The urban-rural polarisation of political disenchantment: an investigation of social and political attitudes in 30 European countries
←
→
Page content transcription
If your browser does not render page correctly, please read the page content below
Cambridge Journal of Regions, Economy and Society 2021, 14, 565–582
https://doi.org/10.1093/cjres/rsab012
Advance Access publication 16 July 2021
The urban-rural polarisation of political
disenchantment: an investigation of social and political
attitudes in 30 European countries
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
Michael Kennya and Davide Lucab,c,
a
Bennett Institute for Public Policy, Cambridge University, 7 West Road, Cambridge, UK,
mk30@cam.ac.uk
b
Department of Land Economy, Cambridge University, 19 Silver Street, Cambridge, UK
c
Middle East Centre, London School of Economics and Political Science, London, UK,
dl622@cam.ac.uk
Received on September 29, 2020; editorial decision on May 24, 2021; accepted on June 28, 2021
Relatively little research has explored whether there is a systemic urban-rural divide in
the political and socioeconomic attitudes of citizens across Europe. Drawing on individual-
level data from the European Social Survey, we argue that there are strong and significant
differences between the populations in these different settings, especially across western
European countries. We suggest that this divide is a continuum, running on a gradient from
inner cities to suburbs, towns and the countryside. The differences are explained by both
composition and contextual effects, and underscore how a firmer appreciation of the urban-
rural divide is integral to future place-based policy responses.
Keywords: urban-rural divide, regional inequality, geography of discontent, political disenchant-
ment, Europe
JEL codes: D72, R20, R58, Z13
Introduction deepening—geographical cleavage, with almost
all large cities being Democratic strongholds
While social scientists for much of the 20th
and rural counties being a cornerstone for the
century tended to assume that political cleav-
Republicans (McKee, 2008; Monnat & Brown,
ages in western democracies revolved around
2017; Scala & Johnson, 2017).
differences in class position and attitudes to-
Across Europe too, notable political events
wards distributional questions and the role
such as the UK 2016 Brexit vote, and the 2018
of the state, in recent decades there has been
Gilet Jaunes protests in France have shed light
a growing emphasis on those associated with
on marked political divergences between
various kinds of group identity and, latterly,
urban and rural places. Yet despite growing
with the importance of place (Kenny, 2014;
evidence from individual countries such as the
Kriesi, 2010). In the US, a large body of work
UK (Goodwin and Heath, 2016; Jennings and
has documented how political differences
Stoker, 2016), France (Guilluy, 2016; Ivaldi and
are increasingly driven by a distinctive—and
© The Author(s) 2021. Published by Oxford University Press on behalf of the Cambridge Political Economy Society.
This is an Open Access article distributed under the terms of the Creative Commons Attribution License
(https://creativecommons.org/licenses/by/4.0/), which permits unrestricted reuse, distribution, and reproduction
in any medium, provided the original work is properly cited.Kenny and Luca
Gombin, 2015) and Italy (Agnew and Shin, in three distinctive ways. First, we show how
2020; Rossi, 2018), relatively little research differences associated with the urban/rural con-
has explored in a systematic way whether the tinuum are significant across a wide range of at-
growing political divide between urban and titudinal dimensions. Most recent studies of the
rural areas mapped in some countries is ap- ‘geography of discontent’ (inter alia, Dijkstra
parent across the whole continent. et al., 2020; Rodríguez-Pose, 2018) consider the
The current article aims to fill this gap. evidence supplied by voting patterns in elec-
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
Drawing on individual-level data from the tions and referendums. Yet such events may
European Social Survey (ESS), it provides well be linked to candidate-related factors and
systematic comparative evidence across 30 contingent political developments, and may
European countries over the period 2002– relate only indirectly to underlying shifts in
2018. We explore links between place of resi- popular attitudes (Abrams and Fiorina, 2012;
dence and attitudes on a range of different Ford and Jennings, 2020). In fact, numerous pol-
socioeconomic and political issues. To antici- itical scientists suggest that electoral campaigns
pate our conclusion, we find that there is a do not change public opinion that much, but ra-
strong and significant divide between the pol- ther ‘activate’ some considerations over others
itical outlooks of urban and rural Europe. But (Mutz, 2018), increasing the extent to which
this divergence is not best seen in binary terms, particular issues matter for voters when they
and is better understood as a gradient run- choose a candidate. It is therefore important
ning from inner cities to metropolitan suburbs, to understand in more depth the factors ex-
towns and the countryside (as anticipated by plaining the underlying dynamics of public
Scala and Johnson, 2017 in the US context). We opinion. We seek to address this challenge using
show how, compared to dwellers in inner urban attitudinal data, and our findings suggest that
cores, respondents living in suburbs, towns and the linkage between the place of residence and
rural areas are more likely to be conservative political attitudes encompasses a wider range
in their orientation, dissatisfied with the func- of political and socioeconomic issues including
tioning of democracy in their country, and less perceptions of political behaviour and trust in
likely to trust the political system, even though political institutions.
they are strikingly more likely to participate in Second, in line with the work of Scala and
it, especially by voting – a finding which has an Johnson (2017) on the US, we show that the
important bearing on current debates about geographical divide in European political atti-
the future of democratic politics (Runciman, tudes should not be thought of as a dichotomy
2018). However, while our analysis highlights between urban and rural places—as suggested
some stark geographical variances in atti- for instance by Cramer (2016) in relation to
tudes towards migration and globalisation, we the US—but conceived instead as a gradient.
do not find significant variation on issues that This finding is in line with some recent analyses
have traditionally been at the core of left-right which underline how inequalities and residen-
cleavages, such as support for welfare state re- tial segregation between inner urban areas and
distribution. And, finally, we show that these suburbs are increasing in many European cities
differences, which are particularly strong across (Musterd et al., 2017).
