Introduction
In many European countries, populist radical right (PRR) mobilization displays clear spatial patterns. The disproportionate success of the Rassemblement National in south-eastern and more recently north-eastern France, or the concentration of the ‘Leave’ vote in English coastal towns and former industrial centres, are prominent examples.
While the phenomenon of spatial patterns in political preferences is hardly a new one, scientific interest in the geography of support for the PRR has grown recently, because such patterns are interpreted as expressions of a broader polarization with which western societies are wrestling – a conflict that pits tolerant and cosmopolitan citizens against their illiberal, nationalist compatriots (Bornschier, Reference Bornschier and Rydgren2018). What sets the current discourse apart from similar but older debates about a rightist backlash against the alleged hegemony of left-libertarian values (see, e.g., Ignazi, Reference Ignazi1992) is that it is explicitly couched in terms of (political) geography. As a result, the role of ‘place’, broadly defined, has recently become a major issue in the study of the PRR. For the period from 2018 to 2022 alone, a cursory search of the Social Science Citation Index for titles that mention ‘radical right’ in conjunction with any mentions of ‘rural’, ‘urban’, ‘local’, ‘geography’, or ‘place’ yields 317 hits.
However, ‘place’ is obviously a complex, multi-faceted phenomenon that can be linked to PRR support in various ways. While most authors focus on proximate causes for contemporary mobilization (see, e.g., Harteveld et al., Reference Harteveld, van der Brug, Lange and van der Meer2021), a smaller segment of the literature aims to explain the present with events and developments in the (sometimes distant) past (see, e.g., Hoerner et al., Reference Hoerner, Jaax and Rodon2019; Haffert, Reference Haffert2022). Similarly, supposedly contextual effects are at the core of many studies, but some others also consider composition effects (see, e.g., Maxwell, Reference Maxwell2019). Finally, while much of the literature is concerned with easily quantifiable, macro variables (e.g., local unemployment rates), others highlight the importance of citizens’ subjective perceptions of and personal ties to their surroundings (see, e.g., Cramer, Reference Cramer2016; McKay, Reference McKay2019; Munis, Reference Munis2020; de Lange et al., Reference de Lange, van der Brug and Harteveld2022; Huijsmans, Reference Huijsmans2023a). This wide array of potential variables, putative mechanisms, and empirical findings can be bewildering.
Thus, while place seems to matter for PRR sentiment, it is far from clear how exactly place affects individual attitudes. In this paper, we therefore aim to make two contributions to this question, one conceptual, the other substantive.
Conceptually, we reconstruct four different yet interlinking meanings or aspects of ‘place’ from extant research. More specifically, we argue that many, if not most, of the variables and mechanisms discussed in the literature can be grouped under the labels of (1) place-related attitudes, (2) place-specific living conditions, (3) socio-demographic composition, and (4) features unique to a place. By distinguishing between these four aspects of place, we hope to structure the convoluted debate on how ‘place’ might be linked to PRR mobilization in principle.
Empirically, we employ geo-referenced data with a fine-grained spatial resolution to assess the relative importance of these four aspects of place for PRR sentiment in Germany. The data were collected in the aftermath of the so-called refugee crisis and during the first phase of the campaign for the 2017 general election – two events that had the potential to activate previously latent attitudes and make them salient.
Although its cold-war division certainly sets it apart from its European neighbours, Germany provides a compelling case study, and most of our findings should be applicable to other European countries. On the one hand, Germany, like other large European states, exhibits regional disparities in economic, political, social, and historical factors influencing PRR sentiment (Fina et al., Reference Fina, Osterhage, Rönsch, Rusche, Siedentop, Zimmer-Hegmann and Danielzyk2019). On the other, Germany has been reunited for almost as long as it was divided, and there are no ethno-linguistic cleavages or parties that campaign for regional independence. Germany features strong redistributive mechanisms and a constitutional commitment to reduce inequalities in living conditions between regions, which makes it more of a challenge to politically exploit such heterogeneity. Evidence for geographic polarization under these circumstances would therefore be particularly interesting.
Unlike some other studies on the geography of the PRR, we focus not on observed or reported behaviour, but instead on PRR sentiment. There are two reasons for this approach. First, such attitudes are believed to be the driving force behind a whole host of behaviours, from everyday racism and voting for PRR parties to taking part in anti-refugee protests, hate speech, and politically motivated violence. Second, the link between attitudes and behaviour itself is often conditional on supply side factors. This is especially clear in electoral behaviour, where a rightist voter’s choices – abstaining, supporting the PRR, or backing a mainstream party – hinge on the organizational strength, candidates, and programmatic profile of PRR and mainstream parties in their constituency (Mudde, Reference Mudde2010). Focusing on attitudes removes at least some of these complications.
The remainder of this paper is structured as follows. In the next section, we develop our analytical framework. More specifically, we discuss and define what we mean by PRR attitudes and how these could be affected by the four aspects of ‘place’ outlined above. Next, we give an overview of the data and the structure of our statistical model. Following that, we present our empirical results. Finally, we discuss the implications of our findings both for politics and for political science research.
Analytical framework
The sub-dimensions of PRR sentiment: nativism, right-wing authoritarianism, populism
The literature on right-wing sentiment is large but can be structured by making use of Mudde’s (Reference Mudde2007) influential conceptualization of PRR ideology. While Mudde was originally concerned with ideas espoused by parties and social movements, his typology, which identifies nativism, right-wing authoritarianism, and populism as core elements of PRR thinking, also applies at the micro-level.
Nativism is a mixture of nationalism and xenophobia which holds that non-native elements are a danger to the homogeneous nation state (Mudde, Reference Mudde2007: 19). At the micro-level, it corresponds to attitudes such as Islamophobia and perceptions of cultural threat (Mudde, Reference Mudde2019: 27–28).
