Content uploaded by Jona de Jong
Author content
All content in this area was uploaded by Jona de Jong on Nov 06, 2024
Content may be subject to copyright.
Enduring Divides?
Social Networks and the Entrenchment of Political
Polarization
Jona de Jong
Thesis submitted for assessment with a view to
obtaining the degree of Doctor of Political and Social Sciences
of the European University Institute
Florence, 31 October 2024
European University Institute
Department of Political and Social Sciences
Enduring Divides?
Social Networks and the Entrenchment of Political Polarization
Jona de Jong
Thesis submitted for assessment with a view to
obtaining the degree of Doctor of Political and Social Sciences
of the European University Institute
Examining Board
Prof. Hanspeter Kriesi, European University Institute (Supervisor)
Prof. Arnout van de Rijt, European University Institute (Supervisor)
Prof. Delia Baldassarri, New York University
Prof. Silja Häusermann, University of Zurich
© Jona de Jong, 2024
No part of this thesis may be copied, reproduced or transmitted without prior
permission of the author
Researcher declaration to accompany the submission of written work
Department of Political and Social Sciences - Doctoral Programme
I Jona de Jong certify that I am the author of the work "Enduring Divides? Social
Networks and the Entrenchment of Political Polarization" I have presented for
examination for the Ph.D. at the European University Institute. I also certify that
this is solely my own original work, other than where I have clearly indicated, in this
declaration and in the thesis, that it is the work of others.
I warrant that I have obtained all the permissions required for using any material
from other copyrighted publications.
I certify that this work complies with the Code of Ethics in Academic Research issued
by the European University Institute (IUE 332/2/10 (CA 297).
The copyright of this work rests with its author. Quotation from it is permitted,
provided that full acknowledgement is made. This work may not be reproduced
without my prior written consent. This authorisation does not, to the best of my
knowledge, infringe the rights of any third party.
I declare that this work consists of 54.654 words.
Statement of inclusion of previous work:
Chapter 2 is an extended version of an article published in Comparative Political
Studies, co-authored with Jonne Kamphorst. I contributed to two-thirds of the work.
Chapter 3 is co-authored with Tarek Jaziri Arjona. I contributed to two-thirds of the
work.
Chapter 4 is co-authored with Delia Baldassarri and is currently under review. I
contributed to 50% of the work.
Chapter 5 has been published in Nature Communications Psychology. I gratefully
acknowledge funding from an EUI Early Stage Research Grant and from the Dutch
Ministry of the Interior.
October 8, 2024
Abstract
This thesis advances a relational approach to study the durability of sociocultural po-
larization between citizens with and without tertiary education in Western democra-
cies, and the severity of partisan animosity in the United States. The broader question
is whether these represent enduring lines of division between cohesive social groups
with clear identities, or more ephemeral phenomena that will not structure politics for
decades to come, let alone cause excessive political conflict.
To understand the durability of educational divides, we lack clarity on what is creat-
ing cohesive collectives out of citizens with similar educational experiences, especially
given the waning of unions and churches that played an important role in creating past
collectives. To understand the severity of mass-level partisan conflict, the question is
whether partisanship is currently eclipsing other social identities in informing social
relationship formation, which can cause widespread social separation and excessive po-
litical division.
This thesis addresses both questions by advancing a relational approach, studying
the importance people attribute to education levels and partisanship in relationship for-
mation, the educational and partisan composition of social networks, and the role played
by social network composition in exacerbating or moderating group-based political di-
vision.
Chapters 2 and 3 focus on educational divides and propose that educationally homo-
geneous social networks have partly supplanted formal organizations in strengthening
and reinforcing initial education-based differences, thereby creating cohesive collec-
tives which consistently care about sociocultural issues, and durably vote for new left
and far right parties, suggesting persistent sociocultural conflict.
Chapters 4 and 5 shift to partisan divides. In contrast to much current literature,
we find little evidence that partisanship supersedes other considerations in real-world
relationship formation. Rather, social networks remain politically heterogeneous and
heterogeneous networks buffer partisan animosity. These results suggest that mass-level
partisan animosity is not as severe as previously thought.
i
.
ii
Contents
Abstract i
Acknowledgements vii
1 Introduction 1
1.1 Theoretical framework . . . . . . . . . . . . . . . . . . . . . . . . . . 4
1.1.1 Durable Division in Democracies . . . . . . . . . . . . . . . . 4
1.1.2 Changing Groups, Waning Organizations and Shallow Identi-
ties: a Turn Away From the Study of Cohesive Collectives . . . 8
1.1.3 A Relational Approach: Studying Relationship Formation, So-
cial Network Composition, and Network Influence to Assess
Group Cohesion and Identification . . . . . . . . . . . . . . . . 12
1.2 Thesisoverview.............................. 13
1.2.1 An Education Cleavage? Social Closure by Education Levels
and Entrenching Sociocultural Polarization . . . . . . . . . . . 13
1.2.2 The Primacy of Partisanship? Cross-Partisan Relationships in
an Era of Partisan Animosity . . . . . . . . . . . . . . . . . . . 16
1.3 DataandMethods............................. 20
2 Separated by Degrees: Social Closure by Education Levels Strengthens
Contemporary Political Divides 25
2.1 Introduction................................ 26
2.2 Theory................................... 30
2.2.1 Durable Political Competition, Social Closure and Social Net-
works............................... 30
2.2.2 Social Closure and Educational Divides . . . . . . . . . . . . . 33
2.2.3 Political Preference Crystallization in Social Networks . . . . . 36
2.3 DataandMethods............................. 38
2.4 Results................................... 46
2.4.1 Educational Homogeneity in Close Relationships . . . . . . . . 46
iii
2.4.2 Are Network Education Levels Associated with Political Atti-
tudesandVoting?......................... 51
2.4.3 Is a Change in Network Education Levels Associated With a
Change in Attitudes and Voting? . . . . . . . . . . . . . . . . . 59
2.5 Discussion................................. 64
A Appendix Chapter 2 67
A.1 More details on the case-control analysis . . . . . . . . . . . . . . . . . 67
A.2 Measures ................................. 69
A.3 Results: Robustness & Sensitivity Checks and Complete Regression
Models for the Cross-Sectional Analysis . . . . . . . . . . . . . . . . . 72
A.4 Results: Robustness & Sensitivity Checks for the Within-Individual Re-
sults.................................... 83
A.5 EthicsStatement ............................. 95
3 Economic Insecurity and Far Right Support: Social Relationships as Buffers
or Catalysts 97
3.1 Introduction................................ 98