western European countries, are explained by Third, we provide a preliminary analysis
both compositional and contextual effects. of some of the factors that may explain the
This article contributes to the literatures in differences that exist along the urban/rural
geography and political science devoted to the continuum. In his analysis of attitudes to immi-
spatial dimensions of political disenchantment gration, Maxwell (2019) argues that differences
566The urban-rural polarisation of political disenchantment
between urban and rural areas are more In the second half of the 20th century this
strongly driven by sociodemographic charac- stark divide faded partially, as sharper polit-
teristics—that is, by compositional effects— ical cleavages, which reflected economic issues,
than by the influence of place of residence upon class divisions and the role of the state in so-
individual outlooks. In contrast, we underline ciety, emerged (Lipset and Rokkan, 1967). Yet
how attitudes vary across sociodemographic across many advanced economies, the processes
and geographical dimensions. While people’s of economic globalisation over the last three
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
attitudes are heavily stratified by key individual decades have generated new socio-economic
observable characteristics, such as age, educa- divides (Ford and Jennings, 2020) and contrib-
tion and occupational status, we uncover a non- uted to the emergence of a new dimension of
negligible correlation between places and their political conflict, cutting across these older di-
inhabitants’ attitudes towards various political visions. Although the extent and nature of this
and social issues. This conclusion has important divide remains contested among social scien-
implications for the ongoing debate about what tists (Norris and Inglehart, 2019), a growing
kinds of policy solutions are best equipped to number of studies show that the classic class-
address growing territorial inequalities, and based Left-Right cleavage in party competition
whether these should be place-sensitive or not is today overlaid by a new division based on
(cf. Barca et al., 2012; Crescenzi and Giua, 2019; education and cultural attitudes. Scholars sup-
Iammarino et al., 2019).The article is structured port this claim with reference to survey data,
as follows. The second section reviews the ex- the positioning of political parties, and the
isting literature on the urban-rural polarisation, composition of party supporters (Norris and
and develops our main, empirical hypotheses. Inglehart, 2019; Piketty, 2018).
Section three describes the data and the analyt- Three accounts figure prominently in this
ical strategy that we have employed. We then debate, each proposing a distinct explanatory
present the key results in section four. In the framework to explain this new cleavage: “ma-
final section, we offer some discussion of their terialism” as opposed to “post-materialism”
implications in relation to ongoing policy de- (Inglehart, 1997); the divide between “winners”
bates, and suggest areas where further research and “losers” of globalisation (Kriesi, 2010); and
would be profitable. a “transnational” conflict of values between “lib-
erals” and “conservatives, authoritarians, and/
Political polarisation along the or nationalists” (De Vries, 2018; Hooghe and
urban-rural divide Marks, 2018). While each approach emphasizes
The polarization of electorates across the a specific trigger, the literature overall points to
urban-rural divide is by no means a new, or re- the increasing salience of geography in relation
cent, phenomenon. At the peak of the indus- to this new attitudinal cleavage, and to growing
trial revolution, between the end of the 19th political disagreements between cosmopolitan,
Century and the beginning of the 20th, many highly educated, and socially progressive urb-
European and North American countries were anites, and nationalist and socially conservative
divided politically between the interests of rural residents of ‘hinterland’ areas.
and small-town dwellers, engaged in agricul- In the US, a significant amount of work has
tural production, and those of urban residents, documented how electoral politics falls in-
experiencing rapid change and a new spatial creasingly into distinctive spatial patterns, with
economic order dominated by manufacturing almost all large cities being Democratic strong-
in large agglomerations (cf. Lipset and Rokkan, holds and rural counties being the cornerstone
1967; Vidal de la Blache, 1913). for the Republicans (Gimpel and Karnes, 2006;
567Kenny and Luca
Monnat and Brown, 2017; Rodden, 2019; Scala systemically rooted urban-rural divide across
and Johnson, 2017). While a broad division be- the whole European continent. And there is
tween ‘blue’ and ‘red’ America has been ob- still considerable disagreement between two
served for some decades,1 the emergence of a broad accounts about the causal dynamics and
salient divide between urban and rural areas processes underpinning this division.
has become more palpable over time and was The first of these relates to composition ef-
particularly clear during the 2016 presiden- fects, that is the spatially heterogenous distribu-
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
tial campaign (Wilkinson, 2018). Analysing tion of individuals with different characteristics.
recent opinion polls and the latest US presi- Research on political disenchantment and
dential elections, Scala and Johnson (2017) populism primarily identifies the archetype of
for example identify a consistent gradient of the anti-system supporter based on gender, age,
conservative sentiment from the most urban education and income (Dijkstra et al., 2020;
to the most rural counties on a wide range of Essletzbichler et al., 2018; Ford and Goodwin,
socio-economic issues. 2014; Goodwin & Heath, 2016). Composition
Across Europe too, there are signs that many effects may be amplified because of increasing
different political systems are adapting to this demographic ‘sorting’ among voters along spa-
new cleavage, and an increasingly spatially div- tial lines (Bishop, 2009)—which occurs pri-
ided electoral geography is emerging (Agnew marily through the dynamic self-selection of
and Shin, 2020; Hooghe and Marks, 2018). younger, more educated and socially liberal in-
France is a much-cited exemplar of this trend. dividuals in large, urban cores. In this changing
There is a growing political divide between the social landscape, large urban areas incubate
‘globalised’ and ‘gentrified’ large urban centres, more economic opportunities and attract those
the banlieues populated by immigrants of recent with greater skills and more liberal-minded,
arrival, and the remaining medium and small- while, conversely, smaller towns, rural areas and
sized cities and rural areas, where long-time cities with an outdated industrial mix become
immigrants and the ‘native’ working classes ex- increasingly ‘left behind’, losing their younger,
perience economic decline and are increasingly more skilled populations and facing economic
disaffected with the political system (Eribon, stagnation or decline (Crescenzi et al., 2016;
2013; Guilluy, 2016; Ivaldi and Gombin, 2015). Lee et al., 2018).
