The uneven impact of inequality on voter turnout in urban and rural Spain

The widening gap between urban and rural populations in terms of political preferences and participation poses a challenge to democracies worldwide (Gimpel et al. 2020; Scala and Johnson 2017; Scala et al. 2015; McKee 2008). Indeed, the growing urban–rural divide in values and priorities is a critical political cleavage (Luca et al. 2023; Cramer 2016; Ford and Jennings 2020; Iammarino 2019; Jennings and Stoker 2019) and is particularly relevant when considering drivers of political participation. While research has explored the link between inequality and voter turnout, it often overlooks the essential role of place of residence. This paper fills this gap by examining how income inequality translates into political participation in small towns and large cities in Spain.

Although inequality across countries has decreased during the last two decades, the within-country gap between the rich and the poor has almost doubled during the same period. The ratio of the average income of the top 10 percent to the bottom 50 percent has increased from about 8.5 to 15 (Chancel et al. 2022). At the same time, global average voter turnout has decreased steadily since the 1990s. From an almost stable 80 percent until the 1980s, it fell in the 1990s to 70 percent and continued its decline to reach 65 percent during the second decade of the twenty-first century (Solijonov 2016). The simultaneous decrease in voter turnout and widening of income and wealth gaps has provoked increasing interest in the association between these variables. This is a critical issue because increasing inequality calls into question the political underpinnings of liberal meritocratic capitalism (Milanovic 2021).

The literature on inequality and voter turnout is characterized by diverging theoretical approaches and mixed empirical results. While some studies, in line with relative power theory (for example, Goodin and Dryzek 1980), find that economic inequality reduces electoral participation, others point to a positive association, as predicted by conflict theory (for example, Meltzer and Richard 1981). A third group finds no significant empirical relation (see, for instance, Stockemer and Scruggs 2012), while others attempt a synthesis by proposing a nonmonotonic relationship with an inverted U shape (Dash et al. 2023). While the association between inequality and political participation has been extensively explored, with varied results, the literature lacks a nuanced understanding of how population size shapes the association.

This paper argues that place of residence significantly influences how income inequality translates to political behavior. By advancing our understanding of the complex interplay between place of residence, income inequality, and turnout, this research contributes to the broader debate on the consequences of rising inequality for democratic societies. It underscores the relevance of economic inclusion, social mobility, and political empowerment for understanding the dynamics of political participation. In line with Dash et al. (2023), we find that in Spain, widening income gaps foster turnout at low levels of inequality. However, there is a tipping point at which further increases are associated with lower electoral participation. In other words, our findings are compatible with an inverted U-shaped relationship. Our work adds to this literature by showing that the shape of this relationship is different for small towns and large cities.

Personal characteristics and demographic differences may drive rural–urban participation splits. Urban residents are more likely to leverage voting to address economic inequalities. Political groups and social movements facilitate the efficient mobilization of diverse urban populations around common grievances, lowering the transaction costs of mobilization.

Indeed, urban areas facilitate the formation of instrumental networks and provide improved access to avenues of participation based on location. Also, globalization and economic hardship have fostered rural resentment (Gimpel and Reeves 2024), which may translate to frustration and political disconnection. In rural and declining industrial communities, for example, people compare themselves with the comfortable lives of urban citizens and feel that politics is skewed toward the latter's interests (Cramer 2016; Gest 2016; Hochschild 2016). Consequently, inequality may affect political participation differently in urban and rural areas.

This paper focuses on Spanish urban and rural areas to assess how inequality influences voter turnout. In Spain, a large and increasing proportion of territory is below critical population density levels, resulting in a downward spiral of diminishing population, aging, and lower economic activity (Pinilla y Sáez 2017). These areas’ socioeconomic and political marginalization has recently gained prominence through the success of platforms that are devoted to the defense of rural life and that achieved symbolic parliamentary representation. The term España vaciada (“Emptied Spain”) has gained political relevance and entered the academic debate (Almendro 2019; Barreira 2021; Di Donato 2019; Fernández 2019; Seco-González, 2020). Despite this success, however, these platforms have emphasized the feeling of exclusion of rural people and the deepening of the rural–urban divide in economic, social, and political terms, a pattern also observed in several other developed countries (Luca et al. 2023; Cramer 2016; Ford and Jennings 2020; Iammarino, Rodríguez-Pose and Storper 2019; Jennings and Stoker 2019). From this perspective, Spain provides a paradigmatic setting to study the political consequences of conflict arising from inequality in urban centers and the role that disillusion, frustration, and disengagement play in rural areas. Results obtained for this country may orient the broader discussion on the inequality-turnout association by framing it within the debate over the rural–urban divide and potentially explaining part of the observed cross-country differences. The prevalence of rural or urban areas and the gap between them may, in fact, provide an explanation of between-country differences.

