Elections and Democracy

The Comparative Study of Electoral Systems (CSES) is a collaborative program of

research among election study teams from around the world. Participating coun-
tries include a common module of survey questions in their post-election studies.
The resulting data are deposited along with voting, demographic, district, and
macro variables. The studies are then merged into a single, free, public dataset for
use in comparative study and cross-level analysis.
The set of volumes in this series is based on these CSES modules, and the vol-
umes address the key theoretical issues and empirical debates in the study of elec-
tions and representative democracy. Some of the volumes will be organized
around the theoretical issues raised by a particular module, while others will be
thematic in their focus. Taken together, these volumes will provide a rigorous and
ongoing contribution to understanding the expansion and consolidation of
democracy in the twenty-first century.

Further information on CSES activities can be obtained from:

CSES Secretariat
Center for Political Studies
Institute for Social Research
The University of Michigan
426 Thompson Street
Ann Arbor, Michigan 481042321
USA

CSES web site: http://www.cses.org

Elections and Democracy
Representation and Accountability

Edited by Jacques Thomassen

1
1
Great Clarendon Street, Oxford, OX2 6DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© The several contributors 2014
The moral rights of the author have been asserted
First Edition published in 2014
Impression: 1
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
198 Madison Avenue, New York, NY 10016, United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 2014930995
ISBN 978–0–19–871633–4
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR0 4YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Series Editors’ Preface

Few topics generate as much interest among observers and practitioners of

politics as the quality of the democratic process. The expansion of democracy
during the twentieth century, which accelerated rapidly after the collapse of
communism in 1990, has meant that a majority of the world’s countries are
now electoral democracies. But not all democracies can be considered equal;
they differ widely in terms of institutional arrangements and practices and in
the levels of public support that they attract. It is the public support for
democracy that the Comparative Study of Electoral Systems (CSES) project is
designed to investigate. This volume series presents the key findings from this
major research project that commenced in 1994.
The first CSES volume, edited by Hans-Dieter Klingemann, has document-
ed much of its historical background, the basic principles of data collection,
and provided sample chapters showing many of the analytical possibilities of
this unique data collection. This volume is based on the first module of sur-
vey questions in the CSES, completed in 2001, which examines the interac-
tion of political institutions and political behaviour regarding attitudes
towards the democratic regime, the political authorities, and the quality of
the political process generally.
As with the first volume, the second one, edited by Russell J. Dalton and
Christopher J. Anderson, addresses the fundamental question of whether the
institutional structure of elections affects the nature of the public’s choices.
The first question looks at explanations of turnout, and how institutions
structure the likelihood of voting. The second question discusses determi-
nants of individual electoral behaviour and examines the role of institutions
in shaping what kinds of political information voters acquire.
The third volume in the series, by Russell J. Dalton, David Farrell, and Ian
McAllister, describes and explains the role of political parties in election cam-
paigns, in forming the electoral choice of voters and their role in government
and opposition. The theoretical arguments relate to the logic of the responsi-
ble parties model. It is the first study that tests these ideas using a comprehen-
sive and comparative design. It demonstrates the importance of the left–right
schema to enable political positioning, political communication, and politi-
cal representation.

v
Series Editors’ Preface

Elections and Democracy: Representation and Accountability, edited by Jacques

Thomassen, constitutes the fourth volume of the series. Inspired by Arend
Lijphart’s theory, the analyses contrast voting and elections in countries of
majoritarian Westminster and consensus democracies. The volume asks the
question: do consensus models of democracy serve the interests of their citi-
zens better than majoritarian systems? The answer is that formal institutions
such as the electoral system matter less than might be expected. What does
matter is the characteristics of the party system, such as the level of polariza-
tion and the clarity of accountability.
Jack Vowles and Georgios Xezonakis have edited the fifth volume, to follow
soon. The authors ask how globalization affects democratic mass politics, and
in particular the political attitudes and behaviour of ordinary citizens and the
policies of political parties—not just governments.
All of the books in the series raise key questions for extending our under-
standing of individual citizen behaviour. Most studies of voting behaviour
have been based on single country studies, often covering just a single elec-
tion. By comparing a wide range of countries, for the first time the CSES pro-
ject enables the institutional environment to be brought into the equation,
enhancing our understanding of the complex relationship between individ-
ual choice and institutional context. Indeed, such analyses were impossible
until the CSES was established.
All of the CSES data are freely available and can be downloaded from our
website <http://www.cses.org>.
Hans-Dieter Klingemann
Ian McAllister
Series Editors

vi
Preface

This book is based upon the Comparative Study of Electoral Systems (CSES). The
most innovative characteristic of CSES is that it collects data from a great
number of countries both at the country level and at the level of individual
citizens. This makes it possible to study the effects of different institutional
arrangements on people’s political behaviour and their perception and evalu-
ation of the democratic process. It hardly needs to be argued how important
it is to know what these effects are. There is a growing body of literature argu-
ing that elections are no longer an effective instrument of democracy as they
are failing to connect the policy preferences of the electorate to government
policy. Also, people allegedly are losing their confidence in politicians, politi-
cal parties, and political institutions. Against this background, trying to find
out which kind of political-institutional arrangements serve electoral democ-
racy best and which institutions are better able to sustain people’s confidence
in them is like a quest for the Holy Grail.
Of course, the political institutions of different countries can differ in many
ways. Fortunately, there is some structure in these differences. Following
Lijphart’s seminal work Patterns of Democracy we make a distinction between
majoritarian and consensus models of democracy. These models are the embod-
iment in institutional structures of two different views of democracy. In the
majoritarian view the single most important function of an election is the
selection of a government. The concentration of power in the hands of an
elected majority government makes it accountable to the people. In consen-
sus models of democracy, or proportional systems, the major function of elec-
tions is to elect the members of parliament who together should be as
representative as possible of the electorate as a whole. The criterion for the
democratic quality of the system is how representative parliament really is.
This book explores how far these different views and their embodiment in
institutional structures influence vote choice, political participation, and sat-
isfaction with the functioning of democracy.
Unfortunately, we did not find the Holy Grail. Our major finding is that
formal political institutions and in particular the distinction between a con-
sensus and a majoritarian system of democracy are far less relevant for peo-
ple’s political behaviour and their perceptions and evaluations of the process

vii
Preface

of democracy than often presumed. Rather than formal political institutions

like the electoral system it seems to be the characteristics of the party system
like polarization and the clarity of responsibility that really matter. This
might be a disappointing message for institutional engineers: whatever the
cure for a poorly functioning democracy or people’s dissatisfaction with the
functioning of democracy is, a makeover of political institutions in all likeli-
hood will not make much of a difference. It is politics that really matters.
This book is the fourth volume in a series of books based on CSES. CSES has
evolved into a collaborative, cross-national programme of electoral behav-
iour among over 60 election study teams from around the world. The first
ideas for bringing together the principal investigators of national election
studies in order to develop a comparative programme of election studies were
developed in the context of the International Consortium for Research into
Elections and Representative Democracy (ICORE) which was founded at the
ECPR Research Sessions in Rimini in 1989. This at first was an organization of
the established election studies in Western Europe, which decided to focus on
the development of a longitudinal common database of European Election
Studies at the Central Archive (now GESIS) in Cologne. Later on this led to
the publication of The European Voter. But ICORE initially did not develop any
plans for a worldwide comparative programme of election studies. It was Ste-
ven Rosenstone, the then principal investigator of the American National
Election Studies who first made the ANES member of ICORE in 1993 and then
immediately took the initiative for CSES. The idea for it was born in the con-
text of ANES, but Steven Rosenstone wanted it to be organized under the
umbrella of ICORE, being afraid that researchers in many parts of the world
would otherwise shy away from what they might consider as American intel-
lectual imperialism. That concern was totally unjustified. Being affiliated
with the famous election studies at the University of Michigan was an impor-
tant incentive for many of them to join. Without Steven Rosenstone CSES
would not have seen the light of day, at least not at any time soon. Co-­
directing the first planning committee with him was an adventurous but
pleasant experience.
The authors of this book are indebted to many people and institutions.
First, we are standing on the shoulders of giants in our discipline, Arend
Lijphart’s in particular. His distinction between majoritarian and consensus
models of democracy is the major independent variable in this book. His
measures were updated and somewhat amended for the countries participat-
ing in CSES by Julian Bernauer, Nathalie Giger, and Adrian Vatter. The second
module of CSES on which this book is almost entirely based was developed by
the second planning committee under the benevolent leadership of Phil
Shively. Dave Howell of the CSES secretariat has been supportive throughout
the project. The Institute of Governance Studies at the University of Twente

viii
 Preface

hosted and funded a research meeting at which first drafts of the chapters
were discussed. Janine van der Woude transformed the chapter drafts into a
single style manuscript. Dominic Byatt of Oxford University Press and the
series editors, Ian McAllister and Hans-Dieter Klingemann, have been sup-
portive and patient from the very beginning. As always, the staff at OUP did
a wonderful job preparing the final manuscript. I’m especially grateful to my
co-authors for their patience and perseverance. They never complained about
yet another revision, at least not in my presence.
Enschede, February 2014
Jacques Thomassen

ix
Contents

List of Figures xiii

List of Tables xv
Contributors xviii

1. Representation and Accountability 1

Jacques Thomassen

2. New Patterns of Democracy in the Countries

of the Comparative Study of Electoral Systems 2 20
Julian Bernauer, Nathalie Giger, and Adrian Vatter

3. Meaningful Choices: Does Parties’ Supply Matter? 38

Bernhard Wessels and Hermann Schmitt

4. Policy-based Voting and the Type of Democracy 60

Diana Burlacu and Gábor Tóka

5. Political Institutions and the Social Anchoring of the Vote 79

Pedro C. Magalhães

6. Political Institutions, Perceptions of Representation,

and the Turnout Decision 99
André Blais, Shane Singh, and Delia Dumitrescu

7. Democratic Structures and Democratic Participation:

The Limits of Consensualism Theory 113
Steven Weldon and Russell Dalton

8. Feeling Policy Represented 132

Sören Holmberg

9. Output-oriented Legitimacy: Individual- and System-level

Influences on Democracy Satisfaction 153
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

10. The Multiple Bases of Democratic Support: Procedural

Representation and Governmental Outputs 181
Mark Peffley and Robert Rohrschneider

xi
Contents

11. Globalization, Representation, and Attitudes towards Democracy  201

Kees Aarts, Jacques Thomassen, and Carolien van Ham

12. Consenting to Lose or Expecting to Win? Inter-temporal Changes

in Voters’ Winner–Loser Status and Satisfaction with Democracy  232
Eric Chang, Yun-han Chu, and Wen-chin Wu

References 255
Index 275

xii
List of Figures

1.1 The CSES research design  6

2.1 Map of democracy using indices of variables loading on the
dimensions of consensus democracy (two-dimensional solution)  31
2.2 Map of democracy using indices of variables loading on the
dimensions of consensus democracy (three-dimensional solution)  32
3.1 Clarity of policy positions of political parties: R , regression
2

of perception of parties’ left–right positions on 14 factor scales

from election manifestos  51
3.2 Marginal effect of proximity on vote choice conditioned
by the degree of clarity of parties’ policy positions  56
3.3 Clarity of policy positions of political parties and mean proximity
to the party voted for  57
4.1 The effect of changes in policy demand and performance
evaluations on the vote share of UK parties in 2005  68
4.2 The effect of changes in policy demand and performance
evaluations on the vote share of Swiss parties in 2003  68
4.3 The party-level effects of changing policy demand and
performance evaluations aggregated with Pedersen index  70
5.1 Plots of structural voting indices (y-axis) against consensual
democracy (executives–parties) index (x-axis), with linear regression
fit lines  93
5.2 Average values of structural voting indices for presidential systems
(1) and parliamentary or semi-presidential systems (0). Error bars
are 90% confidence intervals  94
6.1 Turnout across elections  104
8.1 Policy representation in Sweden, 1969–2006  143
9.1a Average democracy satisfaction in all EU countries, 1976–2006  155
9.1b Average democracy satisfaction in (six) founder EU member states
and (three) 1970s joiners, 1976–2006  155
9.2 Average satisfaction with democracy on 1–4 scale, CSES Wave 2
countries  157

xiii
List of Figures

A.9.1 Variations in democracy satisfaction in Western European countries,

1976–2006  176
10.1 Conceptualizing democratic evaluations  185
10.2 Mean democratic support indicators by country  186
10.3a–c Institutional quality and procedural evaluations of national systems
(mean scores on y-axis)  193
11.1 Trends in political support by socio-economic status  208
11.2 The causal connection between socio-economic status
and political support  208
11.3 Trends in satisfaction with democracy by education in nine European
democracies  216
11.4 Level of completed education, the Netherlands, 1971–2010  218
11.5 Social class self-image, 1971–2010  218
11.6 Level of education and attitudes towards politics  219
11.7 Social class self-image and attitudes towards politics  220
12.1 Predicted probabilities with 95 per cent confidence intervals of being
satisfied with democracies for losers and winners between young and
developed democracies  249

xiv
List of Tables

2.1 Factor loadings of originally ten political-institutional variables

in 34 countries covered by the CSES 2, 1997–2006  29
2.2 Factor loadings of originally 11 political-institutional variables
in 34 countries covered by the CSES 2, 1997–2006  29
A.2.1 Eleven political-institutional variables, their measurement,
and data sources  35
3.1 Factor analysis of election programmes of political parties
in 23 countries  47
3.2 Regression of individually perceived left–right positions of political
parties on factor scores of the dimensional analysis of 114 election
programmes in 23 countries  49
3.3 Conditional logistic regressions: base model and cross-level
interactions  54
3.4 Correlation between macro characteristics (Consensus scale and
clarity) and the change in vote probability by an increase in
proximity  56
4.1 The estimated impact of policy preferences and impact
of performance evaluations across 42 political contexts  71
4.2 Multivariate regressions of the impact of policy preferences
on political system characteristics  73
5.1 Indices of structural voting (SV) in 34 elections  91
5.2 The determinants of structural voting  95
5.3 Marginal effects  96
6.1 Voting and feelings of representation  105
6.2 Voting and feelings of representation; logit estimation  106
6.3 Feelings of representation and political institutions  107
6.4 Voting and institutions; multilevel logit estimation  109
6.5 Voting, institutions, and feelings of representation; multilevel logit
estimation  110
7.1 Levels of political participation  119
7.2 Contextual correlates of political participation  121

xv
List of Tables

7.3 Multilevel logit models of political activism  125

A.7.1 CSES survey variables  129
8.1 Citizens in 35 countries judge the level of issue agreement
between voters and elected representatives in their own country  140
8.2 The feeling of being policy represented by a party or a party
leader among citizens in 35 countries  144
8.3 Assessing policy representation among citizens in old
and young democracies  146
8.4 Assessing policy representation among citizens in countries
with different election systems  146
8.5 Feeling policy represented among citizens in countries along
Vatter’s and Bernauer’s executives–parties dimension  147
8.6 Multilevel regression analysis of the impact of individual-
and system-level characteristics on citizen’s perceptions of being policy
represented by elected representatives in their own countries  150
9.1 Distribution of satisfaction with democracy across CSES Wave
2 countries  156
9.2 Descriptive characteristics of key predictors in democracy
satisfaction model  165
9.3 Bivariate correlations between democracy satisfaction
and individual-level predictors  167
9.4 Difference of means tests on democracy satisfaction (1–4) scale  167
9.5 Ordered logit models of democracy satisfaction  169
9.6 Changes in predicted probabilities derived from the model
reported in Table 9.5 (column C) 173
A.9.2 Country scores on regime characteristics used in the chapter  177
10.1 Conceptualizing the sources of representational judgements  187
10.2 The influence of macro-level variables on public evaluations
of democratic performance and ideals  191
10.3 Predicting democratic performance and democratic ideals,
including individual-level controls  192
  10.4 Predicting democratic performance and democratic ideals,
including individual-level controls and individual-level
evaluations of procedural fairness  195
10.5 Predicting procedural evaluations  196
A.10.1 Countries included in analyses  198
A.10.2 Descriptive variable information about predictors (TBD REDO)  199

xvi
List of Tables

11.1 Dimensions of political conflict  204

11.2 Differences in percentage feeling represented by a political
party between contrast groups  213
11.3 Differences in percentage very/fairly satisfied with the democratic
process between contrast groups  215
11.4 Level of education, and position on income differences
and European integration  223
11.5 Social class self-image, and position on income differences
and European integration  224
11.6 Position on income differences and European integration,
and attitudes towards democracy  225
A.11.1 Construction of CSES social class indicators  227
A.11.2 Construction of CSES socio-economic status indicator  227
A.11.3 CSES Module 2 elections included in the analyses  227
B.11.1 Eurobarometer education indicator  228
B.11.2 Eurobarometer sample: countries and time period included  228
C.11.1 Indicator for ‘feelings of representation’ in DPES  229
12.1 Satisfaction with democracy and support for democracy (%)  243
12.2 Distribution of dynamic winner–loser status (%)  244
12.3 Estimated effects of political support  247

xvii
Contributors

Kees Aarts is Professor of Political Science and Scientific Director of the Institute for
Innovation and Governance Studies (IGS) at the University of Twente, the
Netherlands. His research interests are in democracy, elections, and political
behaviour. He received his BA and Master’s degrees from the University of
Amsterdam, and his PhD from the University of Twente.
Julian Bernauer is Postdoctoral Researcher and Lecturer at the University of Berne,
Switzerland. His research interests include political representation, empirical theory
of democracy, and quantitative methodology. Recent publications have appeared in
Comparative Political Studies, the European Journal of Political Research and Electoral
Studies.
André Blais is Professor in the Department of Political Science at the Université de
Montréal and holds a Canada Research Chair in Electoral Studies. He is the principal
investigator of the Making Electoral Democracy Work project and the Chair of the
Planning Committee of the Comparative Study of Electoral Systems. He is Fellow of
the Royal Society of Canada and past President of the Canadian Political Science
Association. He is a member of the Centre for the Study of Democratic Citizenship.
His research interests are elections, electoral systems, turnout, public opinion, and
methodology.
Diana Burlacu is Postdoctoral Researcher at the Humboldt University of Berlin and
former Marie Curie Fellow in the Electoral Democracy Training Network at Central
European University, Budapest. Her research interests are in comparative politics,
with a particular focus on the consequences of the quality of governance on political
behaviour.
Eric Chang is Associate Professor of Political Science at Michigan State University. He
specializes in comparative political economy, political institutions, political
corruption, and democratization.
Yun-han Chu is Distinguished Research Fellow of the Institute of Political Science at
Academia Sinica and Professor of Political Science at National Taiwan University. He
specializes in politics of Greater China, East Asian political economy, and
democratization.
Harold Clarke is Ashbel Smith Professor in the School of Economic, Political and
Policy Sciences at the University of Texas at Dallas, and Adjunct Professor in the
Department of Government at the University of Essex. He is the author of numerous
studies of British, US, and Canadian politics.

xviii
Contributors

Russell Dalton is Professor of Political Science at the University of California, Irvine.

He has received a Fulbright Professorship at the University of Mannheim, a Barbra
Streisand Center Fellowship, German Marshall Research Fellowship, and a POSCO
Fellowship at the East/West Center. He has written or edited over 20 books and 150
research articles in the fields of comparative political behaviour, political parties, and
empirical democratic theory.
Delia Dumitrescu is Postdoctoral Fellow associated to the Multidisciplinary Opinion
and Democracy Group at the University of Gothenburg. She obtained her PhD from
Ohio State and was previously associated to the Canada Research Chair in Electoral
Studies at the University of Montreal. Her interests are in political psychology and
electoral communication.
Nathalie Giger is Postdoctoral Research Associate at the Department of Political
Science, University of Zürich. Her research interests lie in the linkage between
citizens and political elites, in particular in political representation and the electoral
consequences of public policy. Her work has appeared in the European Journal of
Political Research, West European Politics and European Sociological Review, among
others.
Sören Holmberg is Professor Emeritus of Political Science at the University of
Gothenburg, Director of the Swedish National Elections Studies 1979–2010, and co-
founder of the Society Opinion Media Institute (SOM) and the Quality of
Government Institute (QoG). His main research interests are political representation,
voting behaviour, and good government.
Pedro C. Magalhães is a researcher at the Institute of Social Sciences of ULisbon, one
of the Founders of the Portuguese Election Study, and Member of the Planning
Committee of the Comparative Study of Electoral Systems. He specializes in the
study of public opinion and voting behaviour.
Mark Peffley is Professor of Political Science at the University of Kentucky,
Lexington, USA.
Robert Rohrschneider is Sir Robert Worcester Distinguished Professor of Political
Science at the University of Kansas, Lawrence, USA.
David Sanders is Professor of Government and Pro-Vice-Chancellor (Research) at the
University of Essex. He is former Director of the British Election Study and the author
of various books and articles in the fields of British politics and comparative politics.
Hermann Schmitt holds a Chair in Electoral Politics at the University of Manchester
and is Research Fellow of the MZES and Professor at the University of Mannheim. He
received his doctorate from the University of Duisburg, and holds a venia legendi from
the Free University of Berlin and Mannheim University.
Shane Singh is Assistant Professor in the Department of International Affairs at the
University of Georgia. His research focuses on the institutional and contextual
foundations of political behaviour, opinion, and attitudes.
Marianne Stewart is Professor in the School of Economic, Political, and Policy
Sciences at the University of Texas at Dallas. She is the author of numerous articles
and books on British, Canadian, and US politics.

xix
 Contributors

Jacques Thomassen is Professor Emeritus of Political Science at the University of

Twente. His publications include The European Voter (Oxford University Press 2005),
The Legitimacy of the European Union after Enlargement (Oxford University Press 2009),
and (with Peter Mair) Political Representation and European Union Governance
(Routledge 2011).
Gábor Tóka is Professor of Political Science, Central European University, Budapest.
He specializes in the comparative study of voting behaviour, political attitudes, and
communication, and was a founding member of the CSES Planning Committee.
Carolien van Ham is Postdoctoral Researcher at the University of Twente and Senior
Research Fellow at the Electoral Integrity Project at the University of Sydney. She
received her PhD from the European University Institute in Florence, Italy and has
held a visiting fellowship at the University of Harvard. Her main research interests
are political representation, comparative democratization, elections, and election
fraud. She has published/forthcoming work on election integrity, representation, and
turnout in Electoral Studies, Democratization, and West European Politics.
Adrian Vatter is Full Professor of Political Science and Director at the Institute of
Political Science at the University of Berne. His main fields in research are the Swiss
political system, direct democracy, and consensus and majoritarian democracies in
comparative perspective. His most recent book publication is Das politische System der
Schweiz (Baden-Baden: Nomos UTB 2013).
Steven Weldon is Associate Professor of Political Science at Simon Fraser University.
His research focuses on political representation, participation, and elections in
advanced democracies, and has been published in, among others, the American
Journal of Political Science, the British Journal of Political Science, European Journal of
Political Research, and Party Politics.
Bernhard Wessels is Deputy Director of the research unit Democracy and
Democratization at the WZB Berlin Social Science Center, and Professor at Humboldt
University Berlin. He has published widely on elections and political representation.
He is the co-principal investigator of the German Longitudinal Election Study
(GLES).
Paul Whiteley is Professor of Government and Director of the Centre for Electoral
Studies at the University of Essex. He is former Director of the British Election Study
and the author of various books and articles in the fields of electoral politics, public
opinion, and party politics.
Wen-Chin Wu is Postdoctoral Fellow of the Center for East Asia Democratic Studies
at National Taiwan University. His research interests include comparative political
economy, comparative authoritarianism, and political methodology.

xx
1

Representation and Accountability

Jacques Thomassen

1.1 Introduction

This volume addresses the contrast between the view that elections are a
mechanism to hold government accountable and the view that they are a
means to ensure that citizens’ views and interests are properly represented in
the democratic process. It intends to explore how far this contrast and its
embodiment in institutional structures influence vote choice, political par-
ticipation, and satisfaction with the functioning of the system of political
representation and democracy in general. The volume is mainly based on
data from the second module of the Comparative Study of Electoral Systems
(CSES) which had the same purpose.1 In this introduction we will further
expand on these different models of democracy and summarize how the fol-
lowing chapters contribute to our knowledge of the effects of different insti-
tutional arrangements on people’s political behaviour and their evaluation of
the functioning of democracy.
Ever since Schumpeter defined ‘the democratic method’ as ‘that institu-
tional arrangement for arriving at political decisions in which individuals
acquire the power to decide by means of a competitive struggle for the peo-
ple’s vote’ (Schumpeter 1976, 1942), there is a broad consensus that competi-
tive elections are the identifying property of the contemporary democratic
process (Powell 2000).
However, as much consensus there is on the importance of competitive
elections for democracy, so little agreement there is on the precise function of
elections. Different theories of political representation have different views
on this. Most modern conceptions of democracy agree that the basic function
of elections is to connect the policy preferences of the people to public policy.

1
Jacques Thomassen

However, they differ in what is meant by ‘the people’ and in the mechanism
that is supposed to connect the policy preferences of the people to govern-
ment policy. The two main views of democracy are the majoritarian and the
consensus or proportional view of democracy (Lijphart 1984; Lijphart 1999;
Huber and Powell 1994; Powell 2000). According to Lijphart (1984: 4) the
main characteristic of the majoritarian view is that when the people are in
disagreement and have divergent preferences, the government should pri-
marily be responsive to the majority of the people. In the consensus view
government policy should be responsive to as many people as possible. These
different principles lead to different views on the function of elections and
are embodied in different political institutions.
In this volume we focus on the distinction between the majoritarian and
consensus models of democracy. These models are the embodiment in insti-
tutional structures of the two different views of democracy and reflect a dif-
ferent view on the functions of elections.2 In the majoritarian view the single
most important function of an election is the selection of a government. It
requires that the voters have a clear choice between two competing (groups
of) parties. The concentration of power in the hands of an elected majority
government brings the government under tight control of the majority of the
electorate. This control can be based on two different mechanisms, depend-
ing on voters’ time perspective or rather the considerations they take into
account when they decide how to vote. If voters base their choice on what the
(two) competing parties have on offer in their party manifestos, the winning
party can be said to have a policy mandate from a majority of the electorate
(Powell 2000: 8). This is basically the mechanism assumed by the Responsible
Party Model (Schattschneider 1942; APSA 1950; Thomassen 1994; Klinge-
mann et al. 1994).
However, this model of political representation is very demanding, in par-
ticular with regard to what is required of the voters. Also, a single vote can
hardly provide a policy mandate for a multiple package of issue dimensions.
Therefore, this model is often claimed to be totally unrealistic and unfeasible
(Riker 1982; Thomassen 1994). An alternative and perhaps more feasible
model is based on Schumpeter’s idea of a competitive democracy. In this
model elections are an accountability mechanism, where the sanctions are to
extend or not to extend the government’s tenure (Schumpeter 1976 (1942),
Chapter 22).
The major difference with the policy mandate model is that voters make
their vote choice on the basis of their evaluation of the performance of the
incumbent government. If they are satisfied with that performance, they will
vote for the party or parties in government; if they are dissatisfied, they will
‘throw the rascals out’. This model of accountability is far less demanding of
the voters because all they need to know is which party, or coalition of

2
 Representation and Accountability

parties, is in power and which one is the opposition. Their information about
the content of government policy can be limited. Being satisfied or dissatis-
fied with the government, its policies, or the outcomes thereof is all it takes
(Fiorina 1981).
An essential requirement of this model of accountability at the system level
is the clarity of responsibility (Manin et al. 1999; Powell 2000; Lundell 2011).
Accountability is by definition close to impossible if it is not perfectly clear
who, i.e. which political party or coalition of parties, is responsible for gov-
ernment policy. But not only the incumbent but also the possible alternative
future government must be identifiable. A second requirement is that the
voters’ sanction of the party or parties in power is effective, i.e. that they
really can throw the rascals out without the risk that these (or some of them)
will return to power after having lost the elections. This mechanism can only
work in a majoritarian system where two (blocks of) parties compete for a
majority of the votes and the winner automatically takes (over) government
responsibility. A final requirement is that the voters do indeed hold the gov-
erning party or coalition of parties accountable for their performance while
in office, i.e. when they vote retrospectively.
In consensus models of democracy, or proportional systems, the major
function of elections is to elect the members of parliament who together
should be as representative as possible of the electorate as a whole. The crite-
rion for the democratic quality of the system is how representative parlia-
ment really is. There is no deterministic relationship between the election
outcome and the formation of the government. As a multiparty system is one
of the characteristics of a consensus model of democracy, usually a coalition
of several parties will be needed to form a majority government. Coalitions
will usually be broad, doing justice to the principle that not just a bare major-
ity but as many people as possible can influence government policy.
Between them, the two models of democracy fulfil the two most important
functions elections in a representative democracy have, according to main-
stream normative democratic theory. First, elections allow voters to deter-
mine the political colour of their government, making government
accountable to the judgement of the people. Second, elections produce a leg-
islature that is representative of the distribution of policy preferences amongst
the electorate. However, it may be obvious that there is a certain tension
between these two functions (Manin et al. 1999; Powell 2000; Lord and Pollak
2010; Aarts and Thomassen 2008a; Dalton et al. 2011a; Dalton et al. 2011b).
Electoral and more generally democratic systems cannot optimally serve both
functions at the same time. An important criterion of democracy underlying
the consensus model of democracy is inclusiveness, i.e. the part of the elector-
ate represented in government is as large as possible (Lundell 2011; Kaiser
et al. 2002). However, the better this criterion is met, the lower the clarity of

3
Jacques Thomassen

responsibility and the lower the possibility of an alternation of government,

making it close to impossible to hold accountable and sanction a government
that no longer has the confidence of a majority of the people, a main criterion
of democracy.
The key question then is which model serves democracy best. This, how-
ever, is hard to tell because the two visions of representative democracy rep-
resent two different normative views on democracy and incorporate different
electoral institutions precisely because they are supposed to serve different
purposes or at least different aspects of democracy. As Powell argues: ‘empiri-
cal predictions about the nature of the citizen–policymaker relationship will
focus on dissimilar dependent variables and not really be alternative theories
about achieving the same goal’ (Powell 2000: 7).
We will follow two ways out of this dilemma (cf. Aarts and Thomassen
2008a:7). First we will focus on these different dependent variables. We will
see how well the two models achieve the goals they are supposed to achieve
and to what extent they achieve these goals better than the other model: how
well do majoritarian systems serve the function of electoral accountability and do
they serve this function better than consensual systems? And how well do consen-
sual systems serve the function of representativeness, and do they serve this function
better than majoritarian systems?
A second way out of this dilemma is to transform these dependent variables
into independent variables, and make a comparative assessment of the extent
to which majoritarian and proportional systems of government are instru-
mental for democracy, defined at a higher level of abstraction. This is the
approach taken in several major pieces of previous research. Powell, for
instance, starts from the normative assumption that democratic policymak-
ers should do what their citizens want them to do. The role of elections then
is to link the preferences of citizens to the behaviour of their policymakers
(Powell 2000: 251). His initial empirical findings prove that if this is taken as
the main function of democratic elections, ‘the proportional vision and its
designs enjoyed a clear advantage over their majoritarian counterparts in
using elections as instruments of democracy’ (Powell 2000: 254). More recent
research suggests, though, that this clear advantage applies to most of the
post-war period but vanishes after the mid1990s (Powell 2009; Golder and
Stramski 2010).
In a similar vein Lijphart in his Patterns of Democracy (Lijphart 1999) tries
to assess whether the distinction between majoritarian and consensus
democracy makes a difference for how well democracy works. By comparing
majoritarian and consensus democracies on a number of performance indi-
cators he comes to the conclusion that consensus democracies perform bet-
ter in almost every respect and provide a ‘kinder and gentler’ democracy.
They score better on the best-known indexes of democracy, women are better

4
 Representation and Accountability

represented in parliament, consensus democracies are more egalitarian, turn-

out is higher, and last but not least citizens in consensus democracies are
significantly more satisfied with the functioning of democracy in their coun-
tries than citizens in majoritarian democracies (Lijphart 1999). Also, possible
conflicts in society and politics are subdued in consensus democracies
because of their inclusiveness. The best known example is the difference in
satisfaction with democracy between winners and losers of the elections
among the voters. In all political systems the winners of elections (i.e. the
voters of parties included in the government), are more satisfied with the
functioning of democracy than losers. However, this difference between win-
ners and losers is bigger in majoritarian than in consensus models of democ-
racy. The explanation for this phenomenon is that in majoritarian systems
(the representatives of) the losers of elections are totally excluded from the
policymaking process, whereas in a consensus model with its mechanisms of
power-sharing they are not (Anderson and Guillory 1997; Lijphart 1999;
Anderson et al. 2005).
A clear limitation of many of these studies is that they are limited to the
macro level. In the typical study institutional characteristics like the distinc-
tion between majoritarian and consensus democracies are related to aggre-
gate measures like the level of turnout or the level of satisfaction with
democracy. But what remains hidden is which individual perceptions of the
democratic process make people in one type of democracy more satisfied
with the functioning of democracy than the other, or which individual calcu-
lus explains why turnout is different in different types of democracies.
It is the purpose of this volume to unveil these hidden perceptions and
mechanisms. For this purpose we rely on the second module of the Compara-
tive Study of Electoral Systems (CSES) as our main database. It is the power of
the research design of CSES that it enables us to unravel these hidden mecha-
nisms by relating system characteristics to the attitudes and behaviour of
individual people. The main characteristics of the CSES research design are
illustrated in Figure 1.1.3 Its main innovation compared to traditional elec-
toral research is the possibility it creates to relate micro-level data to macro-
level data. Traditional electoral research tends to focus either on micro–micro
relations (arrow BC) or on macro–macro relations (arrow AC) in which the
dependent variable often is an aggregate variable (for instance the level of
satisfaction with democracy in a country). CSES opens up totally new ave-
nues of research by making it possible to study the interactions between
macro and micro variables. These are mainly of two kinds. A first kind of
interaction occurs when the strength of the relationship between two micro
variables (arrow BC) depends on a macro variable (A). A second kind of inter-
action occurs when a micro variable (B) can be interpreted as an intervening
variable explaining the relationship between a macro and a micro variable

5
Jacques Thomassen

A. Macro characteristics
of democratic systems
Majoritarian vs. consensual
Institutional quality
Policy performance
Clarity of responsibility
Polarization

B. Micro-level independent (or C. Micro-level dependent

intervening) variables variables
social position party/candidate choice
perceptions choice options casting a vote
perceptions institutional quality political participation
policy preferences perceptions of accountability
perceptions of government perceptions of representation
performance satisfaction with democracy
(perceptions of accountability)
(perceptions of representation)

Figure 1.1.  The CSES research design

(AC). Most of the research questions in this volume can only be answered by
taking into account one of these types of interaction.

1.2  Main Findings

As argued above, we try to answer two kinds of questions. First, we examine to

what extent political institutions matter for how well the functions of account-
ability and representativeness are met: do majoritarian systems of democracy
serve the function of accountability better and do consensus systems of
democracy serve the function of representativeness better (­Chapters  3–5)?
The second question is whether consensus democracy leads to a higher qual-
ity of democracy in terms of a higher political participation (Chapters 6–7), a
higher satisfaction with how well citizens are being represented (Chapter 8),
and a higher satisfaction with democracy (­Chapters 9–10). The final question
is to what extent consensus democracy subdues the political effects of social
and political conflicts (Chapters 11–12).

6
 Representation and Accountability

1.2.1  Patterns of Democracy

As explained above, Lijphart’s distinction between majoritarian and consensus
types of democracy and its specific components can be considered as the insti-
tutional embodiment of the two views of democracy. Therefore, the main inde-
pendent variable at the system level throughout this book is the distinction
between majoritarian and consensus democracies. A clear conceptualization as
well as a detailed operationalization of this distinction was first introduced by
Arend Lijphart (1984 and 1999). For our purposes Lijphart’s measurements
needed an update. First, his most recent measurements were over ten years old
and, second, they were not available for all countries represented in CSES. In
Chapter 2 Bernauer, Giger, and Vatter develop an updated measure of majoritar-
ian vs consensus democracy. Their efforts lead to a more recent data set,
improved measurements of variables, and an extended set of countries, cover-
ing 35 of the 36 democratic countries included in the second module of CSES.
They discuss several possible improvements of Lijphart’s original measure-
ments, among others the inclusion of direct democracy as an indicator of con-
sensus democracy. Adding this variable to the original set of variables proposed
by Lijphart leads to a three-dimensional rather than the original two-­dimensional
structure. Leaving out this newly developed indicator, but improving some of
the original measurements, yields a two-­dimensional structure (an executives–
parties and a federal–unitary dimension) very similar to Lijphart’s original typol-
ogy. Therefore, for the sake of comparability with Lijphart’s original findings it
was decided to use this two-­dimensional typology as the main institutional vari-
able throughout this book.

1.2.2  Political Institutions and Party Choice

Our first major research question is how well majoritarian and consensus
systems of democracy serve the functions of electoral accountability and rep-
resentativeness respectively. Chapters 3 and 4 try to answer this question
from the perspective of the role of the voters in the process of political repre-
sentation. The main relevant questions then are: (a) do voters in consensus
systems focus more on which party represents their policy preferences best
and (b) do voters in majoritarian political systems base their vote more on
their perception of the performance of the incumbent government than do
voters in consensus systems?
Chapter 3 starts from Lijphart’s claim that in consensus democracies gov-
ernment policy is more in line with the policy preferences of (the majority of)
the people than in majoritarian democracies. Wessels and Schmitt try to
unravel the possible democratic mechanism behind this finding. They test
the possibility that elections are more meaningful in consensus than in

7
Jacques Thomassen

majoritarian democracies. In their view elections can only be meaningful and

therefore function as an instrument of democracy if (a) political parties offer
meaningful choices, i.e. offer different policy platforms to the voters, and
(b) if the voters behave accordingly, i.e. vote for the party that is closest to
their own policy preferences. They examine to what extent these two condi-
tions are related to each other, i.e. does the extent to which voters vote
according to their policy preferences depend on the clarity of the choices
offered to them? Both the clarity of policy choices and policy voting are oper-
ationalized in terms of the left–right dimension. The clarity of policy choices
is assumed to be higher the stronger the policy stands of political parties are
related to their left–right position as perceived by the voters. Policy voting
means that voters vote for the party that is closest to their own position on
the left–right dimension.
The variation in clarity of the policy positions of political parties between
countries appears to be huge. These differences in clarity are strongly related
to the extent to which people vote according to their policy preferences. In
other words, the extent to which voters perform their democratic duty strong-
ly depends on the political supply side: clear parties produce clear p­ olicy-based
vote choices. Where the policy stands of political parties are unclear, voters
cannot be expected to base their choice on their policy preferences.
The chapter also clearly demonstrates that the extent to which people vote
according to their policy preferences is not related to institutional differenc-
es, i.e. to the distinction between consensus and majoritarian democracies.
Political institutions are probably too far away from the real world of politics,
i.e. the clarity of policy supplies, to have a direct effect on people’s calculus of
voting.
In Chapter 4 Burlacu and Toka continue the discussion about the extent to
which policy-based voting is related to the type of democracy. Their approach
is somewhat different, though, than that of Chapter 3. Rather than focusing
on the effect of institutional differences on the relationship between voters’
policy preferences and their vote choice, the emphasis in this chapter is on
the responsiveness of aggregate election outcomes to shifts in citizens’ pref-
erences. It examines the extent to which this responsiveness is related to
institutional differences and the polarization between political parties. The
authors also try to answer the question whether there is a trade-off between
policy-based voting and performance-based (i.e. retrospective) voting. They
hypothesize—in line with the argument developed above—that Lijphart’s
executive–parties dimension of consensus democracy promotes policy-based
voting partly through ideological polarization between the parties, but also
independently of that. The reverse applies to performance-based voting: the
more majoritarian a country’s institutions, the stronger the impact of voters’
performance evaluations on the vote will be. These expectations are only

8
 Representation and Accountability

partly borne out. There is indeed a trade-off between policy-oriented voting

and performance-oriented voting, but there is hardly any support for the
hypothesis that performance-oriented voting is typical for majoritarian
democracies and policy-oriented voting for consensus democracies. As it
turns out, policy-oriented voting seems to be typical for the multiparty sys-
tems in Western Europe whereas performance-oriented voting occurs more
often in a non-European context. This might at least partly be due to the fact
that left–right orientations are less common in these non-European con-
texts. Still, features associated with the executive–parties dimension make
consensus democracies somewhat more responsive to the electorate’s policy
preferences than majoritarian democracies. This institutional effect is at least
partly mediated by consensus democracy facilitating slightly more ideologi-
cal polarization between parties than majoritarian systems. However, the
statistical evidence of these findings is anything but impressive. Therefore,
this chapter concludes that institutional differences add little to explaining
why the world’s democracies show such a striking variation in policy
responsiveness.
One of the classic questions in electoral research is to what extent people’s
electoral choices are determined by the social structure they are part of. The
extent to which this is the case varies from country to country and within
countries over time. However, what explains why the vote is more socially
anchored in some countries than in others is hardly known. Chapter 5 by
Magalhães addresses this shortcoming in the literature and examines the
extent to which variations in political institutions can explain the variations
between countries in terms of the social anchoring of the vote in legislative
elections. The hypotheses tested are that, first, consensual democracies
should display higher levels of structural voting and, second, presidential
regimes should display lower levels of structural voting. Alternative hypoth-
eses possibly explaining structural voting refer to the level of economic devel-
opment and the length of experience with democracy a country has. The two
hypotheses on the effect of political institutions are borne out. Structural
voting is indeed related to consensual democracy: in democracies where elec-
toral systems are less permissive, where party system fragmentation is lower,
and executive dominance over policymaking is more pronounced, the
anchoring of the vote on religious and gender differences turns out to be
weaker. Also, on average, structural voting is lower in presidential than in
parliamentary or semi-presidential regimes. In presidential systems the insti-
tutional rules that separate the origin and survival of parliaments and execu-
tives seem to create disincentives for the adoption, in legislative elections, of
appeals to socially defined and rooted groups of voters, thus promoting great-
er social heterogeneity of party constituencies. There is hardly any support
for the alternative hypotheses. These findings show the failure of a purely

9
Jacques Thomassen

sociological approach to the explanation of voters’ alignments with parties

and the importance of political institutions for the development of voters’
alignments with political parties.

1.2.3  Consensus Democracy and Political Participation

Chapters 6 and 7 examine the relationship between political institutions and
political participation. As observed above, Lijphart’s claim that consensus
democracy is superior to majoritarian democracy is based on several argu-
ments. One of them is that political participation, voting in particular, is
more widespread in consensus than in majoritarian democracies. Chapter 6
examines to what extent consensus democracies are indeed conducive to
turnout and if so why this is the case. Chapter 7 widens the question to less
institutionalized forms of political participation. Chapter 6, by Blais, Singh,
and Dumitresco, starts from the expectation that the decision to vote or not
to vote strongly depends on people’s belief that one of the parties represents
their views reasonably well. Therefore, a logical further expectation is that as
far as consensus democracy is conducive to turnout this relationship is medi-
ated by feelings of being represented. In other words, consensus democracy
has a positive effect on people’s belief that there is a party representing their
views reasonably well, which in turn has a positive effect on turnout. A fur-
ther expectation is that this is mainly due to the proportional electoral sys-
tem (proportional representation, or PR), one of the main characteristics of a
consensus democracy. PR leads to the presence of more parties running in the
election and this automatically increases the probability of finding a party
that one agrees with. Also, PR produces a more polarized party system. Such
polarization means that a greater diversity of highly differentiated viewpoints
is presented to the electorate and that it is easier to identify at least one party
that represents reasonably well their own political preferences. So the main
hypothesis tested in this chapter is that the relationship between the electoral
system and turnout should disappear when we take into account feelings of
representation. The empirical analysis clearly shows that the propensity to
vote is indeed much higher among those who feel represented than among
those who do not. The hypothesis that citizens are more likely to feel that a
party represents them in a consensual or PR system is also corroborated.
The analysis further demonstrates that it is purely the electoral system that
matters and not the more encompassing dimensions of the consensus model
of democracy. There is no empirical evidence for the beneficial effect of other
aspects of consensualism on turnout in the same manner as found for PR. The
relationship between PR and the feeling of being represented is at least par-
tially mediated by the degree of polarization of the party system. The mere
presence of a greater number of parties is not necessarily conducive to

10
 Representation and Accountability

individuals becoming more prone to say that a party represents their views.
When perceptions of representation are taken into account the—modest—
relationship between PR and turnout loses its significance, i.e. proportional
representation has no direct, independent effect on electoral participation
once political attitudes are taken into account. In other words, the relation-
ship between PR and turnout is indeed mediated by individuals’ ability to
identify a party representing their views.
Chapter 7, by Weldon and Dalton, examines how democratic institutions
affect citizens’ political behaviour beyond the act of voting. The starting
point is again Lijphart’s presumption that consensual institutions incorpo-
rate more citizens into the electoral process and lessen political inequality,
because they give citizens effective voice and representation. This presump-
tion seems to imply that consensualism has spillover effects that stimulate
participation more broadly. However, in contrast to the effect of consensu-
alism on turnout, these further implications of consensualism have hardly
been examined and are not uncontested, not even at a theoretical level.
According to a competing view, consensualism may have the opposite
effect, actually decreasing citizen involvement beyond voting, precisely
because citizens already feel better represented through the electoral pro-
cess in these systems. In addition, majoritarian electoral systems tend to be
candidate-centred, with MPs elected from single-member districts. Because
of this, majoritarian systems offer greater incentives for participation
beyond voting, such as campaigning and contacting public officials. On the
second dimension of political institutions, the federal–unitary dimension,
federal systems may suppress voter turnout, because individual elections
are less critical, but they also offer more opportunities for participation at
the sub-national level than do unitary systems. Because elections occur
more often, civic organizations may stay stronger over time, helping to
mobilize and encourage non-voting forms of participation. These hypoth-
eses are corroborated by the empirical analysis. Participation in all other
political activities than voting is lower in consensual systems when opera-
tionalized on the basis of the executives–parties dimension—often mark-
edly so as in the case of campaign activity, persuading others how to vote,
and contacting a politician. In short, consensual systems seem to demobi-
lize citizen participation beyond casting a ballot. These correlations are
even more outspoken when only the electoral system is taken into account.
As shown in the previous chapter, a proportional electoral system might
encourage people how to vote, but it apparently discourages people from
showing their support during the campaign or trying to persuade others
how to vote. These patterns are unanticipated by the consensualism litera-
ture, which presumes that voting turnout is symptomatic of general involve-
ment in the political process.

11
Jacques Thomassen

When political systems are distinguished on the basis of the federal–­

unitary dimension, all other forms of participation than voting are more
common in federal systems. The participation stimulus of decentralized fed-
eral systems is especially apparent for interpersonal forms of participation
such as trying to persuade others how to vote, contacting a politician, or
working with a group.
Taken together, the two chapters paint an interesting picture of how the
institutional context shapes individual political activity and the develop-
ment of a democratic civil society. Lijphart has long argued that consensual-
ism produces a kinder, gentler form of democracy by bringing more of the
population into the democratic process and representing a broader range of
societal interests and viewpoints. From a formal institutional perspective,
there is some evidence that this is indeed true. But this relationship appears
to be due only to the electoral system, not to the wider characteristics of con-
sensus democracy. Chapter 7 raises more fundamental doubts about consen-
sualism being conducive to a more involved democratic citizenry for
non-voting forms of participation. A constitutional and electoral system
designed to maximize voting turnout may have unintended negative conse-
quences on other forms of political participation.

1.2.4  Political Institutions and Satisfaction with Political Representation

and Democracy
The remaining chapters of the book all focus on how people’s evaluation of
the system of representative democracy is related to the democratic institu-
tions they live under. The main dependent variables are people’s perceptions
of the quality of representation and their satisfaction with the functioning of
democracy in their country. These variables refer to two indicators of the qual-
ity of democracy Lijphart uses as evidence for his general claim that the qual-
ity of democracy in consensus democracies is higher than in majoritarian
democracies. The first one is the quality of representation, operationalized as
the proximity of government policy to the policy preferences of the people.
Referring to earlier work of Huber and Powell, Lijphart claims that this dis-
tance is smaller in consensus than in majoritarian democracies. A second indi-
cator is people’s satisfaction with the functioning of democracy. Referring
again to earlier empirical work, he claims that satisfaction with the f­ unctioning
of democracy is higher in consensus than majoritarian democracies. In this
part of the book these claims will be brought to a further test. ‘Objective’ meas-
ures for the quality of representation, comparing people’s policy preferences
with government policy, are not available for a range of CSES countries.
Instead, we rely on people’s subjective perception of how they are represented
by political parties and political leaders for the assessment of the quality of

12
 Representation and Accountability

political representation, just as in the case of people’s assessment of the quality

of democracy in their country. Obviously, as Holmberg argues in Chapter 8, a
subjective measure of policy representation is not the same thing as an objec-
tive measure. However, it is not necessarily less important. Subjective percep-
tions of how political parties and their leaders represent the people define the
legitimacy of the system of political representation, not objective measures,
however important the latter may be.
Holmberg presents a first straightforward test of the hypothesis that people
in consensus democracies feel better represented than in majoritarian democ-
racies. The outcome of the test is as straightforward as the test itself: there is
hardly any relationship between institutional design and people’s perception
of how they are represented by parliament, political parties, and political
leaders. According to Holmberg, the mechanisms of political representation
in consensus democracies are somewhat different than in majoritarian sys-
tems but the one is not necessarily better or more effective than the other.
What really matters is the age of democracy. People in older established
democracies feel better represented than people in developing democracies.
Chapter 9, by Sanders, Clarke, Stewart, and Whiteley, tries to explain vari-
ations in the satisfaction with the functioning of democracy. It examines the
relative importance of factors at the individual and institutional levels. At the
individual level it was found that satisfaction with democracy is positively
affected by citizens’ assessments of overall government performance, their
feeling of being represented by existing political institutions, and their belief
that the current regime effectively delivers democratic values. From the over-
all perspective of this book the effects of political institutions are the most
interesting ones. In line with Lijphart’s arguments the hypothesis is tested
that plurality rules exert a negative impact on satisfaction with the function-
ing of democracy. This is because of the unfairness of plurality systems by
producing a disparity between parties’ vote shares and their respective seat
shares.
A counter-hypothesis implicitly recognizing the potential practical bene-
fits of plurality electoral systems is that satisfaction with democracy will be
higher where there is greater clarity of (cabinet) responsibility.
The empirical analysis shows that there is no statistically significant effect
of any measure of consensualism on satisfaction with democracy. However,
when the practical electoral outcomes of these institutional rules rather than
the rules themselves are taken into account, the effects are clearly significant.
Both disparity and clarity of responsibility have significant effects in the pre-
dicted directions—negative for disparity and positive for clarity.
This suggests that plurality and consensualism have no direct effects on
satisfaction with democracy but their consequences do. Two main conse-
quences of plurality—disproportionality and clarity—both affect satisfaction

13
Jacques Thomassen

with democracy, but they do so in opposing directions. People tend to dislike

the unfairness associated with disproportionality, but they simultaneously
tend to like clarity of responsibility since it strengthens their ability to iden-
tify the ‘rascals’, who may need to be ‘thrown out’ in subsequent elections.
Ceteris paribus, the former mechanism reduces satisfaction with democracy;
the latter enhances it. As a consequence, the authors of this chapter argue,
would-be institutional reformers find themselves in a somewhat paradoxical
position. If the objective of reform is to maintain, or even to increase, the
overall level of approval for the regime, the abandonment of plurality rules is
very much a double-edged sword. Though a switch to PR, on the one hand,
might serve to increase democracy satisfaction by removing unfairness to
minor parties and ‘wasted votes’, on the other hand it is also likely to reduce
clarity of responsibility and therefore simultaneously to reduce such
satisfaction.
Chapter 10, by Peffley and Rohrschneider, examines the relationship
between the quality of political representation and people’s support for
democracy. It extends the research questions examined in previous chapters
in two ways. First, rather than focusing on the idea of substantive representa-
tion, or the correspondence between the outcomes citizens want and those
produced by the government, it focuses on an equally important component
of representation, its procedural dimension, or its procedural fairness. Proce-
dural fairness is concerned less with outcomes and more with the processes
by which governmental policies are made and administered. Citizens’ evalu-
ations of procedural fairness can be assumed to be based on assessments of
whether authorities are motivated to be fair, are honest, and follow ethical
principles of conduct, whether opportunities for representation are provided,
and whether authorities behave in a biased fashion. Also, this chapter adds
another dependent variable. It examines not only the effect of people’s evalu-
ation of the quality of representation on their satisfaction with the function-
ing of democracy, but extends this question to people’s support for democracy
as such, a level of support that is far more consequential for the stability of
democracy. Citizens’ perceptions of the procedural fairness mainly refer to
their experiences with so-called output institutions, like the police, bureau-
crats, and legal authorities.
The main hypothesis tested in this chapter is that procedurally fair output
institutions increase public evaluations of a regime. This hypothesis is clearly
borne out. The empirical findings unequivocally show that well-functioning
output institutions have a positive effect, not only on people’s satisfaction
with the functioning of democracy but on their support for the ideals of
democracy as well. The latter effect is even stronger than the former one.
Thus, the character of output institutions not only affects democratic perfor-
mance evaluations but public support for democratic ideals as well. Also,

14
 Representation and Accountability

citizens’ evaluations of regime procedures, i.e. their beliefs that human rights
are respected, their perceptions of corruption and the representation process,
are shaped by objective indicators of the quality of output institutions.
The alternative explanation that popular evaluations of regime procedures
are due to the ‘kinder, gentler’ consensus institutions that Lijphart envisioned
is hardly supported by the data. There is some evidence that citizens residing
in federal regimes are more likely to think that human rights are respected in
their country; and they also believe that corruption is lower. But all in all
there is fairly weak evidence in support of the idea that consensus systems
have more procedural integrity than majoritarian institutions.

1.2.5  Does Consensus Democracy Subdue Social and Political Conflicts?

In Chapter 11, by Aarts, Thomassen, and Van Ham, two alternative hypoth-
eses on the development of citizens’ support for representative democracy are
put to a test. According to the modernization thesis, social modernization has
gradually transformed the relationship between citizens and the state. Great-
er political skills and resources—that is, higher levels of cognitive mobiliza-
tion—lead the contemporary electorate towards elite-challenging forms of
political action, which places them in conflict with politicians and govern-
ment officials. These better-educated citizens are inclined to question demo-
cratic politics as it is currently practised not because they do not endorse
strong democratic ideals but because they demand a greater role in the politi-
cal process. Therefore, the hypothesis derived from modernization theory is
that over time people’s confidence in the processes of representative democ-
racy has disproportionally declined among people at the upper end of the
economic order: the better educated, the more skilled, and those with higher
incomes.
The main hypothesis tested in this chapter is exactly the opposite and is
derived from the globalization literature. According to this literature at least
in Western Europe a new structural conflict is developing, the conflict between
‘winners’ and ‘losers’ of globalization. Whereas better-educated cosmopoli-
tan citizens in general profit from the effects of globalization, the people at
the margins of the economic order, the less educated, the less skilled, and
those with lower incomes tend to be the losers of globalization. They feel
threatened by the consequences of globalization, by the open economy lead-
ing to—at least in their perception—the loss of jobs, and the immigration of
people with different economic and cultural backgrounds, often from the
Muslim world. As political elites traditionally tended to be little responsive to
these feelings, the expectation derived from these developments is that not
the better-educated but the lower-educated and less-skilled groups in society
are gradually losing their confidence in the responsiveness of political parties

15
Jacques Thomassen

and political leaders and therefore are becoming less satisfied with the func-
tioning of democracy.
The additional hypothesis tested in this chapter is that in consensus democ-
racies this effect will be subdued because one of the main characteristics of a
consensus democracy is that it is responsive to minority groups. Because of its
proportional electoral system, new political movements based on new con-
flict lines, such as globalization, have easy access to the process of political
representation.
The main hypotheses in this chapter are not borne out. Although in general
less-educated and less-skilled citizens do feel less well represented by any polit-
ical party and are less satisfied with the functioning of democracy, there is no
evidence for a differential development of these feelings: there is no growing
gap in this respect between the winners and losers of globalization. Since no
effect of globalization was found, this by definition means that the political-­
institutional context is not relevant either.
Chapter 12, by Chang, Chu, and Wu, focuses on the well-known winner–
loser gap. Previous research has shown that among voters the losers of elec-
tions tend to be less satisfied with the functioning of democracy than the
winners. It has also been shown that this gap between winners and losers is
larger in majoritarian than in consensus democracies. In these previous stud-
ies electoral winners and losers are distinguished on the basis of a single elec-
tion and the interaction between winners and losers is treated as a one-shot
game. However, since elections in democracies are held routinely, winners
can become losers when electoral alternation occurs. In this chapter two con-
secutive elections are taken into account to define winners and losers. This
logically leads to four categories: two-time winners, two-time losers, winners-
losers, and losers-winners. In this chapter this distinction is connected to
Huntington’s two-turnover test in the comparative democratization litera-
ture. According to Huntington, a nascent democracy is considered consoli-
dated if it has experienced two peaceful electoral alternations. However, the
relevant literature on this subject consequently neglects the importance of
ordinary citizens’ democratic attitudes during electoral alternations in young
democracies. And yet, it is generally agreed that citizens’ attitudes towards
democracy are essential for the legitimacy and stability of democratic regimes,
in particular new democratic regimes. Therefore, this chapter, by examining
how citizens’ experience as winners or losers in successive elections influ-
ences their support for democracy is an important complement to the two-
turnover test literature. Because of their limited experience with democracy,
the expectation is that the gap between winners and losers will be larger in
developing than in established democracies.
The main hypothesis at the individual level in this chapter is that the
experience of having been a winner at least once is more likely to develop a

16
 Representation and Accountability

f­ avourable attitude towards democracy than never having experienced win-

ning before. Only repeated losers are expected to be less satisfied with the
system than other groups. This expectation is borne out. It is the experience
of winning at least once that matters in shaping individuals’ support for dem-
ocratic practices and principles. Being electoral winners matters more to citi-
zens in young democracies than to those in developed democracies. Without
a longer experience with democratic elections, constant losers may easily
blame the democratic system for their loser status. By contrast, as demo-
cratic practices are repeated and consolidated, constant losers may either
have experienced winner status or have more good beliefs of becoming win-
ners in the future. From this finding a clear lesson for developing democra-
cies can be learned. It is important for current losers to realize that the recent
defeat is merely temporary and that sooner or later there will be another
opportunity to determine who holds political power. Therefore, in order to
forge a consensus about new rules of the game after the democratic transi-
tion, it is imperative that current losers will not be permanently excluded
from political power but will have chances to contest and win subsequent
elections.

