The urban-rural polarisation of political disenchantment: An investigation of social and political attitudes in 30 European countries

Despite the prevalent focus upon increasing political divisions between urban and rural Europe, relatively little research has explored whether there is a systemic urban-rural divide in the political and socioeconomic attitudes of citizens across the entire continent. This paper aims to fill this gap. Drawing on individual-level data from the European Social Survey, it explores potential linkages between place of residence and individual attitudes. Our results show that there are strong, and statistically significant, differences between the populations in these different settings. On average, rural dwellers show stronger levels of dissatisfaction with democracy and lower trust in the political system. Yet, while we uncover stark differences in attitudes towards migration and globalisation, we do not find significant variation on some social and economic issues traditionally at the core of left-right cleavages. And our analysis suggests that this spatial divide does not operate in a binary fashion. It is more of a continuum, running on a gradient from inner cities to metropolitan suburbs, towns, and the countryside. The differences are explained by both composition and contextual effects, and underscore the importance of moving beyond ‘standard’ trade-offs between so-called ‘people-based’ versus ‘place-based’ policy approaches to territorial inequality.


Introduction
While, for much of the Twentieth Century, political scientists tended to assume that political cleavages in western democracies revolved around differences in class position and attitudes towards distributional questions and the role of the state, in recent decades there has been a growing emphasis on those associated with various kinds of group identity, and, latterly, with the importance of place (Glaeser and Ward 2006;Kenny 2014;Kriesi 2010). In the US, a large body of work has documented how political differences are increasingly driven by a distinctive -and deepeninggeographical cleavage, with almost all large cities being Democratic strongholds and rural counties being a cornerstone for the Republicans (Gimpel and Hui 2015;McKee 2008;Monnat and Brown 2017;Morrill, Knopp, and Brown 2007;Scala and Johnson 2017;Tam Cho, Gimpel, and Hui 2013).
Across Europe too, notable recent political events, such as the UK 2016 Brexit vote, and the 2018 Gilet Jaunes protests in France have shed light on marked political divergences between urban and rural places. Yet, despite growing evidence from individual countries such as the UK (Becker, Fetzer, and Novy 2017;Carreras, Irepoglu Carreras, and Bowler 2019; Garretsen et al. 2018;Goodwin and Heath 2016;Jennings and Stoker 2016;Lee, Morris, and Kemeny 2018;MacLeod and Jones 2018), France (Eribon 2013;Guilluy 2016;Ivaldi and Gombin 2015) and Italy (Agnew andShin 2017, 2020;Rossi 2018), relatively little research has explored in a systematic way whether the growing political divide between urban and rural areas mapped in some countries is apparent across the whole continent. Indeed, given the prevalent focus upon Anglo-American comparisons within the Anglophone literature, there is a pressing need to consider whether the same phenomena are discernible across Europe.
The current paper aims to fill this gap. Drawing on individual-level data from the European Social Survey (ESS), it provides systematic comparative evidence across 30 European countries over the period 2002-2018. We explore links between place of residence and attitudes on a range of different socioeconomic and political issues. To anticipate our conclusion, we find that there is a strong and significant divide between the political outlooks of urban and rural Europe. But this divergence is not best seen in binary terms, and is better understood as a gradient running from inner cities to metropolitan suburbs, towns, and the countryside (as anticipated by Scala and Johnson, 2017). We show how, compared to dwellers in inner urban cores, respondents living in suburbs, towns and rural areas are more likely to be conservative in their orientation, dissatisfied with the functioning of democracy in their country, and less likely to trust the political system, even though they are -strikingly --more likely to participate in it, especially by voting. However, while our analysis highlights some stark geographical variances in attitudes towards migration and globalisation, we do not find significant variation on issues which have traditionally been at the core of leftright cleavages, such as support for welfare state redistribution. And, finally, we show that these differences are explained by compositional and contextual effects.