3.2 Theory................................... 100
3.2.1 Structural change, Economic Crises and Far Right Support . . . 100
3.2.2 Social Relationships as Buffers or Catalysts of Far Right Appeal 103
3.2.3 Social Influence and Economic Resources . . . . . . . . . . . . 104
3.3 Data, Case Study and Research Design . . . . . . . . . . . . . . . . . . 108
3.3.1 Empirical Strategy . . . . . . . . . . . . . . . . . . . . . . . . 108
3.3.2 Measures ............................. 109
3.3.3 Weighting to Balance Sample . . . . . . . . . . . . . . . . . . 111
3.3.4 Model............................... 112
3.4 Results...................................113
3.4.1 Social Influence and Economic Resources . . . . . . . . . . . . 115
3.5 Discussion................................. 118
B Appendix Chapter 3 121
4 Seeing, not Avoiding the Other Side: the Persistent Role of Heterogeneous
Social Networks in an Era of Partisan Politics 125
4.1 Introduction................................126
4.2 Politically Heterogeneous Social Networks and the Rise of Affective
Polarization................................ 130
4.2.1 Does Partisanship Drive Social Selection? . . . . . . . . . . . . 132
4.2.2 Do Heterogeneous Social Networks Inoculate Against Partisan
Animosity?............................135
iv
4.3 DataandMethods.............................139
4.3.1 Measuring Close-Ties and Acquaintance Networks . . . . . . . 140
4.3.2 The Role of Partisanship in Social Selection . . . . . . . . . . . 141
4.3.3 Measuring Out-partisan Animosity . . . . . . . . . . . . . . . . 144
4.3.4 Measuring Misperceptions and Manipulating Perceived Similarity146
4.4 Results...................................147
4.4.1 Does Partisanship Affect the Formation of Social Relations? . . 147
4.4.2 Close-ties and Acquaintance Networks Heterogeneity . . . . . . 154
4.4.3 Do Heterogeneous Networks Limit In-partisan Selection? . . . 155
4.4.4 Do Heterogeneous Networks Inoculate Against out-partisan An-
imosity?.............................. 159
4.4.5 Seeing-the-other-side: Do heterogeneous Acquaintances Reduce
Misperceptions? .........................161
4.5 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . 164
C Appendix Chapter 4 167
C.1 Sample Characteristics and Controls . . . . . . . . . . . . . . . . . . . 167
C.2 Selection Experiments . . . . . . . . . . . . . . . . . . . . . . . . . . 168
C.3 Selection Experiments by Network Heterogeneity . . . . . . . . . . . . 171
C.4 Out-Partisan Animosity by Network Heterogeneity . . . . . . . . . . . 174
C.5 Relationship Between Heterogeneous Acquaintance Networks and Mis-
perceptions ................................182
C.6 Experimental Manipulation in Support of the Seeing-the-other-side Mech-
anism ...................................184
5 Cross-Partisan Discussions Reduced Partisan Animosity Between UK Vot-
ers, But Less so When They Disagreed 191
5.1 Preface .................................. 192
5.2 Introduction................................193
5.3 Theory................................... 194
5.4 DataandMethods.............................198
5.4.1 Sample, Recruitment and Exclusion . . . . . . . . . . . . . . . 199
5.4.2 Measures ............................. 201
5.4.3 Analysis .............................201
5.4.4 Statistics and Reproducibility . . . . . . . . . . . . . . . . . . 202
5.4.5 TheExperiment..........................204
5.5 Results...................................206
5.5.1 Participant Demographics and Attrition . . . . . . . . . . . . . 206
5.5.2 Affective polarization . . . . . . . . . . . . . . . . . . . . . . . 208
5.5.3 OpinionChange .........................212
5.5.4 Mechanisms ...........................216
v
5.6 Discussion................................. 221
D Appendix Chapter 5 225
D.1 Additional Evidence from Real-World Cross-Partisan Discussions in
theNetherlands .............................. 225
D.2 Main Experiment: Recruitment and Final Sample . . . . . . . . . . . . 229
D.3 Supplementary Analysis . . . . . . . . . . . . . . . . . . . . . . . . . 245
D.3.1 Affective Polarization (Fig. 3 Paper) . . . . . . . . . . . . . . . 245
D.3.2 Discussion Effects by Issue Disagreement (Fig. 4 Paper) . . . . 251
D.3.3 Opinion Change (Fig. 5 Paper) . . . . . . . . . . . . . . . . . . 258
D.3.4 Mechanisms ...........................271
D.3.5 Further Pre-Registered Hypotheses and Deviations from the Pre-
AnalysisPlan...........................275
6 Discussion 285
6.1 Contemporary Cleavage Formation . . . . . . . . . . . . . . . . . . . . 286
6.2 Beyond the Social Side of Partisanship . . . . . . . . . . . . . . . . . . 291
Bibliography ..................................295
vi
Acknowledgements
On the end of young adulthood, Annie Ernaux writes:
Because summers had started to resemble each other (.. . ), because the self-realization
imperative was taking us nowhere fast, by dint of solitude and discussions in the same
cafés; because youth had come to feel like a vague and cheerless time whose end we
could not see (. . . ) (The Years, p. 86)
Before moving to Florence at age 28, I could relate to a version devoid of char-
acteristically French pessimism, having come to believe that in this transition period,
conversations still impress, but less often enthral; new ideas still captivate, but less of-
ten disorient; new friendships still form, but less often last. The last four years have
forever rebuffed these worries. I have many to thank for providing a uniquely stimulat-
ing environment, and for creating a tight-knit community of friends.
This thesis would have been very different had Hanspeter Kriesi not graciously
agreed to take me on as his final supervisee. During your course on Democracy, I
was immediately drawn to your – if you will – Swiss-deliberative, even Socratic ap-
proach to intellectual discussion: in addition to telling me where I took a wrong turn
with much-appreciated bluntness, you would just as often summarize my thoughts and
ask the type of questions that subtly guided me to errors and omissions in my thinking.
Your intellectual breath and endless curiosity provided freshness amidst a slight insti-
tutional tilt towards micro-studies and methodological dogmatism. In my own career,
vii
I will aim to emulate your unwavering dedication to your students, your insistence on
relevance in addition to rigor and your discouragement to too easily follow disciplinary
fads. To close, our multi-hour lunches spent talking everything life and politics were a
high point of my time at EUI.
My other supervisor, Arnout van de Rijt, embodied the second pillar on which this
thesis was built, as well as a model for how I wish to conduct myself in my future career.
A reason I quit my previous job in industry was that I could detect too little passion for
the work among people 10-15 years my senior. If that was to be my future, I’d rather
choose another. The contrast could not have been more stark. Arnout, your passion
for big, bold questions and the admiringly clever ways in which you seek answers are
a consistent source of inspiration. Your contagiously enthusiastic engagement with my
half-formed ideas was instrumental in their realization. I would always look forward to
our meetings – another EUI highlight – where your razor-sharp analytical mind provided
a necessary complement to my tendency toward inductive reasoning, and I greatly en-
joyed us occasionally veering off into witty banter and broader discussions of the future
of the social sciences.