1.3  In Conclusion: Do Institutions Matter?

The purpose of this volume is to assess to what extent consensus systems of

democracy serve democracy better than majoritarian systems and, if so, why
this is the case. We argue that this question can hardly be answered because
the two models of democracies reflect different normative views on democ-
racy and the function of elections. Empirical research cannot decide which
normative view on democracy and the function of elections is best. We found
two ways out of this dilemma. First, we examined how well the two models
achieve the goals they are supposed to achieve and to what extent they do
this better than the other model. Second, we examined how conducive the
two models are for the quality of democracy, measured by indicators recog-
nized from both views as important indicators of the quality of democracy,
political participation and people’s satisfaction with the functioning of repre-
sentative democracy.
Consequently, we try to answer four questions. First, do majoritarian sys-
tems of democracy serve the function of accountability better and do consen-
sus systems of democracy serve the function of representativeness better
(Chapters 3–5)? Second, does consensus democracy lead to a higher degree of
political participation (Chapters 6–7)? Third, does consensus democracy lead
to a higher satisfaction with how well citizens are being represented (Chapter
8) and a higher satisfaction with democracy (Chapters 9–10)? The final

17
Jacques Thomassen

question is to what extent consensus democracy subdues the political effects

of social and political conflicts (Chapters 11–12).
What the findings in all these chapters have in common is that formal
political institutions and in particular the distinction between consensus and
majoritarian system of democracy are far less relevant for people’s attitudes
and behaviour than often presumed.
First, we hardly found any evidence that retrospective voting is typical for
majoritarian democracies and policy-oriented voting for consensus democra-
cies. Policy voting is related to characteristics of the party system, like the
clarity of policy choices and polarization but not to more formal political
institutions.
Second, consensus democracy as such does not seem to have a positive
effect on people’s decision to cast a vote, although a proportional electoral
system, one of its main elements, indirectly does. This is because under a
proportional electoral system people more often feel themselves to be repre-
sented by a political party. Once people’s perception of being represented is
controlled for, the relationship between the electoral system and voting dis-
appears. Perhaps somewhat surprisingly the executive–parties dimension of
consensus democracy seems to have the opposite effect on political participa-
tion beyond casting a ballot. This negative effect is even stronger when only
the electoral system is taken into account.
Third, we found no relationship between institutional design and people’s
evaluation of the system of political representation and democracy in gener-
al. There is no difference between the two systems in how people feel repre-
sented by parliament, political parties, and political leaders. The expectation
that majoritarianism and a plural electoral system have a negative effect on
satisfaction with democracy is based on the unfairness of plurality systems.
They are unfair because they produce a disparity between parties’ vote shares
and their respective seat shares. On the other hand, plurality electoral sys-
tems lead to a greater clarity of responsibility. This in turn might have a posi-
tive effect on satisfaction with democracy. As it turns out, although plurality
and consensualism have no direct effects on satisfaction with democracy,
their consequences do. Disproportionality reduces satisfaction with democ-
racy whereas clarity of responsibility enhances it.
Also, as far as institutions are relevant for people’s satisfaction with democ-
racy, this seems to depend more on their perceptions of the fairness of so-
called output institutions like the police, bureaucrats, and legal authorities,
than on input institutions.
The fourth question we try to answer is to what extent social and politi-
cal conflicts are subdued by the institutions of consensus democracy. Our
findings suggest they do not. The gap in satisfaction with democracy
between different social classes is no less in consensus than in majoritarian

18
 Representation and Accountability

­ emocracies. Still, these findings are too preliminary to justify strong con-
d
clusions, the more so since we did not really (re)examine to what extent
the winner–loser gap is larger in majoritarian than in consensus democra-
cies. Nevertheless, by extending the definition of winners and losers to two
consecutive elections the final chapter seems to deliver strong support for
Lijphart’s initial argument that different societies ask for different political
institutions. If winners and losers do not regularly change places because
solid social cleavages always yield the same political majority, a majoritar-
ian system is disastrous for the legitimacy of the political system among
the (permanent) minority of the people (Lijphart 1984).
But the general conclusion of this volume is that formal political institu-
tions are less relevant for people’s attitudes and behaviour than often pre-
sumed. This is not to say that characteristics of the political system do not
matter. But rather than formal political institutions like the electoral system
it seems to be characteristics of the party system like polarization and the clar-
ity of responsibility that really matter. This is in line with the growing body
of empirical knowledge, mostly based on the Comparative Study of Electoral
Systems, on the relationship between aspects of the political system and the
attitudes and behaviour of individual citizens (Thomassen 2005b; Klinge-
mann 2009; Dalton et al. 2011a; Dalton and Anderson 2011).
This conclusion should be a lesson in modesty for institutional engineers.
If institutional differences have as little effect as this volume suggests, or if
their effects cancel out each other, institutional reforms based on the pre-
sumption that a change in institutions will improve the functioning of
democracy are doomed to lead to frustration among both the reformers and
the citizenry at large.

Notes

1. See <http://www.cses.org/>.
2. The following characterization of the two models of democracy is from Aarts and
Thomassen (2008a: 6–7).
3. This figure was first presented by Bernhard Wessels in 1998 at a CSES conference in
Berlin. Also, see Thomassen (2000) and Klingemann (2009).

19
2

New Patterns of Democracy

in the Countries of the Comparative
Study of Electoral Systems 2
Julian Bernauer, Nathalie Giger, and Adrian Vatter1

2.1 Introduction

As the work of Arend Lijphart (1984, 1999, 2012) has impressively shown by
offering an empirical theory of democracy, a wide range of political-­
institutional characteristics can be subsumed under the concepts of consen-
sus and majoritarian democracy. In a nutshell, the ideal types of consensus
and majoritarian democracy differ in the extent to which power is concen-
trated or shared in the political system. Using principal component analyses,
Lijphart (1999) uncovers two latent (sub-)dimensions of democracy, finding
what he labels an executives–parties dimension and a federal–unitary dimen-
sion. Each construct encompasses five politico-institutional variables. For
each of these, a manifestation on a spectrum from majoritarian to consensual
is possible. Consensus democracy ideally displays multiple parties, multipar-
ty government, a balance of power between executive and legislature, a pro-
portional electoral system, interest group corporatism, a federal structure,
bicameralism, judicial review, a rigid constitution, and an autonomous cen-
tral bank.
The empirical theory of democracy proposed by Lijphart (1999) has
received a fair share of criticism (see below), but given the complex empirical
variety of contemporary democracy, it is widely accepted as a useful concep-
tualization of democracy. The reductive character of the resulting empirical
dimensions of democracy, i.e. the reduction of the number of explanatory
variables in a meaningful way by connecting them to latent constructs, ren-
ders empirical examination of a variety of topics possible. In particular, it
facilitates the analysis of individual political representation, behaviour, and
attitudes under different political-institutional settings.

20
 New Patterns of Democracy

The plan of the chapter is as follows. First, the discussion surrounding the
empirical theory of democracy presented by Lijphart (1999) is briefly reviewed
and the innovations of the study at hand are introduced. These include the
coverage of a different set of countries for the time period 1997–2006 and the
improvement of several single measurements.2 Two alternative sets of dimen-
sions of consensus and majoritarian democracy are presented. The first, two-
dimensional solution aims at comparability with Lijphart’s (1999) original
patterns and yields an executives–parties (covering party systems, electoral
disproportionality, cabinet type, and executive–legislative relations) dimen-
sion and a f­ ederal–unitary (covering constitutional federalism, fiscal decen-
tralization, bicameralism, and judicial review) dimension. It serves as a
baseline for comparison in the chapters to follow, analysing the role of con-
sensus democracy in the realm of representation. The second solution wid-
ens the theoretical scope of the original typology by introducing direct
democracy. The latter changes the expectation regarding the dimensional
structure of consensus democracy. Instead of two dimensions, the variables
should form three factors. Consensual types of direct democracy are hypoth-
esized to interact with cabinet type (Neidhart 1970), as governments are
more inclusive in the face of potential vetoes induced by (consensual) direct
democracy. In sum, we expect a reduced parties–­ elections dimension, a
­federal–unitary dimension, and a new cabinets–direct democracy dimension
once direct democracy is introduced.
With the next step, we introduce the improved and extended dataset and
describe the design of the analysis. The data cover 35 of the 36 democratic
political systems which are represented in the second wave of the Compara-
tive Study of Electoral Systems (CSES). Complete information on 11 variables
between 1997 and 2006 has been assembled. These variables on the party
system, cabinet type, executive–legislative relations, electoral disproportion-
ality, federalism, decentralization, bicameralism, constitutional rigidity, judi-
cial review, central banks, and direct democracy are discussed in turn. Two
factor analyses on different sets of variables report the expected two- or three-
dimensional structure, with direct democracy forming a separate dimension
in conjunction with cabinet type and constitutional rigidity.
The new data allow innovative analyses of the relationship between major-
itarian/consensus democracy and political attitudes, political participation,
the quality of representation, and other relevant aspects of a functioning
democracy (see last chapters in Lijphart 1999). What is more, the develop-
ments in survey research and in the field of methodology offer new opportu-
nities to retest such relationships in a multilevel framework instead of
considering macro-level relationships only. Specifically, survey projects such
as the Comparative Study of Electoral Systems provide comparable individual-­
level and country-level data. Adding to these observations the improved and

21
Julian Bernauer, Nathalie Giger, and Adrian Vatter

extended measurements of Lijphart’s (1999) typology described below, we

believe that all ingredients for intriguing retests of his propositions on the
effects of consensus democracy are now present.

2.2  Patterns of Democracy One Decade On

The typology proposed by Lijphart (1999) has received a fair share of criticism
(to name a few: Kaiser 1997; Bogaards 2000; Grofman 2000; Schmidt 2000;
Tsebelis 2002; Taagepera 2003; Ganghof 2005; Shikano 2006; Freitag and Vat-
ter 2008). This continues to inspire revisions of Lijphart’s (1999) empirical
theory, including improved measurements, in part encouraged by Lijphart
(2003: 20) himself (Keman 2000; Kaiser et al. 2002; De Winter 2005; Flinders
2005; Ganghof 2005; Schnapp and Harfst 2005). The toughest criticism comes
in both empirical and conceptual form. Conceptually, Ganghof (2005)
­challenges Lijphart’s (1999) typology because it allegedly fails to distinguish
between institutions and behaviour. Taagepera (2003) questions the logical
and empirical connections between some of the indicators, such as interest
group corporatism and the first dimension, as well as central bank independ-
ence and the second dimension. Empirically, Shikano (2006: 76–8) replicates
the analysis of Lijphart (1999) using 2000 bootstrap samples and concludes
that three instead of two dimensions should have been taken into account.
Others challenge the selection of countries (Schmidt 2000: 348) or the exclu-
sion of direct democracy (Jung 1996, 2001). We remain within Lijphart’s (1999)
behavioural-­institutional approach towards empirical patterns of democracy,
but taking on board some of the conceptual and empirical critiques.
In two journal articles, Vatter (2009) and Vatter and Bernauer (2009) have
presented improved measurements of the consensual and majoritarian fea-
tures of democracies. They carry out several changes to the original typology.
First, the time frame is more up-to-date compared to Lijphart (1999), ranging
from 1997 to 2006. Second, the selection of countries has been modified
according to the research question at hand. Vatter (2009) focuses on OECD
countries, while Vatter and Bernauer (2009) compare European Union mem-
ber states. Third, several measurements have been improved, including exec-
utive–legislative relations, cabinet type, and decentralization. Fourth, direct
democracy is incorporated into the typology. Direct democracy arguably is a
form of power-sharing in its own right (Vatter 2000; Jung 2001) with increas-
ing relevance. This goes beyond other research on direct democracy as it con-
nects it to the concept of consensus and majoritarian democracy (Hug and
Tsebelis 2002; Qvortrup 2002; LeDuc 2003; Setälä 2006) and is generally in
line with findings by Grofman (2000: 53), who provides evidence for an inde-
pendent third dimension of democracy, comprised of direct democracy alone.

22
 New Patterns of Democracy

Here, two variants of measuring consensus/majoritarian democracy are

presented. The first includes temporal and operational updates of Lijphart
(1999) only, while the second also features an encompassing measure of con-
sensual direct democracy. According to the latter conceptualization, the most
majoritarian form of direct democracy entails, apart from its complete
absence, plebiscites at the discretion of the government, as other forms of
direct democracy always introduce an additional veto player in the form of
the citizens. But there are also varieties of direct democracy which are weak in
terms of power-sharing and therefore less consensual than others (Smith
1976). Against this backdrop, Vatter (2009) develops an index of direct
democracy which discriminates the different forms of direct democracy and
places them on a continuum from high (plebiscites) over medium (manda-
tory referendums) to low (optional referendums and popular initiatives) gov-
ernmental control. The index also discriminates between simple majorities
and supermajorities, with the latter representing more consensual forms of
direct democracy. Finally, the actual use of direct democracy is considered.
The inclusion of direct democracy also changes the expectations regarding
the interactions between different political-institutional elements. While in
Lijphart’s (1999) analysis cabinet type is linked to the electoral structure and
the party system, Vatter (2009) found a strong relationship between the type
of cabinet and the strength of consensual direct democracy. While active
consensual direct democracy goes hand in hand with broadly supported mul-
tiparty governments, purely representative constitutions frequently appear
in conjunction with minimal winning cabinets. In line with the logic of the
veto players theory (Hug and Tsebelis 2002; Hug 2004), we can assume that
the possibility of referendums introduces an additional veto player which
makes significant changes in the status quo for the government more diffi-
cult. In particular, as barriers for the restraint of the executive, optional refer-
endums and popular initiatives take on the function of powerful veto players
which can delay or prevent governmental decisions, thereby serving to shape
the executive’s context of action in a significant way. Generally speaking,
uncontrolled forms of direct democracy introduce a new veto player—the
median voter of the population—into the political game and thus block the
choices of the ruling government (optional referendum) or upset their priori-
ties (popular initiative). Consequently, we can assume that the government
will do its best to reduce the uncertainty caused by uncontrolled referendums
(Vatter 2009). A rational strategy to lessen risks arising from the optional ref-
erendum and popular initiative is to widen the executive formula in order to
encompass all parties likely to make efficient use of the referendum if not
co-opted as partners in the governing coalition (Neidhart 1970). We therefore
presume that the threat of direct democracy from below leads to a boosting
of executive power-sharing.3

23
Julian Bernauer, Nathalie Giger, and Adrian Vatter

2.3  New Patterns of Democracy 1997–2006

2.3.1  Design and Data
For the edited volume on the second wave of the Comparative Study of Elec-
toral Systems, existing databases on consensus and majoritarian democracy
(Vatter 2009; Vatter and Bernauer 2009) have been unified and extended to
cover most of the 36 countries in the CSES Module 2 sample (Vatter and Ber-
nauer 2010b). Of these, sufficient information on 35 could be assembled.
Some data for Taiwan are missing. In total, information on 12 political-­
institutional variables is available for 26 ‘advanced’ democracies, and com-
plete information on 11 variables for 35 countries. The partially missing
variable is interest group corporatism. As emphasized by Taagepera (2003),
interest group corporatism is not logically connected to the other elements of
the executives–parties dimension described by Lijphart (1999). The remain-
ing 11 variables are discussed below.4 For sources, see Appendix A.2.1. The
selection equals the one made by Lijphart (1999), minus interest group cor-
poratism, and plus direct democracy and fiscal decentralization. The data and
a codebook (Vatter and Bernauer 2010b) are available from the authors.
Effective number of legislative parties (party): The first variable on the execu-
tives–parties dimension (Lijphart 1999) is the effective number of parliamen-
tary parties. A higher number of effective parties in parliament indicate
consensus democracy. The measure is based on the formula developed by
Laakso and Taagepera (1979). The index N is computed by taking the inverse
of the sum of the squared seat shares p of the parties i in parliament:

1
N= n

∑p
i=1
2
i

The measure weights the total number of parties by their strength in terms of
seats. It comes close to the raw number of parties when seat shares are distri­
buted equally. The weighting avoids that relatively small parties are counted
as similarly relevant as relatively large ones.
Oversized and minority cabinets (cab): Consensual cabinet types are charac-
terized by power-sharing. Arguably, oversized multiparty cabinets and one- or
multiparty minority cabinets are consensual types. That leaves single- and
multiple-party minimal winning cabinets as the more majoritarian cabinet
formats. In contrast to Lijphart (1999), one-party minority cabinets are treat-
ed as fully consensual rather than majoritarian traits here, as all minority
cabinets have to share power to remain in office (De Winter 2005: 10). The
measure captures the share of oversized and minority coalitions for the time
period under investigation.

24
 New Patterns of Democracy

Executive dominance (exeleg): In Lijphart’s (1999) original study, the power

balance between the executive and the legislative branch of government has
been operationalized by the duration of executives. This particular measure
has received a fair share of criticism (Tsebelis 2002: 109ff; De Winter 2005;
Ganghof 2005). Even Lijphart (2003: 20) himself expresses serious reservations
about the measure. The duration of executives does not capture the power bal-
ance well, and stability can simply result from loyalty of the government to
the parties supporting it (De Winter 2005: 11). Here, the more fine-grained
index of executive dominance proposed by Siaroff (2003) is used.5 The index is
derived from a factor analysis of 27 indicators characterizing parliamentary
systems. Of these, 11 load on a dimension which Siaroff (2003) describes as
executive dominance. The items cover areas such as the agenda-setting preroga-
tives of the government, the cohesion of the government, the electoral struc-
ture, and rights of the plenum and the committees. The values 0, 1, or 2 are
assigned for each item, where higher values indicate more majoritarian traits.
Two items are excluded from the calculations. The electoral structure is already
measured using the Gallagher index (see below). Information on another item,
capturing the influence of committee members on party positions, had to be
omitted due to missing information for a substantial share of countries.6
Electoral disproportionality (elec): One of the most prominent features of
political systems is the electoral system. The character of the electoral rules
determines much of their overall shape. Norris (2004: 209) states: ‘Electoral
systems represent, perhaps, the most powerful instrument available for insti-
tutional engineering, with far-reaching consequences for party systems, the
composition of legislatures, and the durability of democratic arrangements.’
Similarly, Powell (2000) appears to equate consensus democracy and propor-
tional electoral design. From his point of view, electoral systems represent the
two visions of democracy which equal consensus and majoritarian democracy.
From this perspective, it is sufficient to focus on the trigger which influences
the consensual or majoritarian character of political systems to a large degree.
While electoral systems can be generally divided into proportional and
majoritarion ones, a simple dummy indicator would be a misconception of
the wide empirical range of the permissiveness of electoral systems even
among proportional rules. Features such as district magnitude and additional
electoral tiers produce large variation in the strength of proportional electoral
rules (Cox 1997). A widely accepted output measure of the permissiveness of
electoral systems is the Gallagher (1991) index of disproportionality. The
index (G) considers the degree of translation of votes v for party i into their
seats s in parliament, punishing wider margins more strongly:

1 n
G= ∑ (vi − si )2
2 i=1

25
Julian Bernauer, Nathalie Giger, and Adrian Vatter

Federalism (fed) and decentralization (dec): The first two variables on the second,
federal-unitary dimension are the constitutionally and fiscally federal or uni-
tary character of the political system. In Lijphart’s (1999) original specifica-
tion, the two aspects are merged into one factor. Lijphart’s (1999) variable
distinguishes constitutionally federal and unitary states, which are further
divided into centralized and decentralized subclasses. The coding implies that
constitutional federalism is more important than fiscal decentralization, as
centralized federal states are regarded as more consensual than decentralized
unitary ones. We suggest treating both federalism and decentralization as
aspects in their own right, following the distinction between the right to act
(decentralization) and the right to decide (federalism) (Keman 2000: 199). The
first variable refers to the constitutional division of power between levels of
government, ranging from 0 (no federalism) to 2 (strong federalism). The sec-
ond variable is the degree of fiscal decentralization (Lijphart 1984: 178; Cas-
tles 1999; Keman 2000). It measures the leeway during implementation that
state and local governments have as opposed to federal governments. The
share of local and state taxes in total tax revenue is used for this purpose.
Bicameralism (bic): Another variable that logically belongs to the federal-
unity dimension is the bi- or unicameral character of the political system.
Federal and decentralized systems are often complemented by a second
chamber with varying degrees of power. Following Lijphart (1999: 211ff.),
unicameral and bicameral systems are distinguished, with a further quali-
fication of bicameral systems. The most consensual type corresponds to
systems with second chambers at power parity with the first chamber and
partisan incongruence between chambers. Slightly less consensual are
chambers with asymmetric power and partisan incongruence or symmet-
ric power and partisan congruence, while systems with power asymmetry
between chambers and partisan congruence are classified as the most mod-
est form of bicameralism. The index ranges from a majoritarian 1 to a
consensual 4.
Constitutional rigidity (const): Additional features of power-sharing are locat-
ed in the constitution, the judicial system, and the central bank system. Con-
stitutional rigidity refers to the difficulty of changes of the constitution.
A five-point scale, following Lundell and Karvonen (2003), is used to meas-
ure  this aspect of democracy. The most majoritarian configurations occur
when the constitution can be changed by a simple majority in parliament.
­Intermediate rules require elections or referendums in addition to simple
­majorities, qualified majorities, or a combination of qualified majorities with
­referendums. The most consensual configurations foresee, for example, the
­combination of several rounds of qualified majorities in parliament with a
referendum.

26
 New Patterns of Democracy

Judicial review (jud): Judicial courts can be powerful veto players. The meas-
ure of the strength of judicial review ranges from the absence of judicial review
to strong judicial review. A three-category scale of the degree to which laws
can be reviewed by a constitutional court is used. The most majoritarian form
is no or weak judicial review, medium-strength judicial review represents an
intermediate configuration, and strong judicial review the most consensual
trait.
Central bank independence (cent): One of the more controversial variables in
Lijphart’s (1999) selection is central bank independence. Critics argue that it
has no logical connection to the rest of the variables of the federal–unitary
dimension (Taagepera 2003: 12). The empirical analysis below will reveal
more on this issue. The Central Bank Independence (CBI) index, developed
by Cukierman et al. (1992), consists of 16 weighted items which cover the
independence of central banks in the areas of the chief executive officer, pol-
icy formulation, the objectives of the central bank, and limitations on lend-
ing to the government. Higher scores indicate more independent, consensual
central banks.
Consensual direct democracy (dir): The final variable is completely new to the
typology of consensus and majoritarian democracy proposed by Lijphart
(1999). As argued above, direct democracy should be included in a typology
of consensual and majoritarian democracy. The scale ranges from absence of
consensual direct democracy to pronounced consensual direct democracy. It
considers both the forms and the use of consensual direct democracy. Vatter
and Bernauer (2009: 346) explain the index:
The index compiled contains points for the degree of consensualism in the direct
democratic provisions in the constitution and embodied in the decision rules as
well as for the actual use of direct democracy. For those which we have labelled
‘uncontrolled’ referendums, i.e. optional referendums and initiatives, 1 point
each was awarded if prescribed by the constitution. No points were counted for
plebiscites (ad hoc referendums), which are subject to the discretion of the head
of government. As an intermediate form, 0.5 points were given for mandatory
referendums. Turning to decision rules, 0.5 points were awarded for each variant
of direct democracy when a quorum of participation is required and 1 point when
a qualified majority is required. Finally, 1 point was awarded for the actual use of
mandatory referendums, optional referendums and popular initiatives, but not
for plebiscites.

2.3.2  Factor Analysis

In constructing indices of consensus democracy, we largely follow the meth-
od used by Lijphart (1999: 246), who performs a principal component

27
Julian Bernauer, Nathalie Giger, and Adrian Vatter

factor analysis with varimax orthogonal rotation on the single variables to

identify dimensions of democracy, and builds additive indices of the stand-
ardized variables which attach to these dimensions. The factor analysis is
based on average values of the political-institutional variables in 34 of the 36
countries of the CSES 2 for the whole period 1997–2006, as we are interested
in long-standing patterns of democracy.7 Data on Taiwan are not complete,
and Russia is excluded as it is not rated as fully democratic at times during the
observation period, e.g. by Freedom House.8
Instead of the principal component extraction method, ‘real’ factor analy-
sis is used by choosing the principal factor method over principal component
analysis. As the goal is to identify underlying dimensions, hence latent vari-
ables, factor analysis is more appropriate (Costello and Osborne 2005).9 The
analysis proceeds as follows: First, the factor analysis is used to extract factors
and their eigenvalues. Following Lijphart (1999: 246), the factors with eigen-
values above 1 are extracted and rotated using the varimax criterion.10 The
orthogonal (right-angled) rotation technique ensures that the factors are
independent from each other (uncorrelated). An inspection of the eigenval-
ues and alternative estimations with different rotation techniques are used as
robustness tests.11 Variables with split or lacking loadings are excluded from
the final factor analysis.
Two factor analyses are carried out. The first leaves out direct democracy in
order to closely resemble the original set of variables chosen by Lijphart (1999)
and allow comparison with his findings. These variables still feature improved
measurements and exclude a measure of corporatism, which has been criti-
cized as being not logically connected to the other variables (Taagepera 2003).
Improved measures are deemed appropriate, even or exactly when c­ omparison
is the goal, as invalid measures could be one source of divergent findings.
Table 2.1 presents the results of the factor analysis of ten political-­institutional
variables for 34 democracies, excluding direct democracy.
Two factors with eigenvalues above 1 are produced by the analysis. Their
inspection yields values of 1.95, 1.47, and 0.52 for the first three factors.
A  screeplot (not reported) confirms the dramatic drop in the score of the
eigenvalue after factor 2. Hence, the two-dimensional solution is clear-cut.
An alternative estimation using oblique rotation with the promax criterion
(allowing dimensions to be correlated) leaves the configuration of the dimen-
sions intact, and the loadings of the variables remain virtually unchanged. In
sum, the two-dimensional finding is very robust, also against an alternative
specification of the rotation method. The two factors largely resemble the
two dimensions reported by Lijphart (1999)—the executives–parties and the
­federal–unitary dimension—albeit in reversed order. The executives–parties
dimension (second in Table 2.1) encompasses the effective number of parties,
cabinet type (which loads moderately), executive–­legislative relations, and

28
 New Patterns of Democracy

Table 2.1.  Factor loadings of originally ten political-institutional variables in 34 countries

covered by the CSES 2, 1997–2006

Variable Factor I Factor II

Effective number of legislative parties (party) –0.64

Oversized and minority cabinets (cab) –0.41
Executive dominance (exeleg)   0.64
Electoral disproportionality (elec)   0.62
Federalism (fed) 0.85
Decentralization (dec) 0.66
Bicameralism (bic) 0.69
Judicial review (jud) 0.44

Note: Principal factor analysis performed; factors with eigenvalues over 1.0 extracted; varimax orthogonal rotation;
factor loadings above 0.3 reported, loadings above 0.5 bold. Russia and Taiwan excluded. Central bank independ-
ence excluded from the final analysis as it does not load significantly on any dimension; constitutional rigidity exclud-
ed from the final analysis as loadings split between dimensions.

electoral disproportionality with the expected signs. The only variable miss-
ing compared to Lijphart (1999) is corporatism. The federal–unitary dimen-
sion (first in Table 2.1) shares federalism (which has been split here to
explicitly measure constitutional federalism and fiscal decentralization),
bicameralism, and judicial review with Lijphart’s (1999) findings. Central
bank independence did not load substantively on any dimension, while
constitutional rigidity had its loadings split between dimensions. Both vari-
ables have been excluded from the final factor analysis. The second factor
analysis expands Lijphart’s (1999) universe of political-institutional features
and incorporates direct ­democracy (Table 2.2).

Table 2.2.  Factor loadings of originally 11 political-institutional variables in 34 countries

covered by the CSES 2, 1997–2006

Variable Factor I Factor II Factor III

Effective number of legislative parties (party)   0.57

Oversized and minority cabinets (cab)   0.59
Executive dominance (exeleg) –0.82
Electoral disproportionality (elec) –0.63
Federalism (fed) 0.82
Decentralization (dec) 0.70
Bicameralism (bic) 0.72
Constitutional rigidity (const)  0.48 –0.67
Judicial review (jud) 0.41
Consensual direct democracy (dir)   0.51

Note: Principal factor analysis performed; factors with eigenvalues over 1.0 extracted; varimax orthogonal rotation;
factor loadings above 0.3 reported, loadings above 0.5 bold. Russia and Taiwan excluded. Central bank independ-
ence excluded from the final analysis as it does not load significantly on any dimension.

29
Julian Bernauer, Nathalie Giger, and Adrian Vatter

Vatter (2009) and Vatter and Bernauer (2009) have introduced direct
democracy as an additional variable, and hypothesize that it constitutes a
third dimension of consensus democracy in combination with cabinet type.
We found some support for the hypothesis for OECD and (Western) EU coun-
tries. In the CSES sample of 34 countries analysed here, these patterns are
generally also supported by the factor analysis (see Factor 3 in Table 2.2);
however, constitutional rigidity also loads on the dimension. The eigenval-
ues are 2.20, 1.65, 1.07, and 0.60 for the first four factors. A screeplot (not
reported) confirms the considerable drop in the score of the eigenvalue after
the third factor. Hence, while the third factor is retained with limited confi-
dence, we are far away from a four-dimensional solution. An alternative esti-
mation using oblique rotation with the promax criterion (allowing dimensions
to be correlated) does not change the pattern of which variables load most
strongly on which dimension, while the loadings change marginally. These
changes seldom exceed 0.01–0.03 for the loadings of interest, with the excep-
tion of constitutional rigidity with changes up to 0.12 but still has its strong-
est loading on the third factor. The three-dimensional finding is remarkably
robust against the specification of the rotation method. In sum, we take these
results as an encouragement to accept the three-dimensional solution, but
also as a reminder that factor analytical solutions can be somewhat
arbitrary.
We have labelled the new third dimension cabinets–direct democracy. The
other two factors (1 and 2) largely resemble the dimensions found by
Lijphart (1999); the executives–parties (featuring party, exeleg, and elec, but
not cab) and the federal–unitary (featuring fed, dec, bic, and jud) dimension
(again in reversed order). As cabinet type is not primarily loading on the
former dimension, it is labelled parties–elections to distinguish it from
Lijphart’s (1999) executives–parties dimension. Exceptions from the pat-
terns found by Lijphart on the second dimension are central bank inde-
pendence as well as constitutional rigidity. In particular, central bank
independence does not load even modestly on any of the three dimensions
and is excluded from the final factor analysis. This backs up Taagepera’s
(2003) criticism that it constitutes a rather alien element in the set of varia-
bles. Policy convergence caused by the European Central Bank in the Euro-
pean Union further reduces variance on the variable. We therefore
recommend leaving it out in the computation of indices. A more compli-
cated case is constitutional rigidity. Constitutional rigidity has its loadings
split between the first and the third factors, with clearly stronger attach-
ment to the third. While we did not see theoretical reasons connecting con-
stitutional rigidity to the first dimension, it might be reasonable to expect
more amendable institutions where direct democracy is strong. According
to Lijphart (1999), it logically belongs to the second dimension. Given the

30
 New Patterns of Democracy

possible logical connection to the third dimension, and the empirically

stronger attachment to it than to the first, we include constitutional rigidity
in the computation of scores for the cabinets–direct democracy dimension
of consensus democracy in this sample.
The resulting indices of consensus democracy follow the findings of the
factor analyses. For each of the two or three dimensions found, the varia-
bles loading on it in the factor analysis are used to compute an additive
index. After aligning their signs to uniformly indicate consensus democ-
racy, variables are standardized, summed up, and standardized once again
to allow for comparison between indices. The resulting positions for
­countries are displayed in Figures 2.1 and 2.2, respectively. ­Figure 2.1
draws on the two-dimensional solution and displays the positions of coun-
tries on the executives–parties (x-axis) and federal–unitary (y-axis) dimen-
sions. ­Figure 2.2 draws on the three-dimensional solution and shows

DEU
USA
2
CAN

BRA CHE
MEX
Federal-unitary dimension

1
AUS
RUS BEL
POL
ESP DNK
PHL
FRA JPN ITA
0
HUN CZEROU SVN
KOR CHI
BUL SWE
NLD ISR FIN
–1 GBR IRL ISL
PRT NOR
ALB PER
NZL

–2
–2 –1 0 1 2
Executives-parties dimension

Figure 2.1.  Map of democracy using indices of variables loading on the dimensions of
consensus democracy (two-dimensional solution). High values indicate consensus
democracy. The country labels are centred on the scores of the countries. ALB: Albania,
AUS: Australia, BEL: Belgium, BRA: Brazil, BUL: Bulgaria, CAN: Canada, CHE: Switzer-
land, CHL: Chile, CZE: Czech Republic, DEU: Germany, DNK: Denmark, ESP: Spain,
FIN: Finland, FRA: France, GBR: Great Britain, HUN: Hungary, IRL: Ireland, ISL: Iceland,
ISR: Israel, ITA: Italy, JPN: Japan, KOR: South Korea, MEX: Mexico, NLD: Netherland,
NOR: Norway, NZL: New Zealand, PER: Peru, PHL: Philippines, POL: Poland, PRT: Por-
tugal, ROU: Romania, RUS: Russia, SVN: Slovenia, SWE: Sweden, USA: United States.

31
Julian Bernauer, Nathalie Giger, and Adrian Vatter

DEU
USA
2
CAN

CHE BRA
MEX
Federal-unitary dimension

1
AUS
RUS BEL
ESP
POL DNK
PHL
FRA ITA
0 JPN
SVN
HUN ROU CZE
KOR CHI SWE
BUL FIN NLD
ISR
–1 GBR IRL ISL
PRT
NOR
ALB PER
NZL

–2
–2 –1 0 1 2
Parties-elections dimension

Figure 2.2.  Map of democracy using indices of variables loading on the dimensions
of consensus democracy (three-dimensional solution). High values indicate consensus
democracy. Larger circles indicate more consensual traits on the third, cabinets–direct
democracy dimension. The country labels are centred on the scores of the countries.
Country labels, see Figure 2.1.

scores on the parties–elections dimension on the x-axis, and scores on the

federal–unitary dimension on the y-axis. The size of the bubbles in this
graph indicates scores on the third, cabinets–direct democracy dimension.
Note that, unlike in Lijphart (1999), higher scores always indicate stronger
consensus democracy.
The resulting picture largely resembles the findings reported by Lijphart
(1999: 248), in particular regarding the two-dimensional solution
(­Figure  2.1). Prototypes of majoritarian (UK) or consensus (Switzerland)
democracy are found in their respective corners, providing some face
validity. Japan takes a central position as in Lijphart (1999). As to be
expected given electoral reform aiming at higher levels of proportionality
at the beginning of the period under scrutiny, New Zealand has moved
clearly towards stronger consensus democracy on the executives–parties
dimension. While these observations regarding stability and change com-
pared with Lijphart’s (1999) findings are not spurious, note that a precise
longitudinal analysis of change in the type of democracy (see Lijphart

32
 New Patterns of Democracy

1999: 255) would need fully comparable data over time. The purpose of
this database is rather to provide tailor-made measures to complement the
CSES 2 data, with a focus on measurement validity and data availability.
Some countries appear on the map for the first time, including Albania
(rather majoritarian on both the federal–unitary and executives–parties
dimension), South Korea (rather majoritarian on the executives–parties
dimension), and Mexico (taking a central position).
The inclusion of direct democracy into the set of political-institutional fea-
tures yields a third, cabinets–direct democracy dimension and a slightly
altered map of democracy (Figure 2.2). As cabinet type is detached from the
first dimension, which is consequently renamed into parties–elections
dimension, the positions of countries change slightly. The cabinets–direct
democracy dimension (shown using bubbles of different size) provides addi-
tional information about consensus democracy shaped by direct democracy,
cabinet type, and constitutional rigidity. We observe that countries like Swit-
zerland, Denmark, Italy, and several post-communist democracies (e.g. Hun-
gary, Albania, and Poland) display consensual traits on this new dimension,
while others such as Germany, the United States, and in particular Great Brit-
ain are more majoritarian.
Both the two- and the three-dimensional solutions share core variables in
the first two dimensions, but each variant has a slightly different notion. In
any case, having the choice between measures of consensus democracy with
or without direct democracy should enrich our understanding of consensus
and majoritarian democracy and allow a more flexible use of the concepts,
depending on the research interest at hand.

2.4 Conclusion

The chapter has introduced a new dataset on political-institutional context

variables in 35 political systems covered by the second wave of the Com-
parative Study of Electoral Systems. In the spirit of Arend Lijphart (1999,
2012), the first aim was to analyse the (latent) dimensional structure of the
political systems. The innovations offered include a recent data set,
improved measurements of variables, a new set of countries, and the incor-
poration of direct democracy as a full-fledged institutional variable. A first
factor analysis, aiming at comparability with Lijphart’s (1999) solution and
excluding direct democracy, yielded an executives–parties and a federal–­
unitary dimension. Diverging from Lijphart (1999), interest group corpo-
ratism is not part of the first dimension for theoretical reasons (Taagepera
2003) and because of lacking data. The second dimension does not feature

33
Julian Bernauer, Nathalie Giger, and Adrian Vatter

constitutional rigidity and central bank independence, as both variables

have their loadings split between dimensions or low loadings. A second
factor analysis including a measure of direct democracy revealed that three
empirical dimensions of democracy arise from the correlational structure
of the data: a parties–­elections, a federal–unitary, and a cabinets–direct
­democracy dimension. The main difference from Lijphart’s (1999) two-
dimensional solution is that cabinet-type is relocated from the first dimen-
sion and constitutional rigidity from the second to form an independent
dimension with direct democracy. This reflects the expectation of wider
executive formulas in the face of veto threats represented by (potential)
referendums.
Hence, the new database provides an extension and update of the meas-
ures of consensus and majoritarian democracy provided by Lijphart (1999),
improving measurement validity where possible. Of the 36 democratic
countries in the second wave of the Comparative Study of Electoral Systems,
35 are fully covered (counting Russia), including democracies which have
not been mapped before. We find initial evidence of both stability and
change in the location of countries in the two-dimensional solution (approx-
imately) comparable to Lijphart’s (1999) work, and provide a theoretically
derived alternative solution including direct democracy. Combined with the
individual-level CSES data, the indices should facilitate answering a wide
range of research questions regarding the influence of macro-level type of
democracy on individual-level measures of representation, participation and
attitude.
The Lijphartian dimensions of democracy clearly facilitate parsimonious
statistical analysis by aggregating related political-institutional features into
distinct indices. A caveat is that mechanisms can get blurred through the
aggregation, for example, when variables have contradictive effects. We also
need to be aware of the differences between institutions and political behav-
iour (Ganghof 2005) as well as the more or less strong logical connections
between the variables comprising the indices (Taagepera 2003). For these rea-
sons, the chapters in the book will often move on to more disaggregated
forms of the indices. Future research could also attempt to find measures for
longer time periods and apply them to additional countries for a truly encom-
passing, longitudinally comparable database also allowing for the analysis of
change over time. Such research could also invest in new methodology, in
particular testing more powerful approaches to factor analysis such as Bayes-
ian variants (Quinn 2004; Treier and Jackman 2008). For now, the updated
and expanded measures of consensus and majoritarian democracy presented
should facilitate the kind of multilevel analyses of government forms and
performance pursued in this book.

34
 Appendix A.2.1.  Eleven political-institutional variables, their measurement, and data sources
Variable: majoritarian
vs consensual extreme
1) Party system: two-party
systems vs multiparty systems
2) Cabinet type: one-party
cabinets and minimal winning
coalitions vs oversized and
minority coalitions
3) Executive–legislative relationship:
executive dominance vs executive–
legislative power balance
4) Electoral system: disproportional
vs proportional electoral systems
5) Federalism: unitarism vs
federalism
6) Decentralization: centralization
vs decentralization
7) Bicameralism: unicameralism vs
bicameralism
8) Constitutional rigidity:
constitutional flexibility vs
constitutional rigidity
35
Measurement (range)
Effective number of legislative parties: Laakso
and Taagepera (1979) index (1–∞)
Share of oversized and minority cabinets
(0–1)
Modified version of Siaroff’s (2003)
11-element index of executive dominance
(0–18), omitting type of electoral system and
influence of committee members on parties
Gallagher index of disproportionality (0–100)
Degree of constitutional federalism (0–2)
Share of state and local taxes in total tax
revenue (0–1)
Scale of dispersion of legislative power (1–4)
Scale of the majority required for
constitutional amendment (1–5)
Sources
Armingeon et al. 2006; Armingeon and Careja 2004; Gallagher and
Mitchell 2008; Golder 2005; Norris 2002; Reilly 2007a  Comparative
Study of Electoral Systems; own calculations based on online databases1
Armingeon et al. 2006; Armingeon and Careja 2004; Bale and Van Biezen
2007; Katz 2003; Katz and Koole 1999, 2002; Koole and Katz 1998, 2000,
2001; Van Biezen and Katz 2004, 2005, 2006; own calculations based on
online databases1
Siaroff 2003; own expert survey for non-EU/OECD countries; USA,
Switzerland and Eastern Europe: own calculations based on documentary
analysis2
Armingeon et al. 2006; Armingeon and Careja 2004; Gallagher and
Mitchell 2008; Lundell and Karvonen 2003; Reilly 2007b own calculations
based on online databases1
Armingeon et al. 2006; Armingeon and Careja 2004; Lundell and
New Patterns of Democracy
Karvonen 2003
Brazil Ministry of Finance 2002; Llanto 2009; Korea National Tax Service
2009; Korea Ministry of Strategy and Finance 2009; data from Eurostat,
the OECD statistical office and the International Monetary Fund3
Armingeon and Careja 2004; Flinders 2005; Lundell and Karvonen 2003;
Vatter 2005; own calculations for Poland, Slovenia and Czech Republic
Lundell and Karvonen 2003; Israel: Lijphart 1999; Constitutions of
Albania, Brazil, Chile, Taiwan, Republic of China, South Korea, Mexico,
Peru, Philippines, Russia
36

Julian Bernauer, Nathalie Giger, and Adrian Vatter

Variable: majoritarian Measurement (range) Sources
vs consensual extreme

9) Judicial review: absence of
judicial review vs strong judicial
review
10) Central bank independence:
absence of central banks vs
independent central banks
11) Direct democracy: absence of
consensual direct democracy vs
pronounced consensual direct
democracy
Scale of the degree to which laws can be
reviewed by a constitutional court (0–2)
Cukierman index of central bank
independence, incl. European Central Bank
(0–1)
Scale of the forms and use of consensual
direct democracy (0–9.5); see Vatter (2009)
Armingeon et al. 2006; Lundell and Karvonen 2003; Roberts 2006; Siaroff
2005
Cukierman et al. 1992; Cukierman et al. 2002; Sadeh 2005, 2006;
additional sources4
Own calculations based on data from the Centre for Research on Direct
Democracy, Zurich; the Electoral Knowledge Network; Institute for
Democracy and Electoral Assistance; Initiative and Referendum Institute
Asia5

Notes: 1) <http://psephos.adam-carr.net>,<http://www.terra.es/personal2/monolith/00europa.htm> and <http://www.parties-and-elections.de>; 2) Standing orders of parliaments and

constitutions as e.g. published or linked on the assemblies’ websites (directory provided by the Inter-Parliamentary Union: <http://www.ipu.org>) and specialized literature, for Eastern
Europe in particular Kurian (1997), Kraatz and von Steinsdorff (2002), Norton and Olson (2007), and Ismayr (2004); 3)<http://epp.eurostat.ec.europa.eu>, <http://stats.oecd.org/Index.
aspx>, <http://web.worldbank.org/WBSITE/EXTERNAL/TOPICS/EXTPUBLICSECTORANDGOVERNANCE/EXTDSRE/0,contentMDK:20248616~ menuPK:2086395~pagePK:210058~piPK:210
062~theSitePK:390243,00.html#fiscal>; 4) Updated data set retrieved from <http://www-management.wharton.upenn.edu/guillen/files/CBI.Data.xls>; 5)<http://www.c2d.ch/>, <http://
aceproject.org/>, <http://www.idea.int/publications/direct_democracy/index.cfm>, <http://www.iri-asia.net/cms_file.php?id=99>.
 New Patterns of Democracy

Notes

1. The compilation of the database would not have been possible without the help
of Jonas Ertlmaier and Steffen Hurka. Earlier data sets have been collected with the
help of Jens Clasen, Ralph Wenzel, Rainer Stocker, and Stefanie Rall. We would
like to thank several members of the scientific community, including Alex Cukier-
man, Jessica Fortin, Timothy Frye, Joshua A. Tucker, and Tal Sadeh, for valuable
advice and generous access to data, and the numerous respondents of an expert
survey on executive–legislative relations.
2. Note that Lijphart (2012) provides temporal updates of Lijphart (1999), but with-
out significant changes to operationalizations or the sample.
3. This does not mean that we expect the complete absence of a relationship between
cabinet type and the elements of the parties-elections dimension. Future research
might well split cabinet type between the two dimensions.
4. The description of the measurement of single variables strongly draws on Vatter
and Bernauer (2009).
5. The index is developed for the comparison of parliamentary systems. For presi-
dential systems, most but not all items of the index apply, such as the power of the
prime minister. There is no clear answer to the question whether presidential sys-
tems have more powerful legislatures than parliamentary systems (Alonso and
Ruiz-Rufino 2007: 241). The exact consideration of the power of the president in
the index is an issue for further research, as well as the role of presidentialism in
the Lijphartian framework in general.
6. For some countries, responses on a small number of additional items are missing.
The additive index of executive dominance was reweighted using the number of
valid indicators for the countries with missing values in order to have a compara-
ble range of the index across different degrees of coverage.
7. Interest group corporatism is only available for the 26 advanced democracies. For
a factor analysis of 12 political-institutional variables in 26 advanced democra-
cies, including interest group corporatism, see the codebook of Vatter and Ber-
nauer (2010a).
8. With Russia, the value of the third factor drops just below 1, with similar loadings
of the variables.
9. Strictly speaking, factor analysis requires variables with interval levels of measure-
ment (Kim and Mueller 1978: 73). The ordinal variables in the data set do not meet
this requirement; and we have to assume that they are quasi-continuous. Gener-
ally, correlations are robust and can accommodate ordinal variables (De Vaus: 385).
10. Rotation using the varimax criterion causes the factors to be rotated in such a
fashion that the variance of the squared loadings per factor is maximized.
11. Alternative methodologies beyond the scope of this chapter and much farther
away from Lijphart’s (1999, 2012) original methodology but worth future investi-
gation include Bayesian factor analytical and item response theoretical methods,
which have a number of advantages such as the accommodation of both continu-
ous and ordinal variables and the acknowledgement of measurement error (Quinn
2004; Treier and Jackman 2008).

37
3

Meaningful Choices: Does Parties’

Supply Matter?
Bernhard Wessels and Hermann Schmitt

3.1 Introduction

Political representation is the core of democratic systems. Democracy should

lead to a representative parliament and/or a representative government.
Democratic representation has been conceptualized in different ways. There
are two principal alternative visions of representation, differing in how they
work and what they attempt to maximize—the majoritarian vision and the
proportional vision (Huber and Powell 1994: 291; Powell 2000; Wessels
2007). Shepsle called this the great trade-off between governance and repre-
sentation (Shepsle 1988). Many endeavours have been made to judge the
quality of democracy and democratic representation. The most influential
scholar regarding the quality and performance of the two visions of democ-
racy is without doubt Arend Lijphart. He has severely challenged the con-
ventional wisdom that majoritarian democracy is better at governing and
consensus democracy better at representing (Lijphart 1999: 275). Further-
more, he has shown that proximity between voters and governments is
higher in consensus than in majoritarian democracies. He notes that ‘the
smaller these two distances are, the more representative the government is
of the citizens’ policy preferences’ (Lijphart 1999: 288). Thus, his claim is
that—as far as one takes proximity as a measure of representation, as all
major empirical studies do (Miller and Stokes 1963; Converse and Pierce
1986; Esaiasson and Holmberg 1996; Miller et al. 1999; Schmitt and Thomas-
sen 1999; Powell 2000)—political representation works better in consensus
democracies.

38
 Meaningful Choices: Does Parties’ Supply Matter?

However, why should this be the case? Lijphart puts forward ten criteria rang-
ing from characteristics of party systems to those of central banks to differenti-
ate consensus from majoritarian democracies, but these do not offer an
explanation. Also, his chapter on electoral systems does not provide theoretical
insights concerning this question. It is not only self-evident but normatively
required that elections play the pivotal role in this regard. Therefore, one pos-
sible explanation could be that elections are more meaningful in consensus
than in majoritarian democracies. Meaningfulness implies at least two charac-
teristics: political supply should offer real choices in terms of different substance
and election results should translate into a distribution of seats and govern-
ment responsibility corresponding with the political ‘bias’ of the election
results.
That these are a natural characteristic of elections has been challenged in
the recent debates about the quality of democracy. The question arising from
these debates is about the ‘meaningfulness of “democratic elections”’ (Merkel
2004: 34). Introduced in the context of political representation studies some
30 years ago (Dalton 1985), this question is once again being addressed in
new approaches towards measures of democracy (Hadenius and Teorell 2005)
and empirical analyses of democratization processes (Lindberg 2006). Mean-
ingful elections are about the ‘control of the agenda’, i.e. that the demos or
people must have the opportunity to decide what political matters actually
are and what should be brought up for deliberation (Dahl 1989: 107).
At a general level, meaningful elections require at least three things: a) that
the institutional setting translates, in an appropriate way, vote distributions
into political power and policymaking, b) that election results are an indica-
tion of a political mandate with the respective political bias which implies
that political supply offers real and substantive alternatives, and, last but not
least, c) that voters make a reasonable choice between substantively different
policy packages at stake in elections.
In this chapter, an attempt is made to answer the question of under which
conditions elections are most meaningful by looking at the contribution of
institutions (point a), the supply side of politics (point b), and the demand
side (point c). Emphasis is placed on the conditions under which voters can
base their choices on proximity considerations. This is a core aspect in the
idea of ‘elections as an instrument of democracy’ (Powell 2000), because vot-
ing according to the match in political positions of a voter and a party or a
politician is exactly what democratic mandating implies and a properly work-
ing process of democratic representation demands. In this sense, this chapter
is an attempt to complement Lijphart’s perspective with choice-related pre-
requisites of a working democratic process.

39
Bernhard Wessels and Hermann Schmitt

3.2  Meaningful Elections and Meaningful Choices

The factors that could explain the degree to which elections and choices are
meaningful are manifold and located in different dimensions of a political
system. One strand of the discussion on meaningful elections argues from an
institutional perspective. Lijphart puts emphasis on the impact of the type of
democracy, i.e. the type of institutional setting, on the quality of representa-
tion. His central claim is that proportional and, even more, consensus democ-
racies perform better regarding policy performance in the public-good arena.
Thus, elections may be regarded as more meaningful in proportional political
systems and, in particular, consensus democracies.
A second strand lays emphasis on political actors. From this point of view, it is
important that actors (in the first instance parties, but also candidates) offer
meaningful choices in order to make elections meaningful for citizens. Mean-
ingfulness includes that there are choices, i.e. real differences between the
offered policies, and that these policies are regarded as viable, i.e. realistic enough
to be implemented, and the respective actors as capable of implementing them.
These two perspectives, that institutions matter or that actors matter, are
not mutually exclusive. On the contrary, a given set of institutions has impli-
cations for the structure of political competition and, thus, for the shape of
the party system and political-supply structures.
The mechanisms working behind the scenes—securing the link between
citizens and their representatives as well as their governments—can be best
described in terms of the responsible party model (RPM). Thomassen has
pointed out the following conditions on the supply side (Thomassen 1994):
political parties must present different policy alternatives to the voters and
the internal cohesion, or party discipline, of parliamentary parties must be
strong enough to enable them to implement their policy programme (condi-
tion 1 of the RPM). The prerequisites on the supply side are useless when they
are not met by the following prerequisites to be fulfilled at the demand side:
voters must have policy issue preferences (condition 2); they have to be aware
of the policy positions of parties (condition 3); they have to compare their
own position with those of the parties (condition 4); and they have to vote for
the party closest to representing their policy preferences (condition 5) (Pierce
1999: 9). Thus, all in all, the responsible party model demands that party rep-
resentatives develop their manifestos and try to realize the promises they
have made there (Schattschneider 1977) while voters choose accordingly.

3.2.1  Proximity Voting in Majoritarian and Consensus Democracies

In his seminal study, Patterns of Democracy, Lijphart (1999: 288) finds that ‘the
distance between the government’s position on the left–right scale and the

40
 Meaningful Choices: Does Parties’ Supply Matter?

position of the median voter’ and ‘the percentage of voters between govern-
ment and the median citizen’ are smaller in consensus than in majoritarian
democracies (see also Huber and Powell 1994).
This implies either that voters vote more according to their proximity to
political parties or candidates in consensus democracies or that political sup-
ply structures allow a higher proximity to voter groups in consensus than in
majoritarian democracies. The explanation for this is the following. In a
majoritarian one-party government, the median of the party most likely is
not at the median voter. The party may include the median voter. Otherwise
it would not be the party in government. In consensus democracies the domi-
nant form of government is a coalition. If the parties of a coalition are at the
median of their party voters, and their party voters have gained a majority,
the likelihood that the median position of one of the parties is closer to the
overall median voter is higher than in case of a one-party government in a
two-party system.
What are the mechanisms behind the characteristics Lijphart uses to distin-
guish these two types of political systems? His conceptual map of democracy
consists of two dimensions. One is the executive–parties dimension, the other
the federal–unitary dimension. The latter is hardly related to political parties,
competition, and governments and will thus be neglected in our analysis. The
executive–parties dimension accounts for the effective number of parlia­
mentary parties, the existence of minimal winning one-party cabinets, execu-
tive dominance, electoral disproportionality, and interest-group pluralism
(Lijphart 1999: 241–50). The first four characteristics are related to elections
and their outcomes, which is obvious except for the executive dominance.
Executive dominance, however, measures cabinet durability, which should
depend on election outcomes in democracies. How these four elements con-
tribute to proximity and, thus, representation is not discussed by Lijphart, but
general arguments are easy to find. Starting with the simplest one, propor-
tionality of an electoral system stands for the degree of distortion in the trans-
lation of votes into seats. The more proportional, the better the will of the
voters translates into the distribution of political power, and the more likely
sincere voting according to preferences. The effective number of political par-
ties is an indicator of the differentiation of political supply and, therefore, the
number of choices that are reasonable. Both elements are positively correlat-
ed with consensus democracy, whereas one-party governments and longer
cabinet duration are related to majoritarian democracies. In a very general
way, the executive–parties dimension is a measure of the structure of political
competition (centrifugal vs centripetal) as well as of the differentiation of
viable political parties and the rigidity of the electoral system.
A hypothesis on the effect of proximity voting as one possible determinant
of proximity between representatives and those being represented based on

41
Bernhard Wessels and Hermann Schmitt

these considerations reads like this: in consensual democracies proximity

voting is stronger because voting can be sincere (proportionality of the elec-
toral system) and political supply is more differentiated, allowing for a higher
relevance of proximity for the vote.

3.2.2  Proximity Voting and Political Supply

However, as important as quasi-institutional factors may be, the debate
about the shortcomings regarding meaningful elections in liberal democra-
cies is not so much about institutions but rather about actors, and the sup-
ply side in particular. One of the main recurring topics in public and
scientific discourse about the deficits and crisis symptoms of representative
demo­cracy is that parties become indistinguishable because of their charac-
ter of catch-all parties and their tendency to focus on the median voter. This
­critique—valid or not—is not directed at majoritarian systems alone. The
‘run to the middle’ is also a topic in the discussions about multiparty sys-
tems. In election times, statements about boring campaigns, a lack of clarity
concerning the political standpoints of the competitors, and their inclina-
tion to avoid debates on content by concentrating on leading candidates
are the rule rather than the exception. Thus, it is doubtful whether the
political supply really offers policy choices or, instead, elections are rather
like beauty competitions. Concerning the other side of the actor-centred
perspective, the demand side, it is quite often questioned whether voters in
general are acting reasonably or if politics has become so complicated and
citizens so little involved in it that they cannot really make meaningful
choices.
In contrast to studies relying only on voters’ perceptions of parties’ posi-
tions and in contrast to studies on political representation and correct voting
(Lau and Redlawsk 1997) that compare positions of voters and representa-
tives (Lau and Redlawsk 1997; Pierce 1999), we use party stances in election
platforms as the point of reference and yardstick. We are convinced that this
is not only an innovation in investigating the degree to which elections are
meaningful but also a more appropriate way of investigating the working of
the responsible party model. The reason for this belief is that in party democ-
racies parties are the agents of voters and the principals of the representatives.
Parties deliver election programmes that are binding for their representatives.
This is one of the central means to guarantee the chain of delegation by secur-
ing one of the conditions for the working of the responsible party model.
Thomassen has highlighted the relevance of cohesiveness of parties for the
functioning of the responsible party model. On which criteria voters are bas-
ing their choices depends on a variety of conditions of which the supply side
of politics may be the most important one (Schmitt and Wessels 2005). ­Robert

42
 Meaningful Choices: Does Parties’ Supply Matter?

Dahl has emphasized the choice aspect as a precondition for democracy and
democratic participation (Dahl 1971). If there is no choice, elections cannot
function democratically.
For a meaningful choice-set, nominally different choice options are not
sufficient. In addition, those options need to differ in content, because
elections are the means by which citizens collectively decide about the
political agenda to be pursued, the policies to be enacted, and the people
who should be in charge of this. An ideal democratic election provides a
substantive choice between competing policy proposals or political agen-
das. In normative terms, this is the most relevant dimension of elections.
In a meaningful choiceset, the available options differ with regard to the
policies the parties would pursue if they were to be in charge of government
after the election.
On the side of the voters, a sense of what they want to see established as
future policies is necessary in order to be able to evaluate the offers by
political parties, i.e. they must have preferences (condition 2 of RPM). Sec-
ond, they have to establish an expectation on the likely policy of each of
the available choice options. The yardstick for this can be retrospective or
prospective. Retrospectively, voters establish a sense of the course of future
policies on the basis of their experience with the course of past policies (Key
1966; Fiorina 1977). Prospectively, they look at the offers of the parties in
the election campaign. Prospective evaluations of the likely policies of dif-
ferent choice options can only rely on statements of intention as laid down
in election manifestos and other campaign material. Because voting is
about who will govern the people in the future, vote decisions regularly
build on the judgement of statements of intention (prospective) corrected
by retrospective evaluations in order to assess the reliability of an actor’s
offer. This is certainly a rather demanding task in terms of information
gathering and processing. As programme parties offer whole bundles of
issues, this task is even more demanding. Voters have to find a yardstick
that allows them to sum up and choose. From research on electoral behav-
iour it is known that voters use cues and shortcuts in order to reduce the
complexity of political life and choice options. One of the most powerful
cues and general criteria of differentiation between choice options is ideol-
ogy. Differences between concrete policy issues quite often relate to more
general approaches and strategic visions of politics, i.e. ideology. Therefore,
ideology is understood to serve as a proxy for policy in many theories of the
voting decision (Downs 1957). If there are differences between policy
options and voters have policy positions, they will be able to check the cor-
respondence between their demands and the political supply. The better
these conditions are met, the more likely it is that voters express this match
by their respective vote.

43
Bernhard Wessels and Hermann Schmitt

3.3  Variables, Hypotheses, and Design

3.3.1  Proximity Voting
The way in which voters make up their mind concerning choice options is at
the heart of electoral research. With regard to policies, a generalized measure
is quite often used to determine proximity voting, for example. This measure
is the left–right scale. The use of left–right positions as an approximation of
policy stances is not limited to research on voting. Research on political rep-
resentation makes use of it too (Huber and Powell 1994; Powell 2000). The
discussions about the meaning and meaningfulness of left and right in poli-
tics could easily fill a small library. The basic arguments for the importance of
left and right in politics are the following: left–right can be regarded as a
‘super issue’ serving as a generalization of a broad range of diverse issue posi-
tions and policy concerns. ‘Abstract principles like the left–right dimension
and the liberal–conservative dimension are generally seen as instruments
that citizens can use to orient themselves in a complex political world’ (Fuchs
and Klingemann 1989: 203). As a generalized political medium, the left–right
scale facilitates efficient communication and orientation in the political
sphere. This implies that new issues and dimensions of conflict have to be
and can be integrated into this schema (Fuchs and Klingemann 1989: 234).
This point has yet to be demonstrated. Under the condition that this is the
case, the proximity in perceived left–right positions of political parties and
the voter’s individual left–right position should positively influence vote
choice—the closer a party is to the voter’s position the more likely she should
vote for this party. This, however, demands that citizens make use of the left–
right schema. Empirically, there is little doubt that the vast majority of voters
do. On cross-country average, 82.9 per cent of respondents have a position on
the left–right scale. There is variation across countries, from a ‘low’ of 61.3 per
cent in Italy to a high of 96.1 per cent in the Netherlands.

3.3.2  Lijphart’s Consensus Democracy

Lijphart shows that proximities between voters and governments are higher
in consensus than in majoritarian democracies. That Lijphart speaks of gov-
ernments and not of parties—which would be more appropriate in propor-
tional systems—complicates the understanding of the mechanism that
Lijphart may assume to work. By considering for coalition governments in
consensus democracies the composition effect and comparing the likely out-
come to one-party majority government as done above, the mechanism may
work as follows. Higher proximity in consensus democracies can be a first
hint that political supply in consensus democracies is better suited for voters
to find a proximate offer. Inspecting the indicators Lijphart uses to define the

44
 Meaningful Choices: Does Parties’ Supply Matter?

‘executive–parties dimension’, it is obvious that these relate quite closely to

the general and long-term structures of political supply. The effective number
of political parties is an indicator of viable choices, electoral proportionality
an indicator for the degree of the translation of votes into seats—the more
proportional, the less strategic and the more preference based could be vote
choice, cabinet structure, and executive dominance indicators for the possi-
bilities blaming the governing party or parties. Here, Lijphart’s scale will be
used in the updated version by Vatter and Bernauer (see Chapter 2 in this
volume).
In a more structural and formal sense, Lijphart’s dimension thus measures
the general and persistent structure and differentiation of political supply as
it represents itself to the voters in a political system over a longer period of
time. Structure, however, does not tell much about content. Voting decisions
should relate to content in order to produce a reasonable mandate. Thus,
besides the differentiation of supply in mere structural terms, the question is
how the structures are filled with content.