Similarly, England has witnessed a gradual The second explanation focuses on the
‘bifurcation’ (Jennings and Stoker, 2016) in pol- socio-economic trajectory of places and their
itical terms between people with higher educa- contextual role in shaping individual attitudes.
tion and good employment opportunities who Across Europe, there is increasing economic di-
live in metropolitan areas and those living in vergence between core cities and areas that are
‘backwater’ areas associated with economic de- lagging behind in economic terms (Iammarino
cline, hostility to immigration and the EU, and et al., 2019). Commentators hence point to an
a stronger sense of English identity (Garretsen emerging ‘geography of discontent’, reflecting
et al., 2018; Kenny, 2014, 2015). While there is the unhappiness of people living in places which
a strong regional dimension to the geography are stagnating or facing comparative economic
of discontent in Britain (McCann, 2019) in the decline (Garretsen et al., 2018; Los et al., 2017;
UK and elsewhere, the urban-rural fault-line McCann, 2019). Rodríguez-Pose (2018, p. 201)
has become increasingly prominent. for example claims that “[i]t has been thus the
But while many commentators observe places that don’t matter, not the ‘people that
this pattern in a few, paradigmatic countries, don’t matter’, that have reacted”. Rising oppor-
little research has explored whether there is a tunities and growth in thriving urban cores not
568The urban-rural polarisation of political disenchantment
only attract younger, more educated and more think of the urban/rural divide in dichotomous
liberal individuals, but also contributes to shifting terms. For example, even within metropolitan
urban dwellers towards more progressive social areas, there are significant differences between
values and cosmopolitan preferences (Vertovec urban cores and suburbs (Musterd et al., 2017).
and Cohen, 2002). Conversely, declining or stag- Drawing on such insights, our second research
nant material prospects in peripheral towns and hypothesis is the following:
rural places tend to generate a growing sense of H.2 = the urban/rural divide is best under-
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
disaffection, anxiety and resentment, driving citi- stood as a continuum rather than a dichotomy.
zens to adopt a more protective, “zero-sum, ‘us or In addition, we aim to provide a preliminary
them’ frame of mind” (Wilkinson, 2018, p. 5). exploration on the determinants of attitudinal
Place-related grievances are not confined differences across places. We test whether dif-
to economic issues. The differences between ferences in attitudes across the urban/rural con-
urban and rural life, and feelings among rural tinuum might not be exclusively explained by
and town dwellers that their places have been compositional effects. Our last two hypotheses
neglected by economic and political elites, have are the following:
led to growing resentment based on cultural- H.3 = differences in attitudes across the
identity issues which shape a growing sense of urban/rural continuum are explained by com-
mutual alienation (Lichter and Ziliak, 2017). positional effects.
As Cramer (2016) argues, what may look like H.4 = differences in attitudes across the
disagreements over specific policy preferences urban/rural continuum are not exclusively ex-
can often be traced back to this more funda- plained by composition effects and, hence, are
mental difference of worldview, which is rooted linked to some of the intrinsic characteristics of
in questions about identity and contending places.
“ideas about who gets what, who has power,
what people are like, and who is to blame”
(Cramer, 2016, p 5). Research design
In summary, there is considerable evidence Data
within a wide-ranging body of literature to sug- We analyse pooled, cross-sectional individual-
gest that there is a clear political fracture be- level data from the European Social Survey
tween metropolitan and rural (and semi-urban) which, since its inception in 2002, has conducted,
communities. But, as yet, it remains unclear every other year, face-to-face interviews across
whether this pattern works similarly across most participating countries. We draw upon
the European continent. One study—Maxwell data from the EU27 Member States plus the
(2019)—has provided a body of comparative UK, Norway and Switzerland from all the nine
evidence about popular attitudes towards im- available waves, covering the period 2002–2018.
migration in European countries. Our analysis The Survey is representative of all persons aged
builds on his work, seeing to explore a broader 15 and over, regardless of their nationality or
range of socioeconomic and political issues. language (we exclude respondents below 18).
Drawing on the existing literature, our first re- Individuals are selected through a multi-stage
search hypothesis is the following: random probability sampling procedure. The
H.1 = There are discernible differences in the ESS uses sampling designs where some groups
collective outlooks of people who live in urban or regions have higher probabilities of selec-
and rural places. tion. To reduce sampling errors and potential
But, as Scala and Johnson (2017) suggest in non-response bias, we apply country-specific
relation to the US case, it may be misleading to ESS post-stratification weights constructed
569Kenny and Luca
using information on age, gender, educa- common shocks throughout the years. ɛi,c,t is the
tion, and region.2 Furthermore, we also apply error term. We adopt robust standard errors in
country population size weights to account all regressions.
for the fact that countries participating in the Each of the dependent variables J included
ESS have relatively similar net sample sizes in the vector Y is either a dummy or ordinal
(roughly between 900 and 2700 individuals per categorical. In these cases, adopting a linear re-
country) even if the size of their general popu- gression model (cf., for instance, Maxwell, 2019)
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
lation varies considerably (for example from would be inappropriate because the assump-
1.1 million residents in Estonia to 71 million in tions of OLS are violated. We hence estimate
Germany during ESS wave 8). These weights model (1) by means of a logit estimator when
ensure that each country is represented in pro- the outcome is binary, or by means of a propor-
portion to its actual population size. tional odds estimator (ordinal logit) when the
outcome is ordinal categorical. In the second
Model and estimation strategy case, we assume that, for each outcome j, there
is only one model and one set of coefficients,
To test our research hypotheses, we estimate
and the only dependent variable parameter to
the following equation:
change across the values of the explanatory or-
dinal variable are the specific intercepts α (the
Yi,tJ = β1 Ui,t + β2 X L i,t + αc + dt + εi,c,t
(1)
cut-off points) – what is called the proportional
odds assumption. Brant tests, available on re-
Where Y is a vector of ordinal variables quest, confirm the assumption is not violated
measuring individual attitudes on each issue j (significant at the 1% confidence level).
of person i in the ESS wave t. U is our main It is important to bear in mind that this ana-
regressor of interest, and represents a self- lysis does not claim to provide a causal inter-
reported categorical variable indicating pretation of the link between place of residence
whether each respondent resides in a big city and political attitudes. Instead, it seeks to pre-
(the baseline category), in the suburbs/outskirts sent a broad and systematic analysis of a set of
of a big city, in a town/small city, in a country vil- quantitative, stylised facts, which might well be
lage, or in a farm/home in the countryside. Alas, explored in more depth, with the use of more
due to the nature of the data we cannot control advanced causal-inference tools, in future
for more fine-grained geographical determin- research.
ants, nor can we match individual observations
with more objective measures of urbanisation.