It was also Arnout who encouraged me to visit the US, and who introduced me to
Delia Baldassarri, who deserves a full paragraph in these acknowledgements. We joke
sometimes among PhD’s how meeting scholars whose work you admire can be under-
whelming, not living up to the lofty image you constructed based on their brilliance on
a page. In this case, the person fit the scholar. Delia, your scholarship and our col-
laboration have greatly influenced my thinking, and working together while I was still
learning the ropes has honed my research skills like few other experiences. I am deeply
grateful for your efforts in hosting me at NYU, for immediately proposing a collabora-
tion where we hit the ground running, fielding the first survey only two months after I
viii
landed in New York, for your patience with my clumsy latex errors, career advice and
engagement with my work. Thank you, finally, for being on my committee. I look
forward to continuing our joint work.
I am also very grateful to Silja Häusermann for being on my committee. Your schol-
arship has been central to the first three chapters of this thesis, and I deeply appreciate
your thorough engagement with this piece of work.
Several scholars have been kind enough to meet and discuss my ideas during these
last years. I am particularly thankful to Peter Hall, for hosting me at Harvard, Vincent
Buskens, for hosting me in Utrecht, and to Liesbet Hooghe, Stefano Bartolini and Eelco
Harteveld for in-depth discussions. I also thank Peter Marsden, Torben Iversen, Elias
Dinas, Frank van Tubergen, Eva Jaspers, Gary Marks, Natalia Garbiraz-Díaz, David
Lazer, Jonathan Mijs and Jamie Druckman. In addition, this thesis has greatly bene-
fited from the opportunity to present its various chapters in colloquia and workshops at
Harvard, NYU, Utrecht and the EUI. Its introduction owes a lot to the sharp minds of
Anna Clemente and Juliette Saetre. I thank Maureen and Claudia for their help in the
process, and Ruben and Leon for the collaboration. Finally, I thank Koen Damhuis for
being the first to enthuse me about doing a PhD at the EUI. I always look forward to our
bi-monthly dinners spent discussing life in academia, societies´woes and the explana-
tory value of theories developed by a certain French sociologist who shall not be named.
These are the people to whom this thesis is intellectually indebted. But its comple-
tion would have been a much more solitary experience without the rare community we
built in Florence. For a few years, especially during the early stages of the pandemic,
it felt like we owned a slice of this magnificent city. Together, during never-ending
dinners, aperitivo’s and parties in apartments spread across its otherworldly beauti-
ix
ful neighbourhoods, we transitioned from students into scholars – climbing into each
other’s minds, as Grossman calls it, while occasionally collectively regressing back into
young adulthood. Thesis acknowledgements are a rare opportunity to formally thank
friends for their love and inspiration and I am glad for a ritualized excuse to do so.
Odysseas, ela re ma(..), the Greek army is lucky to have you as its postmaster. Your
ability to bring people together – whether it is around mouth-watering lamb dishes in the
flat at Mercato Centrale, the birthplace of our community, or in successful mobilization
against absurdly low wages for researchers from certain countries –, your bottomless
hospitality and interest in others are rare gifts that were crucial to the formation of our
group of friends. These words apply to Luca in equal measure. Thank you for putting
some soul into Boston and for finding a Dutch partner so that your warm company is
only a train-ride away. Anna, I am glad to have found a friend with an equally deep
conviction that academia can and should be a serious force for societal change. Our
hours-long, at times wine-fuelled discussions would reverberate for days and I hope to
keep them going for a long time. Juliette, your witty and warm presence and eternal cu-
riosity make for wonderful conversations about life, art and friendship. Julian, watching
you battle giants was an inspiration; watching you fight windmills a continuous source
of entertainment. I am in awe of your commitment to helping those in need and I hope
you will continue to apply your gifts to making our world a kinder place. Mariusz, I
thoroughly enjoyed our bike rides to Chianti through the luscious forests of Tuscany,
eating pasta cinghiale and discussing qualitative methods whose cutting-edgedness is
as present as the forests. I also thank Jan, Alsena, Susi, Gaia, Casper, Joe, Francesca,
Timo, LJ, Luis, Anica, and many others, for making life in Florence a delight.
Large parts of this work were written in the US, a country with which I have always
had a mixed relationship. When across the Atlantic, I was lucky to be surrounded by
x
several people that made my time worthwhile. I thank Dylan, for Brooklyn-based beers
between the Dutch and English hills, Spanish twins Nerea and Sergi for their friendship,
Lucie, Jill, Roxanne and Alberto for many fun nights, and Meg, for temporarily provid-
ing a home away from home.
Eight years of intermittently living abroad can make one feel uprooted. Close friends
and family are the reason this never happened. I will keep this part shorter: after all these
years, I need few words to tell you how dear you are to me. Timo, Arthur, Thomas,
Felix, Simon, our friendship is one of the most precious things I hold in life. In this
light, I also want to mention Marcel, Sjoerd, Paul, Hans, Nina, Minke, Koos, Noémie
and Misja.
Roots are crucial for the stability they provide and I could not have chosen better
ones than those that were given to me. I thank my late grandfathers, Opa Tijn and Opa
Frans, as well as Wouter, Damaz, Erik, Charlotte, Frank, Annelies, Jonas, Nicole, and
Peer.
My sister Judith, of whom I could not be more proud, is a vital source of sheer joy
in my life. My parents, I thank for a childhood home roaring with laughter and debate,
for their unwavering support and unconditional love.
I dedicate this thesis to my grandmothers. To the memory of Oma Joop, a bedrock
of my childhood whose strength I continue to carry deep within, and to Oma Lot, whose
life and convictions have profoundly shaped my own.
xi
.
xii
Chapter 1
Introduction
"In every book I write, I try to gently tug, in the hope to loosen, the solidified part of
our mind"
(David Grossman in: De Volkskrant. October 30, 2020)
Political polarization, like few other phenomena, has recently captured the imagi-
nation. Divisive sociocultural issues1around immigration, minority rights and cultural
liberalism are increasingly present in political and public discourse, fuelled by elec-
torally successful new left and far right parties, who find much of their support base
among citizens with and without tertiary education, respectively. Simultaneously, espe-
cially in the US, ordinary voters are said to be deeply socially divided by partisanship
(Iyengar, Lelkes, et al. 2019).
Scholars debate whether educational and partisan divides represent enduring lines
of division or more shallow, ephemeral phenomena that may occasionally heat tempers
but will not dominate politics for decades to come, let alone cause profound social and
1For parsimony, I use the term "sociocultural" to contrast these issues from those concerned with the
role of the state in the economy, the level of taxation, and the size and reach of the welfare state. Reality
is more nuanced: cultural issues have economic components and vice versa
1
political division. Enduring divides are worth studying because they typically affect the
policies on the political agenda, the coalitions that parties can form and therewith the
predictability of political competition (Lipset and Rokkan 1967). They can also become
excessively antagonistic, negatively affecting democratic stability (Lipset 1960).
