• 1 Introduction

  Since the mid-2000s, there has been a growing volume of scholarly contributions seeking to investigate non-resident citizens’ voter turnout either from a comparative perspective or using influential case studies. This burgeoning literature has frequently drawn on seminal voter turnout research and has examined the electoral participation of emigrants and their descendants at the aggregate (e.g. Burgess & Tyburski, 2020; Ciornei & Østergaard-Nielsen, 2020) or individual level (e.g. Chaudhary, 2018; Himmelroos & Vento, 2022; Lafleur & Sánchez-Domínguez, 2015; McCann et al., 2019; Mügge et al., 2021; Peltoniemi, 2018). Earlier research has connected non-resident citizens’ voter turnout to (in)formal channels of sociopolitical activism across borders, the activities of political parties abroad, the role of migrant civic associations, the voting methods and the characteristics of electoral districts abroad (e.g. Kernalegenn & van Haute, 2020; Lafleur, 2013; Nemčok & Peltoniemi, 2021; Paarlberg, 2017; Rashkova, 2020; Umpierrez de Reguero & Dandoy, 2021). Yet the influence of compulsory voting – a classic determinant of voter turnout in the domestic arena according to the institutional approach (e.g. Blais & Carty, 1990; Jackman & Miller, 1995; Powell, 1986) – on non-resident citizens’ voter turnout is largely overlooked by the existing literature. There are few exceptions, however, that employ compulsory voting as an independent variable for iterative models of non-resident citizens’ voter turnout in Latin American and Southern European countries (e.g. Umpierrez de Reguero, 2022) or just to mention its importance for tentatively explaining why some cases display higher rates of electoral participation (e.g., Lafleur, 2013).
  To fill this gap, this article exploits a design with aggregate-level observations of migrants from ten Latin American countries in three European countries (Belgium, Luxemburg and the Netherlands). Latin America and the Low Countries represent an important share of the countries implementing compulsory voting worldwide as no less than six of them impose a voting obligation on a majority of their citizens (Belgium, Bolivia, Brazil, Ecuador, Luxemburg and Peru). In addition, these countries are unique regarding the voting obligation across national borders (Umpierrez de Reguero et al., forthcoming) as they provide the only five examples of compulsory voting for external voting rights after voluntary registration at the national level: Belgium, Brazil, Luxemburg, Paraguay and Peru.¹x There are a few other exceptions when it comes exclusively to constitutional referenda such as Equatorial Guinea (see El País, 1982). These facts allow us to configure different treatment-control scenarios to assess the potential influence of the voting obligation on non-resident citizens’ turnout, especially considering other countries in our sample that differentiate among citizens to enforce the vote (e.g., Bolivia and Ecuador) or just set rules for optional voting for all their nationals (e.g. Chile and the Netherlands).
  In what follows, we briefly review the nexus between compulsory voting and non-resident citizens’ voter turnout. Prior to presenting and discussing the results regarding the impact of compulsory voting on non-resident citizens’ voter turnout, we outline the methodological contours of this article considering both origin and residence country contexts. We report a strong connection between the degree of voting obligation and non-resident citizens’ voter turnout.

• 2 Conceptualizing and Measuring Non-resident Citizens’ Voter Turnout

  Non-resident citizens’ voter turnout (also denominated as extraterritorial, transnational or emigrant turnout [Burgess & Tyburski, 2020; Ciornei & Østergaard-Nielsen, 2020; Lafleur, 2013]) comprises the number of votes cast in elections abroad. For first-generation migrants, elections occur in their countries of origin. In the case of second- and third-generation international migrants, these elections stem from the countries of origin of their parents and grandparents. That is why we do not label this type of voter turnout as ‘emigrant voter turnout’, as the category of non-resident citizens is composed of individuals who possess a legal status to participate in homeland elections, obtained by Ius Solis or Ius Sanguinis (Bauböck, 2015; Erdal, 2016).
  Previous research not only makes further efforts to gather and analyse observational turnout data in both countries of origin and of residence but also requires basing its estimations on multiple sources (e.g. private information from diplomatic offices, electoral register of the ministry of interior or electoral commissions, UN migrant stock), owing to the nature of migration (Ciornei & Østergaard-Nielsen, 2020; Umpierrez de Reguero, 2022). On the one hand, scholars can aggregate the number of votes cast in an election. In this scenario, turnout is a figure that summarizes all the votes cast abroad, whether per estimated migrant population (stock) or in different polling stations/countries of residence. By unpacking non-resident citizens’ voter turnout by countries of residence, existing accounts may pose a pathway to embrace political transnationalism and verify whether migrants’ political re-socialization and experiences in diverse countries of residence prompt variation in electoral participation (e.g., Belchior et al., 2018; Ciornei & Østergaard-Nielsen, 2020).
  On the other hand, scholars face various challenges when they have separated the number of votes cast by the different formulae noted previously, considering either one observation per election or several observations because of multiple countries of residence per election. The estimated number of migrants from a country of residence can produce one outcome using the United Nations (UN) migrant stock censuses²x Sources such as the UN migrant stock may lead to over- or underestimating non-resident citizens’ turnout, since its data comprises population under 18 years old (Ciornei & Østergaard-Nielsen, 2020). Also, it does not consider migrants’ descendants who possess the right to participate in mostly all Latin American contexts. and another employing data from individual ministries of interior or electoral commissions (Collyer, 2014b; Vintila et al., 2022). Adapting this dependent variable to the analysis of Global South migration, not every migrant has obtained a regular legal status by the beginning of the year or when scholars and policymakers conduct fieldwork (Acosta, 2018). In other words, there are hundreds of thousands of individuals who emigrate to another country yet are not legally registered in the new jurisdiction, causing missing data for both the countries of origin and residence, including an impact on the estimation of non-resident citizens’ voter turnout.
  Official electoral information provided by diplomatic offices, ministries of interior and/or electoral commissions may also result in methodological limitations, as the processes of registration and deregistration are, in most countries, automatic or require a one-off enrolment (Hutcheson & Arrighi, 2015; Lafleur, 2013). For instance, the number of individuals registered to vote can be anaemic owing to non-resident citizens being unwilling, even afraid, to enrol with the authorities abroad of the country of origin for several motivations related to the migrants’ legal status, authoritarian regimes and a particular interest in residence country’s policies or a combination of them (Bermúdez et al., 2017; Boccagni & Ramírez, 2013; Lafleur & Sánchez-Domínguez, 2015; Mügge et al., 2021; Szulecki et al., 2021). Hence, the use of official data may lead to an over- or underestimation of the population living abroad, contingent on the emigration features.
  Furthermore, state authorities of origin countries have different ways of counting overseas votes (Collyer, 2014a; Umpierrez de Reguero et al., forthcoming). In sum, they can opt to register a single number of overseas votes, disaggregated by the countries of residence and/or polling stations (either in electoral districts abroad or not) or aggregated in a geographical district in their country of origin. In this article, we will only employ non-resident citizens’ voter turnout as the total number of votes divided by the total number of registered voters despite all these methodological limitations since we find this operationalization the best possible means of associating with compulsory voting in a comparative framework.