For authoritarianism , Mudde (Reference Mudde2007: 22–23) explicitly draws on a micro-level approach (Altemeyer, Reference Altemeyer1981), whose concept of right-wing authoritarianism comprises authoritarian submission, authoritarian aggression, and conventionalism, leading to demands for law-and-order policies, strong leadership, and subordination of out-groups (Aichholzer and Zandonella, Reference Aichholzer and Zandonella2016).
Finally, populist attitudes are the micro-level equivalent of an ideational conception of populism. They encompass demand for unrestricted popular sovereignty and a belief in a homogeneous and virtuous people subjugated by a corrupt elite (Akkerman et al., Reference Akkerman, Mudde and Zaslove2014: 1332–1335).
While nativism, authoritarianism, and populism are closely related, they are conceptually and empirically distinct (see, e.g., Harteveld et al., Reference Harteveld, van der Brug, Lange and van der Meer2021). Therefore, we analyse the impact of the various aspects of place on each of them separately.
Four aspects of place, their links to PRR attitudes, and their application to the German case
To fully understand the nexus of geography and PRR attitudes, a nuanced conceptualization of ‘place’ is necessary. In this section, we expand on the heuristic that we sketched in the introduction. Based on a close reading of the literature, we identify four different aspects of ‘place’ that may affect PRR attitudes. We also discuss the ways in which these mechanisms interlink.
Place-related attitudes
This first understanding of place concerns an individual’s feelings towards its place of residence. Two attitudes are particularly important here: localism and place resentment.
Localism is the feeling of belonging to one’s locality. It entails a sense of emotional closeness to the local community, feelings of pride, and striving for more local authority and representation (Fitzgerald, Reference Fitzgerald2018: 10). Localism can easily lead to PRR attitudes because there are structural parallels: like nativism and populism, localism is based on a logic of demarcation between in-groups and out-groups and may entail a ‘feeling of distance from loci of power’ (Fitzgerald, Reference Fitzgerald2018: 6, 11).Footnote 1
The positive feeling of localism may be complemented by the negative sentiment of place resentment: a perception of regional exclusion. It results from the discrepancy between the levels of recognition, representation, or other benefits a locality ought to receive in the respondent’s view on the one hand, and the perception of what it actually gets on the other (Munis, Reference Munis2020: 3–4; de Lange et al., Reference de Lange, van der Brug and Harteveld2022: 2; Huijsmans, Reference Huijsmans2023a: 4). It is directed against outsiders held responsible for or colluding in this situation. In short, place resentment is what people feel in ‘places that don’t matter’ (Rodríguez-Pose, Reference Rodríguez-Pose2018).
In her seminal study of politics in Wisconsin, Cramer (Reference Cramer2016) linked these feelings to a ‘rural consciousness’. But while rural populations in many countries do indeed struggle with severe problems, place resentment is not restricted to rural residents (de Lange et al., Reference de Lange, van der Brug and Harteveld2022; Munis, Reference Munis2020: 2). Rather, place resentment is related to economic, cultural, and political inequalities between regions (Huijsmans, Reference Huijsmans2023a). This is particularly relevant for Germany, where the weak rural-urban cleavage partly intersects with an East-West cleavage (Fina et al., Reference Fina, Osterhage, Rönsch, Rusche, Siedentop, Zimmer-Hegmann and Danielzyk2019).
The connections between place resentment and PRR attitudes are straightforward. Both localism (Fitzgerald, Reference Fitzgerald2018: 27) and place resentment (Cramer, Reference Cramer2016: 89) can be understood through the lens of place-identity theory (Proshansky et al., Reference Proshansky, Fabian and Kaminoff1983), which posits a close connection between place and identity. Perceptions of local deprivation and decline fuel feelings of threat (Huijsmans, Reference Huijsmans2023b: 290). Because of the place-identity link and because residents already feel that their problems are the fault of outsiders, immigrants or Muslims become convenient scapegoats (Cramer, Reference Cramer2016: 85–88; Fitzgerald, Reference Fitzgerald2018: 6, McKay, Reference McKay2019: 8).Footnote 2 If resentment is directed against remote elites or out-groups more generally, populism and authoritarianism ensue (Huijsmans, Reference Huijsmans2023b: 289). In a bid to capitalize on place-related attitudes, radically right populist actors frequently try to make these links salient (Fitzgerald, Reference Fitzgerald2018: 175).
As place-related attitudes are a relatively new concept, research on their role in Germany is limited and focused on the East-West divide. Hildebrandt and Trüdinger (Reference Hildebrandt and Trüdinger2021) show that attachment to East Germany or the GDR provokes dislike of foreigners, concluding that regional identity can indeed have a dark side. Similarly, the underrepresentation of East Germans in the elite seems to be linked to lower political support and higher resentment amongst East Germans (Betz and Habersack, Reference Betz, Habersack, Heinisch, Massetti and Mazzoleni2020: 123–124; Vogel, Reference Vogel2022), fuelling what Gidron and Hall (Reference Gidron and Hall2020) call ‘feelings of social marginalization’.
Place-specific living conditions
A second central understanding of place holds that the way people think and act is shaped by their surroundings, and that an individual’s immediate environment is particularly formative (see, e.g., Johnston et al., Reference Johnston, Propper, Burgess, Sarker, Bolster and Jones2005). Most voters are acutely aware of their local context and interact with it daily, rendering such effects very plausible.
In connection with the PRR, the debate focuses mostly on structural deprivation, with affected regions often being described as ‘left-behind’ (McKay, Reference McKay2019). While deprivation appears in many forms, economic deprivation receives the most attention. Although economic issues are not at the top of the agenda of PRR parties (Golder, Reference Golder2016: 480), economic downturns are thought to reinforce political dissatisfaction and offer an opportunity to blame migrants or ‘the elite’. However, the empirical evidence for this link is inconclusive (Golder, Reference Golder2016: 484; Arzheimer, Reference Arzheimer and Rydgren2018: 155).