According to classic literature in political sociology, political divides can become
enduring when they are rooted in conflict between sizable, cohesive social groups with
clear identities (Lipset and Rokkan 1967; Bartolini and Mair 1990; Schedler 2023). In
contrast, existing work on political polarization – with notable exceptions – has moved
away from the study of conflict between cohesive collectives. Its focus on changing dis-
tributions of individual-level attitudes on policy issues (called ideological polarization
(Fiorina and Abrams 2008; Abramowitz and Saunders 2008)), growing antipathy be-
tween partisan groups (called affective polarization (Iyengar, Sood, et al. 2012; Reiljan
2020)), or average group differences on clusters of issues and party vote (as is common
in the literature on sociocultural conflict in Europe (Hooghe and Marks 2018)) is impor-
tant in its own right, but says little of group cohesion and therefore does not address the
durability of contemporary conflicts.
Concerning the durability of educational divides on sociocultural issues, we know
that citizens with and without tertiary education have distinct social backgrounds, expe-
riences and opportunities. What is less clear is what may be shaping cohesive collectives
out of individuals with some group-based similarities. A complicating factor is that the
formal organizations (e.g. churches, unions) that used to play an important top-down
role in socially isolating citizens, thereby shaping cohesive collectives, have waned in
strength. The question, then, is what may be creating cohesive educational groups today.
Addressing the severity of partisan hostility, US citizens clearly express animosity,
discrimination and avoidance towards out-partisans in survey items. Widespread mu-
2
tual avoidance based primarily on partisanship could indeed mean a split of US society
into opposite, increasingly polarized camps (Baldassarri and Page 2021). Alternatively,
citizens select others based on other sociodemographic dimensions – such as education
– that correlate with partisanship but do not coincide with it (Knoke 1990). We do not
know to what extent partisanship is actually a subjectively meaningful social identity in
people’s everyday life, indeed informing discrimination and avoidance.
To study the durability of sociocultural polarization, and the severity of partisan an-
imosity, this thesis advances a relational approach, focusing on the importance people
attribute to education levels and partisanship in relationship formation, the educational
and partisan composition of social networks, and the effect of social network composi-
tion on political polarization.
Using this approach, we propose that educationally homogeneous social networks
have partly supplanted formal organizations in creating cohesive groups out of citizens
with similar experiences. In today’s individualistic, mobile societies, significant pro-
portions of citizens with and without tertiary education have become mutually insulated
through different, more bottom-up mechanisms, and the resulting homogeneous social
networks serve to strengthen and reinforce initial differences between them. We find
evidence of sizable, insulated educational groups with different political attitudes and
voting behaviour, suggesting persistent sociocultural polarization durably rooted in so-
ciety.
Applying the relational approach to study partisan animosity in the US, we closely
mimic real world processes of relationship formation to study the relative importance
people attach to their partisan identities in social selection. We show that partisanship
does not play a significant role in social selection: it is subordinated to other consider-
3
ations and social identities in informing relationship formation. In addition social net-
works remain politically heterogeneous and these heterogeneous networks play a role
in buffering animosity. The implication is that mass-level partisan conflict may be more
muted than previously thought.
1.1 Theoretical framework
This section first establishes when mass polarization may become durable and poten-
tially excessive. Guided by the premises here developed, I discuss the declining im-
portance attributed to cohesive social groups in the study of political polarization and
the puzzles that arise from it. Finally, I propose a relational approach to the study of
political polarization, which helps solve these puzzles and provides some insight into
the broader durability and potential gravity of contemporary sociocultural conflict and
partisan animosity.
1.1.1 Durable Division in Democracies
Inherent in all democratic systems is the constant threat that the group conflicts which
are democracy’s lifeblood may solidify to the point where they threaten to disintegrate
society (Lipset 1960 pp. 83)
A central framework to study the durability and politicization of social group divides
is provided by cleavage theory. Cleavage theory holds that durable political divides
arise out of critical junctures – revolutions, or large-scale technological or economic
upheavals – that propel long-running conflict between social groups differently affected
by these developments. This conflict in turn structures long-term party political com-
4
petition (Lipset and Rokkan 1967). Societal divisions between sizable (Posner 2004),
cohesive social groups limit parties’ room for manoeuvre in proposing policies and rep-
resenting groups (Rokkan 1999).2Economic issues, for instance, dominated politics for
decades and decades in part because cohesive groups of workers consistently demanded
state intervention and economic redistribution, and cohesive groups of employers con-
sistently opposed these demands.
Not all social divisions become durable or politicized. According to Bartolini and
Mair, a durable divide requires groups to hold different interests or worldviews, have
clear collective identities, and to be somewhat mutually insulated (Bartolini and Mair
1990). Worldviews and interests are based on similar structural positions creating simi-
lar experiences as well as repeated interactions and rituals (Berelson et al. 1954; Jenkins
2014). Moreover, collective identity "has to be made to matter" (Jenkins 2014 pp. 7).
Formal organizations (e.g. unions, churches) and political parties play a role in creating
collectives out of individuals with common interests and experiences (Bartolini 2000;
Bartolini 2005). They often contribute to the insulation of social groups, shielding them
from cross-group influence, which reinforces distinct worldviews and helps create col-
lective identities (Bartolini and Mair 1990). While political actors can thus mobilize and
thereby give subjective meaning to group identities, mobilization is simultaneously con-
strained by the raw material of pre-existing societal divides (Lipset and Rokkan 1967;
McCoy et al. 2018).
Durable group conflict is constitutive to democratic regimes, its lifeblood in the
words of Lipset. Conflict brings groups into the body politic and facilitates democratic
competition over ruling positions, challenges to parties in power, and shifts of parties
2Other work in political science and sociology affirms the crucial importance of sizable cohesive
social groups for durable political divides (Deutsch 1971; Esteban and Ray 1994; DiMaggio et al. 1996).
5
in office (Lipset 1960; Schattschneider 1960). Programmatic competition can energize
electorates (Abramowitz and Saunders 2008) and deter personalistic and clientelistic
politics (Aldrich 1995). In fact, after center-left and center-right parties in Europe pro-
gramatically converged on economic policy (Hall 2022) in effect largely making it the
purview of international organizations and bureaucracies (Mair 2013), scholars saw a
threat in too little polarization rather than too much, fearing the demise of adequate
political representation (Katz and Mair 1994; Mair 2013).
Moderate group-based conflict is thus part and parcel of democratic politics. Yet,
durable divides can become excessive. Some cleavage theorists, as well as scholars
of conflict and classical sociologists, propose that durable divides can escalate into ex-
cessive conflict when the multitude of cleavages that can constitute a society reinforce,
rather than crosscut each other, or when groups come to organize around singular, ex-
clusive identities (Simmel 1908a; Simmel 1908b; Lipset 1960; Coser 1956; Deutsch
1971; Esteban and Ray 1994; Dahl 1961; Blau 1977; Blau and Schwartz 1984; Roc-
cas and Brewer 2002; Baldassarri and Gelman 2008). Crosscutting group affiliations
prevent a single divide from dominating the public sphere, because loyalties are spread
over different groups and groups rely on one another (Simmel 1908a; Simmel 1908b).
