Left Behind Together and Voting for Populism: Regional Out-Migration, Civic Engagement and the Electoral Success of Populist Radical Right Parties

Abstract

According to the academic debate, the populist radical right is particularly successful in regions that have been left behind economically or culturally. Although civic engagement in networks of civil society, a specific form of social capital, seems important, its influence remains ambiguous. In contrast, regional out-migration as a social dimension of being left behind receives limited attention despite the relevance of internal migration to political geography. This study investigates two theoretically possible models to clarify the relationships between regional out-migration, civic engagement, and their impacts on voting for the populist radical right. Using data from the German Socio-Economic Panel (SOEP) and official regional statistics, logistic multilevel analyses are conducted for Germany and the election of the AfD (Alternative for Germany) in the 2017 federal election. The key finding of the cross-sectional analysis is that regional out-migration is a condition that moderates the relationship between civic participation and the election of the AfD. In general, civically involved individuals support established democratic parties, but in regions with high out-migration, they tend to vote for the populist radical right. However, there is no empirical evidence that regional out-migration contributes to the election of the AfD by reducing civic engagement and being mediated by it.

1. Introduction

The Brexit referendum in 2016 on the UK’s leaving the European Union (EU) as well as Donald Trump’s success in the 2016 presidential election in the US have prompted a broad debate in academia. Increasingly, the level of analysis has changed. In explaining the successes of populist radical right parties, candidates, and programs, the sub-national level, its regional variations, and patterns have received more and more attention, so that one can speak of a localist turn in populism research (Chou et al. 2022). The new target of many research activities today is to explain the “Geography of Discontent” and voting behavior in “Left Behind Places”. This literature is based on the assumption that local disparities, in particular, have contributed to the electoral successes of the populist radical right. The contribution of Rodríguez-Pose (2018) uniquely emphasizes this aspect. In regions that do not matter, people tend to feel left behind, and the resulting discontent is expressed in their voting decisions. In particular, people turn away from established democratic parties and candidates who have neglected and ignored the long-term economic decline of their region. In voting against the established system, people in regions that have been left behind are taking revenge.

The debate is dominated by economic arguments (Broz et al. 2021; Dijkstra et al. 2020; Essletzbichler et al. 2018; McKay 2019; Rodríguez-Pose 2018), but also by cultural approaches (Cramer 2016; Fitzgerald 2018; Rodríguez-Pose et al. 2021; Wuthnow 2018), showing how regions have been left behind and where people voice their discontent towards the political establishment. While recent research has shown that internal migration has implications for political geography (Lee et al. 2018; Pickard et al. 2022; Shuttleworth et al. 2021), there is still a lack of clarity about the influences of living in regions that have been abandoned through out-migration and being left behind socially. Therefore, this study explores the question of how regional out-migration affects the voting for the populist radical right, but also the participation of citizens in networks of civic organizations, a specific form of social capital (Putnam 1993, 2000).

Two theoretical models are presented to examine how out-migration from a region, and civic engagement are related and contribute to the election of populist radical right parties. First, regional out-migration may foster the election of populist radical right parties. People feel socially marginalized and ignored by politics and express their dissatisfaction at the ballot box (Gidron and Hall 2020; Rodríguez-Pose 2018). Although civic engagement among individuals encourages the election of established democratic parties (Fitzgerald 2018), however, out-migration may also contribute to a decline in civic engagement (Lesage and Ha 2012; Spina 2017), which therefore explains the electoral success of populist radical right parties. Second, regional out-migration can also be a condition of the relationship between civic engagement and the election of populist radical right parties. Individuals express their dissatisfaction with being left behind in their civic networks, which not only becomes intensified and results in a turning away from established democratic parties, but also provides fertile ground for the populist radical right (Colombo and Dinas 2023; Rodríguez-Pose et al. 2021).

In showing that it is not those people who are left behind who vote for the populist radical right, but those left behind together, this study contributes to the literature on the “Geography of Discontent”, “Left Behind Places” and political geography. It is expanding their perspectives in revealing that the influence of internal migration should not be underestimated. Being left behind is of political importance not only for economic or cultural reasons but also for social matters when others leave the region. Furthermore, this study takes the regional as well as the individual level into account and finally clarifies the theoretical relationship between regional out-migration and participation in civil society.

This paper begins with an overview of the key factors in explaining regional differences and the ways, in which regions are left behind and foster the election of the populist radical right. Thereafter, the AfD (Alternative for Germany) and the German context are highlighted, and the two theoretical models are developed and critically discussed. The presentation of the data is followed by the results of the logistic multilevel models for the federal election (Bundestag) in 2017. After the findings have been described, there is a closing discussion.

2. Causes of the Regional Differences in Favoring the Populist Radical Right

In the academic debate on the “Geography of Discontent” and “Left Behind Places”, economic but also cultural approaches are at the center to explain the regional electoral success of the populist radical right. While the economic arguments offer rather unambiguous insights, the cultural approaches are more heterogeneous and need further research. Another explanatory approach prominent in the political geography literature relates to internal migration. Although it is central to geographic sorting, the relationship between regional out-migration and voting for the populist radical right is ignored.

2.1. Economic Decline of Regions

From an economic perspective, it is argued that many regions in the US, the EU, and the UK are in an economic decline, caused by globalization processes that affect certain sectors of the economy, wealth, income, and employment. An important factor in explaining the support for populist radical right parties is the regional decline of the manufacturing industry. This is how Trump achieved high political victories in the 2016 presidential election in regions that were confronted with such a long-term development but also currently have high manufacturing employment rates, such as the Midwest (Broz et al. 2021; Essletzbichler et al. 2018). A similar development can be observed in European regions. A long-term decline in industrial jobs in the region has led to increased support for populist parties with anti-elite and anti-establishment rhetoric in the 2017 European Parliament election, positioning themselves against the EU system and its integration (Dijkstra et al. 2020). Likewise, the UK leaving the EU has been especially popular in regions with a long industrial tradition that has been slow to adapt to structural change (Essletzbichler et al. 2018). In particular, those regions in Western Europe and the UK that are more exposed to increased global trade and competition, such as the increase in imports from China, are more likely to vote for populist radical right parties or to vote for the UK to leave the EU (Colantone and Stanig 2018a, 2018b).

The loss of economic prosperity in a region and weak growth in household income are other central aspects that have contributed to Trump’s electoral success (Broz et al. 2021). Anti-EU parties are also particularly strong in those regions that, although economically prosperous, have been economically stronger in the past and are currently experiencing a decrease in their wealth (Dijkstra et al. 2020). Compared to other advanced economies, the UK is characterized by one of the highest regional disparities in economic productivity, innovation spending, and income (McCann 2020). Precisely those regions have voted to leave the EU whose economic prosperity and labor income are most dependent on EU markets (Los et al. 2017). It can also be seen for Germany that a long-term, one could also say a historical, decline in regional income benefits the AfD in the 2017 federal election (Greve et al. 2023).