For the empirical part of this paper, we constructed a panel database on voter turnout and income inequality at the municipal level for 1,260 Spanish municipalities. The database covers the period 2004–16, which includes five general elections and three elections to the European Parliament. Each municipality is classified as rural or urban based on the functional urban area (FUA) classification developed jointly by the OECD and the European Commission. To assess the influence of inequality on turnout, we employ a panel-data regression model that includes an interaction between our primary explanatory variable, the Gini coefficient, and a dummy variable for the type of municipality. The model also incorporates a set of control variables deemed relevant in existing empirical studies.

Our results support the notion of an inverted U-shaped relationship between income inequality and voter turnout, consistent with previous research and compatible with conflict theory at low levels of inequality and relative power theory at higher levels. However, our findings show a marked difference between urban and rural municipalities. First, the upward-sloping part of the curve is significantly steeper for urban municipalities, which supports the idea that conflict is more impactful in those areas. Second, for urban municipalities, the association changes sign at higher levels of the Gini index. Third, for sufficiently high inequality, the negative effect of income gaps is larger for rural areas. Disengagement and decreasing political participation affect rural areas with higher inequality the most.

The rest of the paper is structured as follows. The next section reviews the literature and presents the main hypotheses. Section 3 presents the data and the empirical approach. Section 4 presents the results, and Sect. 4 concludes.

The two competing theories usually employed to explain the relationship between inequality and turnout are the relative power and conflict theories. According to the former, economic inequality reduces electoral participation, which is mainly driven by relative income concerns (Anderson and Barandi 2008; Dahl 2006; Galbraith and Hale 2008; Goodin and Dryzek 1980; Lister 2007; among others). The reason, as expressed by Goodin and Dryzek (1980), is that when economic power is unequally distributed, the poor reduce their participation because they find it difficult to get the issues they care about addressed in the political process. In other words, inequality reduces turnout because it reflects a concentration of wealth (and, therefore, power) in the hands of the rich (Seeber & Steinbrecher 2011), and political decision-makers tend to be more responsive to the interests of the rich (Flavin 2012; Flavin & Franko 2017). However, relative power theory might not fully account for mobilization efforts by lower-income groups. Conflict theory predicts that the relationship has the opposite sign. As argued by Meltzer and Richard (1981), higher inequality deepens social conflict, and individuals at both extremes of the income distribution face incentives to participate. The poor demand more redistribution; the rich attempt to maintain the status quo. As a consequence, more citizens turn out when income gaps widen.

The empirical evidence on this relationship is ample and reviewed elsewhere (see, for example, Geys 2006b or Polacko et al. 2021). Here, we only aim to provide some examples of the different strands of this literature. Concerning relative power theory, there is ample evidence in its favor at the micro level for the US, highlighting the lower participation of people at the bottom of the income distribution. In particular, as, among others, Nguyen and Garand (2007) and Solt (20082010) argue, state-level inequality reduces the probability of turning out to vote. This result has also been confirmed in the case of presidential elections (Galbraith and Hale 2008), in macro-level cross-national studies (Mahler 2008; Lister 2007), and for OECD countries. For example, Anderson and Beramadi (2008) conclude that citizens in more unequal societies are less likely to vote. This strand of the empirical literature is thus compatible with a negative effect of inequality on turnout. The concentration of wealth and power in the hands of high-income groups seems to produce a smaller electorate because low-income people abstain from voting (Seeber & Steinbrecher 2011).

