Landholding Inequality and Democratization

What is the effect of inequality on democratization? We follow the rich literature that views democracy as an instrument to redistribute resources (Acemoglu and Robinson Reference Acemoglu and Robinson2006; Albertus Reference Albertus2017; Boix and Stokes Reference Boix and Stokes2003; Przeworski Reference Przeworski2009). Given that poor voters outnumber wealthy ones, we should expect that suffrage extension to poor voters will result in more redistribution. Regarding regime change, this literature suggests that low levels of inequality further the democratization process. When societies become more equal, there are fewer redistributive demands from the poor. Thus, the economic and political elites face lower relative costs for tolerating democracy. When the tax elites would have to pay in a democratic system falls below the costs of the repression needed to maintain authoritarian rule, the elites accept democracy. By contrast, the wealthy elites' resistance to democracy increases with rising levels of inequality.

However, an important distinction between mobile and immobile assets needs to be made. Mobile assets can be moved to evade taxation. By contrast, immobile assets such as land cannot easily be shielded from the threat of taxation. Consequently, landholding inequality (or other forms of geographically concentrated resource endowments) should be a better predictor of opposition to democratization than income inequality because the latter captures mobile assets that can, at least to some extent, evade taxation (Ansell and Samuels Reference Ansell and Samuels2014; Beramendi, Dincecco, and Rogers Reference Beramendi, Dincecco and Rogers2019; Boix Reference Boix2003; Engerman and Sokoloff Reference Engerman and Sokoloff2005; Ross Reference Ross2015).

Such arguments about geographically concentrated resource endowments align with a rich tradition in the literature, which emphasizes the uniquely anti-democratic role played by large landholders because of this group's ownership of immobile assets and its economic dependence on a cheap agricultural workforce (Albertus Reference Albertus2017; Bateman Reference Bateman2018; Moore Reference Moore1966; Rueschemeyer, Huber, and Stephens Reference Rueschemeyer, Huber and Stephens1992; Stephens Reference Stephens1989). For instance, exploring a Prussian parliament vote on a reform that would have transformed the highly inegalitarian three-class voting system, Ziblatt (Reference Ziblatt2008, 612) finds that ‘representatives from electoral constituencies marked by high levels of landholding inequality were more likely to vote against democratic reforms’. By contrast, income inequality did not significantly affect the legislators' voting behaviour. Similarly, Samuels and Thomson (Reference Samuels and Thomson2021) show that, with the spread of mechanization throughout the twentieth century, the role of labour-dependent agriculture substantially declined, as did the need for the landed elites to maintain an autocratic regime that repressed agricultural workers.

However, in recent years, scholars have raised doubts about the hypothesized relationship between landholding inequality and democratization. On the one hand, several contributions have pointed to endogeneity concerns, as concentrated resource endowments might also result from autocratic policies. For instance, autocrats might assist in creating large estates to gain elite support (Albertus Reference Albertus2015; Brooks and Kurtz Reference Brooks and Kurtz2016; David and Wright Reference David and Wright1997; Ramsay Reference Ramsay2011; Stijns Reference Stijns2006). On the other hand, some scholars have questioned the role of distributive conflicts altogether. Instead, these contributions highlight the role of other explanations for democratization, such as international pressure, intra-elite conflicts, or incumbents' missteps (Ahlquist and Wibbels Reference Ahlquist and Wibbels2012; Haggard and Kaufman Reference Haggard and Kaufman2012; Houle Reference Houle2009; Treisman Reference Treisman2020).

This paper adds to this vibrant debate by relying on a novel data set and using multiple instruments to explore the relationship between landholding inequality and democratization. However, we must first discuss the mechanism through which landholding inequality might undermine democratization.

Social Control and Mass Political Behaviour

How does landholding inequality impair democratization? The existing literature typically expects local elites to neutralize mass political behaviour and exercise direct control over political processes using electoral fraud (Lehoucq and Molina Jiménez Reference Lehoucq and Molina Jiménez2002; Mares Reference Mares2015; Ziblatt Reference Ziblatt2009) or outright repression (Acemoglu and Robinson Reference Acemoglu and Robinson2006; Boix Reference Boix2003; Wright, Frantz, and Geddes Reference Wright, Frantz and Geddes2015). Instead, for the reasons we explain below, this analysis focuses on how local elites control the political behaviour of the resident population.