High unemployment in the region is also among the factors that promote the electoral success of the populist radical right. Regions in the US and the UK that are currently struggling with high unemployment, as well as those regions whose labor markets have been particularly affected by the economic shock of the global economic crisis in 2008, are more likely to vote for the populist radical right (Essletzbichler et al. 2018). For the US, it also appears that unemployment caused by the global economic crisis has occurred most severely in regions with a long-term decline in manufacturing (Broz et al. 2021). Equally, in European regions, low opportunities and high demand for employment in the region lead to support for populist parties (Dijkstra et al. 2020).

2.2. Cultural Disparities of Regions

In contrast, cultural approaches argue that the regional economy is not necessarily the decisive force for voters. Instead, regional disparities in place-based identities and social capital are also important for the electoral success of the populist radical right. Studies on place-based identities mainly focus on the rural–urban divide in political preferences and voting behavior (Cramer 2016; Hochschild 2016; Rodden 2019; Wuthnow 2018) recognized in the US (Gimpel et al. 2020), the Netherlands (Harteveld et al. 2022; Huijsmans et al. 2021), and Sweden (Rickardsson 2021). De Lange et al. (2023) provide empirical evidence for the Netherlands that individuals with a strong place-based identity are more likely to express regional resentment, although more so in peripheral rather than rural regions. Moreover, Ziblatt et al. (2020) have demonstrated for Germany that in such peripheral areas, the AfD has achieved particular electoral success. Two prominent arguments explore more deeply the perceptions of cultural differences between rural and urban populations and their worries about the local culture of the community. In her ethnographic research, Cramer (2016) argues that the rural population in the US has developed its own consciousness, from which resentments against urban areas and their inhabitants are generated, which in turn can be exploited by populist candidates and movements for their political purposes. The rural consciousness describes the rural population’s understanding that they participate unfairly in resource allocations despite their contributions to society. In their view, the urban population, which does not understand and has little appreciation for rural ways of life, is favored by the government, which is ignoring the rural population. A similar argument has been articulated by Wuthnow (2018), who is underscoring the cultural origins of the rural–urban divide. Building on his empirical qualitative research, he argues that rural residents live in moral communities in which individuals develop loyalties and obligations to the place, its inhabitants, and ways of life. From their perspective, the urban perceived government is considered to be a threat that either ignores the rural decline or interferes too much in the way of life without knowing the local moral order. Consequently, the rural people’s opposition to the government can be understood as a reaction to the threat to their moral communities. Both arguments provide plausible reasons why the populist radical right is more popular in certain regions than in others.

Another cultural approach suggests that regional differences in the electoral success of populist radical right parties can be explained by the level of social capital and civic engagement. Studies in this line of argument show significant as well as contradictory influences. Fitzgerald (2018) provides the first empirical quantitative evidence of the relationship between support for radical right parties and localism as the person’s feeling of belonging to a local community. Her central finding is that the attractiveness of radical right parties decreases especially for individuals who are actually engaged in community life, while radical right parties are more attractive for individuals who feel nothing more than a strong sense of local attachment. These results are also evident at the local area level. Consequently, high social capital at both the individual and regional levels reduces support for radical right-wing parties. This association is also observed by Chan and Kawalerowicz (2022), who show for the UK that individuals who support the Remain campaign have participated more actively in civic organizations compared to Leave supporters. Further evidence is provided by Bolet (2021), who has also found that a decline in sociocultural centers is associated with support for the radical right. She argues that the closure of community pubs in the UK signals a loss of community and cultural identity and leads to support for UKIP. However, in stark contrast to the previous studies, Rodríguez-Pose et al. (2021) find that low social capital in a region is not at all associated with support for Trump. Rather, it is the combination of strong social capital in the region along with an economic and demographic change that matters. Consequently, a long-term decline in employment and population in regions with relatively strong social capital is the cause of Trump’s electoral success. Although it is clear from these studies that social capital and civic engagement are relevant in explaining regional differences in voting behavior, their impact can vary.

2.3. Internal Migration and Being Left Behind Socially

Another approach for explaining regional differences in voting behavior is based on internal migration. In the research of political geography, it is mainly studied in the context of geographical sorting. By arguing that self-selection into politically compatible areas produces geographic polarization and homogeneous regions where political positions become more extreme, Bishop and Cushing (2008) stimulated a widespread debate. Subsequent studies have shown that people have political preferences regarding where they live and favor living among like-minded people (Gimpel and Hui 2015; Hui 2013). When people live in a region that is not in line with their political values and beliefs, this has negative consequences on their satisfaction with their neighborhood (Gimpel and Hui 2018; Hui 2013; Lütjen and Matschoß 2015), their sense of belonging (Motyl et al. 2014), and social relationships (Chopik and Motyl 2016). Finally, when people move, they choose regions that match their worldviews (Cho et al. 2013; Gallego et al. 2016; Maxwell 2019; McDonald 2011), but it has been observed that people also move to politically incongruent regions (Cho et al. 2019; Mummolo and Nall 2017).

Internal migration is also examined in light of the rejection and support of the populist radical right. In the context of the referendum on the UK leaving the EU, it has been younger people and those who have moved a greater distance who have supported the Remain campaign (Shuttleworth et al. 2021). Interestingly, some impact has also been evident after the referendum. The likelihood of moving away from the region seems to decrease when the regional Brexit referendum result matches individuals’ political preferences. However, individuals sort themselves into politically compatible regions according to their political attitudes if this is not the case (Pickard et al. 2022). In contrast, it has been particularly individuals who are locally rooted and live in the region where they were born who have voted in favor of leaving the EU (Chan and Kawalerowicz 2022; Lee et al. 2018). Similarly, in the French presidential election runoff, those who have lived in the community for at least 10 years are more likely to have voted for populist radical right candidate Marine Le Pen (Rassemblement National) (Patana 2022).

While this extensive research has shown that internal migration has implications for the political geography and the populist radical right, the impact of living in regions with high out-migration and being left behind on a social dimension is still unclear. Against the background of the literature discussed so far, it is important to recognize that regions and individuals may not only be left behind economically or culturally. Due to high out-migration, regions and individuals can also be left behind socially by their former residents or fellow citizens. But there is little knowledge about whether voters in regions with high out-migration are more likely to support the populist radical right or not. Additionally, the previous discussion of the literature also highlights that the participation of individuals in civil society organizations holds a special place in the explanation for voting for established parties. However, due to the different findings, there is also a lack of clarity about whether civic engagement favors or counteracts their electoral successes. Both phenomena need further investigation. But the present study goes beyond the aim of examining only their association with voting behavior. In addition, this paper examines the relationship between regional out-migration and engagement in civil society organizations and how both phenomena contribute to the electoral success of the populist radical right.

3. Populist Radical Right Voting and How It Is Influenced by the Relationship between Regional Out-Migration and Civic Engagement

Before discussing the theoretically possible relationships between outward migration from regions and individuals’ engagement in civic networks and their different influences on the voting decision for populist radical right parties, it is first essential to outline the German context and the central terms on which this study is based. Beginning with the evolution of the AfD as a populist radical right party, a brief outline of its geographic distribution is provided. Subsequently, the central internal migration movements in Germany are shown to demonstrate which regions are particularly affected by regional out-migration. Finally, the definition of civic engagement according to Putnam (1993, 2000) is given.