Some studies, however, provide evidence in favor of conflict theory. Brady (2004) offers an analytical perspective and empirical data to understand why lower-income people might decide to increase their political participation. For example, they may favor a government that adopts policies to reduce inequality, and the upper class might increase their participation to avoid any unfavorable change. Oliver (2001) suggests that in suburbia, wider disparities may favor conflict and political struggle. Jaime-Castillo (2009) shows that political engagement is affected by the availability of political alternatives. When party polarization on economic and social issues is high, turnout increases because citizens attempt to avoid undesired results. In the case of Europe, Horn (2011) and Seeber and Steinbrecher (2011) also conclude that higher income dispersion between the middle and lower classes may drive increased turnout.

This article is mainly related to Dash et al. (2023), which argues that conflict theory prevails at low levels of inequality, while relative power theory comes into its own when inequality exceeds a critical threshold. To explain their empirical findings, the authors make use of the implicit transaction cost motives behind the conflict and the relative power models. The cost of coordination for political action is lower for higher-income individuals. However, the smaller gains by the large number of below-median voters may allow them to overcome their coordination costs. This is true, at least, when the cost of redistribution borne by above-median voters is not high. In fact, the cost of redistribution falls increasingly on those with higher incomes. Therefore, the rich tend not to oppose a small amount of redistribution at lower levels of inequality. However, when the cost of redistribution is sufficiently high, above-median voters are incentivized to turn out, and political action by below-median voters is increasingly ineffective, generating frustration and lower turnout. As proposed below, this argument can be extended to understand how inequality translates to political participation in urban and rural contexts.

Voting patterns, partisanship, and political attitudes have been shown to vary between urban and rural areas (Gimpel et al. 2020; Johnson and Scala 2020; Lin and Lunz Trujillo 2023; Lyons and Utych 2023; Parker et al. 2018; Scala and Johnson 2017). The urban–rural divide may be due to personal characteristics, population density, and geographical closeness to a city, which influences social interactions and social capital and, thus, personal values, personality, and political views (Pena-López et al. 2021; Gimpel and Reeves 2024). For example, bigger cities tend to be populated by more educated and diverse people who work in knowledge-based occupations (Maxwell 2019; King 1996; Rentfrow 2010) and are skewed toward more progressive values (Luca et al. 2023). These demographic differences may drive the rural–urban participation split.

Community ties and strong local networks in rural areas may lead to higher initial levels of participation when inequality is low. Social closeness fosters a sense of mutual assistance and social pressure to participate (Geys 2006a). Therefore, turnout should be higher in rural areas and smaller cities, where noncompliance would entail a loss of social prestige or reputation (Riker and Ordeshook 1968; Overbye 1995) and where politics is more personal (Blank 1974; Davis 1991). In contrast, urban areas might have lower initial turnout at low inequality levels, given the more heterogeneous and impersonal nature of cities, where politics is less personal and people are less likely to feel an immediate social obligation to participate.

As inequality increases, urban areas may see a steeper rise in turnout because of the combination of institutional support for mobilization and increased awareness of inequality’s effects. Uban residents are more likely to leverage institutional pathways (for example, voting for redistributive policies) to address economic disparities. Also, cities provide greater access to government services and charity through formalized channels. Therefore, urban residents may view electoral participation as a practical tool for addressing inequality, especially as social safety nets and public services are more accessible and effective in cities (Florida 2017; Lin and Lunz Trujillo 2023).

Urban contexts also represent more fertile ground for conflict and mobilization, reflected, for example, in the greater availability of political alternatives (Jaime-Castillo 2009). The case of the “Emptied Spain” platform, mentioned in the introduction, can potentially be explained by the Olsonian logic of collective action (Olson 1965). While Olson’s theory posits that smaller, homogeneous groups have an advantage in collective action, urban areas can counteract coordination costs through robust institutional networks. These institutions—such as political advocacy groups, social movements, and digital platforms—facilitate the efficient mobilization of diverse urban populations, effectively lowering the transaction costs Olson associated with large, heterogeneous groups. This is especially pertinent in cities with high socioeconomic diversity, where citizens may mobilize around common grievances such as income inequality. Indeed, Geys (2006a, 2006b) finds that studies on turnout fail to support the correlation between population concentration and turnout, and Lin and Lunz Trujillo (2023) observe similar participation rates in urban and rural areas.