Variables and definitions
While this is a potential limitation of the ana-
lysis, in our approach we follow earlier work on Y is a vector of either binary or ordinal categor-
the urban/rural divide (Maxwell, 2019) and aim ical variables measuring individual attitudes on
to maximise the cross-country coverage offered each issue j of person i in the ESS wave t. We
by the ESS. consider ten issues, grouped along two main
X is a vector of sociodemographic controls L dimensions:
which may affect individual attitudes. European
countries are highly unequal in many geograph- Attitudes towards the political system and
ical (for example land size) and socioeconomic political engagement
aspects. We hence add state fixed-effects (FE) First, we are interested in the link between
αc, which are included to absorb any country- place of residence and individual attitudes to-
specific idiosyncrasies. We also add ESS wave wards the political system, as well as the ways
fixed-effects dt, to account for cross-sectional in which people engage with politics. We focus
570The urban-rural polarisation of political disenchantment
on attitudinal responses and views of modes of respondent resides in the inner part of a large
political behaviour as these are also revealing city (the baseline category), the suburbs/out-
expressions of deeper-lying attitudes towards skirts of a big city, a town/small city, a country
the political system. We explore, specifically, village, or a farm/home in the countryside.
the extent to which people engage via conven- Out of the total pooled sample, 19.46% of re-
tional political channels, such as voting, and the spondents report that they live in a big inner
extent to which respondents evince trust in pol- city, 12.04% in suburbs, 30.59% in towns
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
itical parties, since recent research has identi- or small cities, 31.41% in a country village,
fied a close correlation between discontent with and 6.5% in a farm or isolated home in the
the parties and a deeper mistrust of the polit- countryside.
ical system (Bromley-Davenport et al., 2019; X is a vector of individual sociodemographic
Cramer, 2016). Relatedly, we explore the extent controls L which may affect attitudes, and
to which people feel satisfied with the way in for which micro-level information is avail-
which democracy functions in their country. To able. In particular, X includes the following
provide a comparison, we also present evidence covariates:
on how people feel satisfied about their life
more generally, in order to help us understand Age
better the nature and extent of individual satis- Following the thrust of much recent literature,
faction and dissatisfaction with politics. we may expect attitudes to be highly stratified
by age groups, with younger generations being
Attitudes towards specific issues more likely to embrace cosmopolitan and pro-
gressive views (inter alia: Goodwin and Heath,
We are also interested in exploring how people
2016; Harris and Charlton, 2016) and, at the
respond to specific socioeconomic questions. We
same time, being less engaged in electoral pol-
first consider people’s self-placement along the
itics, given their familiarity with social media
left-right political spectrum, and then explore
and less conventional forms of political en-
their attitudes in three areas: welfare state sup-
gagement (Foa et al., 2020). The variable is ex-
port, which is conventionally treated as integral
pressed in Ln.
to the left/right divide; law and order, and trust in
the police, drawing on the extensive literature on
Gender
the rise of ‘authoritarian values’ (Foa and Mounk,
2016); and attitudes towards globalisation, which We control for the gender of the respondent,
we consider via perceptions of immigrants and since the literature primarily identifies the
the EU, where we draw from an emerging litera- archetype of the anti-system supporter as not
ture on a new ‘transnational cleavage’ in politics only older, but also male, native, and with a
(Ford and Jennings, 2020; Hooghe and Marks, lower level of education (inter alia: Goodwin
2018). Political disenchantment has been widely and Heath, 2016).
interpreted as an essentially populist reaction
against elite politicians who are perceived as Native
being increasingly globalist in their orientation We consequently add a dummy for people
by those more inclined to identify with national born in the country of residence, as we may ex-
identities and social traditions (Goodhart, 2017; pect this variable to affect our outcomes. For
Norris and Inglehart, 2019). instance, on net, we may expect natives to be
As anticipated, U represents a self-reported more prone to express dissatisfaction towards
categorical variable indicating whether each migration and globalisation.
571Kenny and Luca
Educational attainment respondents who, in previous years, have been
Some contributions have established a posi- unemployed for more than three months.3
tive association between lower degrees of
education and higher levels of anti-establish- Partner’s unemployment
ment feeling, as well as more nationalistic/ We also include a dummy if a respondent’s
inward-looking sentiments (inter alia: Dijkstra partner is unemployed.
et al., 2020). We hence control for respond-
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
ents’ highest level of education attainment Benefits
by including dummy variables for each of the
We control for whether the main source of
ISCED (International Standard Classification
household income claims state benefits. We
of Education) levels.
include this variable, as well as the following
ones, to account for potential divides between
Occupation ‘winners’ and ‘losers’ in the contemporary
The literature on political discontent has economy.
linked growing resentment with economic
insecurity in sectors and occupations under Household income feelings
higher threat from automation and trade We create a dummy capturing whether respond-
competition (Colantone and Stanig, 2018). ents feel that life with their present household
We hence additionally include dummies income is difficult or very difficult.
for each different type of occupation. We
follow the International Labour Office’s General economic satisfaction
(ILO) two-digit ISCO-08 (International
The variable captures the overall level of indi-
Standard Classification of Occupations) codes,
vidual satisfaction towards the national economy.
distinguishing between each of the 50 different
Research on the ‘geography of discon-
categories (out of the 96 codes) represented in
tent’ has pointed to a link between political
the ESS sample.
disenchantment and relative regional eco-
nomic status and decline (Rodríguez-Pose,
Employment status 2018). In the final part of the analysis we
We similarly include dummy variables for each will hence also include three regional-level
of the following statuses: employed in paid economic indicators which may affect in-
work, in education, unemployed, inactive, per- dividual attitudes. (While for most coun-
manently sick or disabled, retired, employed tries the ESS matches respondents to their
in community or military service, doing house- NUTS2 level region, in some cases persons
work or looking after children, and other. are matched with either NUTS1 or NUTS3
Following the research on the “winners” and regional identifiers. See Supplementary
“losers” of globalisation (Kriesi, 2010), we may Appendix A.1 for more details.) The vari-
expect disenchantment to be higher among ables we consider are:
those unemployed, inactive or retired.