Deprivation is by no means limited to the economic situation. Regions suffering from a shrinking population or a gender imbalance due to ageing or out-migration are affected by demographic decline, which can create a vicious circle: maintaining infrastructure for a smaller population becomes more expensive, leading to cuts in services, which cause even more people to move away. Demographic decline is a therefore a direct cause of political discontent and ethnocentrism (Salomo, Reference Salomo2019: 109–114). Moreover, because those who move away are mainly the highly educated, high-income earners, women, and people who are more open-minded (Salomo, Reference Salomo2019: 106), demographic decline may indirectly cause higher levels of PRR sentiment through changes in the socio-demographic composition (see the next section).
Immigration is another important structural variable. Group threat theories posit that migrants compete for scarce resources like jobs or cultural dominance, fostering feelings of threat (Blumer, Reference Blumer1958; Blalock, Reference Blalock1967). Such competition should be especially visible at the local level (Bolet, Reference Bolet2020). However, the presence of immigrants does not always induce nativism, as the effect of ethnic diversity is highly context-specific and depends on a number of factors: where the immigrants are from (Rydgren and Ruth, Reference Rydgren and Ruth2013: 712–713; Savelkoul et al., Reference Savelkoul, Laméris and Tolsma2017: 211–212; Bolet, Reference Bolet2020: 827), whether immigrant presence co-occurs with economic deprivation (Bolet, Reference Bolet2020), whether there is positive interaction between ethnic groups (Allport, Reference Allport1954: 264–267; Pettigrew and Tropp, Reference Pettigrew and Tropp2006: 751–752), and whether there is a sudden increase in immigration as opposed to a long history of immigrant presence (Kaufmann, Reference Kaufmann2017).
Economic performance, demographic decline, and immigration (co-)vary systematically across Germany and correlate with other structural factors. Rural regions are often hit harder by demographic and economic decline, whereas urban areas tend to have younger and more ethnically diverse populations and are also better off in economic terms (BMI, 2020: 30–31). However, a more fundamental difference in PRR sentiment between rural and urban regions is posited by many scholars (Cramer, Reference Cramer2016; Maxwell, Reference Maxwell2019; Harteveld et al., Reference Harteveld, van der Brug, Lange and van der Meer2021; Huijsmans et al., Reference Huijsmans, Harteveld, van der Brug and Lancee2021). They argue that even when controlling for deprivation and immigration, there is still an effect of rurality, some kind of ‘cultural deprivation’ which flows from a narrative that equates (big) cities with innovation and modernity while rural areas are portrayed as backward and parochial (Förtner et al., Reference Förtner, Belina and Naumann2021). It is plausible that such cultural deprivation, in tandem with a backlash against this very narrative, could lead to PRR sentiment. Moreover, the population in urban regions is more exposed to diversity and has more heterogeneous personal networks, making it more tolerant against outsiders (Huijsmans et al., Reference Huijsmans, Harteveld, van der Brug and Lancee2021: 3).
Residing in one of Germany’s macro regions (the former FRG/GDR) is a final place-specific living condition that we consider, because east and west are known to differ across a host of social and political attitudes. While their post-war history is specifically German, the existence of such regional disparities is not in any way unusual. Similar attitudinal differences exist between Scotland and England, between Flanders and Wallonia, or even between the north and south of Italy.
While the literature often links East-West differences within Germany to the legacy of the GDR, Germany has now been reunified for almost as long as the GDR existed. Other factors should therefore be taken into account: the anxieties generated by the rapid social and economic transformation of the 1990s, ongoing economic problems, identity conflicts, and the experience of living on the periphery of the country, where almost a quarter of the original population has moved to the western states since 1990 (Mau, Reference Mau2019). As we control for other place-specific living conditions, including objective deprivation, as well as for socio-demographic composition (see next section), we interpret any remaining East-West differences as the result of the specific politics of the eastern Länder since 1990 and thus as structural.
Socio-demographic composition
The socio-demographic composition of a place is linked to PRR sentiment in two ways. First, it may be considered as a part of the place-specific living conditions mentioned above. By providing social cues and opportunities for interaction and social comparison, the demographic situation of a place shapes citizens’ views of politics and society and is therefore a contextual factor in its own right (Alba and Foner, Reference Alba and Foner2017). Second, socio-demographic variables may have a more direct effect. They act as proxies for individuals’ socialization, interests, and resources, and are hence excellent predictors of PRR affinity. Therefore, high levels of PRR sentiment in a given place may simply result from an over-representation of certain socio-demographic groups in the local populace. These groups are well-known, as are the mechanisms linking group membership to holding PRR attitudes.
Low levels of formal education is the most prominent determinant for PRR sentiment and is strongly linked to nativism (Maxwell, Reference Maxwell2019), authoritarianism (Aichholzer and Zandonella, Reference Aichholzer and Zandonella2016: 188), and populism (Spruyt et al., Reference Spruyt, Keppens and van Droogenbroeck2016: 342–343). Manual/working-class occupations are another important predictor (Oesch and Rennwald, Reference Oesch and Rennwald2018). More than other occupational groups, production workers are threatened by digitalization and globalization and fear labour market competition from immigrants (Bolet, Reference Bolet2020). Hence, both education and occupation indicate how vulnerable a person is to social changes and how well that person may adapt to these changes (Bornschier, Reference Bornschier and Rydgren2018). Besides this chiefly economic effect, education and the organizational logic of one’s work also have an impact on one’s values that further contributes to the link between socio-demographics and PRR sentiment (Kitschelt, Reference Kitschelt and Rydgren2013).