A multiplicity of conflicts prevents the formation of fully united fronts (Coser 1956):
allies in one conflict are adversaries in another (Dahl 1983). If, conversely, conflicts
align, then "men and parties come to differ with each other, not simply on ways of
settling current problems, but on fundamental and opposed outlooks. This means that
they see the political victory of their opponents as a major moral threat, and the whole
system, as a result, lacks effective value-integration" (Lipset 1960 pp. 83).
When groups or parties hold incompatible worldviews, and come to see each other
as a threat to the democratic attainment of their interests, democratic cooperation and
6
compromise between political elites becomes more complicated (Lipset 1960; Haggard
and Kaufman 2021; Schedler 2023). Worse, politicians have an incentive to bend or
break democratic norms or rules in order to expand their power and "try to win at all
costs" (Levitsky and Ziblatt 2019).3Polarized voters may no longer punish democratic
transgressions, acting as "partisans first, and democrats only second" (Svolik 2018; Gra-
ham and Svolik 2020 pp. 292). When parties, voters and politicians no longer play by
the democratic rules of the game, democracy is in trouble.
The idea of crosscutting cleavages has been used to explain why historical conflict
became excessive or remained muted. Lipset and Rokkan, comparing the segmented
Netherlands and Austria in the interwar years, write: "if all three of the subcultures
(Catholic, Protestant, Liberal-Secular, in the Netherlands) had developed strong barriers
against each other, the system might conceivably have exploded, much like in the way
the Austrian polity did in 1934" (Lipset and Rokkan 1967, pp. 17.). While especially
conflict along class lines did at times become excessive in this period (Bartolini 2000;
Berman 2019), its worst forms were mostly kept at bay either because elites cooperated
across cleavage lines (Lijphart 1968; Lorwin 1971), or because sizable proportions of
society were cross-cut between, for instance, class-based and religious groups (Lipset
1960; Lipset and Rokkan 1967; Deutsch 1971).4
3In the past, deep antagonisms facilitated the rise of the anti-system parties and movements (Sartori
1976). Today democratic transgressions come primarily from within the party system (Levitsky and
Ziblatt 2019; Haggard and Kaufman 2021)
4While the number of historical cases is low, the broader insights they provide are complemented
by empirical studies at the individual level. They show that people whose group affiliations, or social
relationships pull them in different political directions hold weaker partisan identities (Berelson et al.
1954; Powell Jr 1976; Dassonneville 2022) and display lower levels of partisan animosity (Mason 2016;
Harteveld 2021).
7
1.1.2 Changing Groups, Waning Organizations and Shallow Iden-
tities: a Turn Away From the Study of Cohesive Collectives
Despite these earlier traditions, the study of political polarization as between sizable,
cohesive social groups has lost much of its appeal. This section describes the trends in
the literature that followed, and the questions they leave open.
Already in the 1960s, when Lipset and Rokkan were writing, and accelerating in
the 1990s, when Bartolini and Mair made their contribution, interrelated processes of
rising material prosperity, globalization, deindustrialization, technological innovation
and transnational integration contributed to a decline in the size and political importance
of the industrial class-based and religious groups that had long dominated politics in
Europe. Church attendance fell, as did the electoral prospects of religious parties. The
middle class expanded, collective bargaining declined and many manufacturing jobs
were automated or shored off to lower-wage countries, ushering in a gradual demise of
social-democratic parties.5
Many interpreted the declining electoral strength of these specific groups as a broader
trend towards "dealigned" politics, where cohesive social groups in general had come
to matter less for the structure of political competition (Dalton and Flanagan 1984;
Franklin 1992; Kitschelt and Rehm 2015). According to this perspective, electoral
competition has become fundamentally volatile. Many voters, no longer anchored in
political space through group affiliation, now follow short-term impulses when deter-
mining their party allegiance. The result is an erratic, unpredictable electoral environ-
ment of ever-changing issues, events, parties and party leaders constantly pulling voters
5There has also been a debate in the US literature on the declining social basis of partisanship, which,
as far as I can tell, largely ended in the 90s (Clark and Lipset 1991; Hout et al. 1993; Manza and Brooks
1999)
8
in different directions (Dalton and Flanagan 1984; Franklin 1992; Dassonneville 2022;
De Vries and Hobolt 2020).
The perspective of dealigned politics is alimented by the declining importance of
formal organizations in creating collectives out of individuals with some similarities.
In the past, alliances of trade unions, churches or business organizations and political
parties actively tried to "encapsulate" voters. A classic example of encapsulation is
the Dutch system of verzuiling (pillarization), where Protestant, Catholic and labour
movements in combination with political parties sought to embed prospective voters
in a vast web of affiliated organizations. As a result, sizable proportions of Catholics,
Protestants and workers spent their lives, from cradle to grave, virtually isolated from
one another (Lijphart 1968; Lipset and Rokkan 1967; Rokkan 1977). Mutual insulation
gave these cleavages much of their power, fostering group cohesion and creating loyal
voters. However, secularization and trade union decline severed these strong bonds
between class-based and religious groups, affiliated organizations and political parties
(Dassonneville 2022; Katz and Mair 1994; Lijphart 1968).
Despite general agreement that industrial class and religion now matter less for vot-
ing behaviour, it is debated whether political competition as a whole has indeed become
less rooted in social structure, and therefore less predictable (Kitschelt and Rehm 2015).
An influential counterargument holds that new social groups have emerged out of the
postindustrial knowledge economy, and that these groups increasingly structure political
competition – mainly on sociocultural issues –, between new left and far right parties
(Kriesi, Grande, et al. 2006; Hooghe and Marks 2018). Especially citizens with and
without tertiary education levels are divided on these issues, and increasingly vote for
these parties. Some have taken this as evidence for an "education cleavage" (Stubager
2008; Bovens and Wille 2017).
9
The question that arises is what may be shaping cohesive collectives out of voters
with similar education-based experiences and opportunities in the postindustrial knowl-
edge economy, now that the structuring power of formal organizations has diminished.
Answering this question would shed some light on whether socio-cultural conflict is
durably rooted in social structure, and will therefore remain a feature of political com-
petition in the decades to come.
A second, related question concerns the strength and subjective meaning of these
new group identities. Structural roots, collective interests and experiences, repeated
interactions and rituals, and organizational embeddedness imbue social categories with
meaning and inspire group-based political action. Social categories alone, however, are
not always subjectively meaningful and do not always hold political relevance.
In the literature on political polarization, social group identities have come to be
studied predominantly as feelings of intensity towards social categories (e.g. college-
educated, Republican, Protestant, etc) (Huddy 2001; Zuckerman 2005).6This focus
is reflected in the immense popularity of social identity theory and its minimal group
paradigm in political science. In a series of groundbreaking experiments, Tajfel, Turner
and others found that experimentally assigning individuals to groups based on irrelevant
characteristics – eye colour, art preference, or simply at random – sufficed to induce in-
group favoritism and outgroup discrimination (Tajfel et al. 1979). These findings are
often taken to mean that in-group identification can be easily brought about under mini-
mal conditions, in the absence of collective interests and experiences. Despite criticism
6In the introduction to the book The Social Logic Of Politics, Zuckerman discusses the intellectual
history of this development in the US. He traces it back to the prominence of the Michigan school – which
was the first to use abstract categories to measure group identification – and the subsequent popularity of
rational choice models of voting, which dispensed with social groups as a whole (Campbell 1960; Downs
1957; Zuckerman 2005).