3.1. Alternative for Germany and Its Regional Strongholds

The AfD has been founded as a political response to the European financial crisis in 2013. Based on its election programs, policy positions, and electorate in the German and European Parliament elections in 2013 and 2014, it has not been classified as a complete populist radical right party in the political science literature. Rather, it has been perceived as a Eurosceptic single-issue party due to its rejection of the European currency, even though populist views towards political elites and right-wing positions regarding immigration and family have already been present in its political communication (Arzheimer 2015; Berbuir et al. 2015; Schmitt-Beck 2017). In the wake of the refugee crisis in 2015, a transformation of the party’s content and its electorate has occurred. From now on, anti-immigration positions and attitudes not only complement Euroscepticism but have become the most influential issue and relevant electoral motive, bringing the AfD in line with other populist radical parties in Europe (Arzheimer and Berning 2019; Goerres et al. 2018; Hansen and Olsen 2019, 2022). Since the AfD entered the German Parliament in 2017, this party family has been fully represented in the multiparty system, which is based on proportional representation.

Looking at the regional distribution of AfD voters, there is a clear divide between West and East Germany, where the AfD performs best (Schmitt-Beck 2017; Weisskircher 2020; Arzheimer and Berning 2019; Hansen and Olsen 2019; Martin 2019; Goerres et al. 2018). First, in its election manifestos, the AfD addresses the national and specifically the regional identity in the East (León and Scantamburlo 2022). For another, growing up in the post-socialist East and the former GDR is also a reason (Goerres et al. 2018). Compared to the West, the attachment to political parties is weaker (Schmitt-Beck 2017). Populist and nativist attitudes are also more prevalent, especially among those who were politically socialized in the GDR, while ethnonationalism and xenophobia contribute most to AfD voting (Pesthy et al. 2021). Another difference between West and East Germany can be seen in the rural–urban divide in the AfD’s electoral success. In the East, it is the rural regions with a high over-aging population in which the AfD can mobilize voters (Deppisch et al. 2022; Franz et al. 2018). However, the rural-urban divide should not hide the fact that there are also differences in the AfD’s mobilization potential within cities (Förtner et al. 2021) where it finds particular support in economically, politically, and socially neglected neighborhoods (Mullis 2021).

3.2. Internal Migration and Left Behind Regions

The emergence of socially left behind regions is caused by two predominant patterns of internal migration in Germany: first, the migration between East and West Germany, and second, the migration between rural regions and cities. After the fall of the Berlin Wall in 1989, there have been two major migration movements from East to West Germany, primarily to economically strong regions. The first movement (1989–1990) occurred immediately after the fall of the Wall and decreased until the late 1990s (Heiland 2004). A total of 1.2 million people migrated from the East to the West between 1989 and 1998, equivalent to 7% of the GDR population in 1989 (Kemper 2004). By the turn of the millennium, economic stagnation in the East and increased job opportunities in the West caused a second migration movement (1997–2001), which decreased and converged with in-migration from the West to the East thereafter (Heiland 2004; Stawarz et al. 2020). In contrast to the first movement, the second stream consisted mainly of younger and highly skilled entrants to the labor force from rural regions and, until 2008, mainly of younger or single women (Fuchs-Schündeln and Schündeln 2009; Heiland 2004; Hunt 2006; Kröhnert and Vollmer 2012; Stawarz et al. 2020). In 2017, East Germany experienced its first population increase, and more people moved from the West to the East (Stawarz et al. 2020).

Regarding migration between rural and urban regions, different dynamics in population distribution have been found. Before and after reunification, there has been a spatial population concentration in East Germany and a deconcentration in the West. Over time, the distribution patterns have converged, with suburbanization in both parts of the country in the mid-1990s and reurbanization in the late 1990s (Herfert and Osterhage 2012; Kemper 2004, 2008; Kontuly et al. 1997; Köppen 2008; Siedentop 2008). It has been driven by the economic structural change in cities towards a knowledge-based economy with corresponding labor market and educational opportunities (Gans 2018; Geppert and Gornig 2010; Siedentop 2008). Due to the real estate market with rising rents and housing shortages in cities (Henger and Oberst 2019; Stawarz et al. 2021), it has been replaced by suburbanization in 2011 (Stawarz and Sander 2020). It becomes clear that the two patterns of internal migration from East to West and from rural to urban regions have resulted in disadvantaged regions, with eastern and rural regions being particularly affected.

3.3. Civic Engagement

Given the political and geographical location of the AfD and the regions that have lost most of their population due to regional out-migration, it is further essential to situate the civic engagement of individuals within a conceptual framework. Their participation in networks of civil society can be seen as a special form of social capital. In general, social capital can have many positive effects on societies. Research has shown that countries with high levels of social capital exhibit a greater degree of well-being and life satisfaction (Gómez-Balcácer et al. 2023; Rodríguez-Pose and Von Berlepsch 2014). Also, people are happier when neighborhoods are equipped with large amounts of social capital (Hoogerbrugge and Burger 2018). Furthermore, it is positively associated with the good health of individuals and communities (Ahnquist et al. 2012; Poortinga 2012), enhanced business innovations (Pérez-Luño et al. 2011), and reduced crime and violence (Buonanno et al. 2009; Rosenfeld et al. 2001).

Although there are various concepts of social capital—the contributions of Bourdieu (1986) and Coleman (1988) belong to the classical approaches—the concept of Putnam (1993, 2000) is used in this study to provide greater comparability since his approach is widely used in research. In general, Putnam describes social capital as “connections among individuals—social networks and the norms of reciprocity and trustworthiness that arise from them” (Putnam 2000, p. 19). According to his understanding, these interpersonal relationships have both a private and individual dimension and a public and collective dimension (Portes 2000) and can be expressed in different forms. On the one hand, the organization of relationships can be either informal or formal in nature. On the other hand, their purpose can be based on purely private or public concerns or involve both (Putnam 2000, p. 20f). In this study, the civic engagement of individuals is regarded as (1) a personal good, not a property of the regional community, and (2) a formally organized network in clubs, associations, or initiatives that are (3) focused on public life in society.

3.4. Model 1: Civic Engagement and Its Mediating Function

In the following, two theoretical models are introduced that show how regional out-migration and engagement of individuals in networks of civic organizations are related and foster the election of populist radical right parties. The constellations of their particular relationships to each other are shown graphically in Figure 1. Model 1 illustrates the theoretical possibility that regional out-migration could lead to the election of populist radical right parties among the left behind individuals (a). It is also conceivable that, due to many people moving away from a region, the civic engagement of the remaining citizens will diminish (b). The declining civic engagement, which otherwise has motivated people to vote for parties of the democratic mainstream (c), may enhance the support of populist radical right parties. Consequently, the first model describes the situation in which it is not regional out-migration per se that leads to the election of populist radical right parties but is rather mediated by the associated decline in civic engagement. The relations (a, b, c) outlined in Model 1 can be discussed critically.