Individual economic hardship may also be conducive to PRR affinity. Being unemployed, in particular, is linked to economic deprivation and leads to a loss of social status and opportunities, making it an important determinant of PRR sentiment (Vlandas and Halikiopoulou, Reference Vlandas and Halikiopoulou2019).
Push and pull factors have brought about an unequal distribution of socio-demographic groups across German regions: in poorer areas, a disproportionate number of pupils leave school without qualifications, fewer young people go to university (Fina et al., Reference Fina, Osterhage, Rönsch, Rusche, Siedentop, Zimmer-Hegmann and Danielzyk2019: 30–31), and those who do, will often move to more prosperous regions. For class and employment status, spatial disparities are even bigger (Fina et al., Reference Fina, Osterhage, Rönsch, Rusche, Siedentop, Zimmer-Hegmann and Danielzyk2019: 18–22; BMI, 2020: 40–44, 48–51).
Gender and age/cohort are two further socio-demographic characteristics that represent differences in socialization, resources, and lived experience. With respect to their effect on PRR attitudes, the evidence is somewhat mixed. Although men are substantially more likely to vote for PRR parties, attitudinal differences between men and women are often smaller (Harteveld et al., Reference Harteveld, van der Brug, Dahlberg and Kokkonen2015). For age groups, effects on behaviour are inconsistent (Stockemer et al., Reference Stockemer, Lentz and Mayer2018: 576–577) while attitudinal differences across age groups are closely related to differences in education and class.
Rural areas in Germany (especially in the east) have older populations (BMI, 2020: 28–29) and show a pronounced out-migration of younger women (Mau, Reference Mau2019: 195). Controlling for age and gender is therefore important in itself, but also helps to disentangle contextual and compositional spatial effects.
In short, the unequal distribution of its individual-level determinants can bring about spatial patterns in PRR support. This constitutes a third aspect of place.
Unique features of a given place and the role of historical events
In their search for deeper and broader explanations of right-wing mobilization, scholars have recently (re)discovered the culture and history of a particular place as a potential factor. In a sense, this argument is almost trivial. All the aspects of place that we have discussed so far – the contemporary local living conditions, the current socio-demographic composition of the local community, and even current place-related attitudes – are at least partly shaped by past events. Inevitably, history in general has effects on the present that are mediated through these variables. Once the measurable consequences of ‘history’ – composition, living conditions and place-related attitudes – are controlled for, its explanatory power appears limited.
However, a more subtle variation of the argument points to local or regional political traditions, values, and norms that are transmitted vertically from generation to generation and horizontally through social networks and local organizations (Schulte-Cloos, Reference Schulte-Cloos2019: 70; Veugelers, Reference Veugelers and Ashe2020). Importantly, these effects are analytically distinct from and may operate independently of contemporary living conditions.
Scholars have sometimes been able to find operationalizations for such legacies so that they can be treated as quantifiable properties and hence fall under the rubric of place-specific living conditions. Recent examples in the literature on the German case range all the way from pogroms in the 13th century (Voigtländer and Voth, Reference Voigtländer and Voth2012) to the electoral politics of the early 2000s (Schulte-Cloos, Reference Schulte-Cloos2022), via the traumatic events of the 19th and 20th centuries (Hoerner et al., Reference Hoerner, Jaax and Rodon2019; Haffert, Reference Haffert2022). Although such analyses are intriguing, the number of potential historical causes for a place’s current political makeup quickly becomes unmanageable as one goes further back in time, whereas their empirical effects are often weak in practice.
While case studies can aim to uncover what exact historical and proximate local factors might strengthen or weaken the PRR in a given place (Veugelers, Reference Veugelers and Ashe2020: 310), we argue that in a large-N setup, it makes more sense to bundle all very distant and truly local historical and cultural factors together and treat them as a place-specific residuum, i.e., as unmeasured and unique features of a place that may make it an ‘outlier’. Veugelers’ (Reference Veugelers and Ashe2020: 311) use of the language of regression is particularly illuminating in this respect: in our empirical application, we show that unique features can be efficiently modelled by including random effects. If these are large, they could be used to identify places where a case-based, qualitative investigation into the local history and political culture might be fruitful.
Four aspects of place: interrelationships
While we believe that the four aspects outlined so far provide a useful heuristic for thinking about the role of ‘place’ and investigating its effects, we do not suggest that they exist independently of each other or should be studied in isolation. In the previous sections, we have already alluded to some possible links between the different aspects of place. Figure 1 serves to make these ideas explicit and to present them in a coherent way.
Figure 1. Four aspects of place and their interrelations.
Put simply, we assume that a place’s current socio-demographic composition, living conditions, and other unique factors can directly affect the PRR attitudes of its denizens. Moreover, such structural conditions may also have an impact on place-related attitudes, with the link between deprivation and place resentment being particularly prominent in the literature (e.g., Cramer, Reference Cramer2016; de Lange et al., Reference de Lange, van der Brug and Harteveld2022; Huijsmans, Reference Huijsmans2023a). Therefore, they may have an additional indirect effect on PRR attitudes that is mediated through place-related attitudes.
Finally, the effects of place resentment and localism may be mutually reinforcing. Those who feel more attached to their locality should be more prone to express place resentment (Munis, Reference Munis2020: 3; de Lange et al., Reference de Lange, van der Brug and Harteveld2022: 9–10) and may give more weight to these perceptions. Similarly, place resentment may be more salient and hence more strongly linked to PRR attitudes in those who feel more closely attached to their surroundings.
The following hypotheses summarize these considerations:
PRR attitudes are affected by …
Hypothesis 1a) individual place-related attitudes.
Hypothesis 1b) place-specific living conditions.
Hypothesis 1c) a place’s socio-demographic composition.
Hypothesis 1d) unique features of a given place.
Place-related attitudes (partially) mediate the effects of …
Hypothesis 2a) place-specific living conditions.
Hypothesis 2b) a place’s socio-demographic composition.