This article contributes to the growing literatures in the neighbouring disciplinary fields of geography and political science devoted to the spatial dimensions of political disenchantment in three distinctive ways. First, in line with a growing emphasis in much social scientific research on a widening 'cosmopolitan-nationalist' polarization, we show how differences associated with the urban/rural continuum are significant across a range of attitudinal dimensions. Most recent studies of the 'geography of discontent' (McCann 2019; Rodríguez-Pose 2018) consider the evidence supplied by voting patterns in elections and referendums. Yet, such events may well be linked to candidate-related factors and contingent political developments, and may relate only indirectly to underlying shifts in popular attitudes (Abrams and Fiorina 2012;Ford and Jennings 2019). In fact, many political scientists suggest that electoral campaigns do not change public opinion that much but, rather, 'activate' some considerations over others (Mutz 2018), increasing the extent to which particular issues matter for voters when they choose a candidate. It is, therefore, important to understand in more depth the factors explaining the underlying dynamics of public opinion. We seek to address this challenge using attitudinal data, and our findings suggest that the linkage between place of residence and political attitudes is not confined to attitudes towards iconic and contentious questions such as the EU (Dijkstra, Poelman, and Rodríguez-Pose 2019) or migration (Maxwell 2019). Instead, we conclude, it encompasses a wider range of political and socioeconomic issues including perceptions of political behaviour and trust in political institutions. There are intimations in these results, we suggest, of distinct worldviews that are manifested by spatially segmented parts of the population of many developed democracies.
Second, in line with the work of Scala & Johnson (2017) on the US, we show that the geographical divide in European political attitudes should not be thought of as a dichotomy between urban and rural places -as suggested for instance by Cramer (2016) in relation to the US -but conceived instead as a gradient. This finding is in line with some recent analyses which underline the importance of inequalities within urban places (Baum-Snow, Freedman, and Pavan 2018) and suggest that, as a result, residential segregation between inner urban areas and suburbs is on the rise across many European cities (Musterd et al. 2017). We point to the merits of applying a more finely grained geographical lens to the contemporary urban-rural divide in Europe, arguing in particular for more nuanced, place-sensitive typologies.
Third, we provide a preliminary analysis of the factors that may be most significant in explaining the differences along the urban/rural continuum. In his analysis of attitudes to immigration, Maxwell (2019) argues that differences between urban and rural areas are more strongly driven by sociodemographic characteristics -that is, by compositional effects and the geographical sorting of people with different attributes and outlooks -than by the influence of place of residence over individual outlooks. In contrast, we underline how attitudes vary across sociodemographic and geographical dimensions. We show that while people's attitudes are heavily stratified by key individual observable characteristics, such as age, education, and occupational status, there is a non-negligible correlation between places and their inhabitants' attitudes towards various political and social issues (Agnew 1987). This conclusion has important implications for the ongoing debate about what kinds of policy solutions are best equipped to address growing territorial inequalities, and whether these should be place-sensitive or not (cf. Barca, McCann, & Rodríguez-Pose, 2012;Iammarino, Rodriguez-Pose, & Storper, 2019).
Further research, we suggest, is needed to disentangle potential compositional effects based on unobservable -rather than observable -characteristics such as intrinsic 'cognitive underpinnings' (cf., for example, Lee et al., 2018;Zmigrod, Rentfrow, & Robbins, 2018). And analysts should also explore in more depth the mechanisms through which the kinds of setting in which people live can influence the development of individual traits, especially during early-age socialisation (Bosquet and Overman 2019;Hackman and Farah 2009;Rentfrow, Gosling, and Potter 2008). Yet, irrespective of whether divides in attitudes are driven by compositional effects or by the contextual influence of places on people, marked demographic sorting and the overlapping of territorial and attitudinal cleavages signal a deepening geographical fracture in European societies which, in the long term, may have significant implications for the challenges of generating social cohesion (Wilkinson 2018) and addressing the implications of rising disenchantment with democracy (Foa et al. 2020).
The paper is structured as follows. Section two reviews the existing literature on the urban-rural polarisation, and develops our main, empirical hypotheses. Section three describes the data and the analytical strategy that we have employed. We then present the key results in section four, and discuss the implications of our findings in section five. In the final section, we offer some discussion of their implications in relation to ongoing policy debates, and suggest areas where further research would be profitable.

Political polarisation along the urban-rural divide
The polarization of electorates across the urban-rural divide is by no means a new, or recent, phenomenon. At the peak of the industrial revolution, between the end of the Nineteenth Century and the beginning of the Twentieth, many European and North American countries were divided politically between the interests of rural and smalltown dwellers, engaged in agricultural production, and those of urban residents, experiencing rapid technological and socioeconomic change and a new spatial economic order dominated by manufacturing in large urban agglomerations (cf. Lipset and Rokkan 1967; Vidal de la Blache 1913).