Regarding the relationship between regional out-migration and the support for populist radical right parties (a), two mechanisms can be identified. First, individuals who leave a region may differ in attitudes, values, and beliefs from the individuals who remain in that region. The outward migration of those individuals thus changes the composition of the region without changing the attitudes, values, and beliefs of those left behind as well as those leaving. Previous research addresses the phenomenon that people select or sort themselves into places that are politically compatible according to their political party preferences (Bishop and Cushing 2008; Gallego et al. 2016; Cho et al. 2013). There is empirical evidence suggesting that cities become cosmopolitan places through the in-migration of cosmopolitan-minded individuals, and it seems unlikely that the urban context induces an attitudinal change in individuals (Maxwell 2019, 2020). This is a compositional change, which is to the disadvantage of populist radical right parties in this case, especially since there is a negative correlation between their support and cosmopolitanism (De Vries 2018).

Second, individuals’ political preferences can change as a result of regional out-migration. Adapting the argument of Rodríguez-Pose (2018), high out-migration from a region leads to a sense of being left behind. Regions and individuals can be left behind not only in economic terms, but also in social terms with the departure of friends, colleagues, and family members. According to the “Theory of Social Integration” by Gidron and Hall (2020), this social marginalization can lead to grievance among those left behind. In their approach, if individuals come into contact with others less often, are less socially active, trust others less, and feel less valued by others, a sense of social marginalization and discontent arises due to their lack of social integration. These feelings can be expressed against established democratic parties, policy-makers, and political elites, who seem to neglect and ignore the voters left behind. According to the authors, populist radical right parties can exploit this voter discontent, advocate on their behalf, and in the end generate political capital in this region.

Both mechanisms appear to be relevant in explaining the election of populist radical right parties through regional out-migration. Dancygier et al. (2022) conclude that there is a positive relationship between net migration loss and vote share for populist radical right parties in European regions but also evidence for Sweden that the regional out-migration changes both the composition of the electorate and the political preferences of left behind individuals. Although her study focuses on international out-migration in Central and Eastern European Countries, Lim (2022) comes to a similar conclusion. She finds that in regions with high rates of out-migration to other countries, the vote shares of far-right parties are higher and that the comparatively younger, more educated, and more progressive migrants change the composition of the region of origin with their departure. But she also finds that out-migration has an impact on the political preferences of those left behind in the regions of origin. In this study, a distinction between the two mechanisms is not made, as the central concern is to identify the relationship and its overarching theoretical embedding.

Regional out-migration and its relationship with civic engagement (b) can also be critically discussed, as studies do not come to a definitive conclusion. Putnam (2000) agrees that internal migration can have an impact on the social network in the region and on the individual. However, he concludes that migration from one region to another is not the main reason for the decline of social capital in the US in the last century. Glaeser and Redlick (2009) agree with him, finding no correlation between migration rates and social capital for metropolitan areas in the US. In contrast, Lesage and Ha (2012) provide evidence at the county level that out-migration does harm social capital, while in-migration has a positive effect. In comparison, out-migration even seems to have a greater impact.

Regarding the relationship between regional out-migration and the election of populist radical right parties, the same question arises about whether the composition of civic active individuals or their preference for involvement has changed as a result of internal migration. Hotchkiss and Rupasingha (2021) present findings for the US that relocation in recent years has been unlikely for individuals who are equipped with high personal social capital. This is especially true for individuals whose own personal social capital is greater relative to that of their region. Individuals who, among other types of social capital, are involved in the community, politically engaged, and active seem more likely to stay in their area, which would contradict a change in composition by moving. For the change in preference of being involved in the community, Spina (2017) provides an interesting explanatory approach. Although focusing on external migration and regional outward migration to other countries, he describes the cooperation dilemma of individuals living in regions with high out-migration, which is also applicable to internal migration. The remaining individuals begin to lose confidence in their interactions with others, and exchanges are no longer based on continuity and reciprocity. The generalized trust of individuals left behind toward other fellow citizens is increasingly eroded. Consequently, individuals abandon activities within the community and turn more to their closer social environment, leading to a decline in social capital in the region.

Membership in civil society organizations and its influence on the support of populist radical right parties (c) is also not without controversy. Putnam (1993, 2000) argues that a strongly developed social capital in society has a positive outcome for democracy. In particular, when individuals actively participate in civil society, civic virtues are cultivated among the citizens that keep modern democracies protected from extremist tendencies. Indeed, as mentioned above, Gidron and Hall (2020) show that socially integrated voters tend to prefer populist parties less often and mainstream parties instead. In general, however, empirical studies provide mixed evidence. Rydgren (2009) confirms a negative correlation in Western European countries, but in explaining individuals’ radical right-wing voting behavior, this is not particularly relevant, as is the fact of whether individuals are socially isolated or not. In Central and Eastern European Countries, Rydgren (2011) even finds no correlation. A difference in the voting for radical right-wing parties between civically active individuals and individuals without any civic engagement does not seem to exist here. However, on a contextual level, with a small geographical unit such as Flemish municipalities in Belgium, Coffé et al. (2007) have found that an active associational life in the community does reduce the vote share of the extreme right-wing Vlaams Blok. Concerning the threat to sociocultural centers of a local community, the sympathy of voters for populist radical right parties can rise, as Bolet (2021) illustrates with the closure of community pubs in the UK and the support of UKIP. On the question of whether the individual or contextual level is relevant, current research concludes that both dimensions have an impact. Fitzgerald (2018) reveals that the popularity of populist radical right parties is low, both for individuals who are actively engaged in civic organizations and in areas that are characterized by a strong local civil society.

Despite the controversially discussed relationships above, this study examines the mediating function of civic engagement on the positive relationship between regional out-migration and populist radical right voting. This first theoretical model (see Figure 1) is based on the following hypothesis:

Hypothesis 1.

If individuals live in a region with high out-migration, their engagement in networks of civil society organizations declines, which increases the likelihood to vote for populist radical right parties.

3.5. Model 2: Regional Out-Migration as a Condition

The second theoretical model (Figure 1) describes the possibility that regional out-migration is a condition (d) and influences the relationship between civic engagement and populist radical right voting (c). It is quite plausible that the relationship between individuals’ participation in civil society and the election of populist radical right parties described above (c) can also be positive under certain circumstances. Meaning that people who are active in civil society organizations are more likely to vote for populist radical right parties. For the Weimar Republic, Satyanath et al. (2017) show that in cities with a high density of associations, party entries into the NSDAP were faster and the NSDAP performed better in elections. Social capital apparently had a considerable part in the collapse of the first democracy in Germany.

Colombo and Dinas (2023) formulate a reasonable explanation, called the “Networks of Grievances”. According to their argument, if individuals express their grievances in the social networks to which they belong, their discontent can be intensified on the one hand. On the other hand, their voiced grievances can spread faster in the networks and promote a systematic rejection of established parties. In their study, the authors were able to show that in regions with economic difficulties, the impact of people’s economic discontent on their voting behavior has been influenced by a pronounced social capital in their region and people have been more likely to turn away from established parties. Under the condition of a shrinking economy, social capital increases discontent and leads to a spread of grievances and greater support for populist radical right parties. Rodríguez-Pose et al. (2021) come to a similar conclusion. In the context of the US, they show that in regions experiencing economic decline and a decrease in population size, high social capital within regions generates support for Trump. It is plausible to assume that regional out-migration is an equally important condition (d). Individuals who live in regions with high out-migration and who are involved in social networks due to their engagement in civic organizations can express their discontent about their feeling of being left behind in those networks, which spreads faster and finally provides fertile ground for populist radical right parties. Accordingly, the underlying hypothesis of Model 2 (Figure 1) is:

Hypothesis 2.