Hypothesis 2c) unique features of a given place.
Localism and place resentment interact, i.e., …
Hypothesis 3a) localism strengthens the effect of place resentment.
Hypothesis 3b) place resentment strengthens the effect of localism.
Data and methods
Individual-level data were collected under the auspices of the ScoRE project through a specifically commissioned CAWI survey on PRR attitudes in Germany administered three months before the 2017 federal election. To ensure national representativeness, especially in less populated regions, a large (n = 25,479) quota sample stratified by gender, age, education, and federal state was drawn from a very large (n > 130,000) online access panel, all members of which were recruited offline.
Our dependent variables are individual levels of nativism, right-wing authoritarianism, and populism. The survey included indicators for two sub-dimensions of nativism (three items tapping into perceptions of cultural threat and two items targeting Islamophobia), for populist attitudes (four items), and for authoritarianism (submission and aggression – two items).Footnote 3 For each construct, we built an index by averaging over the non-missing values for the respective items, reversing the scales where necessary so that high index values correspond to more nativist/populist/authoritarian attitudes. All attitudinal items have a range of 1–7. Cronbach’s α varies from 0.62 (perceptions of cultural threat) to 0.85 (Islamophobia). Table A1 in the online appendix shows the exact wording of all items.
We model PRR attitudes as a function of the four aspects of place. To capture the impact of place-related attitudes, we include measures of localism (two items) and place resentment (three items, see Munis, Reference Munis2020). Again, we construct indices by averaging over the non-missing values. Cronbach’s α is 0.89 for localism and 0.81 for place resentment. The indices are essentially uncorrelated (r = −0.09). Because the literature suggests that localism and place resentment may interact, we also include a multiplicative term in the main models.
To model the effect of place-specific living conditions, we map each respondent to their county (Kreis) of residence. Counties correspond to NUTS 3, the lowest level at which most official data are published. As the primary providers of school, social, and emergency services, counties play an important and broadly similar role across Germany’s federal states. Because of the large size of the survey sample, the dataset covers all 401 counties that existed in 2017. The median number of respondents per county is 43, with an inter-quartile range (IQR) of 25–70.
We include several county-level variables that feature prominently in the literature: deprivation, immigration, rurality, and location in either eastern or western Germany.Footnote 4 We measure economic and general deprivation with the share of inhabitants who receive benefitsFootnote 5 and with the remaining life expectancy for males at age 60. The latter is a cross-cutting and very tangible indicator for economic, social, and health inequalities. Demographic decline is operationalized by the share of women aged 25–30, a group whose absence has the strongest effect on community life (Salomo, Reference Salomo2019: 106). Immigration is measured by the share of foreigners.Footnote 6
For rurality, we rely on a typology developed by the Federal Institute for Research on Building, Urban Affairs and Spatial Development (BBSR). The BBSR distinguishes between large cities that are themselves counties (our reference category), other urban or suburban counties, rural counties with some higher-density settlements, and truly rural counties. To capture persistent differences between Germany’s two macro regions, we also include a dummy variable for counties in the eastern states including Berlin.
To control for differences in the socio-demographic composition, we include well-known predictors of PRR attitudes: age, gender, education, (working) class, and occupational status (see Table A2 in the online appendix for descriptive information on all variables).
In the discussion above, we have identified a group of historical and cultural peculiarities that are truly local and unique to a place and may potentially exert an influence on PRR attitudes even after controlling for the other aspects. A simple modelling strategy that accounts for each county’s unique features, and is agnostic as to their specifics, is to estimate a random effect at the county-level and conduct a post-hoc test for spatial autocorrelation (see below).
There is, however, one complication that needs to be addressed. With a median size of 800 km2 and a median population of 154,000, many German counties are rather large and populous. To account for their internal heterogeneity, we also map each respondent to a smaller area, a locality, within their county and include an additional random effect at this level.
Ideally, this mapping would be based on exact geo-coordinates. However, to protect our respondents’ anonymity, only a grid cell reference was supplied for each case. The size of each cell was chosen so that it contains at least six respondents, resulting in a median cell size of 4 × 4 km with an IQR of 2 × 2–9 × 9 km. Therefore, some cells are still relatively large and do not capture respondents’ local surroundings well. For these cells, we use additional informationFootnote 7 that varies within cells to identify groups of respondents living in relative proximity of each other, although their exact co-ordinates remain protected. By doing this, we reconstruct 8737 unique localities that represent the respondents’ immediate environments. The median number of localities per county is 17, and the median number of respondents per locality is 2.
As respondents are nested within localities which are in turn nested within counties, we estimate a series of linear three-level models whose random effects represent unique features affecting the opinion climate in counties and localities. These models already account for simple patterns of spatial heterogeneity and dependency below the county level. It is, however, possible that additional spatial dependencies exist at the county level, resulting in clusters of neighbouring counties exhibiting particularly high or low levels of PRR sentiment even after contextual and individual-level variables are controlled for. Such clusters could result either from the spillover of social, political, and economic shocks from one county to another, or from spatial clustering of omitted individual and contextual variables, including the opinion climate and historical-cultural particularities of a wider region.
Dealing with such potential spatial dependencies is not straightforward. For aggregate and other areal data, there is a wealth of single-level spatial models that can account for both types of dependencies (see, e.g., Darmofal, Reference Darmofal2015). But the development of hybrid multi-level/spatial models is still in its early stages, and some of their complexities are not yet well understood (Wolf et al., Reference Wolf, Anselin, Arribas-Bel and Mobley2018). We therefore make use of an indirect approach and test post-hoc for any spatial autocorrelation of random county-level effects that remains after estimating a classical multi-level model. While this may underestimate the autocorrelation and overstate the precision of the estimates for county-level variables (see Wolf et al., Reference Wolf, Anselin, Arribas-Bel and Mobley2018: 23–26), simulation studies suggest that the spatial structure of random effects (if it exists) emerges ‘regardless of whether or not this structure is formally stated’ (Wolf et al., Reference Wolf, Anselin, Arribas-Bel and Mobley2018: 23), and that classical and hybrid models elicit broadly similar results (Wolf et al., Reference Wolf, Anselin, Arribas-Bel and Mobley2018: 19).