In the second half of the Twentieth Century this stark divide faded partially, as sharper political cleavages, which reflected economic issues, class divisions and the role of the state in society, emerged (Lipset and Rokkan 1967). Yet, across many advanced economies, the processes of economic globalisation over the last three decades have generated new socio-economic divides (Ford and Jennings 2019) and contributed to the emergence of a new dimension of political conflict, cutting across these older divisions. Although the extent and nature of this divide remains contested among social scientists (Norris and Inglehart 2019), a growing number of studies show that the classic class-based Left-Right cleavage in party competition is today overlaid by a new division based on education and cultural attitudes. Scholars support this claim with reference to survey data (Tabellini and Gennaioli 2018), the positioning of political parties (Inglehart and Norris 2016), and the composition of party supporters (Piketty 2018).
Three accounts figure prominently in this debate, each proposing a distinct explanatory framework to explain this new cleavage: "materialism" as opposed to "post-materialism" (Inglehart 1977(Inglehart , 1990(Inglehart , 1997; the divide between "winners" and "losers" of globalisation (Kriesi 1998(Kriesi , 2010; and a "transnational" conflict of values between "liberals" and "conservatives, authoritarians, and/or nationalists" (Hooghe and Marks 2018;De Vries 2018). While each approach emphasizes a specific trigger, the literature overall points to the increasing salience of geography in relation to this new attitudinal cleavage, and to growing political disagreements between cosmopolitan, highly educated, and socially progressive urbanites, and nationalist and socially conservative residents of 'hinterland' areas. The determinants of the recent rise in populism and anti-establishment sentiments across many established democracies are numerous and interlinked. 1 Support for right-wing populist parties relates to a number of variables including age, ethnicity, education and employment.
Yet, these fault lines are characterised by many commentators as linked to a distinctive geographical divide, with political differences among inhabitants in urban and rural places becoming increasingly marked.
In the US, a significant amount of work has documented how electoral politics falls increasingly into distinctive spatial patterns, with almost all large cities being Democratic strongholds and rural counties being the cornerstone for the Republicans  Zaslove, 2013;Ford & Goodwin, 2014;Guiso, Herrera, Morelli, & Sonno, 2019;Inglehart & Norris, 2016;Jens Rydgren, 2010;Kriesi, 2014;Rodrik, 2018;Rooduijn, van der Brug, & de Lange, 2016;Van Gent, Jansen, & Smits, 2014;Zaslove, 2008. America has been observed for some decades (Abramowitz and Saunders 2008), 2 the emergence of a salient divide between urban and rural areas has become more palpable over time, and was particularly clear during the 2016 presidential campaign (McKee 2008;Wilkinson 2018). Analysing recent opinion polls and the latest US presidential elections, Scala and Johnson (2017) for example identify a consistent gradient of conservative sentiment from the most urban to the most rural counties on a wide range of socioeconomic issues. Their results show that differences in outlook remain statistically significant even after controlling for factors such as education, income, age, ethnicity, and religion.
Across Europe too, there are signs that many different political systems are adapting to this new cleavage, and an increasingly spatially divided electoral geography is emerging (Agnew and Shin 2020;Hooghe and Marks 2018). France is a much-cited exemplar of this trend. There is a growing political divide between the large urban centres -'globalised', 'gentrified', and increasingly inhabited by cosmopolitans and 'bobos' (bourgeois bohemians) -and the banlieues populated by immigrants of recent arrival, and the remaining medium and small-sized cities and rural areas, where longtime immigrants and the 'white' working classes experience economic decline and are increasingly disaffected with the political system (Bacqué et al. 2016;Cusin, Lefebvre, and Sigaud 2016;Eribon 2013;Foa et al. 2020;Guilluy 2016;Ivaldi and Gombin 2015).
Similarly, England has witnessed a gradual 'bifurcation' (Jennings and Stoker 2016) in political terms between people with higher education and good employment opportunities who live in metropolitan areas and those living in 'backwater' areas associated with economic decline, hostility to immigration and the EU, and a stronger sense of English identity (Garretsen et al. 2019;Kenny 2014Kenny , 2015. While there is a strong regional dimension to the geography of discontent in Britain (Garretsen et al. 2018;McCann 2016McCann , 2019Tyler et al. 2017), in the UK and elsewhere, the urban-rural fault-line has become increasingly prominent.
Some research into the causes of these spatialized political patterns suggests that the urban-rural political divide has become more intense in recent decades. Exploring a rich dataset on referendum results at the municipal level, Mantegazzi (2018)'s work on Switzerland for example provides a compelling picture of the growing geographical polarisation which, since the 1970s, characterises Swiss politics. This increasingly leads to a distinctive pattern, with voters in large cities positioning themselves in a 'Left-liberal' space as opposed to 'Right-conservative' municipalities that are increasingly clustered in rural areas.