This study examines the voting behaviour of male citizens in Switzerland in four binding, direct democratic votes on suffrage extension to most of the remaining male citizens from 1866 to 1877. Importantly, as we show in the appendix, even in areas with high levels of landholding inequality, most adult males were eligible to vote (Lutz Reference Lutz2000, 35–6) and would have been likely to benefit from redistribution. Hence, for landholding inequality – an indicator of the economic interests of local elites – to have the expected democracy-inhibiting effect in direct democratic votes, local elites must have the means to influence the population's voting behaviour.

In general, the literature on inequality and democratization has not paid much attention to the political behaviour of ordinary citizens (one exception is Montalbo Reference Montalbo2023). Instead, most research relies on macro-level comparisons or analyzing roll-call votes in parliament. As Welzel and Inglehart (Reference Welzel and Inglehart2008, 135) note, this literature often simply assumes that the masses always favour democracy. However, mass support for democracy cannot be taken for granted (Caughey Reference Caughey2018; Claassen Reference Claassen2020; Wuttke, Gavras, and Schoen Reference Wuttke, Gavras and Schoen2022).

We argue that high levels of landholding inequality not only incentivize local elites to make the masses vote in line with their economic interests, it also provides local elites with the means to exert direct economic and political pressure on ordinary citizens. Local elites can do so through processes of social control.

Social control refers to rules that keep individuals bound to conventional standards and, in this way, maintain order in a community. It is typically associated with the exercise of power by the individuals who shape these standards and have the means to enforce compliance with them. We are particularly interested in the traditional patterns of social control that do not require constant compliance enforcement because individuals have come to internalize these standards and accept the underlying power dynamics. In this form of traditional authority, individuals accept rules and standards for reasons of tradition and custom because things have always been this way, and they do not question the resulting hierarchies (Ross Reference Ross2009).

A rich literature argues that landholding inequality is related to such traditional patterns of social control (Mares Reference Mares2015; Moore Reference Moore1966; Rueschemeyer, Huber, and Stephens Reference Rueschemeyer, Huber and Stephens1992; Stephens Reference Stephens1989). For instance, Ziblatt (Reference Ziblatt2008, 615) observes that ‘landholding inequality, unlike income inequality, is a proxy for a particularly pernicious and robust form of preindustrial traditional social power in which prestige, power, and wealth are correlated, giving rise to social norms (for example, invidious hierarchy) that undercut democratization’. Put differently, landholding inequality is associated with forms of social organization that are fundamentally at odds with democracy.

Moreover, landholding inequality gives local elites the resources and means to block democratization efforts. Most notably, traditional patterns of social control allow local elites to protect themselves from redistributive demands by shaping the population's political behaviour.Footnote ¹ How does social control work in practice? Following Lukes (Reference Lukes1974), we can differentiate between the three forms of power that allow the landholding elites to exercise social control over the landed masses.

First, in areas characterized by high landholding inequality, the landed masses are often economically dependent on local elites (Albertus Reference Albertus2017; Baland and Robinson Reference Baland and Robinson2008; Montalbo Reference Montalbo2023). Local elites employ landless agricultural workers; lease land to smallholders; offer loans to small farmers at high interest rates, insured by mortgages; and control access to relevant economic infrastructure. For such reasons, the landed masses often have little choice but to support the local elite's political demands.

Second, local elites typically control political offices because financial compensation for public officials is, at least historically, low, while maintaining large supporting coalitions is costly (Ziblatt Reference Ziblatt2009). In this capacity, local elites control which issues are acceptable for discussion in public forums. This is also true in the case of collective decision-making in public meetings prone to elite capture (Baldwin and Holzinger Reference Baldwin and Holzinger2019; Hinnerich and Pettersson-Lidbom Reference Hinnerich and Pettersson-Lidbom2014). Consequently, local elites typically control the political debates in their regions.