3.3.3  Political Supply and Its Perception

Measuring positions of political parties is one thing. Measuring the relation-
ship between positions of political parties and voters’ perception is another
thing. Two ways of regularly measuring parties’ positions are often used:
expert surveys and content analysis of party platforms. Looking to proximity
between externally determined positions of political parties and a voter leaves
out the crucial factor that only the voter’s perception of a party’s position can
be relevant for proximity considerations of the voter. This calculus of voting,
however, can only lead to effective representation if voters’ perceptions of
parties’ positions are correct. Furthermore, proximity on the super issue left–
right should be a generalization from more concrete policy positions of the
parties. Thus, the question is as to which degree voters respecify policy posi-
tions of political parties on the left–right scale.
The expectation is that left–right proximity as a consideration in voters’
calculus can only count when there is clarity in political supply. Thus, the
more parties are able to package their policies in a way that voters can locate
the party easily on the left–right dimension, the more likely proximity con-
siderations will count for vote choice.
If it is true that policy packages of political parties are not put together by
chance but have a systematic relation to some overarching ideology, it must
be possible to demonstrate this. The Party Manifesto Project has dealt with
this question from its very beginning. Without doubt, it is possible to gener-
alize not only theoretically but in statistical analyses from the policy profiles
laid out in election programmes to positions of political parties in terms of

45
Bernhard Wessels and Hermann Schmitt

left and right (Klingemann et al. 2006: Chapters 1 and 4). However, this is not
sufficient. The voters must also be able to relate concrete policy options to a
general dimension like the overarching left–right schema. Whether this is the
case or not is investigated in the next section. Here, we concentrate on the
question of whether parties offer choices that are constrained by more gen-
eral dimensions.
For CSES 2 countries, the most recent data of the MARPOR project cover
177 election programmes of 23 countries, with 56 standard policy issues
coded by the project. For a dimensional analysis, 177 cases (party platforms)
on 56 issues would not result in reliable solutions (Costello and Osborne
2005). Thus, dimensions are extracted by means of factor analysis for each of
the policy domains, covering: 1. External relations; 2. Freedom and democ-
racy; 3. Political system; 4. Economy; 5. Welfare and quality of life; 6. Fabric
of society; and 7. Social groups (Klingemann et al. 2006: Appendix II, 186–90).
The number of policy categories across these domains ranges from 4 to 16.
For each domain, a factor analysis has been done separately. The criterion
for the selection of a factor as indicative of a generalized approach to policies
is an eigenvalue larger than 1. Applying this criterion, 14 factors can be
extracted covering five policy domains: 1. External relations; 2. Freedom and
democracy; 4. Economy; 5. Welfare and quality of life; and 6. Fabric of society.
Domain 3 (Political system) and domain 7 (Social groups) did not produce
factors meeting the criterion. The number of factors extracted per domain var-
ies from one (domain 2) to four (domain 4). Results are displayed in Table 3.1.
The result shows, on the one hand, that there is not a single dimension like
the super-issue left–right on which the policies can be projected perfectly. On
the other hand, the factor solution can easily be interpreted in terms of con-
tent. We only considered factor loadings of at least 0.5 for interpretation (see
Table 3.1).
That the factor solution is not unidimensional, however, does not imply
that the found policy domains do not relate to the most important ideologi-
cal dimension in politics, namely the left–right scheme. One may have some
informed prejudices about how the policy categories should relate to left–
right. The respective expectation is displayed in the last column of Table 3.1.
These expectations have been checked against the deductively derived left–
right score of the Party Manifesto Project by regressing it on the 14 factor
scores resulting in explained variance of the deductive left–right scale of 69
per cent, which supports the expected relationship. The policy categories
derived are obviously related to the general left–right dimension, thus leav-
ing the chance for voters to respecify what parties offer in policy terms in the
election campaign on the left–right dimension.
Thus, do voters’ perceptions meet with what the finding on election plat-
forms of parties suggests? The left–right schema is a moving target regarding

46
 Meaningful Choices: Does Parties’ Supply Matter?

Table 3.1.  Factor analysis of election programmes of political parties in 23 countries

Factor domain N of items Factor Factor content (criterion: Expl. var. Cumul. Hyp. LR
factor loading ≥ 0.5)

External 10 1 Military: positive 0.219 0.219 +

relations
2 European integration: 0.200 0.419 –
positive
3 Anti-imperialism: 0.186 0.605 –
anti-colonialism
Freedom and 4 1 Democracy: positive 1.000 1.000 –
democracy
Economy 16 1 Economic goals 0.156 0.156 +
2 Productivity: positive 0.143 0.299 +
3 Protectionism: negative 0.134 0.433 +
4 Keynesian demand 0.105 0.538 –
management: positive
Welfare and 7 1 Welfare state 0.344 0.344 +
quality of life limitation: positive
2 Welfare state expansion: 0.330 0.674 –
positive
3 Social justice: positive 0.326 1.000 –
Fabric of 8 1 Law and order: positive 0.292 0.292 +
society
2 National way of life: 0.287 0.579 +
positive
3 Social harmony: positive 0.249 0.828 +

Factor: criterion eigenvalue > 1.

Source: MARPOR data for election manifestos of 177 parties in 23 countries.

its content. It is continually adjusted and new issues and policies are respeci-
fied in terms of left and right soon after they occur. ‘This adjustment is made
possible by a gradual integration of new elements in the established meaning
repertoire of left and right, and the disappearance of elements that have
become irrelevant in competition for power and policy direction. Such a
change of meaning elements can be understood as a function of changes in
(a) their salience and (b) their alignment with one or the other side of the
struggle for political power’ (Van der Eijk and Schmitt 2010).
The basic assumption of Van der Eijk and Schmitt is that if voters use the
left–right schema for evaluating and sorting out the political supply of politi-
cal parties and are aware of the more specific policy offers of the political par-
ties, they would respecify these in terms of left and right in order to make
sense of them. Van der Eijk and Schmitt are realistic enough not to assume
that citizens actually care about election platforms of political parties: ‘While
we focus on parties’ public political pronouncements, and in particular on
the contents of their election manifestos, we do not believe that manifestos
are important as direct sources of information: citizens normally do not read

47
Bernhard Wessels and Hermann Schmitt

election manifestos. Still, those manifestos are important indirectly as their

main messages are spread via the media, the commentaries and the political
behaviour of party elites which itself is informed by and based upon the polit-
ical direction that has been commonly defined in the election manifesto of
their party’ (Van der Eijk and Schmitt 2010).
In line with the general idea of Van der Eijk and Schmitt but on a somewhat
different route (because we are not so much interested in the meaning of left
and right for voters, but in the clarity of political supply), we attempt to deter-
mine the degree to which a party system provides the opportunity for voters
to respecify policy issues on which parties campaign on (as expressed in the
election platforms) in left–right terms.
To the stacked dataset of voters we added the 14 factor scores from the
party manifesto analysis for each available party. These 14 factor-score vari-
ables extracted from the party manifestos serve as determinants of respond-
ents’ perceptions of the left–right position of political parties. Thus, the
left–right position of political parties as perceived by voters is regressed on
the 14 manifesto-content variables. Compared to the analysis of Van der Eijk
and Schmitt with 320 aggregate cases, even at the micro level with about
31,000 real and more than 147,000 stacked cases, the results strongly support
the idea that voters are able to respecify complex policy packages with many
concrete policy proposals on the left–right scale. The factor variables ‘explain’
21 per cent of the variance in the perceptions of left–right positions of politi-
cal parties (Table 3.2). Furthermore, our expectations as to how the 14 poli-
cies expressed by the factors should relate to the left–right scheme finds
almost 100 per cent support. Out of 14 coefficients only one has the wrong
sign (factor 4.1, expressing ‘economic goals’). However, there is the possibil-
ity that this result is not the product of respecification of manifestos’ policy
content on the left–right dimension but a result of the perception at a more
general level, namely the left–right position of parties. In this case, the causal
chain would be the following. Parties have left–right positions. They con-
strain these policies respectively. Thus, there is a relation between the left–
right dimension and policy content produced by the parties themselves. If
voters have a more or less correct perception of parties’ left–right position
this would also result in a correlation of voters’ left–right perception of party
positions and parties’ policy content. In order to check for this indirect effect
which would imply that voters do not respecify at all but parties have done
the job, two more regressions have been performed. The first regression
checks the relationship between voters’ perception of parties’ left–right posi-
tion and the left–right score for parties resulting from party manifestos. We
use here the established RILE measure introduced by Ian Budge and included
in the party manifesto dataset. The second regression includes the 14 factor
variables representing the policy content of the manifestos and the RILE

48
 Meaningful Choices: Does Parties’ Supply Matter?

Table 3.2.  Regression of individually perceived left–right positions of political parties on

factor scores of the dimensional analysis of 114 election programmes in 23 countries

Factors Content  Coef. Std. err. t Emp. LR Hyp. LR Match

f1_1 Military: positive 0.42 0.009 44.86 + + y

f1_2 European integration: –0.22 0.009 –24.92 – – y
positive
f1_3 Anti-imperialism: –0.32 0.006 –50.79 – – y
anti-colonialism
f2_1 Democracy: positive –0.68 0.014 –47.89 – – y
f4_1 Economic goals –0.29 0.013 –23.10 – + n
f4_2 Productivity: positive 0.13 0.008 16.36 + + y
f4_3 Protectionism: 0.11 0.007 14.54 + + y
negative
f4_4 Keynesian demand –0.09 0.011 –8.43 – – y
management:
positive
f5_1 Welfare state 0.20 0.009 22.87 + + y
limitation: positive
f5_2 Welfare state –0.14 0.008 –17.20 – – y
expansion: positive
f5_3 Social justice: –0.60 0.010 –60.72 – – y
positive
f6_1 Law and order: 0.24 0.008 31.48 + + y
positive
f6_2 National way of 0.29 0.006 45.31 + + y
life: positive
f6_3 Social harmony: 0.21 0.008 26.40 + + y
positive
_cons Constant 4.80 0.008 570.94 13 of 14
R-squared 0.21

Note: Number of observations = 147,674; F (14, 31,202) = 213,251; Prob > F = 0.0000; Root MSE = 2.6256 (Std. err.
adjusted for 31,203 clusters in id).

measure as control. If the relationship between voters’ left–right perception

of parties and the 14 policy content variables were an indirect effect, i.e. the
result of the specification of policy content by parties and not of a respecifica-
tion of the policy content on the left–right scale by voters, explained variance
of the regression including the 14 factor variables could not be higher than
from the regression of perceived left–right position of parties on the left–right
position of parties as measured with the RILE scale. Results are as follows.
Regressing left–right perception on RILE shows an explained variance of
0.167. Table 3.2 shows a R2 of 0.210 for regressing the left–right perceptions
of parties on the 14 factor variables representing policy content. Including
the RILE scale in that regression as a control results in a R2 of 0.246. All 14
factor variables except one still show highly significant effects (P>|t| < 0.001).
Only factor variable 6_1 representing law and order policies loses its signifi-
cance (P>|t| < 0.057). These results allow the following conclusions. Because
explained variance for the factor variable model is higher than for left–right

49
Bernhard Wessels and Hermann Schmitt

position of parties, policy-content adds to the explanation of the perception

of left–right positions of parties. This implies that voters indeed respecify
policies offered by parties at least partly by themselves. How much use voters
make of policy contents to arrive at their perception of parties’ left–right posi-
tion and how much results from a general perception of parties’ left–right
position cannot be determined. The regression results offer two interpreta-
tions. Taking the factor variable model and adding the left–right position of
party manifestos increases R2 from 0.210 to 0.246. In this case, generalized
party positions as derived from RILE increase the explained variance by 3.6
percentage points. Taking explained variance of the full model as baseline
factor, variables have a share of 85 per cent of the total explained variance
and RILE the remaining 15 per cent. Taking the parties left–right positions
model (RILE) and adding the 14 factor variables increase R2 from 0.167 to
0.246, i.e. by 7.9 percentage points. In this case, RILE has a share of the total
explained variance of 68 per cent, and the 14 factor variables the remaining
32 per cent. Whereas the first calculation represents the upper bound for the
strength of the effect of policy contents, the second represents the lower
bound. The real contribution of policy contents to the left–right perception
of parties thus may lie somewhere between 32 and 85 per cent of the total
explained variance. Even if clear causality cannot be established, results clear-
ly show that policy content in party manifestos is related to the perception of
left–right positions of parties by voters.
These findings underline the character of the left–right schema as a univer-
sal means of communication in politics and its potential for respecification of
almost any political content. The relevance for our understanding of how
voters deal with the complexity of political life and supply cannot be overes-
timated. In terms of the responsible party model, the results show that voters
are able to structure political issues allowing for a more or less parsimonious
way to check policy offers and policy packages against their own preferences.
Results clearly point in this direction. It should be evident that even if voters
are able to constrain parties’ policy offers on the left–right scale, a check with
their own personal preferences could only work if voters also conceptualize
the latter in left–right terms.
In terms of the clarity of political supply to voters, we assume that clarity is
higher where the relationship between the policy offers of political parties as
expressed in the 14 factor score variables derived from election platforms and
the perception of left–right positions of parties is stronger. Thus, the higher
the explained variance of perceived parties’ positions and the manifestos’
content, the higher the clarity of political supply because left–right packaging
is easier. The measure for clarity of policy positions of political parties is the
adjusted R-squared of the country-wise regression of perceived party position
and the manifesto factor score variables.

50
 Meaningful Choices: Does Parties’ Supply Matter?

0,80
0,76
Adj. R-squared; regression of LR on Party Programms 0,72
0,70 0,66
0,63 0,64
0,57 0,59
0,60
0,52 0,53
0,50
0,50
0,43
0,38 0,38 0,39
0,40
0,33 0,34
0,29
0,30
0,22
0,18 0,20 0,20
0,20

0,10 0,05
0,03

0,00
ROU KOR ITA USA AUS SVN IRE DEU NZE BGR FIN FRA POL PRT CHE NOR NLD DNK HUN ESP PRT SWE CZE

Figure 3.1.  Clarity of policy positions of political parties: R2, regression of perception
of parties’ left–right positions on 14 factor scales from election manifestos

As expected, there is huge variation in our measure of clarity of policy posi-

tions of political parties. It ranges from a low in Romania of only 3 per cent
bound variance to 76 per cent in the Czech Republic (Figure 3.1).

3.4  The Impact of Political Institutions and Political Supply

on Proximity Voting

How much influence can proximity exert on voting, and does this influence
vary across contexts? This is the basic question, and its relevance for voters’
mandate to parties by elections is obvious. Thus, the question here is: Under
which conditions are elections meaningful with regard to two aspects: pro-
viding voters with the opportunity to choose among clearly presented alter-
natives, and providing a substantive mandate for the party or parties elected
regarding the content the election result carries?
Results so far indicate that the conditions for meaningful choices may be
better than the critics of contemporary democracy would suggest. One indi-
cation is that there is differentiation of political supply as expressed in the
election platforms; a second is that the overwhelming majority of voters take
generalized political positions. A third consideration is that political offers at
election times as expressed in election manifestos may not be too complex
and/or remote from citizens’ perceptions because citizens may generalize
from rather specific proposals and policy categories to the abstract left–right
scheme in a rather reasonable way.
The hypotheses arising from these observations are straightforward. One
refers to general features of a political system, namely the degree of consensus

51
Bernhard Wessels and Hermann Schmitt

democracy; the second to the specific political supply offered in election cam-
paigns; a third to expected differences in the effect of the institutional and
the supply characteristics.
Lijphart has found that the left–right distance between voters and govern-
ments is smaller in consensus than in majoritarian systems. He gives little to
no specific argument or reason why that should be the case. From his approach
to political systems and his arguments, however, one may assume that this
has to do with the formal structure of the party system and governments as
an expression of the differentiation of political choices. This argument would
be in line with his more normative inspired claim that consensus democra-
cies are better suited for political representation because of their goal to rep-
resent the many and not only the majority. Because differentiation of political
supply provides the chance of more proximate choices, we may expect a
larger impact of proximity perceptions on the vote in more differentiated
supply structures. The hypothesis is as follows:
H1: In more consensual political systems, proximity between a voter’s left–right
position and the perceived left–right positions of political parties plays a more
important role than in majoritarian systems.

The second hypothesis complements this hypothesis. The argument is that

choices are, or should be, about content, that is, policy concerns at a given
election. This requires a minimal degree of clarity in what parties offer. Clar-
ity is defined here in the sense that the policy options offered by political
parties can be more or less easily respecified on the left–right scale. The meas-
ure for clarity of policy positions of political parties is the explained variance
from regressing voters’ perception of party positions on the 14 factor score
scales derived from election manifestos. Hypothesis 2 reads as follows:
H2: Thus, the higher the clarity of policy positions of political parties, the stronger
are proximity considerations for the vote choice.

The third hypothesis relates to the differences in the strength of the effects of
the two context characteristics for proximity on the vote. Because the institu-
tional or structural characteristics of consensus democracy are not related to
policy content, but are an expression of the general degree of differentiation
of political supply, the viability of the structural alternatives, and strategic
bias demanded in the choice situation, the impact of consensus democracy
on the strength of proximity considerations for voting and its effect should
be limited. Compared to the moderating effect of Lijphart’s Consensus scale,
the moderating effect of clarity of policy positions of political parties should
be stronger.
As our hypotheses are formulated conditionally and we are interested to
know under which macro conditions choice sets are more (or less) meaningful,

52
 Meaningful Choices: Does Parties’ Supply Matter?

the unconditioned results, i.e. the individual-level only model, are of no great
interest here. The question of conditional effects is a multilevel question. As
Franzese has put it: Researchers looking at the interaction of individual-level
and macro-level factors—in our case, how consensus democracy and political-
supply characteristics interact with individual-level determinants—are ­actually
not interested in the constant and micro-level effects on vote choice as such
(Franzese 2005). Thus, the model includes cross-level interactions between
proximity—that is, the closeness between a voters’ left–right self-placement
and her perception of the parties’ positions on the left–right scale—and the
consensus democracy scale and the clarity of policy positions measure. Fur-
thermore, in order to estimate the effect for all parties simultaneously, a stacked
data matrix is used. An individual appears in the data set as often as the choice
options. This allows for a binary model of vote choice. Because the data matrix
is stacked, the logistic model has to account for that. This implies a conditional
logistic regression model, which controls for the fact that an individual appears
as often as the choice options, and for the fact that standard errors are clus-
tered by samples, i.e. countries. Furthermore, the regression coefficients do not
tell much about the variation across contexts, i.e. across the macro variables
introduced in the interactions. In order to evaluate the variation of marginal
effects across contexts, we estimate the average marginal effect of proximity
across the values of the macro variables. Because a binary model is used, simple
linear forms of estimation as proposed by Brambor, Clark and Golder (2006)
cannot be used, but logistic ones. The basis of the conditional logistic regres-
sion model is a rather parsimonious voting model. It includes the spatial com-
ponent of proximity, a measure of partisanship, and an evaluative measure of
representation. For proximity left–right is used as already mentioned. For par-
tisanship, party like–dislike scales are used instead of party identification. The
reason is to take into account the variation in partisanship regarding all choice
options. This would not be possible with party identification. Party identifica-
tion in the stacked data matrix would be a variable with a binary outcome: one
for the party identified with, zero for all others. In order to take into account
the variation in party differentials, party like–dislike scales are used. The evalu-
ative measure regards the representational performance of parties. It carries
the information whether respondents identify one particular party by which
they feel represented.
The two latter measures can be regarded as simple controls because the
variable of interest here is proximity; more specific the variation of the effect
of proximity on vote choice conditioned by the clarity of political supply in
a party system.
Table 3.3, left panel, shows the results. However, our conditional hypothe-
ses concern the moderating effect of macro variables for the micro–micro
relationship between proximity and vote choice. We performed models with

53
Bernhard Wessels and Hermann Schmitt

Table 3.3.  Conditional logistic regressions: base model and cross-level interactions

Micro level only Model with cross-level

interactions

Coefficient P>z Coefficient P>z

Proximity (self to perceived party 0,177 0,000 0,145 0,000

position; 0 to 10)
Representation by a party (1 = yes) 1,967 0,000 1,954 0,000
Party like–dislike scale (–5 to +5) 0,631 0,000 0,636 0,000
Interaction proximity and Lijphart’s 0,105 0,025
Consensus scale*
Interaction proximity and clarity of 0,639 0,006
parties’ policy positions**
Number of observations 96,369 96,369
Prob > chi2 0,000 0,000
Pseudo R-squared 0,654 0,658

Note: Standard errors adjusted for 23 clusters (country).

* Consensus scale: Lijphart’s executive–parties dimension, updated by Vatter and Bernauer (see Chapter 2 in this
volume).
**Clarity: R-squared from regressing perception of parties’ left–right position on 14 factors from parties’ election
manifestos.

interactions of both macro variables with proximity separately, both includ-

ed, and, third, including the three-way interaction of Lijphart’s Consensus
scale, the clarity measure, and proximity in addition. The three-way interac-
tion term fails to reach statistical significance. Two models including only one
interaction term show statistical significance of the interaction in both mod-
els. Including both two-way interactions also shows the significance of both.
The three-way interaction model is a test for a combined effect of consensus
democracy and clarity of supply. Because the interaction is not statistically
significant, we can assume that the combination of the macro characteristics
is not particularly important for mediating the effect of proximity on vote
choice. The model estimates are displayed in Table 3.3, right panel.
Because interactions (not only) in binary models are complicated, interpre-
tation is rather difficult. A more intuitive way for understanding the effects of
interactions is provided by marginal effects and change in proximities under
varying macro conditions. From the model in Table 3.3, marginal effects have
been estimated in a first step. This serves the determination of statistical sig-
nificance in differences in probability change.
Binary models with the logistic function have the positive characteristic
that binary outcomes can be modelled. Compared to linear models they show
the disadvantage that improvements in probability are not linear across
­values of the independent variable. This implies to investigate the change
in  probabilities at every level of the independent variable. The cross-level
interaction model assumes that the change in probability by change in one

54
 Meaningful Choices: Does Parties’ Supply Matter?

unit of the independent variable varies across contexts. For 23 countries and
ten possible one-point changes in probabilities for the ten-point proximity
scale this implies 230 different estimates. In order to reduce this complexity,
average marginal effects over the 23 contexts have been estimated in order to
determine in which contexts a change in probability by a one-point change
in the independent variable, proximity, is statistically significant. This infor-
mation is used to judge the four measures in change in probability of vote
choice by an increase in proximity between a respondent’s left–right self-
placement and the perceived party position on the left–right scale. The four
measures of change in probability are: a) the mean change in probability over
the range of the proximity scale for one-unit change, i.e. between 1 and zero,
2 and 1, until 10 to 9; b) mean changes in probability looking only to values
above the midpoint of the scale, starting with the difference between 6 and
5; c) mean change in probability between 9 and 8 and 10 and 9; and d) change
in probability between 10 and 9.
How do the two contextual variables, i.e. the degree of consensus democ-
racy and the degree of clarity of parties’ policy positions, moderate the impact
of proximity on the vote? The hypotheses formulated above suggest that
both contribute positively but clarity more so. That is, the more consensus
democracy and the more clarity, the larger is the increase in probability of
vote choice for one-unit change in proximity. This would indicate that prox-
imity considerations are more important in consensus democracies, and
more important in a situation of clarity of supply.
Turning to the Lijphart’s Consensus scale as updated by Vatter and Ber-
nauer (see Chapter 2 in this volume), the results show no systematic variation
with the increase in probability. Thus, there is no systematic moderation of
the proximity effect on vote choice. Correlations between the four measures
of change in probabilities and the Consensus scale are not significant and,
furthermore, do not show the same direction across all countries as compared
to only those where average marginal effects show statistical significance
(Table 3.4). Countries for which the average marginal effects are not signifi-
cant include France, Australia, Ireland, and Korea—all countries with majori-
tarian characteristics.
In contrast, clarity of parties’ policy position returns a statistically signifi-
cant and consistent effect (Table 3.4): the higher the clarity, the larger the
increase in probability by one-unit changes in proximity. This is true for the
group of all countries and also for the subgroup for which the average mar-
ginal effects are significant. They are statistically insignificant in Romania,
Korea, Italy, and the USA—those countries which show the lowest clarity in
policy positions.
Figure 3.2 shows the clear structure of the moderating effect of clarity of
parties’ policy positions. Where clarity is low, the change in the probability of

55
Bernhard Wessels and Hermann Schmitt

Table 3.4.  Correlation between macro characteristics (Consensus scale and clarity) and
the change in vote probability by an increase in proximity

Change in probability by one-point Consensus scale* Clarity**

improvement in proximity
All countries/contexts

Correl. Sign. Correl. Sign.

Mean change, all one-point differences 0,06 0,79 0,46 0,03

for proximity between 0 and 10
Mean change, all one-point differences 0,12 0,58 0,47 0,02
for proximity between 5 and 10
Mean change, all one-point differences 0,22 0,31 0,64 0,00
for proximity between 8 and 10
Change by improvement of proximity 0,33 0,13 0,83 0,00
from 9 to 10
Only countries with significant average marginal effects
Mean change, all one-point differences –0,37 0,12 0,78 0,00
for proximity between 0 and 10
Mean change, all one-point differences –0,28 0,24 0,78 0,00
for proximity between 5 and 10
Mean change, all one-point differences –0,18 0,46 0,76 0,00
for proximity between 8 and 10
Change by improvement of proximity –0,05 0,83 0,78 0,00
from 9 to 10

* Consensus scale: Lijphart’s executive–parties dimension, updated by Vatter and Bernauer (Chapter 2 in this volume).
** Clarity: R-square from regressing perception of parties’ left–right position on 14 factors from parties’ election manifestos.

0,25
Mean increase of probability by one-point increase in

Broken line: all differences in probability

y = 0,26x + 0,01; R2 = 0,69
Solid line: only sign. Differences in probability SWE
DNK
0,20 y = 0,34x – 0,03; R2 = 0,61 ESP
HUN CZE
Proximity from 9 to 10

BGR PRT05
0,15
CHE
NLD
NOR
FIN
0,10 ITA
NZE POL
AUS
USA FRA
ROU IRE
0,05 SVN
DEU
KOR
0,00
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8
Clarity of policy position*

Figure 3.2.  Marginal effect of proximity on vote choice conditioned by the degree of
clarity of parties’ policy positions

56
 Meaningful Choices: Does Parties’ Supply Matter?

vote choice is close to zero when proximity increases from 9 to 10. It increases
steadily the higher the clarity. In countries like Sweden, Denmark, the Czech
Republic, and Spain where clarity in policy positions is highest, the change in
proximity from 9 to 10 produces a change in probability of vote choice of
about 20 percentage points.
Does this finding relate to the question of representation? It does for two
reasons. One is that in democracies with more clarity in political supply,
proximity effects on the vote are much stronger compared to party systems
with less clarity, meaning that vote choice is stronger inspired by proximity
considerations and is, thus, more related to a policy mandate. The second
reason is that there is also a systematic relation between proximity and the
clarity of supply. Remember, these are two independent measures: the rela-
tionship between policy stances as put forward in election manifestos and
the perceived left–right position of the parties, and proximity in terms of the
closeness between the left–right self-placement and the perceived party
position.
Results show that there is a systematic increase in proximity of about one
point from systems with low clarity to systems with high clarity (Figure 3.3).

9,5

ESP PRT02
9,0 NLD
DNK SWE
FIN BGR NOR
NZE CZE
DEU CHE HUN
IRE
Proximity to party voted for

8,5 USA PRT05

FRA
KOR SVN
POL
AUS
8,0
ROU

7,5

7,0 ITA

R2 = 0,39
6,5
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8
Clarity of policy position*

Figure 3.3.  Clarity of policy positions of political parties and mean proximity to the
party voted for

57
Bernhard Wessels and Hermann Schmitt

Clarity in political supply not only contributes to a stronger impact of policy

considerations on the vote, it also produces higher proximity. The correlation
of clarity with the mean of proximity is above 0.60. Again, the Consensus
scale fails to produce systematic variation (correlation 0.09).

3.5 Conclusion

Are elections meaningful, and under which conditions? Here, we concen-

trated on two dimensions related to the supply side of elections. One was the
perspective on consensus democracy and Lijphart’s ‘executive–parties dimen-
sion’ in particular. The starting point was his observation that proximity
between voters and governments is higher, thus representation better, in
consensus democracies than in majoritarian systems. The question is what
produces this proximity. Here, we concentrated on the voter side. Only if vot-
ing is meaningful in at least two perspectives does voting contribute to prox-
imity. One condition is that voters find substantive choices; a second, that
they vote for the most proximate political offer. Because Lijphart’s dimension
relates to the structure of political supply in a generalized way, an impact on
substantive differences in supply and on voting is likely. The respective
hypothesis was that proximity considerations are moderated by the charac-
ter of a political system: the more consensual it is, the stronger the impact of
proximity on the vote. This perspective was complemented by a view on the
substantive differences in political supply. Choice structures alone, as
expressed in the consensus measure of Lijphart, do not tell us much about
content. Vote choices must relate to content if elections are to be meaning-
ful. We looked to supply in a new way. Building on an idea and concept of
Van der Eijk and Schmitt, we did not (just) rely on voters’ perceptions of par-
ties’ positions. Rather, we took the policy proposals parties made in their
election platforms as a starting point. Building on data of the MARPOR pro-
ject—the successor of the Party Manifesto Project—we used the 56 standard
categories of their content analysis of election programmes and extracted 14
policy dimensions by means of factor analysis. We demonstrate that there is
a clear relationship between the policy dimensions in election platforms and
the perceived left–right position of a party. We take this as an indication that
the specific policy positions in election platforms can be respecified in terms
of the left–right dimension. There is huge variation across countries to which
degree this is the case and, thus, clarity of political supply differs. The impli-
cation of this finding is that the meaningfulness of choice sets varies accord-
ing to the level of clarity: The more clarity, the more meaningful the choice
set. The respective hypothesis was that the more meaningful a choice set, the
more relevant proximity considerations for vote choice become.

58
 Meaningful Choices: Does Parties’ Supply Matter?

Using pooled binary regression models with cross-level interactions, we

tested the hypotheses by investigating the marginal effects of proximity and
change in vote probabilities under varying conditions of the two macro char-
acteristics, i.e. the Lijphart Consensus scale and our clarity measure.
Our results do not support the assumption that consensus democracies
produce more proximity and better representation as Lijphart claims, because
proximity considerations and the relevance of proximity for vote choice are
higher in consensus democracies than in majoritarian systems. The mecha-
nism by which voting, thus elections, contributes to proximity between elec-
tors and the elected is not related to Lijphart’s measure.
Whereas the first hypothesis failed to gain support, the second clearly is
supported by the results. The degree of clarity of political supply moderates
the impact of proximity on the vote in a systematic and expected way. The
more that policy positions in election manifestos relate to the perceived left–
right position of a party, the more meaningful choices become. This is what
our results suggest. Our results also show that proximity increases the more
clarity there is in parties’ offers. The implication for representation and
democracy should not be underestimated.
For the linkage mechanism between citizens and representatives, our
results produce clear evidence that elections and vote choice contribute to
political representation in the sense of proximity under proper conditions.
Proper conditions include clear policy offers by parties, which can be pack-
aged in terms of left and right. The more clarity there is in what a party system
offers to a voter, the more important proximity considerations become and
the stronger the impact on vote choice. Thus, mandating political parties
works if parties provide the conditions. If the parties fail in this regard, voters
instead downplay policy considerations in their vote. Thus, meaningful elec-
tions in terms of clarity in choice options provide the best opportunity for the
production of political mandates, and they guarantee that the representa-
tional link works. If parties fail in this regard it is likely that representation
fails too.

59
4

Policy-based Voting and the Type

of Democracy
Diana Burlacu and Gábor Tóka

4.1 Introduction

It has often been argued that various features of what Lijphart (1999) calls
consensus democracy, such as proportional representation, multiparty sys-
tems, coalition governments, and significant opposition influence on legisla-
tion promote party–voter linkages based on policy and ideology, while
probably undermining accountability for performance in office. The latter, in
its turn, is often thought to be promoted by features like majoritarian elec-
toral rules, two-party systems, single-party governments, and executive dom-
inance over the legislature. In this chapter we examine these propositions
using the extensive empirical evidence provided by Module 2 of the Com-
parative Study of Electoral Systems (CSES) and a new set of measures for
Lijphart’s political system variables provided by Adrian Vatter and Julian Ber-
nauer for the purposes of this volume and described in Chapter 2. We depart
from an important thread of the prior literature that found that executives
and legislatures are slightly more representative of citizen preferences under
proportional than majoritarian systems (see e.g. Powell 2000; Golder and
Stramski 2010). Instead, we focus on a causally prior factor, namely the inci-
dence of policy-based voting, or, to put it in a way that is conceptually both
more accurate and better linked to macro phenomena, the responsiveness of
aggregate election returns to shifts of policy preferences among citizens. Sec-
tions 4.2 to 4.4 discuss our dependent variable, theoretical expectations, and
statistical models, respectively. Section 4.5 presents the empirical analysis
and section 4.6 concludes.

60
 Policy-based Voting and the Type of Democracy

4.2  Policy-based Voting among Citizens

At least some correspondence between voters’ policy positions and their

vote is widely considered essential for creating a normatively desirable
feature of democracy, namely policy-based linkages between citizens and
their political representatives. This is obviously not the only desirable
type of such linkages, and holding politicians accountable for perfor-
mance (a.k.a. ‘valence’ issues, or competence and integrity) may even
have a trade-off with the degree to which citizen preferences on policy
(a.k.a. ‘positional’, i.e. divisive, issues) impact the vote (Powell 2000: 165).
Yet a stronger impact of citizens’ policy preferences on the vote should be,
and is widely considered, desirable for democracy as a sign of those prefer-
ences exercising a stronger prospective control over policy choices by
elected elites.
However, comparatively evaluating the total impact of citizens’ policy
preferences on vote choices across a large number of democracies is not an
easy task. We approximate it by looking at how much a sizeable shift in
ideological self-placements among citizens on an abstract ideological scale
(left vs right) might have changed the aggregate distribution of votes
across parties. This focus on a single ideological dimension is justified
mainly by previous demonstrations in the extant literature of its useful-
ness for cross-contextual comparisons regarding the impact of policy-­
oriented voting (see e.g. Van der Brug 1997; Van der Eijk et al. 2005),
which in its turn is provided by the ability of the left–right semantics to
absorb whatever the relevant dimensions of policy and ideological differ-
entiation are among the parties in a given party system (see Fuchs and
Klingemann 1989; Kitschelt and Hellemans 1990; Van der Brug et al.
2009). At the same time, this focus on a single ideological dimension
allows us to root our theoretical expectations in the prior literature on
how political institutions impact ideological differentiation among com-
peting political parties. Golder and Stramski’s (2010) analysis, which also
used CSES data, already established that the distribution of left–right posi-
tions in legislatures of ‘proportional’ systems is somewhat more congru-
ent with the distribution of left–right positions in the electorate than the
same distribution in the legislatures of ‘majoritarian systems’. We seek to
add to their test by focusing on responsiveness of aggregate election out-
comes to shifts in citizens’ preferences, i.e. by examining whether such
differences in elite mass congruence between consensus and majoritarian
democracies can conceivably come about as a result of supposed institu-
tional influences on voting behaviour.

61
Diana Burlacu and Gábor Tóka

4.3  Polity-level Determinants

Following Duch and Stevenson’s (2008) classification of structural factors

conducive for retrospective economic voting and Knutsen and Kumlin’s
(2005) discussion of the mechanisms through which ideological polarization
among parties may impact voting behaviour, we expect that political institu-
tions impact policy-based voting among citizens via three key routes:

(a) the political supply—i.e. smaller or bigger political differences among

the parties, which is clearly a major influence on the incidence of ideo-
logical voting among citizens (Lachat 2008);
(b) the political communication of these differences from elite actors to
voters; and
(c) voters’ motivation to act on the ideological differences that they see.

List PR should facilitate greater ideological differentiation among parties

than majoritarian electoral rules (Cox 1990), as well as a more ideological and
less personality- and performance-based electoral competition (Katz 1980),
i.e. a stronger communication of existing differences to voters and the strong-
er priming of the latter on ideology. The same could probably be expected
from multipartism and power-sharing among partisan actors at the national
level, i.e. oversized multiparty coalitions, minority government dependent
on some cooperative opposition parties in the legislature, and a legislative
process that gives significant veto powers to the opposition. Compared to
single-party governments, two-partism, and executive dominance of the leg-
islative process, all these institutions ought to reduce the incentives for ideo-
logical convergence among the parties, help to highlight existing policy
differences, and probably prime citizens more on ideological and policy dif-
ferences than on credit and blame for performance. In line with an extensive
literature going back to Powell and Whitten (1993), we would expect that the
reverse applies to performance-based voting: the more majoritarian a coun-
try’s institutions, the stronger the impact of voters’ performance evaluations
on the vote will be.
It is less clear whether a country’s location along the ‘federal–unitary’
dimension of Lijphart’s typology can affect our dependent variable. Federal-
ism, bicameralism, judicial review, and decentralization should promote hori-
zontal accountability and power-sharing between political actors, which
might be instrumental for clarifying existing policy differences and under-
mining voters’ motivation for performance- (as opposed to policy-) based vot-
ing. Yet these expectations are not very strongly motivated because the
institutional features associated with this dimension of differentiation among
democracies concern relations between territorial units and functionally

62
 Policy-based Voting and the Type of Democracy

s­ eparated actors (such as judicial and executive power), rather than electorally
competing political parties. Hence one could equally expect them to create
divisions within parties and policy convergence across them, thus undermin-
ing policy differentiation between parties as well as its communication to
voters and voters’ motivation to act on perceived ideological differences.
All in all, we hypothesize a trade-off between the impacts of performance-
based and policy-based voting across countries. We further hypothesize that
Lijphart’s (1999) ‘executive–parties’ dimension of consensus democracy pro-
motes policy-based voting partly through ideological polarization between
the parties, but also independently of that. Finally and less clearly, country
locations along Lijphart’s federal–unitary dimension of democratic systems
should probably not affect policy-based voting either through ideological
polarization or independently of that.

4.4  Research Design, Data, and Measures

Our analysis aims at a better understanding of how various characteristics of

the political system explain differences across elections in the extent to which
policy preferences—as captured by citizens’ left–right self-placements—influ-
ence the vote. Given our interest in macro determinants of micro-level behav-
iour, we need observations that are comparable both across different
individuals acting in the same election, and across a significant number of
elections that took place in different political institutional contexts. The com-
parative data set that best meets the above criteria while also featuring suitable
indicators of citizens’ left–right placements and performance evaluations is
Module 2 of the Comparative Study of Electoral Systems project. This module
was administered as part of national post-election surveys to probability sam-
ples of the voting population following national elections that took place on
five continents between 2001 and 2006, yielding a total of 42 political con-
texts and 61,018 voters for our aggregate- and individual-level analyses,
respectively.1
The individual-level analysis consists in deriving comparable measures of
the impact of policy preferences on the vote choice for each of our 42 politi-
cal contexts, while the aggregate-level analysis simply regresses these meas-
ures on quantitative indicators of political system characteristics, taking into
account the statistical error with which these measures were calculated on
the basis of effect parameters in the individual-level models.
We could only do this via a single multilevel statistical model if the dependent
variable in our individual-level analysis had been identically coded across all
political contexts, for instance into ‘left’ and ‘right’, or into ‘pro-government par-
ties/candidates’ and ‘all other parties/candidates’. Such ­standardized measures

63
Diana Burlacu and Gábor Tóka

would, however, hardly do any justice to the complex differences in the nature
of electoral choices between consensus and majoritarian democracies.
Therefore the dependent variable in our individual-level analyses will be
Vote choice, a multinomial variable with a country-specific coding scheme
that simply lists the major alternatives on the ballot in the given election (e.g.
Labour, Conservative, and Liberal Democrat in the 2005 British general elec-
tion). For obvious reasons, the individual-level dependence of such a depend-
ent variable on policy preferences and other factors can only be estimated via
vote functions estimated separately for each of the 42 contexts in the analy-
sis. It would be both technically impossible and substantively meaningless to
estimate macro–micro interactions between the individual-level parameters
of such a model on the one hand, and political system characteristics on the
other. Therefore the individual- and aggregate-level analyses are carried out
in separate steps. The dependent variable in the second step is not a particular
parameter in the micro-level multinomial regressions but a quantity derived
with post-estimation data manipulation. Otherwise, however, the logic of
multilevel analysis remains intact (cf. Achen 2005; Long Jusko and Shively
2005 regarding methodological considerations; and Duch and Stevenson
2008 for a closely related example).
The individual-level analysis involves estimating a vote function separately
for each of the 42 contexts. The model posits a multinomial logit link func-
tion fn between the Vote choice variable on the left-hand side and its presumed
determinants, including Left–right self-placement, on the right-hand side. For
ease of reading, Equation 1 greatly simplifies the tedious notation for such a
model and replaces the maths of the logit link function with a simple and
generic reference to the fn link function:

Vote Choice = fn(B 0 + B1 ( Strength of party identification) +

+B2 ( Left − right self − placement ) + B3 ( Left − right self − placement 2 ) +
+B 4 (Government performance evaluation) +
+B5 ( Strength of party identification)( Left − right self − placement ) +
+B6 ( Strength of party identification)( Left − right self − placement 2 ) +
+B7 ( Strength of party identification)(Government performance evaluation) +
+B8 Female + B9 Age + +B10 Age 2 + B11 ( Education low w ) + B12 ( Education high ) +
+B13 ( Farm job ) + B14 ( Manual work ) + B15 ( Rural residence ) + B16 ( Income ) +
+B1 Devout + B18 ( Minority 1) + B19 ( Minority 2 )) (1)

The first problem that arises with the justification of this model is the notori-
ous dependence of both ideological self-placements and performance evalu-
ations on party preference itself, i.e. that our vote function may suffer from
an endogeneity problem (see e.g. Knutsen 1997 and Evans and Anderson

64
 Policy-based Voting and the Type of Democracy

2006). Since our derivation of the dependent variable for the aggregate-level
analysis crucially depends on the 20 B parameter vectors estimated for each
of the 42 political contexts with the above equation, this is an important
objection. We answer it in three ways.
First, prior cross-national research by Duch and Stevenson (2008: 123ff.)
showed that relative differences between elections in the amount of
performance-­oriented voting remain virtually identical whether or not the
estimates are ‘purged’ of endogeneity in performance evaluations. We
expect that the same applies to our measure of ideological voting, which
closely parallels Duch and Stevenson’s measure of the economic vote in that
they are both based on the net association between vote choice and a single
introspective question about the respondent—how s/he finds the economy
and where s/he is on a left–right scale, respectively—without using any
information about the ‘objective’ responsibility of parties for the economy
or, what would be conceptually equivalent to that, the ‘objective’ left–right
position of the parties. Second, we see no theoretical reason to suspect that
relative ­cross-national differences regarding performance vs policy influ-
ences on vote choices could be severely obscured by cross-national variation
in the extent to which our indicators of performance evaluations and policy
preferences are endogenous to vote choice itself. If anything, our intuition
is that endogeneity (i.e. the tendency to rationalize voting preferences in
performance-­oriented or in policy-oriented ways) increases with the true
effect of performance and policy considerations on vote choices. Should
that be the case, cross-contextual variation in the endogeneity of our meas-
ures of performance- and policy-oriented considerations do not alter rela-
tive cross-­national differences in the influence of these factors on vote
choices. Third, the endogeneity of these measures to party preferences must
logically be a largely non-existent or at least greatly diminished problem
among non-­partisan respondents. Therefore, our actual measures of the
impact of policy preferences and performance evaluations on the vote will,
in the analysis below, be based on estimates that we are making for such
respondents. For this last reason, the vote functions that we estimate with
individual-level data for each of the 42 electoral contexts include the inter-
actions of performance evaluations and policy preferences with Strength of
party identification as shown in Equation (1) above.
A further question is due about whether Left–right self-placement and Govern-
ment performance evaluation really achieve what our model may seem to expect
from them, namely to capture all relevant performance- and policy-related
considerations. We do not think so. To be sure, the left–right semantics is large-
ly avoided in favour of other terminologies (or political dimensions) in some
countries like the USA, and only in the Japanese data set do we have a supposed
local functional equivalent (a ‘progressive–conservative’ scale) to avoid the

65
Diana Burlacu and Gábor Tóka

probable Eurocentric bias in the measurement of the respondent’s ideological

position. Hence our Left–right self-placement variable may have an unequal abil-
ity to capture policy preferences in the different countries in the analysis.
While this line of thinking is worth exploring further, we note that the
validity of our analysis does not in fact depend on the excessively strong
assumption that Left–right self-placement and Government performance evalua-
tion capture all relevant performance- and policy-related considerations in
every country; not even on the slightly weaker assumption that they do so to
an equal extent in all political contexts. Rather, the right question about the
validity of our analysis is whether the undeniable and inescapable measure-
ment problems of our proxies are systematically correlated with political sys-
tem characteristics. Only if that is the case can our findings about the impact
of the latter on the former be misguided. At this point we see neither theoreti-
cal reasons to expect this, nor a feasible empirical way to study whether such
a distortion could emerge. Therefore we leave these important issues for fur-
ther study and proceed to probe the existing data.
Another question regarding Equation (1) concerns the question of whether
an omitted variable bias may impact our estimates about policy- and
performance-­oriented voting in the given systems. We respond to this concern
by including as many socio-demographic control variables in the model as
possible given the CSES 2 data set. We expect these variables to capture such
shared causes of vote choice on the one hand, and policy preferences or perfor-
mance evaluations on the other that, if left uncontrolled in the vote function,
may create spurious correlations between vote choice and the given political
attitudes. The reader may wonder why we do not add similar controls for polit-
ical attitudes. The reason is that our theoretical interest here is in the total
effect of policy preferences and performance evaluations on the vote. What
any other political attitude variable (for instance, issue attitudes, satisfaction
with democracy, or leader evaluations) would add is nothing else but some
rather more specific policy- and/or performance-related considerations. Thus,
these controls would influence the estimated impact of Left–right self-placement
and/or Government performance evaluation on vote choice for an entirely wrong
reason. Therefore the only additional political attitude variable that appears
next to Left–right self-placement and Government performance evaluation in our
model is the Strength of party identification, which appears there only because of
our intention to estimate the impact of the theoretically relevant variables on
vote for non-partisan respondents. As discussed above, this restriction seems
advisable given the inevitable endogeneity of both policy positions and per-
formance evaluations to vote choice among partisan voters.
Technical details about the meaning and coding of the variables entering
Equation (1) are presented in our online appendix.2 Obviously, many of these
variables (particularly income) came with missing values for many r­ espondents

66
 Policy-based Voting and the Type of Democracy

and we were concerned about the impact that dropping these cases might
have on the accuracy and efficiency of the estimates. Therefore we used the
Amelia 2 package of Honaker et al. (2007) to multiply impute all missing val-
ues in our individual-level data sets.3 During all individual-level statistical
analyses we made use of the country-specific socio-demographic weight vari-
ables deposited with the CSES data set, and consistently excluded everyone
who did not give a valid answer (other than did not vote, spoiled the ballot,
or voted blank/invalid) to the question about vote choice.
The first yield of our individual-level statistical analyses was a vast number
of multinomial logistic coefficients for each of 42 political contexts that are
of no theoretical interest here and cannot be presented for reasons of space.
To estimate what impact our two key theoretical variables have on vote choic-
es in the given context, we derived four quantities for each party/candidate
that make use of these initial regression coefficients. These four quantities
reflected the expected fractional share of the party/candidate among all vot-
ers in the analysis under the given model and the observed characteristics of
these respondents except that they all obtain the minimum score on the
Strength of party identification variable and, in the first case, their Left–right self-
placement moves one sample standard deviation to the left of its observed
value; in the second, their Left–right self-placement moves one sample stand-
ard deviation to the right of its observed value; in the third, their Government
performance evaluation drops one sample standard deviation below its observed
value; and, in the fourth, their Government performance evaluation rises one
sample standard deviation above its actual value.
The difference between the second and first of these quantities gives our
estimate about the Impact of policy preferences on the vote share of the given
party/candidate in the given electoral context; the difference between the
fourth and the third provides our estimate about the Impact of performance
evaluations on the vote share of the given party/candidate. To estimate the
statistical error regarding these coefficients that derive from the fact that we
base our estimates on random samples of the relevant population, we boot-
strapped the estimation of these two differences (see Efron and Tibshirani
1993 for a discussion of this method). Specifically, we took 200 random sub-
samples (with replacement) from each of the 42 samples in the analysis, and
re-estimated the parameters of Equation (1), the relevant sample standard
deviations, and the resulting estimates about the Impact of policy preferences
and the Impact of performance evaluations for each resampling. The bootstrap-
ping process provided us with 200 estimates for each of the 230 parties/can-
didates that we could distinguish between in our analysis.
Figures 4.1 and 4.2 give a sense for how these estimates look like using the
example of the countries that are the best (albeit in the majoritarian case
imperfect) examples of what Lijphart (1999) meant by majoritarian and

67
Diana Burlacu and Gábor Tóka

Figure 4.1.  The effect of changes in policy demand and performance evaluations on
the vote share of UK parties in 2005

Figure 4.2.  The effect of changes in policy demand and performance evaluations on
the vote share of Swiss parties in 2003

consensus democracy, respectively. The horizontal axis shows the estimated

impact of policy considerations, and the vertical axis the impact of perfor-
mance evaluations on the vote share of each party. Light dots indicate the
location of the 200 bootstrap estimates for each party. The party acronym is
printed exactly at the mean value of the 200 estimates for the given party.
What the chart signals above all is that very distinctive numerical estimates
are obtained for the different parties in both democracies, and that some par-
ties would greatly benefit or lose (by up to as much as 30 per cent of the total
vote) if a two standard deviations change occurred in everyone’s left–right
placement and/or satisfaction with government. The two figures also illus-
trate the difficulty of making comparisons across political systems using these

68
 Policy-based Voting and the Type of Democracy

party-level estimates. Probably against our theoretical expectations, the

expected impact of such a vast change in policy preferences in the electorate
is slightly bigger (a roughly 30 per cent swing in the case of both Labour and
the Conservatives) on the main ideological antagonists in the UK than in
Switzerland. However, it could be that the UK estimates are higher only
because of a method artefact, namely because the individual parties tend to
be smaller in multiparty Switzerland than in the two-and-a-half party system
of the UK. If so, the Swiss parties, taken individually, could not lose as much
from any swing of public opinion as their UK counterparts. However, since
more parties’ vote is affected by ideological change in Switzerland than in the
UK, the combined shift of the vote may actually be higher in the first than in
the second. Therefore comparisons about effects size may be misleading at
the level of individual parties. Furthermore, since the gains and losses of the
individual parties cancel out within a system, the party-level estimates can-
not be treated as independent observations in an aggregate-level analysis.
Because of these concerns we decided to aggregate the party-level estimates
into a single figure (for each of the 200 resamplings) for each political context
with the help of a modified Pedersen index. Proposed for the measurement of
aggregate electoral volatility in a country k, the original Pedersen index
summed up half the sum of the absolute value of the (positive or negative)
change in each party i’s share of the vote. This index is useful for our present
purposes but needs to be adjusted when applied to bootstrap estimates. In
Taiwan, for instance, the impact of policy preferences is quite close to zero and
some resamplings thus suggest positive while others suggest negative effects
of the same change on the same party’s share of the vote. If we simply summed
up the absolute values of the changes across parties disregarding differences
between estimates that point in totally different partisan directions, we would
obtain exaggerated figures about the total impact of policy preferences on
Taiwanese election outcomes. Indeed, we would inevitably obtain a positive
value of the Pedersen index for all resamplings, which would mistakenly sug-
gest statistically significant effects even where the effects are really not signifi-
cant at all and are entirely inconsistent in direction across the resamplings.
The key dependent variables of our aggregate-level analysis are thus defined
by Equations (2) and (3), which state that the impact of policy preferences/
performance evaluations in election k is calculated, for each resampling p
using the original Pedersen index (as half the sum of the absolute value of the
expected impact on each party i within the same k context), except that the
absolute value of the estimated impact is multiplied by an a kip factor that can
only be either plus one (when the estimated impact for party/candidate i in
context k in resampling p has the same sign as the mean estimate for the same
party across the 200 resamplings) or minus one (when the same two figures
have the opposite sign):

69
Diana Burlacu and Gábor Tóka

j
( Impact of policy preferences )kp = 1 / 2 ∑ ( akip ( Impact of policy preferences)kip ),
i =1

where
200
1 / 200 ∑ ( Impact of policy preferences)kip
p=1 ( Impact of policy preferences)kip
akip = ×
200
( Impact of policy preferences )kip
1 / 200 ∑ ( Impact of policy preferences)kip
p=1
(2)

j
( Impact of performance evaluations)kp = 1 / 2 ∑ ×
i=1

( akip ( Impact of performance evaluations)kip ), where

200
1 / 200 ∑ ( Impact of performance evaluations)kip
p=1
akip =
200
1 / 200 ∑ ( Impact of performance evaluations)kip
p=1

( Impact of performance evaluations )kip

×
( Impact of performance evaluations )kip
(3)

Figure 4.3 displays the 200 bootstrap estimates about the location of the 42
political contexts in the two-dimensional space formed by the impact of pol-
icy preferences and performance evaluations on the vote. Table 4.1 gives
mean estimates and their confidence intervals for each context. There are a
quite a few pairs and even triads of contexts for which the cloud of bootstrap
estimates clearly overlap, suggesting no statistically significant differences

Figure 4.3.  The party-level effects of changing policy demand and performance eval-
uations aggregated with Pedersen index

70
 Policy-based Voting and the Type of Democracy

Table 4.1 The estimated impact of policy preferences and impact of performance

­evaluations across 42 political contexts

Impact of policy Impact of performance

preferences evaluations

Acronym 95% confidence interval

of the bootstrapped mean

Albania (2005) AL 0.34 to 0.42 0.14 to 0.23

Australia (2004) AU 0.19 to 0.28 0.42 to 0.48
Belgium, Dutch speakers (2003) BE-F 0.20 to 0.27 0.29 to 0.34
Belgium, French speakers (2003) BE-W 0.34 to 0.42 0.08 to 0.15
Brazil (2002) BR 0.10 to 0.16 0.17 to 0.22
Bulgaria (2001) BG 0.26 to 0.33 0.12 to 0.17
Canada w/o Quebec (2004) CA-E 0.02 to 0.14 0.20 to 0.31
Quebec, Canada (2004) CA-Q 0.27 to 0.35 0.21 to 0.28
Chile (2005) CL 0.12 to 0.21 0.42 to 0.47
Taiwan (2001) TW01 0.08 to 0.15 0.24 to 0.30
Taiwan (2004) TW04 –0.02 to 0.04 0.48 to 0.53
Czech Republic (2002) CR 0.33 to 0.44 0.24 to 0.37
Denmark (2001) DK 0.40 to 0.44 0.23 to 0.28
Finland (2003) FI 0.37 to 0.43 0.14 to 0.22
France (2002) FR 0.25 to 0.38 0.15 to 0.26
West Germany (2002) DE-W 0.16 to 0.24 0.43 to 0.48
East Germany (2002) DE-E 0.23 to 0.29 0.37 to 0.42
Hong Kong (2004) HK –0.03 to 0.10 0.21 to 0.29
Hungary (2002) HU 0.27 to 0.32 0.32 to 0.38
Iceland (2003) IS 0.36 to 0.42 0.33 to 0.38
Ireland (2002) IE 0.19 to 0.24 0.23 to 0.28
Israel (2003) IL 0.34 to 0.41 0.19 to 0.29
Italy (2006) IT 0.39 to 0.47 0.11 to 0.22
Japan (2004) JP 0.11 to 0.18 0.21 to 0.27
Korea (2004) KR 0.25 to 0.32 0.21 to 0.28
Mexico (2003) MX 0.03 to 0.11 0.21 to 0.27
Netherlands (2002) NL 0.45 to 0.50 0.21 to 0.26
New Zealand (2002) NZ 0.28 to 0.34 0.28 to 0.35
Norway (2001) NO 0.48 to 0.52 0.09 to 0.14
Peru (2006) PE 0.18 to 0.23 0.05 to 0.11
Philippines (2004) PH 0.01 to 0.05 0.31 to 0.37
Poland (2001) PL 0.45 to 0.51 0.07 to 0.15
Portugal (2002) PT02 0.47 to 0.53 0.23 to 0.31
Portugal (2005) PT05 0.26 to 0.32 0.21 to 0.26
Romania (2004) RO 0.08 to 0.27 0.16 to 0.29
Russia (2004) RU 0.11 to 0.16 0.14 to 0.19
Slovenia (2004) SI 0.35 to 0.42 0.25 to 0.32
Spain (2004) ES 0.35 to 0.45 0.19 to 0.26
Sweden (2002) SE 0.42 to 0.47 0.29 to 0.36
Switzerland (2003) CH 0.32 to 0.40 0.18 to 0.25
Great Britain (2005) GB 0.25 to 0.33 0.17 to 0.25
United States (2004) US 0.19 to 0.27 0.32 to 0.39

* Confidence interval of the estimate is obtained through the fifth lowest and the fifth highest estimate across 200
resamplings from the individual-level survey data.

71
Diana Burlacu and Gábor Tóka

between these pairs and triads of cases. In the most extreme instance of such
similarities, we even had to display the acronyms for South Korea (KR) and
Portugal 2005 (PT-05) at some distance from the actual mean value of the
estimates for these contexts because otherwise their acronyms would get
mixed up in the chart with the one for Quebec (CA-Q). Yet by and large the
estimates are very distinctive regarding individual contexts, with each of
them appearing to be significantly different, at least in one of the two dimen-
sions, from a 90+ per cent majority of all other cases in our sample of 42
contexts.
Our key substantive question is whether political system characteristics
influence the incidence of policy-based voting. The aggregate-level analysis
of this question estimates the simple OLS regression model shown in Equa-
tion (4) below. The impact in question is the function of a β0 constant, the
weighted sum of the given election k’s score on an arbitrarily selected n num-
ber of political system characteristics, and an election-specific component (or
in other words residual error of the fitted values from the model). The con-
stant of the model and the election-specific components of the impact have
no theoretical relevance in our case. Instead, it is the weighting of the politi-
cal system characteristics by a set of regression coefficients called βm that
reveals what features of the political institutional context make the impact of
policy preferences bigger or smaller in elections. The model parameters can
be identified and their margin of error can be empirically estimated under the
relatively weak assumption that the εk election-specific error term is nor-
mally distributed around a mean of zero. The validity of the results of course
depends on the significantly more demanding assumption that our sample of
k elections was selected at random and is composed of independent events.

o
( Impact of policy preferences)k = β0 + ∑ βm ( Political system characteristics)mk + εk
m=1
(4)

Two variables on system characteristics—Executive–parties and Federalism–uni-

tarism, respectively—were provided by Adrian Vatter and Julian Bernauer for
this volume to locate a large number of contemporary democracies on
Lijphart’s two conceptual and empirical dimensions of majoritarian vs con-
sensus democracy. Annual values on the elementary variables making up the
two indices were standardized and averaged across the years from 1997 to the
year of the election covered by the CSES 2 survey data in the given country.
We estimate Equation (4) with 39 of our 42 macro-level cases.4 In addition,
our analysis makes use of a measure of Polarization, which shows the standard
deviation of the expert-estimated left–right positions of each country’s rele-
vant parties, with the parties weighted by their number of voters in the CSES

72
 Policy-based Voting and the Type of Democracy

Table 4.2  Multivariate regressions of the impact of policy preferences on political ­system
characteristics

b (s.e.) beta b (s.e.) beta b (s.e.) beta

Executive/parties 0.05*** (0.02) 0.42 0.05** (0.02) 0.37 0.03* (0.02) 0.26
Federal/unitary –0.04** (0.02) –0.35 –0.03* (0.02) –0.23
Polarization 0.08*** (0.03) 0.44
Constant 0.30*** (0.02) 0.30*** (0.02) 0.14** (0.05)
Adjusted R2 0.18 0.30 0.45

*** p ≤ 0.01; ** p ≤ 0.05; * p ≤ 0.10. All statistics aggregated across 200 resamplings for each of the 39 elections in
the analysis using Rubin’s (1987) rules and Carlin et al.’s (2008) ‘mim’ package.