For the case under study, a topical example is offered by two major Spanish cities, Madrid and Barcelona, which have witnessed substantial civil society mobilization in response to inequality. After the 2007–2008 financial crisis, urban movements such as Los Indignados (and their conversion into political parties such as Podemos) rapidly gained traction, galvanizing diverse groups to protest against austerity and inequality. This movement was facilitated by urban contexts that enabled rapid spread through social networks and access to centralized protest locations (Lin and Lunz Trujillo 2023). Indeed, campaigning in social networks, a strategy that started to gain importance after Obama’s triumph in 2008 (Cogburn and Espinoza-Vasquez 2011), was instrumental to the impact and success of these movements (Pavía et al. 2016). This circumstance is especially relevant given Spain’s wide rural digital gap (Malgesini et al. 2022). Such examples demonstrate that despite high heterogeneity, urban areas can achieve significant mobilization by tapping into available political infrastructure and networks, a dynamic less feasible in smaller, dispersed rural areas.

In rural areas, the increase in turnout may be more gradual. A culture of self-reliance and the emphasis of rural communities on self-sufficiency (Gest 2016; Cramer 2016) imply that increasing inequality might not immediately lead to political mobilization. Instead, voters may feel that formal political engagement is less likely to address their issues. Also, in contrast to urban areas, rural residents might see inequality in their daily lives less frequently, which can lead to weaker demand for redistributive policies. In fact, studies on urban empathy and engagement under inequality discuss how, in urban areas, empathy might be more activated by visible inequality in public spaces, which is more pronounced and more easily observed because of the density and diversity of urban settings (Okulicz-Kozaryn and Mazelis 2018). Indeed, recent empirical evidence suggests that rural decline significantly shapes electoral outcomes, intensifying spatial differences in voting patterns (Lago and Lago-Peñas 2025).

Once inequality reaches a certain threshold, rural turnout may drop sharply. Gimpel and Reeves (2024) discuss how globalization and economic decline have animated rural resentment. As mentioned, the different composition of rural and urban contexts in terms of personal characteristics has fueled an increasing divide in values and the perception of politics skewed toward the concerns of the educated urban elites (Luca et al. 2023). High inequality may have a multiplier effect on the frustration of people who already feel marginalized and excluded. The lack of institutional support and a tendency toward disenfranchisement in rural communities mean that rising inequality exacerbates feelings of political disillusionment, leading to rapid disengagement. In contrast, the existing support structures and networks in urban areas allow voters to continue viewing political participation as an effective means of addressing inequality.

The rural–urban divide’s implications complement the argument Dash et al. (2023) develop about inequality and political participation. On the one hand, the conflictual nature and lower coordination costs of urban areas imply that widening income gaps may affect political participation for low levels of inequality. This is true for below-median voters, who bear higher coordination costs, and above-median voters, who are more concentrated in those areas. On the other hand, the multiplier effect of rural resentment may materialize when the threshold level of inequality has been reached. In this situation, frustration may provoke a steeper decline in participation in rural areas and smaller cities.

The arguments outlined above lead us to hypothesize that for low levels of inequality, widening income gaps will translate into a steeper increase in voter turnout in large cities compared to small towns. However, when inequality reaches a high threshold, we expect a steeper decline in turnout in rural areas compared to urban areas.

The empirical part of this article studies the relationship between voter turnout (Turnout), measured as the percentage of eligible voters who cast their ballot, and income inequality at the municipal level, measured by the Gini coefficient (Gini), while testing for possible interactions between Gini and the environment in which each municipality is located (Place of residence), namely urban or rural.

Our database covers the period from 2004 to 2016, which includes five Spanish general elections (March 2004, March 2008, November 2011, December 2015, and June 2016) and three elections to the European Parliament (June 2004, June 2009, and May 2014). Data on voter turnout, our dependent variable, are publicly available from the Spanish Ministry of the Interior. Data on income inequality at the municipal level, our key independent variable, are obtained from Hortas-Rico and Onrubia-Fernández (2014). Their database presents yearly statistics for Spanish municipalities with over 5,000 inhabitants that belong to the Autonomous Communities and Cities of Common Fiscal Regime, which accounts for approximately 82 percent of the population each year.¹

Hortas-Rico and Onrubia-Fernández (2014) use fiscal microdata to produce municipal-level estimates from the annual samples of personal income taxpayers to overcome the limitations—in terms of reliability and territorial representativeness—of household surveys. These samples are published by the Institute for Fiscal Studies, a bureau under the Spanish Ministry of Finance, and the State Agency for Tax Administration, which is responsible for extracting the samples from its administrative census registers. Thus, for each year and municipality, the database contains information on mean and median income, measures of inequality such as the Gini and Atkinson indices, and the structure of the personal income distribution. Although the Gini coefficient was selected as the proxy for inequality, employing alternative inequality indicators yields similar results (available upon request).