Third, local elites can try to influence the thinking of the residents they aim to control (Gaventa Reference Gaventa1980; Ross Reference Ross2009). According to Lukes (Reference Lukes1974), this ideological power can make people want things that are opposed to their self-interest. Acharya, Blackwell, and Sen (Reference Acharya, Blackwell and Sen2018, 22) describe this process as a behavioural path dependence; ‘social and political behaviours are passed down from one generation to the next through parent-child socialization and socialization within communities, schools, churches, and other kinds of local networks’. In the context of this study, churches and schools are particularly important places of socialization, where social hierarchies and the underlying power dynamics are internalized – because things have always been this way. By influencing education systems and the role of religious authorities, local elites play a key role in the socialization of the local population.Footnote ²

We argue that such traditional patterns of social control explain why the masses adopt the local elites' political preferences and oppose suffrage extension. In the empirical section, we explore this mechanism by examining data on population density, the education system, and mutual aid societies at the local level. We expect that social control will prevent the creation of mutual aid societies and be easier to exercise in sparsely populated areas (Tilly Reference Tilly2004b). In addition, high levels of social control should be associated with low levels of public educational investment (Cinnirella and Hornung Reference Cinnirella and Hornung2016). We provide qualitative evidence based on historical research to substantiate the plausibility of this causal mechanism in the appendix. Finally, to further boost confidence in the social control mechanism, we show that alternative mechanisms linking landholding inequality to democratization, namely electoral fraud, pressure from below, and the economic concerns of residents (rather than elites), cannot account for this relationship.

Case Selection: Direct Democratic Votes, 1866–77

This analysis relies on four direct democratic votes in Switzerland to investigate the effect of landholding inequality on democratization. Since the founding of the Swiss Federal State in 1848, full male suffrage had been enshrined in the constitution. However, the decentralized nature of the Swiss polity granted the twenty-five cantons the right to exclude citizens from national, cantonal, municipality-level elections, and popular votes. In response, the cantons created a patchwork of regulations primarily targeting low-income households.

Most notably, the cantons developed various rules that excluded Swiss citizens from voting in national, cantonal, and municipal matters if they resided in a canton other than their canton of origin. These rules were typically tied to a minimum period of residence in a given canton, but the rules differed between cantons. As we demonstrate in the appendix, in this period, domestic migration was primarily motivated by economic reasons and concerned with relocating poor individuals and families from rural areas to centres of industrialization to pursue economic opportunities. However, domestic migrants often struggled to escape poverty in their new domiciles. Given Switzerland's small size and the large rural depopulation movement, approximately 20 per cent of all male citizens over the age of twenty (the minimum voting age) were not allowed to cast a vote (Aubert Reference Aubert1974, 96; Boppart et al. Reference Boppart2013, 246; Lutz Reference Lutz2000, 35–6). However, there are large variations between cantons (see Fig. A3 in the appendix).Footnote ³

Considering these cantonal rules to violate the constitution and in the absence of a constitutional court, the federal government tried to regulate suffrage rights by passing four federal acts that would have abolished these restrictions. These reform proposals, put to direct democratic votes in the period between 1866 to 1877, were part of a larger attempt to regulate and democratize the electoral process (Gruner Reference Gruner1978; Voutat Reference Voutat1996).Footnote ⁴ The differences between the four proposals were minimal (Linder, Bollinger, and Rielle Reference Linder, Bollinger and Rielle2010, 23–6, 38–40, 45–6). In all four cases, the government aimed to restrict the cantons' abilities to limit voting rights in national, cantonal, and municipal matters in cases of inter-cantonal mobility (that is, Swiss citizens residing in a canton different from their canton of origin). The two reform proposals in 1866 regulated voting rights at municipality level and cantonal/national level, respectively, in two separate acts, whereas the reform proposal in 1875 regulated voting rights at municipality, cantonal, and national level in one act. The 1877 reform proposal was identical to the 1875 proposal, except that it would have allowed cantons to exempt mobile citizens for up to a year as a concession to its opponents.

The federal parliament – politically dominated by liberals representing new economic elites – supported all four reforms. The precise voting results in parliament for the two 1866 acts were not recorded, but the 1875 proposal was supported with a sixty to nineteen majority in the lower chamber and the 1877 proposal with a fifty to nineteen majority. However, all four proposals were made subject to a popular vote.Footnote ⁵ In the political campaign before the popular vote, the conservative opponents, mostly representing landed elites, voiced concerns about cantonal autonomy and raised the possibility that the poor mobile population might outnumber and outvote established voters and thus decide on spending issues (Linder, Bollinger, and Rielle Reference Linder, Bollinger and Rielle2010, 23–6, 38–40, 45–6). Importantly, Switzerland featured an exceptionally high level of fiscal decentralization at the time. There were no direct taxes at the federal level, whereas most public spending, including spending on social matters, occurred at the subnational level (Emmenegger, Leemann, and Walter Reference Emmenegger, Leemann and Walter2021, Reference Emmenegger, Leemann and Walter2022).