• 3 Compulsory Voting and Non-resident Citizens’ Voter Turnout

  By compulsory voting, we refer to the obligation to vote that citizens within a given polity must comply with. The consequences of non-participation in elections range from an economic sanction often in the form of a fine (e.g. in Ecuador), non-access to some privileges or administrative acts (e.g. in Brazil) or to the absence of sanctions (e.g., in Belgium). Despite this legal obligation and possible sanctions, compulsory voting is often accompanied by a low turnout among non-resident citizens and by a high dissatisfaction with being obliged to vote (see e.g., Frizzo & Mascitelli, 2017 in the Brazilian case).
  By adapting this legal-based conceptualization to migrants’ political participation, we consider two elements for the study of compulsory voting across national borders. First, the degree of compulsory voting and, second, the dual context where this variable can be altered by the influence of migrants’ political (re-)socialization.
  The scholarly literature that refers to a ‘cultural’ explanation stands out, characterized by the belief that electoral attitudes are the product of socialization. Political socialization frames how individuals engage in political development and learning and how they construct relationships with the political contexts in which they live (Sapiro, 2004; Superti & Gidron, 2021). According to this perspective, compulsory voting creates a civic duty to vote that develops as a consequence of early socialization processes. The voting obligation is internalized at some point in the early stages of life and translates into predispositions for or against voting in elections (Galais & Blais, 2016).
  Similarly, several studies have suggested that non-resident citizens’ voter turnout is partly influenced by prior political socialization (pre-emigration) (e.g., Boccagni, 2011; Lafleur & Sánchez-Domínguez, 2015; McCann et al., 2019). Given that civic duty to vote precedes adulthood and might be fostered by political socialization in the family and at school (Blais & Galais, 2016; Feitosa et al., 2022; Galais, 2018), pre-emigration political socialization may explain why some countries exhibit relatively high rates of non-resident citizens’ voter turnout even if only resident citizens are obliged to vote.
  Another perspective posits that some citizens experience external social pressure to conform to the norm of voting, and that norm may be internalized through the process of political socialization (Doherty et al., 2019). It is these external social influences that affect individuals’ judgement of their own behaviour, in particular regarding the civic duty to vote. Applied to non-resident citizens, this socialization becomes a matter of social capital, as these residents are supposed to be less integrated into public life (Lineira & Vallès, 2014). While the first perspective focuses on family and school (i.e. socialization in the country of origin), this perspective focuses on the importance of peer groups, migrant organizations and community (i.e. re-socialization in the country of residence). Hence, the theoretical importance of studying the impact of compulsory voting settings in both countries of origin and residence.
  In our sample, four countries worldwide oblige non-resident citizens to take part in the elections in the country of origin after voluntary registration (i.e. Belgium, Brazil, Luxemburg and Peru).³x Paraguay also belongs to that group of countries but is not included in our analyses as it did not establish polling stations either in Belgium or in Luxemburg or the Netherlands. In Bolivia and Ecuador, voting within the country is compulsory, while voting is voluntary for national citizens residing abroad. The remaining set of countries sampled (i.e. Chile, Colombia, Costa Rica, Dominican Republic, the Netherlands, Mexico and Venezuela) invite their non-resident and resident citizens to vote without any type of enforcement or sanction.
  As this Special Issue deals with the political integration of migrants in the Low Countries (Camatarri & Baudewyns, 2022) and the literature has proven that voting in the country of origin may affect the electoral behaviour of the migrants towards the country of residence and vice versa (see Ahmadov & Sasse, 2016; Chaudhary, 2018; Goerres et al., 2021; Paul, 2013; Szulecki et al., 2021; Tsuda, 2012; White et al., 2008), we also consider the dual context in which compulsory voting can be implemented. As a result, we create a 2 × 2 typology combining voting obligation with international migration loci to design four mutually exclusive quadrants. As international migrants seek to balance their political engagement between two autonomous territories identifying in which of the two, in both or neither they are able to and want to electorally participate (Erdal & Oeppen, 2013; Finn, 2020), the first quadrant of our typology comprises a scenario where both countries – of origin and residence – apply voluntary voting for (non-)resident citizens. In turn, the second quadrant implies a setting in which the countries of residence enforce the vote for their (non-)citizen residents, while the countries of origin (of the non-resident citizens) do not. Conversely, the third scenario supposes a context where the legal framework of the countries of origin induces compulsory voting, while the electoral rules of the countries of residence do not. Lastly, the fourth quadrant encompasses a scenario in which both countries – of origin and residence – set compulsory voting for their elections.

  Table 1 The Dual Context of Compulsory Voting for Non-resident Citizens

  		Country of Residence
  		Voluntary voting	Compulsory voting
  Country of Origin	Voluntary voting	(1)	(2)
  	Compulsory voting	(3)	(4)

  Considering the degree and the dual context of compulsory voting from a transnational perspective, we expect to observe that (H1) non-resident citizens’ turnout is higher in countries where voting is compulsory for (non-)residents voters (as compared with countries with voluntary voting for (non-)residents voters). This hypothesis can be subdivided when one distinguishes the countries of origin and residence: non-resident citizens’ turnout is higher in countries of origin where voting is compulsory (H1a), and non-resident citizens’ turnout is higher in countries of residence where voting is compulsory (H1b). Likewise, we hypothesize that (H2) non-resident citizens’ turnout is higher when both countries apply compulsory voting for (non-)resident voters (as compared with when voting is compulsory in one out of two countries or when voting is voluntary in both).

• 4 Data and Method

  To test our hypotheses, we construct a cross-sectional time-series database with electoral results of Latin American migrants voting in polling stations located in Belgium, Luxemburg and the Netherlands. It comprises different types of elections at the national and supranational levels (i.e. presidential, legislative, Andean⁴x We only compiled information of Ecuador and Peru, since Paraguay has no polling stations in Belgium, Luxemburg or the Netherlands. Similarly, Venezuela does not disclose non-resident citizens’ voter turnout for the MERCOSUR Parliament. In the rest of the Latin American cases with active external voting rights, the existing provisions enable the electoral participation of their non-resident citizens in supranational elections. and national referendum), from 2005 to 2022 (N = 213). Rules concerning compulsory voting in the analysed countries do not vary by election type. Our unit of analysis has been structured from the combination of institutional contexts in the countries of origin and residence. Following Ciornei and Østergaard-Nielsen (2020), we multiplied the number of elections in ten countries of origin from Latin America⁵x Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, Mexico, Peru and Venezuela. As of the beginning of 2022, Cuba, Nicaragua and Uruguay do not apply a provision of external voting rights. Non-resident citizens’ voter turnout data from Argentina, Guatemala, Honduras, Panama, Paraguay and El Salvador in Belgium, Luxemburg and the Netherlands is inaccessible or non-existent. Owing to data availability constraints, several single election data at the polling station level are similarly missing, as, for instance, the 2014 presidential elections in Brazil or the 2016 legislative elections in the Dominican Republic. Even if the 2019 Bolivian elections were annulled, we kept the record of its turnout figures in our database. per three countries of residence (Belgium, Luxemburg and the Netherlands).⁶x In many cases, the electoral management body does not open a polling station for its national residents in Luxemburg. For instance, there are no records of separate polling stations in Luxemburg for the Brazilian elections. In doing so, we built a data set suitable to unwrap to what extent non-resident voters from Latin America in the Low Countries behave similarly or differently, as the former set of countries provides different voting obligation. Voting is compulsory for at least a majority of the voters in four countries of origin (Bolivia, Brazil, Ecuador and Peru) and in two countries of residence (Belgium and Luxemburg), while voting is voluntary in six countries of origin (Chile, Colombia, Costa Rica, Dominican Republic, Mexico and Venezuela) and in one country of residence (Netherlands).
  Our datasets rely on primary and secondary sources. We gathered information provided by online archives and existing official documents from the electoral management bodies of each Latin American country in our sample. Once we collected the information on non-resident citizens’ voter turnout from Latin America in Belgium, Luxemburg and the Netherlands that was available, we consulted the Electoral Rights of Migrants Dataset (Umpierrez de Reguero et al., forthcoming) from the Global Citizenship Observatory (GLOBALCIT). This enabled us to operationalize our explanatory and control variables based on the eligibility and access conditions of migrants’ electoral rights. In addition, we coded other controls that are a by-product of contextual factors and relied on the UN census of international migrant stock (UNDESA, 2020) to run robustness checks.