2 survey. The source of the expert estimates about party locations was the
Macro Data Set accompanying the CSES 2 survey.5 Relevant OLS regression
estimates are displayed in Table 4.2 and will be discussed below.

4.5  Empirical Findings

The online appendix provides test results regarding the statistical significance
of the various effects in Equation 1 that involve left–right self-placement as
an indicator of policy preferences. It suffices here to summarize these results
just briefly. A simple direct effect of left–right self-placement on vote choice
is clearly significant in all but a few non-European electoral contexts, which
are the English-speaking provinces of Canada, Hong Kong, the Philippines,
and Taiwan (where the effect is of borderline significance). The additional
effects of the squared value of left–right self-placements are also statistically
significant in well over half the electoral contexts, especially often where the
number of parties/candidates in the analyses is relatively large. This makes
good theoretical sense as in conventional spatial models based on policy-
based voting some parties in the more complicated multiparty systems may
have the highest probability of support not on the extremes of the ideological
spectrum but somewhere closer to the centre. Therefore we feel that the
inclusion of this squared term in the common model for all political contexts
is well justified.
Taken together, the two interaction effects of left–right self-placement and
its squared term on the one hand and strength of partisanship on the other
are statistically insignificant in a majority of the 42 contexts. However, they
appear to register significant effects in a lot more cases than we would expect
to occur just by chance. In nearly all these latter contexts, the effects of policy
preferences tend to increase with the strength of partisanship, which is con-
sistent with our expectation that we see here a spuriously inflated effect

73
Diana Burlacu and Gábor Tóka

among partisans due to a greater endogeneity of ideology to party preference

in this group. This seems to support our decision that the estimates for non-
partisans should provide the best guide to the actual effect of policy prefer-
ences on vote choice. Last but not least, the comparison of the likelihood
ratio statistics presented in the online appendix regarding models 1 and 4
suggests that at least some of the above effects of policy preferences are statis-
tically significant in every single context except Taiwan in 2001 and, on a
closer call, in the Philippines.
Another and probably more interesting way of looking at the significance
of these effects is look at the bootstrapped estimates about the expected
impact of a two standard deviation shift in everyone’s policy preferences on
the vote share of individual parties. Since the Impact of policy preferences may
be either underestimated (due to the imperfect measurement of policy prefer-
ences) or overestimated (if left–right self-placement is endogenous to vote
choice even among the non-partisan), it is instructive to consider this issue in
a relative perspective, i.e. whether policy preferences or performance evalua-
tions exercise a bigger effect on the vote share of parties/candidates. Consid-
ering all 230 parties/candidates in the analysis, we find that a two sample
standard shift in performance evaluations would change (positively or nega-
tively) the average party’s vote share by 9.7 percentage points; while a simi-
larly large shift in policy preferences would change the same share by 10.4
percentage points (data not shown). Considering this evidence, one may
want to conclude that policy preferences and performance evaluations are
equally well (or equally poorly) reflected in election outcomes.
Of course, the picture varies considerably across individual parties/candi-
dates. Figures 4.1 and 4.2 illustrate the general trends that emerge. Appar-
ently there are rather big positive effects of the simulated two standard
deviation shift to the right among citizens on the vote share of right-wing
parties like the Conservatives in Britain or the Swiss People’s Party (SVP/
UDC). This is mirrored by the large negative effect of such changes on the
vote share of left-wing parties like Labour in Britain and the Social Demo-
cratic Party (SP/PS) in Switzerland.6 More centrist parties like the Christian
Democrats (CVP) in Switzerland or the Liberal Democrats in Britain are little
affected by this change: they may get a more or less different set of individu-
als voting for them when such massive ideological shift occurs in society, but
their overall share of the vote changes little in the process. In this sense the
vote share of these centrist parties says rather little about policy preferences
in society compared to what is revealed by the ups and downs of the dis-
tinctly left- and right-wing parties.
Similar trends emerge regarding the impact of performance evaluations.
The simulated increase in satisfaction boosts the vote shares of the UK gov-
ernment (provided by Labour at the time of the survey) and undermines the

74
 Policy-based Voting and the Type of Democracy

vote of the main opposition party. Meanwhile the vote share of the Lib Dems
would barely be affected by this change, presumably because they attract
medium-satisfied voters, whose proportion does not change much under the
given scenario. Most interesting is that in Switzerland, where all the parties in
the chart save the Greens (GPS/PES) and the ‘others’ have been part of the
federal government all the time since 1943, the vote share of parties is still
quite sensitive to shifts in performance evaluations, even if not quite as much
as in Britain. Apparently not all Swiss government parties were equally will-
ing and able to claim credit from voters satisfied with the government, and at
least one government party (the SVP/UDC) managed to attract a relatively
dissatisfied electorate in the 2003 election. Hence even under such a vastly
oversized coalition government as the one in office in Switzerland, elections
can still remain a barometer of public opinion with respect to satisfaction
with performance. Conversely, even in the relatively majoritarian democracy
of the UK, election results show more than just citizens’ satisfaction with
government. Since the parties stake out distinct ideological positions and the
voters apparently respond to that accordingly, their left–right position sub-
stantially influences the distribution of the vote over and above whatever
influence is exercised by (dis)satisfaction with government performance.
For reasons explained in section 4.4, we believe that the impact of political
system characteristics is better examined at a higher level of aggregation
than those of the individual parties, and for this reason we cumulate the
party-level effects with our slightly modified Pedersen index shown in Equa-
tions (2) and (3) into the 200 times 42 context-level estimates displayed in
Figure 4.3 and Table 4.1. The first remarkable finding here is the apparent
negative relationship between the location of electoral contexts along the
two dimensions of Figure 4.3 (significant at the p = 0.03 level).7 The negative
correlation (r = –0.34) is consistent with theoretical expectations and sug-
gests a weak trade-off between the extent to which citizens’ policy prefer-
ences and performance evaluations are reflected in election outcomes. The
contexts around the top-left corner of Figure 4.3 are characterized by strong
performance-oriented and weak policy-oriented voting, and include a con-
spicuously large number of non-European cases: the two Taiwanese elec-
tions, the Philippines, Hong Kong, Mexico, English-speaking Canada, Japan,
and Chile. On the opposite bottom-right quadrant of the figure we see an
equally remarkable concentration of European multiparty systems, with
Norway, Poland, the French-speaking part of Belgium, Italy, Finland, the
Netherlands, and Portugal in 2002 as probably the most extreme cases on
this end. It is tempting to infer that ideological self-placements (particularly
on the left–right scale) are probably much less appropriate proxies of policy
preferences in these non-European contexts, but at this point this remains
merely a speculation.

75
Diana Burlacu and Gábor Tóka

Table 4.2 presents our statistical evidence about how type of democracy
may alter the responsiveness of election outcome to policy preferences in the
electorate. Here, we regress the bootstrapped estimates of the Impact of Policy
Preferences on various sets of our key independent variables. Note again that
the level-1 margin of error regarding the true Impact of Policy Preferences in
individual elections is factored in the analysis using Rubin’s (1987) rule and
the ‘mim’ package of Carlin et al. (2008), which makes these results technical
equivalents of estimated macro–macro effects in a single-step multilevel
model.
The findings are largely consistent with our theoretical expectations. The
executive–parties dimension of consensus democracy is positively and sig-
nificantly associated with the responsiveness of election outcomes to policy
preferences in the electorate, and this effect retains borderline statistical sig-
nificance at p = 0.06 even when ideological polarization between the parties
is controlled for (data not shown), and p remains 0.06 when both polariza-
tion and country location on the federal–unitary dimension are controlled
for (see the rightmost panel in Table 4.2). The bivariate impact of the execu-
tive–parties dimension on polarization is positive (with R2 = 0.09; data not
shown) and borderline significant (p = 0.06). Polarization has a substantial
and significant direct effect on the Impact of Policy Preferences net of other
model variables. However, a little over half the total effect of the executive–
parties dimension is independent of polarization. This may support our
expectation that country locations on the executive–parties dimension also
impact the clarity with which party positions are communicated to voters
and/or voters’ motivation to respond to party positions on divisive policy
issues.
Somewhat unexpectedly, country locations on the federal–unitary dimen-
sion show a similarly sizeable effect on the Impact of Policy Preferences (see
Table 4.2) as well as on Polarization (data not shown). Federalist traits are
associated with less policy voting and less polarization than unitary ones,
though the differences regarding polarization do not reach conventional lev-
els of statistical significance in either the bivariate or multivariate specifica-
tions that we tried (data not shown). The negative effect of federalism on
Impact of Policy Preferences is also just borderline significance (see Table 4.2).

4.6 Discussion

The institutional features of consensus democracy appear to affect voting

behaviour, and there is a moderately strong trade-off between performance-
and policy-based voting across democracies. Both theoretical and empirical
uncertainty remains, though, about the influence of country locations along

76
 Policy-based Voting and the Type of Democracy

the federal–unitary dimension on the responsiveness of national election

outcomes to policy preferences in the electorate. If there is an effect at all,
then, according to our analysis, the impact is more likely to be negative than
positive. In contrast, features associated with the executive–parties dimen-
sion should, on theoretical grounds, make consensus democracies more
responsive to electoral preferences than majoritarian democracies. Our
empirical analysis suggests that this is indeed the case. The effect is at least
partly mediated by consensus democracy facilitating slightly bigger ideologi-
cal polarization between parties than majoritarian systems. Yet, while the
statistical significance of our findings is not very impressive, on both theo-
retical and empirical grounds we are inclined to think that consensus democ-
racy also influences the degree of policy-based voting in the electorate
independently of party polarization.
The 0.05 regression coefficient shown in the leftmost panel of Table 4.2
suggests, though, that these normatively desirable effects of consensus
democracy are very modest indeed. On this basis, we would expect the differ-
ence on the executive–parties dimension between Britain (with a score of
–2.44) on one extreme and Belgium (with a score of 1.62) on the other to
produce just about 1 per cent bigger aggregate change in election results
among non-partisan voters in the latter country when a massive two stand-
ard deviation change occurs in the left–right policy preferences of the elector-
ate. It is hard to see by what standard this tiny expected difference would be
worth attention from institutional designers.
As always, the possibility remains that these results merely reflect the
impossibility of identifying, across such a heterogeneous sample of democra-
cies, either the extent of policy-based voting with the help of the left–right
dimension, or clear effects of institutional design. However, in further analy-
ses we failed to find significantly stronger effects of the executive–parties
dimension on the Impact of Policy Preferences either in European rather than
non-European countries, or in old rather than new democracies. Neverthe-
less, we suspect that the relative irrelevance of the left–right semantics out-
side Europe does influence our findings: even with Israel and the Anglo-Saxon
world included, the impact of the executive–parties dimension on our meas-
ure of policy-based voting is essentially zero, and, in fact, negative. Yet, while
the Impact of Policy Preferences appears notably higher within Europe, it does
so almost irrespectively of institutional set-up, with even the bivariate impact
of the executive–parties dimension dropping to 0.03 and staying there when
we look only at Europe’s older democracies.8
All in all, the theories explored in this chapter receive some support from
the empirical data, but add little to explaining why the world’s democracies
show such a striking variation in policy responsiveness as Figure 4.3 suggest-
ed above.

77
Diana Burlacu and Gábor Tóka

Notes

1. We kept elections in the analysis even with a lower level of democracy as long as
they appeared to have been sufficiently competitive and influential for institution-
al variables related to the distinction between consensus and majoritarian democ-
racies to have any impact on the incidence of performance- or policy-oriented
voting behaviour. Hence, we only dropped one of the CSES 2 elections—the 2005
presidential election in Kyrgyzstan—from the analysis, because of the insufficient
variation on the vote choice variable caused by a candidate who collected about 90
per cent of all votes and doubts about whether vote counting was sufficiently fair
to award voters real influence on government formation. We treat as two separate
cases—though occurring in identical contexts as far as consensual vs majoritarian
traits of democracy are concerned—the elections for the French- and Dutch-­
speaking lists in Belgium, Eastern and Western Germany, and Quebec and the Eng-
lish-speaking provinces of Canada, because the major within-country differences
in their party systems would have made the estimation of identical vote functions
meaningless for these regions. While these cases are hardly as independent of each
other as elections in two neighbouring countries, they are substantially different
with respect to the key dependent variables in our aggregate-level analysis (see Fig-
ure 4.3). Introducing a correction for clustering for just three pairs of cases out of a
total of 42 seemed too much ado for too little and was thus avoided in the
analysis.
2. See <http://www.personal.ceu.hu/departs/personal/Gabor_Toka/Policy/>.
3. We created five imputed data sets for each political context separately. The imputa-
tions were based on a slightly larger set of variables than those listed in the online
appendix. Technical details are available from the authors.
4. Two elections in Taiwan and one in Hong Kong had to be dropped from this analy-
sis for lack of data on some institutional features.
5. Missing expert judgements were single-imputed using information about each par-
ty’s left–right position in all CSES 2 countries as estimated via the mean value of all
voters’ placements of these parties on a left–right scale in the CSES 2 survey, or, if
that was also missing, via the mean self-placement of each party’s voters in the
survey.
6. Of course, the negative and positive signs of the changes in Figures 4.1 and 4.2
merely indicate which parties would be differently affected by homogeneous move-
ments of the electorate in a particular dimension. The size of the effect would be the
same but the sign reversed if we simulated the impact of a similar movement in a
leftward (or, in terms of performance evaluations, dissatisfied) direction.
7. As in all aggregate-level analyses reported in this chapter, here too we report esti-
mates aggregated across the 200 sets of bootstrapped estimates following Rubin’s
(1987) rules, i.e. as if we were dealing with a data set with 200 multiply imputed
values for each observation.
8. The associated p-value is 0.11 in the analysis of all 26 European elections in our
analysis, and 0.13 when only the 13 elections in older European democracies are
considered.

78
5

Political Institutions and the Social

Anchoring of the Vote
Pedro C. Magalhães

5.1 Introduction

Does the fact that individuals share certain important social, demographic, or
group-membership features make them likely to vote in the same way? How
strong is structural voting—‘the extent to which party choice is determined
by voters’ structural positions’ (Van der Brug 2010)—in different countries?
And what explains why the vote is more socially anchored in some countries
than in others?
There are at least three main things that, in the last decades, electoral
research has established with regard to these issues. First, there seems to have
been a point in time in the history of (at least some) democracies where vot-
ing behaviour and party systems had strong social anchors. Famously, Lipset
and Rokkan (1967) suggested that social cleavages along religious, class, and
other lines had contributed, during democratization processes and in histori-
cally contingent combinations, to define groups of voters with conflicting
interests and values, as well as corresponding party labels and organizations
that represented them. Such alignments between voters and parties were
made stable through both political socialization and parties’ organizational
linkages to society, particularly through the role of unions and organized
religion. By the time Lipset and Rokkan wrote, it was argued that those align-
ments, reflecting processes that had taken place many decades before, had
become ‘frozen’, a diagnostic confirmed by subsequent works (Rose and
Urwin 1969). This seemed to square rather well with a particular view about
what allowed parties to perform a representative role in a democracy: a popu-
lar claim to representation (Saward 2010), based on the notion that mass

79
Pedro C. Magalhães

parties, rooted in social cleavages, could function as ‘agents’ or ‘vehicles’ for

stable and enduring social interests.
The second thing research has shown us is that, as ‘social cleavage ­theory’
was being proposed, this supposedly frozen ground was already thawing.
The signs became clear since the early 1970s and have accumulated with
time. The most telling indications were the increase in electoral volatility
detected in Western democracies (Pedersen 1979; Budge 1982; Dalton,
McAllister and Wattenberg 2002) and the declining importance of social
membership variables as predictors of vote choices in studies based on
post-election surveys (Franklin, Mackie and Valen 1992). To be sure, the
latter diagnostic is not entirely unanimous. It has been argued that several
fundamental social markers of one’s resources, values, and interests con-
tinue to be or have even increasingly become—as in the case of gender—
significant correlates of the vote in several countries (Manza and Brooks
1998; Evans 2000; Brooks, Nieuwbeerta and Manza 2006). However, when
we move our attention from the effects of particular variables to our overall
ability to explain vote choices in elections on the basis of socio-structural
variables, most of the evidence points to the notion that our combined
knowledge of voters’ occupational status, organizational memberships,
religious affiliation, and religiosity, for example, has become decreasingly
useful to account for the choices they make. Whether one treats the vote as
a choice between a party of the left or of the right (Dalton, Flanagan and
Beck 1984; Franklin, Mackie and Valen 1992; Oskarson 2005), as a discrete
choice without any pre-imposed unidimensional left–right structure (Nieu-
wbeerta and De Graaf 1999; Dalton 2002; Knutsen 2004; Van der Brug,
Hobolt and De Vreese 2009; Van der Brug 2010), or both (Franklin 2009),
the observed trend is, generally speaking, one of declining importance of
socio-structural variables.
The third basic finding that emerges from this literature is that the extent
to which the vote is socially anchored seems to vary widely between political
systems, independently of any underlying secular decline. All of the above-
cited studies that detected a ‘decline in cleavage politics’ have also observed
important cross-national variations. They have been less successful, howev-
er, in finding explanations for them. Much of the difficulty arises from the
limited sources of data that have been available so far. Until the emergence
of projects such as the Comparative Study of Electoral Systems (CSES), avail-
able evidence about the strength of socio-structural voting was limited to a
restricted number of comparatively well-studied advanced electoral democ-
racies with an established tradition of national election studies. Although
such ‘most similar systems’ design was well suited to the detection of trends
over time in a particular set of countries, it was less well suited for the exp­
lanation of broad cross-national differences. Furthermore, the degrees of

80
 Political Institutions and the Social Anchoring of the Vote

freedom available to scholars wishing to test hypotheses in any systematic

fashion remained small.
The main goal of this chapter is to address this shortcoming in the litera-
ture. Along the lines of the theoretical framework of this volume, I will exam-
ine the extent to which variations in political institutions help to explain the
extent to which countries vary in terms of the social anchoring of the vote in
legislative elections. In particular, I am interested in the role of those institu-
tional rules that structure how votes are converted into seats and govern poli-
cymaking and executive–legislative relations. It has often been argued that
those rules are likely to be consequential for structural voting. As Kitschelt
puts it, the decisions made by voters in elections are ‘refracted’ by the institu-
tional rules ‘that codify in a democracy what preference articulations count
(votes) and how they count in the choice of representative bodies and execu-
tives’ (Kitschelt 2010: 661). In other words, rules organizing elections and
policymaking are likely to have consequences for the extent to which stable
alignments between parties and socially defined groups of voters can be
formed and sustained through time. However, as I will explain later, the
results of the few existing empirical studies have been somewhat contradic-
tory (Norris 2004; Huber 2011). Furthermore, they have neglected to system-
atically examine the importance of institutional factors other than electoral
systems. Finally, I will confront findings about the importance of institu-
tional contexts for structural voting with other conventional hypotheses,
namely those focusing on the role of social modernization and a country’s
experience with competitive electoral democracy.
This chapter takes advantage of the CSES data in order to address these
questions using a relatively large—albeit only in comparison with most previ-
ous studies—number of democracies and elections. It provides, as far as I
know, the first multilevel analysis of the correlates of variations in structural
voting across countries. I proceed as follows. The next section—section 5.2—
presents and discusses theoretical arguments about the kind of fundamental
system-level characteristics of polities that may affect why the social anchor-
ing of the vote should be deeper in some countries. Section 5.3, after discuss-
ing the measurement problems involved in the study of structural voting,
presents the empirical analysis. In a first stage, I follow Huber’s (2011)
approach, originally developed for the study of ethnic voting, to estimate
indices that capture the extent to which, in a particular national election, an
individual’s vote choice can be predicted by simply knowing that individual’s
gender, whether he or she belongs to a union, his or her socio-economic sta-
tus, and his or her religious denomination and frequency of church attend-
ance. Then, in the second stage of the analysis, I examine the extent to which
political institutions and other system-level factors explain variations in
those indices. Section 5.4 concludes.

81
Pedro C. Magalhães

5.2  Why System-level Differences in the Social Anchoring

of the Vote?

What political, institutional, and social macro-level factors explain why, in

some countries, the social characteristics of voters are better predictors of the
vote than in others? In this section, I present four main hypotheses. While
some have already been explicitly advanced as explanations of cross-national
variations in structural voting, others, to my knowledge, have not, and
deserve to be carefully investigated.

5.2.1  Electoral Rules and Types of Democracy

In their seminal work on this subject, Lipset and Rokkan pointed out that, to
a great extent, political institutions were shaped by the configuration of
interests in society. The choice of an electoral system, for example, reflected
existing cleavages and the efforts of established parties to ‘consolidate their
position’ (Lipset and Rokkan 1967: 30). However, Lipset and Rokkan also
provided illustrations of how institutions ended up, in turn, affecting the
incentives of political and social actors to engage in alliances or to preserve
pre-existing divisions, and thus the particular shape of party systems and
party alignments with society. Electoral rules, by imposing different thresh-
olds for political representation of emerging social movements, created dif-
ferent incentives for alliances with already established parties (Lipset and
Rokkan 1967: 31). Similarly, Lipset and Rokkan called attention to the way in
which particular ‘traditions of decision-making’—more or less centralized,
more or less accommodating of conflicts—affected the likelihood that new
inputs into the political system might be converted into policy and, thus, the
adoption of more or less confrontational and divisive strategies of emerging
social interests and movements (1967: 26). In sum, Lipset and Rokkan were
not really proposing a sociological determinism in what concerned the shape
and anchoring of party systems, and were quite willing to concede something
that later research was to establish more clearly: that the nature of party sys-
tems and their alignment with social interests are the result of interaction
between pre-existing cleavage structures and established electoral rules (Ord-
eshook and Shvestova 1994; Neto and Cox 1997).
This suggests possible paths for exploring the relationship between elec-
toral rules and the extent to which structural voting may prevail in a particu-
lar democracy. One commonly advanced argument has been that, in
majoritarian systems, in order to secure the electoral majorities that are neces-
sary to obtain power, parties have incentives to extend their appeals beyond
the confines of narrowly defined social groups. Conversely, if more permissive
rules allow parties to form and survive by exploring socially defined electoral

82
 Political Institutions and the Social Anchoring of the Vote

niches—as tends to be the case in PR systems—clearer alignments between

parties and social groups become more likely (Horowitz 1993; Norris 2004).
In fact, taking into account the logic of ‘winner-take-all’ politics and Lipset
and Rokkan’s own arguments about ‘traditions of decision-making’, a broad-
er argument could be plausibly made about different ‘types’ of democracy,
rather than just electoral systems. Lijphart’s seminal distinction between
majoritarian and consensual democracies (Lijphart 1984, 1999) suggests that
electoral laws combine with other institutional features to form two different
types of democracy, majoritarian and consensual, organized around the basic
principles of, respectively, concentration of power and its diffusion. PR sys-
tems, higher levels of party system fractionalization, strong parliaments, and
coalition cabinets all tend to go together in existing democracies, fostering a
general diffusion of power. Conversely, majoritarian electoral systems, lower
levels of party system fractionalization, weak ‘arena’ parliaments, and major-
ity governments combine to foster concentration of power. This concentra-
tion of power seems inimical to the preservation of party appeals directed at
particular social groups or to the generation of socially homogeneous party
constituencies. The more the system is governed by a winner-take-all logic—
with little room for accommodating minority interests through coalition
governments and a lack of checks on executive power—the more parties are
left with few avenues to seek their goals besides winning elections. This, in
turn, requires them to abandon narrow appeals aimed at mobilizing particu-
lar social segments of the electorate. Besides, if levels of party fractionaliza-
tion are lower—another element of majoritarian politics—the existence of
cross-cleavages becomes more likely, leading parties to de-emphasize certain
cleavages and to blur conflicts in those dimensions where their bases are
divided (Zielinski 2002).
What has been the empirical support for these hypotheses? Dalton, for
example, finds the predicted relationship between the number of parties and
‘class voting’: the more fractionalized party systems are, the stronger the
­relationship between social class and vote choices (Dalton 2008). More gener-
ally, Norris (2004) shows that, under majoritarian electoral systems, socio-­
structural features of voters tend to explain less variance in the vote than in
PR or mixed systems. However, not all findings point in this direction. Huber
(2011), for example, in the context of the study of ethnic voting, finds it to be
less prevalent in PR than in majoritarian systems. He speculates that this may
result precisely from how easy it is, in PR systems, to mobilize voters on dif-
ferent issues. Given the heterogeneity of preferences in members of any social
group, parties aiming to attract voters on the basis of an ethnic appeal soon
find competition from other parties who try to attract them on the basis
of  other appeals, something that contributes to ‘diffuse the cohesiveness
of group voting behaviour’ (Huber 2011). Therefore, the empirical evidence

83
Pedro C. Magalhães

concerning the relationship between majoritarian or consensual political

institutions or some of its components—such as the electoral or party s­ ystems—
and structural voting has remained somewhat contradictory.

5.2.2  Presidentialism
There is more to say about the relationship between political institutions and
structural voting than focusing exclusively on the electoral rules that prevail
in legislative elections or, more broadly, on the extent to which a democracy
is ‘majoritarian’ or ‘consensual’. There is an additional hypothesis that, as far
as I know, has not yet been systematically investigated: it relates presidential-
ism to a lower social anchoring of the vote.
Of course, some of the possible consequences of presidentialism are, in a
sense, partially captured by the notions of ‘consensual’ vs ‘majoritarian’
democracy. If we focus strictly on the composition of executive power, presi-
dentialism’s election rules and unipersonal executive office already ensure,
by definition, ‘majoritarianism’. However, there might be additional aspects
of presidential systems that, in comparison with other systems—regardless of
whether they are majoritarian or consensual—are likely to make structural
voting less prevalent.
The crucial aspect here concerns the separate origin and survival of the
executive in relation to the legislature that characterizes presidentialism. On
the one hand, it is not indifferent whether legislative elections coexist with
presidential elections that determine the composition of the executive, as in
presidential systems, or whether legislative elections solely contribute to
determine the executive’s composition, as in parliamentary systems. In the
former case, voters are systematically exposed to two different sorts of elec-
toral appeal: from parties, in legislative elections, and from presidential can-
didates, in presidential elections. Given the incentives provided by majority
rule for the election of presidents, policy positions and electoral appeals on
the part of candidates tend to be more personalized, more centrist, and more
aimed at the median voter (Wiesehomeier and Benoit 2009). This also means
that, in presidential systems, even legislative elections run under PR can be
contaminated by a majoritarian logic. There is empirical evidence of such
contamination. Samuels and Shugart (2010), for example, show that in the
rare instances where we are able to observe moves towards rules that promote
a separation between origin and/or survival of the executive and legislature,
we also tend to observe a change in the organization and behaviour of parties
that leads—at least for the larger parties—towards ‘vote-seeking’ strategies,
greater personalization of politics, and a lower importance of ideology in
legislative elections. In other words, presidentialism tends to ‘presidentialize’
party appeals in legislative elections and infuse them with a majoritarian

84
 Political Institutions and the Social Anchoring of the Vote

logic, even if the particular institutional rules that govern legislative elections
would not lead us to predict such outcome.
There is a second potential consequence of presidentialism. The separation
in the origin and survival of executives and legislatures undermines party
unity and loyalty in the legislature, giving MPs lower incentives to behave
cohesively and breaking linkages between legislators and the party leadership
(Carey and Shugart 1995). From this point of view, presidentialism affects
electoral party politics by allowing candidates in legislative elections to dif-
ferentiate from each other and to serve their specific constituencies, rather
than adopting national party platforms and building linkages with broadly
defined social groups and interests. There is also considerable evidence of this
phenomenon. Presidential systems tend to be characterized by greater intra-
party divergence, by ‘personal vote’, and by a low level of nationalization of
politics (Carey and Shugary 1995; Morgenstern and Swindle 2005; Morgen-
stern, Swindle and Castagnola 2009). These developments, by being inimical
to the establishment of strong and stable links between parties and social
groups, should lead us to expect presidential systems to be characterized by
lower levels of structural voting.

5.2.3  Social Modernization

Not all hypotheses relating macro contextual features of a political system
with the prevalence of structural voting focus on institutional rules. In fact, the
most conventional idea on this issue is that social modernization should have
consequences: the more socially and economically developed a country is, the
lower structural voting should be. As the argument goes, modernization brings
about changes in class structures and social stratification that are thought to
generate new conflicts of interests, to disturb previous alignments between
parties and social groups, and to weaken ties between individuals and those
organizations—such as churches or unions—that in the past had given expres-
sion to collective identities and worked as intermediaries between parties and
society (Dalton, Flanagan and Beck 1984; Clark and Lipset 2001; Dalton 2002;
Norris 2002). Secularization and the increase of citizens’ cognitive skills pro-
duce changes in their generic values and attitudes towards politics, leading to
the emergence of new value conflicts that cross-cut previous alignments and
undercut the impact of social cleavages—such as class or religion—in the vote.
Although there is disagreement on the real nature of those new value clea­
vages,1 it is relatively clear that they have generically emerged as a result of the
transition from industrial to post-industrial societies and economies.
As it was formulated, this is basically an account of why structural voting is
thought to have experienced a secular decline in a particular set of advanced
post-industrial democracies. But this account has also been extended to

85
Pedro C. Magalhães

become a hypothesis about what may explain cross-national differences: in

more socially and economically developed societies, voting choices should be
less anchored in traditional social cleavages (Norris 2004). Empirical support
for this hypothesis, however, remains limited. In what is, to our knowledge,
the broadest cross-national comparison in the literature, Norris compared the
strength of ‘cleavage politics’—as captured by the variance in an ordinal
measure of ‘left–right vote’ explained by a model containing variables meas-
uring age, gender, education, income, union membership, religiosity, and
belonging to a ‘linguistic majority’—in 37 elections that took place between
1996 and 2002 in 32 countries, using the CSES’s Module 1 data. Instead of
lower levels of explained variance in the ‘post-industrial’ societies, Norris
found, in fact, that the explanatory power of social characteristics of voters
was, on average, higher in the post-industrial than in the industrial democra-
cies. Of course, the fact that the models were not the same in all countries and
elections (with variables such as union membership, religiosity, or belonging
to a linguistic majority missing on some of them) and the inclusion of an
attitudinal variable (left–right self-placement) as part of this ‘social cleavage’
model raises questions about the soundness of the comparison and what it
might be really capturing. But in any case, the main point is that, even with
the use of a (comparatively large) set of countries, support for a modernization
theory of cross-national differences in structural voting seems to be absent.

5.2.4  Democratic Experience and Timing of Democratization

A final generic hypothesis I will address here relates a country’s experience
with democracy (or the timing of its democratization) with the social anchor-
ing of the vote. There are good reasons to believe this relationship should
exist. Mainwaring and Zoco (2007), in their study of electoral volatility in old
and new democracies, propose two mechanisms through which one country’s
democratization may influence the extent to which parties are able to build
stable alignments with voters. One argument is that the longer the history of
democratic political competition in a particular country, the more likely it is
that voters have been able to form partisan attachments and ‘that parties win
over some relatively stable clientele groups, routinise their electoral appeals
and build a more stable base’ (Mainwaring and Zoco 2007: 161). An alterna-
tive mechanism relates the existence of such alignments to the timing of
democratization. The process described by Lipset and Rokkan (1967) took
place in a particular historical period and in a set of countries experiencing an
opening up of their political structures to competition and participation as
they were undergoing crucial ‘national revolutions’. In those cases, parties
became both the main vehicles for the aggregation of social preferences and
the source of social and political identities, reinforced by socialization and

86
 Political Institutions and the Social Anchoring of the Vote

links to organizations such as churches and unions (Pizzorno 1981). In con-

trast, countries that have democratized later have done so in very different
social and political contexts. In these countries—so-called ‘third wave’ democ-
racies—new democratic parties were formed after the emergence of modern
mass media (particularly television) as the main channels of political interme-
diation, parties played a less central role in the expansion of citizenship, and
their formation took place in a context of already weakened links between
individuals and secondary organizations (Mainwaring and Torcal 2006: 209;
see also Scarrow 2010). As a consequence, party system institutionalization,
citizens’ attachments to existing parties, and, thus, the social anchoring of the
vote are likely to be lower in ‘third wave’ democracies.
One of the problems of dealing with cross-sectional data, such as the one
we have available for this chapter, is that it becomes very difficult to distin-
guish the effects of length of experience with democracy from the effects of
timing of democratization. Without panel data, such as used by Mainwaring
and Zoco, measures of the former will correlate almost perfectly with meas-
ures of the latter. However, there is a deeper question that needs to be posed
in this regard. There is a powerful counterargument to the notion that voting
behaviour in older democracies should be more socially anchored. Instead of
contributing to forge increasingly stronger links between social groups and
parties, a longer experience with democracy may in fact perform the opposite
role. The expansion of the suffrage and the transfer of social conflicts to the
institutionalized setting of democratic representation, deliberation, and poli-
cymaking may precisely serve to defuse the very cleavage-based conflicts that
gave origin to those alignments in the first place. As generations succeed and
such conflicts find political resolutions, previous loyalties tend not to be
transmitted and to gradually weaken (Franklin, Mackie and Valen 1992). In
fact, it may be the case that the relationship between length of democratic
rule and structural voting turns out to be the opposite of conventional expec-
tations: one of the consequences of the conflict-resolution process allowed by
political representative democracy may be that ‘in new and consolidating
democracies, cleavages are often of greater importance than in today’s estab-
lished democracies’ (Van der Eijk and Franklin 2009: 98).

5.3  Socio-structural Voting in 34 Legislative Elections

5.3.1  Measurement
Studies of the socio-structural anchoring of the vote have remained a battle-
ground between perspectives that focus on the strength of the relationship
between specific social/group membership variables and the vote and those
that focus on the overall relevance of social differences as explanations of vote

87
Pedro C. Magalhães

choices. It is a battleground between what have been described as, respective-

ly, ‘sociological’ and ‘political science’ traditions (see Franklin 2009 and 2010).
In this paper, our basic research question is clearly linked to the latter tradi-
tion. However, even within this generic line of inquiry, many different analyti-
cal and methodological choices need to be made, namely in what concerns
the manner in which the impact of social variables on the vote is to be gauged.
How can we ensure comparability between countries and elections? Surely,
coding voting choices in terms of ‘left’ vs ‘right’ (Franklin, Mackie and Valen
1992; Norris 2004) allows comparability. However, imposing a left–right uni-
dimensionality is unlikely to capture the social anchoring of vote choices in
democracies where alignments between parties and voters may have devel-
oped in other historically contingent ways or where the concepts of ‘right’ or
‘left’ are more fluid and indeterminate. Alternatively, one might treat indi-
viduals in each country and election as having made discrete choices between
different parties or party families, without imposing any left–right dimen-
sionality or aggregation, and then use measures of fit resulting, say, from
multinomial or conditional logit models, in order to compare the social
anchoring of the vote in different countries (Knutsen 2004). However,
although pseudo R-square and other measures of fit are useful in comparing
different discrete choice models predicting the same outcomes and using the
same samples, they do not provide comparable estimates across data sets and
with different choice sets (Tabachnick and Fidell 2007).
How to proceed, then? In this chapter, we follow Huber’s (2011) approach,
developed in the context of the cross-national study of ethnic voting. Con-
sider a country with Blue voters and Green voters, and three parties: Choco-
late, Vanilla, and Strawberry. How well can one predict whether a randomly
selected voter will choose Chocolate, Vanilla, or Strawberry on the basis of
knowing whether he or she is Blue or Green? How can we, so to speak, meas-
ure the ‘colour voting’ phenomenon in this country? Huber’s approach is
based on the Gallagher’s (1991) least-squares index to measure the dispropor-
tionality of election systems. However, instead of calculating differences
between votes and seats for parties, Huber uses it to calculate indices captur-
ing the difference between vote shares for the electorate as a whole and vote
shares for a particular group (the percentages of vote for each party among,
say, Blue voters). Vgj If is the proportion of individuals in group g that supports
party j, Vj the proportion of individuals in the electorate that support j, and p
is the number of parties, then ‘structural voting’ (SV) for a particular group in
a given election is obtained by:

1 p
∑ (Vgj − Vj )
2
SVg =
2 j=1

88
 Political Institutions and the Social Anchoring of the Vote

To obtain an overall measure of cleavage voting for the country as a whole,

cleavage voting for each party can be summed and weighted by the size of
each group. If G is the number of groups and sg the proportion of group g in
the electorate, then:
G
SV ’ = ∑ ( SVg × sg )
g =1

However, the theoretical maximum of SV’ is below 1 and it is sensitive to

number of groups. Thus, Huber proposes to weigh SV’ by a function of the
number of groups:

1 G
SV =
G−1
∑ ( SV
g =1
g × sg )

2G

The properties of SV are particularly useful for our purposes. First, it ranges
from 0 to 1. In our example, if in a given election Blue and Green voters dis-
tribute themselves among the parties in the exact same way, SV = 0. Con-
versely, if all Blue voters choose one party while all Green voters choose
another, SV = 1. Second, the measure is sensitive to group size. Imagine that
Blue and Green voters distribute themselves by parties similarly in two differ-
ent countries. However, in country 1, Blues and Greens represent equal pro-
portions of the electorate, while in country 2 Greens are only 10 per cent of
the electorate. In the latter case, SV will be lower. Finally, weighting by a func-
tion of the number of groups ensures that, for any number of groups in any
given election, SV will always be equal to 0 if the distribution of the vote for
each group is the same.
While groups can be defined in ethnic or ethno-linguistic terms, as in
Huber’s study, they can also be defined in other ways. For example, members
and non-members of unions form two groups, and Huber’s measure can be
used to assess the extent to which knowledge of a voter’s membership in a
union helps us predict, in any given election, his or her vote choice. Men and
women form two different groups on the basis of gender. Individuals defined
in terms of the frequency of their religious attendance can also be thought of
as forming different groups, and the same occurs with those belonging to dif-
ferent social classes or religious denominations. Thus, using Huber’s measure,
we can extend our analysis to other manifestations of structural voting.2

5.3.2  Indices of Structural Voting

For this analysis, we use 34 post-election surveys conducted in 33 countries,
which are contained in the CSES Module 2 dataset. We have not considered

89
Pedro C. Magalhães

elections that have taken place in non-democratic regimes, which led to the
exclusion of Kyrgyzstan, Hong Kong, and Russia. Furthermore, we study here
exclusively legislative elections, which led to the exclusion of surveys in
Chile and Peru, which studied vote choices in presidential elections.
Considering the relevant socio-demographic variables available in the
CSES surveys, we focus our analysis on class voting and on its organizational
dimension (i.e. trade union membership), on religious voting (both in terms
of religious denomination and religiosity), and on gender voting. Ideally, one
would like to be able to include a measure of ethnic voting too. However, a
question about ethnicity was asked in only 16 of our 34 election surveys, forc-
ing us to exclude ethnic voting from the analysis. Conversely, of the 34 CSES
surveys, all contained a question determining whether the respondent
belonged to a trade union. Socio-economic status is measured by a nominal
variable with four categories, based on answers to questions about the
respondent’s occupation. The four categories are ‘white-collar’ (non-manual
employees), ‘worker’ (workers engaged in manual labour), ‘self-employed’
(covering entrepreneurs, shopkeepers, and professionals), and farmers. This
variable is available for 28 of the 34 elections considered. In the case of religi-
osity, for the 32 surveys where the question was available, we distinguish
between individuals who report attending religious services at least once a
week (coded as 1) from all other individuals (coded as 0).3 Questions about
religious denomination were asked in 29 elections. Finally, we look at gender,
for which we have measures in all 34 surveys.
Table 5.1 displays 157 indices. To improve readability, cells with higher
indices are displayed with darker colours. Purely for presentational purposes,
countries are sorted by their average levels of structural voting across the five
measures.
The first thing that emerges even from a cursory observation of Table 5.1 is
the fact that some countries seem to consistently display comparatively high-
er levels of structural voting across the board. The fact that Switzerland, Swe-
den, or Israel, for example, emerges with comparatively higher levels of
structural voting, or that the Czech Republic also stands out among East
European countries in this respect, is not particularly surprising in the light
of previous studies (Norris 2004; Van der Brug 2010). Conversely, countries
such as the Philippines, Mexico, and Taiwan appear here, also as in other
studies (Norris 2004), with consistently very low levels of structural voting.
This suggests the possibility that some common underlying factors may be
driving the social anchoring of the electorate down or up, regardless of the
particular socio-demographic variable with which one is concerned.
At the same time, however, it is also clear that generic measures of s­ tructural
voting are likely to miss out on relevant differences between countries. In Swe-
den, for example, social class and union membership appear as comparatively

90
 Political Institutions and the Social Anchoring of the Vote

Table 5.1.  Indices of structural voting (SV) in 34 elections

Social class Union Religious Religiosity Gender

membership denomination

Czech Republic 0.14 0.06 0.21 0.15 0.12 

Netherlands 0.07 0.06 0.27 0.14 0.06
Sweden 0.20 0.15 – 0.04 0.07
Switzerland – 0.11 0.16 0.12 0.06
Slovenia – 0.02 0.16 0.13 0.08
Canada 0.06 0.06 0.22 0.15 0.05
Finland 0.18 0.09 0.10 0.10 0.04
Bulgaria 0.01 0.30 0.05 0.04
Norway 0.10 0.09 – – 0.09
Israel 0.08 0.07 0.12 0.13 0.06
Italy 0.12 0.08 – 0.08 0.07
Hungary 0.08 0.10 0.10 0.11 0.04
Belgium 0.09 0.08 0.14 0.07 0.04
Germany 0.08 0.08 0.13 0.10 0.03
Australia 0.06 0.14 0.12 0.04 0.03
New Zealand 0.06 0.07 0.10 0.10 0.06
Romania – 0.06 0.14 0.04 0.06
Spain 0.14 0.02 0.07 0.12 0.02
Albania 0.11 0.04 0.06 0.05 0.11
Poland 0.12 0.04 0.01 0.13 0.06
United States 0.07 0.05 0.11 0.09 0.04
United Kingdom 0.08 0.10 0.12 0.02 0.04
Iceland 0.10 0.06 – 0.01 0.12
Japan 0.12 0.06 0.02 0.08 0.06
Brazil 0.07 0.04 0.13 0.04 0.05
Portugal 2005 – 0.06 0.07 0.09 0.03
France – 0.04 0.10 0.05 0.05
Korea 0.08 0.02 0.13 0.01 0.05
Ireland 0.08 0.04 0.04 0.10 0.03
Denmark 0.11 0.04 – 0.01 0.06
Taiwan 0.05 0.04 0.06 – 0.04 >=.20
Portugal 2002 0.06 0.06 0.00 0.08 0.03
Mexico 0.08 0.03 0.04 0.04 0.03 <.20[
Philippines 0.06 0.02 0.02 0.04 0.02
Average 0.09 0.06 0.11 0.08 0.05 <.10

strong predictors of the vote. In the 2002 elections under analysis, both the
Swedish Social Democrats and the Left Party did considerably better among
the (large number of) union members than, in contrast, the Moderate or the
Christian Democratic parties. Similarly, while the Social Democrats did par-
ticularly well among working-class voters, those who were self-employed or
white collar were rather more likely to vote for the Moderate or the Liberal
People’s Party, while farmers flocked to Centre Party. In contrast, religiosity
seems to be of almost no consequence in Sweden. The Netherlands exempli-
fies the opposite situation. In the 2002 elections, Catholic and (to a lesser
extent) Protestant (Calvinist) voters were much more likely to choose the

91
Pedro C. Magalhães

Christian Democratic Appeal party, while the Christian Union party also
­managed to attract an important share of Protestants (but, predictably, not
­Catholics). At the same time, individuals professing no religion were rather
more likely to vote for the Labour or the Green parties, while the small share
of Muslim voters massively concentrated their vote in the Labour party. In
contrast, occupational variables played little role in structuring the vote in the
Netherlands 2002 election. Although farmers, for example, did seem to dis-
play very different voting patterns from other groups defined in terms of
socio-economic status, their very small number in the sample ends up result-
ing in an index of class voting for the Netherlands that is among the lowest in
our sample of countries.
Several of these variations have been known for some time. In Rose’s semi-
nal study of 12 Western democracies (1973), Scandinavian countries emerged
as those where occupational cleavages contributed more to structure the
vote, which in turn appeared to be of little to no relevance in countries such
as the Netherlands, Ireland, Canada, the UK, or the USA. Similarly, he found
that the Netherlands stood out as a country where religion played a crucial
role, a finding replicated by the more recent comparative study by Van der
Brug (2009). Our results, albeit expanding the cross-national scope of com-
parison, generally tend to replicate these patterns. In other words, even if it is
the case that structural voting is experiencing a secular decline, as the litera-
ture has documented, many of the known differences between countries
seem to be highly resilient. What explains them?

5.3.3  The Determinants of Structural Voting

Equipped with indicators of structural voting that describe the relationship
between socio-demographic features of individuals and their vote choices, we
can move to the second and central stage of our multilevel analysis: testing
hypotheses concerning the relationship between macro-level features of poli-
ties and the prevalence of structural voting. The hypotheses resulting from
the discussion in section two are the following:
H1: More consensual democracies should display higher levels of structural
voting.
H2: Presidential regimes should display lower levels of structural voting.
H3: Higher levels of socio-economic development should be associated with lower
levels of structural voting.
H4a: Longer experience with democracy/more distant democratization should
result in higher levels of structural voting.
H4b: Shorter experience with democracy should result in higher levels of struc-
tural voting.

92
 Political Institutions and the Social Anchoring of the Vote

Hypothesis 1 is tested by using the indicator of consensual democracy intro-

duced in Chapter 2 in this volume. More specifically, we use the Consensual
democracy on the executive–parties dimension index (Vatter and Bernauer 2009),
which is constructed by averaging the standardized scores of variables captur-
ing the disproportionality of the electoral system, the effective number of
parliamentary parties, and the dominance of the executive over the legislature
in the period from 1997 up to the date of each election. Values range from a
minimum of –2.45 (United Kingdom) to a maximum of 1.97 (Belgium). For
presidentialism, we use a dummy variable identifying the presidential regimes
in the sample with value 1: Brazil, Korea, Mexico, the Philippines, and the
United States.
A first look at bivariate relationships is rather encouraging in what con-
cerns the role of political institutions in increasing or decreasing structural
voting. Take Figure 5.1, for example, which displays plots and simple regres-
sion lines concerning the relationship between consensual democracy and
the indicators of structural voting in our countries. In all cases, the relation-
ship is positive. Similarly, Figure 5.2 shows the average levels of structural
voting for presidential (1) and non-presidential regimes (0): in all cases, struc-
tural voting is, on average, lower in presidential than in parliamentary or
semi-presidential regimes.
To what extent do these relationships resist multivariate analysis? We
regressed all structural voting indices not only on the consensual democracy
and presidentialism variables, but also on the remaining variables relevant
for our hypotheses. The notion that structural voting should be lower in the

Class voting Union membership Denomination

Religiosity Gender

Figure 5.1.  Plots of structural voting indices (y-axis) against consensual democracy
(executives–parties) index (x-axis), with linear regression fit lines

93
Pedro C. Magalhães

Class voting Union membership Denomination

.15

.10

.05

.00
0 1 0 1 0 1

Religiosity Gender
.15

.10

.05

.00
0 1 0 1

Figure 5.2. Average values of structural voting indices for presidential systems (1)
and parliamentary or semi-presidential systems (0). Error bars are 90per cent confi-
dence intervals

more developed democracies is captured by using GDP per capita at constant

2000 US dollars for each election year in each country (source: World Bank,
World Development Indicators). Since the variable is highly skewed—­ranging
from $1,073 (Philippines 2005) to $38,246 (Japan 2004)—we use the natural
log of GDP per capita. The variable Years since democratic election captures the
number of years elapsed until the election year since a particular country first
had a value of 6 or above on Polity IV’s democracy. Since the variable is also
highly skewed, with values ranging from 204 (United States) to 6 years (Mex-
ico), we use its natural log. Besides, following Huber (2011), we add a control
variable to all models, the level of fractionalization in s­ ociety along the differ-
ent socio-demographic variables. As Huber notes, the extent to which socie-
ties are actually diverse along the different socio-demographic variables—or,
instead, highly homogeneous—will obviously affect how likely they are to
serve as predictors of the vote. For this purpose, we use the aggregate distribu-
tions found in the different surveys for union membership, socio-economic
status, religiosity, and religious denomination, and estimate fractionalization
indices for all countries along all variables, using Fearon’s formula (2003).4
Finally, for the case of gender voting, we take into account Manza and Brooks’s
(1998) suggestion that higher levels of female labour participation rate

94
 Political Institutions and the Social Anchoring of the Vote

(source: World Bank, World Development Indicators) should increase gender

voting.
We use OLS to regress all structural voting indices on the previously
described independent variables. Since the indices vary between 0 and 1 but
OLS may lead to predictions outside the range, we employ a logistic transfor-
mation of the dependent variable, log (CV/(1-CV). Table 5.2 presents the
estimates.
Some variables perform generally as predicted. Higher social fractionaliza-
tion along the different socio-demographic attributes of individuals leads to a
greater strength of structural voting (with the exception of union member-
ship, albeit the sign is positive). Higher participation of women in the labour
market strengthens gender voting, confirming and generalizing Manza and
Brooks’s (1998) hypothesis. We also have a few negative findings. Table 5.2
shows that hypothesis 3, which related socio-economic development with
structural voting, finds no empirical support. Similarly, the contradictory
expectations concerning the relationship between length of experience with
democratic rule and the strength of structural voting seem to result mostly in
the absence of a relationship between the two variables. Only in the case of
union membership voting do we find that it tends to be more prevalent in
older democracies. We tried alternative codings of this variable, namely, a
simple dummy variable identifying ‘third wave’ democracies (i.e. countries
that had become democratic only since 1976) and a measure of years of

Table 5.2.  The determinants of structural voting

Social class Union Religious Religiosity Gender

membership denomination

Consensual 0.07 0.07 0.34*** 0.28*** 0.10**

democracy (0.06) (0.09) (0.10) (0.09) (0.05)
Presidentialism –0.44*** –0.50** –0.65* –1.28*** –0.36*
(0.12) (0.20) (0.34) (0.32) (0.18)
(ln) GDP per 0.10 0.02 –0.17 –0.08 0.01
capita (0.08) (0.12) (0.26) (0.14) (0.13)
(ln) Years since –0.01 0.23** 0.04 0.10 –0.13
democratic (0.09) (0.09) (0.18) (0.17) (0.10)
elections
Fractionalization 2.50*** 0.58 2.99*** 3.99*** –
(0.72) (0.94) (0.92) (1.02)
Female labour – – – – 0.03**
participation (0.01)
Constant –4.63*** –4.00*** –2.05 –3.20*** –4.08***
(0.81) (1.02) (2.13) (0.90) (1.08)
N 28 34 29 32 34
R-squared 0.42 0.45 0.47 0.50 0.30

*p < 0.10; **p < 0.05; ***p < 0.0.01; White Heteroskedasticity-consistent standard errors in parenthesis.

95
Pedro C. Magalhães

continuous democracy until the date of the election. However, results remained
fundamentally unchanged for all models. These null results are also not due
to multicollinearity problems: the highest variance inflation factor (VIF)
found for any coefficient in any model is 3.02, below the conventional cut-off
levels above which multicollinearity starts being a problem.
In contrast, political institutions go a long way in accounting for why struc-
tural voting is stronger in some countries than in others. Early on, we had
seen that some countries seemed to display low levels of structural voting
across the board, suggesting the existence of common underlying factors that
decreased the overall social anchoring of the vote from all points of view.
Table 5.2 shows that presidentialism is such a factor. Hypothesis 2 is sup-
ported in all cases: all coefficients are negative, as expected, and all are statisti-
cally significant at conventional levels. Hypothesis 1 receives a more qualified
support. Consensual institutions—with more permissive electoral systems, a
larger number of parties, and stronger parliaments—are indeed associated
with greater levels of religious and gender voting. In the cases of class or
union membership voting, although the coefficient is positive, it is also short
of conventional levels of statistical significance.
What can we say about the magnitude of these institutional effects? Given
that the dependent variables had been transformed for the OLS analysis,
Table 5.3 displays calculated marginal effects, capturing the change in the value
of the different dependent variables produced by a one-unit change in the
value of our two institutional variables, consensual democracy and presiden-
tialism. For the case of consensual democracy, we also show the predicted effect
of a change from the minimum (UK) to the maximum (Belgium) values.
In comparison with parliamentary or semi-presidential systems, presiden-
tialism decreases the predicted value of the index of religiosity-based voting
by about two standard deviations and the predicted value of all remaining
indices by about one standard deviation. Similarly, if we focus on the effect of
a move from the minimum to the maximum value of consensual democracy,
voting based on religious denomination is predicted to increase by almost
three standard deviations, two standard deviations in what concerns religios-
ity, and one standard deviation in gender voting. In sum, the resilient differ-
ences in terms of social anchoring of the vote between modern democracies

Table 5.3.  Marginal effects

Social class Union Religious Religiosity Gender

membership denomination

Consensual 1-unit change – – 0.04 0.02 0.01

Democracy From min to max – – 0.18 0.09 0.04
Presidentialism 1-unit change/ –0.04 –0.03 –0.07 –0.09 –0.02
From min to max

96
 Political Institutions and the Social Anchoring of the Vote

seem to result, to a considerable extent, from something that is itself equally

resilient in any political system: the kind of institutional rules shaping elec-
tions, policymaking, and executive–legislative relations.

5.4 Conclusion

Lipset and Rokkan (1967) are often presented as the seminal source for a
‘sociological’ approach to the explanation of voters’ alignments with parties.
However, as Franklin reminds us, this does little justice to Lipset and Rokkan’s
thinking: they ‘expressly recognized that different party systems in different
countries resulted from different historical developments and different insti-
tutional settings’ (Franklin 2010: 655). Among those institutional settings,
they argued, electoral and policymaking rules played an important role
(Lipset and Rokkan 1967: 26–31). And if such factors were consequential for
the formation of voter–party alignments in early decades of the twentieth
century, there are good reasons to believe not only that their legacies might
still be visible, as Lipset and Rokkan argued was still the case in the late 1960s,
but also that they may continue to be relevant for how parties appeal to par-
ticular segments of society and how voters respond to those appeals.
Indeed, we found that cross-national differences in the extent to which
voting behaviour is socially anchored seem to be largely driven by institu-
tional factors. Given the Western European bent of most comparative studies
on ‘cleavage politics’ and its decline, presidentialism had not yet been exam-
ined as one of such factors. However, we show that institutional rules that
separate the origin and survival of parliaments and executives seem to create
disincentives for the adoption, in legislative elections, of appeals to socially
defined and rooted groups of voters, thus promoting greater social heteroge-
neity of party constituencies. Presidentialism is negatively related to all kinds
of structural voting, including the role of occupational and social class vari-
ables, religion, and gender. A second institutional factor that seems to be rel-
evant is, less surprisingly, a country’s type of democracy, in Lijphart’s sense.
In democracies where electoral systems are less permissive, where party sys-
tem fragmentation is lower, and executive dominance over policymaking is
more pronounced, the anchoring of the vote on religious and gender differ-
ences turns out to be weaker.
At the same time, we found that other commonly advanced hypotheses
about expected declines (or rises) in structural voting do not seem to translate
particularly well as accounts of cross-national variations. It has been suggest-
ed, for example, that a country’s continued experience with democracy is
likely to affect (positively for some, negatively for others) parties’ alignments
with particular social groups. But with the exception of the relationship

97
Pedro C. Magalhães

between union membership and the vote—apparently a feature more preva-

lent in older democracies—we found no evidence that this explains varia-
tions in structural voting across the board. It has also been suggested that
social modernization has decreased structural voting. But again, although
this hypothesis may work well in explaining shifts in the strength of struc-
tural voting within countries, we found little evidence that it is helpful in
accounting for cross-national variation, something that previous studies
already suggested (Norris 2004). In sum, that ‘electoral behaviour is primarily
political behaviour that is shaped by the supply side of politics at least as
much as by autonomous processes in society’ (Thomassen 2005a: 265) is a
message that seems to remain true as the scope of cases under analysis is
expanded beyond Western European democracies.

Notes

1. See, for example, Inglehart and Rabier (1986) and Inglehart (1997), on the one
hand, and Kitschelt (1994) and Kriesi (1998), on the other.
2. An alternative would be using propensity to vote (PTV) questions for each main
party (Van der Eijk and Franklin 1996). This approach has been put to good use in
a few analyses of structural voting (see, for example, Van der Brug 2010). However,
although the CSES surveys, in Modules 1, 2, and 3, contain one set of questions that
alluringly approximate a PTV measure, i.e. a set of like–dislike scores vis-à-vis all
major parties in the party system, Van der Eijk and Marsh (2007) show that like–­
dislike scores have much worse properties than PTV scores, especially in what con-
cerns the crucial aspect of the relationship with actual vote choices.
3. In the few surveys where no church attendance question was asked but a religiosity
question was (‘How religious are you?’), we coded as 1 those who responded ‘very
religious’.
4. The exception is gender, where ‘fractionalization’ is basically close to a constant in
all countries.

98
6

Political Institutions, Perceptions

of Representation, and the Turnout Decision
André Blais, Shane Singh, and Delia Dumitrescu

6.1 Introduction

The chapter examines the relationship between individuals’ perceptions of

being represented by a party and the decision to vote or not to vote in nation-
al legislative elections. Our focus is on how political institutions influence
such perceptions, as well as how closely or loosely these perceptions affect
turnout.
We assume that the decision to vote or not to vote can be construed as basi-
cally expressive (Brennan and Lomasky 1993). We also assume that what peo-
ple wish to express in an election is their support for a particular party. We
recognize that some people cast a personal vote for either a leader or a local
candidate (Wattenberg 1991; Blais et al. 2003; Clarke et al. 2004; Aarts et al.
2011). We would still argue that, for the great majority of people, an election
is first and foremost a contest among parties, and so how people feel about
the parties is crucial in the turnout decision.
The simple intuition that we test is that a person votes in a national elec-
tion if she actively believes that one of the parties represents her views rea-
sonably well. We expect a strong relationship between such a feeling and
turnout. The first step of our analysis is to determine whether this prediction
holds up. Our expectation is that feeling represented by a party is a sufficient
condition for an individual to express her support for that party. The absence
of such a feeling does not necessarily lead to abstention, as there may be other
non-expressive motivations for voting, especially normative ones such as the
belief that voting is a civic duty that the ‘good’ person ought to fulfil (Blais
and Achen 2009), but it does substantially weaken the propensity to vote.

99
André Blais, Shane Singh, and Delia Dumitrescu

Once we have established that the perception of being represented is a

strong predictor of the decision to vote or not to vote, we examine the sources
of these perceptions. The focus in this paper is on their macro determinants,
and so the questions are: In which countries are perceptions of being repre-
sented highest and lowest? Why?
Our initial interest, in this chapter as in the whole book, lies in the poten-
tial impact of consensus democracy on voting and on its relationship to feel-
ings of representation. We first determine whether a consensual form of
democracy fosters feelings of representation and, indirectly, a higher turnout.
We then look more specifically at the role of the electoral system. We make a
simple distinction between PR and non-PR systems. Ideally, we would like to
make finer distinctions, especially between plurality and majority systems
(Blais and Loewen 2009), but the simple PR/non-PR dichotomy appears to be
the most crucial distinction to be made. Mixed systems are considered to be
PR if they are of a compensatory nature (Massicotte and Blais 1999). Propor-
tional representation is generally construed to be a major component of con-
sensus democracy. Our objective is to ascertain whether consensual democracy
writ large or more simply a PR electoral system (or neither) is associated with
feelings of being represented and a higher turnout.