Hence, suffrage extension might have shifted the median voter's preferences and increased the demand for redistribution. Equally problematic from the point of view of the conservative opponents, suffrage extension would have enfranchised citizens who, due to their inter-cantonal mobility, had not been subjected to the same extent to the inter-generational socialization and institutional reinforcement that characterized the long-term residents. For these reasons, the reform would have significantly increased the cost of exercising social control for local elites. Ultimately, in all four cases, the opponents' framing of the issue proved to be powerful. A narrow majority of voters followed the conservative opponents of suffrage extension. They rejected the four proposals with no-shares of 56.9 per cent and 51.9 per cent (1866); 50.6 per cent (1875); and 61.8 per cent (1877), respectively.

The Swiss case offers a unique setting to explore the relationship between landholding inequality and suffrage extension. We can examine four direct democratic votes on suffrage extension in a political context where approximately 80 per cent of all adult male citizens could cast a vote. This setting enables us to explore whether landholding inequality also affects the voting behaviour of the population that does not directly benefit from concentrated resource endowments. Like Montalbo (Reference Montalbo2023), we examine the effect of landholding inequality on mass political behaviour and thus explore whether landholding inequality can undermine democratization through social control. However, unlike Montalbo (Reference Montalbo2023), who explores electoral support for pro-democratic parties, we can directly examine mass political behaviour on the question of suffrage extension.

The voting population was asked a straight yes/no question in all four cases. There is thus no risk of complications due to package deals or interdependencies between different elements of the reform, which make the democratizing effect of the reform often difficult to appraise (Albertus Reference Albertus2017). Similarly, the four votes were not embedded in a long series of negotiations shaped by incumbency/opposition dynamics, as is often the case in roll-call votes on democratizing reforms.

This paper thus follows the recent literature that increasingly focuses on developments within single countries, which helps to reduce the number of unobservable and unquantifiable historical differences, because all units were similarly affected by broader trends (Pepinsky Reference Pepinsky2019). Hence, the focus on within-country variation increases confidence that the results are not driven by omitted variable bias or changing units of analysis. Despite providing a more homogeneous sample, we have a large number of observations, as we work with district-level data.

Research Design and Data

The empirical section investigates the relationship between landholding inequality and popular support for suffrage extension. Following the literature, we expect a linear relationship (Boix Reference Boix2003), but we also consider the possibility of a non-linear one (Acemoglu and Robinson Reference Acemoglu and Robinson2006). Our statistical analysis proceeds in two steps. First, we examine the link between geographical instruments, landholding inequality, and support for suffrage extension. Second, we examine how landholding inequality is linked to different mechanisms that might explain mass opposition to suffrage extension; for example, social control, electoral fraud, popular pressure from below, and economic concerns of the resident population. Tables A2 and A3 in the appendix provide the summary statistics.

Our dependent variable is the share of ‘yes votes’ in the four direct democratic ballots between 1866 and 1877 at the district level. We have data for all 183 Swiss districts for the votes in 1875 and 1877. For the two votes in 1866, the results could not be retrieved from the archives for the districts in the cantons of Ticino and Basel-Country, leaving us with data from 171 districts only.

The data for measuring landholding inequality at the district level comes from Switzerland's first agricultural farm census (Statistisches Bureau des eidgenössischen Departement des Innern, 1910). To capture landholding inequality, we calculated the Gini index using data on plot sizes.Footnote ⁶ Similar to other countries (Ziblatt Reference Ziblatt2008), the first census with detailed subnational data was conducted at the turn of the twentieth century – unfortunately, about thirty years after the popular votes on suffrage extension. However, this should not be a problem for our analysis. Research has shown that landholding inequality displays a high degree of autocorrelation over the nineteenth and early-twentieth centuries in all countries for which data is available (Mares and Queralt Reference Mares and Queralt2015; Vanhanen Reference Vanhanen2003). This is also the case for Switzerland, as Fig. A4 in the appendix shows (using the Vanhanen data). Unfortunately, the Vanhanen data are unavailable at the subnational level, but it is assumed that this high degree of autocorrelation also applies to the subnational level. We are thus confident that the time difference between the popular votes and the agricultural census does not pose a significant issue. Yet, even if our inequality measure is mismeasured, the OLS estimates are likely downward biased and our instruments correct for the measurement error (Hausman Reference Hausman2001).