Using tax data to analyze the income distribution has several advantages over other data sources. First, since the personal income tax is universally applied in Spain, using tax data allows for the collection of microdata on a broader population compared to interview-based surveys. This advantage is particularly noteworthy in our case, as it allows us to obtain representative samples of the income distribution for more than a thousand municipalities, thereby increasing the granularity of our analysis. Second, previous studies (for example, Feldman and Slemrod 2007; Slemrod and Weber 2012; Meyer et al. 2015) highlight that fiscal data are not subject to the issues that plague survey-based data, such as unit and item nonresponse or the possibility of misreporting relevant variables. Also, top incomes tend to be grossly misrepresented in surveys compared to our source.

While tax data offer a large sample size across municipalities, limitations exist. First, income is reported at the individual, not household, level. Second, minimum income thresholds may exclude some individuals. While some low-income households may still file (Hortas-Rico & Onrubia-Fernández 2014), their data might be less accurate. Finally, tax evasion and deductions may cause actual income levels to be underestimated. We acknowledge these limitations, but tax data remain the best option for such a large-scale study.

We employ the OECD and European Commission’s FUA classification to categorize all municipalities in the sample as urban or non-urban. This enables us to test for differentiated behavior between these two groups. According to this taxonomy, FUAs are formed based on population size, density, and commuting patterns rather than arbitrary legal or administrative boundaries (Dijkstra et al. 2019). For this analysis, municipalities are categorized into two groups based on their urban status: those that are part of an urban center or part of an FUA (urban) and those entirely outside of any FUA (rural). This classification results in a total of 660 rural municipalities and 600 urban municipalities grouped into 81 FUAs.

In addition to the core variables—income inequality and place of residence—our model incorporates a set of demographic and socioeconomic factors identified in previous studies as potentially influencing voter turnout. These include population size (expected to have a negative relationship with turnout because smaller municipalities are more homogeneous—Kostadinova & Power 2007; Górecki & Gendźwiłł 2020), median income (expected to have a positive relationship—for example, Steiner 2010), average age and its square (to capture a potentially nonlinear relationship—Dash et al. 2023), percentage of female population (expected correlation unclear because of changing trends—Norris 2002), number of seats allocated in the constituency (expected to have a positive relationship because of increased representation—Jackman 1987), unemployment rate of the province (expected impact on turnout uncertain—Dash et al. 2023), and a dummy variable distinguishing between general and European elections (expected to show lower turnout for European elections—Mattila 2003; Johnston et al. 2007; Matsubayashi & Wu 2012; Braun & Schäfer 2022). Finally, the model includes a set of (17) regional dummy variables to account for potential differences in voter turnout across regions and a time variable to capture a potential long-term trend in voter turnout over the period under analysis, in line with the observations of Solijonov (2016). Table 1 lists the variables used in the study and descriptive statistics.

Because of its importance for the correct interpretation of results, it should be noted that the number of urban and rural municipalities varies greatly across the inequality spectrum. As shown in Table 2, the sample is greatly reduced for Gini coefficients lower than 0.2 and higher than 0.7.

We estimate a random‐effects generalized least squares (GLS) model with municipality‐level intercepts and include region dummy variables to account for time‐invariant differences across regions. Given the potential for intragroup correlation within the municipalities over time, we employ clustered standard errors at the municipality level. This approach adjusts the standard errors to account for potential heteroskedasticity and autocorrelation within each municipality, thereby providing more reliable and robust estimations. The following equation expresses the models being estimated:

Here, \(\beta_{0}\) is the global intercept; \(\gamma_{region\left[ i \right]}\) is a fixed effect (dummy) for each region; V is a vector of control variables; \(u_{i}\) is a random intercept for each municipality i; \(\varepsilon_{it}\) is idiosyncratic error. We estimate three models, which all include the complete set of control variables. The first model only considers inequality in quadratic form (the first term in brackets). Second, we study the effect of adding the place of residence (the second term in brackets). Finally, we include the interaction terms (in the third set of brackets).