But whilst many commentators observe this pattern in a few, paradigmatic countries, little research has explored whether there is a systemically rooted urban-rural divide in political and socioeconomic attitudes across the whole European continent. And there is still considerable disagreement about the causal dynamics and processes underpinning this division. Two broad explanatory accounts of his deepening divide, respectively focused on composition and contextual effects, are most notable within current academic literature. The second explanation focuses on the socio-economic trajectory of places and their contextual role in shaping individual attitudes. Across Europe, there is increasing economic divergence between core cities and areas that are lagging behind in economic terms (Iammarino, Rodriguez-Pose, and Storper 2019). Commentators hence point to an emerging 'geography of discontent', reflecting the unhappiness of people living in places which are stagnating or facing comparative economic decline (Garretsen et al. 2018;Los et al. 2017;McCann 2019). Rodríguez-Pose (2018, p. 201) for example shows how "[i]t has been thus the places that don't matter, not the 'people that don't matter', that have reacted". Rising opportunities and growth in thriving urban cores not only attracts younger, more educated and more liberal individuals, but also contributes to shifting urban dwellers towards more progressive social values (Friedman 2005) and cosmopolitan preferences (Vertovec and Cohen 2002).
Conversely, declining or stagnant material prospects in peripheral towns and rural places tend to generate a growing sense of disaffection, anxiety and resentment, driving citizens to adopt a more protective, "zero-sum, 'us or them' frame of mind" (Wilkinson, 2018, p. 5). Lee et al. (2018), for instance, show that in the Brexit referendum, support for Leave was notably higher for those living in the same county in which they were born than those who had left their county of birth. Importantly, the correlation between residential immobility and Leave voting was only strong in peripheral places where wages had stagnated after the 2008 crisis (Crescenzi, Luca, and Milio 2016), or in areas which had experienced steep recent increases in the number of immigrants moving to them. In times of growing economic inequality, residents of rural areas and small towns seem to have moved towards increasingly socially conservative and anti-liberal positions.
Place-dependent grievances are not confined to economic issues. The differences between urban and rural life, and feelings among rural and town dwellers that their places have been neglected by economic and political elites, have led to growing resentment based on cultural-identity issues which shape a growing sense of mutual alienation (Lichter and Ziliak 2017). As Cramer (2016) argues, what may look like disagreements over specific policy preferences can often be traced back to this more fundamental difference of worldview, which is rooted in questions about identity and contending "ideas about who gets what, who has power, what people are like, and who is to blame" (Cramer, 2016, p 5). In her (ibid.) account, 'rural consciousness' develops out as an attachment to place, and is linked to a sense that politicians and decision-makers systematically ignore peripheral places, failing to provide them with their fair share of resources, and to acknowledge their distinct lifestyles and values.
Identities rooted in peripheral places seem to have, at least partly, come to shape the political preferences of a much larger group of people.
In summary, there is considerable evidence within a wide-ranging body of literature to support the contention that there is a clear political fracture between metropolitan and rural (and semi-urban) communities. But, as yet, it remains unclear whether this pattern works similarly across the European continent. One study -Maxwell (2019)has provided a body of comparative evidence about popular attitudes towards immigration in European countries. Our analysis builds on his work, seeing to explore a broader range of socioeconomic and political issues, and testing whether differences on these are also reflective of spatial differences. Drawing on the existing literature, our first research hypothesis is the following: H.1 = There are discernible differences in the collective outlooks of people who live in urban and rural places.
But, as Scala & Johnson (2017) suggest in relation to the US case, it may be misleading to think of the urban/rural divide in dichotomous terms. 'Rural America' is an elusive and deceptive term, incorporating a great diversity of types of place (ibid.). Similarly, even within metropolitan areas, there are significant differences between urban cores and suburbs (Baum-Snow, Freedman, and Pavan 2018; Musterd et al. 2017). Drawing on such insights, our second research hypothesis is the following: H.2 = the urban/rural divide is best understood as a continuum rather than a dichotomy.
In addition, we aim to provide a preliminary exploration on the determinants of attitudinal differences across places, distinguishing between composition and place effects, based on our attitudinal data. We test whether differences in attitudes across the urban/rural continuum might not be exclusively explained by compositional effects -that is, by the sorting of younger, more educated and economically better off individuals in large urban centres. Our last two hypotheses are the following: H.3 = differences in attitudes across the urban/rural continuum are explained by compositional effects.
H.4 = differences in attitudes across the urban/rural continuum are not exclusively explained by composition effects and, hence, are linked to some of the intrinsic characteristics of places.