6.2  Theoretical Background

We start with the simple expectation that basic positive identification of a

party representing one’s views is a strong predictor of the decision to turn out
to vote, even when other attitudinal and socio-economic status predictors are
taken into account. Democracy, in theory, should grant citizens a voice, and
perceptions of representation and responsiveness are forcefully related to
democratic attitudes (Almond and Verba 1963; Abramson and Aldrich 1982).
If we conceive of representation in the classical principal–agent model (Mans-
bridge 2003), then ‘principals’ (voters) who feel that there is a party (‘agent’)
in the system representing their views should feel particularly empowered to
go to vote in order to see their preferred policies put in place. Feeling exclud-
ed from the political process, alternatively, can lead to disenchantment with
democracy (Anderson et al. 2005: 23–6) and, ultimately, withdrawal from the
political process.
The relationship between perceptions of representation and turnout need
not be entirely policy driven. Given that most voters are ‘cognitive misers’
(Fiske and Taylor 1991) who make decisions on the basis of limited informa-
tion, we would expect few of them to make a judgement call of representa-
tion based uniquely on party policies. Rather, we work under the assumption
that for most people, expressed perceptions of representation are indicative

100
 Political Institutions, Perceptions of Representation, and Turnout

of their overall feeling about a party at a given point in time. Compared to the
long-standing psychological identification with a party, we expect percep-
tions of representation to be more ‘fluid’ (Oakes 2002; Huddy 2002). That is,
we expect them to be less stable over time, and more susceptible to be acti-
vated by the debates taking place at election time.
Our attention then focuses on the contextual factors that are more suscep-
tible to activate these perceptions. We first examine the potential impact of
consensus democracy. As consensual democracy is generally associated with
the presence of many parties, one may infer that citizens in this type of
democracy find it easier to identify a party that matches their views about
what the government should or should not do. At the same time, however,
the search for consensus and the depoliticization of disagreements may lead
to the perception that no specific party really defends one’s particular inter-
ests or values. Furthermore, depoliticization may produce boring election
campaigns with few mobilization efforts, and this may well contribute to a
lower turnout.
In a second step, we look at the effect of the electoral system on represen-
tation. As per Powell (2000), there are two broad visions of representation:
the majoritarian vision and the proportional vision. In the majoritarian
vision, individuals should be allowed to choose the government, thus max-
imizing responsiveness and accountability, while in the proportional vision
voters choose agents to do their bidding, which should lead to policy that
represents the largest possible amount of voters. The most critical way in
which these two very different visions are institutionalized is via the elec-
toral system, with non-PR systems pushing the majoritarian vision and PR
systems promoting the proportional vision. Assuming that an individual’s
perception of party representation is a function of both policy representa-
tion and, more generally, political identity, there is some reason to expect
PR systems to be more conducive to stronger individual perceptions of
representation.
Wessels and Schmitt (2006) examine the relationship between the charac-
teristics of a country’s electoral context and the likelihood of identifying a
party perceived as representing one’s views. They find that ‘where supply
structures are meaningful [in the sense that there are a lot of political options
available to voters], voters find it easier to identify a party which represents
them’. Similarly, McAllister (2005) finds a weak, but significant, negative cor-
relation between the type of electoral system (majoritarian or not) and the
percentage of individuals who believe that there is a party representing their
views in the system. In the same vein, Banducci et al. (1999) find that atti-
tudes about government responsiveness among individuals in New Zealand
became more positive once the country switched from a first-past-the-post to
a more proportional mixed-member compensatory system.

101
André Blais, Shane Singh, and Delia Dumitrescu

There is thus some empirical ground to expect proportional representation

to foster stronger feelings of representation. There could be two main reasons
for such an effect, corresponding to two mediating factors. The first is simply
based on the number of viewpoints likely to be present in the system. PR
leads to the presence of more parties running in the election (Lijphart 1994)
and this automatically increases the probability of finding a party that one
agrees with.
The second reason is that PR produces a more polarized party system. Such
polarization means that a greater diversity of highly differentiated viewpoints
is presented to the electorate and that it is easier even for cognitive misers to
identify at least one party that represents reasonably well their own ideology.
In non-PR systems, voters positioned toward or at the extremes of ideological
space are less likely to feel represented, as parties tend to converge towards
centrist positions, whereas in PR systems parties are more likely to stake out a
range of political positions (Downs 1957; Dow 2011), as it is possible to win
seats in such systems by cultivating the votes of various subsets of the elector-
ate. Polarization also makes policy differences and party identities more sali-
ent at election time, thereby activating fluid political identities even among
those generally less strongly attached to a party.
These relationships are far from obvious, however. It is true that propor-
tional representation is likely to foster a more polarized multiparty system.
But a multiparty system often implies the formation of coalition govern-
ments in which the various parties have to make compromises, which means
the abandonment of some prior commitments made in the election cam-
paign, and some supporters may feel betrayed by the parties. Voters may also
perceive the electoral contest as being among two or three main ‘camps’ (left
and right, for instance), and they may feel that no specific party really repre-
sents them.
Further, under coalition governments voters are less able to discern which
party is responsible for policy outcomes (e.g. Powell and Whitten 1993;
Anderson 2000; Fisher and Hobolt 2010). Thus, voters may have difficulty
determining if any party, whether in government or not, is truly representing
their interests. There is evidence that individuals are less satisfied with coali-
tion governments. Listhaug and Wiberg (1995), for example, demonstrate
that multiparty coalition governments tend to be viewed negatively, and,
connecting attitudes about coalition government to broader attitudes toward
institutions, Karp and Bowler (2001) show that negative attitudes toward coa-
lition government led some New Zealanders to be less supportive of that
country’s PR system.
We will thus ascertain whether PR systems are conducive to more posi-
tive perceptions of representation. If there is such a relationship, we will
then sort out whether this is due to the tendency for the party system to

102
 Political Institutions, Perceptions of Representation, and Turnout

become more polarized under PR, to the presence of more parties under
PR, or to both factors.
Finally, a number of analyses have found that PR tends to foster higher
turnout (Jackman 1987; Blais and Carty 1991; Franklin 1996; Radcliff and
Davis 2000. For a more sceptical perspective, see Blais and Aarts 2006; Blais
2006). Given the expected independent effect of the electoral system on
perceptions of representation, and the mechanisms linking these to
turnout, we theorize that this relationship between the electoral system
and turnout should disappear when we take into account feelings of
representation.

6.3  The Data

The data come from Module 2 of the Comparative Study of Electoral Sys-
tems (CSES). Case selection was based on countries conducting legislative
elections in the CSES sample and data availability. We end up with 32 elec-
tions across 31 countries,1 and 35,980 individuals. There are an average of
1,124 individuals per election, with a minimum of 393 and a maximum of
1,916. Figure 6.1 depicts actual turnout in each of the elections we cover. To
empirically examine the above theory, we construct a model using the fol-
lowing variables, which are measured at both the individual and election
levels.

6.4  Individual-level Variables

Our primary dependent variable is whether somebody voted or not. The vari-
able voted is a dichotomous variable that indicates whether an individual
reported voting: coded 1 if she voted and 0 otherwise.
Our second key variable is feel represented, a variable measured with a ques-
tion that inquires whether the respondent feels represented by any particular
party.2 The variable is coded 1 for those who feel represented and 0
otherwise.
We control for the following two political variables: efficacy (measuring the
degree to which the respondent feels that her vote makes a difference, with
higher values corresponding to greater efficacy)3 and party ID (coded 1 for
those who feel close to a party and 0 otherwise).4
We also include the following variables at the individual level: age (meas-
ured in years), gender (coded 1 for females), income (measured in quintiles)
and college education (a dichotomous variable coded 1 for individuals with a
college education and 0 otherwise).

103
André Blais, Shane Singh, and Delia Dumitrescu

Mexico 2003
Switzerland 2003
Poland 2001
Albania 2005
United States 2004
Japan 2004
Czech Republic 2002
Romania 2004
S. Korea 2004
Slovenia 2004
Canada 2004
United Kingdom 2005
Portugal 2002
Ireland 2002
Portugal 2005
Taiwan 2001
Bulgaria 2001
Israel 2003
Finland 2003
Hungary 2002
France 2002
Norway 2001
Spain 2004
New Zealand 2002
Germany 2002
Sweden 2002
Brazil 2002
Italy 2006
Denmark 2001
Iceland 2003
Peru 2006
Australia 2004

40 60 80 100
Turnout (%)

Figure 6.1.  Turnout across elections

104
 Political Institutions, Perceptions of Representation, and Turnout

6.5  Election-level Variables

At the election level there are three key independent variables. Two of them
represent primary dimensions of consensus democracy. The first dimension,
the Consensus, executives–parties index, encompasses the effective number of
parties, cabinet type, executive–legislative relations, and electoral dispropor-
tionality. The second dimension of consensus democracy, the Consensus, fed-
eral–unitary index, encompasses federalism, bicameralism, and judicial review.
Both variables are discussed at length in Vatter and Bernauer (2010b). Our third
key independent variable, Proportional, is coded 1 for proportional electoral
systems or mixed systems with a compensatory component and 0 for all else.
We also include the following controls: compulsory (a dichotomous variable
that equals 1 for Australia and Peru,5 the two countries in our sample with
compulsory voting with some enforcement of sanctions (Birch 2009: 36;
Singh 2011)), age of party system (the average age of political parties in a coun-
try, measured in years; based on Kittilson and Anderson (2011)), polarization
(Dalton’s (2008) party polarization index)6, and the number of parties (using
Laakso and Taagepera’s (1979) effective number of electoral parties index).7
Individuals who are more interested in politics (and more prone to vote)
are more inclined to participate in political surveys. Furthermore, people
sometimes incorrectly report having participated in an election, perhaps due
to faulty recollection or social desirability bias (Karp and Brockington 2005).
To correct for these biases, we have reweighted the data set so that the report-
ed turnout in each election study corresponds to the official turnout.

6.6  Feelings of Representation and Turnout

We first look at the overall relationship between feelings of representation

and turnout. As shown in Table 6.1, the propensity to vote is much higher
among those who feel represented (78 per cent) than among those who don’t
(57 per cent). We must revise our prediction that feeling represented is a suf-
ficient condition for voting; as many as one out of five among those who say
that a party represents them reasonably well still abstain.

Table 6.1.  Voting and feelings of representation

Respondent does not feel represented Respondent feels represented

Did not vote 43 22

Voted 57 78

Note: Cell entries are column percentages of respondents reporting having voted or not. Survey weights are used to
correct for over-reporting of turnout.

105
André Blais, Shane Singh, and Delia Dumitrescu

Table 6.2.  Voting and feelings of representation; logit estimation

Model A

Variable Coef. p-value

Age 0.188 0.000

Female –0.043 0.199
Income 0.138 0.000
Education 0.456 0.000
Feel represented 0.519 0.000
Party ID 0.632 0.000
Efficacy 0.227 0.000
Constant –1.785 0.000
N 35,980
Prob > F 0.000

Note: Dependent variable is whether one voted. Survey weights are used to correct for over-reporting of turnout.
P-values are two-sided.

Table 6.2, which summarizes a regression of turnout on the covariates,

shows that feeling represented is a strong predictor of turnout, even after we
control for age, gender, education, income, party identification, and political
efficacy. All in all, according to the estimation of Model A, the likelihood of
voting increases by 10 percentage points when one feels represented.8

6.7  Feelings of Representation and Political Institutions

As we have now demonstrated that feeling represented matters, the next

hypothesis that we wish to examine is that citizens are more likely to feel that
a party represents them in a consensual or PR system. The hypothesis is tested
in Table 6.3. We are interested here in aggregate patterns, and so the depend-
ent variable is the percentage of respondents in a given country who indicate
that one of the parties represents their views reasonably well.
The main independent variables are first the two major dimensions of con-
sensus democracy as identified by Vatter and Bernauer in Chapter 2, the exec-
utive–parties and the federal–unitary factors, and then the simple PR dummy
variable. We add as a control variable the age of the party system; it is easier
for voters to come to the view that one particular party best corresponds to
their interests or values when the set of options remains relatively stable over
time.
Model B of Table 6.3 shows that when we relate the percentage of respond-
ents who feel represented by a party in a country to the age of the party sys-
tem and the two indicators of consensus democracy, the former variable has
the expected positive effect, but that there appears to be no association
between consensus democracy and perceptions of representation.9

106
 Political Institutions, Perceptions of Representation, and Turnout

Table 6.3.  Feelings of representation and political institutions

Model B Model C Model D

Variable Coef. p-value Coef. p-value Coef. p-value

Consensus, 0.027 0.295

executives–parties
Consensus, federal–unitary –0.030 0.242
Proportional 0.131 0.039 0.077 0.223
Age of party system 0.003 0.000 0.003 0.000 0.004 0.000
Polarization 0.069 0.007
Number of parties –0.006 0.724
Constant 0.471 0.000 0.350 0.000 0.173 0.172
N 30 32 32
R2 0.463 0.451 0.607
Prob > F 0.001 0.000 0.000

Note: Dependent variable is proportion of respondents that feel represented. P-values are two-sided.

In Model C, we substitute the PR dummy variable for the dimensions of

consensus democracy. We do find a significant positive correlation between
the presence of a PR system and the proportion of people who say that they
are represented by a party. Everything else being equal, the percentage of
respondents who feel represented is 13 points higher under a PR than a non-
PR system. This is a rather substantial difference.
We want to better understand why this is the case. We see two potential
reasons. The first is simply that PR produces more parties and that it is thus
easier to find a party that suits one’s views when there are more of them.
The second is that PR produces parties with more distinct platforms, which
means more highly differentiated policies and more salient identities at
election time, allowing for an easier identification of a party representing
one’s views. In Model D of Table 6.3 we add these two contextual variables.
We see that the direct effect of PR is substantially reduced and that party
system polarization has a significant effect on feelings of representation.
In short, it seems that the relationship between PR and the feeling of being
represented is at least partially mediated by the degree of polarization of
the system. The results suggest that the mere presence of alternatives in a
system (i.e. a greater number of parties) is not necessarily conducive to
individuals becoming more prone to say that a party represents their
views. Instead, the strong positive coefficient for polarization suggests
that, consistent with the view of perceptions of representation as a
fluid identity that gets activated by contextual factors, individuals are
more likely to harbour these feelings when these alternatives are highly
differentiated.

107
André Blais, Shane Singh, and Delia Dumitrescu

6.8  Turnout, Feelings of Representation, and the Electoral

System

In the last stage of our analysis, we investigate how political institutions and
feelings of representation combine to affect the decision to vote or not to
vote. The individual-level variables are those already considered above: feel-
ing represented, party identification, political efficacy, age, gender, educa-
tion, and income. The main contextual variables are again consensual
democracy and the electoral system, but we control for whether voting is
compulsory or not. Because we are now interested in the effects of both indi-
vidual-level and contextual factors, a multilevel estimation is warranted. We
thus fit a random intercept to each election. Due to the dichotomous nature
of the dependent variable, the covariates are mapped to individual turnout
with a logistic link function. To estimate the models, we employ gllamm
(Rabe-Hesketh and Skrondal 2005).
The results are presented in Table 6.4. Model E includes socio-economic
characteristics and contextual variables, that is, the electoral system and
compulsory voting. We can see that those in proportional or compensatory
mixed systems do have a higher probability of voting. Post-estimation predic-
tions show that the propensity to vote is two percentage points higher among
individuals in our sample living under a PR system.
As mentioned in the introduction, there is the possibility that what matters
is not the electoral system as such but a broader political arrangement of
which PR is just one component: the consensus model of democracy. We put
this alternative possibility to a test in Model F of Table 6.4, in which we
replace the PR dummy variable with the two main dimensions of consensus
democracy: executives–parties and federal–unitary.10 As expected, we do not
find evidence for the beneficial effect of ‘consensualism’ on turnout in the
same manner that we find for PR. If anything, there is a negative relationship
between the federal–unitary consensualism dimension and the likelihood of
voting.
We further expect that the relationship between PR and turnout loses its
significance when perceptions of representation are taken into account.
Model G in Table 6.5 presents a more complete model with the addition of
perceptions of representation and the other individual-level variables intro-
duced in Model A.
The findings show that even when controlling for contextual effects, feel-
ing represented by a party does increase the propensity to vote. Results also
indicate that compulsory voting substantially contributes to a higher turn-
out, irrespective of one’s feeling of representation (and party identification
and political efficacy). Everything else being equal, among the individuals in
our sample, the average likelihood of voting is 22 percentage points higher in

108
 Political Institutions, Perceptions of Representation, and Turnout

Table 6.4.  Voting and institutions; multilevel logit estimation

Coefficient estimates and significance levels

Model E Model F

Variable Coef. p-value Coef. p-value

Individual level
Age 0.026 0.000 0.025 0.000
Female –0.031 0.415 –0.045 0.267
Income 0.179 0.000 0.173 0.000
Education 0.518 0.000 0.568 0.000
Institutional
Proportional 0.268 0.000
Consensus, executives–parties 0.044 0.295
Consensus, federal–unitary –0.079 0.055
Compulsory 2.354 0.000 2.006 0.000
Constant –1.101 0.000 –1.015 0.000

Random Effects Parameters

Estimate of Standard error Estimate of Standard error

variance variance

Constant 0.175 0.016 0.211 0.034

Number of observations 35,980 32,948
Number of elections 32 30
Prob > χ2 0.000 0.000

Note: Dependent variable is whether one voted. Survey weights are used to correct for over-reporting of turnout.
P-values are two-sided.

countries where voting is mandatory and abstention is credibly sanctioned.

These same results support the expectation that proportional representation
has no direct, independent effect on electoral participation once political
attitudes are taken into account.
Finally, Model H in Table 6.5 tests the presence of interaction effects
between feelings of representation and our two contextual variables. We find
no statistically significant interaction effects, which indicates that feeling
represented likely affects turnout in a similar positive manner across institu-
tional contexts.

6.9 Conclusion

Our analysis provides evidence in favour of the importance of feeling repre-

sented by a party for the individual decision of turning out to vote. While the
issue of representation in democracies has received significant attention over
the years (Pitkin 1967; Przeworski, Stokes and Manin 1999; Mansbridge

109
André Blais, Shane Singh, and Delia Dumitrescu

Table 6.5.  Voting, institutions, and feelings of representation; multilevel logit estimation

Coefficient estimates and significance levels

Model G Model H

Variable Coef. p-value Coef. p-value

Individual level
Age 0.024 0.000 0.023 0.000
Female 0.005 0.883 –0.004 0.920
Income 0.159 0.000 0.160 0.000
Education 0.412 0.000 0.402 0.000
Feel represented 0.681 0.000 0.490 0.000
Party ID 0.588 0.000 0.617 0.000
Efficacy 0.211 0.000 0.214 0.000
Institutional
Proportional 0.014 0.837 0.122 0.427
Compulsory 1.643 0.000 1.621 0.000
Interactions
Feel represented × 0.211 0.210
proportional
Feel represented × –0.267 0.187
Compulsory
Constant –2.317 0.000 –0.244 0.000

Random Effects Parameters

Estimate of Standard error Estimate of Standard

variance variance error

Constant 0.194 0.025 0.195 0.025

Number of observations 35,980 35,980
Number of elections 32 32
Prob > χ2 0.000 0.000

Note: Dependent variable is whether one voted. Survey weights are used to correct for over-reporting of turnout.
P-values are two-sided.

2003), the focus of the debate has been more on how representation is
achieved. In this paper we take the voter’s perspective and examine the con-
textual determinants and consequences of how representation is perceived
using CSES data from 32 elections in 31 countries.
In a nutshell we have found the following. First, the simple fact of perceiv-
ing that there exists a party representing one’s views is a strong predictor of
whether one will turn out to vote or not. This relationship holds even when
controlling for other individual-level and contextual variables. Second, PR is
generally more conducive to an individual feeling represented by a party.
However, this relationship is mediated, as least partly, by the polarization of
the party system. Finally, the positive relationship between PR and turnout
attenuates sharply when perceptions of being represented are introduced in

110
 Political Institutions, Perceptions of Representation, and Turnout

the analysis. In other words, the effect of the PR system is mediated by indi-
viduals’ ability to identify a party representing their views.
The importance of perceived representation by a party raises theoretical
questions about the nature of this perception, requiring further investigation.
Given that voters are cognitive misers, we work under the assumption that
these perceptions are partly grounded in actual agreement with party policies
and partly grounded in more fluid partisan feelings, both of which depend on
the amount and clarity of partisan information that individuals can easily
access. While perceptions of being represented and partisan identification are
not one and the same, theoretical advances in identity theory (e.g. Oakes
2002) suggest that the difference might lie in the role that contextual factors
play in their activation. In particular, perceptions of representation might
require a polarized context (conducive to clearer information on differenti-
ated alternatives) to be activated. This is consistent with our findings with
regard to the mediating role of polarization.
With respect to institutional variables, we have confirmed that PR contrib-
utes to a higher turnout, though it should be kept in mind that the effect is
quite modest (two percentage points). More importantly, we have shown that
the effect takes place, in part, through a more polarized party system. We
have also shown that what is at play is a specific institutional rule, the elec-
toral system, rather than a whole conglomeration of approaches to democ-
racy subsumed under the consensual model of democracy.
Finally, our analysis bears certain limitations. As we use individuals’
responses to just one question, these results should be taken are preliminary.
Further analyses should incorporate a measure of the actual congruence
between individuals’ policy views and those advocated by parties (e.g. Giger
et al. 2009).

Notes

1. Portugal is surveyed twice in Module 2 of the CSES.

2. Question wording: ‘Would you say that any of the parties in [country] represents
your views reasonably well?’
3. Question wording: ‘Some people say that no matter who people vote for, it won’t
make any difference to what happens. Others say that who people vote for can
make a difference to what happens. Using the scale on this card, where would you
place yourself?’
4. Question wording: ‘Do you usually think of yourself as close to any particular
political party?’
5. Belgium and Chile, which are included in CSES Module 2, also have credibly sanc-
tioned compulsory voting, but are excluded from our sample due to missing data.

111
André Blais, Shane Singh, and Delia Dumitrescu

2
( xi − x )
6. ∑ vi ×
5
where vi is the proportion of votes for party i in a given

election, xi is that party’s ideological position, determined by aggregated individ-

ual perceptions of the parties, and x is the mean party position.
1
7. where vi is the proportion of votes for party i in a given election.
∑ vi 2
8. All predicted probabilities calculated with the covariates held at their means.
9. The number of observations is 30 rather than 32 because data on both consensual-
ism indices are not available for Portugal in 2005 and Taiwan in 2001.
10. The number of observations is 32,948 rather than 35,980 because data on consen-
sualism indices are not available for Portugal in 2005 and Taiwan in 2001.

112
7

Democratic Structures and Democratic

Participation: The Limits of Consensualism
Theory
Steven Weldon and Russell Dalton

7.1 Introduction

Democracy requires a politically active citizenry. Sidney Verba and Norman

Nie (1972: 3), for example, state that political participation ‘is at the heart of
the democratic political formula in the United States’. It is through discus-
sion, popular interest, voting, and other political involvement that societal
goals should be defined and carried out. Without public involvement in the
process, democracy loses both its legitimacy and its guiding force.
Like others in this volume, we are interested in how democratic institu-
tions affect citizen political behaviour. We focus on how institutions shape
the patterns of citizen participation, especially beyond voting. Following
Arend Lijphart (1999), our starting point is the distinction between consen-
sual and majoritarian systems. Among his claims of the kinder and gentler
benefits of consensus democracy, Lijphart (1999: 307) maintained that it
stimulates electoral turnout, minority representation, and ‘an underlying
consensual and commutarian culture’. Consensual institutions presumably
incorporate more citizens into the electoral process and lessen political ine-
quality, because they give citizens effective voice and representation (Lijphart
2001).
While the relationship between consensual institutions and voting turnout
has been well researched (Lijphart 1999; Norris 2002; Blais 2006; Kittilson
and Anderson 2011), only recently have scholars begun to explore their
impact on other forms of political engagement (see Karp and Banducci 2008;
Van der Meer, Van Deth and Scheepers 2009). This chapter contributes to this
research, using evidence from the Comparative Study of Electoral Sys-
tems (CSES) to examine how consensual and majoritarian structures affect

113
Steven Weldon and Russell Dalton

political participation. In addition to electoral participation (voting and par-

ticipation in campaigns), we examine four examples of non-electoral action
(trying to persuade others how to vote, contacting a party or politician, work-
ing with a political group, and attending a protest or demonstration). Con-
sistent with the approach in this volume, we focus on consensualism’s key
dimensions as identified by Lijphart (1999; see also Bernauer, Giger and Vat-
ter, this volume), the executive–parties and federal–unitary dimensions, as
well as other institutional features. Starting with national-level relationships,
we develop a series of multilevel models to predict each type of participation.
Our findings lead to a discussion of how consensual institutions and other
national-level structures affect participation, and the implications of these
relationships for contemporary democratic politics.

7.2  Macro-level Context: Consensualism and Political

Participation

Lijphart has long hypothesized that consensual democracies are inherently

more inclusive than majoritarian systems, and that this inclusiveness helps
stimulate voter participation. Their proportional representation (PR) elector-
al systems allow more political parties to gain legislative seats, which has at
least three important implications. First, more citizens feel represented by the
democratic process, thereby stimulating turnout (Blais, Singh and Dumitres-
cu, this volume; Lijphart 1999). Second, there is lower distortion in the trans-
fer of votes into seats, which means individuals who support smaller parties
feel less alienated from institutionalized politics and more efficacious about
voting (Norris 2004; Karp and Banducci 2008). Finally, PR systems may fuel
partisan attachments to political parties and further encourage turnout
(Bowler, Lanoue and Savoie 1994).
What about participation beyond voting? Lijphart implies that consensual-
ism has spillover effects that stimulate participation more broadly. For exam-
ple, he shows that consensualism is positively related to Dahl’s measure of
polyarchy and Vanhanen’s index of democracy, both of which are based on
broader definitions of participation (Lijphart 1999: 276–80; also Lijphart
2001).
Implicit in this perspective is that participation is habit-forming. This may
be because the psychological predictors of voting, which are higher in PR
systems, are the same for other forms of participation (Karp and Banducci
2008), or because voting (or participation of any form) has feedback effects
on the same predictors of initial participation, including party identification,
efficacy, and political knowledge and interest (see Leighley 1995; Parry, Moy-
ser and Day 1992: Chapter 13). Alternatively, participants may acquire new

114
 Democratic Structures and Democratic Participation

psychological and social resources through initial participation, so-called

process incentives that stimulate further participation, such as feeling good
about oneself, meeting like-minded people, or developing a sense of civic
duty (Whiteley and Seyd 2002; Corrigall-Brown 2011).
A competing view suggests that consensualism may have the opposite
effect, actually decreasing citizen involvement beyond voting. Adopting a
rational actor approach, Kitschelt (1986) argues that citizens protest when
they believe the state is not sufficiently responsive to their interests. If the
state is open and inclusive, then it is not necessary to turn to the streets.
Although Kitschelt does not use the consensus and majoritarian terminology,
he identifies the characteristics of consensual systems as being more open,
including more parties and a strong legislature (also see Powell 1982). Accord-
ing to this perspective, citizen interests are relatively constant, and participa-
tion, which is simply a means to achieve those interests, depends on the
(institutional) opportunity structure. We can therefore extend Kitschelt’s
logic to all modes of participation—individuals will engage in progressively
more intensive forms of participation when leaders are unresponsive to con-
ventional, less intensive acts. Thus, consensual systems may diminish other
forms of participation indirectly, precisely because citizens already feel better
represented through the electoral process in these systems (see also Van der
Meer et al. 2009).
In addition, majoritarian electoral systems also tend to be candidate-­
centred with MPs elected from single districts. Paradoxically, this means they
provide more ‘opportunity points’ for citizens to mobilize for change and
affect the democratic process (Whiteley and Seyd 2002). Citizens can work to
‘capture’ individual seats and gain direct representation. They also allow citi-
zens to hold individual representatives accountable. In this sense, majoritar-
ian systems offer greater incentives for participation beyond voting, such as
campaigning and contacting public officials. When successful, it often results
in an identifiable benefit, such as the legislator’s vote on a certain bill or assis-
tance with government programmes (Cain, Ferejohn and Fiorina 1990).
Lijphart’s (1999) second dimension of democratic structures contrasts fed-
eral and unitary systems (Bernauer et al., this volume). Decentralized systems
may suppress voter turnout, because individual elections are less critical, rep-
resenting just one of several veto points in the policymaking process (Tsebelis
2002). Yet federalism also creates more opportunities for participation at the
sub-national level. This brings politics closer to the grassroots. Moreover,
because there are elections more often than just every four or five years, civic
organizations may stay stronger over time, helping to mobilize and encour-
age non-voting forms of participation (see e.g. Vrablikova 2010).
While the role of consensual and majoritarian institutions is our primary
interest, we also examine other contextual factors that may affect ­participation

115
Steven Weldon and Russell Dalton

patterns. For example, previous research suggests that a party system’s level of
left–right ideological polarization can influence turnout and citizen attitudes
(Karp and Banducci 2008; Kittilson and Anderson 2011). Larger ideological
differences among parties may lead voters to see greater consequences in elec-
toral outcomes (Dalton 2008). This is also likely to stimulate more intensive
forms of political action. Similarly, citizens may be more likely to participate
when the election itself is of greater importance. We control for this, differen-
tiating among countries that held legislative elections only, a presidential
election only, and joint legislative and presidential elections. We expect par-
ticipation is higher when a country holds legislative and presidential elec-
tions simultaneously.
We also expect that participation differs across established and developing
democracies (Norris 2002). Established democracies have institutionalized
party systems that are well experienced in running campaigns and mobilizing
voters, who tend to have stronger party ties. Established democracies also typi-
cally have a more robust civil society, which can stimulate group activity, con-
tacting, and even direct forms of action (Welzel, Inglehart and Deutsch 2005;
Dalton, Van Sickle and Weldon 2010). In a similar logic, wealthier countries
should have higher levels of citizen participation (Verba, Nie and Kim 1978).
Affluence, a highly skilled public, and citizens freely engaging in voluntary
associations create a resource environment that can support political action.
Thus, we expect a positive relationship between national affluence and vari-
ous forms of participation (Norris 2002; Dalton, Van Sickle and Weldon 2010).
We recognize that there may be other contextual factors that shape politi-
cal engagement beyond those described here, especially for specific types of
political activities. The literature on voting turnout, for example, has identi-
fied several constitutional and electoral institutions that can affect turnout
(Blais 2006). However, the above discussion provides a theoretical base for
examining how key institutional factors, including consensus versus majori-
tarian institutions, affect a broader range of political activities.

7.3  Individual-level Theories of Political Participation

While our chief interest lies in how institutions and other contextual-level fac-
tors shape patterns of political participation, participation is ultimately an indi-
vidual act. Thus, we need to link our expectations about the role of institutions
directly to micro-level theories of collective action.1 We start with Verba,
Schlozman and Brady’s (1995) general individual-level model of political par-
ticipation—the Civic Voluntarism Model. Verba and his colleagues succinctly
argue that people participate because ‘they want to’, ‘they can’, and ‘somebody
asked’.

116
 Democratic Structures and Democratic Participation

The first factor recognizes that political action requires motivation. Indi-
viduals must have a spur to action: a grievance or a good that they desire. Put
differently, ambivalence inhibits participation. We examine this with
respondents’ satisfaction with democracy, which typically increases electoral
participation and decreases contentious action. Also, the more that individu-
als care about an issue, the more likely they will act to achieve it. That is,
intensity of preferences matters. We test this with two indicators: ideological
extremism and party attachment. We hypothesize that the more ideologically
extreme and strong party identifiers are more likely to engage in electoral
politics, although the impact of these factors may vary for non-electoral
participation.
Individuals must also believe that action has the potential to achieve the
desired outcome. If they believe political elites or the system as a whole are
non-responsive, then the perceived (instrumental) benefits of participation
decrease. We test for this with the belief that it matters which parties are in
government, or what might be termed power efficacy. We hypothesize that it
relates positively to all forms of participation, but particularly voting and
other conventional acts (Almond and Verba 1963). Similarly, if one feels rep-
resented by a party, one should be more likely to turn out to support that party
(see Blais et al. in this volume). However, this might also limit engagement in
other activities, particularly unconventional activities not directly related to
the success of the favoured party in elections. Indeed, as discussed above in
regard to Kitschelt’s opportunity theory of participation, one who does not
feel represented by a political party may be more likely to engage in elite-
challenging behaviour such as protest.
The second factor focuses on how individual and collective resources are
necessary to overcome the costs of political action (Verba, Nie and Kim 1978;
Rosenstone and Hansen 1993; Norris 2002). People participate when they
have time, are better educated, and have higher levels of income. We test for
education and income with the expectation that they correlate positively with
all forms of participation. We also hypothesize, however, that they relate
more strongly to resource-intensive forms of activism, such as group activity,
contacting political officials, and protest. In many ways, voting, a less inten-
sive activity, may help level the playing field between the resource-rich and
resource-poor.
Third, social networks and mobilization efforts can increase political action
(Gerber, Green and Larimer 2008). They may increase awareness of problems
and the need for action. They also may lower the costs of political engage-
ment by communicating necessary information, such as the location of pro-
test sites or informing one of the organizations already working on an issue.
Furthermore, social networks can increase the costs of not participating,
because of social pressure, informal sanctions from acquaintances (those

117
Steven Weldon and Russell Dalton

often doing the mobilizing), and reciprocity expectations for certain types of
activities. Given that non-voting modes of activism tend to be even more
social in nature, we expect mobilization has an even stronger influence on
these activities. Due to limited indicators in the CSES, we are able to examine
just one type of mobilization in the current study, whether the respondent
was contacted by a political party.2

7.4  National-level Analyses

We base our empirical analyses on the Comparative Study of Electoral Sys-

tems (CSES). Module 2 of the CSES was conducted between 2001 and 2006
and includes 33 nations that are the bases for our multilevel analyses.3
We begin this section by describing the levels of participation across the
nations in the CSES. Then we examine the correlation between national char-
acteristics and each form of participation, with a special focus on consensual-
ism and its related dimensions.

7.4.1  Levels of Participation

Module 2 included six new questions tapping different forms of political par-
ticipation (the full question wordings are in Appendix A.7.1):
– Voted in the election;
– Campaign activity or showing party support;
– Tried to persuade others how to vote;
– Contacted a politician or government official;
– Worked with a group on a common concern;
– Attended a protest or demonstration.

These six items do not span the full range of possible political activities, but
the data do provide several distinct advantages. The survey questions include
examples of each of Verba and Nie’s (1972) four modes of political action—
voting, campaign activity, contacting, and communal action—along with a
fifth mode of political protest. The timing of the CSES surveys following
national elections produces more comparability for the election-related ques-
tions than surveys staggered at different points in the electoral cycle. The
battery of non-electoral questions has a longer timeframe (activity in the past
five years) and so timing is less important for these items. The breadth of the
CSES nations also provides the contextual variation we need to examine the
correlates of participation patterns.
Table 7.1 displays the cross-national distribution of political activity.4
While the patterns of voting turnout are well known, the cross-national

118
 Democratic Structures and Democratic Participation

evidence for other forms of political action are more limited (also see Norris
2002; Van der Meer 2009). For example, some nations with the highest levels
of turnout (often because voting is mandatory) have relatively low levels of
campaign activity. Thus, more than 80 per cent of Australians and Belgians
voted in the election, but these nations rank near the cross-national median
in levels of campaign activity. Conversely, the United States and Switzerland
are regularly noted for their very low election turnout, but campaign activity
is very high in the United States and close to the cross-national average in

Table 7.1.  Levels of political participation

Nation Voted Campaign Persuade Contact Work with Protest

activity others politicians group

Albania 59.5 45.5 33.6 24.9 49.0 37.7

Australia 82.4 16.2 32.4 29.1 22.7 13.8
Belgium 85.9 7.2 12.1 8.8 15.0 10.0
Brazil 68.4 18.0 36.6 11.8 20.0 8.1
Bulgaria 72.1 6.6 7.0 3.3 11.7 3.6
Canada 55.3 34.7 65.0 36.2 33.0 14.3
Chile 66.5 12.0 22.6 12.1 16.8 12.3
Czech 59.0 20.4 26.0 6.3 14.8 9.1
Denmark 84.3 7.7 22.2 20.4 34.4 11.3
Finland 70.0 11.4 12.6 14.6 18.1 6.0
France 69.9 6.9 29.0 12.1 19.9 25.2
Germany 73.5 6.4 27.7 12.4 25.6 12.2
Hungary 55.9 9.8 15.4 4.9 13.4 2.8
Iceland 89.1 16.3 22.1 21.4 22.0 12.6
Ireland 67.0 8.2 13.1 21.6 17.5 7.6
Israel 76.1 10.9 32.4 10.8 14.4 13.7
Italy 82.1 8.4 9.1 4.4 4.1 6.7
Japan 59.0 4.3 12.3 4.2 3.6 1.3
Korea 59.5 4.2 20.6 – – –
Mexico 43.4 12.8 8.9 11.3 10.3 7.9
Netherlands 76.8 7.0 12.3 13.8 7.1 9.6
New Zealand 72.5 5.8 21.5 21.7 21.1 7.9
Norway 75.0 6.6 17.5 14.2 36.0 11.4
Peru 84.1 16.2 28.8 8.0 19.5 17.1
Philippines 71.5 27.2 25.7 14.1 31.9 3.5
Poland 47.6 4.2 7.2 4.5 4.2 2.8
Portugal ’02 68.6 7.3 9.6 6.0 8.9 4.7
Portugal ’05 69.2 – 10.6 6.1 12.4 7.4
Romania 62.3 7.5 17.3 3.8 19.6 4.4
Russia 62.2 3.4 25.7 2.8 2.6 1.5
Slovenia 61.1 5.3 6.7 4.2 12.7 5.5
Spain 79.8 5.7 7.7 4.2 14.3 27.8
Sweden 78.0 3.1 12.9 13.5 14.7 10.9
Swiss 37.3 6.0 15.1 12.1 24.5 14.9
Taiwan ’01 66.2 7.5 15.6 8.7 5.4 2.6
Taiwan ’04 78.3 16.7 24.4 13.9 8.1 9.2
UK 58.3 13.2 18.2 19.0 13.4 7.0
United States 56.7 29.9 44.1 28.3 34.8 5.7
Average 68.0 11.9 20.6 12.7 17.8 10.0

Source: Comparative Study of Electoral Systems, Module 2; vote turnout from IDEA (voting-age public).

119
Steven Weldon and Russell Dalton

Switzerland. The table demonstrates the great variation in the levels of politi-
cal activity both across nations and across different modes of action.
It is difficult to determine the interrelationships between political activities
across nations from all the statistics in Table 7.1. We might expect participa-
tion levels to be positively correlated across nations, since an engaged demo-
cratic public is likely to participate in many ways. Alternatively, political
institutions or cultural traditions may produce distinct patterns of action
across different modes. For example, the French follow an expected pattern
of relatively high levels of protest activity, but modest levels of campaign
activity. In additional analyses (not shown) we explored the bivariate correla-
tions between these six activities.5 For turnout we used voting-age public
(VAP) calculated by IDEA from national voting statistics, and self-reported
turnout from the CSES surveys. Both measures are strongly related across
nations (r = 0.74), with self-reported turnout substantially higher than IDEA
calculations.6 We also found that political participation tends to be general-
ized across different modes of action. In nations where people are actively
showing their support for a party during the campaign, for example, they are
also more likely to try to persuade others how to vote (r = 0.71). However,
voting stands somewhat by itself. Other forms of participation are only mod-
estly related to self-reported voter turnout, and almost unconnected to
national statistics on turnout from the IDEA.7 This is a first indication that
past research focusing on the effects of institutional structures on voting
turnout may not accurately describe the impact of these same institutions on
other forms of political activity.

7.4.2  National Characteristics and Participation

The previous literature on consensualism and other institutional structures
argues that the institutional context provides incentive structures that may influ-
ence overall engagement and participation in specific forms of action. Certainly
there is a wide variety of institutional structures and national characteristics that
might affect the cross-national patterns of political participation described
above.≈This section systematically evaluates these possibilities and compares the
impact of the consensualism dimensions to other contextual factors.
The first row in Table 7.2 correlates two measures of consensus democracy with
voting turnout from the IDEA and the six participation items reported in  the CSES
survey (Bernauer et al., this volume). The first measure reflects Lijphart’s most
basic consensualism dimension, the executive–parties dimension, contrasting
multiparty parliamentary systems with strong executive systems. Consensual sys-
tems display higher levels of election turnout, especially for the IDEA measures.
Lijphart implies that this mobilization effect will carry over to activities beyond
voting. However, this appears not to be the case. In fact, participation in all other

120
 Table 7.2.  Contextual correlates of political participation

Predictor IDEA vote CSES vote Campaign activity Persuade others Contact politician Work in group Protest

Consensualism indices

Democratic Structures and Democratic Participation

Executive–parties 0.18 0.05 –0.37 –0.36 –0.39 –0.09 –0.13
Federal–unitary –0.32 –0.09 0.13 0.37 0.13 0.10 –0.10
Electoral system
Prop. representation 0.22 0.03 –0.43 –0.55 –0.51 –0.27 –0.06
District magnitude 0.10 0.12 –0.18 –0.01 –0.16 –0.31 –0.08
Party system characteristics
Eff. no. electoral parties 0.29 0.17 –0.26 –0.08 –0.27 –0.18 –0.02
Party-centred system 0.22 0.04 –0.15 –0.25 –0.44 –0.28 0.01
Party system polarization –0.07 –0.10 –0.21 –0.29 –0.26 –0.06 0.19
Development
Established democracy 0.30 0.18 –0.16 0.01 0.39 0.15 0.15
Yrs of democracy 1955– 0.33 0.27 –0.07 0.17 0.49 0.25 0.16
GDP/capita 0.17 0.09 –0.18 0.02 0.39 0.06 0.00
Human development index 0.18 0.11 –0.25 –0.02 0.36 0.01 0.09

Source: Comparative Study of Electoral Systems, Module 2.

Note: The table presents the Pearson’s r correlations between national characteristics and the percentage active in various measures of political participation. N = 38 in most correlations
121
Steven Weldon and Russell Dalton

political activities is lower in consensual systems—often markedly so, as in the

case of campaign activity (r = –0.37), persuading others (r = –0.36), and contacting
a politician (r = –0.39). In short, consensual systems seem to demobilize citizen
participation beyond casting a ballot.
The second measure of consensus democracy captures the distinction
between federal and unitary systems. Turnout in national elections is lower
in federal systems, but all other forms of participation are more common in
federal systems. This pattern thus also suggests there are underlying differ-
ences in the factors that bring people to the polls on election day and those
that engage them in other forms of political action. The participation stimu-
lus of decentralized federal systems is especially apparent for interpersonal
forms of participation such as trying to persuade others how to vote (r = 0.37),
contacting a politician (r = 0.13), or working with a group (r = 0.10).8
The next panel in the table correlates characteristics of the electoral system
with political activity. As we should expect from previous research (Blais
2006; Kittilson and Anderson 2011), turnout is higher in proportional repre-
sentation systems and those with a large district magnitude. This is most
clearly evident for the IDEA measure of turnout, with weak positive correla-
tions for the CSES self-reported turnout. However, most striking are the oppo-
site relationships for all other forms of political action! A proportional
electoral system might encourage people to vote (r = 0.22), but it apparently
discourages people from showing their support during the campaign (r =
–0.43) or trying to persuade others how to vote (r = –0.55). There are also
negative effects for non-electoral activities such as contacting politicians and
working with a group.
The third panel in Table 7.2 displays the relationships between party sys-
tem characteristics and participation. Nations with a large number of effec-
tive parties have higher levels of voting turnout—but lower participation in
most other forms of political activity. Another aspect of the party system is
the centrality of political parties in the electoral process.9 Similar to propor-
tional representation, party-centred systems have slightly higher levels of
voting turnout, but lower levels for most other forms of political action.
Highly polarized party systems display lower levels of participation in all
forms of action, but these tendencies are fairly modest.
These patterns are unanticipated by the consensualism literature, which
presumes that voting turnout was symptomatic of general involvement in the
political process.10 The explanation may lie in how consensual electoral sys-
tems set expectations for the role of the citizen. These are typically party-cen-
tred elections, where political parties control candidate selection, c­ ampaign
funding, and the organization of the campaign. Often a cadre of formal card-
carrying party members is instrumental in the organization of the c­ ampaign
(Scarrow 1996; Whitely and Seyd 2002). In these institutional contexts,

122
 Democratic Structures and Democratic Participation

parties encourage their supporters to come to the polls. This is also apparent
in comparing party-centred versus candidate-centred party systems. Yet party
mobilization, and implicitly public norms of citizen participation, often seems
to stop at casting a ballot in these party-centred systems.
The final panel in Table 7.2 shows the relationships between development
(political and economic) and the modes of participation. Beyond the impact
of consensualism variables, the democratic development of a nation may also
affect participation patterns. As other scholars have demonstrated (Norris
2002), more democratic nations tend to have higher levels of turnout—
although these patterns are reversed for campaign activity. Being an estab-
lished democracy is positively related to non-electoral participation in
contacting politicians, working with a group, and even protesting. Socio-­
economic development is also positively related to both turnout statistics,
but its relationship with most other forms of political action is less distinct.
There is actually a negative relationship with campaign activity, and a strong
positive relationship with contacting a politician. We suspect that the effects
of socio-economic conditions overlap with other factors, and thus the impact
is diffused in these simple bivariate correlations.

7.5  Multilevel Regression Analyses

The aggregate results above give us an initial indication of how the national-
level political and socio-economic contexts shape citizen participation. In
this section, we further flush out these relationships with multilevel logistic
regression analyses, while also controlling for individual-level predictors of
participation. We present two models for each type of participation.11 Model
1 for each participation type includes the micro-level variables and just the
macro-level factors tapping Lijphart’s distinction between consensual and
majoritarian systems (Bernauer et al., this volume), the exception being voter
turnout where we also control for compulsory voting laws.12 Model 2 for each
type adds additional macro-level factors. Because of multicollinearity and
overlapping theoretical expectations, we use a subset of the other macro-level
variables not captured well by the two Lijphart dimensions. Specifically, we
test for party system polarization, the distinction between new and estab-
lished democracies, GDP per capita (1,000s), and the type of ­election in the
study (legislative only, presidential only, and legislative and presidential
simultaneously; legislative only is the reference category).
At the micro level, we include indicators for each of the major theoretical
perspectives discussed in the literature review section: motivation/­
psychological engagement and efficacy (democratic satisfaction, feel repre-
sented by a party, party identification, left–right ideological extremism, and

123
Steven Weldon and Russell Dalton

power efficacy); individual resources (income and education); and mobiliza-

tion (contacted by a political party or candidate). We also control for age and
gender since these are often related to participation. While some indicators
are more oriented to voting than other forms of participation, especially con-
tacted by a political party (see endnote 2), they nonetheless provide a good
test of the competing theories and allow us to control for such factors in iso-
lating the contextual-level effects.
The results in Table 7.3 and the reported values are logistic coefficient esti-
mates. The first two columns present the findings for voter turnout; consist-
ent with previous research the key micro-level variables are statistically
significant and in the expected direction. The only variables not significant
are ideological extremism and gender. The strongest factors are those related
to psychological engagement, while income, education, and having been
contacted by a party also are key determinants of turnout. The macro-level
factors are also in the expected direction with turnout more likely in strong
party and unitary systems; however, the relationships do not reach statistical
significance (see also Blais et al. in this volume).
Our main interest, however, is to examine the findings for voting in com-
parison to the other political activities. First, the most basic orientation toward
the political system, democratic satisfaction, appears to have differential
effects on the types of participation. Those satisfied with democracy are more
likely to vote, but they are less likely to engage in more intensive and elite-
challenging types of activism. This indicates the rational elements of partici-
pation—as long as one is satisfied with democracy, it is not necessary to engage
in elite-challenging behaviour and activities that go beyond periodic voting.
Other individual-level factors, however, do not follow this pattern; rather,
they appear to stimulate all forms of participation. Indeed, party i­ dentification
is the strongest predictor of institutionalized participation, including cam-
paign activity and trying to persuade others, but it is also a strong predictor
of non-institutionalized modes, including protest. Feeling represented by a
party in the democratic process has a similar effect. Ideological extremeness
also increases the likelihood of engaging in intensive forms of participation;
however, it does not appear to affect the decision to vote. Taken together, the
results indicate the complexity of psychological factors on activism. Dissatis-
faction with the political process appears to discourage voting but to increase
the likelihood of engaging in elite-challenging behaviour. In contrast, a posi-
tive disposition to political issues and formal, institutionalized actors stimu-
lates conventional and unconventional participation alike.
Second, the resource and demographic control factors tend to have similar
effects on other forms of participation as they do on voting. Education is
positively related to all types of participation, but the results indicate it has a
greater effect on non-electoral participation, contacting public officials,

124
 Table 7.3.  Multilevel logit models of political activism

Variables Vote Persuade Campaign Contact Group activity Protest

1 2 1 2 1 2 1 2 1 2 1 2

Constant –1.30** –1.98** –3.14** –3.49** –4.66** –3.99** –4.60** –4.96** –3.45** –3.58** –3.46** –4.62**
Satisfaction with democracy 0.10** 0.10** –0.05** –0.05** –0.03 –0.03 –0.14** –0.14** –0.16** –0.16** –0.20** –0.20**
Feel represented 0.59** 0.59** 0.49** 0.49** 0.50** 0.50** 0.09* 0.09* 0.24** 0.24** 0.14** 0.14**

Democratic Structures and Democratic Participation

Party ID 0.53** 0.52** 0.69** 0.69** 1.09** 1.09** 0.49** 0.49** 0.48** 0.48** 0.44** 0.44**
Ideological extremism –0.02 –0.02 0.24** 0.24** 0.38* 0.38* 0.25** 0.25** 0.36** 0.36** 0.37** 0.37**
Power efficacy 0.16** 0.16** 0.14** 0.14** 0.11** 0.11** 0.11** 0.11** 0.11** 0.11** 0.10** 0.10**
Age (in 10s) 0.25* 0.25* –0.07** –0.07** 0.03** 0.03** 0.10** 0.10** –0.03** –0.03** –0.18** –0.18**
Male –0.05 –0.05 0.24** 0.24** 0.23** 0.23** 0.23** 0.23** 0.17** 0.17** 0.07* 0.07*
Income 0.13** 0.13** 0.02 + 0.02 + –0.03 + –0.03 + 0.07** 0.07** 0.04** 0.04** –0.05** –0.05
Education 0.09** 0.09** 0.08** 0.08** 0.06** 0.06** 0.16** 0.16** 0.14** 0.14** 0.22** 0.22**
Contacted by party 0.41** 0.41** 0.74** 0.74** 1.03** 1.03** 0.94** 0.94** 0.71** 0.71** 0.59** 0.59**
Macro variables
Executive–parties index –0.11 –0.14 –0.25* –0.19 + –0.15* –0.21* –0.19 + –0.10 –0.02 0.00 –0.06 –0.04
Unitary–federalism index –0.24 + –0.21 0.22* 0.14 0.18 0.13 0.03 0.01 0.00 –0.07 –0.09 –0.06
Party polarization index 0.08 –0.05 –0.15 + –0.22* –0.13 0.08
New vs established democracies 0.73 0.23 –0.09 0.38 –0.08 0.45
GDP per capita (1,000s) –0.02 –0.00 –0.03 0.03 0.03 0.01
Presidential election (dummy) –0.31 0.51 –0.93* –0.70* –0.63 –0.22
Leg. & pres. election (Dummy) 0.45 0.77* 0.51 + 0.09 0.57 + 0.09
Compulsory Voting 1.13** 1.09**
N (macro/micro) 33/30967 33/30183 33/31053 32/30284 32/30152 32/30212

Source: Comparative Study of Electoral Systems, Module 2.

Note: **p < 0.01; * p < 0.05; + p < 0.10. Values are logistic coefficient estimates. For election type, ‘legislative election only’ is the reference category
125
Steven Weldon and Russell Dalton

group activity, and protest. There are some exceptions. Consistent with previ-
ous research, older people are more likely to vote but generally less likely to
engage in other forms of participation. It also appears males, while no more
likely to vote, are more likely to engage in all other types of participation.
Finally, party mobilization efforts are strongly linked to all forms of partici-
pation. This may be counterintuitive when thinking of institutionalized ver-
sus non-institutionalized types of participation. Taken together with those
for party identification and feeling represented by a political party, this find-
ing suggests that participation of all types, including protest activity, is large-
ly linked to and channelled through institutionalized political actors—that
is, it is the citizens who are connected to the political system in one form or
another who are most likely to engage in all forms of political activity. This
holds even if citizens are dissatisfied with government or the major political
party offerings.
Turning now to the macro-level effects, the findings are consistent with
those from the aggregate analyses above. The results for the executive–parties
dimension all point in the same direction—that is, consensualism, while pos-
sibly stimulating voter turnout, likely diminishes other forms of participa-
tion. Consensual democracies not only have lower levels of group activity
and protest, but more remarkably, they also have less campaign activity, and
citizens in these systems are less likely to try to persuade others how to vote.
These findings hold even when controlling for other macro-level factors. Fed-
eralism, in contrast, appears to stimulate non-voting forms of political action,
albeit the results do not reach conventional levels of statistical significance.
Nonetheless, there is some evidence that more elections with multiple veto
points encourage citizens to take a more active role in politics and engage in
high-intensity activities.
In terms of the other aspects of the macro-level context, we see that party
polarization appears to have little effect on participation patterns, once we
control for the individual-level predictors. This is somewhat surprising given
recent research pinpointing the significance of polarization; however, that
research suggests its effects may be largely conditional, interacting with
micro-level predictors to affect participation, which we do not have the space
in this chapter to explore fully (see Karp and Banducci 2008). Consistent with
previous studies, participation is also higher in established democracies, even
after controlling for individual-level factors. In contrast, we find little evi-
dence that the level of economic development affects participation. Finally,
the type of election clearly matters for participation. Except for persuasion,
participation is lowest in presidential-only elections and tends to be highest
when presidential and legislative elections are held simultaneously. This sug-
gests that participation increases with more important elections and more
opportunities or separate offices being contested.

126
 Democratic Structures and Democratic Participation

Finally, again, we should note that we have only tested for the direct contex-
tual effects in the current analysis. Previous research suggests that the key effects
of some of these factors may be either conditional or indirect, meaning they
primarily work to strengthen the psychological determinants of political partici-
pation among the electorate, such as party identification, feelings of efficacy, and
the feeling of being represented (see Karp and Banducci 2008; Van der Meer et al.
2009; Blais et al., in this volume). Nonetheless, the results from both the aggre-
gate and multilevel analyses provide an initial indication that consensualism has
different effects on voting versus non-voting modes of participation.

7.6  Consensualism and Democratic Participation

Our findings point to a bigger picture about how institutional arrangements

affect the political process and expectations about citizens’ role in that pro-
cess. Consensual party-based electoral systems may stimulate turnout in
­elections—but they also appear to discourage an active and more robust dem-
ocratic citizenship (see also Van der Meer et al. 2009). Since democracy
expects more from its citizens than an occasional trip to the polling booth,
this may erode the larger foundations of democratic participation.
Further research is required to replicate these results and to begin to explore
the potential causal forces at play. We can, however, offer some initial theo-
rizing based on our findings. When parties control the elections, campaigns
may focus on mobilizing existing party identifiers to vote. Not only is non-
electoral activity lower in consensual systems, it is also lower in strong party
systems. Having voted, citizens may feel they have fulfilled their democratic
duties. This might produce a passive orientation toward campaign activity,
which is seen as the domain of formal party members and a centralized party
organization. In turn, since there is less emphasis on the conversion of voters
to a different party, there is less reason to try to persuade others, display cam-
paign support, or even contribute to a political party (activities which are
typically financed from public sources). In other words, consensual systems
may encourage voters to be semi-spectators, whose participation focuses on
turning out to vote and little more.
In contrast, majoritarian and candidate-based systems may involve a differ-
ent process. The US system is the archetypical example. Candidates must
assemble their personal campaign teams and their own financing; they cannot
base their campaigns on card-carrying party members. Voting choices are more
fluid and candidates try to convert voters (typically independents) to their
cause, so persuasion and shows of campaign support become more important
elements of a campaign. People see campaigns as more than just voting. At the
same time, the individualized and ad hoc nature of candidate-based campaigns

127
Steven Weldon and Russell Dalton

makes it more difficult for campaigns to mobilize voters, which decreases

turnout.
The patterns for campaign activity also carry over to non-electoral forms of
political action. Although the findings could be more robust and significant,
they all point in the same direction. Contacting politicians, working in a
group on a common concern, and protest activity are more common in
majoritarian electoral systems, with small district magnitudes and a candidate-­
centred party system. The personalization of politics in candidate-­centred
systems may be especially important in encouraging political contacting.
This again suggests that consensual systems may focus the norms of active
citizenship on voting to the detriment of other forms of involvement.13
The unitary–federalism dimension shapes political action in a similar
pattern—­federal systems have lower levels of voter turnout in national elec-
tions, but higher levels of persuasion, group activity, and protest, as well as
possibly campaign activity. Federal systems, such as the United States, Cana-
da, and Switzerland, disperse power among different levels of government,
creating more veto or opportunity points in the democratic process. This
means that any single election is less important or consequential to policy
outputs, which may explain the lower rates of voter turnout in such systems.
However, the greater number of opportunity points appears to stimulate
most other forms of participation. We suspect this is also supported by a more
robust set of political associations and interest groups in federal systems. Mul-
tiple, staggered elections, rather than one election every four years, mean
that politics is more constant. Mobilization efforts need to be sustained, and
it seems likely this fosters high-intensity modes of political activism.
Taken together, the findings paint an interesting picture of how the institu-
tional context shapes individual political activity and the development of a
democratic civil society. Lijphart has long argued that consensualism pro-
duces a kinder, gentler form of democracy by bringing more of the popula-
tion into the democratic process and representing a broader range of societal
interests and viewpoints. From a formal institutional perspective, there is
little debate that this is indeed true. Consensual democracies often have a
greater number of political parties, spread out more along the left–right ideo-
logical spectrum. Coalition and minority governments further ensure that
this broader range of viewpoints is reflected in the policymaking process. One
might expect this also to produce a more politically active, engaged citizenry,
and this may be the case when it comes to turnout (see, though, Blais 2006).
Yet this chapter raises more fundamental doubts about consensualism being
conducive to a more involved democratic citizenry for non-voting forms of
participation. A constitutional and electoral system designed to maximize
voting turnout may have unintended negative consequences on other forms
of political participation.

128
 Appendix A.7.1.  CSES survey variables

Variable Question wording Coding

Voted B3004: Did the respondent vote in the election? 0 = no, 1 = yes
Persuade others B3001_1: Here is a list of things some people do during elections. Which if any did you do during the 0 = no, 1 = yes
most recent election? . . . talked to other people to persuade them to vote for a particular party or
candidate?
Campaign activity B3001_2: [During the recent elections]. . .showed your support for a particular party or candidate by, 0 = no, 1 = yes
for example, attending a meeting, putting up a poster, or in some other way?