The polynomial specification of the inequality variable implies potential nonlinearities of the effects. As a consequence, we follow the best practices proposed by Mize (2019) for estimating, interpreting, and presenting nonlinear interaction effects. In particular, we ignore the coefficient estimates and provide tests for first and second differences in the average marginal effects. This approach allows us to investigate how the impact of income inequality on voter turnout varies depending on the type of municipality (urban vs. rural).

To conduct a comprehensive evaluation of the interactions in the presence of nonlinear relationships and categorical variables, we estimate average marginal effects and test their differences. This approach allows for the approximation of the impact of explanatory variables at varying levels of the interacting variable, which, with the help of graphical analysis, facilitates visualization of changes across the range of the interacting variables.

Finally, we conduct supplementary Mundlak random-intercept panel regressions, controlling for province-level group means of inequality. This approach helps to capture unobserved heterogeneity across groups that might be correlated with the explanatory variables, enhancing our findings’ robustness.

Table 3 presents the results of our random-effects GLS panel regression model.

Model 1 confirms the existence of a nonlinear (inverted U-shaped) relationship between inequality and voter turnout, in line with Dash et al. (2023). The result is robust to the inclusion of the place-of-residence dummy variable (Model 2) and interaction effects (Model 3). In Model 2, a municipality's status as urban is not associated with voter turnout. However, the coefficient of this variable turns out to be significantly different from zero when the interaction terms are included. This result indicates the characteristic relationship between the variables of interest, as discussed below.

The coefficients of the interaction terms are not significantly different from zero (see Model 3). However, to ascertain whether the interactions between municipality category and inequality in its polynomic form are statistically significant, it is not adequate to merely examine the sign, value, or statistical significance of the estimated coefficients in Table 2. Indeed, the magnitude and direction of an interaction effect can vary considerably across independent variables’ values, potentially leading to erroneous conclusions (Mize 2019). Therefore, a post-estimation analysis based on the Wald test is performed to study average marginal effects and their differences. This analysis reveals two clearly differentiated curves for urban and rural municipalities, as shown in Fig. 1.

Bild vergrößern

These findings suggest a complex relationship between income inequality, urbanization level, and voter turnout. The average marginal effect of Gini on voter turnout (0.0817) is positive for urban municipalities but negative for rural municipalities (− 0.0289). To understand this finding, observe that the curves, which both indicate an inverted-U-shaped relationship, differ in both the level of inequality at which they achieve the maximum turnout and their slopes at different levels of inequality. In particular, the level of inequality above which the relationship turns negative is higher in urban municipalities (Gini coefficient slightly above 0.602) than in rural municipalities (Gini = 0.405). Therefore, in urban settings, widening income gaps boost turnout throughout a larger part of the range of the Gini coefficient, peaking around 0.6. However, the estimated marginal effects are smaller in rural settings at low levels of inequality, which suggests that urban voters react more to identical increases in inequality compared to non-urban voters—at least up to a certain level of inequality. The turning point in the relationship between inequality and turnout likely reflects the contrasting dynamics of mobilization in urban and rural areas. In cities, the conflict generated by widening income gaps can be more easily channeled into political participation because of lower coordination costs and the presence of diverse political options. However, in rural areas, high inequality acts as a multiplier on feelings of frustration and disenfranchisement, leading to a steeper decline in turnout when the system is perceived as unresponsive to voter needs.

This result is supported by the tests presented in Tables 4 and 5. Table 4, obtained from the estimation of Model 3, presents the predicted turnout for different levels of the Gini coefficient by place of residence (urban or rural). Figure 1 is the graphical representation of these results. The two curves do not overlap, except at their intersection around the value of the Gini coefficient of 0.4—and at both ends, where the sample size is extremely small (see Table 2). Table 5 presents the results of the Wald test for the first and second differences of the average marginal effects. The results show that the two curves exhibit a markedly dissimilar slope throughout most of the Gini coefficient range, which implies that a similar marginal change in inequality will have vastly different effects depending on the (rural or urban) setting of the municipality. Indeed, the estimated marginal effects for rural municipalities signify lower reactivity to increases in inequality on the upward-sloping segment of the curve.