If people live in a region with high out-migration, their engagement in networks of civic organizations increases the likelihood to vote for populist radical right parties.

4. Data and Methods

4.1. Individual and Regional Data

In order to test the two theoretical models, survey data from the German Socio-Economic Panel (SOEP) (Goebel et al. 2019) are used at the individual level. At the regional level, the analysis is based on information about internal migration from the Federal Statistical Office of Germany and the Statistical Offices of the Länder1 and regional data from the Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR) (BBSR 2021). The SOEP is a representative and annually repeated longitudinal study of private households and their members in Germany since 1984. The survey includes data on voting decisions in the 2017 federal election, civic engagement, and the socioeconomic situation of individuals. Based on the spatial identifiers, it is possible to assign respondents to the region in which they live. The geographical unit used for the data analysis are regions at the NUTS 3 level (Nomenclature des Unités territoriales statistiques). In Germany, this corresponds to the administrative unit of “Kreise und kreisfreie Städte” or counties.

The Federal Statistical Office of Germany and the Statistical Offices of the Länder provide data on annual internal migration between counties. These data refer exclusively to migration behavior within the German federal territory and exclude external migration across German state borders. All long-term and officially registered outward movements from and inward movements to counties are considered. Information is available on the citizenship, age group, and gender of people moving both in and out of the county for the period from 1991 to 2017. Following the territorial reforms in this period, the data have been harmonized by the BBSR. The BBSR also offers annual and harmonized data on counties, which provide information on the population structure as well as the economic conditions of the counties. Unfortunately, the territorial status is not the same for all data sources. While the data from SOEP and the Federal Statistical Office Germany and the Statistical Offices of the Länder are at the territorial status for the year 2017, the data from BBSR are at the territorial status for the year 2019.2

4.2. Variables

4.2.1. Dependent Variable

Support for the populist radical right represents the outcome variable and is measured by the vote for the AfD.

Table 1. Descriptive statistics.

	Obs	Mean	SD	Min	Max
Individual level
AfD voting	14,329	0.087	0.282	0	1
Civic engagement	14,329	0.264	0.441	0	1
Male	14,329	1.469	0.499	1	2
Age	14,329	52.237	16.521	18	102
Education
Low level	14,329	0.256	0.436	0	1
Medium level	14,329	0.285	0.452	0	1
High level	14,329	0.156	0.363	0	1
Highest level	14,329	0.303	0.459	0	1
Log HH net income (EUR 1000)	14,329	1.115	0.567	−1.609	6.908
Occupational position
White collar	14,329	0.097	0.296	0	1
Blue collar	14,329	0.414	0.493	0	1
Civil servants	14,329	0.055	0.228	0	1
Self-employed	14,329	0.063	0.243	0	1
No status	14,329	0.371	0.483	0	1
Economic worries
No worries	14,329	0.435	0.496	0	1
Some worries	14,329	0.451	0.498	0	1
Great worries	14,329	0.114	0.318	0	1
Migration background	14,329	0.117	0.321	0	1
Regional level
Internal migration
In-migration	14,329	0.321	0.467	0	1
Balanced migration	14,329	0.379	0.485	0	1
Out-migration	14,329	0.300	0.458	0	1
GDP per capita (EUR 1000)	14,329	36.264	15.478	15.068	144.092
Unemployment rate (%)	14,329	6.558	2.769	1.356	14.504
Employment 2nd sector (%)	14,329	24.942	9.305	6.728	54.178
Regional type
Urban	14,329	0.505	0.500	0	1
Suburb	14,329	0.216	0.412	0	1
Rural	14,329	0.278	0.448	0	1
East Germany	14,329	0.195	0.396	0	1

Note: unweighted; data: German Socio-Economic Panel (v37) (Goebel et al. 2019), Federal Statistical Office Germany and the Statistical Offices of the Länder, Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR 2021); own calculations.

provides descriptive details such as the mean, standard deviation, minimum and maximum value of the voting behavior, and the other variables used in the data analysis. In the 2018 survey, the participants of the SOEP have been asked which party they voted for in the election of the German Parliament on 24 September 2017. Multiple party combinations are possible, as voters have two votes in the federal election. A variable with two categories has been constructed for the voting decision. Category 0 contains all persons who have not given any of their votes to the AfD and at least one of their votes to a party that has been represented in the German Parliament after the election. This concerns the parties of the Christian Union (CDU/CSU), Social Democrats (SPD), Greens (Bündnis 90/Die Grünen), The Left (Die Linke), and Liberals (FDP). In contrast, category 1 includes all persons who have placed at least one of their votes for the AfD. The focus is on parliamentary parties only, therefore other party combinations or non-voters are not taken into account.

4.2.2. Independent Variables

Civic engagement, the most important independent factor at the individual level, has been constructed with two items. In the survey in 2017, SOEP respondents have been asked: “Which of the following activities do you take part in during your free time? Please check off how often you do each activity”. For the activities “Volunteer work in clubs or social services” and “Involvement in a citizens’ group, political party, local government” respondents have been able to indicate whether they engage in this activity at least once a week, at least once a month, less often, or never. For the variable civic engagement, a variable with two values has been created. The value 0 is assigned to all persons who are never or less often voluntarily or politically active. People who are active in a civic or political function at least once a month or week are assigned the value 1.

Regional out-migration is the most important independent factor at the context level, which is based on the internal migration balances of the counties. For persons with German citizenship, they are calculated as the difference between annual inflow and outflow across the county border and within the federal territory, based on the county’s population, and multiplied by 1000.3 According to their average internal migration balances between the years 2013 and 2017, the counties are divided into three equally sized categories, so that one-third of the counties are in each category. The first category includes counties that have a surplus of in-migration. Counties with a balanced migration are grouped in the second category (reference). The last category includes counties that have experienced more out-migration on average in the last five years before the election.