10
that identities created in experimental settings are often weak or non-existent in ev-
eryday life, and that there is a crucial distinction between belonging to an objective
social category and internalizing its meaning (Huddy 2001), social categories became
the dominant proxy for social group belonging in the study of group-based politics.
The study of social categories instead of subjectively meaningful group identities is
especially visible in the study of partisan animosity. Much of this literature takes social
identity theory as its point of departure in explaining in-partisan attachment and out-
partisan hostility (Iyengar, Sood, et al. 2012; Iyengar, Lelkes, et al. 2019).7Some argue
that partisanship has become a "mega-identity, that renders opposing partisans different
from, even incomprehensible to each other" (Mason 2018; Finkel et al. 2020). If this is
indeed the case, it means that US citizens may still be cross-cut between different social
identities, but that these other social identities no longer foster moderation because they
are subordinated to their partisan identities. The consequence would be a split of US
society into two antagonistic camps.
These claims are largely based on survey items and survey or field experiments in
which people indeed express out-partisan dislike as well as a clear willingness to avoid
out-partisans. However, these items ask about feelings towards abstract social cate-
gories. As a result, it remains unclear whether partisanship is a meaningful social iden-
tity for people, indeed informing real world relationship formation and discrimination.
This is an important question, because widespread mutual avoidance based primarily on
partisanship may indeed cause significant social division (Baldassarri and Page 2021;
Wagner 2024).
7Though others emphasize that partisan animosity is also related to policy disagreement (Orr and
Huber 2020; Dias and Lelkes 2021)
11
1.1.3 A Relational Approach: Studying Relationship Formation,
Social Network Composition, and Network Influence to As-
sess Group Cohesion and Identification
To assess the durability of sociocultural polarization, and the severity of partisan ani-
mosity, we thus need to understand what is shaping cohesive collectives out of similar
individuals in the absence of formal organizations, as well as the subjective importance
that people attach to their educational and partisan group affiliations. This thesis stud-
ies these questions by adopting a relational approach, looking at relationship formation
across educational and partisan lines, the educational and partisan composition of so-
cial networks, and the role played by social network composition in exacerbating or
moderating group-based political division.
A relational approach can, firstly, provide an answer to the question of what is cre-
ating cohesive collectives out of citizens with similar education levels. We propose that
educationally homogeneous social networks strengthen and reinforce initial education-
based distinctions stemming from social background, experiences and material opportu-
nities, thereby creating cohesive groups durably anchored in political space. In an indi-
vidualistic, mobile society and a postindustrial economy, network homogeneity may be
driven less by top-down, deliberate organizational encapsulation, and more by bottom-
up, organic processes related to material opportunities and interaction preferences.
Second, a relational approach can help assess the subjective meaning people attach
to their educational and partisan identities. In societies where people exercise substantial
choice in shaping their social environment, in changing their identities or the salience
thereof (Giddens 1991; Bauman 2013), subjectively meaningful identities should be
reflected both in the composition of real-world relationships and in interaction prefer-
12
ences. Closely studying the role of partisanship and education in everyday processes of
relationship formation can reveal whether these are meaningful social identities.
This wider approach builds on a rich tradition in political sociology that studies
individual political behaviour as a function of the social context in which citizens are
embedded (Berelson et al. 1954; Lazarsfeld et al. 1944; Huckfeldt and Sprague 1995;
Zuckerman 2005; Baldassarri 2007; Ternullo 2022). This tradition has been overshad-
owed by atomistic approaches to political behaviour, and has not been applied much
in the political polarization literature. In addition, most of this work is concerned with
the partisanship and political attitudes of social relations, and whether and how they
affect individual attitudes and voting behaviour (Lazarsfeld et al. 1944; Huckfeldt and
Sprague 1995; Beck et al. 2002; Lazer et al. 2010; Foos and De Rooij 2017). In contrast,
the chief concern of this thesis is the social (specifically: educational) composition of
people’s relationships, which we think informs partisanship and political attitudes in the
first place. In addition, and closer to existing work in the US, we study whether hetero-
geneous partisan ties still affect political behaviour in the contemporary hyper-partisan
US political context.
1.2 Thesis overview
1.2.1 An Education Cleavage? Social Closure by Education Levels
and Entrenching Sociocultural Polarization
Over the course of recent decades, sociocultural issues have come to divide electorates
throughout Western democracies (Hall et al. 2023). Issues like immigration, transna-
13
tional integration and cultural liberalism feature increasingly prominently in party plat-
forms, debates and the media. New left and far right parties, who mobilize predomi-
nantly on these issues, experience increasing electoral success, forcing mainstream par-
ties to reposition themselves, often adopting stances similar to those of their competitors
on the flanks (Abou-Chadi and Krause 2020).
Education levels, more than any other socio-structural variable, are strongly predic-
tive of people’s positions on these issues, and their likelihood to vote for these parties
(Abou-Chadi and Hix 2021; Gethin et al. 2022) in some cases increasingly so (Marks
et al. 2023). In addition, there is clear evidence that tertiary-educated citizens’ distinct
social backgrounds (Lancee and Sarrasin 2015), distinct educational experiences (Scott
2022) and distinct material opportunities (Kriesi, Grande, et al. 2008) induce a pro-
gressive worldview. These patterns are stable across countries with different historical
legacies, political institutions and types of market economies. Differences are becom-
ing so pronounced that several scholars have started speaking of an education cleavage
(Stubager 2008; Bovens and Wille 2017).
Chapter 2 proposes that educationally homogeneous social networks are durably
anchoring citizens with and without tertiary education in political space. The large-
scale technological and economic developments that set in motion the rise of sociocul-
tural conflict in the first place, also contributed to social separation between educational
groups. The tertiary-educated increased in size, rising to between 35-50% of West-
ern electorates (OECD 2019), a sufficiently sizable social group to form the basis for a
cleavage (Posner 2004). Moreover, they increasingly clustered in cities, neighbourhoods
and places of work and learning, partnering up with similarly-educated others (Kalmijn
1998; Mijs and Roe 2021), and potentially creating relatively homogeneous close re-
lationships (Smith et al. 2014). We hypothesize that these educationally homogeneous
14
social relationships serve to strengthen and reinforce initial distinctions between citizens
with and without tertiary education (see also (Zollinger and Attewell 2023; Bornschier,
Haffert, et al. 2024)).