  4.1 Measures

  Our dependent variable is non-resident citizens’ voter turnout. As stated in the previous section, we employ non-resident citizens’ voter turnout as the total number of votes divided by the total number of registered voters. As many electoral systems with compulsory voting have a considerable share of spoiled votes (>10%), we include a twofold operationalization to disaggregate emitted from valid votes (following Dandoy & Kernalegenn, 2021) to examine the impact of compulsory voting on non-resident citizens’ voter turnout (see Models 1-4). Correspondingly, we create two different, albeit related, measurements: the first comprises the total number of valid votes as the numerator of non-resident citizens’ voter turnout, while the second considers the total number of emitted votes as the numerator to estimate how many registered voters eventually voted.
  A consequence of this research strategy is that there is a varying number of missing data, depending on the type of the dependent variable used in the models. Some election management bodies publish the overall number of emitted votes, while others only make the total number of valid votes public. While our database contains information about 213 Latin American elections in any of the three Low Countries, it varies from 194 observations in models using turnout calculated using the total number of valid votes (Models 1 and 2 in Table 3) to 147 observations in models using turnout calculated using the total number of emitted votes (Models 3 and 4).⁷x See Table A1 in the appendix for the number of observations and descriptive statistics of the different operationalisations of the dependent variable, including the ones used in the robustness checks.
  By operationalizing the dependent variable in two ways, we can differentiate the 28.8% of the Peruvians residing in Belgium that effectively chose a political party in the supranational election of 2021 of their country of origin from the 46.4% that cast the ballot in the same electoral contest (either by electing a party or by spoiling their vote). Despite examples such as the 2021 Peruvian Andean election, the means (as calculated by different numbers of data set observations) seem to be quite standardized at least between the baseline and the alternative formulae. On average, the equation estimated over the registered voters duplicates, even in some cases such as Bolivia, triplicates the percentage of the equation based on migrant stock. This is unsurprising since each of the Latin American countries in the sample requires its non-resident citizens to register before voting (Umpierrez de Reguero et al., forthcoming; Wellman et al., 2022).

  Average of non-resident citizens’ voter turnout (NRVT) per country of origin
  Notes: N = 194 for NRVT (Valid Votes/Registered); N = 147 for NRVT (Emitted Votes/Registered); N = 194 for NRVT (Valid Votes/Migrant Stock); N = 146 for NRVT (Emitted Votes/Migrant Stock). See Table A1 in the Appendix.

  Our main explanatory factor is the compulsory voting settings in a particular country. First, we code ‘0’ if neither the country of origin nor the country of residence has compulsory voting; ‘1’ if only resident citizens are obliged to vote but not citizens residing abroad; and ‘2’ if both resident citizens and citizens residing abroad are obliged to vote. The various Latin American countries in our dataset are distributed as follows: voluntary voting in Chile, Colombia, Costa Rica, Dominican Republic, Mexico and Venezuela; partially compulsory voting in Bolivia and Ecuador; and fully compulsory voting in Brazil and Peru. Overall, the total number of elections is relatively balanced among the three categories of response.
  In addition, we use an alternative operationalization of our main independent variable by constructing a 2 × 2 typology (see Table 1) where compulsory voting is implemented or not in the country of origin and/or in the country of residence. Ergo, we configure different combinations based on the four quadrants of voting obligation (i.e. voluntary versus compulsory voting in the countries of origin and residence). Table 2 shows 6 combinations where the vote is not enforced in both countries as ‘0’; then, the 12 combinations in which voting is voluntary in the country of origin but compulsory in the country of residence as ‘1’; thereafter, the 4 combinations where voting is compulsory in the country of origin but otherwise in the country of residence as ‘2’; and, finally, the eight remaining combinations in which voting is enforced in both countries as ‘3’.

  Table 2 The Dual Context of Compulsory Voting for Non-resident Citizens

  		Country of Residence
  		Voluntary voting	Compulsory voting
  Country of Origin	Voluntary voting	CL-NL, CO-NL, CR-NL, DO-NL, MX-NL, VE-NL	CL-BE, CL-LX, CO-BE, CO-LX, CR-BE, CR-LX, DO-BE, DO-LX, MX-BE, MX-LX, VE-BE, VE-LX
  	Compulsory voting	BO-NL, BR-NL, EC-NL, PE-NL	BO-BE, BO-LX, BR-BE, BR-LX, EC-BE, EC-LX, PE-BE, PE-LX

  Notes: Belgium (BE), Bolivia (BO), Brazil (BR), Chile (CL), Colombia (CO), Costa Rica (CR), Dominican Republic (DO), Ecuador (EC), Luxemburg (LX), the Netherlands (NL), Mexico (MX), Peru (PE), Venezuela (VE). The fact that in Peru the electoral management body passed a resolution to exonerate non-resident citizens for the economic sanction derived from the compulsory voting in 2006 but their vote legally remain enforced does not constitute a reason for a different code after and before 2006-2007.⁸x See Constitution (1993, Art. 31) and Organic Law of Elections (1997, Arts. 9 and 240). Law 27369 (2000, Art. 25). Since 2006, the authorities of the Peruvian electoral management body decided that non-resident citizens are exonerated from paying a fine if they are registered and do not vote (see Law 28859, August 3, 2006; Art. 4).

  We also include different control variables in the statistical models and robustness checks further on. We considered access-based conditions to external voting rights. On one side, we incorporated a binary coding of electoral registration and, on the other, the voting method. Since our sample varies from one-off active registration to non-permanent registration, we code the former as ‘0’ and the latter as ‘1’. Voting registration relates to the issue of self-selection, especially in the countries where that transnational voting becomes compulsory after voluntary registration. In particular, it addresses the question of why a citizen would register in the first place when he/she can be subject to sanctions thereafter. Most-restrictive registration is used in only two countries in our sample (Chile⁹x Chileans residing abroad shall change their domicile if they want to vote from their countries of residence “one hundred and forty days prior to each election or on the date of publication of the decree calling for a plebiscite, resuming from the first day of month following the election or plebiscite” (Law 20,960, 2016, Art. 28). Additionally, there is a residence-based eligibility for foreign-born Chileans; they need to have a certificate of past residence (certificado de avecinamiento) (see Umpierrez de Reguero et al., 2020). and Venezuela¹⁰x In 2004, Venezuelan Electoral Management Body enacted a provisional resolution to request non-resident citizens a series of documents to enrol and thereafter to vote, including a residence permit of visa. Although the type of registration is similar to that of its neighbours (i.e. one-off registration), Venezuelan authorities made de facto more difficult the registration procedure since 2004, particularly to dual citizens, to control external voting rights (Umpierrez de Reguero et al., 2021).), both countries where voting is voluntary. Similarly, we control for voting modalities, and we code in-person voting at diplomatic offices as ‘0’ and postal voting as ‘1’. While almost all the countries in our dataset use in-person voting for their non-resident citizens, postal voting is used in Mexico (where voting is voluntary).¹¹x Internet and postal voting were used in some polling stations in the 2021 elections in Ecuador but not in polling stations located in the Low Countries.
  We also computed a dummy variable to mirror the characteristics of overseas electoral districts. Some states reserve seats in their national or local legislatures for non-resident citizens’ direct representation. The number of seats can be ruled by the same electoral law or rely on demographic features of overseas districts as well as electoral outcomes in proportional representation electoral systems. In that respect, most countries with the special representation of non-resident voters establish differentiated overseas districts by geographical criterion (e.g. Dominican Republic and Ecuador), but there are also examples such as Colombia and Peru¹²x This is applicable only in the 2021 legislative election at the national level. that put together all overseas votes in a stand-alone electoral district (e.g., Escobar, 2007; Fliess, 2021; Palop-García, 2018). In turn, the vast majority of countries around the globe that extend de facto emigrant enfranchisement merge non-resident citizens’ voter turnout with domestic votes. Accordingly, we code ‘1’ when a country establishes an overseas district(s) to count the overseas votes, otherwise ‘0’.
  Similarly, we code several control variables associated with the types of election as well as electoral contexts. First, we included a categorical variable to reflect on the type of election: presidential, legislative, supranational or national referenda. Second, we computed a dummy variable to control non-resident citizens’ voter turnout in the middle of the Covid-19 pandemic. Since many of our data set observations depend on the way Belgium, Luxemburg and the Netherlands have faced the pandemic, we coded all the elections carried out from 2020 to 2022 as ‘1’ and ‘0’ otherwise.