Our finding suggests that for urban municipalities, turnout increases with inequality until relatively high levels, when it starts to slowly decrease. In contrast, in rural municipalities, turnout slowly increases until a lower threshold is reached and decreases afterward with a steeper slope.

Considering the distribution of municipalities by level of inequality and setting permits us to understand better the contrast between urban and rural settings. As shown in Table 2, about 74 percent of rural municipalities have a Gini coefficient higher than 0.405, while about 97 percent of urban municipalities have a Gini coefficient lower than 0.602. Therefore, for most rural municipalities, increasing inequality is associated with decreasing turnout. In contrast, widening income gaps for almost all urban municipalities translate into higher turnout. In other words, since most urban municipalities have lower inequality than the threshold, we observe a positive relationship between inequality and political participation, a result best explained by conflict theory. In contrast, the relationship between inequality and voter turnout looks consistent with relative power theory for rural municipalities because most of them exhibit higher inequality.

The analysis of control variables yields several interesting observations, mostly in line with the literature. A higher percentage of women in a municipality is associated with increased voter turnout, reflecting the closing of the gender gap in recent decades (Norris 2002; Kostelka et al. 2019). The relationship between age and turnout follows an inverted U shape, peaking around 41 years. This suggests that younger and older voters may be less likely to participate than those in their middle years. Consistent with prior research, larger population size is linked to slightly lower voter turnout (Stockemer 2017). This can be attributed to the perception that one’s vote has a lower impact when the population is large. Conversely, a greater number of seats available in an electoral district leads to a slight increase in turnout, likely because of the chance of a wider variety of parties gaining representation, which can motivate more voters to participate. Our findings also confirm the expected positive correlation between income and voter turnout, with municipalities boasting higher median incomes exhibiting higher participation rates. Furthermore, the economic climate plays a role, as evidenced by lower turnout in Spanish regions with persistently high unemployment rates, suggesting a dampening effect on political engagement. As expected, the dummy variable for European Parliament elections indicates a significantly lower turnout for those elections than national general elections. The former elections are generally viewed as less critical and consistently elicit lower participation. Finally, the time variable reveals a concerning trend: a gradual decline in overall voter turnout in Spain that is not fully explained by the other variables in the model.

Our regression models account for a significant proportion of the variance in the dependent variable, with an overall R-squared value exceeding 89 percent in all three iterations of the model. The within R-squared value indicates that our models account for more than 94 percent of the variation within municipalities over time, while the between R-squared value suggests that around 56 percent of the variation between municipalities is explained by the model’s regressors.

The random-effects GLS specification includes region dummy variables to capture time-invariant regional heterogeneity and a random intercept at the municipality level. This combination helps control for region-level unobserved factors (for example, cultural or institutional differences between Spanish regions) and unobserved municipality-specific factors that are stable over time. The estimated \(\hat{\theta }\) for our model is approximately 0.67, as reported in Table 3. Following Jordan and Philips (2023), this statistic measures the degree of quasi-de-meaning occurring in our random-effects model—in other words, how closely our estimation strategy approaches a fixed-effects (no pooling) model rather than a fully pooled model.

A \(\hat{\theta }\) of 0.67 suggests that about two-thirds of the variation comes from within-municipality variation, while approximately one-third arises from between-municipality variation. That the value of \(\hat{\theta }\) is closer to 1 than 0 indicates that the effects we estimate predominantly reflect changes within municipalities over time rather than differences between municipalities. Thus, our results about the nonlinear (inverted U-shaped) relationship between inequality and voter turnout—and its moderation by the urban–rural divide—are primarily informed by how changes in inequality within each municipality over time affect turnout.

However, that \(\hat{\theta }\) is not extremely close to 1 (full fixed effects) underscores the importance of considering some between-municipality differences. This partial pooling allows the model to retain valuable cross-municipality variation, reinforcing our finding that the structural features of municipalities—such as their urban or rural status—significantly moderate how inequality influences political behavior. In other words, our analysis balances the benefits of controlling for unobserved municipal heterogeneity while still effectively capturing meaningful variation across municipalities.