4.2.3. Control Variables

At the individual level, several control variables have been taken into account. Earlier studies indicate clearly that the AfD has a greater appeal to male voters compared to female voters (reference) (Hansen and Olsen 2019, 2022; Arzheimer and Berning 2019; Goerres et al. 2018; Dilling 2018). However, there are inconclusive findings regarding the influence of age. While some studies show no significant effect (Schmitt-Beck 2017; Goerres et al. 2018; Wurthmann et al. 2021), others suggest that younger individuals (Hansen and Olsen 2019, 2022; Arzheimer and Berning 2019) or those in middle age (Dilling 2018) are more likely to support the AfD. In this study, age and squared age are considered. Furthermore, this study considers socioeconomic factors such as education, income, and occupation, as well as economic worries. For the educational level of voters, some investigations have found no effect (Goerres et al. 2018; Hansen and Olsen 2019), but others have provided evidence that lower levels of education lead to higher levels of support for the AfD (Arzheimer and Berning 2019; Hansen and Olsen 2022). Therefore, the educational level is taken into account. It is based on the Comparative Analysis of Social Mobility in Industrial Nations (CASMIN) scale and reduced into four groups. The first and lowest educational group (reference) includes persons who have basic vocational training and/or inadequately or adequately completed general elementary education. Persons with an intermediate general and/or vocational qualification constitute the second group with a medium level of education. Persons with a general and/or vocational maturity certificate belong to the third educational group. The last and highest educational group includes persons who have lower or higher tertiary education. Persons who are still in school are not included. Furthermore, people with lower incomes are more likely to vote for the AfD (Hansen and Olsen 2019, 2022), although it has been shown that income does not always have to make a difference (Goerres et al. 2018). In the models, the log household net income (EUR 1000) is used. Regarding the impact of occupations, previous research has not found significant effects (Arzheimer and Berning 2019; Goerres et al. 2018; Schmitt-Beck 2017), but the argument from Adorf (2018) seems plausible that blue-collar workers favor the AfD due to its welfare chauvinist position. The included occupational positions are divided into five groups: blue-collar workers (1st group, reference), white-collar workers (2nd group), civil servants (3rd group), self-employed persons (4th group), and persons without such a status like unemployed, retired or trainees (5th group). The concerns about the voters’ own economic situation also seem to have an impact on voting for the AfD (Goerres et al. 2018), but sometimes they do not (Hansen and Olsen 2019). Three groups have been added to the models. The first group has no worries (reference), the second has moderate worries and the last group has great worries about their economic situation. Lastly, recent studies have found that ethnic Germans who migrated from the former Soviet Union States (“late re-settlers”) are more likely to lean toward the AfD (Hansen and Olsen 2020; Spies et al. 2023). Therefore, the migration background is considered, which is given if one of the parents or the person in question is not born with German citizenship. At the regional level and in line with the literature on the “Geography of Discontent” and “Left Behind Places” (Broz et al. 2021; Essletzbichler et al. 2018; Dijkstra et al. 2020), it is useful to control for the economic prosperity of the region, the labor market, and, in particular, the share of manufacturing and its impact on the election of the AfD. A relevant indicator of regional prosperity and economic performance is the Gross Domestic Product (GDP) per capita, measured in EUR 1000 per inhabitant. The unemployment rate among the labor force (%) provides information on the state of the local labor market. Previous research on AfD showed inconclusive results. For weak labor markets, it could be observed that the AfD achieved high (Martin 2019) but also low election results (Goerres et al. 2018). A good measurement of occupational groups whose activities are affected by modernization is provided by employment in the second sector of the economy. It describes the share (%) of employees who are subject to social insurance contributions and work in the manufacturing sector in relation to the total number of employees subject to social insurance contributions at the workplace. For all variables, the average values have been calculated from 2013 to 2017. Based on the previous discussion on the geographic distribution of the AfD in Germany, it is further necessary to look at the rural–urban divide (Deppisch et al. 2022; Ziblatt et al. 2020) and the disparities between East and West Germany (reference) (Schmitt-Beck 2017; Weisskircher 2020; Goerres et al. 2018). In order to control for differences between urban, suburban, and rural regions, regional types have been constructed. These are based on population density per square kilometer for the five years before the election.4 Urban counties (reference) have a population density of 300 or more inhabitants per square kilometer. They also include independent cities which are not assigned to a county but represent a county of their own (“Kreisfreie Städte”). Suburban counties have a density of 150 to less than 300 inhabitants per square kilometer, while rural counties have a density of fewer than 150 inhabitants per square kilometer. Furthermore, certain individuals and counties are excluded from the data analysis. These are persons without the right to vote, such as persons under the age of 18 and persons without German citizenship, as well as the six counties mentioned above that are affected by territorial reforms. A total of 14,329 observations in 394 counties are included in the calculations.

4.3. Methods

For the cross-sectional analysis, logistic multilevel regression models are used since the dependent variable is dichotomous and the data set has a hierarchical data structure in which individuals are nested in counties. In order to examine the first theoretical model, a mediation analysis is conducted in which the civic engagement of individuals represents the mediator. The aim is to test whether the effect of regional out-migration on the election of the AfD can be explained by civic engagement, in the sense that regional out-migration reduces participation in civic organizations, which in turn would affect the voting for the AfD. Therefore, two statistical models are calculated, one without taking into account the civic involvement of individuals and one with their civic engagement. To ensure the comparability of the two models, this study follows the recommendation of Best and Wolf (2012) and uses the method of Karlson et al. (2012) in combination with average marginal effects (AME) as output. However, for a clearer understanding of the two models, the logistic regression Equations (1) and (2) are given as the logit of the odds for voting the AfD. They are as followed:

$$\mathit{ln}\left(\frac{P\left({AFD}_{ij}=1\right)}{P\left({AFD}_{ij}=0\right)}\right)={\gamma }_{00}+{\gamma }_{10}{RESIDUALS}_{ij}+{\gamma }_{k0}{K}_{ij}+{\gamma }_{01}{MIGRATION}_j+{\gamma }_{0k}{K}_j+{\mu }_{0j}$$

(1)

$$\mathit{ln}\left(\frac{P\left({AFD}_{ij}=1\right)}{P\left({AFD}_{ij}=0\right)}\right)={\gamma }_{00}+{\gamma }_{10}{CIVIC}_{ij}+{\gamma }_{k0}{K}_{ij}+{\gamma }_{01}{MIGRATION}_j+{\gamma }_{0k}{K}_j+{\mu }_{0j}$$

(2)

where ${\gamma }_{00}$ represents the fixed intercept, ${\gamma }_{k0}{K}_{ij}$ are the coefficients with the level 1 control variables of individual $i$ in county $j$, ${\gamma }_{01}$ is the coefficient for regional internal migration balance ${MIGRATION}_j$ of county $j$, ${\gamma }_{0k}{K}_j$ are the coefficients with the level 2 control variables of county $j$, and ${\mu }_{0j}$ is the random intercept. In line with Karlson et al. (2012), the first equation contains ${\gamma }_{10}{RESIDUALS}_{ij}$ as the coefficient for the residuals of individual $i$ in county $j$, obtained from an OLS regression with civic engagement as the dependent variable and the other variables as independent variables (except for voting the AfD). Compared to Model 1, ${\gamma }_{10}{RESIDUALS}_{ij}$ is excluded in the second equation. Instead, ${\gamma }_{10}{CIVIC}_{ij}$ is included, which is the coefficient with the civic engagement of individual $i$ in county $j$.