Findings
To examine whether meaningful attitudinal differences exist across Germany, we start by mapping the county averages of the four PRR variables. We also map the county averages of place-related attitudes, as an apparent spatial pattern in the former could be a result of spatial clustering in the latter.
While all six variables have considerable ranges, some counties have low numbers of respondents, which makes extreme values more likely. The IQRs and standard deviations suggest only moderate variation for most variables, with the clear exception of place resentment (see Table A3 in the appendix). Nonetheless, the maps (see Figure A1 in the appendix) suggest regional clusters where PRR attitudes are either notably widespread or uncommon.
This impression is supported by Moran’s R, a measure for spatial association, which indicates significantFootnote 8 positive autocorrelations for all variables, meaning that counties with high (low) values tend to be surrounded by counties whose values are also high (low). Spatial autocorrelations should not be interpreted as Pearson correlations because their value depends on the neighbourhood structure of the underlying map, rendering comparisons across geographies invalid.Footnote 9 It is, however, possible to compare autocorrelations of different variables on the same map. With an R of 0.486, place resentment displays by far the highest degree of clustering. For the other attitudes, Moran’s R is markedly lower with values between 0.119 (localism) and 0.188 (cultural threat perceptions).
The maps themselves show some interesting disparities, and the territory of the former GDR seems to stand out in many of them. But such patterns may be misleading, and the wealth of information contained in the maps can be distracting. Therefore, we also plot local cluster maps, which are a valuable tool for quickly identifying larger patterns. Following Anselin (Reference Anselin1995), the maps are constructed in two steps. First, a local indicator of spatial association (here: local Moran’s R) is calculated for each county, and all counties without significant local autocorrelation are filtered out. Then, the remaining counties are divided up into four groups. Counties with values above (below) the mean surrounded by counties below (above) the mean form the ‘high – low’ and ‘low – high’ groups. They are spatial outliers that are atypical for their environment. The ‘high – high’ and ‘low – low’ groups, sometimes called ‘hotspots’ and ‘coldspots’ in the literature, are usually more interesting. Hotspots consist of counties with above-average values whose neighbours also have above-average values. For coldspots it is the other way around. Importantly, hotspots/coldspots show only the cores of larger clusters, as most of their neighbours will have above (below) average values, too (Fig. 2).
Figure 2. Cluster maps of PRR attitudes, localism, and place resentment (county level).
Large cores of nativism stand out in Saxony and Thuringia in the south of the former GDR, and in eastern Bavaria (cultural threat). There are also cores of populism and authoritarian submission/aggression in the south-west of the former GDR (Thuringia and Saxony-Anhalt), and two populism coldspots in the north (Hamburg) and north-west (Münsterland and Emsland regions).
Localism is common across Germany, but there is a core of high-localism counties in Thuringia and a core of low-localism counties in the rural west. Finally, place resentment is particularly concentrated in the former GDR, with the notable exception of Berlin and the surrounding counties, but also in the Ruhr district, a large agglomeration of (former) industrial cities in the west that have been struggling to adapt to structural change since the 1970s. Conversely, there is a huge southern cluster of counties with very low levels of place resentment. It is bounded by the cities of Karlsruhe and Freiburg in the west, Munich and its surroundings in the south-east, and Nuremberg in the north, and complemented by a smaller cluster in the Rhein-Main area around Frankfurt. These are all growing, prosperous areas dominated by modern industries.
Having established that there are indeed some significant spatial disparities in PRR sentiments and place-related attitudes, we next estimate a series of random-intercept-only models to see how attitudes vary over individuals, localities, and counties. For localism and the PRR attitudes, between 97 and 99% of the variation occurs at the person-level (see Table 1). The variation across localities and counties is much smaller, which suggests that contextual variables and unique effects of place play only a limited role compared to micro-level variables including place-related attitudes. This is in line with previous research (Gallego et al., Reference Gallego, Buscha, Sturgis and Oberski2016; Maxwell, Reference Maxwell2019).
Table 1. Partitioning of variance in random-intercept-only (empty) models
The picture is very different, however, for place resentment. A substantial share of almost 8% of the variation in place resentment occurs at the county level, and there is also somewhat more variation across localities within counties.
Because the place-related attitudes may act as mediators, we regress them on place-specific living conditions and on the socio-demographics in an intermediate step. We also include random effects that reflect unique factors at the county and locality levels.
As all dependent variables are measured on the same scale and because coefficients are unstandardized, the interpretation is straightforward. As Table 2 shows, demographic decline is associated with more resentment and less localism, although one must consider the limited variation (IQR 0.5) of the independent variable. Low life expectancy for older males (again with a small IQR of 1.1) is also closely associated with place resentment, but not with localism. This mirrors findings by de Lange et al. (Reference de Lange, van der Brug and Harteveld2022) and Huijsmans (Reference Huijsmans2023a) for the Netherlands. Moreover, even after controlling for other contextual factors, both resentment and localism are considerably (0.2 points) higher on average in the eastern states. Conversely, immigration, benefit dependency, and even rurality have no or very weak effects. Again, this is in line with research that points out the importance of (cultural) peripherality (e.g., de Lange et al., Reference de Lange, van der Brug and Harteveld2022).
Table 2. Multi-level of place-related attitudes in Germany
Notes: *P < .05, **P < .01, ***P < .001.