Data
We analyse pooled, cross-sectional individual-level data from the European Social Survey. Since its inception in 2002, the Survey has conducted, every other year, faceto-face interviews across most participating countries. We exploit data on the EU27 Member States plus the UK, Norway and Switzerland from all the nine available waves, covering the period 2002-2018. The Survey is representative of all persons aged 15 and over, regardless of their nationality or language (we exclude respondents below 18). Individuals are selected through a multi-stage random probability sampling procedure. The ESS uses sampling designs where some groups or regions have higher probabilities of selection. To reduce sampling errors and potential non-response bias, we apply country-specific ESS post-stratification weights constructed using information on age, gender, education, and region. 3 Furthermore, we also apply country population size weights to account for the fact that countries participating in the ESS have relatively similar net sample sizes (roughly between 900 and 2700 individuals per country) even if the size of their general population varies 3 Cf. https://www.europeansocialsurvey.org/methodology, accessed on 5 July 2019.
considerably (e.g. from 1.1 million residents in Estonia to 71 million in Germany during ESS wave 8). These weights ensure that each country is represented in proportion to its actual population size.

Model and Estimation strategy
To test our research hypotheses, we estimate the following equation: Where is a vector of ordinal variables measuring individual attitudes on each issue of person in the ESS wave . is our main regressor of interest, and represents a categorical variable indicating whether each respondent resides in a big city (the baseline category), in the suburbs/outskirts of a big city, in a town/small city, in a country village, or in a farm/home in the countryside. Alas, due to the nature of the data we cannot control for more fine-grained geographical determinants. While this is a potential limitation of the analysis, in our approach we follow earlier work on the urban/rural divide (Maxwell 2019) and aim to maximise the cross-country coverage offered by the ESS.
is a vector of sociodemographic controls which may affect individual attitudes.
European countries are highly unequal in many geographical (e.g. land size) and socioeconomic aspects. We hence add state fixed-effects (FE) αc, which are included to absorb any country-specific idiosyncrasies. We also add ESS wave fixed-effects dt, to account for cross-sectional common shocks throughout the years. ɛi,c,t is the error term.
We adopt robust standard errors in all regressions. It is important to bear in mind that this analysis does not claim to provide a causal interpretation of the link between place of residence and political attitudes. Instead, it seeks to present a broad and systematic analysis of a set of quantitative, stylised facts, which might well be explored in more depth, with the use of more advanced causalinference tools, in future research.

Variables and definitions
is a vector of either binary or ordinal categorical variables measuring individual attitudes on each issue of person in the ESS wave . We consider ten issues, grouped along two main dimensions: Attitudes towards the political system and political engagement. We are, first, interested in the link between place of residence and individual attitudes towards the political system, as well as the ways in which people engage with politics. We focus on attitudinal responses and views of modes of political behaviour as these are also revealing expressions of deeper-lying attitudes towards the political system. We explore, specifically, the extent to which people engage via conventional political channels, such as voting, and the extent to which respondents evince trust in political parties, since recent research has identified a close correlation between discontent with the parties and a deeper mistrust of As anticipated, represents a categorical variable indicating whether each respondent resides in the inner part of a large city (the baseline category), the suburbs/outskirts of a big city, a town/small city, a country village, or a farm/home in the countryside. Out of the total pooled sample, 19.46% of respondents report that they live in a big inner city, 12.04% in suburbs, 30.59% in towns or small cities, 31.41% in a country village, and 6.5% in a farm or isolated home in the countryside.
is a vector of individual sociodemographic controls which may affect attitudes, and for which micro-level information is available. In particular, includes the following covariates: Age. Following a pattern clearly established in the literature, we may expect attitudes to be highly stratified by age groups, with younger generations being more likely to embrace cosmopolitan and progressive views (inter alia: Goodwin & Heath, 2016;Harris & Charlton, 2016) and, at the same time, being less engaged in electoral politics, given their familiarity with social media and less conventional forms of political engagement (Foa et al. 2020). The variable is expressed in Ln.
Gender. As customary, we control for the gender of the respondent. Native. We add a dummy for people born in the country of residence, as we may expect this variable to affect our outcomes. Partner's unemployment. We also include a dummy if a respondent's partner is unemployed.
Benefits. We control for whether the main source of household income claims state benefits. We include this variable, as well as the following ones, to account for potential divides between "winners" and "losers" in the contemporary economy.
Household income feelings. We create a dummy capturing whether respondents feel that life with their present household income is difficult or very difficult.