In order to test the second theoretical model, a moderation analysis is performed with regional internal migration balance as a moderator. The aim here is to assess whether regional out-migration is a condition for the effect of participation in civic organizations on the election of the AfD. A third statistical model is calculated in which a product term between the civic involvement of the individuals and the internal migration balance of the counties has been added. Since this is a cross-level interaction, the recommendation of Heisig and Schaeffer (2019) is followed and a random slope term for the variable at the first level of the interaction is included in the model, in this case, civic engagement. As for models 1 and 2, the results of Model 3 are given as AME, and Equation (3) is also provided for the log odds for casting a vote in favor of the AfD:

$$\begin{array}{ll}\mathit{ln}\left(\frac{P\left({AFD}_{ij}=1\right)}{P\left({AFD}_{ij}=0\right)}\right)=& {\gamma }_{00}+{\gamma }_{10}{CIVIC}_{ij}+{\gamma }_{k0}{K}_{ij}+{\gamma }_{01}{MIGRATION}_j+\\ & {\gamma }_{11}{CIVIC}_{ij}{MIGRATION}_j+{\gamma }_{0k}{K}_j+{\mu }_{1j}{CIVIC}_{ij}+{\mu }_{0j}\end{array}$$

(3)

Here, ${\gamma }_{11}{CIVIC}_{ij}{MIGRATION}_j$ stands for the interaction term between the civic engagement of individual $i$ in county $j$ and the internal migration of county $j$, while ${\mu }_{1j}{CIVIC}_{ij}$ depicts the random slope for the civic engagement of individual $i$ in county $j$. Robust standard errors are clustered for the countries for all three statistical models. The data analysis has been carried out with Stata 14.1.

5. Results

The effects on voting the AfD are presented as average marginal effects in

Table 2. Logistic multilevel analysis for the AfD vote in the 2017 German parliament election.

	Model 1		Model 2		Model 3
	AME	SE	AME	SE	AME	SE
Individual level
Civic engmt			−0.035 ***	(0.005)	−0.056 ***	(0.009)
Male	0.046 ***	(0.004)	0.047 ***	(0.004)	0.047 ***	(0.004)
Age	0.005 ***	(0.001)	0.006 ***	(0.001)	0.006 ***	(0.001)
Age (squared)	−0.000 ***	(0.000)	−0.000 ***	(0.000)	−0.000 ***	(0.000)
Education (low level)
Medium level	−0.035 ***	(0.009)	−0.032 ***	(0.009)	−0.032 ***	(0.009)
High level	−0.079 ***	(0.009)	−0.075 ***	(0.009)	−0.075 ***	(0.009)
Highest level	−0.082 ***	(0.009)	−0.077 ***	(0.009)	−0.078 ***	(0.009)
Log HH net income (EUR 1000)	−0.025 ***	(0.005)	−0.023 ***	(0.005)	−0.023 ***	(0.005)
Occupation (white collar)
Blue collar	0.029 **	(0.009)	0.028 **	(0.009)	0.028 **	(0.009)
Civil servants	−0.017	(0.012)	−0.016	(0.012)	−0.015	(0.012)
Self-employed	0.006	(0.008)	0.008	(0.009)	0.008	(0.009)
No status	0.002	(0.007)	0.004	(0.007)	0.003	(0.007)
Economic worries (no worries)
Some worries	0.030 ***	(0.006)	0.029 ***	(0.006)	0.029 ***	(0.006)
Great worries	0.054 ***	(0.008)	0.051 ***	(0.008)	0.051 ***	(0.008)
Migration background	0.010	(0.009)	0.007	(0.009)	0.007	(0.009)
Regional level
Int. migration (balanced)
In-migration	0.007	(0.008)	0.006	(0.008)	0.001	(0.009)
Out-migration	−0.006	(0.008)	−0.006	(0.008)	−0.014	(0.009)
Int. migration × civic engmt
In-migration × civic engmt					0.030	(0.016)
Out-migration × civic engmt					0.048 **	(0.015)
GDP per capita (EUR 1000)	−0.000	(0.000)	−0.000	(0.000)	−0.000	(0.000)
Unemployment rate (%)	0.004 *	(0.002)	0.004 *	(0.002)	0.004 *	(0.002)
Employment 2nd sector (%)	0.001 *	(0.001)	0.001 **	(0.001)	0.001 **	(0.001)
Regional type (urban)
Suburb	0.001	(0.010)	0.002	(0.010)	0.002	(0.010)
Rural	−0.010	(0.009)	−0.008	(0.009)	−0.008	(0.009)
East Germany	0.058 ***	(0.010)	0.057 ***	(0.010)	0.057 ***	(0.010)
N (level 1)	14,329		14,329		14,329
N (level 2)	394		394		394
AIC	7492.464		7492.464		7487.476
BIC	7681.715		7681.715		7699.437
Log-likelihood	−3721.232		−3721.232		−3715.738
McFadden’s pseudo-r2	0.098		0.098		0.099

Note: *** p < 0.001, ** p < 0.01, * p < 0.05; shown are average marginal effects; (robust standard errors clustered for counties in parentheses); null model: AIC (8251.207), BIC (8266.347), log-likelihood (−4123.603); unweighted; data: German Socio-Economic Panel (v37) (Goebel et al. 2019), Federal Statistical Office Germany and the Statistical Offices of the Länder, Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR 2021); own calculations.

. Model 1 shows a non-significant effect for regional out-migration. According to the first theoretical model, out-migration also should have a positive effect on the election of the AfD, but the effect is negative. For this reason, it cannot be claimed that counties with high out-migration in the last five years before the election are more likely to favor the AfD compared to counties with a balanced internal migration. Whether individuals live in a region with high out-migration or high in-migration is irrelevant to their propensity to vote for the AfD. At this point, it is already clear that individuals’ civic engagement does not function as a mediator. For the assumptions of the first theoretical model to hold and mediation to occur, there must be a significant effect of regional out-migration, but this is not the case. Thus, it cannot be said that regional out-migration decreases individuals’ civic involvement and thus promotes the election of the AfD.

In Model 2, the results indicate that the negative and insignificant effect of regional out-migration remains unchanged when individuals’ participation in civic organizations is taken into account and included in the model. As expected, the civic engagement of individuals is shown to have a significant effect on voting decisions. The effect on voting for the AfD is negative, which implies that individuals who are involved in civic organizations voted less often for the AfD and more often for established democratic parties in the federal election than those who are not civically engaged. Although the involvement in networks of civil society organizations seems to have an important role in advancing values and beliefs that oppose the AfD, the first theoretical model and its associated Hypothesis 1 must nevertheless be rejected.

In Model 3, an interaction between the internal migration balance of the county and individuals’ civic engagement is included to examine the moderating effect of regional internal migration on the relationship between civic involvement and individuals’ voting decision. The results demonstrate that the interaction term between regional out-migration and civic engagement is positive and significant. On the one hand, this means that regional out-migration moderates the influence of civic engagement on voting decisions. On the other hand, this also means that individuals involved in networks of civic organizations and living in counties with high out-migration have a stronger preference for the AfD and are more likely to have voted for this party than for established democratic parties compared to those not civically engaged and living in counties with balanced internal migration. These are interesting findings, as the results reveal that in general, civically active individuals vote less often for the AfD than those who are not engaged, but under the condition of living in counties with high out-migration, individuals who participate in civil society organizations favor and support the AfD more strongly in the federal election. Consequently, Hypothesis 2 cannot be rejected, and the second theoretical model is empirically confirmed.