A major concern for causal identification is that the relationship between landholding inequality and support for suffrage extension might be endogenous (Albertus Reference Albertus2015; Brooks and Kurtz Reference Brooks and Kurtz2016; Ramsay Reference Ramsay2011). To tackle this problem, we employ different instruments that encompass geographical features. These features pose natural limits to returns to scale and increase transportation costs, thereby acting as exogenous factors for land demand and optimal farm size. More precisely, our first instrument is climate suitability for wheat production. Climatic characteristics that are associated with commodities grown on family farms (for example, wheat) have promoted the growth of a sizeable middle class, which is associated with low levels of landholding inequality (Baten and Juif Reference Baten and Juif2014; Easterly Reference Easterly2007; Sokoloff and Engerman Reference Sokoloff and Engerman2000). Conversely, poorer climatic conditions are linked to large-scale cattle husbandry (Emmenegger, Leemann, and Walter Reference Emmenegger, Leemann and Walter2021). The second instrument is nutrient storage capacity as a measure of soil quality. In line with previous research (Cinnirella and Hornung Reference Cinnirella and Hornung2016; Easterly Reference Easterly2007; Frankema Reference Frankema2010), soil quality is another natural constraint that affects agricultural output independently of policies. The third instrument is terrain ruggedness. An extensive literature has linked ruggedness to large transportation and construction costs (Nunn and Puga Reference Nunn and Puga2012), making small-scale farming more costly (Baten and Hippe Reference Baten and Hippe2018). Fig. 1 shows that the relationship between these geographical characteristics and landholding inequality in Switzerland aligns with the existing literature.

Figure 1. Geography and landholding inequality in Switzerland.

Valid instruments have to satisfy the exclusion restriction. Put differently, instruments must affect the outcome via a specified endogenous variable and not via another confounder. While this condition can never be conclusively proven, we can test alternative channels that might link instruments and outcomes. In this case, the geographical instruments could affect support for suffrage extension via economic development instead of landholding inequality. For instance, soil quality, terrain ruggedness, and climate might limit agricultural output and urbanization or lead to higher levels of infant mortality and rural depopulation; therefore, we provide a correlation matrix in the appendix, which shows that our instruments are unrelated to urbanization, infant mortality, and migration patterns (see Fig. A5). Given that our instruments are unrelated to indicators normally used as proxies for economic development in historical research, we are confident that the exclusion restriction holds with regard to the theoretically most prominent channel between geography, modernization, and democratization.

Fig. 2 shows the geographical distribution of landholding inequality and support for suffrage extension for the votes in 1875 and 1877. The maps for both votes in 1866 are in the appendix (see Fig. A2). The upper left map shows that landholding inequality is less pronounced in a belt stretching from Northeast to Western Switzerland. By contrast, the Alpine regions in the South and Southeast and some parts of the Northwest display high levels of landholding inequality. Support for suffrage extension is more heterogeneously distributed across the country, as the two lower maps show. However, Fig. 2 clearly shows that the major opposition against suffrage extension in 1875 and 1877 came from the southern Alpine regions (see also Fig. A2 in the appendix), which runs counter to claims that the Alpine regions were the most equal areas and supportive of democratization (Boix Reference Boix2003, 112–8).

Figure 2. Landholding inequality and support for suffrage extension at the district level.

Fig. 2 suggests a positive correlation between landholding inequality and popular opposition to suffrage extension. Is social control the mechanism through which landholding inequality operates? Three sets of indicators are used to shed some light on this question. First, as outlined above, landholding inequality is expected to correlate with social control. To exercise social control, landholding elites need to monitor the behaviour of ordinary citizens, which is more difficult in densely populated regions (Beissinger Reference Beissinger2022; Caselli and Falco Reference Caselli and Falco2022; Gonzalez-Ocantos et al. Reference Gonzalez-Ocantos2020; Hirschi Reference Hirschi1969). At the same time, population density raises the potential for collective action and organized resistance to elites (Thomson Reference Thomson2018; Tilly Reference Tilly2004b). Therefore, we use the density of eligible voters per km² in 1875 to gauge whether landholding inequality is linked to sparsely inhabited regions.Footnote ⁷

