.215, SD = 0.0723) appears to foster complacency, reducing national turnout. Conversely, net earnings significantly boost EU turnout (0.852, SD = 0.335, p < 0.05), indicating that financial security enhances engagement in supranational elections. Education positively influences national turnout (0.291, SD = 0.102, p < 0.01), emphasizing its role in fostering civic participation. These findings highlight the economic and institutional drivers of turnout at national and EU levels.

________________

Table 7: Disaggregated Pooled OLS by Group (Group 0 vs. Group 1)

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Table 7 presents the results of Model 1, an OLS regression analysis examining the relationship between economic and electoral factors and voter turnout, disaggregated by Group 0 and Group 1 countries. Group 1 consists of smaller, more vulnerable, and less stable economies, which exhibit distinct patterns of voter participation that are largely shaped by economic conditions and electoral development. When breaking the analysis down by country groups, more nuanced patterns emerge:

Debt and National Turnout: In Group 0 (more stable economies), the relationship between debt and national turnout remains positive but insignificant (0.213, SD = 0.248). In Group 1 (less stable economies), the coefficient is much smaller (0.0262, SD = 0.0850), indicating little to no relationship between public debt and national turnout in these contexts. The stronger (though still insignificant) coefficient of 0.213 in Group 0 suggests that in wealthier economies, higher debt levels might weakly correlate with increased voter mobilization – possibly due to concerns over fiscal policy.

Debt and EU Turnout: Interestingly, debt appears to have a much stronger (though still statistically insignificant) positive relationship with EU turnout in Group 0 (0.578, SD = 0.440) compared to Group 1 (0.0722, SD = 0.0899). This suggests that in wealthier economies, higher debt may trigger increased participation in EU elections, potentially due to concerns over EU-level economic governance or fiscal policies. However, the large standard errors make it difficult to draw definitive conclusions. Further tests that expand the panel dataset across time could help reduce standard errors and smooth noisy series.

In Group 1, economic hardship – measured by CPI, unemployment, and low electoral development – depresses national turnout, while higher net earnings significantly boost participation (0.609, p < 0.01), underscoring the mobilizing effect of financial stability. In contrast, Group 0 countries, with stronger economies, exhibit greater turnout volatility. Here, while unemployment negatively affects participation, EDI becomes a positive predictor (1.099**, SD = 0.438), suggesting that in developed democracies, institutional strength fosters engagement. This refines earlier findings, showing that the negative EDI-turnout relationship in Table 6 was largely driven by Group 1’s weaker democracies.

At the EU level, turnout in Group 1 remains sensitive to economic shifts, with CPI reducing participation (-0.398***, SD = 0.144) and net earnings increasing it (0.631***, SD = 0.203). In contrast, Group 0 countries show no significant EU-level effects, suggesting their voters perceive national economic conditions as more detached from EU governance.

These findings suggest that Group 1 countries, as newer and more economically fragile EU members, exhibit greater reliance on supranational governance, making their turnout more responsive to economic fluctuations. Meanwhile, Group 0 countries, with stronger institutions and economic stability, experience less economically driven voter behavior, particularly at the EU level.

________________

Table 8: Fixed Effects Model by Group (Group 0 vs. Group 1)

Robust standard errors in parentheses

*** p<0.01, ** p<0.05, * p<0.1

Debt and National Turnout: The effect of debt on national turnout remains insignificant in both groups, suggesting that after controlling for time-invariant country characteristics, public debt does not systematically drive national electoral participation. However, in Group 0, the sign remains positive (0.128), whereas in Group 1, it is close to zero (0.0123). This reinforces the earlier observation that in wealthier countries, concerns about debt may marginally influence national turnout, but the effect is too weak to reach significance. Additionally, EDI exhibits a significant negative association with national turnout in Group 0, reinforcing earlier findings that high levels of electoral development may inadvertently foster civic complacency, reducing voter participation. A similar trend is observed in Group 1, where EDI remains a strong negative predictor of turnout (-0.267***, SD = 0.0874). This effect is further compounded by the adverse impact of high inflation (CPI: -0.453***, SD = 0.133) on national turnout. However, economic stability appears to counterbalance these trends, as both higher Net Earnings (0.605**) and Education (0.442*) are positively correlated with increased voter participation, suggesting that financial security and educational attainment play key roles in mobilizing voters.