Average regional per-capita GDP
Unemployment spells This is included to account for the overall eco-
While employment status captures current un- nomic development of the region where re-
employment, we further include a dummy for spondents live.
572The urban-rural polarisation of political disenchantment
Regional per-capita GDP growth into people’s self-placement on the left-right
We control for changes in GDP levels over spectrum, while the last five models focus on
the previous four years.4 We may expect dis- specific socioeconomic issues. Models six and
enchantment to be higher in regions where seven, respectively, focus on attitudes towards
growth has been sluggish or negative. welfare state support and trust in the police.
Finally, the last three models report results re-
% ratio regional per-capita GDP / lating to the ‘transnational cleavage’, namely
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
national average perceptions towards immigrants (columns eight
and nine) and attitudes towards the EU (trust
We add a measure of relative regional eco-
in the European Parliament, column ten). The
nomic wealth. We add this variable following
results broadly confirm our prior assumptions,
the empirical research that suggest how polit-
and provide strong evidence in support of hy-
ical discontent is higher in areas which are suf-
pothesis H.1. Across most issues covered, there
fering relative, rather than absolute, economic
are stark and statistically significant differences
stagnation or decline (Dijkstra et al., 2020; Lee
between urban and rural places. Besides, in line
et al., 2018).
with hypothesis H.2, Table 1 shows that the
Supplementary Appendix A.2 reports
divide in attitudes is a gradient linked to urban
key weighted summary statistics, while
density, running on a continuum from inner
Supplementary Appendix A.3 provides a de-
cities to suburbs, towns, villages, and isolated
tailed description for each variable.
rural houses.
Controlling for country and year idiosyn-
Results crasies, respondents living outside large inner
This section presents the baseline results of cities are, on average, significantly more likely
our analysis. In Table 1, we explore the overall to vote. At the same time, however, they tend
differences in attitudes that we find along the to show less trust towards the political system.
urban/rural continuum, when not controlling For instance, the odds of somebody voting, or
for composition effects. For each issue j, the reporting a one-unit higher level of trust in par-
table presents the proportional odds ratios ties (which is measured on a scale 0–10), if they
(that is the exponentiation of the ‘raw’ logit/or- live on a farm or in an isolated rural area (the
dinal logit coefficients) of respondents living in last category), are, respectively, 33.5% higher
each of the geographical categories compared and 16.2% lower than those of an average resi-
to respondents residing in large urban cores, dent of a large urban core. These results sug-
the baseline category. In all models, we include gest that, while levels of trust in the political
country and year fixed-effects. system are lower in rural areas, in these places
The first four columns of Table 1 report out- traditional modes of political engagement are
puts for attitudes towards modes of political more prevalent.5 Our results, more generally,
engagement and the political system, and illu- confirm that the residents of these places are
minate the extent to which people engage via far less likely to engage in non-conventional
‘traditional’ political channels such as voting political behaviours, like signing petitions and
(column one), whether they trust political par- boycotting products.6 But they are also, para-
ties (column two), or whether they are satis- doxically, more sceptical than their urban coun-
fied with democracy in their country (column terparts about the political system and the
three). And, as a point of comparison, we choices it presents them with.
also report respondents’ satisfaction with life In line with characterisations of a cosmopol-
(column four). Column five provides insights itan/conservative divide between large urban
573574
Kenny and Luca
Table 1. Place of residence and individual attitudes: robust logit / ordinal logit estimates.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Attitudes towards the political system Attitudes towards specific issues
Voted in Trust in Satisfaction Satisfaction L-R place- Welfare state Trust in Migration Migration good Trust in
elections parties with democracy with life ment support police good for for culture EU Parl.
economy
Suburbs 1.102*** 0.902*** 0.961* 1.049** 1.169*** 1.012 1.016 0.837*** 0.806*** 0.830***
(0.038) (0.021) (0.022) (0.024) (0.026) (0.024) (0.023) (0.019) (0.019) (0.018)
Town 1.029 0.884*** 0.895*** 1.038** 1.180*** 0.941*** 1.035** 0.716*** 0.701*** 0.802***
(0.027) (0.016) (0.016) (0.018) (0.021) (0.017) (0.018) (0.013) (0.013) (0.014)
Village 1.150*** 0.821*** 0.849*** 1.169*** 1.366*** 0.937*** 1.063*** 0.633*** 0.602*** 0.710***
(0.030) (0.015) (0.015) (0.020) (0.024) (0.017) (0.018) (0.011) (0.011) (0.012)
Rur. house 1.335*** 0.838*** 0.829*** 1.271*** 1.572*** 0.967 1.075** 0.619*** 0.544*** 0.649***
(0.069) (0.027) (0.026) (0.041) (0.051) (0.032) (0.033) (0.020) (0.018) (0.020)
Observations 244,690 222,293# 244,690 244,690 244,690 244,690 244,690 244,690 244,690 244,690
Country FE yes yes yes yes yes yes yes yes yes yes
Year FE yes yes yes yes yes yes yes yes yes yes
Ind. ctrls no no no no no no no no no no
Econ. status no no no no no no no no no no
Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. # Not available for ESS wave 1. For each issue j, the table presents the proportional
odds ratios (that is 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.
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021The urban-rural polarisation of political disenchantment
centres and elsewhere (Guilluy, 2016), the re- such as cognitive traits and personality types.
sults of column five show that people living in Nevertheless, we work from the assumption
urban fringes, towns and the countryside are that any residual correlation between place of
significantly more likely to identify as conser- residence and attitudes that is not explained by
vatives, tending to place themselves on the personal socioeconomic characteristics might
right of the political spectrum. As an example, well be related to places, and their contextual
the odds of a person living in a country house effects. With the exception of life satisfaction
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
feeling one-unit closer to the political right on and, partially, also for trust in the police, for all
the left-right spectrum (which is measured on a other outcomes the differences across places
scale 0–10, where zero is left and 10 is right) is after conditioning on individual covariates re-
57.2% higher.7 duce in magnitude and significance, lending
Interestingly, however, we do not find any support to hypothesis H.3.