Democratic Structures and Democratic Participation

Contact politician B3042_1: Over the past five years or so, have you done any of the following things to express your 0 = no, 1 = yes
views about something the government should or should not be doing? . . . contacted a politician or
government official either in person, or in writing, or some other way?
Protest B3042_2: Over the past five years or so, have you . . .taken part in a protest, march or demonstration? 0 = no, 1 = yes
Worked with others B3042_3: Over the past five years or so, have you…worked together with people who shared the 0 = no, 1 = yes
same concern?
Democratic satisfaction B3012: On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not at all satisfied 0 = Not at all satisfied
with the way democracy works in [country]? 1 = Not very satisfied
2 = Fairly satisfied
3 = Very Satisfied
Feel represented B3023: Would you say that any of the parties in [country] represents your views reasonably well? 0 = no, 1 = yes
Party attachment B3028: ‘Do you usually think of yourself as close to any particular political party?’ 0 = no, 1 = yes
Ideological extremism B3045: In politics people sometimes talk of left and right. Where would you place yourself on a scale Recoded as follows:
from 0 to 10, where 0 means the left and 10 means the right? 0,1,2, 8, 9, and 10 = 1
3–7 = 0
Power efficacy B3013: Some people say it makes a difference who is in power. Others say that it doesn’t make a 0 = It makes a difference
difference who is in power. Using the scale on this card, where would you place yourself? 4 = Doesn’t make a difference
Contacted B3003: During the last campaign did a candidate or anyone from a political party contact you to 0 = no, 1 = yes
persuade you to vote for them?
Income B2020: Household income quintile 0 = Lowest quintile
4 = Highest quintile
Education B2003: Education of the respondent 0 = none
7 = university degree
Gender B2002: Gender 0 = female, 1 = male
129

Age B2001: Age Coded in discrete years, 18+

Steven Weldon and Russell Dalton

Notes

1. We examine only the direct effects of context variables, but current research sug-
gests that there are often strong interactions between context and individual-level
predictors of participation.
2. Being contacted by a political party is a very narrow definition of mobilization,
especially for non-conventional modes of participation. The CSES does query
membership in unions, professional associations, and religious attendance, but
these questions were not asked in every country, leading to significant missing
data. The ‘party contact’ variable does allow us to examine whether electoral and
non-electoral modes of political participation alike are influenced by party activ-
ity, but we should be cautious in interpreting as an indicator of social networking
effects more generally.
3. We use the June 2007 release of the data that we acquired through the CSES web-
site (<http://www.cses.org>). The data reflect the combined efforts of the scholars
who collected these surveys, but we are responsible for the analyses and interpre-
tations presented here. We do not include Kyrgyzstan or Hong Kong because these
were non-democratic elections; we also deleted the entries for non-electoral par-
ticipation in Korea because of apparent errors in these data. The multiple regres-
sion models excluded the following countries for the reasons in parentheses:
Taiwan 2001 and 2004 (federal–unitary dimension missing, and feel represented
by a party); Belgium 2003 (income missing); and Japan 2004 (contacted by party
missing). South Korea has missing data for the following dependent variables—
contacting a politician or party, working in a group, and protesting. We also cor-
rected errors in the New Zealand ‘persuade others’ question in the CSES release.
This leaves us with 33 countries for the multilevel models of voting, campaign
activity, and persuading others, and 32 countries (excluding South Korea) for con-
tacting a politician, working in a group, and protesting.
4. The participation questions are not identical across nations, which can distort the
specific national percentages. Readers should consult the CSES documentation on
their website for the specific question variations (<http://www.cses.org/>). In
addition, the framework of the survey (panel versus cross-section, and the mode
of interviewing) may affect participation statistics. Panel surveys, for instance,
typically have higher participation rates because of differential response rates by
people interested in politics.
5. These results are available from the authors upon request.
6. The average turnout from IDEA is 68 per cent; self-reported turnout in the CSES
averages 84 per cent.
7. A principal components analysis yielded the following loadings for a first unro-
tated dimension (49.5 per cent of total variance):

Voting (CSES) 0.57

Voting (IDEA) 0.20
Worked with group 0.84
Contacted politician 0.83
Campaign activity 0.81

130
 Democratic Structures and Democratic Participation

Persuade others 0.81

Protest 0.63

8. In preliminary analyses, we also examined a third consensual institutional dimen-

sion, tapping direct democracy, which Vatter and Bernauer (2009) identify in their
three-factor solution approach. Our findings indicated that all forms of political
action tend to be lower in countries with extensive direct democracy provisions,
even controlling for Switzerland as an outlier. However, the results were not statis-
tically significant in the multivariate models.
9. This index combines five characteristics measuring the extent to which the ballots
involve a candidate or party vote, party control of the ballot, government funding
for parties, free campaign advertising on television, and legal recognition for
political parties. See Dalton, Farrell and McAllister (2011b, Chapter 2).
10. Van der Meer and van Ingen (2009) have analysed participation patterns in the
CSES. Although he does not focus on the consensual model, his findings are
broadly consistent with the analyses presented here.
11. Multilevel models allow us to properly control for the fact that citizens in the
same country are not independent observations, and therefore, the appropriate
sample size for calculating the standard errors of the country-level factors is the
number of level-two units (33 countries). For voter turnout, we weight at the indi-
vidual level to control for over-reporting by using IDEA’s estimates of turnout
based on voting age population (VAP). The models were estimated using the lmer
package in R.
12. We identified countries with compulsory voting without distinguishing if it is
enforced or there is a penalty for non-compliance.
13. Partial evidence comes from the 2004 International Social Survey Program. We
constructed a three-item index of norms supporting participation beyond voting
(Dalton 2009). Nations that are higher in consensualism tend to have lower levels
of participatory norms (r = –0.20, N = 22) and federal systems have stronger par-
ticipatory norms (r = 0.33, N = 22).

131
8

Feeling Policy Represented

Sören Holmberg

8.1 Introduction

Representative rule based on democratic elections and political parties with

voting rights for all grown-ups have been around for about 100 years. Longest
in North America and Northern Europe, more recent in the rest of the world.
Time-honoured mechanisms that might seem a bit old-fashioned today,
much like steam engines. They work but make a lot of noise at the same time
as the reliability and pulling power are less impressive.
That, however, is in all likelihood a wrongful image. Elections and repre-
sentative democracy are not outdated and becoming less and less common.
On the contrary, they are hot, and not only among political scientists. The
World Bank, United Nations, international aid organizations—all talk about
the importance of elections and support the creation of electoral democracies
around the world. Never before have there been as many working electoral
democracies in the world as today (Diamond 1999; Freedom House 2010). In
Freedom House’s latest assessment, 116 of the countries of the world were
classified as electoral democracies. That means that in 2009, a majority of all
nation states employed elections and practised representative democracy (60
per cent). Not bad for old steam engines.
But what about effects? Mostly window dressing and empty smoke puffs or
can we notice more evident outcomes? The question of effects, what democ-
racy is good for, is obviously of great importance. In this context we will
address it by focusing on a limited albeit very central area. That area is citi-
zens’ subjective judgement of whether their democracy delivers what it is
expected to deliver—some form of agreement between rulers and ruled on
the policies that rule the land. Our focus is on how voters around the world

132
 Feeling Policy Represented

perceive how their political views are represented by the politicians they elect
to represent them.

8.2  Consent or Dissent

The historical cornerstone of representative democracy is consent (Manin

1997). In contrast to ancient Athenian decision-making or to modern forms
of direct democracy, citizens are not supposed to take a very active part in
how the representative polity conducts its business—apart from at election
times. The inventors of representative democracy in the late 1700s—James
Madison in the USA, Edmund Burke in England, Charles-Louis de Montes-
quieu in France—did not envisage representative rule as a kind of approxima-
tion to direct democracy (Ober 2008). Athenian democracy was not their
ideal. Quite to the contrary, they perceived their invention as something
qualitatively new.
Division of labour, leadership, and efficiency were their guiding principles.
Voter participation was essential but relegated to election times when people
were supposed to give or not to give their consent to the rule. Madison’s and
the others’ idea was that elected representatives should lead and that people
should approve or disapprove come election time. Representation from
above, not from below, was the idea (Holmberg 2011).

8.3  Mandate or Sanction

Consequently, the main task of voters in a representative democracy is to

consent or to dissent. If voters consent to a programme involving policy ideas
for the future a mandate is given. If, on the other hand, voters consent to
policies pursued during a past election period we talk of a sanction not
delivered.
Similarly, when voters do not like what they see looking forward—when
they dissent—no mandate is given, or if they look backward and disapprove,
a sanction is handed down. Consequently, what has been branded the Man-
date Model focuses on the future and puts emphasis on the opinions of voters
and the policy programmes of parties. Voters are supposed to support or not
support the parties’ future-oriented policy programmes. The Sanction Model,
on the other hand, puts past actions of parties and enacted policies in the
limelight and relies on how voters assess actions and policies. Voters are sup-
posed to hold parties and governments accountable for their past actions.
Observe, however, that the views of voters are central in both models, but
that the views are of a different character. In the Mandate Model, voters’

133
Sören Holmberg

policy wishes for the future are what count, while voter evaluations of past
policies are most important in the Sanction Model. But in a broader perspec-
tive both models concur. It is the will of the people that should be
represented.
In a party-based Mandate Model, usually called The Responsible Party Model,
political parties formulate future-oriented programmes ahead of elections
which voters choose between (APSA 1950; Thomassen 1994). People are issue
voting on the party closest to their own views. They give their consent to a
party’s programme. The parties in turn are expected to work and try to imple-
ment the policies they have promised to pursue. Focus is on issue opinions.
Elections are supposed to bring about a high degree of congruence between
party policies and party voters’ views. In the model, voter consent to party
mandates is the mechanism that ideally brings about policy representation
and assures that the will of the people be done.
The approach of the Sanction Model is somewhat different. Focus is on
retrospective evaluations of government and party actions. Voters give con-
sent or dissent to what has passed, not to what is promised. Yet in this process
voters’ political views are not irrelevant. On the contrary, people’s opinions
affect how they judge what has been done, in much the same way as people’s
opinions affect how they judge the programme of parties in the Mandate
Model. Political views play a decisive role when people give a government or
a party a positive or a negative sanction.
Governments who successfully anchor their policies among voters increase
their possibilities of receiving a positive sanction come election time. Parties
and governments do the anchoring through opinion formation and policy
positioning between elections in order to maximize policy congruence with
voters, thus minimizing the risk of a negative sanction at the polls. But the
process requires that politicians know what people want and how people
value what they receive.
Anticipation is the name of the technique politicians employ when they
try to know public opinion and to foresee how people react. Being successful
at knowing the people is a necessary art form in a democracy. As a matter of
fact, elite anticipation in combination with elections is the device that in the
best of worlds assures that the will of the people is implemented in the Sanc-
tion Model. Politicians read what people want and, if need be, adjust their
policies to avoid losing in the next election. The threat of losing makes elect-
ed representatives amend their policies to secure re-election.
Consequently and indirectly, a high degree of policy agreement between
voters and politicians becomes a vital ingredient even in the Sanction Model;
arguably not as central as in the Mandate Model, but still essential. The main
purpose of the Mandate Model as well as of the Sanction Model is the num-
ber-one goal of a democratic system—to fulfil the will of the people.

134
 Feeling Policy Represented

8.4  Majoritarian or Consensus Models of Democracy

The distinction between the Mandate and Sanction Models is related to the
difference between two other celebrated models in democracy research.
Those models are the Majoritarian and the Consensus Models, most famous-
ly proposed by Arend Lijphart (1975, 1984, 1999).
The Majoritarian Model, which has the selection of majority government as
the fundamental building block, is a distant cousin to the Sanction Model. In
both models clarity of the choice and the possibility to hold parties and gov-
ernment accountable play an essential role (Thomassen, this volume,
­Chapter 1). And like in the Sanction Model, majority governments governing
with the consent of the people and with policies congruent with what people
want run a small risk of being served a negative sanction and being defeated
come election time. Thus, knowing what people want and trying to pursue
policies in accordance are of the essence in the Majoritarian Model.
The Consensus Model has a distant cousin too and that is the Mandate
Model. The major function of the Consensus Model is to bring about a close
resemblance between elected politicians and the electorate. Parliaments and
parties should be as representative of the voters as possible. Ideally, voters’
social characteristics as well as policy views should be mirrored among the
elected representatives in parliament. And when that is the case, representa-
tive bodies rule with the implicit consent of the people; and they have a
mandate to act on behalf of the people.
As Jacques Thomassen points out in Chapter 1, the prime purpose of elec-
tions in the Majoritarian Model is the accountability function, while the
prime purpose is the representation function in the Consensus Model. But in
both models, policy agreement between voters and politicians is important;
admittently somewhat more so in the Consensus Model than in the Majori-
tarian Model. However, and also pointed out by Thomassen, in majoritarian
systems focus is on mandate to and accountability of the government while in
consensus systems representativeness of and mandate to elected members of
parliament is the primary objective.

8.5  Two Main Hypotheses and Two Auxiliary Hypotheses

Having established the centrality of policy representation in democratic the-

ory, the task is now to empirically study the extent to which policy represen-
tation is related to a number of basic features of the working of democracies.1
The first of these are how long democratic practices have been in use in a
country—the age of democracy. The hypothesis is that experience hones
skills—older democracies who have applied their electoral mechanisms

135
Sören Holmberg

­ uring a longer period are expected to exhibit higher degrees of issue congru-
d
ence between voters and elected representatives compared to newer democ-
racies where elections are a more recent tool. An electoral democracy is not
born fix and ready. It takes time and practice to make the democratic engine
work well. And this irrespective of what kind of democratic system we
deal with.
Our second main hypothesis, however, is not unrelated to how the dem-
ocratic system is constructed. On the contrary, we want to systematically
contrast the democratic models we have discussed previously. In much of
the relevant literature, the Consensus Model is found to perform better in
most democratic respects than the Majoritarian Model (Lijphart 1999;
Powell 2000). Examples of democratic aspects being studied are the link
between public opinion and public policies, satisfaction with the working
of democracy, electoral turnout, and women’s representation. Conse-
quently, our hypothesis is that systems employing the Consensus Model
should on average exhibit higher degrees of policy representation between
legislators and the public than systems characterized by the Majoritarian
Model.
In the empirical test we operationalize the Majoritarian Model as countries
using a plurality/majority electoral system while countries having propor-
tional voting systems are classified as belonging to the Consensus Model fam-
ily. Possible causal mechanisms are more cohesive parties and a stronger
emphasis on representation according to the Mandate Model in consensus
systems compared to a clearer focus on individual candidates and strong gov-
ernments in majoritarian systems.
Another alternative operationalization that also will be employed builds on
one of Vatter and Bernauer’s indices of consensus–majoritarian ­democracy—
the executive–parties dimension. It is a composite index encompassing the
effective number of parties, cabinet-type, executive–­legislative relations, and
electoral disproportionality (Vatter and Bernauer 2011; see also Chapter 2).
The two auxiliary hypotheses deal with the difference between representa-
tion through political parties and representation through political leaders.
Which is better for high degrees of policy representation, collective represen-
tation via parties or individual representation via leaders? Notwithstanding
the relevance of the question, we will not be able to answer it in any direct
way. However, indirectly it is possible to shed some light on it. We can do that
by entertaining two hypotheses. The first being that policy representation
among citizens in majoritarian systems is perceived as better channelled
through leaders than through parties. In consensus systems, we expect the
opposite—that voters should perceive better policy congruence via political
parties than via political leaders.

136
 Feeling Policy Represented

Our second auxiliary hypothesis may be somewhat controversial. It states

that among citizens in new democracies, policy representation through lead-
ers should be viewed as more successful than policy representation via politi-
cal parties. Among citizens in old established democracies our expectation is
reversed. Policy representation through parties should be viewed as better
than policy representation through political leaders.
The background to the hypothesis is how political rule has evolved histori-
cally. Rule through strong leaders and dominant individuals came first and is
the traditional form of ruling with roots way back in ancient times and car-
ried through the ages by the institution of chieftains, kings, queens, kaisers,
and czars; in modern times supplemented by authoritarian dictators.
Political parties as we understand them today are a more recent invention
going back only some 100–150 years in Western Europe and the USA, and
even fewer years in the rest of the world. Forming and sustaining cohesive
and competitive parties are a demanding process that requires resources. Sys-
tems based on competitive party rule are more modern and ‘advanced’ than
systems ruled by strong leaders. Hence, we expect voters in old, more mature
democracies to view representation through parties as more successful than
representation through leaders, while citizens in new, more recent democra-
cies have the opposite assessment—leader rule yields a better policy represen-
tation then party rule.

8.6  What Do We Already Know?

The answer to the question is in the headline—not much. Empirical studies

of issue agreement between voters and elected officials are rare. Comparative
investigations across different political systems are even rarer. The reason is
twofold. First, good studies of issue congruence are expensive to do. They
require comparable surveys among citizens as well as among elected politi-
cians. And surveys do not come cheap. Second, and this is the more impor-
tant factor, politicians in most political systems have a strong tendency not
to participate in surveys involving themselves. Elite access is a big problem in
studies involving elected representatives. The refusal rate is most often more
than 60 or 70 per cent, sometimes even worse.
In the largest comparative study of issue congruence between voters and
politicians ever done—it was done in Europe in 1996–7—the following refus-
al rates were encountered: in the European Parliament 50 per cent, in the
national parliaments in Italy 85 per cent, in Greece 80, in Portugal 76, in
France 75, in Spain 63, in the Netherlands and in Ireland 57, in Germany and
Luxembourg 53, in Belgium 42, and in Sweden 10 per cent (Schmitt and

137
Sören Holmberg

Thomassen 1999). Not very uplifting results. It is only the Swedish outcome
with a low refusal rate of 10 per cent which is fully acceptable. The results for
the other countries run the risk of ending up unduly shaky with such large
refusal rates among the politicians.
Things are no better across the Atlantic. In a comparative study in the 1980s
in which the USA participated, the refusal rate among members of the House
of Representatives was 71 per cent (Miller et al. 1999): about as bad as for
many European countries.
A Nordic study in the 1990s yielded somewhat better outcomes, although
not quite good enough in all countries. The refusals among legislators were
39 per cent in Finland, 37 per cent in Denmark, 28 per cent in Iceland, 12 per
cent in Norway, and 4 per cent in Sweden (Esaiasson and Heidar 2000). All
outcomes are decent but it is only the participant rate in Norway and Sweden
which is good.
The high rates of elite participation in Swedish parliamentary surveys are
not unique to the two referred studies. In a series of eight Riksdag Surveys that
the Swedish National Election Studies have conducted since the late 1960s,
the refusal rate has never exceeded 11 per cent (Holmberg 1974; Roth 1996;
Esaiasson and Holmberg 1996; Brothén and Holmberg 2010). The most recent
study was done in 2010 (Wängnerud, Esaiasson, Gilljam and Holmberg 2010).
Such is the comparative research situation on issue representation—a very
limited number of empirical studies and potentially unreliable results due to
a lack of good elite access. Yet, if we have to appreciate what we have, what
are the results? The simple answer is that the differences we find in levels of
policy congruence between citizens and politicians are not very pronounced
across different national systems (Miller et al. 1999; Schmitt and Thomassen
1999; Esaiasson and Heidar 2000).
A similar conclusion is valid when we compare countries with different
electoral systems. Issue agreement in the USA with a plurality-based electoral
system is no lower than in countries like Sweden and the Netherlands which
employ proportional systems. France, with a majoritarian system, exhibits
about the same degree of issue congruence as Germany, with its mixed elec-
toral system (Holmberg 1999). Thus, the hypotheses that countries using
majoritarian electoral systems should have lower levels of issue agreement
between voters and politicians compared to countries with proportional sys-
tems is not founded given these results.
Other studies in the tradition of opinion-policy research have indicated a
better fit between public opinion and policy outcomes in proportional sys-
tems compared to majoritarian systems, thus rendering support to the
hypotheses (Powell 2000). But there are also studies pointing at better poli-
cy representation in majoritarian systems than in proportional (Pierce
1999; Wessels 1999). In a very ambitious and pathbreaking study of how

138
 Feeling Policy Represented

r­epresentative democracy works in the three majoritarian systems of the

United States, Canada, and Britain, Stuart Soroka and Christopher Wlezien
find high degrees of policy representation. Most of the time, the authors
show, citizens get what they want from government (Soroka and Wlezien
2010).
The conclusion cannot be any other than that we do not have enough
comprehensive studies or good enough data to be able to give an evidence-
based verdict. The jury is still out on this one.
The same conclusion must even more decisively be drawn when it comes
to the hypothesis about new democracies being less successful in producing
high-level policy representation compared to old experienced democracies.
We simply do not have enough useful representation studies done in new or
emerging democracies. Most studies so far have been performed in the West.
We lack good investigations in the new democracies in Eastern Europe and in
the emerging democracies in Asia, Africa, and Latin America.

8.7  The Feeling of Being Policy Represented

An alternative and less expensive way of measuring policy representation is

not to measure policy representation. Instead we could measure the feeling of
being policy represented among citizens. That means that we will not study
‘real’ policy representation, but the extent to which people recognize that
their points of view are represented by elected politicians. Also, such a meas-
urement has the added advantage of not involving reluctant politicians
unwilling to participate in surveys.
But obviously, there is a downside as well. A ‘subjective’ measure of policy
representation is not the same thing as an objective measure. However, a
subjective measure is not irrelevant as of its own. Independently of how the
objective level of policy representation looks, it is relevant to study the extent
to which people feel that they are being represented by their elected officials.
The subjectively felt degree of policy representation says something about
how people judge the functionality of their own representative democracy. It
is indicative of how politically legitimate citizens perceive their rule to be
(Gilley 2009). Ceteris paribus, it is positive for a system if many of its citizens
perceive that they are well represented by their politicians. The opposite—
that most citizens feel that their politicians do not represent their views—is
more negative.
Our hypotheses are relevant even if we restrict our study to feelings of
being represented. We expect people in older democracies and in consensus
democracies as well as in countries with a proportional election system to
perceive higher levels of policy representation than citizens in newer

139
Sören Holmberg

­ emocracies, in majoritarian democracies, and in countries having m

d ­ ajority/
plurality elections.
The empirical test of the hypotheses commences in Table 8.1. The informa-
tion comes from The Comparative Study of Electoral Systems (CSES) data set
compiled in 2002–4 for 35 countries. The interview question asks about the

Table 8.1. Citizens in 35 countries judge the level of issue agreement between voters
and elected representatives in their own country

Country Election system Very well Rather well Sum Mean Percentage
percentage Don’t know

  1. Denmark PR 14 66 80 2,09 5
  2. USA Plurality/majority 13 59 72 2,20 3
  3. Ireland PR 8 57 65 2,34 8
  4. Spain PR 4 61 65 2,39 9
  5. Belgium PR 4 60 64 2,41 11
  6. Netherlands PR 6 52 58 2,41 2
  7. Switzerland PR 2 57 59 2,42 9
  8. Sweden PR 2 56 58 2,42 8
  9. Australia Plurality/majority 7 49 56 2,44 0
10. France Plurality/majority 4 56 60 2,45 1
11. Philippines Mixed 7 51 58 2,46 0
12. New Zealand Mixed 6 49 55 2,47 0
13. Iceland PR 4 51 55 2,48 16
14. Bulgaria PR 14 38 52 2,49 43
15. Taiwan Mixed 4 47 51 2,52 14
16. Hungary Mixed 3 48 51 2,52 6
17. Great Britain Plurality/majority 3 46 49 2,55 0
18. Finland PR 2 46 48 2,55 3
19. Poland PR 6 43 49 2,56 12
20. Romania PR 10 28 38 2,56 41
21. Germany Mixed 1 47 48 2,58 0
22. Chile PR 6 38 44 2,59 4
23. Mexico Mixed 10 37 47 2,61 7
24. Italy Mixed 3 43 46 2,63 17
25. Israel PR 4 43 47 2,64 7
26. Albania Mixed 4 40 44 2,65 4
27. Canada Plurality/majority 2 39 41 2,65 0
28. Russia Mixed 1 42 43 2,67 29
29. Portugal PR 1 37 38 2,70 25
30. Peru PR 9 30 39 2,74 8
31. Kyrgyzstan Plurality/majority 2 33 35 2,74 20
32. Slovenia PR 3 25 28 2,78 1
33. Japan Mixed 1 24 25 2,86 0
34. Czech Republic PR 1 27 28 2,90 15
35. Brazil PR 4 25 29 2,94 5
Mean 35 countries 5 47 52 2,55 10

Question: Considering how elections in (country . . . ) usually work in your view, to what extent do elections result in
members of parliament having views mirroring what voters want very well, rather well, not particularly well, not well
at all?”
Comments: The results come from The Comparative Study of Elections (CSES) involving voter surveys in the years
2002–2004. The election systems have been classified according to IDEA’s handbook Electoral System: The New Inter-
national IDEA Handbook (2005). The means can vary between very well (1) and not well at all (4); thus, low means
indicate high levels of issue agreement. Percentages have been computed among respondents who answered the
question.

140
 Feeling Policy Represented

extent to which citizens perceive that elections bring about issue agreement
between voters and elected representatives. The response alternatives are four
from very and rather well to not particularly well and not well at all. Observe
that the question, strictly speaking, measures perceptions of representation,
not feelings of representation. Yet the border between perception and feeling
is arguable.
A reasonable level for a passing grade is that a majority of citizens in a coun-
try perceive that they are policy represented by their elected politicians. In
our case this is operationalized as when a majority of citizens recognize that
their parliamentarians very well or rather well represent the views of the
voters.
The results show that a majority of countries did not achieve a passing
grade. It was close to a majority, though. In 16 out of 35 countries a majority
of citizens did perceive their elected legislators as mirroring the views of vot-
ers very or rather well. However, the average result across the 35 countries is
just over the passing bar with a mean of 52.
At the top of the list we find countries like Denmark and the USA with
80 and 72 per cent, respectively, of their citizens indicating that elected
politicians represent the views of people very or rather well. At the bottom
are countries like Brazil, Slovenia, the Czech Republic, and Japan. In these
countries, only between 25 and 27 per cent of the citizens feel that their
parliamentarians represent the opinions of people very or rather well.
Toward the middle of the list we find countries like Great Britain and Fin-
land, but also Hungary and Taiwan. About 50 per cent of the citizens in
these countries perceive their politicians to be very or rather representa-
tive of the views of people. Sweden is located on the upper half (rank 8)
with 58 per cent of the respondents indicating that the members of the
Riksdag represent the opinions of Swedes very or rather well. Neighbour-
ing Russia on the other hand ends up in the lower half (rank 28) with only
43 per cent of Russians feeling that their politicians’ views represent them
very or rather well.
Eyeballing the results does not reveal any evident patterns except perhaps
that a number of old established democracies tend to be placed high on the
list, while many new democracies end up toward the bottom. The more for-
mal hypothesis testing later in Table 8.3 will discern that this impression is
warranted. Citizens in older democracies do indeed feel more policy repre-
sented than people in emerging democracies.
Politicians in established democracies tend to concur. They too have a posi-
tive, although perhaps exaggerated, opinion of how well they represent their
voters. The CSES question on perceptions of the level of policy representation
has as well been put to members of the parliaments in Germany and Sweden
(Wessels 2004; Brothén and Holmberg 2010). Among members in Sweden as

141
Sören Holmberg

many as 81 per cent answered that the policy representation was very or
rather good. In Germany the result was 79 per cent. Voters in the two coun-
tries were less positive with 58 per cent among Swedes and 48 per cent among
Germans recognizing that their politicians were very or rather representative
of the views of the voters. To a degree members have a tendency to stray
towards wishful thinking (Holmberg 1999). Voters are more cynical or per-
haps more realistic.
The operative mechanism behind the fact that older democracies more
than young democracies have citizens who feel that they are better policy
represented by their elected representatives does not necessarily have any-
thing to do with the time factor or more experience. Notwithstanding that
practice increases skills, it is naïve to presuppose that higher levels of policy
representation would evolve in any automatic fashion over time or as a result
of voters and politicians encountering each other often in elections. There is
a multitude of other factors besides experience that play a role. It is rather
self-evident that levels of policy representation can increase as well as decrease
across time.
Sweden can serve as an illustrative example. The world’s longest time series
on policy representation between citizens and elected parliamentarians is
available for the Swedish case. And what we are talking about now is not
measures of feelings of being represented. What we are talking about are ‘real’
measures of policy representation between elite and mass. Beginning in
1968–9 and fairly regularly since then, members of the Riksdag and eligible
voters have been surveyed on identical issue questions numbering between
12 and 20 on the different study occasions.
Methodologically, the results from the oldest study in 1968–9 are not quite
comparable with later results. All issue questions were not identically formu-
lated and face-to-face interviews were used instead of mail questionnaires as
in the later studies. Thus, the comparatively low level of policy congruence
registered for 1968–9 must be interpreted with some caution, although the
result does not differ that much from the outcome of the later studies (Holm-
berg 1974). The comparability is better for the later studies starting in 1985
and going up to 2006.
The results (see Figure 8.1) indicate a fairly stable level of policy representa-
tion between people and politicians in the old established Swedish democ-
racy over the last 40 years. If there is any trend, however, it is not toward
increased levels of policy representation. Quite the opposite, there is a slight
tendency toward less policy representation if we compare the results in the
1980s and early 1990s with the last results in 2002 and 2006. Issue congru-
ence between citizens and elected politicians in Sweden has not improved
during the last 25 years. It has become slightly worse.

142
 Feeling Policy Represented

Percent Issues with the

Same Majority Position
Among Voters and Members
High 100
90
80 75 75
70 69
70 65 67 67
63
60
50
40
30
20
10
Low 0
Number 1969 1985 1988 1994 1996 1998 2002 2006
of issues 20 20 12 20 16 12 18 19

Figure 8.1.  Policy representation in Sweden, 1969–2006

8.8  Represented Through a Party or a Party Leader

The interview question on how people perceive the level of policy representa-
tion in their country is focused on the national parliament as a whole and on
all members. The question is not focused on what people de facto vote for, e.g.
political parties, candidates, and possibly also party leaders. Since it is quite
feasible to judge policy representation through parliament as a whole as want-
ing at the same time as policy representation via a party, a candidate, or a leader
is perceived as acceptable or even good, we might have a problem.2 However,
the problematique is noticed and to a degree covered in the CSES project. Two
special interview questions on party and party leader representation are includ-
ed in the questionnaires. Yet a third possible question covering representation
through individual candidates is not included in the CSES study.
The questions were phrased thus: ‘Would you say that any of the parties in
(country . . .) represent your points of view reasonably well?’ followed by
‘Irrespective of what you think about the parties, would you say there is any
party leader that represents your views reasonably well?’ For both questions,
the response alternatives were dichotomous yes or no. The results presented
in Table 8.2 confirm the suspicion that it would be common to perceive par-
ties and party leader as more policy representative than the entire parliament.
After all, you vote for a party or a leader, not for the parliament as a whole.

143
Sören Holmberg

Table 8.2.  The feeling of being policy represented by a party or a party leader among
­citizens in 35 countries

Percentage yes Party leader

Country Election system Party Party leader Difference

  1. Germany Mixed 91 75 +16

  2. Switzerland PR 87 80  +7
  3. Denmark PR 84 73 +11
  4. Australia Plurality/majority 83 79  +4
  5. Norway PR 82 72 +10
  6. New Zealand Mixed 80 83  –3
  7. Ireland PR 78 78  ±0
  8. Sweden PR 78 64 –14
  9. Czech Republic PR 78 56 +22
10. USA Plurality/majority 74 77  –3
11. Spain PR 74 73  +1
12. Great Britain Plurality/majority 73 67  +6
13. Hungary Mixed 73 81  –8
14. Canada Plurality/majority 69 68  +1
15. Israel PR 68 57 +11
16. Iceland PR 64 56  +8
17. Finland PR 64 51 +13
18. Albania Mixed 63 72  –9
19. France Plurality/majority 58 60  –2
20. Japan Mixed 57 53  +4
21. Portugal PR 56 59  –3
22. Mexico Mixed 48 38 +10
23. Bulgaria PR 46 44  +2
24. Romania PR 45 48  –3
25. Chile PR 44 70 –26
26. Italy Mixed 43 41  +2
27. Brazil PR 40 64 –24
28. Poland PR 40 39  +1
29. Russia Mixed 39 61 –22
30. Taiwan Mixed 37 46  –9
31. Peru PR 34 46 –12
32. Slovenia PR 29 35  –6
33. Philippines Mixed 29 31  –2
34. Korea Mixed 25 22  +3
35. Kyrgyzstan Plurality/majority 24 61 –37
Mean 35 countries 59 59  ±0

Question: Would you say that any of the parties in (country . . . ) represent your views reasonably well? Response alter-
natives: yes or no.
Question: Disregarding what you think of the parties, would you say that any party leader represents your views rea-
sonably well? Response alternatives: yes or no.
Comments: Based on CSES data; see Table 8.1. Percentages have been computed among respondents who answered
the questions. Non-response is on average below 10 per cent for both questions. The countries have been rank-
ordered according to the results on the party question.

144
 Feeling Policy Represented

On average across 35 countries, a majority of people acknowledge that

there is a party that represents their views (59 per cent). Results in the 50+
range are found in 21 of our 35 countries, that is, in a clear majority of the
investigated countries. Parties as successful vehicles for policy representation
are most frequently mentioned by citizens in Germany (91 per cent), Switzer-
land (87), and Denmark (84). The USA and Sweden also place themselves
high with rank numbers 10 and 8, respectively. At the bottom we find coun-
tries like Kyrgyzstan, Korea, and the Philippines with only some 24–29 per
cent of voters recognizing any political party as representing their views.
Another country located toward the bottom at rank 29 is Russia.3
Substantial majorities of people also claim that, besides the parties, there
are party leaders who represent their points of view. The average is once again
59 per cent and the number of countries with 50+ results is 25 out of 35.
Thus, across our 35 countries leaders are perceived as at least as useful for
achieving policy representation as parties. As a matter of fact, however, in
most countries, it is more common to deem party representation more suc-
cessful than leader representation. That is the case in 19 countries versus 15
countries where representation through leaders is more often mentioned
than representation via parties. Admittedly, a small difference. The most
appropriate conclusion is that there is a draw between party and leader
representation.
Countries that especially stick out as party-oriented in their voters’ percep-
tions of who represents them best are the Czech Republic (+22 points more
mentioning of party over leaders), Germany (+16), Sweden (+14), and Fin-
land (+13). Among countries where voters most clearly have the opposite
perception—finding leaders more useful than parties in providing policy
representation—a couple of new emerging democracies are most visible.
Most leader-oriented are citizens in Kyrgyzstan (+37 points more mentioning
of leader over party), followed by citizens in Chile (+26), in Brazil (+22), and
in Russia (+ 22).4

8.9  Testing Our Hypotheses on the Aggregate Level

As is readily noticeable in Table 8.1, old established democracies like Den-

mark, the USA, and Switzerland rank high when citizens assess whether their
parliamentarians represent their views or not. At lower ranks we tend to find
newer emerging democracies like Slovenia, Kyrgyzstan, and Brazil. Thus,
there are clear indications that our first main hypothesis is confirmed. Policy
representation through the national parliament is perceived as more success-
ful among citizens in old experienced democracies than in young inexperi-
enced democracies. The more systematic tabulations in Table 8.3 prove the

145
Sören Holmberg

point. Among citizens in old established democracies, on average 59 per cent

judge the policy mirroring of their members of parliament as very or rather
good. The comparable average result for citizens in new emerging democra-
cies is more modest, only 42 per cent.5
Looking at Table 8.1, the outcome for our second main hypothesis is less
evident. We find PR countries at the top (Denmark and Spain) as well as at the
bottom (Slovenia and Czech Republic). And similarly, we find high-ranked
majoritarian countries (USA) and low ranked (Kyrgyzstan). This lack of a
transparent outcome is borne out in the systematic test in Table 8.4. We con-
ventionally distinguish between three types of election systems—plurality/

Table 8.3.  Assessing policy representation among citizens in old and young democracies

Average percentage of citizens who:

Type of democracy Perceive that citizens’
opinions are very or
rather well represented
by members of national
parliaments
Consider that a party Consider that a party
represents their views leader represents their
views

Old established 59 75 70
democracies
New established 45 65 60
democracies
New emerging 42 43 51
democracies

Comments: See Tables 8.1 and 8.2. Classified as old established democracies are: Denmark, USA, Ireland, Belgium,
France, Switzerland, Sweden, Australia, New Zealand, Iceland, Finland, the Netherlands, Great Britain, Canada (14
countries). New established democracies are: Spain, Germany, Portugal, Israel, Italy, Japan (six countries). New
emerging democracies are: Bulgaria, Taiwan, Hungary, Poland, Mexico, Brazil, Czech Republic, Philippines, Chile,
Albania, Russia, Peru, Kyrgyzstan, Slovenia, Romania (15 countries). The percentages in the table are averages.

Table 8.4. Assessing policy representation among citizens in countries with different

­election systems

Average percentage of citizens who:

Election system Perceive that citizens’
opinions are very or
rather well represented
by members of national
parliaments
Consider that a party Consider that a party
represents their views leader represents their
views

Plurality/majority 52 65 69
Mixed 47 53 55
Proportional 50 61 59

Comments: See Tables 8.1 and 8.2. The number of plurality/majority systems is six, the number of mixed systems is
ten, while the number of proportional systems is 19. The percentages in the table are averages.

146
 Feeling Policy Represented

Table 8.5. Feeling policy represented among citizens in countries along Vatter’s and
­Bernauer’s executives–parties dimension

Average percentage of citizens who:

Perceive that citizens’ Consider that Consider that a party

opinions are very or a party represents leader represents
rather well represented their views their views
by members of national
parliaments

High values/ 52 64 60
consensus democracy
Middle values 44 58 61
Low values/majoritarian 55 60 59
democracy

Comments: Consensus–majoritarian democracy on the executives–parties dimension is measured adding up the

standardized scores of the effective number of parties, cabinet type, executive–legislature relations, and electoral dis-
proportionality using their average values over the period 1997–2006 (see Vatter and Bernauer 2011). High-value
countries are Switzerland, Brazil, Belgium, Denmark, Italy, Slovenia, Sweden, Netherlands, Finland, Israel, and Nor-
way. Middle-value countries are Germany, Poland, Chile, Spain, Russia, Hungary, Peru, Japan, Czech Republic, Roma-
nia, Bulgaria, Portugal, and New Zeeland. Low-value countries are Canada, USA, Australia, Great Britain, Korea,
France, Mexico, Philippines, Albania, Ireland, and Iceland.

majoritarian, proportional, and mixed. The results reveal no big differences

between the three election systems. No matter what election system is used,
about 50 per cent of citizens tend to feel being policy represented by their
elected politicians. But contrary to the hypothesis, the average result is high-
est among citizens in majoritarian systems (52 per cent), and not in propor-
tional systems (50 per cent). Consequently, our main hypothesis number two
is not confirmed.
Much like when we study ‘real’ policy representation, feelings of being pol-
icy represented are not more common among citizens in proportional sys-
tems, just like they are not less common among voters in majoritarian
systems. The main result is that there are no major differences between differ-
ent election systems when it comes to levels of policy representation between
citizens and their nationally elected politicians. In this context, election sys-
tem does not matter much.
Testing hypothesis number two by measuring the consensus–majoritarian
model via Vatter and Bernauer’s executive–parties dimension does not lead to
any different conclusion. Citizens living in countries with high (= consensus
democracies) or low (= majoritarian democracies) values on the executive–
parties dimension tend, on average, to feel to be policy represented to about
the same extent by their national parliaments. If we look more closely, it
turns out that the average results once more point in the opposite direction
than is spelled out in the hypothesis. Differences are small, but feelings of
being policy represented are not more common among people living in

147
Sören Holmberg

consensus democracies (52 per cent). They are somewhat more prevalent
among citizens in majoritarian democracies (55 per cent). There is no doubt.
Our hypothesis number two has not survived the empirical tests.
The same strong conclusion cannot be drawn when it comes to how citi-
zens assess policy representation through parties or through leaders; but
almost. There are some differences between how people living in countries
with different election systems judge policy representation via parties or via
leaders, but the differences are not very impressive. Yet what differences there
are, are in accordance with the hypothesis (Table 8.4). Citizens in majoritari-
an systems to a somewhat larger extent perceive successful leader representa-
tion than successful party representation. Citizens in proportional systems
tend to do the opposite evaluation. On average, they marginally put party
representation ahead of leader representation. Thus, our first auxiliary
hypothesis is confirmed, although just barely.
The word ‘barely’ should perhaps be emphasized. Studying the results from
the test using the executive–parties measurements reveals an even weaker
support for the hypothesis (see Table 8.5). Here, as before, more people in
consensus democracies view party representation as successful compared to
leader representation (64 vs 60 per cent). But among citizens in majoritarian
democracies feelings of being policy represented are not more common when
it comes to leaders (59 per cent) than when it comes to parties (60 per cent).
The difference is miniscule, but opposite to the hypothesis. Consequently,
our first auxiliary hypothesis is only partially and barely supported.
The support for the second auxiliary hypothesis is stronger. On average,
citizens in old established democracies put party representation ahead of
leader representation, while citizens in new emerging democracies tend to
put leader representation ahead of party representation. The evaluation of
party representation differs most markedly. Among citizens in old established
democracies an average of 75 per cent perceive party representation as suc-
cessful in achieving policy representation between voters and elected politi-
cians. The corresponding result for citizens in new emerging democracies is
much lower, only 43 per cent.
The result underscores that it takes time to establish a well-functioning
party-based democracy. Policy representation through parties is less ‘natural’
and more complicated to bring about than representation through leaders.
Thus, it takes more time and resources to make it work.

8.10  Multilevel Tests

Our finding that the age of democracy plays a decisive role when citizens get
to be policy represented by their elected leaders needs a tougher test. Before

148
 Feeling Policy Represented

we accept the finding, it should be proven resilient after relevant controls

have been applied in multivariate analyses. In Table 8.6 we try to do just that
in a series of multilevel regression analyses where citizens’ perceptions of
being policy represented by members of their national parliament are
regressed on a number of individual and system characteristics. On the sys-
tem level, age of democracy is challenged by our variable for the election
system design (Table 8.4) and the Vatter and Bernauer’s measure of the con-
sensus–majoritarian divide (Table 8.5). On the individual level, three controls
are applied. First a left–right congruence variable measured as the distance
between citizens’ self-placement on an 11-point left–right scale and their
placement of the party voted for. Second a support of government variable
indicating whether respondents voted for/prefer any governing party or an
opposition party. Third a variable for whether or not people have a party
identification. Our expectation, based on previous research, is that feelings of
being policy represented should increase if you have a short ideological dis-
tance to your favourite party, if you support a presently governing party, and
if you have a party identification.
The outcome of the multilevel tests vindicates our previous conclusion that
was based on only rather simple bivariate tests. Age of democracy survives the
controlled tests.6 Even after controlling for the three individual-level varia-
bles, age of democracy has a relevant and statistically significant effect on
how people feel policy represented. The two competing system-level varia-
bles—how the election system was set up and the extent of consensus versus
majoritarian rule—proved not to have any visible or significant impact on
how citizens perceived the extent to which their elected politicians were mir-
roring the views of the voters.
Similar tests using the same models and variables but substituting the
dependent variable for the variables measuring policy representation of party
and party leader, respectively, yield the same result. After relevant controls,
age of democracy matters, but not design of the electoral system or the con-
sensus–majoritarian divide.

8.11  No Need for Alarmism

It may seem like a rather sad outcome—especially to constitutional engi-

neers—that the differences between election systems used in democracies are
not more evident when it comes to levels of policy representation between
citizens and elected politicians, as well as when it comes to the feeling of
being policy represented among people. In this context election system
design seems not to matter much. Majoritarian, proportional, or mixed sys-
tems do about equally well. The political science constitutional tool box is

149
150

Sören Holmberg
Table 8.6.  Multilevel regression analysis of the impact of individual- and system-level characteristics on citizens’ perceptions of being policy
­represented by elected representatives in their own countries

Bivariate Multivariate

Model 1 Model 2 Model 3 Model 4

Regr Std Regr Std Regr Std Regr Std Regr Std
Coeff Err Coeff Err Coeff Err Coeff Err Coeff Err
System-level variables
Age of democr. rule –0.23*** 0.05 –0.23*** 0.06 –0.24*** 0.06 – – –0.19** .06
Electoral system design + 0.04 0.08 –0.02 0.07 – – – – – –
Consensus–majorit. democr. + 0.05 0.08 – –  0.05 0.07 – – – –
Individual-level variables
Left–right congruence  0.15*** 0.02 – – – –  0.11*** 0.03 0.11*** .03
Government party support –0.19*** 0.01 – – – – –0.15*** 0.01 –0.15*** .01
Party ID –0.19*** 0.01 – – – – –0.14*** 0.01 –0.14*** .01
Constant – – 20.69*** 20.65*** 20.61*** 20.71***
Number of countries – – 35 33 29 29
Number of individuals – – 51,553 48,208 25,280 25,117
Appr. R2 0.14 = 22% 0.15 = 17% 0.16 = 11% 0.14 = 22%

*p < 0.05, **p < 0.01, ***p < 0.001.

Comments: Data come from CSES Module 2. The dependent variable is scaled between 1 and 4 (see Table 8.1). All independent variables are scaled between 0 and 1. The three system-level
variables are trichotomized as in Tables 8.3–8.5 with old democracies scored 1, new established democracies 0.5, and new emerging democracies 0. Proportional electoral systems are
scored 1, mixed 0.5, and majoritarian systems 0. Consensus democracies have been given a score of 1, middle-value systems 0.5, and majoritarian democracies 0. The left–right congruence
is measured as the distance between citizens’ self-placement on an eleven-point left–right scale and the placement of the party voted for. The shorter the distance (= congruence) the lower
the score. The government party variable indicates whether respondents voted for/prefer any governing party (coded 1) or an opposition party (coded 0). Party identification is coded 1 for
respondents who answered yes to a question on whether they usually think of themselves as close to any particular party. Respondents who answered no were given the code 0. Approxima-
tive R2 is computed according to Kreft and De Leeuw (1998) and Dahlberg (2009). Thanks to Stefan Dahlberg for help with specifying the models and doing the data runs.
 Feeling Policy Represented

essential, but apparently less so in fine-tuning how the conduct of elections

affects policy representation between rulers and ruled in democracies.
Here, it appears, if not anything goes, at least most things go. The Majori-
tarian Model and the Consensus Model are not miles apart when it comes to
how to bring about feelings of being policy represented among citizens.
Instead, the two models are very close.
If the feeling among citizens of being policy represented is weakly related
to election system design in democracies, how old and mature the demo-
cratic system is in which people live is much more consequential. Policy rep-
resentation is deemed as functioning well by more people in old established
democracies than in new emerging democracies. And this goes for represen-
tation via parliaments, as well as for representation through political parties
and party leaders. Age and experience matter. The feeling of being policy
represented is more common among people in old democracies than in new
democracies.
The importance of this finding can hardly be overrated. Imagine the oppo-
site result, that more citizens in new democracies would be more content
with their level of policy representation than citizens in old established
democracies. Then, we would have begun to ponder what was going on and
wonder what was wrong. Speculations—and probably also ambitious
research projects like the Crisis of Democracy project in the 1970s—would be
launched (Crozier, Huntington and Watanuki 1975; see also Klingemann
and Fuchs 1995). Talk of wear-and-tear processes in old democracies would
surface. Petrifying symptoms would be looked for in the old party-based
democracies.
But now all this is not called for. Citizens in older established democracies
still feel better policy represented than citizens in young emerging democra-
cies. Furthermore, clear majorities of citizens in old democracies feel that
their elected politicians represent their views. Apparently, elections still
work as intended in the old democracies, at least when it comes to the essen-
tial task of bringing about a feeling of being policy represented among
people.
Thus, there is no need to be alarmist. Established consensus or established
majoritarian democracies are not in crisis. Elections and party-based repre-
sentative democracy deliver in both cases. The old steam engines still work.
The situation is less positive in many emerging democracies, however.
Consequently, if any form of alarmism is called for it should primarily be
directed toward new democracies. And here our best hope is that time and
more experience will do the job, since the political science tool box will be of
little help. Election system design does not matter much in this context.
What really matters is long experience with democratic elections. Different
election designs matter less.

151
Sören Holmberg

Notes

1. An analysis of this character based on CSES data is performed in Holmberg 2006,

although only for about 20 countries.
2. Looking at the average results for the Swedish parties, policy representation between
members and voters is higher than the comparable results for the entire parliament.
In the Swedish Representation Studies, the percentage issues with identical majori-
ties among members and voters have been the following on average for the parties
and for the parliament as a whole: 1968–9, 81 per cent/65 per cent; 1985, 86/70;
1988, 79/75; 1994, 81/75; 1996, 78/67; 1998, 74/69; 2002, 82/67; 2006, 71/63.
3. In most countries—in 23 out of 33—citizens more often acknowledge some form of
policy representation through parties than they perceive that their national parlia-
ment represents their opinions vary or rather well. Among the nine countries with
an opposite result—more people see the parliament as representative than on aver-
age see the parties as representative—we find seven emerging democracies (Philip-
pines, Bulgaria, Taiwan, Poland, Russia, Peru, and Kyrgyzstan) and two established
democracies with very unstable party systems (France and Italy). The result shows
a tie for one country (Chile).
4. As for parties, more countries have citizens who recognize the existence of policy
representation via party leaders than have citizens who judge their national parlia-
ment’s policy representation to be very or rather good. That is the case for 24 coun-
tries. Seven countries have citizens who place parliament representation ahead of
leader representation. Those countries are five emerging democracies (Philippines,
Bulgaria, Taiwan, Poland, Mexico) and Denmark and Italy. The result is a tie in one
country (France).
5. Old established democracies are: Denmark, USA, Ireland, Belgium, France, Switzer-
land, Sweden, Australia, New Zealand, Iceland, Finland, the Netherlands, Great
Britain, and Canada. As new established democracies we have classified: Spain,
Germany, Portugal, Israel, Italy, and Japan. New emerging democracies are: Bul-
garia, Taiwan, Hungary, Poland, Mexico, Brazil, Czech Republic, Philippines, Chile,
Albania, Russia, Peru, Kyrgyzstan, Slovenia, and Romania.
6. The variable for election system design and the Vatter/Bernauer consenus–majori-
tarian variable is highly correlated at 0.57, and is run separately against the age of
democracy variable in Table 8.6. However, the results are not changed if both are
included in an all-inclusive model. No matter how we specify our models, the elec-
tion system design and consensus–majoritarian variables have only weak and non-
significant effects on the extent to which citizens feel policy represented by their
politicians.

152
9

Output-oriented Legitimacy:
Individual- and System-level Influences
on Democracy Satisfaction
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

9.1 Introduction

It has long been recognized that popular evaluations of democratic regimes—

their legitimacy—respond to the performance of governments. The more
that successive incumbents—in Easton’s terms, the ‘political authorities’—
are able to deliver generally desired policy outcomes for their citizens, the
more positively they are likely to evaluate the regime itself.
This chapter considers a series of hypotheses about the cross-national
sources of satisfaction with democracy and the ways in which institutional
structures affect the relationship between support for incumbents and evalu-
ations of regime performance more generally.
These hypotheses examine the relative importance of three sets of factors:
those at the individual level; those relating to the macro-level characteristics
of the regime; and those that involve interactions across these two levels. The
various theoretical claims that are articulated are evaluated using CSES data
from 35 countries. The results show that the individual calculus of demo-
cratic satisfaction operates remarkably evenly across a range of different insti-
tutional contexts, with assessments of government performance playing a
predictably key role.
The results also show that, while the ‘unfairness’ of electoral system out-
comes that is typically associated with plurality rules reduces satisfaction lev-
els, this effect is more than offset by the positive effects of the greater ‘clarity
of responsibility’ that is also typically associated with plurality systems. This
finding has potentially important implications for any (re)design of electoral

153
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

systems aimed at increasing the overall level of regime satisfaction. The first
part of the chapter (Section 9.2) describes the core measure of democracy
satisfaction that is employed here and outlines the way in which it varies
both across countries and over time. Section 9.3 reviews the various theoreti-
cal claims that have been made about the sources of regime and democracy
satisfaction among mass publics. Section 9.4 specifies a model of democracy
satisfaction that enables these claims to be systematically evaluated. In
­Section 9.5 we report our empirical results.

9.2  Aggregate Variations in Democracy Satisfaction

Within established democracies, the public’s degree of ‘satisfaction with

democracy’ is often regarded as a good indicator of its evaluations of the
regime—of its assessments of the system of political rules and practices that
underpins the day-to-day cut and thrust of politics (Fuchs, Guidorossi and
Svensson 1995; Toka 1995). Levels of democracy satisfaction have been meas-
ured in a wide variety of national contexts, most typically using a four-point
scale derived from the survey question ‘On the whole, are you very satisfied,
fairly satisfied, not very satisfied, or not at all satisfied with the way democ-
racy works in (country)?’.
The most extensive over-time, cross-national source of measures of satisfac-
tion with democracy derives from the Eurobarometer series, where this ques-
tion has been asked annually of representative samples in all EU member
states since 1976 (with the exception of 1996).
Figure 9.1a shows the year-by-year variations in average democracy satis-
faction across the European Union between 1976 and 2006, the most recent
year for which comprehensive data are available.1 The graph clearly indicates
that, although satisfaction with democracy varies over time, there is no
­obvious linear trend either upwards or downwards—satisfaction has been
more or less constant across the EU for over three decades. One limitation of
­Figure 9.1a is that the average level of satisfaction is calculated across differ-
ent sets of countries over time, reflecting the various waves of EU accession
that occurred from the mid 1980s. Figure 9.1b accordingly shows how satis-
faction levels have changed since 1976 among the same set of nine countries
that were members of the EU from 1973. Again, as the graph indicates, there
is little evidence of secular change. Although average satisfaction levels across
the nine were slightly higher in the early twenty-first century than in the mid
1970s, the overall pattern is broadly one of trendless fluctuation.2
Table 9.1 reports the marginal distribution of satisfaction with democracy
across surveyed in the 2002–6 CSES wave. As the table shows, well over half
of respondents were either ‘very’ or ‘fairly’ satisfied with the democratic

154
 mean of demsat mean of demsat

0
1
2
3
0
1
2
3

19 19
7 7
19 5 19 6
7 7
19 6 19 7
7 78
19 7 19
7 7
19 8 19 9
7 8
19 9 19 0
8 8
19 0 19 1
8 8
19 1 19 2
8 8
19 2 19 3
8 8
19 3 19 4
8 8

(three) 1970s joiners, 1976–2006

19 4 19 5
8 8
19 5 19 6
8 8
19 6 19 7
8 8
19 7 19 8
8 89
19 8 19
8 9
19 9 19 0
9 9
19 0 19 1
9 9
19 1 19 2
9 9
19 2 19 3
9 9
19 3 19 4
9 9
19 4 19 5
9 9
19 5 19 6
9 9
19 6 19 7
9 9
19 7 19 8
9 9
19 8 20 9
9 0
20 9 20 0
0 0
20 0 20 1
0 0
20 1 20 2
0 0
20 2 20 3
0 0
20 3 20 4
0 0
20 4 20 5
0 0
20 5 20 6
0 07
20 6
Figure 9.1a.  Average democracy satisfaction in all EU countries, 1976–2006

07
Output-oriented Legitimacy

155
Figure 9.1b.  Average democracy satisfaction in (six) founder EU member states and
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

Table 9.1.  Distribution of satisfaction with

democracy across CSES Wave 2 countries

Percentage

Not at all satisfied 11.1

Not very satisfied 32.1
Fairly satisfied 48.6
Very satisfied  8.3
N of cases 56653

Question: On the whole, are you very satisfied, fairly sat-

isfied, not very satisfied, or not at all satisfied with the
way democracy works in (country)?

process in their respective countries, with only 11 per cent declaring them-
selves ‘not at all satisfied’. Figure 9.2 shows how the distribution of satisfac-
tion varied across the CSES sample of countries, presenting average scores on
the 1–4-point democracy satisfaction scale. As the figure indicates, Denmark
(mean = 3.27) displays the highest overall satisfaction levels, followed by Aus-
tralia (2.98) and the United States (2.97). The top ten countries are all either
Western European or North American. The countries with the lowest satisfac-
tion levels are mostly ‘new’ democracies like Bulgaria (1.83) and Brazil (2.05),
though Italy (2.26), Israel (2.26), and Switzerland (2.33) all score relatively
poorly given their relatively long-standing status as democratic systems.
­Germany also scores relatively modestly (2.45) given its strong democratic
record since the late 1940s. However, it is evident from Eurobarometer data
that overall German satisfaction levels fell after unification: prior to 1990,
(West) German satisfaction levels were close to the West European average
(see Appendix A.9.1).

9.3  The Sources of Democracy Satisfaction

It is clear from the evidence reported in Figures 9.1 and 9.2 that even within
democratic systems, public satisfaction with democracy varies over both time
and space. The key question is why these variations occur. As intimated
above, we consider these sources of variation under three general headings:
individual perceptions and characteristics; institutional features of the
regime; and interactions between perceptions and institutions.