  4.2 Estimation Strategy

  To answer whether the degree and the dual context of compulsory voting impact significantly on non-resident citizens’ voter turnout, we employ a set of linear panel data models. Panel data regressions are broadly used in social science to examine two-dimensional or cross-section time-series data sets such as the one we analyse in this article. The first dimension that crosses our data set is the country of origin, and the other is the country of residence (i.e. ten countries of origin by three countries of residence). Importantly, the dimension that considers the country of origin is time-sensitive, making unique configurations per data set observation.

  A typical panel data regression in comparative politics looks like the following:

  Y_jt = α + β x_oc … β n_oc + e_oc,

  where Y is the dependent variable (in this case, non-resident citizens’ voter turnout), x the explanatory variable(s) (i.e. compulsory voting), α and ß the coefficients, as well as o and c, that represent the country of origin and the country of residence, respectively. The e, which depicts the statistical error of the model and is also calibrated by o and r, is relevant in panel data models since it tallies the regression towards pooling, fixed or random effects – the three most used approaches in panel data analysis. Panel data model assumptions vary according to the error term, and, thereby, the researcher can decide between fixed and random effects or, contrarily, a pooling method. In a fixed effects model, e_oc should vary in a non-stochastic way over o and r. Conversely, in a random effects model, e_oc is assumed to stochastically change over o and r. If it is impossible to follow these two assumptions associated with the error term and there are no unique attributes of o within the assessment set, then the effects are not universal across r, and pooling regression should be used.

• 5 Results

  Our four baseline models are highly significant (p < 0.001) and display relatively high explanatory power. As indicated, all models are panel data regressions. Yet Models 1 and 3 use random effects, whereas Models 2 and 4 are pooled regressions. This decision was based exclusively on the coefficients of the Lagrange multiplier (with time as an effect) and Hausman tests.¹³x Both tests are commonly used to determine the nature of the regression analysis with panel data. When selecting random effects, the models were statistically significant and insignificant, respectively. By contrast, we chose pooled models when the Lagrange multiplier test was insignificant. The same methodological approach was adopted for the robustness checks when selecting what type of panel data regression to run (Models 5-8).
  Models in Table 3 rely on two alternative operationalizations of the dependent variable: turnout as the share of valid votes from the registered voters in Models 1 and 2 (N = 194) and turnout as the share of emitted votes from the registered voters in Models 3 and 4 (N = 147). While we examine the degree of compulsory voting along with several control variables in Models 1 and 3, we run other models considering the migrants’ dual context of political (re-)socialization, namely the countries of origin and residence in Models 2 and 4.
  Both alternatives of the explanatory variable are highly significant, meaning a strong positive influence of compulsory voting settings on non-resident citizens’ voter turnout. Remarkably, our results indicate a meaningful statistical relation, regardless of the operationalization of the dependent variable as well as the variation in the number of data set observations in the different models. Our main findings suggest that the higher the degree of voting enforcement, the higher the probabilities for non-resident Latin Americans to vote in Belgium, Luxemburg and the Netherlands.
  Models 1 and 3 confirm that compulsory voting in the country of origin has a positive impact on the turnout of Latin American non-resident voters in the Low Countries, independently of whether it concerns partially or fully compulsory voting settings. Models 2 and 4 allow us to include variables that specify compulsory voting settings in both countries of origin and residence to verify our hypotheses (compulsory voting in both countries is used as a reference category). Our first hypothesis is confirmed: turnout for non-resident citizens is higher in countries where voting is compulsory (as opposed to countries with voluntary voting), and the positive effect of compulsory voting is direct for both types of countries. Interestingly, turnout is higher when voting is compulsory in the country of origin compared with when voting is compulsory in the country of residence. This may indicate that the socialization effects in the country of origin (pre-migration) and in migrant groups composed of individuals sharing the same citizenship are more important than the re-socialization (in the country of residence), for compulsory voting.
  Finally, our second hypothesis is also confirmed as the effect of voluntary or compulsory voting seems to be more important when both countries apply the same rules. We observe in Models 2 and 4 that turnout is higher when both countries apply compulsory voting (as compared with when voting is compulsory in one out of two countries or when voting is voluntary in both). Non-resident citizens who live in Belgium or Luxemburg (two countries with compulsory voting) and are registered to vote in national or supranational elections in a Latin American country with compulsory voting (i.e. Brazil, Bolivia, Ecuador and Peru) tend to be more prone to vote as compared with other combinations detailed in Table 2.
  Regarding control variables, registered voters residing in the Low Countries that emigrated from Latin American countries that require a non-permanent active residence are more likely to vote than registered voters that came from countries with one-off active or automatic registration. In other words, tighter registration ensures a larger interest in voting as compared with a more flexible enrolment. Coinciding with the assumption of the importance of offering remote voting method(s) to non-resident voters, our variable ‘postal voting’ is positively associated with high rates of non-resident citizens’ turnout as opposed to in-person voting. Unsurprisingly, the type of elections plays a role in attracting voters and leads to higher or lower turnout figures. Presidential elections seem to attract more non-resident voters, while referendum and supranational elections are negatively associated with turnout.
  The Covid-19 pandemic has no consistent statistical impact on turnout. Yet it would be interesting to observe the interaction between compulsory voting and remote voting in times of the Covid-19 pandemic. As organizing an election domestically in the middle of a pandemic is challenging, organizing an electoral process externally is even more difficult given the permanent need to coordinate with countries of residence when the voting method is in-person. In that sense, remote voting could be a solution for increasing the electoral participation of non-resident voters. Several countries in our data set are currently considering implementing some forms of remote voting for their next national elections (see e.g. Mexico or Peru). In the 2021 Ecuadorian elections, the Internet and postal voting led to a significant increase in turnout among Ecuadorians living in North America (Dandoy & Umpierrez de Reguero, 2021).