substantial difference in support for the wel- In Supplementary Appendices A.4 and A.5
fare state (model six), an issue which has trad- we break down the results of Table 2, respect-
itionally played an important role in left-right ively controlling for only sociodemographic or
political cleavages in western democracies, or, only economic observables, to explore the ex-
indeed, in the trust they place in the police. It tent to which composition effects are linked
may well be that in the era of populism, worries to demographic factors such as education, age,
about welfare and law and order are no longer a and indigeneity, as opposed to labour market
source of significant divergence between those and economic factors. The results suggest that
who live in different parts of a country. both groups of regressors are important in
By contrast, results from the last three explaining attitudinal differences along the
models provide clear evidence of a stark urban/ urban/rural continuum (in fact, including either
rural divide on issues associated with the new group leads to relatively similar reductions in
‘transnational cleavage’ (Hooghe and Marks, the size of the urban/rural coefficients). Among
2018; Kriesi, 2010). The results of columns the economic regressors, additional tests we
eight and nine show significant differences in ran suggest that only employment status and
attitudes towards international immigration. sector of occupation play a role in mediating
As an example, compared to an inner-city the link between place of residence and indi-
dweller, the odds for a rural home resident re- vidual attitudes, while proxies for individual
porting a one-unit higher level of belief in the deprivation such as being dependent on public
positive role of migration in enriching the na- benefits, anxiety about household income and
tional culture (measured on a scale 0–10) are overall satisfaction with the economy, have a
more than 55% lower. A very similar picture very minor mediating effects.8
emerges with respect to attitudes towards the In Supplementary Appendices A.6 and A.7
EU (column ten). we stratify the sample of Table 2, respectively
In Supplementary Appendix B.1 we plot the distinguishing between Western European
predicted probabilities for models three and countries (EU14 Member States plus UK,
eight from Table 1. The graphs provide visual Norway, and Switzerland) and the 13 coun-
evidence of the differences in attitudes across tries which joined the EU in the 2004/07 en-
the urban/rural continuum. largements, most of which were formerly part
In Table 2 we test hypotheses H.3 and H.4, of the Eastern Bloc. The outputs suggest how
and present the results, controlling for in- attitudinal heterogeneity along the urban/rural
dividual observable characteristics. We are continuum is particularly pronounced across all
unable to control for unobservable factors the countries of Western Europe. By contrast,
575576
Kenny and Luca
Table 2. Place of residence and individual attitudes: robust logit / ordinal logit estimates controlling for sociodemographic and economic individual characteristics.
(1) (2) (3) (4) (5) (6) (7) (8) (9) (10)
Attitudes towards the political system Attitudes towards specific issues
Voted in Trust in Satisfaction Satisfaction L-R Welfare state Trust in Migration Migration Trust in
elections parties with democracy with life placement support police good for good for EU Parl.
economy culture
Suburbs 1.068* 0.934*** 0.994 1.064*** 1.129*** 1.043* 1.021 0.905*** 0.882*** 0.888***
(0.038) (0.022) (0.022) (0.024) (0.025) (0.025) (0.023) (0.020) (0.020) (0.020)
Town 1.058** 0.947*** 0.971* 1.110*** 1.144*** 1.019 1.071*** 0.841*** 0.824*** 0.893***
(0.029) (0.017) (0.017) (0.020) (0.021) (0.019) (0.019) (0.015) (0.015) (0.016)
Village 1.189*** 0.882*** 0.919*** 1.252*** 1.272*** 1.022 1.094*** 0.793*** 0.769*** 0.811***
(0.033) (0.016) (0.016) (0.022) (0.023) (0.019) (0.019) (0.014) (0.014) (0.015)
Rur. house 1.262*** 0.903*** 0.897*** 1.392*** 1.389*** 1.051 1.094*** 0.800*** 0.738*** 0.768***
(0.068) (0.030) (0.029) (0.046) (0.046) (0.035) (0.035) (0.026) (0.024) (0.025)
Observations 244,690 222,293# 244,690 244,690 244,690 244,690 244,690 244,690 244,690 244,690
Country FE yes yes yes yes yes yes yes yes yes yes
Year FE yes yes yes yes yes yes yes yes yes yes
Ind. ctrls yes yes yes yes yes yes yes yes yes yes
Econ. status yes yes yes yes yes yes yes yes yes yes
Robust standard errors in parentheses *** p < 0.01, ** p < 0.05, * p < 0.1. # Not available for ESS wave 1. For each issue j, the table presents the proportional
odds ratios (that is 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.
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021The urban-rural polarisation of political disenchantment
EU13 Members show significantly less marked respondents’ region of residence. Besides, not
differences. all individuals are matched with a regional iden-
In Supplementary Appendices B.2, B.3, tifier, while we do not have regional economic
and B.4 we then plot the predicted probabil- data for the latest ESS wave 9. We are hence
ities for models three and eight of Table 2. In able to merge regional-level characteristics
each of the appendices we plot probabilities only to waves 4–8 and a sub-set of respondents.
distinguishing between age groups, educational (For comparison, Supplementary Appendix
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
attainments, and occupation, and holding A.8 re-estimates the regressions of Table 2 on
other variables constant at their means. As the restricted sample. With the exception of
the results suggest, the role of age, education ‘Satisfaction with democracy’, which partially
and occupation in explaining variation in at- loses significance, results are overall similar to
titudes is significantly larger than the residual those from the full sample.) Results, reported
variation attributable to idiosyncratic place in Supplementary Appendix A.9, suggest that
effects. Hence, the findings from Table 2 sug- controlling for regional economic dynamics has
gest that attitudes are significantly stratified by only a minor effect on the link between place
sociodemographic measures, as suggested, for of residence and individual socio-political atti-
instance, by Maxwell (2019). Nevertheless, we tudes. Additional tests we ran equally suggest
underscore how, even after controlling for in- that the regional economic variables do not
dividual observable characteristics, places still act as moderators, as their interactions with
have a non-negligible correlation with people’s place of residence are statistically insignifi-
political attitudes, especially on the dimensions cant.9 Some recent analyses of the ‘revenge of
of voting behaviour (column 1 of Table 2), left- places that don’t matter’ (Dijkstra et al., 2020;
right placement (column 5), and migration and Rodríguez-Pose, 2018) have underlined a link
globalization (that is those relating to the ‘new between contemporary electoral political griev-
transnational cleavage’, columns 8 to 10). For in- ances and territorial economic stagnation and
stance, compared to an inner-city dweller, even decline. While our methodology and data are
after controlling for individual observables, the not closely comparable, our findings underscore
odds for a rural home resident reporting a one- how differences in attitudes along the urban-
unit higher level of satisfaction with democracy rural continuum may be also linked to broader
in their country (measured on a scale 0–10) are cultural-identity issues, as highlighted for ex-
more than 10% lower, while the odds of them ample by Cramer (2016) on the US. Future re-
reporting a one-unit higher level of belief in search may explore through individual country
the positive role of migration in enriching the studies the conclusions of our paper, and con-
national culture (measured on a scale 0–10) are sider in more depth the extent to which cultural-
more than 26% lower. In other words, we do identity and territorial economic factors interact
not fully reject hypothesis H.4, but instead con- in determining political disenchantment in atti-
clude that, while compositional effects are very tudes and at the ballot box.10
important in shaping attitudes, they are not suf-
ficient to explain the urban/rural divide in pol-
itical views in these European countries.