Overall, the value of \(\hat{\theta }\) supports our methodological choice of combining municipality random effects with regional dummies since this approach retains substantial within-municipality information, which is essential in order to test our hypotheses about the contextual dynamics shaping voter turnout in response to changing inequality.

To strengthen the robustness of our findings, we also conduct supplementary Mundlak random-intercept panel regressions. These models control for province-level group means of inequality, addressing potential unobserved heterogeneity across provinces. The results, presented in the Electronic Supplementary Material, consistently support the main findings.

This paper has examined the complex relationship between income inequality and voter turnout in Spain, focusing on the urban–rural divide. Our results strongly support the central idea that the relationship between inequality and voter turnout is not linear. The inverted U shape found in the data aligns with Dash et al.’s (2023) argument and the concept of a critical threshold of inequality. At lower inequality levels, conflict theory seems dominant, with increasing inequality driving up turnout, possibly because of social conflict and the mobilization of both the rich and poor. When inequality passes a certain point, the relationship becomes compatible with relative power theory. The decrease in turnout at high inequality levels reflects the idea that extreme inequality can disenfranchise and demotivate lower-income voters.

However, the shape of this relationship differs significantly between towns and cities, suggesting that the social and economic contexts in which people live play a crucial role in how they respond to economic disparities. In particular, the varying slopes and peaks of the inverted U shape for urban and rural areas highlight how the same levels of inequality can have drastically different impacts on turnout depending on the context. Rising income inequality initially spurs greater political participation. The finding that urban voters are initially more responsive to inequality aligns with the idea that urban areas can foster political engagement through improved coordination and diverse political alternatives. Urban characteristics such as higher population density, economic diversity, and access to a broader range of political organizations facilitate the mobilization of citizens at low levels of inequality. Higher density makes coordination easier, while economic diversity and political infrastructure offer multiple channels for participation, allowing urban residents to more effectively mobilize around inequality-driven grievances. These elements contribute to the steeper initial increase in urban turnout in response to rising inequality.

In small towns, the initial positive effect of inequality on turnout is less pronounced and diminishes at lower levels of inequality compared to urban areas. The steeper decline in turnout for rural areas and smaller cities at high inequality levels supports the hypothesis that rural resentment and feelings of marginalization can be exacerbated by extreme inequality. Also, we identified distinct thresholds of inequality for urban and rural areas (Gini coefficients of 0.6 and 0.4, respectively), which further supports our main argument: Different contexts may have varying tipping points at which inequality’s impact on turnout shifts.

Some weaknesses limit the external validity of this study and may guide future research. First, our analysis potentially oversimplifies the urban–rural divide. The binary classification of municipalities as urban or non-urban might not fully capture the nuances of urbanization. A more nuanced approach, considering different degrees of urbanization or the trend of population size, could provide a more accurate picture of the complex relationship between place of residence and voting behavior. The use of the FUA classification may introduce selection bias if unobserved characteristics differ systematically between classified and unclassified areas. However, FUA provides a functional distinction between urban and rural areas that considers commuting and population density, offering a more standardized urban–rural categorization compared to traditional administrative boundaries. Future research could explore alternative classification approaches to assess potential biases in the FUA-based classification.

Second, while the model includes several relevant control variables, it could be enhanced by incorporating additional factors such as educational attainment and political polarization. Indeed, unobserved factors might influence both inequality and turnout. Polarization, which we have not considered because of the limitations of our data set, appears especially relevant because of its close association with participation in marginalized areas. Inequality and economic hardship may affect political support for mainstream parties. Therefore, the joint consideration of turnout and participation may reveal critical differences between urban and rural settings. Third, as mentioned in Sect. 3, our data set potentially underrepresents lower-income people.

In any case, future research should delve deeper into the mechanisms underlying the observed differences between urban and rural areas. This line of research could involve investigating the role of social capital, political trust, and attitudes toward government in shaping the relationship between inequality and voter turnout. Additionally, comparative studies across different countries and regions could shed light on the generalizability of our findings and the extent to which the urban–rural divide shapes political behavior in diverse contexts.