Debt and EU Turnout: A striking result emerges in Group 0, where debt is now negatively associated with EU turnout (-0.568**, SD = 0.198, p < 0.05). The change suggests that in Model 1, variation in debt was partially capturing other country-specific factors that were not explicitly controlled for. Model 2 corrects for these influences, isolating the independent effect of debt on turnout. This new result suggests that in wealthier economies, higher debt levels may discourage participation in EU elections, possibly reflecting voter disillusionment with supranational economic governance. These patterns align with earlier arguments regarding responsibility attribution within multilevel governance systems. In wealthier democracies, voters may hold supranational institutions accountable for fiscal constraints, reducing the perceived value of participation, while in less stable contexts, debt may mobilize voters who view the EU as a corrective to domestic economic mismanagement. In contrast, in Group 1, debt is positively correlated with EU turnout (0.368*, SD = 0.201, p < 0.1), indicating that in less stable economies, higher debt may lead to greater electoral engagement at the EU level. This could reflect reliance on EU institutions for financial support or dissatisfaction with national-level economic management.

It's important to note how debt coefficients shift sharply from Model 1 (pooled OLS) to Model 2 (fixed effects), highlighting the impact of unobserved heterogeneity.

Table 7: Disaggregated Pooled OLS by Group (Group 0 vs. Group 1) 

Table 8: Fixed Effects Model by Group (Group 0 vs. Group 1)

The most noticeable difference between the two specifications is that, for Group 0, the coefficient on debt jumps from being positive but insignificant in Table 7 (0.578) to being negative and significant at the 5% level in Table 8 (-0.568**). There are a number of possible explanations for these diverging coefficients. First, Table 8 presents a regression with fixed effects, controlling for time-invariant country-specific factors like political culture and institutional strength, which may alter coefficient magnitudes. The shift from 0.578 (Model 1, OLS) to -0.568 (Model 2, FE) suggests that failing to account for these fixed factors in Model 2 biased the debt coefficient upwards. The dramatic shift from a positive to negative coefficient also suggests that in Model 1, some variables may have been positively correlated with both debt and EU parliamentary turnout (fiscal policy, political trust…) that were confounding our estimations. Once fixed effects were included (Model 2), the true within-country effect of debt emerges as negative, meaning that for a given country, when debt increases over time, EU parliamentary turnout tends to decline. This could be due to political disengagement, austerity fatigue, or declining trust in supranational governments. The key takeaway is that fixed effects uncover a strong negative impact of debt on turnout in Group 0, which was obscured in pooled OLS models. This suggests that debt-induced political disengagement is a within-country phenomenon rather than a cross-country effect.

To test which countries (if any) are driving the negative relationship between debt and turnout, I iteratively drop each country in Group 0 from the regression and observe how the coefficient on debt changes.

For most countries in Group 0, removing them from the regression results in a relatively small change in the coefficient on debt, typically between -0.10 and -0.15 When Denmark is dropped, the coefficient on debt changes to -0.431, and when Germany is dropped, it changes to -0.427. These small shifts suggest that Denmark and Germany are not strongly influencing the relationship between debt and EU Parliamentary Turnout. Four countries stand out as main drivers of the relationship between debt and the outcome variable, because their removal results in much larger changes in the coefficient:

(1) Austria: The coefficient on debt becomes -0.381, indicating that Austria is contributing to the negative relationship between debt and the outcome, but its removal moderately weakens the negative association between debt and turnout by 0.187 percentage points.

(2) Belgium: The coefficient on debt becomes -0.361, again showing a similar moderate change, meaning Belgium is also playing a significant role in creating a negative relationship between turnout and debt.

(3) Italy: The coefficient changes to -0.335, reflecting Italy's substantial impact on the debt-turnout relationship.

(4) France: The coefficient becomes -0.603, the most extreme shift and the only one to make the coefficient MORE negative, suggesting that France has a particularly strong influence on the relationship between debt and the outcome. It also points to the fact that France could be an outlier benefitting from a “special case” status - not only are its economic and turnout trends not aligned with other Group 0 countries, French voters seem to be even less responsive to Debt than most others in Group 0. This could reflect the high levels of democratic development and stability in the country, as well as a possible trend of high voter participation in response to economic troubles. France’s unique economic or political circumstances might have been weakening the observed negative relationship between debt and turnout in your model. The result indicates that once France is excluded, the debt-turnout relationship among the remaining Group 0 countries becomes stronger and more consistent.