Second, we investigate whether landholding inequality is associated with investments in education. The vast majority of the democratization literature links education to several factors that erode traditional patterns of social control, such as geographical and labour market mobility and pro-democratic attitudes (Inglehart and Welzel Reference Inglehart and Welzel2005; Lipset Reference Lipset1959). Thus, the landholding elites have a vested interest in keeping the local population's education level low (Baten and Hippe Reference Baten and Hippe2018; Baten and Juif Reference Baten and Juif2014; Cinnirella and Hornung Reference Cinnirella and Hornung2016; Goñi Reference Goñi2023). To capture the supply of schooling, we explore data from Kinkelin (Reference Kinkelin1875) at district level on total public school expenditure per pupil and the share of teachers with poor training. To capture educational outcomes, we use the results of the first pedagogical examination of Swiss military conscripts from 1875–9 (Statistisches Bureau des eidgenössischen Departement des Innern, 1876). These examinations, mandatory for all males in the relevant age cohort, are the first publicly available and standardized measure of educational attainment, which allows education levels across the country at the district level to be compared. More precisely, we compute the average grade of all conscripts in each district across all tested subjects: reading, writing, numeracy, and history (‘Vaterlandskunde’). The grades ranged from 1 (best) to 4 (worst).

Third, traditional patterns of social control should prevent the creation of civil society organizations. We account for such groups by including the existence of local sections of the Grütli association (‘Grütliverein’) and the location of their meetings. In this period, the Grütli association was Switzerland's largest and most important mutual aid society in the country (Gruner Reference Gruner1968). As an organization of non-elite actors, the Grütli association was a proponent of suffrage extension. Importantly, the Grütli association did not endorse the 1877 proposal because it considered the law not far-reaching enough (Linder, Bollinger, and Rielle Reference Linder, Bollinger and Rielle2010, 45–6). Although this was primarily a tactical move to increase pressure, we focus exclusively on sections and meetings before 1875.

We also examine multiple alternative mechanisms highlighted in the existing literature to increase our confidence in the social control mechanism.

First, local elites can engage in electoral fraud (Lehoucq and Molina Jiménez Reference Lehoucq and Molina Jiménez2002; Mares Reference Mares2015; Ziblatt Reference Ziblatt2009). Failing to align the local population's voting behaviour with their economic interests, local elites can try to manipulate the political process to achieve the desired outcome; for example, through their control over voting officials. As a result, ballot outcomes might not accurately indicate citizen preferences.Footnote ⁸

We employ two strategies to rule out electoral fraud as a mechanism at play. We examine systematic differences in turnout across space. Given that nationwide majorities win referendums, there is little incentive for opponents of suffrage extension to stop citizens from voting. Rather, to overcome majorities for suffrage extension in other areas of the country, opponents of suffrage extension might use electoral fraud to increase turnout. We should, therefore, observe turnout levels to increase with landholding inequality. In addition, we examine data on complaints about electoral fraud in the four popular votes. Such complaints had to be directed at the federal government, which provides reports on all ballots. As a supporter of suffrage extension, the government had no incentive to withhold complaints that would have undermined the political legitimacy of referendums against the government's proposals.

Second, democratization might be the result of pressure from below. Several contributions emphasize how mass support for reform and the threat of social unrest can lead to democracy (Aidt and Franck Reference Aidt and Franck2019; Collier Reference Collier1999; Tilly Reference Tilly2004a); for instance, by increasing the cost of repression. Importantly, citizens' demands for democratization vary with the level of inequality. When there is greater inequality, citizens have more to gain from regime change, making a push for democratization more likely. Combining citizens' demands for democracy and the elites' aversion to it (both increasing with the level of inequality), Acemoglu and Robinson (Reference Acemoglu and Robinson2006) argue that there is an inverted U-shaped relationship between inequality and democratization, with democratization being most likely at medium levels of inequality.