The effects of some socio-demographics are even stronger. Being male, having a technical or manual occupation, and, above all, being unemployed/sick are associated with higher levels of place resentment, whereas higher levels of formal education have a sizable effect in the opposite direction. Localism generally increases with age but is slightly lower in males, workers, and people with higher levels of education. The strong negative effect for the unemployed is unexpected and may reflect the necessity to find work elsewhere.
Finally, the residual variance at both the county and the locality levels is very small. Taken together, these findings suggest that place-specific living conditions and the socio-demographic composition of a place may have additional indirect effects on PRR attitudes that are mediated by place-related attitudes.
Before we return to this point, we will discuss the full models for the PRR attitudes (see Table 3).
Table 3. Multi-level models of PRR attitudes in Germany
Notes: *P < .05, **P < .01, ***P < .001.
To aid the interpretation of the joint impact of place-related attitudes, we predict values for the four dependent variables from the fixed part of the model, conditional on levels of localism and place resentment. On the x-axis, place resentment is varied from its lower to its upper decile. The three lines represent the lower decile, median, and upper decile of localism (Fig. 3).
Figure 3. The interactive effect of place resentment and localism on PRR attitudes.
The graphs show that place resentment has a strong positive effect on all PRR attitudes that is virtually constant across all levels of localism. Conversely, localism has no significant effects on populism and perceptions of cultural threat. Its effect on Islamophobia is weak and conditional on low levels of place resentment. Its effect on authoritarianism is somewhat stronger, particularly when place resentment is low.
In sum, Hypothesis 1a) is partly confirmed with localism exerting no or weak effects on PRR sentiments but place resentment strongly affecting all PRR attitudes. Conversely, hypotheses Hypothesis 3a) and Hypothesis 3b) are rejected, as the strong effects of place resentment are not moderated by localism and the direction of localism’s moderation by place resentment runs in the opposite direction to expectations, possibly pointing at a ceiling effect.
Hypothesis 1b) about the effects of place-specific living conditions is also only partly confirmed. Differences between cities and other county types are small and, apart from a single exception, statistically insignificant, so rurality per se does not seem to play a role. Migration has a consistently positive impact on PRR attitudes, but in substantive terms, the effect is rather weak: moving a county from the lower (6.9) to the upper (13.1) quartile shifts the predicted attitude by less than 0.1 scale points.Footnote 10 The share of the population receiving benefits has an unexpectedly negative but similarly weak effect. Demographic decline and male life expectancy at 60 have the expected effects on most attitudes, but again their impact is almost negligible given their respective IQRs.
Conversely, the macro region exerts a substantial effect. Everything else being equal, the expected difference between respondents living in an eastern county and their western counterparts varies between 0.1 and 0.2 scale points.
In line with Hypothesis 1c), socio-demographic variables mostly have the expected effects. Men’s cultural threat perceptions do not differ from women’s, but men are slightly more Islamophobic, authoritarian, and populist than women. The effect of formal education is much stronger: Having a secondary school leaving certificate reduces the expected value of Islamophobia, cultural threat perceptions, and authoritarianism by about half a scale point. For populism, the reduction is still strong at about 0.3 points. To allow for non-linear effects, age is measured with dummy variables for five age groups (reference category: 18–29). Ceteris paribus, populism, and Islamophobia are more prevalent in older respondents. For authoritarianism and cultural threat perceptions, there are no consistent patterns.
In line with previous research, class and employment are important, too: a technical or manual occupation raises the expected value on most attitudinal scales by about 0.2 points. Being unemployed or unable to work has no effect on nativism and authoritarianism, but has a substantial impact on populism. Collectively, the findings suggest that differences between counties in their socio-demographic composition are a crucial aspect of place.
Turning to the unique features of a given place, the remaining random variances at the locality and county level are very small. For completeness’ sake, we nonetheless estimate their impact by calculating the best linear unbiased predictors (BLUPs) for each county and each attitude. For Islamophobia, populism, and authoritarianism, all BLUPs have absolute values <10–8, so we refrain from any further analyses. For cultural threat perceptions, their IQR is 0.01 with maximum values of ±0.05. The spatial autocorrelation of the BLUPs is 0.068, a fraction of the value for the raw county averages.Footnote 11 That suggests that effects that are truly unique to a given place play no relevant role for PRR sentiment in Germany once the other aspects of place are accounted for. Therefore, Hypothesis 1d) is rejected.
In sum, the four aspects of place do not contribute equally to an understanding of the spatial pattern of PRR attitudes in Germany. Instead, the geography of PRR attitudes is primarily shaped by differences in a place’s socio-demographic makeup and by place resentment. Contextual factors play a role, too, with the macro region being more important than immigration and deprivation.
However, this only sums up the direct effects of the structural aspects of place. To look into indirect effects that are mediated by localism and place resentment, we re-estimate the models, leaving out the place-related attitudes.
The results are virtually identical. Coefficients differ by 0.05 or less, with the single exception of education, where the effects estimated for the reduced models are about 0.09 points stronger. This, together with the estimates in Table 2, suggests that contrary to Hypothesis 2a)–Hypothesis 2c), the structural aspects of place have no substantively relevant indirect effects on PRR attitudes. To conserve space, we therefore refrain from conducting a full multi-level mediation analysis and instead present the reduced models in the appendix (Table A5).
Conclusion and outlook
The links between the characteristics of a given place and the affinity of its inhabitants to the PRR have recently become a point of major political and scientific interest. However, ‘place’ is a multi-faceted phenomenon that can potentially affect PRR sentiment in multiple ways. Our aim in this contribution was twofold: to gain a better understanding of how researchers apply the notion of ‘place’ when they study the geography of the PRR, and to gauge the relative importance of these different aspects of place.