For the control variables at the individual level, the analyses show that gender, age, educational level, the log household net income, occupational position, and economic worries affect the voting decision. Men have a stronger preference for AfD than women, who are more likely to vote for established democratic parties. Age does not seem to have a linear effect on the voting decision and the small influence appears to be negligible. Individuals with higher levels of education and log household net income have been more likely to vote for mainstream democratic parties, while the AfD is more popular among people with a lower level of education, income, and greater concerns about their economic situation. A person’s occupational position seems to be partly relevant. It can be said that blue-collar workers favor the AfD more strongly than white-collar workers, while there are no differences in preferences for the AfD or established democratic parties between civil servants, self-employed persons, persons without a professional position, and white-collar workers. Whether a person has a migration background or not is irrelevant. People with a migration history do not favor or reject the AfD more strongly than people without a migration background.

For the control variables at the regional level, unemployment in a county has a significant positive effect on the AfD vote. In counties with a higher lack of employment, the AfD is more likely to be elected. Also, the magnitude of the second sector of the economy has an impact on the voting decision of individuals in the county. Although the influence is minimal, individuals tend to support the AfD if there are more employees in the secondary sector in their county. Furthermore, it makes a considerable difference in the voting decision depending on whether a person lives in the old or the new federal states. In particular, the AfD is more strongly supported in Eastern German counties than in Western German counties. This effect is independent of the county’s economic situation, and labor market over the last five years before the election. This suggests that other factors, such as cultural differences or historical events like the German Unification and their processes afterward, might still play an important role in the decision for or against the AfD. Interestingly, the findings show that the economic performance of a county has no influence and that there are no differences in voting preference between urban, suburban, and rural counties. In this respect, all counties support or reject the AfD to the same extent.

For the overall evaluation of the goodness of fit, the McFadden’s pseudo-r² of 0.098 and 0.099 show moderate improvement or goodness of fit for the models, relative to the null model. This is less than perfect but an acceptable outcome. A comparison of the goodness of fit between the models is difficult to make. On the one hand, Models 1 and 2 cannot be compared with each other since the applied method according to Karlson et al. (2012) produces the same AIC and BIC values. On the other hand, their comparison with Model 3 does not provide any clear results, since it cannot be concluded whether the addition of the interaction improves the model or not. In the model comparison, the AIC value decreases, while the BIC value increases, and this to extremely small degrees, so that no model seems to be better than the other one.

6. Discussion

This study aims to analyze the electoral success of populist radical right parties in regions that have been left behind. In this context, the economic decline of regions and the feelings of being culturally ignored have been addressed extensively in the literature on the “Geography of Discontent” and “Left Behind Places”. Internal migration, however, has been largely neglected in this debate. This is surprising since internal migration is not only relevant to the political geography but also has an important social dimension for the remaining individuals who are left behind by fellow citizens moving away from the region. Until now, it has been unclear to what extent such regional out-migration and the involvement of individuals in networks of civil society organizations are related and foster or constrain the election of populist radical right parties. This study closes this research gap and outlines two theoretical models explaining the relationships.

The key finding of this research is that, in a certain sense, it is those individuals who have been left behind together that vote for the populist radical right. One theoretical explanation refers to the situation in which internal migration between regions moderates the relationship between individuals’ involvement in civic organizations and the election of populist radical right parties. For Germany, this theoretical model finds empirical evidence. In counties with high regional out-migration in the last five years before the election, left behind individuals who are active in civil society have a stronger tendency to vote for the AfD than individuals who are not civically engaged and live in counties with a balanced migration. The impact of the individuals’ civic engagement is dependent on the condition of internal migration. Consequently, it is not only the economic decline of a region or its population decline that provokes regions with high social capital to vote for the populist radical right (Colombo and Dinas 2023; Rodríguez-Pose et al. 2021), this phenomenon can also be observed in regions with high out-migration. Furthermore, populism is not only related to social and local disintegration (Fitzgerald 2018; Gidron and Hall 2020) but also, under a certain condition, to social integration in specific networks. Not by any means do civically engaged citizens always follow democratic principles (Satyanath et al. 2017). It is important to emphasize that this condition is independent of both the economic position of individuals and their economic worries, as well as the economic situation of regions. Argued with the findings of Colombo and Dinas (2023) it seems plausible that the discontent about regional outflows and feelings of being left behind are also articulated in the networks of civil society organizations, intensifying and spreading in this network and resulting in systemic disapproval of established democratic parties and providing fertile ground for the populist radical right.

The other theoretical explanation describes the situation in which regional out-migration leads to the support of populist radical right parties. The civic engagement of individuals, which usually promotes the election of established democratic parties, is reduced by regional out-migration, and thus serves as a mediator for outward migration and populist radical right voting behavior. The findings of this study show that these theoretical relationships are not empirically supported. In contrast to other studies (Dancygier et al. 2022; Lim 2022), regional out-migration does not lead to a higher voting preference for the AfD. Opposed to previous studies (Bishop and Cushing 2008; Gallego et al. 2016; Cho et al. 2013), this suggests firstly that there has been no selection or sorting of individuals with certain values and beliefs, and consequently no change in regional composition which would favor the election of the AfD. Second, there is no evidence to suggest that regional out-migration has produced a general social marginalization of all those left behind in the region, causing discontent and a change in their political preferences that are directed against established democratic parties, although this could have been expected based on the literature (Gidron and Hall 2020; Rodríguez-Pose 2018). Furthermore, there is no indication that regional out-migration lowers the civic engagement of individuals (Glaeser and Redlick 2009; Putnam 2000). It seems neither likely that individuals who have been civically engaged have also moved away along with regional out-migration (Hotchkiss and Rupasingha 2021). Nor is it likely that out-migration has created a cooperation dilemma, in which people who have been left behind lose trust in their fellow citizens and no longer engage in collective activities with them as, as demonstrated by earlier research (Spina 2017). Nevertheless, this study confirms that in general, the involvement of individuals in civil society organizations leads to the rejection of the populist radical right and support for established democratic parties (Fitzgerald 2018; Gidron and Hall 2020).

This study has its limitations. Despite a high number of observations, it is not possible to rule out the eventuality that the dropouts of individuals and counties from the sample are not random. This has implications for the representativeness of the results and generalizability, which are limited. One critical point is that the logistic multilevel models conducted represent a cross-sectional analysis and that no conclusions can be made about causal relationships. Another critical point is that changes at the regional level have not been investigated, and only statements about the conditions of the counties can be made. However, this study illustrates that the moderating impact of internal migration is relevant and requires further attention. Future research should be encouraged to validate and extend the finding with further studies. First, to corroborate the evidence, it is reasonable to examine more specific forms of internal migration and to clarify whether the living conditions, region of origin, or destination of those leaving are important. In this study, only the internal migration of persons with German citizenship is considered, but additional information on the characteristics of internal migrants would be useful. Second, it would be worth exploring in more detail what kind of civil society organizations are in question and to what extent other forms of social networks and aspects of social capital, such as generalized trust or norms of reciprocity, are affected. Third, studies are needed that address the specific causes of what exactly lies behind the relationship that socially integrated individuals in left behind regions are more likely to support the populist radical right parties. It is possible that, in addition to the discontent over social marginalization, individuals have also formed a local identity that populist radical right parties are addressing and using for mobilization. This knowledge would not only be scientifically relevant but also of societal interest. After all, democracies are at risk when civically involved individuals in left behind regions reject established democratic parties, while populist radical right parties make political capital out of their situation.