The second noticeable insight is that we observe the opposite effect among countries in Group 1: the coefficient on debt jumps from being almost 0 and insignificant in Table 7 (0.0722) to being much larger and significant at the 10% level in Table 8 (0.368*). This stark jump in magnitude suggests that incorporating fixed effects controls for unobserved heterogeneity reduces a significant amount of omitted variable bias, and therefore produces far more accurate estimates of the impact of debt at the country level. As noted before, it seems that debt impact on EU parliamentary voter turnout is a within-country, rather than a cross-country phenomenon. Two more things: first, we emit a few hypotheses as to why Group 1 countries have a positive association between debt and turnout, in contrast to Group 0 countries. Countries that are in Group 1 - Croatia, Latvia, Romania, Hungary - tend to face greater volatility and unemployment, weaker institutions, and lower levels of engagement and economic performance. These are also countries that are relatively more recent members of the EU, and that have historically relied on and benefited from EU aid and financial stimulus packages. Debt - which is a country-level indicator of economic well being, is also impacted by supranational governance. It may be the case that both Group 0 and Group 1 voters are sensitive to public debt, but if for Group 0 voters it translates with disillusionment with supranational governance (many French citizens - and other members of more stable countries, express criticism towards ‘weaker’ and more reliant member of the EU) and austerity fatigue, Group 1 voters actually trust and turn to supranational bodies of governance for debt management and overall stability support. Another possibility is that Group 1 countries might have a higher proportion of politically active demographics (eg: younger voters, middle-class citizens) who see debt as a critical election issue, while group 0 countries may and do have an older, politically apathetic voter base, where economic hardship reduces the willingness to vote. Lastly, in Group 1, higher debt might increase concerns about EU fiscal policies, leading to greater voter engagement in European elections. In Group 0, voters might feel that EU policies are ineffective in addressing debt, leading to apathy.

Second, we want to observe - like we did for Group 0 - if any countries in particular are driving this positive association between debt and turnout.

For most countries in Group 1, removing them from the regression results in a moderate drop in the coefficient on debt, typically by about 0.2, often making it insignificant. This suggests that the relationship between debt and EU Parliamentary Turnout in Group 1 is more fragile and dependent on the inclusion of all countries. However, a few countries stand out as key drivers of the observed positive association between debt and turnout, as their removal results in much larger changes in the coefficient estimate:

(1) Malta: The coefficient on debt drops to 0.103, and the p-value rises to 0.726, making it highly insignificant. This suggests that Malta plays a role in sustaining the positive relationship, but its removal destabilizes the estimate. The high p-value indicates that the remaining countries do not exhibit a strong debt-turnout relationship.

(2) Poland: After Poland is excluded, the coefficient drops to 0.096, but the within-R² decreases to 0.0776, and the regression’s overall significance weakens. This suggests that Poland contributes to the strength of the relationship, though not as dramatically as other countries.

(3) Spain: The coefficient drops to 0.064, and while the model fit slightly improves (within-R² = 0.1421), the coefficient remains insignificant. Spain is an important contributor to the positive association; its exclusion weakens the estimated effect of debt on turnout.

The results indicate that unlike in Group 0, where some countries (e.g., France, Italy) strongly influenced the debt-turnout relationship in a consistent direction, the relationship in Group 1 is more fragile and heterogeneous. The fact that removing any individual country makes the coefficient insignificant suggests that the positive association between debt and turnout in Group 1 is not robust across all countries but is likely driven by a subset of influential cases. Future analyses could focus on exploring why some countries exhibit this relationship while others do not, potentially by examining country-specific economic conditions, political institutions, or voter attitudes toward debt.

This sensitivity in Group 1 could stem from the small sample size; since there are only 17 countries in Group 1 (62 observations), removing one significantly alters the estimates. Country-specific factors may also be coming into play, with certain countries like Poland or Malta having unique political or economic dynamics that strengthen the observed relationship, while others do not. Finally, there may simply be voter behavior differences across all group 1 countries, making the overall trend unstable.