Taking stock of extant research, we first identified four different aspects of ‘place’ that undergird much of the current literature: place-related attitudes, place-specific living conditions, socio-demographic composition, and unique features. This perspective provides analytical leverage and allows us to go beyond many existing studies: instead of looking at one or two aspects in isolation, using geo-referenced data from Germany we can assess their relative importance by jointly modelling their effects. While Germany’s history of post-war division and reunification is unique, its current social, economic, and political diversity is not. Comparable territorial divisions exist in other major European states. Germany’s strong redistributive mechanisms and lack of ethno-linguistic conflicts may make the politicization of regional disparities more difficult than in these countries. Nevertheless, we find clear patterns of spatial polarization. This makes us confident that our findings are relevant beyond the German case.
Our results show, first, that PRR attitudes are unevenly distributed. They are more prevalent in the former GDR (particularly in Thuringia and Saxony) and in eastern Bavaria (only cultural threat). Conversely, parts of northern and north-western Germany stand out as ‘coldspots’, where levels of PRR sentiment are particularly low.
Second, different aspects of place contribute to this spatial pattern to varying degrees. While the effects of localism are weak and inconsistent, place resentment and the socio-demographic make-up of a place explain a substantial part of the spatial variance of all PRR attitudes. Put differently, respondents who perceive their region as excluded and/or who have socio-demographic characteristics that are associated with PRR attitudes tend to cluster in regions such as the Ruhr or the south of the former GDR, and this clustering produces a similar clustering of PRR attitudes.
Place-specific living conditions are of somewhat lesser importance. Structural variables such as demographic decline, migration, and overall deprivation have significant but relatively small effects on most attitudes. Net of these, location in the east, representing both the lingering effects of the former regime and the current conditions of cultural periphery and perceived inferiority, still has a substantial influence. Conversely, we find no evidence of an independent effect of rurality. Finally, there is little evidence that the unique features of a place, i.e., its local history and culture, contribute substantially to the spatial pattern of attitudes, once other aspects of place are controlled for.
Taken together, these findings reflect a paradox of (German) structural policy: Each year, billions of Euros are transferred to rural or otherwise disadvantaged regions, especially in the east. As a result, objective disparities are comparatively small, and the south of the former GDR in particular is doing relatively well in terms of employment and economic growth. Yet it is here where place resentment and, as a result, PRR attitudes, are more prevalent than in many other places.
More generally, these results speak to several of the debates in the literature on the role of ‘place’ for the PRR. First, the importance of socio-demographic composition and the weak effects for immigration and objective indicators of deprivation chime with the main finding in the important study by Maxwell (Reference Maxwell2019), who shows that cosmopolitan immigration attitudes in large cities result chiefly from sorting processes, not from contextual effects.
Second, the very strong effect of place resentment confirms the importance of this novel concept, originally developed for the USA (Munis, Reference Munis2020) and then applied in the Netherlands by de Lange et al. (Reference de Lange, van der Brug and Harteveld2022) and Huijsmans (Reference Huijsmans2023b), in a second European country. While Munis relates place resentment to racial resentment, Huijsmans and we find strong links to a broader range of PRR attitudes that focus on a different set of out-groups. This is an important result, because it strongly supports the notion of place resentment as a concept grounded in the classic social psychology of groups (Munis, Reference Munis2020: 3).
Third, we find that the strong effect of place resentment is mostly constant across various levels of localism, underlining its central importance for PRR attitudes, while localism itself has no relevant effects at any level of place resentment. This finding supports Fitzgerald’s (Reference Fitzgerald2018) claim that localism is a complex concept, whose components may pull voters into opposite directions and whose exact effects depend on its politicization. Future studies should therefore focus on developing a more nuanced operationalization of localism. It may also be worthwhile to take the politicization attempts of local and national actors into account.
Fourth, after we control for place-related attitudes, contemporary living conditions including former GDR status, and social-democratic composition, there is nearly no random county or locality level variance, which represents the unique features of place, left. While we appreciate the attempts to find deeper explanations for PRR sentiment, this suggests that researchers should focus their efforts on more proximate causes, including the sources and consequences of place resentment.
Of course, our analysis is subject to the same set of limitations that afflict all studies that rely on administrative data and boundaries: our findings may depend on the chosen scale and/or the spatial units used for aggregation. More specifically, one could argue that counties are relatively large units, and that smaller neighbourhoods (ideally tailored to each respondent’s home, see, e.g., Johnston et al. Reference Johnston, Propper, Burgess, Sarker, Bolster and Jones2005) would provide a more relevant frame of reference for most citizens. However, the small variances of the random effects at the locality level give us some confidence that we are not missing too much small-scale variation.
Beyond the scholarly debate on the PRR, our results also have some implications for policymakers: the socio-demographic composition of a place is an essential driver of regional PRR attitudes. While this composition is difficult to alter by political means in the short term, this finding highlights the need for long-term demographic policies and structural planning. Moreover, our results imply that a lack of positive recognition is linked to PRR affinity. While we already knew this for the individual level, our paper shows that this applies equally to local communities. Hence, politicians should consider strengthening ‘places that don’t matter’, but not just in economic terms. The representation of such places that are far away from metropolises, university towns, and other cultural centres, and the attention that is paid to them might be just as relevant. The notion of place therefore needs a differentiated understanding, as place does matter – especially in places that do not seem to matter.
Supplementary material
The supplementary material for this article can be found at https://doi.org/10.1017/S1755773923000279.
Data availability statement
Replication code and data are available from the corresponding author’s dataverse: https://doi.org/10.7910/DVN/VZTWPI
Acknowledgements
We gratefully acknowledge the financial support for our research that we received from the German Science Foundation (DFG). We would also like to thank the editors and anonymous reviewers, Carl Berning and our other colleagues in the SCoRE project, and Daniel Ziblatt and the other participants of a colloquium held in 2022 at the WZB Berlin Science Center.
Competing interests
Both authors have no conflicting interests to report.
Open access