We employ two strategies to rule out that pressure from below is the mechanism at play. Following Acemoglu and Robinson (Reference Acemoglu and Robinson2006), we examine whether support for suffrage extension is highest at medium levels of landholding inequality. In addition, we consider the possibility that the share of citizens that would be enfranchised in case of an approval of the vote was the main cause of opposition among the elites. Here, the assumption is that pressure from below increases with the share of excluded citizens. All four proposals were primarily concerned with adult Swiss males who held citizenship from another canton, who were excluded from suffrage due to inter-cantonal mobility. We, therefore, explore the relationship between landholding inequality and the share of disenfranchised adult Swiss males.

Third, the landed masses might oppose suffrage extension, independent of social control exercised by local elites, because they do not want people who have been living a short amount of time in the region to vote on municipal, cantonal, or national matters and influence the lives of long-term residents, especially in matters of taxation and spending. Put differently, the already enfranchised voters might have had their own reasons to oppose franchise extension. We would expect such opposition to increase with the share of newcomers and levels of poverty because of growing concerns about the division of scarce resources among newcomers and long-term residents (Ardanaz and Mares Reference Ardanaz and Mares2014; Stasavage Reference Stasavage2022).Footnote ⁹ In the empirical analysis, we consider this possibility in two ways. Next to the share of Swiss males excluded from suffrage due to inter-cantonal mobility (see above), we also look at the share of non-cantonal citizens and data on infant mortality (as an indicator of poverty).

It could also be argued that citizens with more authoritarian attitudes move into areas with high levels of landholding inequality. However, as shown in the appendix, domestic migration primarily concerned the movement of people from rural areas to centres of industrialization in this period. There is no evidence of migratory movements in the other direction for political reasons. More plausibly, it could be argued that citizens with more democratic attitudes leave rural areas, especially those with high levels of landholding inequality. However, such migratory movements perfectly align with the argument about social control. Resisting social control and supporting democracy, such individuals are expected to incur economic and social sanctions, which might ultimately force them to leave their home areas. Finally, it could be argued that citizens in small remote villages in mountainous areas might generally be more hostile to aliens from other cantons, although the chance of being outvoted by newcomers is comparatively small. While this possibility cannot be completely ruled out, it remains unclear why such conservative attitudes should co-vary with landholding inequality. Moreover, the existing literature on mountainous areas and democratization has typically argued the opposite, emphasizing how mountainous areas are conducive to democracy, although there is no comprehensive evidence in favour of this argument (see Gerring et al. Reference Gerring2022, 313–319).Footnote ¹⁰

OLS and instrumental variable estimation are relied on to estimate the effect of landholding inequality on support for suffrage extension. Given that idiosyncratic factors might drive voting behaviour in individual votes, we pooled all four votes and added vote fixed effects to all of our models. We used cluster-robust standard errors to account for the non-independence of our observations within districts. We expect our treatment and outcome to be spatially correlated, as Fig. 2 suggests. To avoid bias induced by spatial correlation, we employ a spatial two-stage least square model (S-2SLS) that provides unbiased estimates under spatial interdependencies within instruments and outcomes (Betz, Cook, and Hollenbach Reference Betz, Cook and Hollenbach2020). We specify the neighbouring structure by using the inverse distance between districts and then row-standardized the connectivity matrix. This gives us the weighted average for yes shares for all districts with higher weights for districts that are closer.

Moreover, we include several control variables that might confound the relationship between landholding inequality and support for suffrage extension. Most notably, we control for urbanization, a widely used indicator to capture differences in economic development (Stasavage Reference Stasavage2022, 182). This is particularly important because landholding inequality might be associated with higher agricultural productivity and, thus, economic development (Montalbo Reference Montalbo2023). By controlling for urbanization, we ensure that our indicators of landholding inequality do not simply capture differences in economic development.Footnote ¹¹ Additional controls include population size and whether a district is predominantly Catholic and/or German-speaking. Table A1 in the appendix shows an overview of the operationalization. Finally, in the appendix, we provide qualitative evidence based on historical research to substantiate the plausibility of the social control mechanism. The historical literature on the social and political conditions in regions characterized by high levels of landholding inequality offers rich depictions of social hierarchies, which could govern the behaviour of ordinary citizens.