Lastly we would like to run a regression analysis with interaction terms to identify if other economic variables actually mediate the impact of debt on turnout - focusing on EU parliamentary turnout since that is where we observe a somewhat significant relationship.

________________

Table 9: OLS regression of EU Turnout on Interacted Economic Variables

p<0.05; * p<0.01

The results indicate that net earnings have a strong positive effect on turnout (β = 0.7616, p = 0.002), suggesting that higher earnings are associated with increased participation. Inflation has a significant negative effect (β = -0.3974, p = 0.015), implying that higher consumer prices may discourage turnout. The interaction term between debt and unemployment is also significant (β = 0.0626, p = 0.023), indicating that as unemployment rises, the impact of debt on turnout changes. However, other interaction effects, including those involving GDP, net earnings, EDI, and CPI, are not statistically significant, suggesting that these economic factors do not strongly moderate the relationship between debt and turnout. The model as a whole is significant (F = 3.10, p = 0.0087), indicating that at least some of the included variables contribute meaningfully to explaining EU parliamentary turnout.

The interaction between debt and unemployment shows that labor market conditions fundamentally shape how public debt influences voter turnout. In economies with higher unemployment, like those in Group 1, public debt may be perceived as a crucial policy tool for stabilizing the economy, funding social programs, and supporting job creation. Voters in these contexts may respond positively to higher debt levels, seeing them as necessary government intervention during economic hardship, which in turn boosts political engagement and turnout. Conversely, in low-unemployment economies like those in Group 0, the negative association between debt and turnout suggests that higher debt may be viewed as fiscal irresponsibility or a risk to economic stability, leading to political disengagement or disillusionment with government policies. This difference aligns with the only significant interaction term (Debt × Unemployment, p = 0.023), indicating that where unemployment is high, the negative effect of debt on turnout diminishes or even turns positive. This reinforces the idea that the economic narrative surrounding public debt is not universal – it is conditioned by the broader labor market, where debt can either mobilize voters seeking government action (high-unemployment economies) or discourage participation when fiscal discipline is prioritized (low-unemployment economies).

7. Conclusion

This paper finds that the relationship between public debt and voter turnout is neither direct nor uniform across EU states. Instead, it is shaped by broader macroeconomic conditions, voter perceptions of fiscal policy, and the level of economic stability within a country. Across all models, public debt is not a significant predictor of national turnout, suggesting that while it may be politically salient, it does not independently drive participation once structural factors are accounted for.

However, disaggregating by economic stability reveals a meaningful divergence in how debt drives participation. In more stable economies (Group 0), higher debt is weakly associated with increased turnout in national elections—but decreased participation in European Parliament elections. This suggests that voters perceive rising debt as a domestic issue necessitating national-level political engagement while simultaneously fostering disillusionment with EU governance. Conversely, in less stable economies (Group 1), debt has no meaningful effect on national elections but is positively associated with EU turnout, indicating that voters in more vulnerable contexts may turn to the EU for economic solutions, thus reinforcing supranational engagement.

Interaction effects further illuminate these dynamics. The significant and positive interaction between debt and unemployment suggests that, where unemployment is high, voters view public debt as a stabilizing force that sustains government intervention and social protections. In contrast, in wealthier countries with lower unemployment, debt may be viewed less favorably – either as a sign of fiscal mismanagement or as a future burden on economic stability – resulting in reduced political engagement at the EU level. This finding offers a compelling explanation for why we observe a positive relationship between debt and turnout in Group 1 but a negative relationship in Group 0: in economies where unemployment is high, debt may be perceived as a necessary policy tool, whereas in more stable economies, it may instead generate political disillusionment or complacency.

Nevertheless, these findings must be interpreted with caution. The relatively small sample size reduces statistical power, making it more difficult to detect small but meaningful effects. Additionally, while the use of fixed effects helps to control for unobserved heterogeneity across countries, the possibility of omitted variable bias remains. Other macroeconomic or political factors, such as levels of political trust, public attitudes toward fiscal policy, or the salience of economic issues in election campaigns, could play a critical role in shaping the observed relationships. Moreover, these models primarily capture short-term dynamics, raising the question of whether the political consequences of rising debt may unfold over longer time horizons. Future research should explore whether the effects of debt on turnout are persistent or whether they evolve in response to shifting economic conditions, particularly in the context of economic crises or prolonged periods of fiscal austerity.