General economic satisfaction. The variable captures the overall level of individual satisfaction towards the national economy.
Research on the "geography of discontent" has pointed to a link between political disenchantment and relative regional economic status and decline (McCann 2019; Rodríguez-Pose 2018). In the final part of the analysis we will hence also include three regional-level economic indicators which may affect individual attitudes. (While for most countries the ESS matches respondents to their NUTS2 level region, in some cases persons are matched with either NUTS1 or NUTS3 regional identifiers. See Appendix A.1 for more details.) The variables we consider are: Average regional per-capita GDP. This is included to account for the overall economic development of the region where respondents live.
Regional per-capita GDP growth. We control for changes in GDP levels over the previous four years. 5 % ratio regional per-capita GDP / national average. We add a measure of relative regional economic wealth.
Appendix A.2 reports key weighted summary statistics, while Appendix A.3 provides a detailed description for each variable.

Baseline results
This section presents the baseline results of our analysis. In Table 1, we explore the overall differences in attitudes that we find along the urban/rural continuum, when not controlling for composition effects. For each issue , the table presents the proportional odds ratios (i.e. the exponentiation of the 'raw' logit/ordinal logit coefficients) of respondents living in each of the geographical categories compared to respondents residing in large urban cores, the baseline category. In all models, we include country and year fixed-effects.
The first four columns of Table 1 report outputs for attitudes towards modes of political engagement and the political system, and illuminate the extent to which people engage via 'traditional' political channels such as voting (column one), whether they trust political parties (column two), or whether they are satisfied with democracy in their country (column three). And, as a point of comparison, we also report respondents' satisfaction with life (column four). Column five provides insights into people's self-placement on the left-right spectrum, while the last five models focus on specific socioeconomic issues. Models six and seven, respectively, focus on attitudes towards welfare state support and trust in the police. Finally, the last three models report results relating to the 'transnational cleavage', namely perceptions towards immigrants (columns eight and nine) and attitudes towards the EU (trust in the European Parliament, column ten). The results broadly confirm our prior assumptions, and provide strong evidence in support of hypothesis H.1. Across most issues covered, there are stark and statistically significant differences between urban and rural places. Besides, in line with hypothesis H.2, Table 1 shows that the divide in attitudes is a gradient linked to urban density, running on a continuum from inner cities to suburbs, towns, villages, and isolated rural houses.
Controlling for country and year idiosyncrasies, respondents living outside large inner cities are, on average, significantly more likely to vote. At the same time, however, they tend to show less trust towards the political system. For instance, the odds of somebody voting, or reporting a one-unit higher level of trust in parties (which is measured on a scale 0-10), if they live on a farm or in an isolated rural area (the last category), are, respectively, 33.5% higher and 16.2% lower than those of an average resident of a large urban core. These results suggest that, while levels of trust in the political system are lower in rural areas, in these places traditional modes of political engagement are more prevalent. 6 Our results, more generally, confirm that the residents of these places are far less likely to engage in non-conventional political behaviours, like signing petitions and boycotting products. 7 But they are also, paradoxically, more sceptical than their urban counterparts about the political system and the choices it presents them with.
In line with characterisations of a cosmopolitan/conservative divide between large urban centres and elsewhere (Guilluy 2016), the results of column five show that people living in urban fringes, towns and the countryside are significantly more likely to identify as conservatives, tending to place themselves on the right of the political spectrum. As an example, the odds of a person living in a country house feeling oneunit closer to the political right on the left-right spectrum (which is measured on a scale 0-10, where zero is left and 10 is right) is 57.2% higher. 8 Interestingly, however, we do not find any substantial difference in support for the welfare state (model six), an issue which has traditionally played an important role in left-right political cleavages in western democracies, or, indeed, in the trust they place in the police. It may well be that in the era of populism, worries about welfare and law and order are no longer a source of significant divergence between those who live in different parts of a country.
By contrast, results from the last three models provide clear evidence of a stark urban/rural divide on issues associated with the new 'transnational cleavage' (Hooghe and Marks 2018;Kriesi 2010). The results of columns eight and nine show significant 6 Levels of trust in parties are virtually identical to levels of trust in politicians. By contrast, levels of trust in the national parliament are slightly lower, consistent with the hypothesis that disenchanted rural dwellers may be more trustful of individuals or specific political parties that the political system overall. These additional results are available on request.