  Table 3 Coefficients (std. err.) on the nexus of compulsory voting on non-resident citizens’ voter turnout

  		Model 1	Model 2	Model 3	Model 4
  Degree of compulsory voting
  	Voluntary voting (ref.)	-		-
  	Compulsory voting (within the country)	23.35*** (2.93)		25.11*** (3.56)
  	Compulsory voting (external voting rights)	10.97** (4.00)		11.12** (3.26)
  Migrants’ dual context on compulsory voting
  	Voluntary voting in both countries		-27.44*** (3.09)		-23.18*** (3.35)
  	Compulsory voting (only in the country of residence)		-24.65*** (2.97)		-17.59*** (3.39)
  	Compulsory voting (only in the country of origin)		-13.08*** (2.65)		-8.69** (3.08)
  	Compulsory voting in both countries (ref.)		-		-
  Overseas electoral districts		-8.36* (3.74)	-2.31 (2.76)	-7.09* (3.05)	-5.20† (2.89)
  Non-permanent active registration		31.88*** (5.39)	37.38*** (4.40)	30.72*** (4.33)	31.45*** (4.19
  Postal voting		41.95*** (6.68)	45.41*** (5.26)	45.24*** (4.95)	46.41*** (4.63)
  Migrant Stock (per country of residence, log)		-9.95*** (2.02)	-10.60*** (2.29)	-12.51*** (1.88)	-10.57*** (2.18)
  Presidential Elections (ref.)		-	-	-	-
  Legislative Elections		1.31 (3.23)	1.64 (2.69)	-3.61 (3.25	-5.21* (3.09)
  Referendum		-16.25*** (3.22)	-18.21*** (2.71)	-16.30*** (2.83)	-14.99*** (2.64)
  Supranational Elections		-12.10* (4.85)	-14.79*** (4.00)	-5.01 (4.85)	-9.62* (4.48)
  Covid-19 pandemic		-3.52 (2.90)	-4.85* (2.40)	5.11* (2.51)	4.29 (2.40
  Intercept		74.94*** (6.98)	98.92*** (7.43)	81.47*** (6.25)	94.37*** (7.09)
  Pooling effects		No	Yes	No	Yes
  Random Effects		Yes	No	Yes	No
  R-square		0.550	0.641	0.734	0.740
  Adj. R-Square		0.531	0.620	0.715	0.728
  N		194	194	147	147

  Notes: Dependent variable based on registered voters. p-value † < 0.10; * <0.05; ** <0.01; ***<0.001.

  We finally ran some robustness checks using the estimated population of the Latin American countries sampled in Belgium, Luxemburg and The Netherlands and excluded some data set observations that may distort the baseline results.¹⁴x In parallel, we estimated non-resident citizens’ voter turnout as the baseline models but excluded four data set observations corresponding to elections in Venezuela because of their autocratic classification, as well as the Bolivian 2019 electoral contest, which was ex post annulled. Comparing these findings with the baseline models (Models 1-4), there is no significant change. In Models 5 to 8 (see Table A2 in the Appendix), non-resident citizens’ turnout was calculated over the migrant stock per country of origin, instead of the number of registered voters.¹⁵x See Figure A1 to know why we opt for a log function. Results confirm the conclusion of our baseline models: turnout is higher when the degree of compulsory voting increases while turnout is higher when both countries of origin and residence implement compulsory voting.

• 6 Conclusion

  While we know that compulsory voting has a broad and positive impact on resident voters’ turnout (see, for instance, Blais, 2006; Cancela & Geys, 2016; Geys 2006; Stockemer, 2017), its impact on non-resident citizens’ voter turnout remains largely understudied (for an exception, see Frizzo & Mascitelli, 2017). This lack of academic interest is partly due to the novelty of the phenomenon (in most countries, non-resident citizens received voting rights only recently [e.g., Chile in our sample]) and to the scarcity of publicly available data on non-resident citizens’ voting behaviour at the residence country level. Yet in modern political systems, where significant proportions of the population reside in another country, the analysis of their decision to participate – or not – in elections is increasingly relevant.
  This article drew on an aggregate-level design, observing the voting behaviour of migrants from ten Latin American countries in polling stations located in Belgium, Luxemburg and the Netherlands. These sets of countries of origin and residence provided a diverse set of treatment-control scenarios where the combination of voluntary and compulsory voting created an original typology of voting obligation. Based on a quantitative analysis of the voting behaviour of Latin American migrants in the Low Countries, this article aimed to test whether different electoral contexts had an impact on turnout figures of these voters.
  Results allowed us to falsify the null hypotheses and demonstrated a direct and significant influence of compulsory voting on non-resident citizens’ turnout. Both contexts, namely the country of origin and the country of residence, affect migrants’ political participation meaningfully. Compulsory voting in either of the two countries has a positive impact on turnout (as compared with voluntary voting), while the effect is larger when voting is compulsory in countries of origin and in both countries in tandem.
  Our results shed light on the explanatory capacity of the theory of political (re-) socialization in electoral contexts. Migrants are submitted to a double process of political socialization: a primary socialization (e.g. in family and school contexts) in their country of origin, followed by a socialization in their groups of peers in their country of residence; and a second political socialization in their country of residence. This article demonstrated that – in the case of participation in elections in the country of origin – these two socialization processes combine their positive effects rather than cancel each other out. More research has to be done on the impact of the dual institutional context that an international migrant is immersed in, particularly in the field of electoral engineering.
  Similarly, future research on non-resident citizens’ turnout could be developed in three directions. First, it would be interesting to enlarge the present analysis by including countries that – unlike Belgium, Luxemburg and the Netherlands – present different voting rights for non-resident citizens to participate in their elections. As a result, we could observe diverging patterns and trends of political integration and discuss more deeply the current debate of migrants’ political (re-)socialization versus a complementary approach.
  Second, this article briefly touches on the relevance of some elements of the electoral system for understanding non-resident citizens’ turnout, such as the process of voter registration and the implementation of remote voting (be it postal or Internet voting). While the interaction between compulsory voting and (compulsory) registration or between compulsory voting and remote voting for resident voters has often been investigated in the case of resident citizens, it largely remains a black box for non-resident citizens.
  Finally, one cannot deal with the issue of compulsory voting without addressing its normative implications. One needs to disentangle the academic debate on the benefits versus the inconveniences of compulsory voting in the national arena from the identical debate in the transnational arena. While the objective of a high(er) voting turnout may be universal, the institutional and political tools to reach it may vary largely, depending on the type of voters. In addition, this normative debate would be enriched by a multiplication of empirical data (e.g. based on surveys or in-depth interviews) on the perception of compulsory voting by non-resident citizens.

• Supplementary material

  Replication files are available at Harvard Dataverse: https://doi.org/10.7910/DVN/KFZAJX.

References

• Acosta, D. (2018). The National Versus the Foreigner in South America. New York: Cambridge University Press.

• Ahmadov, A. K. & Sasse, G. (2016). A Voice Despite Exit: The Role of Assimilation, Emigrant Networks, and Destination in Emigrants’ Transnational Political Engagement. Comparative Political Studies, 49(1), 78-114. doi: 10.1177/0010414015600468.

• Bauböck, R. (2015). Morphing the Demos Into the Right Shape. Normative Principles for Enfranchising Resident Aliens and Expatriate Citizens. Democratization, 22(5), 820-839. doi: 10.1080/13510347.2014.988146.

• Belchior, A. M., Azevedo, J., Lisi, M. & Abrantes, M. (2018). Contextual Reasons for Emigrants’ Electoral Participation in Home Country Elections: The Portuguese Case. Journal of Contemporary European Studies, 26(2), 197-214. doi: 10.1080/14782804.2017.1413641.

• Bermúdez, A., Lafleur, J. M. & Escrivá, Á. (2017). Contribuyendo a la democracia en países de origen: el voto externo de los migrantes andinos. América Latina Hoy, 76, 35-54. doi: 10.14201/alh2017763554.

• Blais, A. (2006). What Affects Voter Turnout? Annual Review of Political Science, 9, 111-125. doi: 10.1146/annurev.polisci.9.070204.105121.

• Blais, A. & Carty, R. K. (1990). Does Proportional Representation Foster Voter Turnout? European Journal of Political Research, 18(2), 167-181. doi: 10.1111/j.1475-6765.1990.tb00227.x.

• Blais, A. & Galais, C. (2016). Measuring the Civic Duty to Vote: A Proposal. Electoral Studies, 41, 60-69. doi: 10.1016/j.electstud.2015.11.003.

• Boccagni, P. (2011). Reminiscences, Patriotism, Participation: Approaching External Voting in Ecuadorian Immigration to Italy. International Migration, 49(3), 76-98. doi: 10.1111/j.1468-2435.2011.00691.x.