9.3.1  Individual Perceptions and Characteristics

A large number of studies have argued that many of the political choices
made by ‘ordinary’ citizens are rooted in some sort of rational calculation
(Downs 1957; Stokes 1963 and 1992). This is not to suggest that, in order to

156
 Output-oriented Legitimacy

3,5
3
2,5
2
1,5
1
0,5
0
denmark
australia
usa
ireland
norway
sweden
spain
canada
britain
finland
new zealand
chile
iceland
japan
belgium
netherlands
taiwan
france
philipines
germany
romania
hungary
czech
switzerland
portugal
poland
israel
italy
peru
albania
russia
korea
mexico
brazil
bulgaria
Figure 9.2.  Average satisfaction with democracy on 1–4 scale, CSES Wave 2 countries

maximize their expected utility, most citizens must devote large amounts of
time to collecting and analysing all of the information that might be relevant
to the choices that they make. Rather, rational citizens frequently use heuris-
tics or cognitive shortcuts in order to make decisions where they have rela-
tively limited information (Popkin 1991; Lupia and McCubbins 1998). Thus,
for example, a rational voter might decide between competing parties on the
basis of her/his assessment of the likely competence of the rival party leaders.
In these circumstances, rather than making a detailed analysis of the policy
positions and delivery capabilities of the rival parties, s/he would be using the
heuristic of ‘leader images’ in order to make a ‘limited information’ but none-
theless rational choice (Lupia, McCubbins and Popkin 2000).
Using these sorts of cognitive shortcuts, one obvious way in which citizens
might evaluate the quality of the democratic process in their respective coun-
tries is through their assessments of the performance of the incumbent govern-
ment. In Easton’s terms, the more that the ‘political authorities’ can deliver
desirable policy goals, the more likely it is that citizens will develop positive
evaluations of the regime itself (Easton 1965; Kornberg and Clarke 1992).
There is certainly evidence from individual country studies to suggest that
‘valenced’ judgements about incumbent party policy delivery are positively
linked to satisfaction with democracy (see, for example, Clarke et al. 2004,
2009). The simple hypothesis tested here is that, independently of national
context, democracy satisfaction will be positively affected by citizens’ assess-
ments of overall government performance (H1.1). A second set of potential
individual-level influences on democracy satisfaction relates to citizens’ eval-
uations of the extent to which domestic political institutions represent their
views and interests. The sense that citizens are ‘represented’ by those in polit-
ical power is a central tenet of all democratic theory (Weale 2005). If citizens
feel that existing political institutions fail to reflect and represent their inter-
ests, then the legitimacy of the political system as a whole is likely to be
brought into question. In contrast, if citizens believe that these institutions

157
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

do effectively represent their concerns, then they are more likely to consider
that the state wields legitimate authority—they are more likely to be satisfied
with the democratic process. The hypothesis here is that the more an indi-
vidual feels represented by existing political institutions, the more likely s/he
is to be satisfied with democracy (H1.2). A third set of individual influences
focuses on citizens’ beliefs about the ability of the state to deliver desirable
democratic goals or values, such as the protection of human rights and the pre-
vention or avoidance of political and financial corruption (Powell 2000). The
more citizens perceive that the existing regime delivers these values, the more
likely they are to regard the existing regime as legitimate and accordingly to
lend it their democratic support. In short, there should be a positive relation-
ship between democracy satisfaction and the belief that the current regime
effectively delivers democratic values (H1.3). Note finally that individual
demographic characteristics—age, sex, education, religion, and so on—might
also affect people’s satisfaction with democracy. In general, we make no par-
ticular assumptions about the form that any such effects might take, though,
as discussed in the next section, our operational model takes full account of
the effects of these characteristics as control variables.3

9.3.2  Institutional Features of the Regime

The potential influences on democracy satisfaction identified in the previous
subsection all involve a degree of limited information rationality. Since we
consider that people in general are broadly rational in many aspects of their
attitudes and behaviours, we expect these individual-level influences to oper-
ate across all of the countries in the CSES sample. This said, it would be decid-
edly odd, given the known tendency for institutional arrangements to
condition individual attitudes and behaviour, if the varying institutional
contexts of the CSES countries did not exert some sort of effect on individual
citizens’ propensities to be satisfied with democracy (see, for example, Cox
and Amorim 1997; Whiteley et al. 2010).
Indeed, there are good reasons to suppose that democracy satisfaction
might well be affected by various regime characteristics. Perhaps the most
obvious is the form of the electoral system (Taagepera and Shugart 1989;
­Gallagher and Mitchell 2008). The Anglo-Saxon first-past-the-post ‘plurality’
system, in which the winner in each constituency is the individual who
receives the most votes, typically produces a ‘clear’ election-winning party in
the national assembly (Lijphart 1994 and 1999). However, first-past-the-post
also tends to produce governments with clear parliamentary majorities that
have been elected on far less than 50 per cent of the votes—with the implica-
tion that a powerful government has been elected when more than 50 per
cent of voters failed to support it. Moreover, first-past-the-post also tends to

158
 Output-oriented Legitimacy

concentrate electoral competition in a relatively small number of ‘marginal’

constituencies, thereby rendering the voting choices of many citizens in non-
marginal seats largely irrelevant in terms of electoral outcomes. In contrast
with the plurality system, the various forms of proportional representation
(PR) actively seek to ensure that governments will find it difficult to achieve
a secure parliamentary majority without receiving at least 50 per cent of the
popular vote. In addition, most PR systems are also designed to maximize the
chances that all individuals’ votes will count equally—that a vote for Party X
will be of equal importance regardless of the constituency in which it is cast
(Balinski and Young 2001). It seems likely in these circumstances, given the
clear injustices associated with plurality electoral rules, that democracy satis-
faction will be lower in countries with plurality rules than it is in countries
with PR electoral systems (Norris 2004). In short, net of other effects, it is
expected that plurality rules will exert a negative impact on democracy satis-
faction (H2.1).
Note, however, that the potential effects of plurality refer to the possible
consequences for public attitudes of formal electoral rules. It could be argued
that the practical effects of official rules on the composition of parliaments
and of governments are rather more important than the actual rules them-
selves. One possible way of assessing these practical effects involves examin-
ing the relationship between the vote shares and the parliamentary seat
shares that parties receive. It is obviously the case that the greater the overall
disparity between parties’ vote shares and their respective seat shares, the
more ‘unfair’ the outcome of the electoral process. A situation where Party X
receives 35 per cent of the vote and 60 per cent of the seats is clearly less fair
than one where Party X receives 35 per cent of both votes and seats. It in turn
seems reasonable to suppose that the more unfair the electoral outcome, the
less citizens will be satisfied with the democratic process. In short, we hypoth-
esize that vote/seat share disparities should exert a negative effect on democ-
racy satisfaction (H2.2).
A second way of assessing the practical effects of electoral rules involves
considering the extent to which there is ‘clarity of responsibility’ among the
parties in government (Powell and Whitten 1993). This ‘practical effect’
clearly relates to one of the possible positive features of plurality rules. In a
situation where a single party with well under 50 per cent of the votes can
easily obtain a clear majority of parliamentary seats, there is no need for coali-
tion government. Single-party government—in which only one party is rep-
resented in terms of cabinet posts—implies absolute ‘clarity of responsibility’.
Only one party is running the country; that party can clearly be held respon-
sible for whatever policy successes or failures occur during a given parliamen-
tary term; and voters can clearly hold the governing party to account in an
election—either re-electing a successful party or ‘throwing the rascals out’ in

159
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

the event of policy failure. Multiparty government, in contrast, implies a lack

of clarity of responsibility. The more parties that share cabinet posts, the less
clear it is which party is responsible for particular (or even general) failures of
policy. It is accordingly much more difficult to reward one particular party for
‘good’ performance or to punish another for ‘bad’. In short, in may be impos-
sible for voters to decide ‘who the rascals are’, let alone to ‘throw them out’ in
an election. Given the clear advantages of democratic electorates being able
to ‘throw the rascals out’, our expectation is that democracy satisfaction will
be higher where there is greater ‘clarity of (cabinet) responsibility’ (H2.3). Note
that this hypothesis implicitly recognizes the potential practical benefits of
plurality electoral systems and that as such it provides a clear counter-hypoth-
esis to H2.1 above.
A further ‘institutional feature’ relates to the age of the regime itself. The
group of countries analysed here includes both ‘new’ and ‘old’ democracies.4
Given that regime legitimacy—the positive evaluation of regime ­performance—
is something that is built progressively over time, it is reasonable to suppose,
ceteris paribus, that newer democratic systems will display lower levels of sat-
isfaction with democracy than older systems, which have had longer to devel-
op a sense of popular legitimacy. The simple hypothesis here, therefore, is that
being a ‘new’ democracy should exert a negative effect on democracy satisfac-
tion (H2.4). Our final ‘institutional feature’ is the general performance of the
national economy. Previous research (e.g. Clarke et al. 2009) has established
that democracy satisfaction tends to be higher (lower) when the economy is
doing well (badly). Using the unemployment rate as a proxy for (poor) eco-
nomic performance, we hypothesize that unemployment is likely to exert a
negative effect on democracy satisfaction (H2.5).

9.3.3  Interactions between Individual Perceptions and Regime

Characteristics
Thus far we have considered the effects of individual calculation and insti-
tutional contexts separately. We have assumed that in arriving at their judge-
ments about the adequacy of the democratic process, individual citizens will
use various cognitive shortcuts based on limited information rationality. We
have also assumed that their judgements will be influenced by certain insti-
tutional features—relating to both rules and outcomes—of the countries in
which they live. There is clearly a further set of possibilities, however. It is
conceivable that some institutional contexts may produce stronger (or
weaker) individual-level effects on democracy satisfaction than others. Here,
we consider three such individual/context interactions, all of which relate
to the potentially context-varying effects of perceptions of government
performance.

160
 Output-oriented Legitimacy

As noted above, one of our core individual-level hypotheses is that satis-

faction with government performance, other things being equal, is likely to
build regime legitimacy and support—to increase people’s satisfaction with
democracy. It is possible, however, that the magnitude of this effect varies
with institutional context. Consider, first, how the effect of government per-
formance on democracy satisfaction might vary according to the maturity of
the democratic system. In ‘old’ democracies, citizens are likely to be able to
distinguish fairly clearly between their feelings about the democratic system
generally and their evaluations of the government of the day. In contrast, in
‘new’ democracies, citizens are less likely to distinguish between govern-
ment and regime performance, since the character of the (new) regime is in
large measure defined by what the current government does. The implica-
tion here is that although individuals’ assessments of government perfor-
mance will affect democracy satisfaction in both old and new democracies,
the magnitude of the effect is likely to be stronger in ‘new’ regimes than it is
in ‘old’ (H3.1).
This principle of potentially differential effects can be extended to the
other institutional features of regimes that were indicated above—clarity of
responsibility, vote/seat disproportionality, and plurality electoral rules.
Consider, first, the possible interaction between government performance
and clarity of responsibility. If government performance is poor in a situa-
tion where clarity of responsibility is relatively low, citizens are relatively
unlikely either to be able to identify the precise ‘rascals’ who are responsible
for that poor performance or to use their votes in order to remove those ras-
cals from office. In these circumstances of relative impotence, dissatisfaction
with government is more likely to translate into dissatisfaction with democ-
racy more generally. In contrast, if government performance is poor in a situ-
ation where clarity of responsibility is high, if they are so minded, voters can
remove the governing party from office precisely because they know ‘who
the rascals are’. In these circumstances, dissatisfaction with government is
less likely to translate into democracy dissatisfaction because appropriate
remedial change can be instituted. In sum, where clarity of responsibility is
relatively high, the effects of assessments of government performance on
democracy satisfaction are likely to be smaller than the equivalent effects in
low clarity systems (H3.2).
The opposite pattern is expected for the interaction between assessments of
government performance and vote/seat disproportionality. In highly dispro-
portional systems, dissatisfaction with government is more likely to translate
into dissatisfaction with democracy because people are more likely to con-
sider that an unfair electoral system has in part been responsible for the poor
government performance that is being experienced. This in turn implies that
where disproportionality is high the effects of government performance on

161
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

democracy satisfaction are likely to be greater than the equivalent effects in

low disproportionality systems (H3.3).
Finally, we consider the possible interaction between assessments of gov-
ernment performance and plurality electoral rules. Since plurality systems are
held to produce both high clarity of responsibility and greater vote/seat dis-
proportionality, it is possible that the interaction effects anticipated in
hypotheses H3.2 and H3.3 could cancel each other out. Equally, it is also possi-
ble that one of the effects could dominate the other. We accordingly investi-
gate the possible interaction effect between government performance effects
and plurality versus non-plurality systems without any prior expectation as
to the direction of the effect (H3.4).

9.4  Specifying a Model of Democracy Satisfaction

There are broadly two ways of specifying models using the sort of multilevel
data typical of the CSES dataset. Traditional multilevel models involve, in
effect, the specification of two statistical models—one for the effects on the
dependent variable of predictors that are measured at the individual level;
and one for the effects on the coefficients of the individual-level model of
predictors that are measured only at the aggregate, country level. A second,
simpler approach—which is adopted here—is to estimate the effects of both
individual-level and aggregate-level effects on individual variations in the
dependent variable simultaneously, but to take explicit account of the fact
that the data are clustered by country.
This latter approach ensures that the correct (robust) standard errors are
used for estimating the significance of aggregate-level effects. It also has the
advantage of interpretative simplicity, particularly in terms of comparing the
relative explanatory power of individual and aggregate effects on the depend-
ent variable.5
Taking account of the various hypotheses that were advanced above, our
model of democracy satisfaction takes the following general form:
Democracy Satisfactioni = b0 + Σb1–5 Individual Political Perceptionsi
+ Σb6–8 Individual Demographic Controlsi
+ Σb9–k Institutional/System Characteristicsj
+ Σbk–m Interactions between Individual Perceptions
of Government Performance and Institutional/System
Characteristicsij + eij (1)

where i subscripts denote individual-level predictors; j subscripts denote

country-level predictors; ij subscripts denote cross-level interactions; and eij is
a random error term.

162
 Output-oriented Legitimacy

Our modelling strategy considers three different variants of this general

model specification. We develop and test each of these variants separately
because of the limited degrees of freedom available at country level (N = 35 for
the data analysed here) in the CSES dataset. The first variant considers the
potential effects of the Lijphart-inspired measures of consensus versus majori-
tarian democracy that have been employed throughout this volume. It will
be recalled that Lijphart distinguishes between an executive dimension of con-
sensualism (which describes the extent of the constraints on the formation of
single-party government) and a federal dimension (which describes the extent
of the constraints on governments to implement their chosen policy plat-
forms). The first variant of our general specification accordingly includes terms
for executive–consensualism and federal–consensualism in the ‘­institutional/
system characteristics’ segment. We also include cross-level interactions
between government performance assessments and each of these consensual-
ism terms. As in our earlier discussion of the interaction between government
performance assessments and plurality election rules, we are agnostic as to the
signs on these interaction terms. Following the logic of our discussion of the
effects of plurality, however, we expect both measures of consensualism to
exert significant positive effects on democracy satisfaction. Our first variant of
our general specification is accordingly:
Democracy Satisfactioni = b0 + b1 General Government Performancei
+ b2 Voters Represented in Electionsi + b3 Party Represents
Respondent’s Viewsi
+ b4 Government Respects Human Rightsi
+ b5 Corruption is Widespread in Respondent’s Countryi
+ b6 Agei + b7Male/noti + b8 Educationi
+ b9 National Unemployment Ratej
+ b10 Executive Consensualismj + b11 Federal Consensualismj
+ b12 GovernmentPerformance* Executive Consensualismij
+ b13 GovernmentPerformance* Federal Consensualismij + eij (2)

where the i and subscripts and eij are as defined in [1]; b1–b5 are expected to be
significant and positive; there are no expectations for the signs or significance
levels for the demographic controls b6–b8; a negative sign is expected for b9;
positive signs are expected for the two consensualism terms, b10 and b11; and
there are no expectations for signs or significance levels of the interaction
terms, b12 and b13. Note that the coefficients b12 and b13 on the interaction
terms indicate the shift away from the ‘parent’ government performance
coefficient, b1, for the specified macro-level characteristic.6
The second variant of [1] considers the effects of formal electoral rules and
whether or not a country is a ‘new’ democracy. We accordingly substitute
plurality/not and new democracy/not for the two Lijphart measures of

163
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

consensual democracy in the third and fourth segments of the model as

follows:
Democracy Satisfactioni = b0 + Σb1–5 Individual Political Perceptionsi
+ Sb6–8 Individual Demographic characteristicsi + b9 National
Unemployment Ratej
+ b10 Plurality/notj + b11 New Democracy/notj
+ b12 GovernmentPerformance*Plurality/notij
+ b13 GovernmentPerformance*New Democracy/notij + eij(3)

where all terms and expectations for b1–b9 are as in [2]; b10 and b11 are both
expected to be negative; there are no prior expectations for b12; and b13 is
expected to be positive.
Our final variant focuses on the consequences of electoral rules rather than
on their character. Accordingly we substitute terms for disproportionality
and clarity of responsibility in the third and fourth segments as follows:
Democracy Satisfactioni = b0 + Σb1–5 Individual Political Perceptionsi
+ Σb6–8 Individual Demographic characteristicsi + b9 National
Unemployment Ratej
+ b10 High Disproportionality/notj + b11 High Clarity/notj
+ b12 GovernmentPerformance*High Disproportionality/notij
+ b13 GovernmentPerformance*High Clarity/notij + eij(4)

where all terms and expectations for b1–b9 are as in [2]; b10 and b13 are both
expected to be negative; and b11 and b12 are expected to be positive.7
Table 9.2 indicates how each of the terms in equations [2]–[4] relates to the
hypotheses that were advanced earlier and reports the marginal distributions
of the operational variables across the CSES sample of countries considered
together. As the table indicates, H1.1 is tested using the respondent’s general
assessment of government performance. Hypotheses H1.2 and H1.3 are each
operationalized using two indicator measures: H1.2 by assessments of the con-
sequences of elections for representation and by whether or not the respond-
ent considers her/his own views to be represented by a particular political
party; and H1.3 by perceptions of the prevailing level of respect for individual
freedom and by assessments of political corruption. The remaining hypoth-
eses are tested using a single measure. H2.1 uses the standard CSES coding of
plurality electoral rules, where 1 denotes plurality rules and zero otherwise.8
The disproportionality index used in H2.2 reflects the difference between par-
ties’ shares of the popular vote and the shares of seats that they obtain in the
lower national assembly. The index ranges from a minimum 0.3 (for the
Netherlands), which indicates a high correspondence between vote and seat
shares and a low level of disproportionality, to a maximum of 7.1 (for the
UK), which indicates a high level of disproportionality.9 In order to simplify

164
 Table 9.2.  Descriptive characteristics of key predictors in democracy satisfaction model

Very bad job (1) Bad job (2) Good Job (3) Very Good job (4) N

H1.1 General government 11.9 35.1 48.5 4.4 53183

performance
Not at all Well (1) Not Very Well (2) Quite Well (3) Very Well (4)
H1.2 Elections ensure voters’ views 10.2 40.7 44.0 5.1 48951
are represented by major
parties
No (0) Yes (1)
Does any party represent 42.2 57.8 53253
respondent’s views?
None (1) Not Much (2) Some (3) A Lot (4)
H1.3 Respect for individual freedom 8.7 27.2 48.4 15.7 56645
and human rights in (nation)
Hardly happens at Not Very Widespread (2) Quite Widespread (3) Very Widespread (4)
all (1)
Corruption among politicians 6.0 23.2 37.6 33.2 56143
in (nation) is . . .
No (0) Yes (1)
H2.1 Live in plurality system 72.6 27.4 60651

Output-oriented Legitimacy
H2.4 New democracy 70.5 29.5 60651
Minimum Maximum Mean St. Deviation
H2.2 Disproportionality of votes 0 7.1 2.41 1.60 36
to seats
H2.3 Clarity of responsibility in 0 1 0.51 0.35 36
cabinet
H2.5 Unemployment rate 3 19.4 7.02 3.69 36
Lijphart federal–consensualism –1.43 2.25 0.00 1.00 35
Lijphart executive–consensualism –2.42 1.47 0.00 1.00 36

Note: Cell entries except where indicated are row percentages. Numbers in (brackets) indicate numerical codes assigned to each category in constructing independent variable measures.
Country scores on plurality, disproportionality, clarity of cabinet responsibility and new/old democracy are provided in Appendix A.9.2.
165
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

the interpretation of the cross-level interaction terms, for our model estima-
tions we convert the disproportionality index into a dummy variable in
which countries above the mean (high disproportionality) are scored 1 and
those below the mean are scored 0. The clarity of responsibility variable used
to test H2.3 is a dummy, where 0 indicates that cabinet seats are shared among
more than one party and 1 indicates that only one party is represented in
cabinet and holds all the seats.10 Finally, H2.4 is tested using a simple dummy
variable that reflects whether the respondent lives in a new or an old democ-
racy, while H2.5 is tested using the national unemployment rate for the year in
which the relevant CSES survey was conducted.11
Table 9.2 also shows that the distributions of most of the predictor variables
are reasonably well dispersed. For example, in relation to general government
performance, roughly 53 per cent of respondents take a positive view com-
pared with 47 per cent who take a negative view. Similarly, some 71 per cent
consider that corruption is either ‘very’ or ‘quite’ widespread, compared with
29 per cent who think it is not. The table also shows (see the figures in brack-
ets) the numerical values assigned to the categories of each of the independ-
ent variables. Since the variables are all at quasi-interval level, all measures are
scored so that high values reflect agreement with or positive attitudes towards
the concept specified. Thus, for example, the belief that ‘Elections ensure vot-
ers views’ are very well represented by major parties’ is scored as 4, whereas the
belief that voters views’ are not at all well represented is scored as 1. Similarly,
the conviction that ‘Corruption among politicians’is very widespread is scored
as 4, whereas the view that corruption hardly happens at all is scored as 1. The
remaining, binary variables are scored as 0/1 dummies. Appendix A.9.2 reports
the actual country-by-country scores for each of our macro-level variables.

9.5  Empirical Results

Tables 9.3 and 9.4 report the bivariate relationships between democracy sat-
isfaction and, respectively, the individual-level and macro-level predictor
variables from equations [2]–[4]. Table 9.3 shows the simple bivariate correla-
tions between the four-point democracy satisfaction scale and each of the
individual-level predictors. The correlations are all significant at p < 0.0001,
though this is to be expected with such a large number of cases.
The observed relationships are consistent with all of the individual-level
hypotheses outlined in H1.1–H1.3. The government performance measure cor-
relates positively with democracy satisfaction, providing initial support for
H1.1. Similarly, the two representation measures both correlate positively with
democracy satisfaction, indicating preliminary support for H1.2. This pattern
of preliminary support also extends to the regime performance measures. As

166
 Output-oriented Legitimacy

Table 9.3. Bivariate correlations between democracy satisfaction and individual-level

predictors

Democracy satisfaction

H1.1: General government performance 0.40

H1.2: Elections ensure voters are represented by parties 0.30
H1.2: A party represents respondent’s views 0.22
H1.3: Respect for freedom and human rights in (nation) 0.40
H1.3: Corruption among politicians in (nation) –0.34

N = 40056. All correlations significant at p = 0.0000.

H1.3 anticipates, the ‘respect for freedom and human rights’ scale correlates
positively with democracy satisfaction, whereas the ‘corruption is wide-
spread’ scale correlates negatively.
Table 9.4 yields a more modest set of preliminary conclusions with regard
to the macro-level correlates of democracy satisfaction. Since the macro char-
acteristics are all dummy variables, the table reports simple difference of
means t-tests. For comparability, the unemployment and Lijphart consensu-
alism variables referred in [1] are transformed into dummies, in which values
above the mean on each variable score 1 and all other values score 0. The
results, at best, lend only weak support to our initial set of macro-level hypoth-
eses. As the table shows, although there are minor variations in democracy

Table 9.4.  Difference of means tests on democracy satisfaction (1–4)

scale

Mean Score N of cases

‘Old’ democracies 2.66 24

‘New’ democracies 2.27 11
Difference .39
Non-plurality systems 2.58 24
Plurality systems 2.48 11
Difference .10
Low disproportionality of seats to votes 2.56 12
High disproportionality of seats to votes 2.53 23
Difference .10
Low clarity of responsibility in cabinet 2.63 13
High clarity of responsibility in Cabinet 2.40 22
Difference 0.23
Low federalism–consensualism 2.51 17
High federalism–consensualism 2.60 18
Difference 0.09
Low executive–consensualism 2.52 20
High executive–consensualism 2.59 15
Difference 0.07

Notes: Cases are countries. No differences significant at p < 0.05.

167
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

satisfaction levels across the various types of system (e.g. PR systems on aver-
age score .10 points higher on the 1–4 scale than plurality systems, while
‘new’ democracies score 0.39 higher than ‘old’ ones), none of the differences is
statistically significant at conventional levels. In short, although the differences
shown in Table 9.4 are generally in the theoretically predicted direction, none
of them is substantial enough to provide clear support for any of the theoreti-
cal claims advanced earlier about direct macro-level effects.
The key question in these circumstances is whether or not any of these appar-
ently weak effects are strengthened when they are considered in a multivariate
context. Table 9.5 examines the relevant models—one for the Lijphart consen-
sualism terms (Model A); one for electoral rules and new/old democracy status
(Model B); and one for electoral outcomes (Model C). Since the democracy sat-
isfaction scale, strictly, is at ordinal level, the models respectively estimate equa-
tions [2]–[4] using ordered logit. As noted above, significance levels are estimated
using robust standard errors, with the data clustered by country.12 The table
reports one-tailed significance tests for all relationships where a clear direction-
al effect is hypothesized; two-tailed tests are reported otherwise. The models are
all reasonably well-determined, producing pseudo-R2 values of 0.17.13
Several general conclusions are suggested by the results reported in Table 9.5.
First, with regard to the hypothesized individual-level effects, H1.1 to H1.3 are
clearly confirmed in all three model specifications. As predicted, the govern-
ment performance terms (which vary between b = 0.82 and b = 0.84) and both
of the sense of representation terms provide significant positive coefficients.
The ‘Elections ensure voters are represented’ takes the same value (b = 0.51)
across all three equations and the ‘Party represents the respondent’s view’
term varies between b = 0.35 and b = 0.40. In relation to regime performance,
also as predicted, the ‘Respect for . . . human rights’ terms consistently yields
a positive effect (b = 0.68 across all three speci­fications) while the ‘Corruption’
term produces a negative effect (which varies between b = –0.36 and b = –0.42).
These results indicate that the individual-level calculus of satisfaction with
democracy is relatively straightforward. Satisfaction is maximized when citi-
zens perceive that their governments deliver on key policy goals; when they
feel they are represented by the major political parties; and when they con-
sider that regimes deliver respect for human rights and minimize corruption.
The second set of conclusions suggested by Table 9.5 relates to the macro-
level and cross-level sources of individual variations in democracy satisfaction.
Here, however, the pattern of significance in the coefficients varies consider-
ably across the three specifications. The results shown in Model A clearly
indicate that neither of the measures of consensualism exerts a significant
effect on democracy satisfaction. The non-significance of the corresponding
interactions between government performance and the executive and federal
consensualism terms also indicates that these regime characteristics do not

168
 Table 9.5.  Ordered logit models of democracy satisfaction

A: Lijphart-inspired executive– B: Objective features model— C: Electoral outcomes model—

and federal–consensualism electoral rules and new/old disproportionality and clarity of
model democracy responsibility

Coefficient St error p Coefficient St error p Coefficient St error p

Individual-level effects, H1.1–H1.3

H1.1: Government performance .82 0.09 0.000 0.82 0.09 0.000 0.84 0.09 0.000
H10.2: Sense of representation
Elections ensure voters are represented 0.51 0.05 0.000 0.51 0.05 0.000 0.51 0.05 0.000
A party represents respondent’s views/not 0.38 0.05 0.000 0.35 0.04 0.000 0.40 0.04 0.000
H1.3: Regime performance
Respect for freedom and human rights 0.68 0.04 0.000 0.68 0.04 0.000 0.68 0.04 0.000
in (nation)
Corruption among politicians in (nation) –0.42 0.05 0.000 –0.36 0.04 0.000 –0.39 0.05 0.000
Macro-level effects, H2.1–H2.5
Executive–consensualism 0.11 0.13 0.165
Federal–consensualism –0.04 0.17 0.457
H2.1: Plurality electoral rules/not –0.28 0.37 0.226
H2.4: New democracy/not –0.62 0.55 0.131

Output-oriented Legitimacy
H2.2: High disproportionality/not –0.93 0.48 0.027
H2.3: High clarity of responsibility/not 0.97 0.44 0.014
H2.5: National unemployment rate in survey –0.02 0.02 0.149 –0.00 0.03 0.451 –0.01 0.02 0.251
year
Cross-level interaction effects, H3.1–H3.4
Government performance* –0.04 0.05 0.147
executive–consensualism
Government performance* 0.06 0.07 0.314
federal–consensualism
H3.1: Government performance* new 0.11 0.20 0.296
democracy/not
169
170

David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

Table 9.5.  (continued)

A: Lijphart-inspired executive– B: Objective features model— C: Electoral outcomes model—

and federal–consensualism electoral rules and new/old disproportionality and clarity of
model democracy responsibility

Coefficient St error p Coefficient St error p Coefficient St error p

H3.4: Government performance*plurality 0.01 0.15 0.470

electoral rules/not
H3.3: Government performance* high 0.32 0.17 0.036
disproportionality/not
H3.2: Government performance* high –0.33 0.18 0.031
clarity /not
Demographic controls
Age 0.00 0.00 .588 0.00 0.00 0.418 0.00 0.00 0.345
Male/not 0.01 0.03 0.822 0.00 0.03 0.884 0.01 0.03 .822
Education 0.02 0.02 0.144 0.03 0.02 0.108 0.03 0.02 .115
Cut 1 1.48 0.28 1.59 0.33 1.69 0.32
Cut 2 3.85 0.32 3.97 0.38 4.08 0.34
Cut 3 7.18 0.37 7.30 0.44 7.40 0.41
Pseudo R2 0.17 0.17 0.17
N 39798 39798 39798

Note: Robust Standard Errors reported for 35 clusters of countries. Significance levels report one-tailed test for all directional hypothesized relationship; two-tailed tests for others.
 Output-oriented Legitimacy

interact with assessments of government performance in the determination

of satisfaction with democracy. A similar set of conclusions is suggested by
the macro- and cross-level results shown in Model B. Once again, both the
plurality rules and new/old democracy terms are clearly non-significant; and
the same pattern of clear non-significance extends to the interactions between
these two terms and government performance assessments. Note, finally,
that in both Models—and in Model C—the unemployment rate term is con-
sistently non-significant, implying that we can unambiguously reject H2.5.
This general pattern of macro-level non-significance does not extend, how-
ever, to Model C, which tests the effects of practical electoral outcomes rather
than institutional rules or the age of democratic institutions. In Model C we
find that both disproportionality and clarity of responsibility have significant
effects in the predicted directions—negative for disproportionality (b = –0.93)
and positive for clarity (b = 0.97). We also find, as predicted in H3.2, that the
interaction between government performance and clarity produces a signifi-
cant negative coefficient (implying that the effects of government perfor-
mance on democracy satisfaction are attenuated when clarity is high) and, as
predicted in H3.3, that the interaction between government performance and
disproportionality produces a significant positive effect (implying that the
effects of government performance on democracy satisfaction are greater
when disproportionality is high).14
What all of this suggests is that, when appropriate statistical controls are
applied, the key macro drivers of democracy satisfaction appear to revolve
around the practical consequences of electoral rules (disproportionality and
clarity) rather than the rules themselves (consensualism or plurality elections).
Plurality and consensualism appear to have no direct or interacting effects on
democracy satisfaction but their consequences do. The key insight that Model
C provides is that two main consequences of plurality—­disproportionality and
clarity—both affect democracy satisfaction, but they do so in opposing direc-
tions. People tend to dislike the unfairness associated with disproportionality,
but they simultaneously tend to like clarity of responsibility since it strength-
ens their ability to identify the ‘rascals’ who may need to be ‘thrown out’ in
subsequent elections. Ceteris paribus, the former mechanism reduces democ-
racy satisfaction; the latter enhances it. This result has resonance both for
those who argue in favour of proportional representation electoral systems
and those who oppose them. For PR’s protagonists, the reductive effect on
democracy satisfaction of disproportionality clearly supports their claims that
the ‘unfairness’ of plurality rules, in terms of ‘wasted votes’, weakens regime
support. For PR’s opponents, the positive effect of clarity clearly supports their
claims that democracy works most effectively when electoral rules produce
clear winners who can be held properly to account by the electorate at some
later date.

171
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

Two other sets of comments are in order about the Model C results. The first
is simply that the demographic control variables all fail to yield significant
coefficients—as they also fail to do in Models A and B. The second concerns
the precise interpretations of the cross-level interaction terms. Recall that
these interactions were included in the model specifications to incorporate
the idea that the effects of government performance vary across different
institutional contexts. That none of the cross-level terms in Models A and B
is significant means we can safely ignore the possible effects that they repre-
sent. In the context of Model C, it was expected that the effects of govern-
ment performance on democracy satisfaction would be stronger in high
disproportionality systems (H3.3) and weaker in high clarity systems (H3.2).
These predictions imply: (1) that the interaction term for government
performance*disproportionality should be significant and positive; and (2)
that the interaction term for government performance*clarity of responsibil-
ity should have a significant negative coefficient that is smaller in magnitude
than the (positive) ‘parent’ government performance coefficient.15 The results
shown in the table indicate that these expectations are borne out. The effect
of government performance on democracy satisfaction in low disproportion-
ality systems is b = 0.84; in high disproportionality systems it is b = (0.84 +
0.32 =)1.16. The countervailing effect of clarity in low clarity systems is b =
0.84; in high clarity systems it is b = (0.84–0.33=)0.51. In short, H3.2 and H3.3
are strongly supported by Model C.
Note, however, that one of the difficulties of using ordered logit as an estima-
tion tool—as with any form of logistic regression—is that the relative magni-
tudes of the various coefficients are not easy to interpret directly. For this reason,
Table 9.6 reports, for each independent variable in Table 9.5 Model C, the
changes in probabilities associated with moving from the predictor variable’s
minimum to maximum values whilst holding all other predictors constant at
their respective means. The results are highly instructive since they enable us to
assess the relative explanatory power of the different hypotheses that were
advanced earlier. The most obvious feature of the table is that the individual-
level predictors generally have larger impacts on the changes in probability
than the system-level structural features. The three largest effects—for govern-
ment performance (dp = 0.55), elections ensure representation (dp = 0.35), and
respect for human rights (dp = 0.36)—are all part of the individual-level calculus
of democracy satisfaction. For example, as an individual moves from registering
1 to registering 4 on the general government performance scale, s/he increases
her/his probability of being either very or fairly satisfied with democracy by
p = 0.55. Similar effects (though not quite as large) are also evident with regard
to the terms for regime performance and sense of representation. This is not to
say, however, that structural features of the regime do not matter. It is clear from
the dp values for disproportionality (–0.23) and for clarity of responsibility (0.21)

172
 Output-oriented Legitimacy

Table 9.6. Changes in predicted probabilities derived from the model reported in

Table 9.5 (column C)

Range dp value: Change in predicted

probability of being Satisfied
with Democracy

Government performance
H1.1: Government performance assessments 1–4 0.55
Sense of representation
H1.2: Elections ensure voters are represented 1–4 0.35
H1.2: A party represents respondent’s 0–1 0.10
views/not
Regime performance
H1.3: Respect for freedom and human rights 1–4 0.36
in (nation)
H1.3: Corruption among politicians in 1–4 –0.27
(nation)
Regime characteristics
H2.2: Disproportionality of seat share/vote 0–1 –0.23
share
H2.3: Cabinet clarity of responsibility 0–1 0.22
Cross-level interactions
H3.2: Government performance*high 1–4 0.27
disproportionality
H3.2: Government performance*high clarity 1–4 –0.31
of responsibility
Demographic controls
Age 16–100 0.04
Male 0–1 0.00
Education 1–4 0.04

Note: Change in probability figures record the change in the probability that an individual will be satisfied with
democracy (either fairly or very satisfied) given a change in the independent variable from its minimum to its maxi-
mum possible value, holding all other variables constant at their respective means. Estimation using CLARIFY for
STATA (Tomz et al. 1999).

that these structural factors also have important effects. The disproportionality
effect implies that a person living in a high disproportionality system would be
p = 0.23 less likely to be satisfied with democracy than someone with an identi-
cal attitudinal and demographic profile in a majoritarian system. Yet before it is
immediately concluded that democracy satisfaction could straightforwardly be
increased by a move to a more proportional selectoral system, it needs to be
recognized that a similar effect in the opposite direction is evident with regard
to the impact of clarity of responsibility (dp = 0.22). Here, an increase from 0 to
1 on the clarity scale produces an increase in the probability of an individual
being satisfied with democracy. Given that greater clarity of responsibility is
typically associated with plurality rules (even though the bivariate correlation
between our measures of plurality and clarity is only r = 0.19), it is evident that
any consequent loss of clarity associated with the abandonment of plurality
would simultaneously serve to cancel out any benefits to democracy satisfac-
tion that might have been expected to derive from that abandonment.16

173
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

9.6  Summary and Conclusions

Democratic regimes retain their legitimacy over time by securing positive

evaluations of regime performance among their respective mass publics. By
measuring the extent to which citizens feel satisfied with the democratic pro-
cess in their respective countries, a plausible assessment of the extent to
which different regimes enjoy popular approval can be made. It is clear from
Eurobarometer and CSES data that satisfaction with democracy—the positive
evaluation of regime performance—varies both over time within specific
countries and across different political systems. The key question is why these
variations occur, and in particular how they relate to governmental perfor-
mance. Using a cross-sectional comparative design, this chapter has explored
the individual- and system-level sources of these variations.
The empirical results that we have reported suggest support for a number of
hypotheses. At the individual level, it is clear that satisfaction with democra-
cy responds most directly to variations in government performance. Govern-
ments build (or lose) approval for the regime by providing citizens with the
policy outcomes that they prefer; the more (less) successful that they are in
this enterprise, the greater (smaller) the level of democracy satisfaction. In
addition to the actions of the government of the day, however, democracy
satisfaction is also conditioned by citizens’ perceptions of the performance of
the regime more generally. The more that people feel parties and elections
perform some sort of representation function, the more likely they are to dis-
play high levels of democracy satisfaction. Similarly, the more citizens believe
that the regime delivers on two of the key features of contemporary liberal
democracy—respect for individual freedom and human rights, and the mini-
mization of political corruption—the more positively they will evaluate that
regime. Crucially, the results presented here show that these relationships
hold, controlling for a wide range of standard demographic variables, across
all of the CSES countries and across various model specifications.
The individual calculus of satisfaction with democracy, however, only tells
part of the story. There is also a role for political—and particularly electoral—
institutions. Critics of the first-past-the-post electoral system frequently point
out that (a) the unfairness to minority parties and (b) the ‘wasted votes’ of citi-
zens who live in ‘safe’ constituencies that are engendered by plurality rules
mean that satisfaction with democracy is typically lower in plurality systems
than it is in systems based on proportional representation. At the purely
descriptive level, this observation carries some weight: it is undoubtedly the
case that democracy satisfaction is on average lower in plurality systems than it
is under PR. However, this simple observation taken in isolation ignores anoth-
er very important fact: when we include theoretically relevant individual-level
controls (as in Table 9.5, Model B), plurality rules exert no effect on democracy

174
 Output-oriented Legitimacy

satisfaction whatsoever. Moreover, when we consider the key twin conseq­

uences of plurality rule—disproportionality and clarity of ­responsibility—in an
appropriately specified model (as in Table 9.5, Model C), we find that they exert
significant, but opposing, effects on democracy satisfaction. Disproportionality
damages democracy satisfaction because it is unfair; clarity of responsibility
enhances it because it means that the ‘rascals’ responsible for government per-
formance can be readily identified and, if necessary, voted out of office. Since
clarity of responsibility and disproportionality are often low in countries with
PR, where multiparty coalition governments are common, would-be institu-
tional reformers find themselves in a somewhat paradoxical position. If the
objective of reform is to maintain, or even to increase, the overall level of
approval for the regime, the abandonment of plurality rules is very much a
double-edged sword. Though a switch to PR, on the one hand, might serve to
increase democracy satisfaction by removing unfairness to minor parties and
‘wasted votes’, on the other hand it is also likely to reduce clarity of responsibil-
ity and therefore simultaneously to reduce such satisfaction. Indeed, the results
of our ordered logit model suggest that these two countervailing effects would
in all probability cancel each other out. This suggests that advocates of electoral
reform need to be very careful in specifying their precise aims and priorities if
their putative reform proposals are to avoid producing unintended (and poten-
tially undesirable) consequences.
One other institutional feature that has figured in some accounts of democ-
racy satisfaction is the age of the democratic system itself. Again, it is cer-
tainly the case at the simple descriptive level that democracy satisfaction is
lower in ‘new’ as opposed to ‘old’ democracies. The results reported here,
however, show conclusively that when appropriate individual-level controls
are applied, the apparent (reductive) effects of ‘being a new democracy’ disap-
pear. The implications of this finding are both important and straightfor-
ward. New democracies exhibit lower levels of democracy satisfaction than
their older counterparts not simply because they are new but because their
citizens are more critical of current government policy delivery, regime per-
formance, and representation.
Our results show, finally, that there are two important cross-level interactions
between institutional characteristics and the individual calculus of democracy
satisfaction. First, where there is high clarity of responsibility in policymaking,
the effects of government performance on democracy satisfaction—though
still highly significant—are muted; the effects are much greater where clarity is
low. The reason for this effect, we infer, is that in high clarity systems it is much
easier to identify and remove the culprits when policy performance is poor,
which reduces the extent to which government dissatisfaction translates into
regime dissatisfaction. In contrast, in low clarity systems, because it is more dif-
ficult to identify and remove ‘rascals’, dissatisfaction with government

175
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

performance converts more directly into dissatisfaction with the regime and
with democracy itself. The second effect works in the opposite direction. Where
disproportionality is high (low), the effects of government performance on
democracy satisfaction are greater (smaller). We infer here that this reflects the
tendency for citizens in highly disproportional systems to attribute poor (good)
performance to the unfairness of the electoral system, which in turn translates
into greater dissatisfaction (satisfaction) with the existing democratic system.
In any event, these twin effects offer a further caution to would-be electoral
reformers bent on removing the injustices of first-past-the-post.

Appendix A.9.1:  Variations in democracy satisfaction in Western

European countries, 1976–2006

3 Belgium 2.5
Britain

2
mean of demsat

mean of demsat

2
1.5

1
1

.5

0 0
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

4 Denmark 3 Germany

3
mean of demsat

mean of demsat

1
1

0 0
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

France 3 Ireland
2.5

2
mean of demsat

mean of demsat

1.5

1
1

.5

0 0
1975
1976
1977
1978
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

1975
1976
1977
1978
1977
1979
1980
1981
1982
1983
1984
1985
1986
1987
1988
1989
1990
1991
1992
1993
1994
1995
1996
1997
1998
1999
2000
2001
2002
2003
2004
2005
2006
2007

176
 Chile
mean of demsat

Brazil
mean of demsat mean of demsat

0
1
2
3
0
1
2
3
0
.5
1
1.5
2
2.5

Albania

Canada
Belgium

Bulgaria
Australia

Denmark
1975 1975
1976 1976 1975
1977 1977 1976
1978 1978 1977
Italy

1979 1979 1978

Spain

Czech Republic
1980 1980 1979

Greece
1981 1980
1981 1981
1982 1982
1983 1982
1983 1983
1984 1984

in the chapter
1985 1984
1985 1985
1986 1986 1986
1987 1987 1987
1988 1988 1988
1989 1989 1989
1990 1990 1990
1991 1991 1991
1992 1992 1992

0.3
2.7
3.3
5.7
3.6
1.1
1.6
3.9
4.2
1993 1993 1993
1994 1994 1994
1995 1995 1995
1996 1996 1996
1997 1997 1997
1998 1998 1998
1999 1999 1999

Disproportionality
2000 2000 2000
2001 2001 2001
2002 2002 2002
2003 2003 2003
2004 2004 2004
2005 2005 2005
2006 2006 2006
2007 2007 2007

mean of demsat mean of demsat

0.09
0.32
0.15
1.00
0.70
0.49
0.04
0.32
0.53
0
1
2
3
0
1
2
3

Clarity of
1975
1976 1975

responsibility
1977 1976
1978 1977
1979 1978
1980 1979
1981 1980
Portugal

1982 1981
1983

(0)
1982
1984
Netherlands

1983
1985 1984
1986 1985
1987 1986
1988 1987

0
0
0
0
0
0
0
0
0
1989 1988
1990
1991 1989
1992 1990
1993 1991
1994 1992
1993

Plurality(1) / not
1995
1996 1994
1997 1995
1998 1996
1999 1997
2000 1998
2001 1999
2002 2000
2003 2001
2004 2002
2005 2003
2006 2004

0
1
1
0
1
1
0
0
1
2007 2005
2006
2007

(1) / not (0)

Appendix A.9.2:  Country scores on regime characteristics used

New democracy
Output-oriented Legitimacy

177
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

Disproportionality Clarity of Plurality(1) / not New democracy

responsibility (0) (1) / not (0)

Finland 1.4 0.22 0 0

France 2.52 1.00 0 0
Germany 1.7 0.27 0 0
Iceland 1.2 0.00 0 0
Ireland 2.4 0.37 0 0
Israel 0.9 0.39 0 0
Hungary 4.1 0.23 0 1
Italy 1.5 0.31 0 0
Japan 2.3 0.44 1 0
Korea 6.5 1.00 1 0
Mexico 1.0 0.30 1 0
Netherlands 0.3 0.20 0 0
New Zealand 0.8 0.45 1 0
Norway 1.4 0.25 0 0
Peru 5.3 1.00 0 1
Philippines 2.3 0.27 1 1
Poland 2.3 0.45 0 1
Portugal 1.8 0.36 0 0
Romania 4.5 0.25 1 1
Russia 2.3 1.00 0 1
Slovenia 1.3 0.23 0 1
Spain 1.8 1.00 0 0
Sweden 0.4 1.00 0 0
Switzerland 1.1 0.11 0 0
Taiwan 2.2 1.00 0 0
UK 7.1 1.00 1 0
USA 1.5 1.00 1 0

Notes

1. The index is the mean, constructed by scoring the response categories as very
satisfied = 4, fairly satisfied = 3, a little dissatisfied = 2, and very dissatisfied = 1.
2. Appendix A.9.1 reports country-by-country variations in democracy satisfaction
for these nine countries. Three broad patterns are evident: trendless fluctuation
­(Belgium, France, Netherlands, Luxembourg); slight over-time increase (Italy, Ire-
land, Denmark, Britain); and steep decline (Germany, where, after unification, the
lower levels of satisfaction in the east produced an overall national decline). For
similar evidence, see Norris (2011).
3. One other individual-level variable, not considered here, that potentially could
affect democracy satisfaction is whether or not the individual voted for (one of)
the winning party(ies) in the previous election (e.g. Anderson et al. 2005). We
suspect that our government performance assessments variable captures much of
this effect and we accordingly exclude it from our analysis.
4. Among the countries analysed here, we define the following countries as new
democracies: Albania, Brazil, Bulgaria, Chile, Czech Republic, Hungary, Philip-
pines, Poland, Romania, Russia, and Slovenia.

178
 Output-oriented Legitimacy

5. As we report below, we also specified and estimated a formal multilevel version of

the ‘clustered’ model summarized here. As we also report below, the substantive
results from this multilevel model are identical in all important respects to those
reported here.
6. Note that the interaction between government performance and clarity of respon-
sibility is constructed by first splitting the clarity term into a dummy variable
where 1 denotes a clarity score equal to or greater than the mean and 0 denotes a
clarity score below the mean. This form of construction renders the interaction
coefficient much more interpretable and makes it directly comparable with the
other interaction terms in the model, which are constructed from government
performance and a dummy variable.
7. High disproportionality refers to those cases where disproportionality is greater
than the mean; high clarity refers to those cases where clarity of responsibility
scores 1, signifying that only one party is represented in government.
8. Plural systems in the CSES dataset are Japan, South Korea, Mexico, New Zealand,
Philippines, Romania, USA, and UK. Including the two ‘mixed’ systems (Germany
and Slovenia) in either the plurality or the non-plurality category makes no differ-
ence to the statistical results reported here.
9. The disproportionality index is constructed as Σ(abs(%Vote Sharei– %Seat
Sharei))/N of Parties where i denotes the ith party.
10. The original clarity of responsibility index from which our dummy is constructed
was defined as Σ(abs(%Share of Cabinet Seatsi– %Share of Cabinet Seatsj))/N of
Parties; where any pair of parties are subscripted i, j, and where i is not equal to j.
A score of 0 on this index indicates that cabinet seats are distributed evenly among
the parties in a coalition government (for example, Iceland’s cabinet was com-
posed of 12 ministers, six from each of two coalition partners); a score of 1 indi-
cates that only one party is represented in cabinet and holds all the seats.
11. Identical models to those reported in Table 9.5 below were also estimated using
the annual change in the unemployment rate for the year in which the CSES sur-
vey was conducted. These alternative specifications produced substantively iden-
tical results to those reported in Table 9.5: none of the unemployment terms
achieved statistical significance in any of the estimated equations and the other
coefficients were unaffected.
12. Three of the CSES countries are excluded from the analysis reported here, all on
the grounds of missing data: these are Hong Kong, Kyrgyzstan, and Peru. The CSES
2 dataset contains more than one election survey for some countries (e.g. Taiwan).
In these cases, data from only the second election are included in the estimation,
to prevent the results from any one country counting disproportionately in the
analysis.
13. We tested the robustness of the findings reported in Table 9.5 by estimating the
models as formal multilevel models using HLM. For each model, this involved
specifying a fixed effects two-equation model. The level 1 model predicted
­individual-level democracy satisfaction from the set of individual-level predictors
indicated in equations [2]–[4]. The level 2 model predicted the Intercept term of
the level 1 model from the macro-level predictors in [2]–[4]. The level 2 model also

179
David Sanders, Harold Clarke, Marianne Stewart, and Paul Whiteley

included cross-level interaction terms for government performance*plurality;

government performance*disproportionality; government performance*clarity
of responsibility; and government performance*new democracy/not. The pat-
terns of coefficient signs and significance levels are virtually identical in the Table
9.5 and the formal multilevel model specifications.
14. The negative coefficient on the clarity term in this multivariate specification is not
an artefact of multicollinearity between clarity and plurality. The bivariate corre-
lations among plurality, disproportionality, and clarity are as follows: clarity–­
plurality: r = 0.27; plurality–disproportionality: r = 0.30; clarity–disproportionality:
r = 0.37.
15. A negative interaction coefficient indicates a weaker effect in the subset of cases
interacted than for the ‘parent’ term; a positive interaction coefficient indicates a
stronger effect.
16. One further feature of Table 9.6 that merits brief attention are the changes in prob-
ability associated with the government performance*disproportionality and gov-
ernment performance*clarity interaction terms. The positive sign on the former,
for example, means that, for high disproportionality countries, the effect of a
change in government performance from a rating of 1 to 4 increases the probabil-
ity that an individual will be satisfied with democracy by dp = (0.55 + 0.27) = 0.82.
The effect for low disproportionality countries remains dp = 0.55. A similar inter-
pretation applies for the latter: in high clarity countries, the change in probability
of being satisfied with democracy increases by dp = (0.55–0.31) = 0.24. The effect
for low clarity countries remains dp = 0.55.

180
10

The Multiple Bases of Democratic

Support: Procedural Representation
and Governmental Outputs
Mark Peffley and Robert Rohrschneider

10.1 Introduction

How does the ability of a democracy to adequately represent its citizens influ-
ence public support for the regime? Since the essence of a democracy is to
represent its citizens, there is little doubt that citizens’ evaluations of how
well they are being represented will influence their support for democratic
government. But as we argue below and as our analysis will show, the rela-
tionship between representation and democratic support is more complex
than it seems at first blush because both political representation and demo-
cratic support assume many forms, some more consequential than others. If
citizens feel poorly represented, for example, they are likely to downgrade
their evaluations of the performance of a democratic regime almost as a mat-
ter of course. A more telling question, however, is whether inadequate repre-
sentation leads to a more serious, and presumably long-term, decline in
support for democracy as a form of government. Once citizens begin to ques-
tion whether democracy as an ideal is flawed in one way or another, the legiti-
macy of the regime and the reservoir of support for it during times of stress
begin to erode.
Therefore, in this chapter, we are interested in how political representation
impacts both forms of democratic support—i.e. democratic performance
evaluations, or public satisfaction with how well the current democratic gov-
ernment is working, as well as democratic ideals, or public support for democ-
racy as a form of government.
Just as democratic support assumes many different forms, so does political
representation. Although most representation studies focus on the idea of
substantive representation, or the correspondence between the outcomes

181
Mark Peffley and Robert Rohrschneider

citizens want (e.g. policies, goods, performance) and those produced by the
government, an equally important component of representation is its proce-
dural dimension. Citizens expect governmental procedures and the processes
by which government works to be fair. Procedural fairness is concerned less
with outcomes and more with the processes by which governmental policies
are made and administered (Tyler 1988; Dahl 1989). According to Tyler (1988:
103), irrespective of the policies and the outcomes generated by the govern-
ment, citizens’ evaluations of procedural fairness are based on assessments of
whether authorities are motivated to be fair, are honest, and follow ethical
principles of conduct, whether opportunities for representation were provid-
ed, and whether authorities behaved in a biased fashion. Importantly, judge-
ments of procedural fairness or unfairness have heady consequences for
citizens’ behaviour beyond mere expressions of dissatisfaction with elected
officials. Widespread perceptions of procedural unfairness undermine citi-
zens’ willingness to obey laws and authorities, as well as their fundamental
perceptions of governmental legitimacy (Tyler 1990; Hibbing and Theiss-
Morse 2002).
Thus, unlike the previous chapter by Sanders et al., which focuses more on
the impact of economic outputs on satisfaction with how well the current
democratic government is working or performing, our focus is primarily on
the impact of citizens’ evaluations of procedural fairness on both forms of
democratic support—i.e. democratic satisfaction and democratic ideals, with
a particular interest in the latter. In other words, the primary independent
and dependent variables of theoretical interest in our study are quite different
from those of the Sanders et al. chapter.
The importance of citizens’ evaluations of procedural unfairness has been
demonstrated in a variety of contexts. In the USA, for example, studies find
that one of the most important reasons why the public ‘hates’ the US Con-
gress is that the national legislature is perceived to violate expectations of
procedural fairness (e.g. Hibbing and Theiss-Morse 1995, 2002). More gener-
ally, comparative studies find that the procedural fairness of judiciaries and
bureaucracies has an important impact on citizens’ evaluations of parties and
elections (Sanches-Cuenca 2000; Rohrschneider 2005), and even how they
evaluate the EU (Rohrschneider and Loveless 2010). Publics in nations where
so-called arbitrating institutions (i.e. judiciaries and bureaucracies) work well
are much more likely to feel represented by the government than citizens in
countries with poorly working judiciaries and bureaucracies. This finding
comports well with a large volume of research by Tyler and others (e.g. Lind
and Tyler 1988) documenting how citizens’ direct experiences with authori-
ties in their everyday lives—with police, bureaucrats, and legal authorities—
has a major impact on how they view government. Specifically, when people
are treated unfairly by authorities, they quickly generalize those lessons in

182
 The Multiple Bases of Democratic Support

formulating their views of the fairness of the broader system (e.g. Peffley and
Hurwitz 2010). For most citizens, encounters with police, bureaucrats, and
legal authorities are as close as they come to an experience with the
government.
Experiences with the authorities of arbitrating or ‘output’ institutions are
doubtless even more important outside the USA, given the limited number of
elections during a typical parliamentary cycle. Thus, it comes as no surprise
that previous research in advanced democracies finds that citizens’ experi-
ences with ‘output’ institutions influence how citizens evaluate ‘input’ insti-
tutions, such that when the former work well, publics also believe that input
institutions—parties and politicians—represent the public interest (Rohrsch-
neider 2002, 2005). In the current study, we focus on more far-reaching con-
sequences of poorly functioning arbitrating institutions for regime
evaluations: does their procedural quality influence citizens’ support for
democratic government—not just their satisfaction with the way democracy
is working, but their more fundamental support for democracy as a form of
government?
We will contrast the influence of the procedural quality of output institu-
tions on regime evaluations with an important alternative explanation for
regime support that emphasizes the degree to which national institutions
are consensual rather than majoritarian (Lijphart 1998; Powell 2000).
Because regimes that rely on consensual institutions, such as federalism and
proportional representation, disperse power broadly to a greater number of
groups—often providing opposition parties with some degree of representa-
tion, consensual democracies are expected to gain the support of a relatively
broad crosssection of the public. By contrast, because majoritarian institu-
tions concentrate political power in the hands of fewer veto players, majori-
tarian democracies are more likely to be supported by those who voted for
the winning majority party than those who sided with a losing party. In
short, democratic support is likely to be more pervasive in regimes character-
ized more by consensual than majoritarian institutions and our study will
compare this institutional explanation of democratic support with one out-
lined previously—the importance of the procedural quality of output
institutions.
We will proceed as follows. We first locate the properties of our two forms
of democratic support—i.e. evaluations of regime performance and regime
ideals—on a ladder of increasing abstraction ranging from concrete evalua-
tions of the performance of a regime to support for the constitutional ideals
of a democracy. Following that, we consider the character of substantive and
procedural representation, which suggests several hypotheses that will be
tested in the empirical analysis, after which we draw more general implica-
tions from the results of our study.

183
Mark Peffley and Robert Rohrschneider

10.2  Levels of Regime Support: A Conceptual Ladder

of Abstraction

As a preliminary matter, it is useful to clarify the nature of various types of

regime support—the dependent variable in our analyses, and their attendant
measures (see also the Appendix A.10.1). In Figure 10.1 (Rohrschneider and
Schmitt-Beck 2002: 37), we array six different types of regime support from
the most concrete assessment at the bottom of the ladder, labelled ‘demo-
cratic reality’ and defined as assessments of the performance of the incum-
bents of a regime, to ‘constitutional ideals’, at the top of the ladder, which is
indicated by the public’s more abstract evaluations of the democratic ideals
of a regime. One advantage of conceptualizing regime support in this way is
that it allows us to locate the plethora of measures of institutional support
scholars have studied to gain a more comprehensive understanding of their
likely antecedents and consequences. Measures designed to tap constructs at
the bottom of the ladder, for example, are more likely to be shaped by various
governmental performance criteria, such as how government handles the
economy or foreign policy, whereas political values and other, non-perfor-
mance criteria are expected to exert more influence as one moves up toward
the ideational pole of the continuum. And, as noted by a variety of scholars,
the consequences of a government experiencing a shortfall in public support
at different rungs of the ladder are likely to vary enormously. It is one thing
for the public to withdraw its support for an incumbent at the bottom of the
hierarchy, an utterly common occurrence in democracies, where elections
and other mechanisms offer a number of opportunities for government turn-
over. In contrast, the implications of low public support for constitutional
ideals (at the top of the ladder)—i.e. the constitutional principles of the exist-
ing regime—are much more severe, since the remedy may require major
reforms in the constitutional order and, in this case, perhaps even entertain-
ing alternatives to a democratic government.
Our focus in this chapter is on two very different types of democratic sup-
port delineated in the figure that are assessed in the CSES: one located at the
performance end of the continuum, at level 5 of the ladder, and another
question located at the level of constitutional reality, or level 1.1 The satisfac-
tion with democracy question (level 5), clearly calls on respondents to assess
the overall performance of a democratic regime at a given point by asking,
‘On the whole, are you very satisfied, fairly satisfied, not very satisfied, or not
at all satisfied with the way democracy works in (country)?’ (e.g. Canache,
Mondak and Seligson 2001). By contrast, the other question (at level 1) cap-
tures the essence of Winston Churchill’s famous statement that people sup-
port democracy as the ‘lesser evil’ compared with any other form of
government, by asking respondents the extent to which they agree with the

184
 The Multiple Bases of Democratic Support

Constitutional
Ideal

1. Constitutional Ideals: Is some form of democracy

preferred over other forms of governance?

2. Constitutional Reality I: Is another Constitutional

Regime preferred over the existing constitution?

3. Constitutional Reality II: Is the existing

Constitution viewed positively?

4. Trust in Institutions: Do institutions work well in

the long run?

5. Satisfaction with Democracy: Does the current

system work well?

6. Evaluation of Governments: Does the current

Government perform well?

Democratic
Reality

Figure 10.1.  Conceptualizing democratic evaluations

statement, ‘Democracy may have problems buts it’s better than any other
form of government,’ on a four-point scale ranging from strongly agree to
strongly disagree.
These questions clearly produce different responses in a way that is theo-
retically meaningful (Figure 10.2). There is virtually a consensus in the 21
nations in our study that democracy is the best form of government. The
mean level of agreement is at or above the ‘agree’ threshold (point 3) in every
country save one (Bulgaria’s mean is 2.9) In percentage terms, more than 90
per cent of respondents in the pooled sample agree that democracies are the
best form of government. The country patterns generally mirror this result,
except for Bulgaria, where a surprisingly large proportion (32 per cent) indi-
cate that another form of government is preferable over a democracy. Other
countries where a significant minority remains sceptical about democracies
are Hungary and Italy (18 per cent), and Slovenia (15 per cent). With the
exception of Italy, these are newer post-communist democracies which sug-
gests that some residual scepticism over democratic ideals exists in these
nations. On the whole, though, these patterns indicate that an overwhelm-
ing proportion of most publics accept democratic ideals.

185
Mark Peffley and Robert Rohrschneider

Bulgaria 2001
Italy 2006
Poland 2001
Portugal 2005
Slovenia 2004
Hungary 2002
Romania 2004
Germany 2002
France 2002
Japan 2004
New Zealand 2002
Iceland 2003
Finland 2003
Great Britain 2005
Canada 2004
Spain 2004
Sweden 2002
Switzerland 2003
Ireland 2002
US 2004
Australia 2004
Denmark 2001

0 1 2 3 4

Democratic satisfaction (1= not at all, 4 = very satisfied)

Democratic ideal (1= str disagree, 4 = str agree)

Figure 10.2.  Mean democratic support indicators by country

These patterns clearly contrast with the more negative evaluations of the
current performance of democracies. Here the mean evaluations for most
countries range between being ‘not very satisfied’ (point 2) to ‘fairly satisfied’
(point 3). In percentage terms, two-thirds of the pooled sample is at least
fairly satisfied with the democratic process. At the extremes, citizens in Den-
mark are clearly the most satisfied (less than 7 per cent are dissatisfied),
whereas 79 per cent of Bulgarian citizens are dissatisfied. All in all, the two
indicators behave as expected in light of Figure 10.1: the performance of a
regime is evaluated much more critically than the ideal of a democracy.