7 These additional results are available on request.
8 Results not presented but available on request equally show statistically significant differences on matters such as family issues and women rights.
differences in attitudes towards international immigration. As an example, compared to an inner-city dweller, the odds for a rural home resident reporting a one-unit higher level of belief in the positive role of migration in enriching the national culture (measured on a scale 0-10) are more than 55% lower. A very similar picture emerges with respect to attitudes towards the EU (column ten).
In Appendix B.1 we plot the predicted probabilities for models three and eight from In Appendices A.6 and A.7 we stratify the sample of Table 2  Nevertheless, we underscore how, even after controlling for individual observable characteristics, places still have a non-negligible correlation with people's political attitudes, especially on issues such as voting behaviours (column 1 of Table 2), leftright placement (column 5), and migration and globalization (i.e. the dimensions relating to the 'new transnational cleavage, columns 8 to 10). For instance, compared to an inner-city dweller, even after controlling for individual observables, the odds for a rural home resident reporting a one-unit higher level of satisfaction with democracy in their country (measured on a scale 0-10) are more than 10% lower, while the odds of them reporting a one-unit higher level of belief in the positive role of migration in enriching the national culture (measured on a scale 0-10) are more than 26% lower. In other words, we do not fully reject hypothesis H.4, but instead conclude that, while compositional effects are very important in shaping attitudes, they are not sufficient to explain the urban/rural divide in political views in these European countries.
To understand what may explain the link between place and individual attitudes, we finally estimate equation (1) controlling for sociodemographic observables and economic status while also adding regional economic characteristics. Before wave 4 the ESS did not report respondents' region of residence. Besides, not all individuals are matched with a regional identifier, while we do not have regional economic data for the latest ESS wave 9. We are hence able to merge regional-level characteristics to only waves 4-8 and a sub-set of respondents. (For comparison, Appendix A.8 reestimates the regressions of Table 2 on the restricted sample. With the exception of 'Satisfaction with democracy', which loses significance, results are overall similar to those from the full sample.) Results, reported in Appendix A.9, suggest that controlling for regional economic dynamics has only a minor effect on the link between place of residence and individual socio-political attitudes. Additional tests we ran equally suggest that the regional economic variables do not act as moderators, as their interactions with place of residence are statistically insignificant. 10 While some recent analyses of the "revenge of places that don't matter" (Dijkstra, Poelman, and Rodríguez-Pose 2019; Rodríguez-Pose 2018) have underlined a link between contemporary political grievances and territorial economic stagnation and decline, we conclude that differences in attitudes along the urban-rural continuum are primarily linked to cultural-identity issues, as highlighted for example by Cramer (2016). This challenge connects with the second main implication of these results. This concerns the growing political focus upon the 'left behind' -a category that is both ubiquitous and contested. Our analysis suggests that there are degrees of 'left behindness' in terms of political worldviews, and that a firmer appreciation of the geographical specificities of different rural areas, towns and cities is integral to a more contextually informed and tailored policy responses to the challenges posed by interand intra-regional inequality (Iammarino, Rodriguez-Pose, and Storper 2019;Rossi 2018). For instance, research has shown how the European Cohesion Policy has contributed to generating economic growth in rural areas close to urban agglomerations, but not in those farther away from cities (Gagliardi and Percoco 2016).

Discussion and implications for policy
Hence, 'place-sensitive' development policies require a deeper and more contextual appreciation of the different patterns of economic development apparent in different places (Dijkstra, Garcilazo, and McCann 2013;Iammarino, Rodriguez-Pose, and Storper 2019). And, following recent debates about how best to tackle inter-territorial inequalities, our findings underscore the importance of moving beyond 'standard' trade-offs between so-called 'people-based' and 'place-based' policy approaches (Barca, McCann, and Rodríguez-Pose 2012;Kline and Moretti 2014) and support the idea of integrated policy frameworks that bring together individual (micro) and territorial (meso) logics (Iammarino, Rodriguez-Pose, and Storper 2019).
Finally, this paper offers a challenge to the ingrained tendency in the worlds of Anglophone scholarship and media punditry to foreground the UK and the US as the leading exemplars of the political disruption associated with rapid techno-economic change in the contemporary period. Our results lead instead to the conclusion that there are important trends and dynamics at work across nearly all developed European economies that are exercising a powerful impact upon the public attitudes and social orientations of people resident in these different contexts. Of course, there are still key differences of political economy, history and institutional structure at work in these different countries and regions, and these need to be given due analytical consideration. But the common patterning of these results, and the close linkages they suggest between the size and kinds of community in which many people live, and how they feel about their country and its politics, lead us to conclude that the Anglo-American bias inherent in much discourse on the 'left behind' needs to be challenged.