• Boccagni, P. & Ramírez, J. (2013). Building Democracy or Reproducing ‘Ecuadoreanness’? A Transnational Exploration of Ecuadorean Migrants’ External Voting. Journal of Latin American Studies, 45(4), 721-750. doi: 10.1017/s0022216x13001181.

• Burgess, K. & Tyburski, M. D. (2020). When Parties Go Abroad: Explaining Patterns of Extraterritorial Voting. Electoral Studies, 66, 102-169. doi: 10.1016/j.electstud.2020.102169.

• Camatarri, S. & Baudewyns, P. (2022). Demand- and Supply-side Approaches to Migrant Political Integration: A Focus on the Low Countries. Politics of The Low Countries.

• Cancela, J. & Geys, B. (2016). Explaining Voter Turnout: A Meta-Analysis of National and Subnational Elections. Electoral Studies, 42, 264-275. doi: 10.1016/j.electstud.2016.03.005.

• Chaudhary, A. R. (2018). Voting Here and There: Political Integration and Transnational Political Engagement Among Immigrants in Europe. Global Networks, 18(3), 437-460. doi: 10.1111/glob.12171.

• Ciornei, I. & Østergaard-Nielsen, E. (2020). Transnational Turnout. Determinants of Emigrant Voting in Home Country Elections. Political Geography, 78, 102-145. doi: 10.1016/j.polgeo.2020.102145.

• Collyer, M. (2014a). A Geography of Extra-Territorial Citizenship: Explanations of External Voting. Migration Studies, 2(1), 55-72. doi: 10.1093/migration/mns008.

• Collyer, M. (2014b). Inside Out? Directly Elected ‘Special Representation’ of Emigrants in National Legislatures and the Role of Popular Sovereignty. Political Geography, 41, 64-73. doi: 10.1016/j.polgeo.2014.01.002.

• Dandoy, R. & Kernalegenn, T. (2021). Internet Voting from Abroad: Exploring Turnout in the 2014 French Consular Elections. French Politics, 19, 1-19. doi: 10.1057/s41253-021-00148-8.

• Dandoy, R. & Umpierrez de Reguero, S. (2021). E-voting and Non- Resident Citizens’ Voter Turnout. A Quasi-Experiment in Ecuador. In: K. Piet & M. Mario (Eds.), Proceedings of the 14th International Conference on ICT, Society and Human Beings 2021. IADIS Press, 30-36.

• Doherty, D., Dowling, C. M., Gerber, A. S. & Huber, G. A. (2019). Should I Cast an Ill-Informed Ballot? Examining the Contours of the Normative Obligation to Vote. American Politics Research, 47(5), 1081-1099. doi: 10.1177/1532673X18821359.

• El País (1982, August). Guinea Exuatorial votará su nueva Constitución el próximo día 15. https://elpais.com/diario/1982/08/11/internacional/397864814_850215.html.

• Erdal, M. B. (2016). Juxtaposing Pakistani Diaspora Policy with Migrants’ Transnational Citizenship Practices. Geoforum, 76, 1-10. doi: 10.1016/j.geoforum.2016.08.005.

• Erdal, M. B., & Oeppen, C. (2013). Migrant Balancing Acts: Understanding the Interactions between Integration and Transnationalism. Journal of Ethnic and Migration Studies, 39(6), 867–884. doi: 10.1080/1369183X.2013.765647

• Escobar, C. (2007). Extraterritorial Political Rights and Dual Citizenship in Latin America (Derechos Políticos Extra-Territoriales y Doble Ciudadanía en América Latina). Latin American Research Review, 43-75.

• Feitosa, F., Stiers, D. & Dassonneville, R. (2022). A Cause or a Consequence? A Quasi-Experimental Test of the Duty-Participation Nexus. Journal of Elections, Public Opinion and Parties, 32(2), 439-448. doi: 10.1080/17457289.2020.1819299.

• Finn, V. (2020). Migrant Voting: Here, There, in Both Countries, or Nowhere. Citizenship Studies, 24(6), 730-750. doi: 10.1080/13621025.2020.1745154.

• Fliess, N. (2021). Campaigning Across Continents: How Latin American Parties Link Up with Migrant Associations Abroad. Comparative Migration Studies, 9(1), 1-19.

• Frizzo, D. & Mascitelli, B. (2017). Brazilians Abroad: Emigrant Voting and Political Engagement. Cambridge Scholars Publishing.

• Galais, C. (2018). How to Make Dutiful Citizens and Influence Turnout: The Effects of Family and School Dynamics on the Duty to Vote. Canadian Journal of Political Science, 51(3), 599-617. doi: 10.1017/S0008423918000021.

• Galais, C. & Blais, A. (2016). Beyond Rationalization: Voting Out of Duty or Expressing Duty After Voting? International Political Science Review, 37(2), 213-229. doi: 10.1177/0192512114550372.

• Geys, B. (2006). Explaining Voter Turnout: A Review of Aggregate-Level Research. Electoral Studies, 25, 637-663. doi: 10.1016/j.electstud.2005.09.002.

• Goerres, A., Elis, J., Mayer, S. & Spies, D. (2021). Integration and Transnational Political Interest among Immigrant-Origin Voters: An Analysis of Germans from Turkey and their Children in the Immigrant German Election Study 2017. Retrieved from SSRN: https://papers.ssrn.com/sol3/papers.cfm?abstract_id=3927086

• Himmelroos, S. & Vento, I. (2022). The Tie that Binds? A Comparison of Ethnicity-Based Party Ties among Emigrated and Resident Citizens. Comparative Migration Studies, 10(1), 1-20.

• Hutcheson, D. S. & Arrighi, J. T. (2015). “Keeping Pandora’s (ballot) box half-shut”: A Comparative Inquiry into the Institutional Limits of External Voting in EU Member States. Democratization, 22(5), 884-905. doi: 10.1080/13510347.2014.979161.

• Jackman, R. W. & Miller, R. A. (1995). Voter Turnout in the Industrial Democracies During the 1980s. Comparative Political Studies, 27(4), 467-492. doi: 10.1177/0010414095027004001.

• Kernalegenn, T. & van Haute, É. (2020). Political Parties Abroad: A New Arena for Party Politics. Routledge.

• Lafleur, J. M. (2013). Transnational Politics and the State: The External Voting Rights of Diasporas. Routledge.

• Lafleur, J. M. & Sánchez-Domínguez, M. (2015). The Political Choices of Emigrants Voting in Home Country Elections: A Socio-Political Analysis of the Electoral Behaviour of Bolivian External Voters. Migration Studies, 3(2), 155-181.

• Lineira, R. & Vallès, J.M. (2014). Abstencion Diferencial en Cataluña y en la Comunidad de Madrid: Explicación Sociopolitica de un fenómeno urbano, Revista Española de Investigaciones Sociológicas, 146(1), 69-91. doi: 10.5477/cis/reis.146.69.

• Nemčok, M., & Peltoniemi, J. (2021). Distance and Trust: An Examination of the Two Opposing Factors Impacting Adoption of Postal Voting among Citizens Living Abroad. Political Behavior, 1–25. Doi: 10.1007/s11109-021-09709-7

• McCann, J. A., Escobar, C. & Arana, R. (2019). Mexicans and Colombians at Home and Abroad: A Comparative Study of Political Engagement. Latin American Research Review, 54(1), 16-34. doi: 10.25222/larr.191.

• Mügge, L., Kranendonk, M., Vermeulen, F. & Aydemir, N. (2021). Migrant Votes ‘Here’ and ‘There’: Transnational Electoral Behavior of Turks in the Netherlands. Migration Studies, 9(3), 400-422. doi: 10.1093/migration/mnz034.