Conclusion and implications
To understand what may explain the link for policy
between place and individual attitudes, we Drawing on individual-level data from the
re-estimate equation (1), controlling for European Social Survey (ESS), this art-
sociodemographic observables and economic icle explores linkages between place of resi-
status, while also adding regional economic char- dence and attitudes on a range of different
acteristics. Before wave 4 the ESS did not report socioeconomic and political issues, providing
577Kenny and Luca
systematic comparative evidence across 30 products). This last finding, in particular, has
European countries over the period 2002– an important bearing on current debates
2018. Confirming earlier research on North about the future of democracy (Runciman,
America (Gimpel & Karnes, 2006; Monnat & 2018), and the potential risk that democratic
Brown, 2017; Rodden, 2019; Scala & Johnson, politics may become ‘eroded from within’ by
2017), we show that there is a strong and sig- individuals who engage with elections while,
nificant divide also between urban and rural at the same time, distrust the political system
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
Europe, a finding that carries significance for and are drawn to populist, anti-system politics
debates in relation to two broad phenomena. (Levitsky & Ziblatt, 2018).
One is the growth of political disenchantment The analysis also contributes to an ongoing
in many non-metropolitan locations across debate about whether people’s outlooks are ul-
Europe, and the fertile soil this provides for timately shaped by sociodemographic character-
nationalist and populist parties and causes. The istics—that is, by compositional effects and the
other is the debate about what kinds of policy geographical sorting of people with different at-
agenda and political response are required in tributes and outlooks—or by place effects (Abreu
order to re-engage the inhabitants of what are & Öner, 2020; Maxwell, 2019). Even though atti-
commonly termed ‘left-behind’ places. tudes are highly stratified by individual charac-
In relation to current debates about the teristics such as age, educational attainments, and
underpinnings and scope of political disen- occupation, we underline how place still appears
chantment, our findings suggest the import- to have a non-negligible correlation with values
ance of a place-sensitive conception of this and outlooks after controlling for individual
phenomenon, and simultaneously serve to covariates. More work is needed to understand
undermine overly generalised characterisa- better the mechanisms through which this rela-
tions of ‘rural consciousness’ or ‘left-behind’ tionship works. Recent research has shown how
disillusion (Cramer, 2016; Harris & Charlton, place of birth and the context where individuals
2016). The clear gradient that we identify in spend their ‘impressionable years’—that is the
terms of political attitudes and social values, period of late adolescence and early adulthood
and their correlation with different spatial during which people form durable political atti-
scales and kinds of community—ranging from tudes—have a significant influence in moulding
metropolitan centres at one end of the spec- both observable characteristics such as education
trum through to more remote, rural areas at (Bosquet and Overman, 2019) and unobservable
the other—suggest the need for a more de- cognitive capacities (Rentfrow et al., 2008). Even
tailed and contextual understanding of the in some of the most dynamic and developed
geography of disillusion. Our analysis shows economies in the world, it appears that where
how, compared to inner urban core residents, you are born and grow up is one of the most im-
respondents living in suburbs, towns and rural portant facts about the life of any citizen, and
areas are more likely to have anti-immigration this should give policy-makers food for thought.
and anti-EU views, to be conservative in There are large numbers of people resident in
their orientation, dissatisfied with the func- areas where trust in politics and the political
tioning of democracy in their country, and system is low, and where socially liberal values
less likely to trust the political system, even if have only a thin presence. Yet, successful major-
they are strikingly more likely to participate itarian politics require that parties of the political
in it through voting (while, by contrast, people mainstream find ways to win the support of many
living in inner urban areas are more likely to of these voters, while also pursuing policies—in
engage in non-conventional political behav- areas like climate change or migration—which
iours, like signing petitions and boycotting may well be anathema to many of them.
578The urban-rural polarisation of political disenchantment
This challenge connects with the second This article does not claim to provide a causal
main implication of these results. Our ana- interpretation of the link between place of resi-
lysis suggests that a firmer appreciation of the dence and political attitudes, but, instead, seeks
geographical specificities of different rural to present a broad and systematic analysis of a
areas, towns and cities is integral to a more set of quantitative, stylised facts. Each of these
contextually informed and tailored policy re- could be explored in more depth in the future.