Understood as a wider phenomenon, we are much more likely to grasp the underlying economic and cultural dynamics that are driving and perpetuating these spatially embedded patters of political disillusion.

Conclusion
Drawing on individual-level data from the European Social Survey (ESS), this paper explores linkages between place of residence and attitudes on a range of different socioeconomic and political issues, providing systematic comparative evidence across 30 European countries over the period 2002-2018. We find both that there is a strong and significant divide between urban and rural Europe, and that this is not best seen in binary terms but as a gradient running from inner cities to metropolitan suburbs, towns, and the countryside. We also show that these differences are best explained by a mixture of composition and contextual effects, rather than through an emphasis on either of these kinds of factor alone.
Our analysis shows how the link between place of residence and political attitudes is not confined to attitudes towards iconic and contentious questions such as the EU (Dijkstra, Poelman, and Rodríguez-Pose 2019) or migration (Maxwell 2019). Instead, we conclude, it encompasses a wider range of political and socioeconomic issues. In particular, we show how, compared to dwellers in inner urban cores, respondents living in suburbs, towns and rural areas are systematically more likely to identify as right-wing, dissatisfied with the functioning of democracy in their country, and less likely to trust the political system. But they still tend to be more likely to vote, a finding which has an important bearing on current debates about the future of democratic politics (Runciman 2018 policy approaches (Barca, McCann, and Rodríguez-Pose 2012;Kline and Moretti 2014) and, by contrast, point to the need to develop integrated frameworks combining both individual (micro) and territorial (meso) logics (Iammarino, Rodriguez-Pose, and Storper 2019).
The paper does not claim to provide a causal interpretation of the link between place of residence and political attitudes, but, instead, seeks to present a broad and systematic analysis of a set of quantitative, stylised facts. Each of these could be explored in more depth, and with the use of more advanced causal-inference tools.   Discrete, 0/1 "Have you ever been unemployed and seeking work for a period of more than three months?" Partner's unemployment Discrete, 0/1 "Which of the descriptions on this card applies to what he/she has been doing for the last 7 days? Unemployed and actively looking for a job".
Benefits Discrete, 0/1 "Please consider the income of all household members and any income which may be received by the household as a whole. What is the main source of income in your household?" [Unemployment, redundancy benefits, any other social benefits or grants].

Household income satisfaction
Discrete, 0/1 "Which of the descriptions on this card comes closest to how you feel about your household's income nowadays?" [Difficult / very difficult on present income].

General economic outlook
Discrete, 0/10 "On the whole how satisfied are you with the present state of the economy in [your country]?" Regional per-capita GDP Continuous Ln average regional per-capita GDP. Data from Cambridge Econometrics. Change in regional per-capita GDP

Continuous
Overall percent variation over the previous 4 years of the regional per-capita GDP, divided by three terciles (below 1.33%, between 1.34 and 7.17%, above 7.18%). Own calculation on data from Cambridge Econometrics.
Regional/national per-capita GDP % ratio Continuous % ratio of average regional-per capita GDP over national average. Own calculation on data from Cambridge Econometrics.
Voted in elections Discrete, 0/1 "Some people don't vote nowadays for one reason or another. Did you vote in the last [country] national election in [month/year]?" Satisfaction with democracy Discrete, 0/10 "On the whole, how satisfied are you with the way democracy works in [your country].

Satisfaction with life
Discrete, 0/10 "All things considered, how satisfied are you with your life as a whole nowadays?" Trust in political parties Discrete, 0/10 Using this card, please tell me on a score of 0-10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust…. Political parties?
L-R placement Discrete, 0/10 "In politics people sometimes talk of "left" and "right". Using this card, where would you place yourself on this scale, where 0 means the left and 10 means the right?" Trust in the police Discrete, 0/10 "Using this card, please tell me on a score of 0-10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust… The police?" Welfare state support Discrete, 0/5 "Using this card, please say to what extent you agree or disagree with each of the following statements. The government should take measures to reduce differences in income levels".
Migration good for economy Discrete, 0/10 "Would you say it is generally bad or good for [country]'s economy that people come to live here from other countries?" Migration good for culture Discrete, 0/10 "And, using this card, would you say that [country]'s cultural life is generally undermined or enriched by people coming to live here from other countries?"

Trust in European Parliament
Discrete, 0/10 "Using this card, please tell me on a score of 0-10 how much you personally trust each of the institutions I read out. 0 means you do not trust an institution at all, and 10 means you have complete trust… The European Parliament?"