• Paarlberg, M. A. (2017). Transnational Militancy: Diaspora Influence Over Electoral Activity in Latin America. Comparative Politics, 49(4), 541-562.

• Palop-García, P. (2018). Contained or Represented? The Varied Consequences of Reserved Seats for Emigrants in the Legislatures of Ecuador and Colombia. Comparative Migration Studies, 6(1), 1-20.

• Paul, R. (2013). Space, the Final Frontier of Political Socialisation Research: Geopolitical Contexts, Migrant Resocialisation and Political Remittances. In S. Abendschön (Ed.), Growing into Politics: Contexts and Timing of Political Socialisation (pp. 183-216). ECPR Press.

• Peltoniemi, J. (2018). Transnational Political Engagement and Emigrant Voting. Journal of Contemporary European Studies, 26(4), 392-410. doi: 10.1080/14782804.2018.1515727.

• Powell, G. B. (1986). American Voter Turnout in Comparative Perspective. American Political Science Review, 80(1), 17-43. doi: 10.2307/1957082.

• Rashkova, E. R. (2020). The Party Abroad: A New Modus Operandi for Political Parties. Parliamentary Affairs, 73(4), 839-855. doi: 10.1093/pa/gsaa039.

• Sapiro, V. (2004). Not Your Parent’s Political Socialization: Introduction for the New Generation. Annual Review of Political Science, 7, 1-23. doi: 10.1146/annurev.polisci.7.012003.104840.

• Stockemer, D. (2017). What Affects Voter Turnout? A Review Article/Meta-Analysis of Aggregate Research. Government and Opposition, 52(4), 698-722. doi: 10.1017/gov.2016.30.

• Superti, C. & Gidron, N. (2021). Too Old to Forget: The Dynamics of Political Trust Among Immigrants. Political Studies, 70(3), 624-654. doi: 10.1177/0032321720980899.

• Szulecki, K., Bertelli, D., Erdal, M. B., Coşciug, A., Kussy, A., Mikiewicz, G. & Tulbure, C. (2021). To Vote or Not to Vote? Migrant Electoral (dis) Engagement in an Enlarged Europe. Migration Studies, 9(3), 989-1010. doi: 10.1093/migration/mnab025.

• Tsuda, T. (2012). Whatever Happened to Simultaneity? Transnational Migration Theory and Dual Engagement in Sending and Receiving Countries. Journal of Ethnic and Migration Studies, 38(4): 631-649. doi: 10.1080/1369183X.2012.659126.

• UN DESA. (2020). International Migrant Stocks. United Nations, Department of Economic and Social Affairs.

• Umpierrez de Reguero, S. (2022). Making Nonresident Citizens’ Votes Count : External Voting in Latin American and Southern European Countries (1962–2021). Leiden University.

• Umpierrez de Reguero, S. & Dandoy, R. (2021). Should We Go Abroad? The Strategic Entry of Ecuadorian Political Parties in Overseas Electoral Districts. Representation, 1-18. doi: 10.1080/00344893.2021.1902850.

• Umpierrez de Reguero, S. A., Finn, V., Cartagena, V., & Baquerizo, G. (2020). El sufragio multiterritorial. Evidencia de migrantes andinos en Ecuador. Revista Andina de Estudios Políticos, 10(1), 126-151. doi: 10.35004/raep.v10il.187

• Umpierrez de Reguero, S. A., Yener-Roderburg, I. Ö., & Cartagena, V. (2021). Political Regimes and External Voting Rights: A Cross-National Comparison. Frontiers in Political Science, 3. Doi: 10.3389/fpos.2021.636734

• Umpierrez de Reguero, S., Wegschaider, K., & Bauböck, R. (forthcoming). Globalcit Dataset of Electoral Rights of Migrants. European University Institute.

• Vintila, D., Pamies, C. & Paredes, M. (2022). Electoral (Non)Alignment between Resident and Non-Resident Voters: Evidence from Spain. European Political Science.

• Wellman, E. I., Allen, N. A. & Nyblade, B. (2022). The Extraterritorial Voting Rights and Restrictions Dataset (1950-2020). Comparative Political Studies. doi: 10.1177/00104140221115169.

• White, S., Nevitte, N., Blais, A., Gidengil, E. & Fournier, P. (2008). The Political Resocialization of Immigrants: Resistance or Lifelong Learning? Political Research Quarterly, 61(2), 268-281. doi: 10.1177/1065912908314713.

Legislation References

• Constitution of Peru (1993). www.constituteproject.org/constitution/Peru_2021.pdf?lang=en.

• Law 26.859 (1997). Ley Orgánica de Elecciones. https://www.gob.pe/institucion/congreso-de-la-republica/normas-legales/368389-26859

• Law 20.960 (2016). Regula el derecho a sufragio en el extranjero. www.bcn.cl/leychile/navegar?idNorma=1095759.

• Law 28859 (August 3, 2006). Reglamento de Multas por Omisión al Ejercicio del Sufragio, Inconcurrencia o Negativa a la Instalación de la Mesa de Sufragio, Negativa a su Confirmación. https://portal.jne.gob.pe/portal_documentos/files/procesoselectorales/Documentos%20%20Procesos%20electorales/Normativas%20y%20Resoluciones%20-%20Elecciones%20Regionales%20y%20Municipales%202006/res_4247-06-reglamento.pdf

• Law 31032 (July 23, 2020). Ley Que Modifica El Artículo 21 De La Ley 26859, Ley Orgánica De Elecciones, Para La Creación De La Circunscripción De Electores Peruanos Residentes En El Extranjero. https://spijweb.minjus.gob.pe/ley-n-31032/

Appendix

• List of covered elections

  Bolivia
  Presidential elections: 2014, 2019, 2020
  National referendum: 2016

  Brazil
  Presidential elections: 2006, 2010, 2014, 2018

  Chile
  Presidential elections: 2017, 2021
  National referendum: 2020 (2)

  Colombia
  Presidential elections: 2010, 2014, 2018, 2022
  Legislative elections (senate): 2014, 2018, 2022
  Legislative elections (lower house): 2014, 2018, 2022
  Andean elections: 2014
  National referendum: 2018 (2)

  Costa Rica
  Presidential elections: 2014, 2018, 2022

  Dominican Republic
  Presidential elections: 2008, 2012, 2016, 2020
  Legislative elections: 2012, 2020

  Ecuador
  Presidential elections: 2006, 2009, 2013, 2017, 2021
  Legislative elections (national seats): 2007, 2009, 2013, 2017, 2021
  Legislative elections (reserved seats): 2007, 2009, 2013, 2017, 2021
  Andean elections: 2009, 2013, 2017, 2021
  National referendum: 2017, 2018

  Mexico
  Presidential elections: 2012, 2018
  Legislative elections (senate): 2018

  Peru
  Presidential elections: 2006, 2011, 2016, 2021
  Andean elections: 2006, 2011, 2016, 2021
  National referendum: 2018