sponses to the challenges posed by regional Further research could, in particular, explore
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
inequality and discontent (Iammarino et al., how and why the kinds of setting where people
2019; Rossi, 2018). Place-sensitive policies live can influence the development of individual
will have to be developed in a way that in- traits (Bosquet and Overman, 2019; Rentfrow
tegrates an understanding of regional forms et al., 2008). Besides, to our best knowledge,
of inequality, but also the degree of urban- there is still very limited evidence on whether
isation and proximity to urban agglom- the growing political divide between urban and
erations. For instance, with reference to rural areas mapped on both sides of the North
economic processes, research has shown that Atlantic is also prevalent around other parts of
the European Cohesion Policy has contrib- the world. Future research should explore this
uted to generating economic growth in rural issue, expanding the analysis to include countries
areas close to urban agglomerations, but not from both the ‘Global North’ and the ‘Global
in those farther away from cities (Gagliardi South’. Last but not least, further work is needed
and Percoco, 2016). Similarly, recent work to disentangle potential compositional effects
from both the US and Europe underlines based on unobservable—rather than observ-
how many rural areas face social and demo- able—characteristics such as intrinsic ‘cognitive
graphic challenges which undermine govern- underpinnings’ (cf., for example, Rentfrow et al.,
ments’ efforts to deliver public services and, 2008). Yet, overall we find that irrespective of
ultimately, the wellbeing of residents as well whether divides in attitudes are driven by com-
as their perception of being excluded by the positional effects or the contextual influence of
broader society (Accordino, 2019; OECD, places on people, the overlapping of territorial
2021). Hence, ‘place-sensitive’ public pol- and attitudinal cleavages signals a deepening
icies require a deeper and more contextual geographical fracture in European societies
appreciation of the different patterns of dis- which, in the long term, may have significant im-
enchantment apparent in different places, as plications for the challenges of generating social
well as an understanding of how economic cohesion (Wilkinson, 2018) and addressing the
and non-economic factors interact in driving implications of rising disenchantment with dem-
individual dissatisfaction. ocracy (Foa et al., 2020).
Finally, our results lead to the conclusion
that there are common important trends and
dynamics at work across the continent and, es- Supplementary material
pecially, across western European countries. Of Supplementary material is available at Cambridge
course, there are still key differences of political Journal of Regions, Economy and Society
economy, history and institutional structure at Journal online.
work in these different countries and regions.
Yet, understood as a wider phenomenon, we are
Endnotes
much more likely to grasp the underlying eco-
nomic and cultural dynamics that are driving 1
Some scholars have challenged the view of America
and perpetuating these spatially embedded as a country divided into two clearly distinct and polit-
patters of political disillusion. ically homogeneous areas (cf. Abrams & Fiorina, 2012).
579Kenny and Luca
2
Cf. https://www.europeansocialsurvey.org/method- Abreu, M. and Öner, Ö. (2020) Disentangling the
ology, accessed on 5 July 2019. Brexit vote: the role of economic, social and cul-
tural contexts in explaining the UK’s EU ref-
3
Data on long-term unemployment is unfortunately
erendum vote, Environment and Planning A:
missing for the majority of respondents. Economy and Space, 52: 1434–1456.
4
We calculate variations over an even-numbered Accordino, J. (2019) Introduction to bridging the
interval of years so that the measure coincides with “urban–rural divide”, State and Local Government
ESS waves. Review, 51: 217–222.
Downloaded from https://academic.oup.com/cjres/article/14/3/565/6322445 by guest on 16 December 2021
Agnew, J. and Shin, M. (2020) Mapping Populism.
5
Levels of trust in parties are virtually identical to Taking Politics to the People. London: Rowman &
levels of trust in politicians. By contrast, levels of trust Littlefield.
in the national parliament are slightly lower, con- Barca, F., McCann, P. and Rodríguez-Pose, A. (2012)
sistent with the hypothesis that disenchanted rural The case for regional development intervention:
dwellers may be more trustful of individuals or spe- place-based versus place-neutral approaches,
cific political parties that the political system overall. Journal of Regional Science, 52: 134–152.
These additional results are available on request. Bishop, B. (2009) The Big Sort. Boston, MA:
Houghton Mifflin.
6
These additional results are available on request. Bosquet, C. and Overman, H. G. (2019) Why does
7
Results not presented but available on request birthplace matter so much?, Journal of Urban
equally show statistically significant differences on Economics, 110: 26–34.
matters such as family issues and women rights. Bromley-Davenport, H., MacLeavy, J. and Manley, D.
(2019) Brexit in Sunderland: the production of dif-
8
These additional results are available on request. ference and division in the UK referendum on
9
They are available on request. European Union membership, Environment and
10
We thank one anonymous referee for providing Planning C: Politics and Space, 37: 795–812.
Colantone, I. and Stanig, P. (2018) The trade origins
such suggestion.
of economic nationalism: import competition and
voting behavior in Western Europe, American
Acknowledgements Journal of Political Science, 62: 936–953.
Cramer, K. J. (2016) The Politics of Resentment:
The authors thank the Special Issue editors and Rural Consciousness in Wisconsin and the Rise
the anonymous referees for helpful comments and of Scott Walker. Chicago: University of Chicago
suggestions, as well as participants at Birmingham Press.
University’s City REDI Research Seminar, the Crescenzi, R. and Giua, M. (2019) One or many
University of Barcelona AQR Research Seminar, the Cohesion Policies of the European Union? On the
differential economic impacts of Cohesion Policy
University of Cambridge’s Land Economic Work-in-
across member states, Regional Studies, 54: 10–20.
progress Seminar, the Italian Regional Science an- Crescenzi, R., Luca, D. and Milio, S. (2016) The geog-
nual 2020 Conference, and the 2020 European Week raphy of the economic crisis in Europe: national
of Regions. The authors are particularly grateful to macroeconomic conditions, regional structural
Raquel Ortega-Argilés, Paola Proietti, Roberto Foa factors and short-term economic performance,
and Enrique López-Bazo. All errors and omissions Cambridge Journal of Regions, Economy and
are our own. Society, 9: 13–32.
De Vries, C. E. (2018) The cosmopolitan-parochial
divide: changing patterns of party and elect-
Data statement oral competition in the Netherlands and beyond,
Journal of European Public Policy, 25: 1541–1565.
The data used in the research is available on request. Dijkstra, L., Poelman, H. and Rodríguez-Pose, A.
(2020) The geography of EU discontent, Regional
Studies, 54: 737–753.
References Eribon, D. (2013) Returning to Reims. Los Angeles,
Abrams, S. J. and Fiorina, M. P. (2012) The big sort CA: Semiotext(e).
that wasn’t: a skeptical reexamination, Political Essletzbichler, J., Disslbacher, F. and Moser, M.
Science and Politics, 45: 203–210. (2018) The victims of neoliberal globalisation and
580You can also read