Ultimately, these results highlight that the political implications of fiscal conditions depend heavily on the broader institutional and societal environment. In the context of Europe’s ongoing concerns about declining electoral participation, weakening trust in representative institutions, and the rise of rhetoric that challenges both national and EU-level authority, understanding how public debt shapes citizens’ engagement is paramount. By clarifying when and how fiscal pressures influence turnout across different layers of governance, this study contributes to a deeper understanding of the democratic strains facing contemporary Europe.

________________

References

Arcelus, F., & Meltzer, A. H. (1975). The Effect of Aggregate Economic Variables on Congressional Elections. The American Political Science Review, 69(4), 1232-1239. https://www.jstor.org/stable/1955283

Burden, B. C., & Wichowsky, A. (2014). Economic Discontent as a Mobilizer: Unemployment and Voter Turnout. The Journal of Politics, 76(4), 887-898. https://www.jstor.org/stable/10.1017/s0022381614000437

Bureau of Labor Statistics. (n.d.). EMPLOYMENT AND UNEMPLOYMENT AMONG YOUTH — SUMMER 2024. New Release - Bureau of Labor Statistics. chrome-extension://efaidnbmnnnibpcajpcglclefindmkaj/https://www.bls.gov/news.release/pdf/youth.pdf

Cassel, C. A., & Luskin, R. C. (1988). Simple Explanations of Turnout Decline. The American Political Science Review, 82(4), 1321-1330. https://www.jstor.org/stable/1961763

Department of the Treasury and the Bureau of the Fiscal Service. (2012). Understanding the National Debt | U.S. Treasury Fiscal Data. Fiscal Data. Retrieved November 9, 2024, from https://fiscaldata.treasury.gov/americas-finance-guide/national-debt/

Georgian, A. (2024, October 11). French public debt reaches record high: Is France the sick man of Europe? - Talking Europe. France 24. Retrieved November 9, 2024, from https://www.france24.com/en/tv-shows/talking-europe/20241011-french-public-debt-reaches-record-high-is-france-the-sick-man-of-europe

Grant, C., & Csaky, Z. (2025, May 9). Does EU enlargement require voting reform? Centre for European Reform. https://www.cer.eu/insights/does-eu-enlargement-require-voting-reform

Hobolt, S. B., & Tilley, J. (2014). Blaming Europe? Responsibility Without Accountability in the European Union. OUP Oxford.

Hume, D. (1752). Of Public Credit. Hume Texts Online. https://davidhume.org/texts/pld/pc

Kitidi, K. (Director). (2011). Debtocracy [Film]. https://www.youtube.com/watch?v=2SdfGsUDkC4

Merrifield, J. (1993). The institutional and political factors that influence voter turnout. Public Choice, 77, 657–667. https://link.springer.com/article/10.1007/BF01047865

Radcliff, B. (1992). The Welfare State, Turnout, and the Economy: A Comparative Analysis. The American Political Science Review, 86(2), 444-454. https://www.jstor.org/stable/1964232

Reif, K., & Schmitt, H. (1980). NINE SECOND-ORDER NATIONAL ELECTIONS – A CONCEPTUAL FRAMEWORK FOR THE ANALYSIS OF EUROPEAN ELECTION RESULTS. European Journal of Political Research, 8(1), 3-44. https://doi.org/10.1111/j.1475-6765.1980.tb00737.x

Rosenstone, S. J. (1982). Economic Adversity and Voter Turnout. American Journal of Political Science, 26(1), 25–46. https://www.jstor.org/stable/2110837?origin=crossref

Sigelman, L., Sigelman, C. K., & Bullock, D. (1991). Reconsidering Pocketbook Voting: An Experimental Approach. Political Behavior, 13(2), 129-149. https://www.jstor.org/stable/586038?seq=1

World Bank Group. (n.d.). What Explains Global Inflation. World Bank. Retrieved November 9, 2024, from https://www.worldbank.org/en/research/brief/global-inflation

Bertilsson