We test these expectations with the use of ego-network data from the Netherlands
and the US, and measures of family and partner education levels in twelve European
countries. We find that social networks in the Netherlands exhibit substantial levels of
educational homophily and that especially tertiary-educated citizens express clear pref-
erences to form relationships with similarly-educated others. Higher (lower)-educated
citizens embedded in educationally homogeneous social networks, in turn, hold more
progressive (conservative) attitudes and vote in higher numbers for green and social-
liberal (or far right) parties, compared to citizens in heterogeneous networks. Moreover,
we find that a change in network education levels is predictive of a change in attitudes
and voting, suggesting that homogeneous networks indeed reinforce initial distinctions,
and heterogeneous networks moderate them. These findings show the existence of siz-
able, distinct and insulated educational groups who express significant support for par-
ties and issues on opposite sides of the sociocultural divide. They suggest that socio-
cultural conflict between new left and far right parties remains durably, and predictable
rooted in social structure, and will be here to stay for the foreseeable future.
Chapter 3 builds on and extends these findings by studying the buffering role of
heterogeneous social relationships during a crucial period of cleavage formation: the
rise of far right parties during the Great Recession. Cleavages crystallize gradually but
crises can catalyze the electoral success of especially the far right, mostly in Northern
and Western Europe (Hutter and Kriesi 2020; Scheiring et al. 2024). Amidst heightened
insecurity, the far right’s particular anti-immigrant, anti-establishment message is well-
15
received among especially structurally vulnerable citizens (i.e. those without tertiary
education, or those engaging in manual labour) (Mudde 2007; Damhuis 2020). How-
ever, citizens make sense of the world, and changes therein, in conjunction with others,
and citizen-level discussions can buffer the effect of elite-level messages (Berelson et
al. 1954; Druckman and Nelson 2003). We expect that presence in the network of less
structurally vulnerable individuals, who have different crises experiences and different
pre-existing attitudes, will moderate the relationship between radical right appeal and
economic uncertainty.
Following individuals during the first years of the Great Recession in the Nether-
lands – a period of widespread economic insecurity which also saw the surge of the far
right Party for Freedom – we find that structurally vulnerable individuals embedded in
heterogeneous close-tie networks, or active in bridging voluntary organizations, are less
likely to respond to increased economic insecurity by coming to vote for the far right.
Disentangling influence and economic mechanisms, we find no evidence that citizens in
homogeneous or heterogeneous networks were differently affected by the crisis. Rather,
informal discussions with close ties explain part of the effect.
1.2.2 The Primacy of Partisanship? Cross-Partisan Relationships
in an Era of Partisan Animosity
Growing educational divides may still constitute politics as usual: new parties represent-
ing the interests of new social groups. In contrast, US partisan animosity is often said
to be dire, with prominent articles not shying away from describing the phenomenon
as "fear and loathing" (Iyengar and Westwood 2015; Reiljan 2020) between "warring
tribes" (Mason 2018) with "dark consequences" (Finkel et al. 2020).
16
At the same time, political divides have always had ideological and affective ele-
ments (Schedler 2023; Häusermann and Bornschier 2023; Wagner 2024) and past stud-
ies which use similar measures do not interpret expressions of partisan animosity as
necessarily dangerous (Richardson 1991).8
What is potentially worrying in the US today, is the suggestion that partisan iden-
tity has eclipsed other social identities in shaping self-identification as well as out-
group discrimination and avoidance. In survey-experimental settings that ask respon-
dents to choose between dating profiles (Huber and Malhotra 2017), roommate pro-
files (Shafranek 2021), and residential neighborhoods (Mummolo and Nall 2017), they
strongly prefer co-partisans over out-partisans, even when other social identities are con-
trolled for. People also discriminate against out-partisans in behavioural games (West-
wood et al. 2018), economic decisions (McConnell et al. 2018) and hiring decisions
(Gift and Gift 2015).
Many take these findings to mean that "party identity has in many cases come to
dominate social cleavages" (Westwood et al. 2018 pp. 334) and that "partisanship may
loom largest (emphasis in original) as a consideration (in determining social selection)"
(Shafranek 2021 pp. 288). If this is true, contemporary US society may be exiting the
world of crosscutting cleavages inducing moderation. While citizens still hold multiple,
crosscutting social identities, these are subordinated to their partisanship in informing
social behaviour and the formation of social relationships. As a result, society may split
into opposite camps and widespread partisan division would be inevitable (Baldassarri
and Page 2021). Alternatively, the study of partisanship as a social category, instead of
a subjectively meaningful identity, overstates the importance that people attach to it in
social life. This is one of the claims we are making in Chapter 4.
8Though see (Almond, Verba, et al. 1963 pp. 123 and 134-35)
17
In this chapter, we study the extent to which partisanship informs the formation of
social relationships as well as the potential buffering effect of heterogeneous relation-
ships on out-partisan animosity. In our view, the many studies that manipulate par-
tisanship in survey settings, find that respondents express hostility or avoidance, and
take these findings as evidence of real world avoidance, miss two important aspects of
real world relationship formation. Unlike in survey settings, partisanship is often not a
visible trait in early stages of relationship formation. Instead, it usually emerges only
after individuals have initially bonded over shared experiences or common characteris-
tics (Minozzi et al. 2020), and by the time it does, its impact on established relation-
ships is likely milder than at their outset (Balietti et al. 2021). Second, when thinking
about Republicans and Democrats in abstract survey settings, people imagine stereotyp-
ical, ideologically extreme, politicized others (Ahler and Sood 2018; Druckman, Klar,
et al. 2022). Real world encounters, in contrast, are usually with political moderates,
which will likely lead to less extreme reactions. In this chapter, we therefore deploy
survey-experimental settings that take into account these core aspects of the process of
relationship formation. Namely, we study encounters when partisanship is known and
unknown, and with political moderates and extremists.
We find that people do not look for partisanship in early encounters. When they
are not explicitly provided with a partisan label, partisanship is subordinated to a range
of socio-demographics and character traits. In contrast, when we provide respondents
with partisan labels in selection experiments, we find – just like in most other stud-
ies – that people strongly prefer co-partisans to out-partisans. However, these effects
become much less pronounced when people are presented with a moderate (counter-
stereotypical) out-partisan, compared to a strong (pro-stereotypical) out-partisan. We
provide additional suggestive evidence that people are often unaware of partisanship
18
in real-life encounters. Moreover, using several measures of social networks, we find
that social relationships in US society retain significant levels of partisan heterogeneity
(Lee, Lee, et al. 2023). These findings suggest that partisanship is indeed subordinated
to other considerations in informing relational patterns. Expressed willingness to avoid
out-partisans, captured by survey questions, should not be taken as evidence for real-
world avoidance. The implication is that partisanship has not yet taken over people’s
social life, and mass-level partisan division may be more muted than often assumed.
In fact, we show that the partisan heterogeneity of people’s social environments
works to buffer the widespread elite and media polarization that people are increasingly
subjected to. We advance two mechanisms for this contention. First, in line with canoni-
cal literature, we show that politically heterogeneous close relationships reduce hostility
through cross-partisan discussions (Mutz 2002; Huckfeldt, Mendez, et al. 2004). In ad-
dition, we demonstrate that politically diverse acquaintances serve to counter the kind
of distorted out-partisan perceptions that partly drive political animosity, and thereby
inoculate against partisan hostility.