Quantitative Evidence

Table 1 displays the OLS and instrumental variable (IV) estimations for the effect of landholding inequality on support for suffrage extension. The first two models provide OLS estimates for the relationship between landholding inequality and support for suffrage extension. As the second model shows, a 0.1 increase in landholding inequality is associated with a decrease of 3.6 percentage points in support for suffrage extension. In the remaining models, we forward IV estimates using soil quality, climate suitability for wheat, and ruggedness. In line with Fig. 1, the first-stage results demonstrate a strong relationship between the instruments and landholding inequality, even if we account for different confounders and spatial dependence. The second stage results show that the IV estimates are significant with the expected sign. More precisely, the results suggest that a 0.1 increase in landholding inequality leads to a decrease of 6.2 or 8.4 percentage points in support of suffrage extension.

Table 1. Landholding inequality and support for suffrage extension

***p < 0.001; **p < 0.01; *p < 0.05; ^†p < 0.1.

Having provided evidence on the link between landholding inequality and support for suffrage extension, Tables 2 and 3 provide OLS estimates to shed light on the causal mechanism linking landholding inequality and opposition to democratization.Footnote ¹² We start by focusing on indirect measures of social control. Landholding inequality is negatively and significantly linked to education supply and outcomes. Put differently, landholding inequality is associated with less education spending per pupil (‘Education Spending’), a higher share of poorly trained teachers (‘Education Teacher’), and a weaker educational performance (‘Education Performance’). In addition, the relationship between landholding inequality and voter density is also negative (‘Voter Density’). Moreover, Table 2 shows that civil society organizations (‘Associations’) were less likely to exist in regions with high levels of landholding inequality, which provides further evidence of social control. Overall, the results demonstrate that voters residing in districts with higher landholding inequality could be more easily monitored by traditional elites, were less educated, and displayed lower levels of civil society organization. These findings suggest that social control can explain mass political opposition to suffrage extension in regions characterized by high levels of landholding inequality.

Table 2. Landholding inequality and support for suffrage extension: social control mechanisms

Heteroscedasticity-robust standard errors in parenthesis. ***p < 0.001; **p < 0.01; *p < 0.05; ^†p < 0.1.

Table 3. Landholding inequality and support for suffrage extension: alternative mechanisms

Heteroscedasticity-robust standard errors in parenthesis. ***p < 0.001; **p < 0.01; *p < 0.05; ^†p < 0.1.

What about the alternative explanations for the relationship between landholding inequality and mass opposition to suffrage extension? As mentioned above, local elites might engage in electoral fraud if they cannot exercise sufficient social control. Table 3 shows the relationship between landholding inequality and turnout in the votes in 1875 and 1877 (‘Turnout 1875’ and ‘Turnout 1877’). Turnout data for the two votes in 1866 is, unfortunately, not available. As discussed above, given the logic of referendums (nationwide majorities), electoral fraud by opponents of suffrage extension should lead to higher turnout (to compensate for supporting majorities in other areas). However, the results provide no evidence that turnout was systematically associated with landholding inequality.

Qualitative evidence provides no support for the electoral fraud mechanism either. The federal government received no complaints about electoral fraud regarding the two popular votes in 1866 and 1875 and only one complaint in the case of the 1877 vote. For the sake of comparison, there were eighteen complaints in 1875, seven in 1876, and eight in 1877 about the violation of electoral laws in elections (Bundesrat 1866, 67–120; 1875, 151–67; 1876, 456–68; 1877, 428–46). Given that the federal government had no incentive to withhold information about complaints, these results suggest that landholding inequality did not operate via electoral fraud.

What about pressure from below? Table A4 in the appendix shows that there is, in contrast to Acemoglu and Robinson (Reference Acemoglu and Robinson2006), little evidence for an inverted U-shaped relationship between landholding inequality and opposition to suffrage extension. In addition, Table 3 shows that landholding inequality is not systematically associated with the share of excluded citizens (‘Excluded’). Similarly, Table 3 offers no evidence that the landed masses opposed suffrage extension (independent of social control exercised by local elites) because they did not want people from outside to influence local political matters. The share of non-cantonal citizens (‘Non-Cantonal’) and the poverty level (‘Poverty’) in a given district are not systematically associated with landholding inequality.

In sum, the evidence presented in Tables 2 and 3 points to social control as the likely mechanism linking landholding inequality to mass opposition to suffrage extension. In the appendix, we provide additional qualitative evidence based on historical research to substantiate the plausibility of this causal mechanism. More specifically, we show that in areas with high levels of landholding inequality, local elites had the means to exercise social control by keeping investments in education low, suppressing opposition groups, and exercising control over institutionalized forms of political participation.