  Venezuela
  Presidential elections: 2006, 2013

  Table A1 Descriptive statistics on non-resident citizens’ voter turnout

  	Valid votes / registered		Emitted votes / registered		Valid votes / migrant stock		Emitted votes / migrant stock
  	N	Mean (Std. Dev.)	N	Mean (Std. Dev.)	N	Mean (Std. Dev.)	N	Mean (Std. Dev.)
  Bolivia	8	67.2 (8.4)	8	68.8 (9.3)	8	8.5 (3.1)	8	8.7 (3.2)
  Brazil	8	46.5 (2.9)	8	49.4 (3.1)	8	25.9 (6.7)	8	27.5 (7.0)
  Chile	12	66.2 (5.5)	12	66.9 (5.4)	12	13.6 (3.2)	12	13.7 (3.3)
  Colombia	42	26.7 (19.7)	40	27.8 (21.4)	42	8.1 (7.0)	40	8.6 (7.3)
  Costa Rica	12	43.3 (13.0)	12	43.9 (13.1)	12	13.5 (4.4)	12	13.7 (4.4)
  Dominican Rep.	10	32.9 (9.0)	10	28.3 (12.0)	10	10.7 (5.8)	9	9.9 (5.0)
  Ecuador	56	48.7 (19.4)	14	36.2 (9.1)	56	13.8 (7.6)	14	14.0 (4.2)
  Mexico	9	87.0 (6.4)	9	87.3 (6.6)	9	19.9 (28.2)	9	20.0 (28.2)
  Peru	33	43.5 (16.4)	30	43.2 (12.9)	33	27.0 (25.5)	30	29.8 (29.1)
  Venezuela	4	65.2 (4.4)	4	65.6 (4.3)	4	25.5 (14.6)	4	25.6 (14.8)
  Total	194	45.8 (21.9)	147	44.4 (22.3)	194	15.4 (14.9)	146	16.6 (17.7)

  Table A2 Robustness Checks on the nexus of compulsory voting on non-resident citizens’ voter turnout

  		Model 5	Model 6	Model 7	Model 8
  Degree of compulsory voting
  	Voluntary voting (ref.)	-		-
  	Compulsory voting (within the country)	0.29*** (0.06)		0.33** (0.10)
  	Compulsory voting (external voting rights)	0.43*** (0.09)		0.51*** (0.09)
  Migrants’ dual context on compulsory voting
  	Voluntary voting in both countries		-0.61*** (0.06)		-0.68*** (0.08)
  	Compulsory voting (only in the country of residence)		-0.34*** (0.06)		-0.42*** (0.08)
  	Compulsory voting (only in the country of origin)		-0.28*** (0.05)		-0.30*** (0.07)
  	Compulsory voting in both countries (ref.)		-		-
  Overseas electoral districts		-0.17* (0.08)	-0.29*** (0.05)	-0.09 (0.09)	-0.22** (0.06)
  Non-permanent active registration		0.18 (0.12)	0.10 (0.09)	0.21 (0.12)	0.12 (0.09)
  Postal voting		0.21 (0.14)	0.13 (0.11	0.22 (0.14)	0.16 (0.11)
  Registered voters (ln)		0.07 (0.04)	0.15*** (0.04)	-0.01 (0.04)	0.10* (0.04)
  Presidential Elections (ref.)		-	-	-	-
  Legislative Elections		-0.08 (0.07)	-0.09 (0.05)	-0.13 (0.09)	-0.15* (0.07)
  Referendum		-0.19** (0.07)	-0.26*** (0.05)	-0.23** (0.08)	-0.29*** (0.06)
  Supranational Elections		-0.21* (0.11)	-0.21* (0.08)	-0.16 (0.14)	-0.14 (0.10)
  Covid-19 pandemic		0.16* (0.06)	0.17** (0.05)	0.21** (0.07)	0.22*** (0.05)
  Intercept		0.76*** (0.12)	1.11*** (0.11)	0.95*** (0.12)	1.28*** (0.13)
  Pooling effects		No	Yes	No	Yes
  Random Effects		Yes	No	Yes	No
  R-square		0.286	0.491	0.351	0.553
  Adj. R-Square		0.247	0.461	0.303	0.516
  N		194	194	146	146

  Notes: Dependent variable based on migrant stock. p-value † < 0.10; * <0.05; ** <0.01; ***<0.001.

  Non-resident citizens’ voter turnout calculated over the migrant stock
  Notes: For running these histograms, the number of observations was standardized and based on the same version of the data set as the one to run Models 5 and 6 (N=194). To improve the normality of the data, we apply a log function to Figure A1(a), obtaining Figure A1(b).

Noten

• 1 There are a few other exceptions when it comes exclusively to constitutional referenda such as Equatorial Guinea (see El País, 1982).

• 2 Sources such as the UN migrant stock may lead to over- or underestimating non-resident citizens’ turnout, since its data comprises population under 18 years old (Ciornei & Østergaard-Nielsen, 2020). Also, it does not consider migrants’ descendants who possess the right to participate in mostly all Latin American contexts.

• 3 Paraguay also belongs to that group of countries but is not included in our analyses as it did not establish polling stations either in Belgium or in Luxemburg or the Netherlands.

• 4 We only compiled information of Ecuador and Peru, since Paraguay has no polling stations in Belgium, Luxemburg or the Netherlands. Similarly, Venezuela does not disclose non-resident citizens’ voter turnout for the MERCOSUR Parliament. In the rest of the Latin American cases with active external voting rights, the existing provisions enable the electoral participation of their non-resident citizens in supranational elections.

• 5 Bolivia, Brazil, Chile, Colombia, Costa Rica, Dominican Republic, Ecuador, Mexico, Peru and Venezuela. As of the beginning of 2022, Cuba, Nicaragua and Uruguay do not apply a provision of external voting rights. Non-resident citizens’ voter turnout data from Argentina, Guatemala, Honduras, Panama, Paraguay and El Salvador in Belgium, Luxemburg and the Netherlands is inaccessible or non-existent. Owing to data availability constraints, several single election data at the polling station level are similarly missing, as, for instance, the 2014 presidential elections in Brazil or the 2016 legislative elections in the Dominican Republic. Even if the 2019 Bolivian elections were annulled, we kept the record of its turnout figures in our database.

• 6 In many cases, the electoral management body does not open a polling station for its national residents in Luxemburg. For instance, there are no records of separate polling stations in Luxemburg for the Brazilian elections.

• 7 See Table A1 in the appendix for the number of observations and descriptive statistics of the different operationalisations of the dependent variable, including the ones used in the robustness checks.

• 8 See Constitution (1993, Art. 31) and Organic Law of Elections (1997, Arts. 9 and 240). Law 27369 (2000, Art. 25). Since 2006, the authorities of the Peruvian electoral management body decided that non-resident citizens are exonerated from paying a fine if they are registered and do not vote (see Law 28859, August 3, 2006; Art. 4).

• 9 Chileans residing abroad shall change their domicile if they want to vote from their countries of residence “one hundred and forty days prior to each election or on the date of publication of the decree calling for a plebiscite, resuming from the first day of month following the election or plebiscite” (Law 20,960, 2016, Art. 28). Additionally, there is a residence-based eligibility for foreign-born Chileans; they need to have a certificate of past residence (certificado de avecinamiento) (see Umpierrez de Reguero et al., 2020).

• 10 In 2004, Venezuelan Electoral Management Body enacted a provisional resolution to request non-resident citizens a series of documents to enrol and thereafter to vote, including a residence permit of visa. Although the type of registration is similar to that of its neighbours (i.e. one-off registration), Venezuelan authorities made de facto more difficult the registration procedure since 2004, particularly to dual citizens, to control external voting rights (Umpierrez de Reguero et al., 2021).

• 11 Internet and postal voting were used in some polling stations in the 2021 elections in Ecuador but not in polling stations located in the Low Countries.

• 12 This is applicable only in the 2021 legislative election at the national level.

• 13 Both tests are commonly used to determine the nature of the regression analysis with panel data.

• 14 In parallel, we estimated non-resident citizens’ voter turnout as the baseline models but excluded four data set observations corresponding to elections in Venezuela because of their autocratic classification, as well as the Bolivian 2019 electoral contest, which was ex post annulled. Comparing these findings with the baseline models (Models 1-4), there is no significant change.

• 15 See Figure A1 to know why we opt for a log function.