Chapter 5, finally, is a stand-alone paper, which strengthens some of the contentions
put forward in Chapter 4. I run a cross-partisan discussion experiment in the UK where
I invite 300 pairs of Labour and Conservative supporters to briefly discuss a statement
on immigration or redistribution in an online chatroom. Results show that the discus-
sion immediately reduces affective polarization, durably increases willingness to discuss
across partisan lines, but does not change opinions. In line with findings from Chapter 4,
part of these effects seem to be driven by reduced misperceptions. Treatment effects are
larger for respondents randomly matched to out-partisans with a counter-stereotypical,
or moderate opinion (e.g. a Conservative voter in favour of immigration), and disappear
19
when opinions are too far apart. In addition, in an open-response question, about one in
five respondents indicated to have expected an ideologically extreme partner, or a heated
discussion. To their surprise, they and their partner agreed on the issue, and the discus-
sion was civil. For these respondents, treatment effects are much larger. These findings
suggest that even brief contact experiences can reduce animosity. The chat platform is
scalable, having been used by over 100.000 Dutch citizens by now.
These results suggest the existence and potential growth of societally entrenched
educational divides, but more muted mass-level partisan conflict. They underline the
importance of a relational approach to polarization to understand the endurance, and
potential escalation, of group-based political divides.
1.3 Data and Methods
Several chapters of the thesis are concerned with network effects. We expect that the
educational and partisan composition of people’s social relationships affects their po-
litical attitudes and behaviour. Determining network effects is complicated by people
self-selecting into networks. We use several empirical strategies to mitigate this con-
cern.
Past work that considered the role of formal organizations in shaping group co-
hesiveness studied the association between union membership and voting for the left
(Berelson et al. 1954; Bartolini and Mair 1990; Dassonneville 2022). Analogously, we
use cross-sectional models that regress the political outcomes of interest on the educa-
tional and partisan composition of various types of networks, using conservative sets of
control variables. Moreover, we use the only longer-running, nationally representative
20
panel dataset that contains an established measure of close-tie ego networks (Marsden
1987). This allows us to estimate the effects of changes in network composition, and to
follow individuals embedded in different networks over time, to see how they respond
politically to economic change. Finally, other chapters study real-world interactions,
or experimentally manipulate the salience of existing networks. While especially the
within-individual analyses represents a real methodological improvement of past work
on social environments and politics (Bramoullé et al. 2020), results should still be in-
terpreted with caution as social networks can evolve endogenously and individual out-
comes can be influenced by both individual and peers’ time-invariant characteristics.
In addition to network effects, several chapters study educational and partisan net-
work composition, in combination with interaction preferences, as indications of mean-
ingful subjective identities. Existing studies of educational identification use three types
of measures: attitudes towards people with similar and different education levels (cold,
warm, close, not close), perceived social distance (e.g. "would you mind if someone
with a different education level moved into your neighbourhood/married your daugh-
ter") and behaviour towards them in a behavioural game. They show that perceived
social distance, and behaviour in a trust game do not differ based on education (Hel-
bling and Jungkunz 2020). When it comes to warm/cold feelings, tertiary-educated
respondents give higher scores to similarly-educated others, but still largely positive
scores to differently-educated others (Kuppens, Spears, et al. 2018). The same goes for
reported feelings of closeness: respondents report to feel closer to their own educational
group (low, middle, high), but overall feel reasonably close to all educational groups
(Zollinger and Attewell 2023; Bornschier, Haffert, et al. 2024). These studies thus show
some evidence for the existence of education-based identities but effects seem small.
In the measures discussed above, social desirability can be an issue: people may feel
21
the need to report close feelings to all groups, which could explain the relatively small
differences between in and outgroup rating in some of the studies mentioned above.
Second, it is necessary to determine the relative weight that people attach to their poten-
tially many group affiliations, to see which one people prioritize in social interactions.
Finally and more generally, intergroup attitudes are often only weakly predictive of in-
tergroup behaviour (Brauer 2024).
To deal with these concerns, I use more unobtrusive measures of social networks
that first ask respondents to name a certain number of people "with whom they discuss
important issues" and only then ask for these people’s characteristics (e.g. education
level, partisanship, relationship, closeness, etc) (Marsden 1987). In sociology, these
"name generators" are commonly used as an indicator of social division/cohesion, in-
tegration/segregation and interaction preferences (McPherson and Smith-Lovin 1987;
DiPrete et al. 2011; Smith et al. 2014).
However, social networks can be homogeneous as a result of structure, choice, or
both (Homans 1950; Feld 1981; McPherson, Smith-Lovin, and Cook 2001). Therefore,
I study network composition in combination with interaction preferences. We present
respondents with a fictitious relationship (a partner in Chapter 2; an officemate in Chap-
ter 4), and ask them what aspects of this person they would like to know more about
before deciding whether they select this person. Then, they are presented with pairs
of characteristics, randomly drawn from a list, and asked to choose the one they find
more important. This design thus assesses the relative salience of certain group affilia-
tions. In Chapter 4, where we are interested in the relative importance people attach to
partisanship in social selection, the pairwise comparison design means to mimic a situ-
ation in which partisanship is unknown, to see whether people are actively looking for it.
22
The questions posed in Chapters 2 and 3 are mostly studied in the Netherlands. The
Netherlands is a typical case of electoral realignment, with strong past class and espe-
cially religious divides slowly being supplanted by education, place of residence and
postindustrial class categories. The country has a sizable tertiary-educated population
and clear electoral mobilization of sociocultural issues by long-standing green, social-
liberal and far right parties (Kriesi, Grande, et al. 2006, pp. 936). For external validity,
we run further analyses using European Social Survey and original data collected in the
US. The comparison between the US and the Netherlands fits especially the question
of the effect of educational network composition on attitudes and voting. The countries
differ in important ways that may confound this relationship (e.g. political institutions,
population diversity, segregation patterns, inequality, public/private ownership, educa-
tion system, heterogeneity of the college experience), but have been similarly affected
by the structural transformations that set in motion educational divides (Iversen and
Soskice 2020).
In Chapter 4, we use original data collected in the US, an extreme case of partisan
animosity and thus a most likely case to find that partisanship has overtaken other con-
siderations in informing relationships, and a least likely case to find that heterogeneous
relationships buffer animosity. Chapter 5 then moves to the UK for the cross-partisan
discussion experiment. Like the US, the UK has reasonably high levels of inter-partisan
animosity (Westwood et al. 2018) and essentially a two-party system.
23
.
24
Chapter 2
Separated by Degrees: Social Closure
by Education Levels Strengthens
Contemporary Political Divides
Abstract Across Western democracies, education levels are predictive of immigration
attitudes and voting for new left or far