10.3  Representation and Regime Evaluations

What explains this variation in support for regime performance and ideals?
Typically, students of regime evaluations use two basic perspectives to explain
democratic support (Table 10.1). One research tradition focuses on the policy
outcomes of the regime, finding that when the economic performance of a
regime is strong, as indicated by higher GDP per capita or lower inflation
rates, publics are more positive about the performance of a nation’s constitu-
tional reality than when economic conditions are weak (e.g. Clarke, Dutt and
Kornberg 1994; Canache, Mondak and Seligson 2001). In representation
terms, this mechanism reflects substantive representation: when a ‘democratic

186
 The Multiple Bases of Democratic Support

Table 10.1.  Conceptualizing the sources of representational judgements

Macro‐Level Micro‐Level

Substantive
Representation
Macro‐Economic Conditions Evaluations of Economic
Conditions

Quality of Arbitrating Institutions Evaluations of Procedural

Fairness
Procedural
Representation

government provides an orderly and peaceful process by means of which a

majority of citizens can induce the government to do what they most want it
to do and to avoid doing what they most want it not to do’ (Dahl 1989: 95).
The idea is that when the preferences of citizens are met, they are likely to
support a regime that delivers those goods they prefer. Empirically, substan-
tive representation can be measured at either the national level (GDP per
capita, inflation, and unemployment rates are the favourites) or at the indi-
vidual level (perceptions of personal or national economic conditions, either
past, present, or future).
But, as we noted above, representation is not limited to the notion of sub-
stantive representation. As we argued earlier, following Tyler (1988: 103) and
others (e.g. Hibbing and Theiss-Morse 2002), irrespective of the policies and
the outcomes generated by the government, citizens’ support for democracy
is also based on their assessments of procedural fairness, which in turn are based
on their evaluations of whether authorities are motivated to be fair, are hon-
est, and follow ethical principles of conduct, whether opportunities for repre-
sentation are provided, and whether authorities behave in a biased fashion.
How do citizens know that a regime is procedurally ‘fair’? We argue that the
way so-called arbitrating institutions—bureaucracies and judiciaries—func-
tion conveys important information about the way that a ‘political regime’
works. Consider that in most parliamentary democracies outside the USA,
citizens only have rare occasions to participate in elections, perhaps as few as

187
Mark Peffley and Robert Rohrschneider

five times over a four-year interval. They can cast a vote in local, regional,
national, and EU elections, and perhaps in a referendum (e.g. in Switzerland,
or over EU-related matters). But that is usually the extent of citizens’ direct
experience with the procedural mechanisms that constitutes the ‘input’ insti-
tutions of a regime. Contrast this with their exposure to the output institu-
tions of a regime. They no doubt have to contact a local, regional, or even
national bureaucracy or judiciary more frequently than they participate in an
election. Moreover, these contacts often involve highly personal, and there-
fore individually salient, matters (e.g. divorce, disputes with neighbours,
building permits), so that fair or unfair treatment constitutes an important
piece of information about the way that these output institutions function.
This personal experience is no doubt reinforced by impressions from friends,
neighbours, and family members. For example, living in a nation with a cor-
rupt bureaucracy and judiciary, as in Bulgaria and Italy, no doubt contributes
to dissatisfaction with the entire regime (Anderson and Tverdova 2003). In
contrast, residing in a country with well-functioning output institutions
clearly helps boost the regime’s overall evaluations, as in the Netherlands and
Denmark (Rohrschneider 2002, 2005). Our overall argument, then, is that the
procedural quality of output institutions provides an important basis on
which publics evaluate the political regime.
For this reason, Table 10.1 contains two cells for procedural representation,
one located at the individual level and another at the macro level. Prior
research mostly uses a micro-level approach to assess the fairness of institu-
tions (Mishler and Rose 2001; Seligson 2002). This research generates impor-
tant insights but a shortcoming of it is that judgements about a regime’s
fairness and satisfaction with it are likely endogenous—those who are satis-
fied with a regime are also those that believe a regime delivers goods even-
handedly, and vice versa. For this reason, some studies use a macro-level
measure for the fairness of national output institutions, such as the corrup-
tion measures from Transparency International (Anderson and Tverdova
2003) or the World Bank measures of the fairness of bureaucracies and judici-
aries (Rohrschneider and Loveless 2010). To the extent that these indicators
rely on judgements by experts, and aggregate this information by country,
they avoid the endogenous relationship between indicators measured in one
survey at the individual level. To avoid the bias associated with one method
or another, we will rely on both macro and micro assessments of institutional
quality in ways discussed below.
In sum, the perspective of substantive representation predicts that regime
evaluations will improve when economic conditions and economic percep-
tions become more favourable, whereas the procedural perspective hypothe-
sizes that the procedural quality of output institutions influences citizens’
evaluations of a regime’s performance and ideals. Formally, we focus the cen-

188
 The Multiple Bases of Democratic Support

tral hypothesis on the procedural dimension because the Sanders et al. chap-
ter examines economic performance factors in detail:
H1: Procedurally fair output institutions increase public evaluations of a regime.

10.4  Consensus Institutions and Regime Evaluations

Another important mechanism relating institutions to procedural evalua-

tions is rooted in the degree to which political institutions are consensual or
majoritarian. The central argument harks back to the work of Lijphart (1998)
and Powell (2000), who have sparked numerous investigations into the way
that a nation’s regime type influences citizens’ perceptions of the democratic
process, such as their views about institutions’ performance (Anderson and
Guillory 1997), political representation (Miller et al. 1999; Rohrschneider
2005), and the ideological placements of political parties (Powell 2000).
Lijphart has forcefully argued that proportional regimes are ‘kinder, gentler’
institutions when compared to majoritarian regimes (1999: 275). Powell
broadly echoes this conclusion when he concludes in his detailed investiga-
tion of ideological congruence in proportional and representational regimes
that the former are much better in representing citizens. Thus, it would
appear that consensus democracies, on the whole, produce a greater amount
of satisfaction among publics than majoritarian institutions.
More recently, however, the relevance of consensus institutions has been
undermined by studies suggesting their influence on ideological agreement
between parties and citizens is vanishing, including by one author who ini-
tially established this relationship (Powell 2009). Others have shown that
institutions do not appear to play the central role in affecting citizens’ evalu-
ations of the democratic process as was initially argued (Elkins and Sides
2007; Ezrow 2010). For these reasons, we hypothesize that proportional
regimes bring about greater satisfaction with a democratic regime’s perfor-
mance and enhance support for democratic ideals but we remain agnostic as
to the precise strength of the effect:
H2: Publics in proportional regimes are more positive about the regimes’ demo-
cratic performance and ideals than citizens in majoritarian systems.

10.5  Data and Measurement

We use the CSES, second wave, to test these hypotheses. We included all
countries of the EU, and democracies from other regions of the world
(Table A.10.1). We excluded nations that are not fully democratic because we

189
Mark Peffley and Robert Rohrschneider

wanted to focus on developed democracies in order to test whether output

institutions influence the way that citizens evaluate regimes within a demo-
cratic framework. Missing data further reduced available countries where
some questions were not asked at all.
In order to measure institutional quality, we use the World Bank indicators
that assess the procedural integrity of bureaucracies, judiciaries, and the
extent of corruption. These measures are based on a number of sources—pub-
lic opinion data, surveys of those who conduct business in nations, and rat-
ings by country experts (Kaufmann et al. 2006). This set of measures pools all
available evidence about the integrity of output institutions, and thus pro-
vides important information about how ordinary citizens likely experience
the procedural quality of a regime. These measures have the clear advantage
that they are removed a considerable distance from individual-level evalua-
tions of democratic performance and ideals so that we lower the degree of
endogeneity between these evaluations. Clearly, countries that are more
affluent (as measured by the United Nation’s Human Dimension Index) are
also more likely to have procedurally fair institutions. At the same time, the
relationship is far from perfect (r = 0.82). Several affluent nations vary quite a
bit in terms of their institutional quality. Thus, both procedural quality and
the capacity of regimes to produce national affluence need to be considered
in the multivariate analyses reported below.
As institutional measures, we include Lijphart’s federalism scores as well as
the parties–executive dimension. Together, they tap important institutional
features such as the concentration of power, the number of veto players, and
the extent to which institutional arrangements lead to the group representa-
tion of social interests. Our argument is that the quality of output institutions
continues to affect the way that citizens evaluate the performance and ideals
of national democracies when we control for these other sources of opinion
formation about these objects. Finally, we included the length of time a coun-
try has been a democracy, which may influence how citizens evaluate demo-
cratic regimes and their procedural properties. Long-standing democracies
likely ingrain publics with a sense of what is right and wrong, and what to
expect from a regime. Publics in newer democracies may not have sorted this
out yet and it maybe the length of regime experience, more than other macro
factors, that influences citizens’ evaluations about regime procedures.
Our analysis will proceed as follows. We will first present the influence of
macro variables on public evaluations of democratic performance and ideals.
In a second step, we will add several individual-level control variables. We
started out with a larger set than presented here but, in the interests of keep-
ing the N number of cases as large as possible—especially at the country
level—this chapter presents the results for age, education, political interests,
and governing performance.2 The point of these analyses is to test whether

190
 The Multiple Bases of Democratic Support

the institutional quality variable continues to predict democratic attitudes

when these control variables are included; and second, to prepare the next
step. For a final set of analyses we will then add three variables that measure
procedural perceptions at the individual level: evaluations of corruption,
respect for human rights, and evaluations of representation. Here the goal is
to test whether the macro indicator of institutional quality exerts its influ-
ence through individual-level perceptions of democratic procedures, as our
argument assumes.3

10.6 Results

Table 10.2 presents the results for the macro-level variables. The results une-
quivocally show that the institutional quality variable is the only (!) signifi-
cant predictor of both democratic performance and ideals. Clearly, when
nations have well-functioning output institutions, citizens are much more
likely to give positive performance evaluations.4 Somewhat surprising is the
even stronger impact of institutional quality on democratic ideals—­
individuals are more intense in their preference for a democracy over its alter-
natives when output institutions work well, independently of national
affluence. Given the range of the democratic ideals variable for most c­ ountries
(Figure 10.2), it must be that most of the explained variation is across agree
versus strongly agree response categories rather than the agree versus disagree

Table 10.2. The influence of macro-level variables on public evaluations

of democratic performance and ideals

VARIABLES Democratic Performance Democratic Ideals

Macro variables:
Institutional quality 0.081* 0.067**
(0.040) (0.024)
HDI 1.37 –0.11
(1.77) (1.03)
Parties–executive dimension 0.015 0.036
(0.047) (0.027)
Federalism 0.015 0.030
(0.044) (0.026)
Length of democracy 0.00021 0.00087
(0.0013) (0.00077)
Constant 1.01 3.06**
(1.52) (0.89)
Observations 29,686 28,809
Number of groups 21 21

Note: Entries are unstandardized coefficients estimated using xtmixed in Stata 11.
+, *, ** indicates significance at the p = 0.10, 05, and 01 levels, respectively.

191
Mark Peffley and Robert Rohrschneider

categories. Thus, an important initial finding is that the character of output

institutions not only affects democratic performance evaluations but publics’
support for democratic ideals. It is remarkable that these patterns emerge
despite the fact that we control for national affluence and institutional
arrangements which are usually thought of as the main influences on these
individual-level judgements.
In a second step, we include several individual-level control variables
(Table 10.3), namely respondents’ age, education, and level of political inter-
est. However, theoretically, the most important control variable is the per-
ceived performance of a government,5 which, as hypothesized, should be
connected to citizens’ support for democracy. Expectedly, perceived govern-
ment performance is strongly related to both measures of democratic
­support—i.e. democratic satisfaction and democratic ideals. This clearly sup-
ports the substantive perspective on representation which argues that sup-
port for democracy depends on the government’s capacity to deliver goods
that citizens prefer. It is also clear that the impact of institutional quality is

Table 10.3.  Predicting democratic performance and democratic ideals, including individual-­
level controls

VARIABLES Democratic performance Democratic ideals

Macro variables:
Institutional quality 0.059+ 0.055*
(0.034) (0.024)
HDI 0.64 –0.29
(1.49) (1.06)
Parties–executive dimension 0.016 0.047
(0.039) (0.028)
Federalism 0.049 0.026
(0.037) (0.026)
Length of democracy –0.00046 0.00026
(0.0011) (0.00078)
Individual-level variables:
Political interest 0.030** 0.092**
(0.011) (0.011)
Government performance 0.22** 0.069**
(0.0032) (0.0033)
Female –0.028** –0.042**
(0.0086) (0.0086)
Age 0.000079 0.0034**
(0.00026) (0.00027)
Education 0.017** 0.070**
(0.0027) (0.0027)
Constant 0.71 2.49**
(1.28) (0.90)
Observations 23,793 22,992
Number of groups 21 21

Note: +, *, ** indicates significance at the p = 0.10, 05, and 01 levels, respectively.

192
 The Multiple Bases of Democratic Support

diminished in Table 10.3 after including individual-level predictors. Govern-

mental performance evaluations undoubtedly capture some of the influence
of the procedural integrity of output institutions. But while the influence of
the institutional quality variable on democratic support is reduced some-
what, we note that it remains statistically significant despite the inclusion of
stringent controls.6
In a third step, we push our analyses further to determine whether the
quality of national institutions influences citizens’ support for democracy via
individual-level assessments of the procedural fairness of the government. If

Denmark 2001
Canada 2004
Human Rights are Respected (1 = No 4 = Yes)

Poland 2003
3.2
Australia 2004
Germany 2002
US 2004 Finland 2003
3 Great Britain 2005
Japan 2004 Sweden 2002
Hungary 2002 Ireland 2002 New Zealand 200
2.8 Iceland 2003
Spain 2004

France 2002
2.6
Italy 2006
Slovenia 2004
2.4 Poland 2001

Romania 2004
2.2
Bulgaria 2001

0 2 4 6 8
Institutional Quality

3.5 Romania 2004 Poland 2001

Bulgaria 2001
Italy 2006
Perceived Corruption (1 Low 5 High)

Germany 2002
Slovenia 2004 Japan 2004
France 2002
Hungary 2002
3
Ireland 2002

Canada 2004
US 2004
Spain 2004
2.5
Switzerland 2003
Great Britain 2005
Australia 2004
Finland 2003
Iceland 2003
New Zealand 200
Sweden 2002
2 Denmark 2001

0 2 4 6 8
Institutional Quality (Low to High)

Figure 10.3 a–c.  Institutional quality and procedural evaluations of national systems
(mean scores on y-axis)

193
Mark Peffley and Robert Rohrschneider

1
Party Represents Citizens (0 No 1 Yes)
Switzerland 2003
Australia
Denmark2004
2001
.8 New Zealand 200
Ireland 2002
Sweden 2002
Spain 2004 Great Britain 2005
Hungary 2002 US 2004 Canada 2004
Finland 2003
.6 Iceland 2003
Japan 2004 Germany 2002
France 2002

Bulgaria 2001
Romania 2004
Italy 2006
.4
Poland 2001

Slovenia 2004

.2

0 2 4 6 8
Institutional Quality (Low to High)

Figure 10.3 a–c.  (Continued)

this is true, as our argument stipulates, then the inclusion of measures that
clearly tap individual-level perceptions of the procedural quality of a political
regime should reduce the impact of the macro-level quality indicator. As
revealed in the graphs of Figure 10.3, there is a fairly strong relationship at
the aggregate level between institutional quality, on the one hand, and the
average country scores for measures of the degree to which the government
is perceived to engage in corruption and respect human rights and how well
the parties represent one’s views (see the Appendix A.10.1 for the survey
measures). The first graph plots the quality of national institutions on the
x-axis and the mean country scores on government corruption perceptions
on the y-axis. Including this variable in the multilevel model should there-
fore absorb some of the impact of the institutional quality indicator on dem-
ocratic evaluations. Similarly, when publics believe that human rights are
respected, then this too should capture part of the procedural quality of out-
put institutions. Therefore, the inclusion of individuals’ evaluations of the
state of human rights should also lower the impact of the institutional qual-
ity variable. Finally, we also included citizens’ perceptions of whether parties
represent them in the political process. This captures clearly both procedural
and substantive aspects of representation and this indicator should also con-
tribute to a reduction of the relevance of the institutional quality indicator.
Consistent with this expectation, our next model (Table 10.4) shows that
the institutional quality indicator is now insignificant. This indicates that
its influence is mediated by individual-level evaluations of regime

194
 The Multiple Bases of Democratic Support

Table 10.4.  Predicting democratic performance and democratic ideals, including individual-­
level controls and individual-level evaluations of procedural fairness

VARIABLES Democratic performance Democratic ideals

Macro variables:
Institutional quality 0.0082 0.028
(0.030) (0.025)
HDI 0.81 –0.16
(1.30) (1.10)
Parties–executive dimension 0.0069 0.45
(0.034) (0.029)
Federalism 0.039 0.014
(0.032) (0.027)
length of democracy –0.00030 0.00037
(0.00096) (0.00081)
Individual-level variables:
Political interest –0.0082 0.059**
(0.011) (0.011)
Government performance 0.17** 0.041**
(0.0034) (0.0035)
Female 0.0031 –0.027**
(0.0087) (0.0089)
Age –0.00087** 0.0028**
(0.00027) (0.00028)
Education 0.00070 0.060**
(0.0027) (0.0028)
Party representation 0.14** 0.13**
(0.010) (0.010)
Respects freedoms and rights 0.23** 0.15**
(0.0063) (0.0065)
Perceived corruption –0.12** –0.037**
(0.0061) (0.0063)
Constant 0.72 2.28*
(1.11) (0.94)
Observations 21,183 20,697
Number of groups 21 21

procedures in theoretically meaningful ways: evaluations of human rights,

perceptions of corruption, and feelings of representation capture at the indi-
vidual level what the World Bank governance scores express at the country
level. This is an instance where the insignificance of a key ­predictor—insti-
tutional q­ uality—is theoretically predicted and supports a substantively
important argument.
But is it true, one might reasonably ask, that the institutional quality indi-
cator drives popular evaluations of human rights, corruption, and represen-
tation? One might wonder, reasonably so, whether popular evaluations of
regime procedures are due to the ‘kinder, gentler’ consensus institutions that
Lijphart envisioned. Moreover, it may well be that it is not institutional
quality but socio-economic affluence that drives mass evaluations of demo-
cratic procedures about human rights, corruption, and representation. In

195
Mark Peffley and Robert Rohrschneider

order to address this argument, we conducted several analyses with the three
­procedural evaluations as the dependent variable. That is, we estimated
three multilevel models which include the predictors shown in Table 10.4
(minus government performance),7 with the dependent variable being pop-
ular evaluations of human rights, corruption, and representation. The goal
is to examine whether institutional quality is indeed the main reason for
these evaluations, or whether its influence is superseded by the two Lijphart
indicators or nations’ socio-economic affluence (Table 10.5).
The empirical patterns are surprisingly clear: a nation’s institutional quality
significantly shapes mass evaluations of regime procedures. Higher quality
increases beliefs that human rights are respected in a country, perceptions of
corruption are lower, and the representation process receives more positive
evaluations than in countries that have lower governance scores. Evidently, a
well-functioning set of governance institutions goes a long way toward bring-
ing about a belief that national democratic procedures of a nation are in
order. What is remarkable about these results is that they emerge even though
we control for a country’s socio-economic affluence, which never reaches
statistical significance. In addition, these results also control for the two

Table 10.5.  Predicting procedural evaluations

VARIABLES Human rights Corruption Parties represent citizens

Macro variables:
Institutional quality 0.11** –0.19** 0.26**
(0.027) (0.046) (0.092)
HDI 0.49 1.40 –3.33
(1.19) (2.03) (4.06)
Federalism 0.084** 0.13** –0.013
(0.029) (0.050) (0.10)
Parties–executive dimension –0.00024 –0.082 0.022
(0.031) (0.053) (0.11)
Length of democracy –0.00056 –0.0015 0.00021
(0.00088) (0.0015) 0.26**
Individual-level variables:
Political interest 0.025* –0.024* 0.96**
(0.011) (0.012) (0.040)
Female –0.058** 0.11** –0.10**
(0.0083) (0.0087) (0.027)
Age 0.00093** –0.0024** 0.014**
(0.00025) (0.00026) (0.00083)
Education 0.050** –0.035** 0.078**
(0.0026) (0.0027) (0.0085)
Constant 1.66 2.64 1.42
(1.02) (1.74) (3.48)
Observations 29,754 28,570 28,081
Number of groups 21 21 21

Note: Models 1 and 2 were estimated using xtmixed in Stata 11.0; Model 3 is estimated using xtlogit.

196
 The Multiple Bases of Democratic Support

Lijphart dimensions which are mostly insignificant, once again suggesting

that the difference between majoritarian and consensus institutions is not
particularly important for popular perceptions of a regime. There are, how-
ever, two exceptions: citizens residing in federal regimes are more likely to
think that human rights are respected in their country; and they also tend to
believe that corruption is lower. Of the six coefficients,8 however, these two
are the exception; the other four are insignificant. All in all, this is fairly weak
evidence in support of the idea that consensus systems have more procedural
integrity than majoritarian institutions.

10.7 Conclusions

All in all, these analyses clearly show that the procedural quality of national
institutions influences the way that citizens evaluate the performance of a
regime. More importantly, however, is the result that the quality of output
institutions also affects the way that individuals judge democratic ideals. It is
perhaps not surprising that individuals residing in nations where corruption
is rampant among bureaucrats and the judiciary (to name but one example)
also conclude that other parts of a regime—parties, politicians, and
­governments—are not performing well. For these reasons, it is regrettable
that few studies link the procedural integrity of output institutions to public
evaluations of democratic institutions. This is all the more regrettable as the
quality of these output institutions also affects the way that democratic ideals
are viewed. Perhaps the most important, and distressing, finding is that poor-
ly functioning output institutions are associated with a strong tendency for
citizens to question whether a democracy is preferable to other forms of
­government—perhaps the most enduring and consequential evaluation of
democracy in Figure 10.1. This fact alone should spawn further studies of the
connection between the procedural quality of output institutions and public
evaluations of democratic ideals.

Appendix A.10.1: Measurement

Democratic satisfaction: ‘On the whole, are you very satisfied, fairly satisfied,
not very satisfied, or not at all satisfied with the way democracy works in
(country)?’ Response categories range from not satisfied at all (1) to very
­satisfied (4).

Democratic ideals: ‘Democracy may have problems buts it’s better than any
other form of government’, on a four-point scale ranging from strongly agree

197
Mark Peffley and Robert Rohrschneider

to strongly disagree. Response categories are from strongly agree (1) to strong-
ly disagree (4). We reversed the polarity.

Evaluations of human rights: ‘How much respect is there for individual free-
dom and human rights nowadays in (country)? Do you feel there is a lot of
respect for individual freedom, some respect, not much respect, or no respect
at all?’

Evaluations of corruption: ‘How widespread do you think corruption such as

bribe taking is amongst politicians in (country): very widespread, quite wide-
spread, not very widespread, it hardly happens at all?’

Representation through parties: ‘Would you say that any of the parties in (coun-
try) represents your views reasonably well?’

Government performance: We combined the responses to two indicators (alpha =

0.80): 1) Performance on most important problem: ‘And thinking about that
issue, how good or bad a job do you think the government/president in (capi-
tal) has done over the past (number of years between the previous and the
present election OR change in govt.) years. Has it/he/she done a very good
job? A good job? A bad job? A very bad job?’; 2) Performance in general: ‘Now
thinking about the performance of the government in (capital)/president in

Table A.10.1.  Countries included in analyses

Nation Freq. Percentage

Australia (2004) 1,769 4.29

Bulgaria (2001) 1,482 3.59
Canada (2004) 1,674 4.06
Denmark (2001) 2,026 4.91
Finland (2003) 1,196 2.90
France (2002) 1,000 2.42
Germany (2002) 2,000 4.85
Hungary (2002) 1,200 2.91
Iceland (2003) 1,446 3.51
Ireland (2002) 2,367 5.74
Italy (2006) 1,439 3.49
Japan (2004) 1,977 4.79
New Zealand (2002) 1,741 4.22
Poland (2001) 1,794 4.35
Portugal (2005) 2,801 6.79
Romania (2004) 1,913 4.64
Slovenia (2004) 1,002 2.43
Spain (2004) 1,212 2.94
Sweden (2002) 1,060 2.57
Switzerland (2003) 1,418 3.44
Great Britain (2005) 860 2.09
USA (2004) 1,066 2.58
Total 41,242 100.00

198
 The Multiple Bases of Democratic Support

Table A.10.2.  Descriptive variable information about predictors (TBD REDO)

Variables Obs. Mean Std. dev. Min max

Macro variables:
Institutional quality 41242 4.21539 2.226138 –.56 6.72
Human Dimension Index 41242 .92551 0.0451706 .802 .963
Federalism (Lijphart) 38441 .0710335 1.118332 –1.611294
2.451343
Executive–parties dimension (Lijphart) 38441 .066794 1.038367 –2.372764
1.690315
Individual-level variables:
Age 40948 47.81169 17.16458 17 102
Education 40844 5.126604 1.723941 1 8
Political interest 40825 .216632 0.4119547 0 1
Government performance 31187 4.610415 1.451503 2 8
Human rights respected 38197 2.825719 0.7879873 1 4
Corruption 38157 2.808056 0.9041523 1 4
Representation 30252 1.180881 0.7748019 0 2

general, how good or bad a job do you think the government/president in

(capital) has done over the past (number of years between the previous and
the present election OR change in govt.) years. Has it/he/she done a very good
job? A good job? A bad job? A very bad job?’

Age: Coded in years.

Education: An eight-point indicator ranging from low to high.

Political interest: ‘Here is a list of things some people do during elections.

Which if any did you do during the most recent election? Q1a. . . .talked to
other people to persuade them to vote for a particular party or candidate?’
No = 0, Yes = 1.

Gender: Male = 0, Female = 1.

Notes

1. See Rohrschneider and Schmitt-Beck (2002) for a detailed discussion of each of the
six levels of democratic evaluations.
2. A more extensive set of control variables, including class and left–right perceptions,
does not substantially change the results but has the undesirable consequence of
reducing the number of cases at both the individual level and, especially regretta-
ble, the countries that we can include in our analyses.
3. We also imputed missing data for our multilevel data using Stata 11’s Multiple
Imputation method for clustered data. The results for the multiply imputed data,
however, are virtually identical to those presented here, but are available from the
authors upon request.

199
Mark Peffley and Robert Rohrschneider

4. Both variables are statistically significant when democratic performance evalua-

tions are regressed on them individually.
5. As described more fully in the measurement Appendix A.10.1 government perfor-
mance evaluations is an additive index based on individuals’ assessments of how
well the government is doing managing the most important problem and how well
the government is doing in general.
6. When regressing government performance evaluations on institutional quality
and HDI, the coefficient associated with the former variable is statistically signifi-
cant whereas the latter is not.
7. Government performance is as much a consequence as it is a source of procedural
evaluations. We therefore excluded it from these analyses. Note, however, that its
inclusion does not affect the results.
8. There are three dependent variables in this analysis and two indicators for Lijphart’s
typology.

200
11

Globalization, Representation,
and Attitudes towards Democracy
Kees Aarts, Jacques Thomassen, and Carolien van Ham

11.1 Introduction

One of the main challenges for modern democracies is their supposedly

declining support among the citizenry, often referred to as the ‘crisis of legiti-
macy of representative democracy’. In recent years, in the public as well as
the academic debate, citizens’ growing disillusionment with politics has been
a topic of increased concern. Signalling widespread democratic malaise, erod-
ing political support, and ever declining trust in political leaders and politics,
the question ‘Why we don’t trust government?’ or even ‘Why we hate poli-
tics’ has been frequently researched (Nye et al. 1997; Norris 1999a, 2011;
Pharr and Putnam 2000; Dalton 2004; Hay 2007; Zmerli and Hooghe 2011).
However, how convincing the evidence for declining support is, is still a mat-
ter of dispute (Klingemann and Fuchs 1995; Norris 1999a; Norris 2011). In
addition, the causal mechanisms behind the supposed negative trend in peo-
ple’s support for democracy and its institutions are the subject of an ongoing
debate as well (Kaase and Newton 1995; Nye et al. 1997; Dalton 2004; Kriesi
et al. 2008; Norris 2011).
Some observers have pointed to a growing gap between citizens’ demands
of the political system and the system’s ability to deliver, resulting in ‘system
overload’ or ‘legitimacy gaps’ (King 1975; Norris 2011). Others have argued
that the changing nature of political parties and the media, acting as interme-
diaries between citizens and the political system, has negatively affected citi-
zens’ political support (Newton 1999, 2006; Mair 2009). Hence, explanations
of political support can be found at the micro level of individual citizens and
their aspirations for democracy (demand side); at the macro level of political
systems and their performance in terms of input and output (supply side);

201
Kees Aarts, Jacques Thomassen, and Carolien van Ham

and at the meso level of political parties and media that mediate between citi-
zens and system, transmitting citizens’ preferences as well as shaping them.
At the micro level of individual citizens, the two most important explana-
tions for declining political support are the modernization thesis on the one
hand and the globalization thesis on the other hand. The modernization the-
sis was first introduced by Inglehart (1971, 1977), and subsequently elaborated
and empirically tested by Dalton (2004) and Norris (2011). In his 2004 study
Dalton rephrases this thesis as the positive effects thesis, and contrasts it to the
negative effects thesis. Both hypotheses refer to the effects of the socio-economic
(and concomitant cultural) transformation of advanced industrial societies.
According to the positive effects thesis the second half of the twentieth cen-
tury was a period of rising affluence, expanding education, and improving
social opportunities for most citizens. This social modernization gradually
transformed the relationship between citizens and the state. Greater political
skills and resources—that is, higher levels of cognitive mobilization—led the
contemporary electorate towards elite-challenging forms of political action,
which places them in conflict with politicians and government officials.
These better-educated citizens are inclined to question democratic politics as
it is currently practised not because they do not endorse strong democratic
ideals but because they demand a greater role in the political process. There-
fore, the positive effects thesis claims:

The greatest loss in support should be located among those who are at the upper
end of the economic order: the better educated, the more skilled, and those with
higher incomes. (Dalton 2004: 86)

The negative effects thesis argues, however, that the development of advanced
industrial societies created new social and economic problems that have
eroded the public support of the citizenry. Social groups on the periphery of
the economy, such as the less educated and the unemployed, are becoming
increasingly marginalized by the labour structure of advanced industrial soci-
eties. Moreover, it is these marginalized citizens that are most strongly affect-
ed by processes of globalization, as we will further elaborate below (Dalton
2004; Kriesi et al. 2008; Bovens and Wille 2009). Hence, the negative effects
hypothesis posits that:

The greatest loss in support should be located among those who are at the margins
of the economic order: the less educated, the less skilled, and those with lower
incomes. In contrast, upper-status groups also might have become more sceptical
about government, but not to the same degree as lower-status groups. (Dalton
2004: 85)

Dalton’s own research and Norris’s (2011) more recent findings leave little
doubt about the empirical validity of these two alternative hypotheses. Both

202
 Globalization, Representation, and Attitudes towards Democracy

authors find evidence for the ‘positive effects thesis’ that locates declining
legitimacy among the wealthier, higher educated ‘critical citizens’. Dalton
(2004) moreover shows that over time the relationship between education
and political trust has gradually changed. In most of the countries repre-
sented in Dalton’s study this relationship has weakened whereas in some
countries (the USA and Germany) it has even reversed. In these countries the
higher educated citizens are now less trustful than lower-educated citizens.

11.2  Globalization and a New Structural Conflict?

However, we think that recent developments in many modern democracies

justify a further discussion of the validity of these alternative hypotheses. In
particular in Western Europe the libertarian revolution of well-educated
youngsters in the second half of the twentieth century now seems to be suc-
ceeded by a populist revolution that is anything but libertarian and is dispro-
portionally supported by the less well-to-do and less educated part of the
population. Across the Western world populist parties from the left and in
particular from the right successfully appeal to the frustrations of people who
feel threatened by the consequences of globalization, by the open economy
leading to—at least in their perception—the loss of jobs, and the immigration
of people with different economic and cultural backgrounds, often from the
Muslim world. This differential effect of globalization seems to be more in
line with the negative effects hypothesis, and it has recently been addressed
by Kriesi and his collaborators. They argue that at least in Western Europe a
new structural conflict is developing, the conflict between ‘winners’ and ‘los-
ers’ of globalization, as the consequences of globalization are not the same for
all members of a national community:

The ‘losers’ of globalization are people whose life chances were traditionally pro-
tected by national boundaries. They perceive the weakening of these boundaries
as a threat to their social status and their social security. (Kriesi et al. 2008: 5)

The likely winners of globalization, in contrast, are those ‘who benefit from
the new opportunities resulting from globalization, and whose life chances
are enhanced’ (Kriesi et al. 2008: 5). In terms of social groups, ‘winners’
include entrepreneurs and qualified employees in sectors open to interna-
tional competition, as well as cosmopolitan citizens, while ‘losers’ include
entrepreneurs and qualified employees in traditionally protected sectors, all
unqualified employees, and citizens who strongly identify themselves with
their national community (Kriesi et al. 2008: 8).
According to Kriesi et al., the antagonism between winners and losers of
globalization is leading to a new structural conflict between integration and

203
Kees Aarts, Jacques Thomassen, and Carolien van Ham

demarcation. They distinguish between an economic and a cultural dimen-

sion of the integration–demarcation divide. In the economic domain, a neo-
liberal free trade position is opposed to a position in favour of protecting the
national markets. In the cultural domain, a universalist, multiculturalist or
cosmopolitan position is opposing a position in favour of protecting national
culture and citizenship. The authors expect that this new structural conflict
will be embedded in the two main dimensions of conflict originally distin-
guished by Rokkan: a cultural (religion) and a social-economic one (class).
The substantive meaning of these dimensions had already been changed by
the cultural revolution of the 1960s. The cultural dimension was transformed
from a dimension mainly defined in terms of religious concerns to one oppos-
ing culturally liberal or libertarian concerns, on the one side, and the defence
of traditional (authoritarian) values and institutions on the other. Similarly,
Kriesi et al. hypothesize that the new demarcation/integration conflict will be
embedded into this transformed two-dimensional basic structure, transform-
ing it once again.
On the social-economic dimension, the new conflict can be expected to
reinforce the classic opposition between a pro-state and a pro-market posi-
tion while giving it a new meaning. The pro-state position is likely to become
more defensive and more protectionist, while the pro-market position is like-
ly to become more assertive in favour of the enhancement of national com-
petiveness on world markets. On the cultural dimension, they expect
enhanced opposition to the cultural liberalism introduced by the cultural
revolution of the second part of the twentieth century as a result of the eth-
nicization of politics. In other words, the defence of tradition is increasingly
expected to take on an ethnic or nationalist character. Furthermore, new
issues should be integrated into the cultural dimension. In the Western Euro-
pean context, central among these are the issues related to European integra-
tion, which correspond to the new political and cultural forms of competition
linked with globalization. The demarcation pole of the new cultural cleavage
should be characterized by opposition to the process of European integration
and by restrictive positions with regard to immigration. This two-­dimensional
space can be represented as in Table 11.1.

Table 11.1.  Dimensions of political conflict

Social-economic dimension

Cultural dimension Demarcation Integration

Demarcation A B
Integration C D

204
 Globalization, Representation, and Attitudes towards Democracy

11.3  Globalization and Political Representation

Now, how might globalization affect citizens’ attitudes towards democracy?

The causal mechanism connecting globalization processes and citizens’ atti-
tudes is not that straightforward, and requires further specification. We pro-
pose that the connection might run via changes in citizens’ issue positions on
the new conflict dimensions discussed above. We expect that those who are at
the margins of the economic order: the less educated, the less skilled, and those with
lower incomes will mainly be located in quadrant A of Table 11.1. Being less
skilled, they will tend to be on the demarcation side of the social-economic
dimension; being less educated they will tend to be on the demarcation side
of the cultural dimension as well. But why then should being on this position
erode political support, as the negative effect thesis claims? A possible expla-
nation was presented more than 50 years ago by Lipset. He pointed to the
location of the working class in this two-dimensional space and the problem
of political representation people in this category encounter:

The poorer strata everywhere are more liberal or leftist on economic issues; they
favour more welfare state measures, higher wages, graduated income taxes, sup-
port of tradeunions, and so forth. But when liberalism is defined in noneconomic
terms—as support of civil liberties, internationalism, etc.—the correlation is
reversed. The more well-to-do are more liberal, the poorer are more intolerant.
(Lipset 1966: 101–2)

And:

In some nations working-class groups have proved to be the most nationalistic

sector of the population. In some they have been in the forefront of the struggle
against equal rights for minority groups, and have sought to limit immigration or
to impose racial standards in countries with open immigration. (Lipset 1966: 99)

Since the political elites tend to belong to the well-to-do and in particular to
the better educated, the logical consequence is a problem of political repre-
sentation among the parties on the left where political elites combine a leftist
attitude on social-economic issues with a libertarian attitude on immaterial
issues whereas their voters—assuming they vote according to social-­economic
issues—combine a leftist attitude on social-economic issues with an authori-
tarian or conservative stand on immaterial issues. As a consequence political
elites will hardly be representative of their rank and file on immaterial issues:

The gradual realization that extremist and intolerant movements in modern soci-
ety are more likely to be based on the lower classes than on the middle and upper
classes has posed a tragic dilemma for those intellectuals of the democratic left
who once believed the proletariat necessarily to be a force for liberty, racial equal-
ity, and social progress. (Lipset 1966: 97)

205
Kees Aarts, Jacques Thomassen, and Carolien van Ham

The cultural dimension introduced by Kriesi et al. is based on the same liber-
tarian–authoritarian dimension Lipset referred to, and his argument still
applies.1 If this is correct, most working class voters will be in quadrant A
where—at least in many countries, and at least until relatively recently—no
political parties are, whereas the political parties and political elites on the
left are in quadrant C.2 Moreover, people in quadrant A might presently have
even better reasons to feel poorly represented than at the time when Lipset
made his observations, for two reasons. First, in the 1990s most of the tradi-
tional social-democratic parties have moved towards a more liberal position
on the social-economic dimension, leaving their traditional electorate
behind. Second, and more importantly, over the last decade the cultural
dimension has become more salient (Pellikaan, De Lange and Van der Meer
2007; Aarts and Thomassen 2008b). Therefore, among voters on the demarca-
tion side of the cultural dimension, feelings of not being well represented can
only be expected to increase.3 As a consequence they will have little reason to
be satisfied with the functioning of democracy. Of course, it is precisely for
this reason that populist parties from the left (positioning themselves in
quadrant A) and in particular from the right (positioning themselves in quad-
rant B) are extremely successful in most Western European countries and in
particular in consensus democracies with their proportional electoral system
where they can easily get access to parliament. This is in line with the general
expectation that consensus democracies mainly because of their proportional
electoral system and their multiparty system are more responsive to the
(changing) needs of the citizens. Thus one might argue that in those coun-
tries voters no longer have a reason to feel unrepresented as there are now
parties available to represent them on both dimensions, or at least on the
increasingly important cultural dimension.
Therefore, it is our expectation that in line with the negative effects hypoth-
esis those who are at the margins of the economic order: the less educated, the less
skilled, and those with lower incomes feel gradually less well represented by any
of the political parties and therefore are increasingly dissatisfied with the
functioning of democracy. We also expect that this effect will be smaller in
consensus democracies than in majoritarian democracies.4 The well-educated
group of the population that was responsible for the cultural revolution of
the second part of the twentieth century is in a totally different position.
Inglehart’s (1977) argument that one of the reasons why well-educated
youngsters rebelled against the political establishment of those days was that
the established parties did not represent them on post-material issues like the
protection of the environment, women’s rights, etc. hardly applies anymore.
They often found close allies in the established parties of the left and, in due
course, spawned a new set of parties—the New Left and Green parties. Mainly
located at the borderline between quadrants C and D, these well-educated,

206
 Globalization, Representation, and Attitudes towards Democracy

libertarian, and cosmopolitan citizens are among the winners of globaliza-

tion and are well represented by the libertarian political parties of the left. If
they are dissatisfied with the functioning of democracy it is for different rea-
sons, possibly because they are the dissatisfied democrats Dalton and others
have been referring to (Klingemann 1999; Dalton 2004). As a consequence we
also expect a growing gap between the two groups in terms of how well they
feel themselves to be represented and their satisfaction with democracy.
In the framework of this chapter, we approach these propositions in three
steps. First of all, we test whether there is a difference in attitudes towards
democracy between citizens ‘at the margins’ and citizens ‘at the upper end’ of
the economic order, using data from the Comparative Study of Electoral Sys-
tems. Second, since the modernization and globalization hypotheses refer to
changes in attitudes towards democracy, we test whether the two groups have
grown apart over time using data from the Eurobarometer. Third and finally,
we evaluate to what extent the causal mechanism outlined above connecting
globalization to attitudes towards democracy might hold, using the case of
the Netherlands as empirical illustration.

11.4 Hypotheses

As outlined above, we expect to find empirical support for the negative effects
hypothesis rather than the positive effects hypothesis and more so in majori-
tarian than in consensus democracies. Hence, our first two hypotheses are:

H1: Those who are at the margins of the economic order—the less educated, the
less skilled, and those with lower incomes—will have more negative attitudes
towards democracy than those that are at the upper end of the economic order:
the better educated, the more skilled, and those with higher incomes.
H2: This difference will be larger in majoritarian than in consensus democracies.

Note that while hypothesis 1 is static, specifying differences between social

groups, both the positive and negative effects hypotheses seek to explain
long-term over-time changes in attitudes towards democracy. Here, the posi-
tive effects (modernization) hypothesis posits that it is the higher educated
citizens that have become increasingly critical of politics, while lower edu-
cated citizens have experienced little or no change in their attitudes towards
democracy. Hence, the historical gap in political support between high and
low educated citizens has become increasingly smaller, and in some coun-
tries highly educated citizens now even are less supportive of politics than
low educated citizens. This is the process Dalton (2004) finds evidence for.
On the other hand, the negative effects (globalization) hypothesis posits that
it is the lower educated citizens that have become increasingly critical of

207
Kees Aarts, Jacques Thomassen, and Carolien van Ham

Graph A: Graph B:
modernization globalization

Political support
Political support
High SES

Low SES
Low SES
High SES

Time Time

Figure 11.1.  Trends in political support by socio-economic status

politics, while higher educated citizens have experienced little or no change

in their attitudes towards democracy. Hence, the gap in attitudes towards
democracy between high and low educated citizens is becoming increasingly
larger over time.5 These opposing dynamics are illustrated in Graphs A and B
in Figure 11.1.
If the globalization argument holds, we expect to find empirical support for
our third hypothesis (i.e. trends as outlined in Graph B in Figure 11.1):
H3: Over time, those who are at the margins of the economic order (i.e. the less
educated, the less skilled, and those with lower incomes) have increasingly nega-
tive attitudes towards democracy, while attitudes towards democracy of those that
are at the upper end of the economic order have remained stable.

While in the positive effects hypothesis the causal mechanism leading well-
off citizens to be more critical of politics is rather straightforward, in the nega-
tive effects hypothesis the causal mechanism connecting globalization
processes to attitudes towards democracy is less obvious and requires further
specification. In our discussion above we have proposed one way in which
globalization may be connected to attitudes towards democracy, by arguing
that citizens with different socio-economic status are differentially affected by
globalization, which subsequently alters their issue positions. The degree to
which such positions are catered for by political parties subsequently affects
the degree to which these citizens feel represented by parties and are satisfied
with democracy. This causal mechanism is drawn in Figure 11.2 below.
Note that the link between issue positions and feeling represented is once
again mediated by both the party system and the electoral system (not shown

Socio- Feelings of Satisfaction

Issue
economic being with
positions
status represented democracy

Figure 11.2. The causal connection between socio-economic status and political

support

208
 Globalization, Representation, and Attitudes towards Democracy

in Figure 11.2), as the availability of parties that match one’s positions as well
as the chances of such a party entering parliament are likely to affect feelings
of representation.6 In our empirical analyses, however, we will only test the
links between socio-economic position and issue positions, as well as issue
positions and feelings of representation and satisfaction with democracy; as
a first step to elaborate the causal mechanism proposed. Since the globaliza-
tion thesis refers to over-time change, hypotheses 4 and 5 are formulated
dynamically as well, as follows:
H4: Those who are at the margins of the economic order (i.e. the less educated, the
less skilled, and those with lower incomes) are increasingly found in quadrant A
of the socio-economic and cultural issue dimensions.
H5: Those who are in quadrant A of the socio-economic and cultural issue dimen-
sions feel increasingly less represented by political parties and are less satisfied
with democracy.

11.5  Data and Methods

11.5.1  Data
To test hypotheses 1 and 2, we use the second module of the Comparative
Study of Electoral Systems (CSES).7 This CSES module contains measures of
some of the key concepts needed for our purposes, although not all relevant
variables have been included in all participating countries. Since the CSES
data are crosssectional, to test hypothesis 3 we use Eurobarometer data on
socio-economic status and satisfaction with democracy for nine European
democracies from 1973 until 2010. Finally, to test hypotheses 4 and 5, we
rely on the data of the Dutch parliamentary election studies (DPES) from
1971 until 2010 (for specific variables, the time series are shorter and some-
times interrupted). The DPES contains several recurring indicators of citi-
zens’ attitudes towards politics. In addition, it also contains a variety of
questions on political issues, and information on people’s social and eco-
nomic background.8

11.5.2  Concepts and Measures

The theoretical arguments outlined in the previous section require that at the
very least three sorts of concepts are analysed. These are: a person’s socio-
economic status, issue positions on the socio-economic and cultural dimen-
sion, and attitudes towards democracy. For testing hypotheses 1, 2, and 3,
only socio-economic status and attitudes towards democracy are needed; for
testing hypotheses 4 and 5, we will also consider issue positions. The follow-
ing sections describe the indicators used to measure these concepts.

209
Kees Aarts, Jacques Thomassen, and Carolien van Ham

11.5.3  Cross-sectional Comparative Analysis: CSES

In CSES, which has a relatively high comparability of measures within each
module, three indicators are available for socio-economic position: social class,
income, and the level of education. Social class is derived from the respond-
ent’s position on the labour market. Kriesi et al. (2008: 349) developed a clas-
sification of social class in nine categories, which aims at capturing
simultaneously the vertical dimension of social class (lower versus higher and
managerial positions in the labour market) as well as the horizontal dimen-
sion that is of crucial importance for one’s position in a globalizing economy
(traditional industrial sectors versus services-oriented sectors). Using the data
available in the CSES Module 2, we build on the categories developed by
Kriesi et al. (2008) to create an indicator of social class.
The information used is contained in two variables in the CSES module: the
respondent’s current employment status and his/her main occupation. The
first variable is used to distinguish between persons who are currently in the
labour market and those who are out of it (because they are currently unem-
ployed, or still studying, or retired, or for other reasons). The second variable,
main occupation, classifies a person’s type of employment according to the
internationally developed ISCO coding.9 With the help of this code, the hori-
zontal dimension of socio-economic status, i.e. the sector of the economy
where one is employed, can be measured.10 The resulting classification of social
class consists of nine categories (see Table A.11.1 in the Appendix A.11.1).
Education and income have been used to create another indicator of socio-
economic status. In the CSES study, the level of education of the respondent
has been categorized so as to generate maximum comparability across coun-
tries, but the international comparability of education remains problematic. In
this chapter a rough distinction is made between respondents with post-sec-
ondary (often higher trade or vocational schools) and university-level educa-
tion (with or without a degree) on the one hand, and all other types of education
on the other hand. Household income, which is even less comparable across
countries than education, has been coded in quintiles. A simple combination
of education and household income results in an indicator with three catego-
ries. The middle category serves as a reference; the other categories contain
respondents with a clearly lower or higher socio-economic status than the ref-
erence group. Details can be found in Table A.11.2 in the Appendix A.11.1.
In conclusion, in the analyses we use two distinct measures of socio-­
economic status: one that is based on the respondent’s type of employment,
and one that is based on the combined measurement of education and
income.11 We will investigate the relationship between these two measures of
socio-economic status and two indicators of attitudes towards democracy in
the CSES module. The first indicator, feeling of representation, assesses whether

210
 Globalization, Representation, and Attitudes towards Democracy

any of the political parties in the country represents the respondent’s views
reasonably well. The response to this question can be yes or no. The second
indicator is satisfaction with democracy. This variable has been measured
directly in CSES Module 2, with responses on a 1–4 scale running from ‘very
satisfied’ to ‘not at all satisfied’.
Using the indicators for social class and for status implies that not all elec-
tion studies included in the CSES Module 2 can be analysed, leaving us with
28 elections in the analysis (see Table A.11.3 in Appendix A.11.1). With the
CSES study, the main propositions to be tested are hypotheses 1 and 2, i.e.
those who are at the margins of the economic order—the less educated, the
less skilled, and those with lower incomes—feel relatively poorly represented
by any of the political parties and are, perhaps therefore, relatively dissatis-
fied with the functioning of democracy. Moreover, we expect this gap to be
larger in majoritarian democracies than in consensus democracies.

11.5.4  Longitudinal Comparative Analysis: Eurobarometer

In the Mannheim Eurobarometer trend file, five indicators are available for
socio-economic position: type and sector of occupation, subjective social
class, income, and level of education. However, the Mannheim trend file only
runs until 2002, and hence for the current analyses we use a more recent
trend file of Eurobarometer data created by the University of Amsterdam, that
includes data until 2010.12 The latter dataset only includes education as an
indicator of socio-economic status. Education is measured as years at school,
and categorized into low, medium, and high education for the purpose of the
analysis (see Table B.11.1 in Appendix A.11.1). Attitudes towards democracy
are measured by satisfaction with democracy in one’s country, which in the
Eurobarometer is measured on a four-point scale ranging from ‘very satisfied’
to ‘not at all satisfied’. The data set includes 16 Western European democra-
cies. However, for only nine of these (the then member states of the EU) data
are available from 1973 onwards, allowing for thorough analyses of over-time
trends. Hence in our analyses we focus on the data from these nine democra-
cies, collected almost annually from 1973 until 2010 (33 timepoints). With
the Eurobarometer data, we will test hypothesis 3, i.e. that the gap in atti-
tudes towards democracy between those who are the margins and those who
are at the upper end of the economic order has increased over time, due to
decreasing support among those at the margins.

11.5.5  Unpacking Causal Mechanisms: DPES

In the DPES, socio-economic status is operationalized in two ways: a person’s
highest level of completed education, and the self-image of his/her social class.

211
Kees Aarts, Jacques Thomassen, and Carolien van Ham

Throughout 1971–2010, education has been measured with several scales (and
the school systems in the Netherlands have changed as well). It is possible
though to use a consistent rough distinction between those respondents who
have completed only primary and/or basic secondary school (usually circa four
more years of school after primary school), and those who obtained a diploma
from a more advanced type of school or university.13 Social class self-image has
been asked consistently throughout the 1971–2010 period, with a five-point
scale. In the analyses, a trichotomy has been applied between those respond-
ents considering themselves as (upper) working class, middle class, and upper-
middle or upper class. Attitudes towards democracy are measured with two
indicators: satisfaction with democracy and external efficacy. External efficacy
approximates the CSES item ‘feelings of being represented’ best, and is con-
structed as an index based on three survey items (see Appendix A.11.1). These
variables have been included in every election study since 1971. Satisfaction
with democracy has been asked in the Dutch studies only since 1998, and there-
fore only a relatively short time series is available for this key indicator.
In addition to socio-economic position and attitudes towards democracy,
in order to test hypotheses 4 and 5, we also need issue positions. We have
picked two political issues that measure respondents’ positions on the socio-
economic and cultural dimensions outlined in the theoretical section, i.e. the
desirability of larger or smaller income differences, and the desired extent of
European integration. Respondents have been asked to place themselves on
these issues, which have together been included in the DPES from 1994
onwards. Based on these issue positions, we can position respondents in one
of the four quadrants outlined in Table 11.1. Using the DPES data, we will test
hypotheses 4 and 5, i.e. those with lower socio-economic status are increas-
ingly found in quadrant A of the socio-economic and cultural issue dimen-
sions, and those in quadrant A of the socio-economic and cultural issue
dimensions, feel increasingly less represented by political parties and are less
satisfied with democracy.

11.6 Results
11.6.1  Cross-sectional Comparative Analysis: CSES
In the cross-sectional comparative analyses we test hypotheses 1 and 2, focus-
ing on the effects of socio-economic status on feelings of being represented
and satisfaction with democracy.
We first look at the feeling of being represented. In Table 11.2 the percent-
age differences are presented between contrast groups on the indicators of
social class and education/income (see above and Appendix A.11.1). In order
to be able to test hypothesis 2, consensus and majoritarian democracies are

212
 Globalization, Representation, and Attitudes towards Democracy

Table 11.2. Differences in percentage feeling represented by a political party between

contrast groups

High income economies Middle and low income economies

Election % difference % difference Election % difference % difference

social class education– social class education–
income income

Consensus democracies Consensus democracies

Czech Rep (2002) –0.9 Brazil (2002) 3.9 12.7**
Denmark (2001) 10.3*** Romania (2004) 23.9***
Finland (2003) 4.0 8.2*
Israel (2003) 3.4 7.8
Italy (2006) 10.8
Japan (2004) 9.6* 6.5
Netherlands (2002) 0.1 0.6
Norway (2001) 3.6 + 5.3 +
Poland (2001) 10.3* 10.2*
Slovenia (2004) –9.8* –8.7
Sweden (2002) 2.7 –2.9
Switzerland (2003) 4.0 –2.6
Average 3.10 3.72 Average 18.30

Majoritarian democracies Majoritarian democracies

Australia (2004) –0.7 –4.6 + Albania (2005) 14.5**
Canada (2004) 9.3** 13.7*** Bulgaria (2001) 13.8**
France (2002) 0.8 Chile (2005) 10.5*
Germany (2002) 8.5** 14.3*** Mexico (2003) –2.6 2.7
Great Britain (2005) 3.5 1.8 Peru (2006) 9.6**
Hungary (2002) 2.6 22.3*** Philippines (2004) 9.5 15.0***
Iceland (2003) 3.0 7.9 + Russia (2004) 0.2
Ireland (2002) 3.1 2.2
New Zealand (2002) 6.1* –7.1*
Portugal (2002) 0.3 12.2
Portugal (2005) 7.0* 18.3***
South Korea (2004) 8.8 12.5**
Spain (2004) 0.3 –7.0
Taiwan (2001) 1.8 3.6
Taiwan (2004) 3.1 4.5
United States (2004) 5.2 8.7*
Average 4.13 6.51 Average 3.45 9.47

Notes: Bivariate logit regression results. P-values: + 0.1, * 0.05, ** 0.01, *** 0.001.

presented separately. If our indicators are valid and it is true that persons in
disadvantaged socio-economic positions also feel less well represented, we
expect to find positive percentage differences in Table 11.2.
A large majority of the percentage differences is indeed positive, as expect-
ed.14 This means that the better-off social classes and those with higher scores
on the income-education indicator do feel represented by a political party
more often than persons on the other side of these social divides. The con-
trast is especially clear in Canada and Germany. In other countries, however,
the  differences are very small or negligibly close to zero. These include

213
Kees Aarts, Jacques Thomassen, and Carolien van Ham

Ireland, Britain, and the Netherlands. Grouping these countries in consen-

sual and majoritarian democracies, using the executive–parties dimension of
the two-dimensional classification developed by Bernauer et al. in this vol-
ume, does not seem to demonstrate significant differences in the effect of
social class on feelings of representation. Apparently, other factors, not
included in our analyses, are (also) important for the extent to which class
differences occur in the extent to which people find themselves represented
by a political party. Therefore, hypothesis 2 is rejected.
The second dependent variable in the analyses of CSES data is satisfaction
with the way democracy works in one’s country. In Table 11.3, percentage
differences corresponding to those reported in Table 11.2 are shown for the
percentage of respondents who state that they are very satisfied or satisfied
with the way democracy works in their country. Again, it is expected that the
less well-off in social-economic terms will also be less satisfied with the way
democracy works in their country, resulting in positive differences. All the
percentage differences in Table 11.3 are indeed positive—except in Israel and
Ireland—but again the differences in majoritarian democracies are not larger
than in consensus democracies. Therefore, there is some support for hypoth-
esis 1 but not for hypothesis 2.15
In summary, the evidence of differing feelings of representation, and of dif-
ferent levels of satisfaction with the functioning of democracy among differ-
ent social-economic groups, by and large conforms to our expectations. This
lends support to the negative effects hypothesis outlined in hypothesis 1, i.e.
those at the margins of the economic order are less positive about the func-
tioning of representative democracy than those at the upper end of the eco-
nomic order. However, those differences are not larger in majoritarian than in
consensus democracies as hypothesis 2 would predict.

11.6.2  Longitudinal Comparative Analysis: Eurobarometer

As we outlined in the hypotheses, the globalization argument is an over-time
argument, predicting decreasing political support among those at the margins of
the economic order and as a consequence, an increasing gap over time between
citizens of low and high socio-economic status in terms of their attitudes towards
democracy. To investigate whether this hypothesis (hypothesis 3) holds, we ana-
lysed Eurobarometer data for nine European democracies from 1973 until 2010.
The trend graphs presented in Figure 11.3 appear to show that, first, the
differences between educational groups in satisfaction with democracy are
very small in most countries. Contrasting high educated with low educated
citizens, the average difference in proportion of satisfied citizens varies
between1.3 per cent and 12.2 per cent. In six out of nine countries, on aver-
age higher educated citizens are more satisfied (once more corroborating

214
 Globalization, Representation, and Attitudes towards Democracy

Table 11.3. Differences in percentage very/fairly satisfied with the democratic process

between contrast groups

High income economies Middle and low income economies

Election % difference % difference Election % difference % difference

social class education– social class education–
income income

Consensus democracies Consensus democracies

Belgium (2003) 0.5 Brazil 14.1*** 12.5**
 (2002)
Czech Rep (2002) 15.5** Romania –0.7
 (2004)
Denmark (2001) 6.8***
Finland (2003) 8.1* 12.7**
Israel (2003) –6.6 + –6.5
Italy (2006) 3.3
Japan (2004) 1.0 3.9
Netherlands (2002) 7.2* 14.0**
Norway (2001) 1.3 10.3**
Poland (2001) 17.3*** 29.3***
Slovenia (2004) 8.1 + 17.6*
Sweden (2002) 5.7 + 1.4
Switzerland (2003) 5.4 + 13.3***
Average 4.8 10.13 Average 5.9

Majoritarian democracies Majoritarian democracies

Australia (2004) 3.7 4.5 Albania 8.1 +
 (2005)
Canada (2004) 3.2 2.7 Bulgaria 21.0***
 (2001)
France (2002) 9.9* Chile 6.2
 (2005)
Germany (2002) 10.1** 16.4*** Mexico 1.9 –6.3 +
 (2003)
Great Britain (2005) 3.3 13.5** Peru (2006) 3.6
Hungary (2002) 6.7 –3.5 Philippines 7.6 2.3
 (2004)
Iceland (2003) 1.7 5.3 Russia –7.0
 (2004)
Ireland (2002) –4.7* 0.8
New Zealand (2002) 1.8 3.1
Portugal (2002) 13.1** 21.1*
Portugal (2005) 10.2** 19.6***
South Korea (2004) –3.0 –1.8
Spain (2004) –4.2 –4.9
Taiwan (2001) –3.7 –6.1 +  
Taiwan (2004) –0.1 –3.2
United States (2004) –5.8 2.5
Average 2.15 4.99 Average 4.75 3.99

Notes: Bivariate logit regression results. P-values: + 0.1, * 0.05, ** 0.01, *** 0.001.

215
216

Kees Aarts, Jacques Thomassen, and Carolien van Ham

100 Belgium 100 Denmark 100 France

80 80 80

60 60 60

40 40 40

20 20 20

0 0 0

1973

1978

1983

1988

1993

1998

2003

2008

1973

1978

1983

1988

1993

1998

2003

2008

1973

1978

1983

1988

1993

1998

2003

2008
Low educated Medium educated Low educated Medium educated Low educated Medium educated
High educated High educated High educated

100 Germany 90 Ireland 100 Italy

80 80 80

60 70 60

40 60 40

20 50 20

0 40 0
1973

1978

1983

1988

1993

1998

2003

2008

1973

1978

1983

1988

1993

1998

2003

2008

1973

1978

1983

1988

1993

1998

2003

2008
Low educated Medium educated Low educated Medium educated Low educated Medium educated
High educated High educated High educated

100 Luxembourg 100 Netherlands 100 United Kingdom

80 80 80

60 60 60

40 40 40

20 20 20

0 0 0
1973

1978

1983

1988

1993

1998

2003

2008

1973

1978

1983

1988

1993

1998

2003

2008

1973

1978

1983

1988

1993

1998

2003

2008
Low educated Medium educated Low educated Medium educated Low educated Medium educated
High educated High educated High educated

Figure 11.3.  Trends in satisfaction with democracy by education in nine European democracies
 Globalization, Representation, and Attitudes towards Democracy

hypothesis 1). However, in three out of nine countries differences between

educational groups are either very small or reversed, with the lower educated
on average being more satisfied.
Considering trends over time, Figure 11.3 appears to show that trends fol-
low a rather similar pattern for all educational groups, suggesting that the
explanatory power of education is quite moderate and that other causes
explain the over-time variation in satisfaction with democracy. When the
difference in satisfaction with democracy between high educated and low
educated citizens is regressed on time, this gap between educational groups
appears to have increased significantly in six of nine cases. In all these cases
the gap appears to be rather small, though. These analyses and Figure 11.3
provide some support for hypothesis 3, but the graphs make clear that this
support is feeble.

11.6.3  Unpacking Causal Mechanisms: DPES

We argued above that globalization might affect the connection between
socio-economic status and attitudes towards democracy by changing the
issue positions of citizens at the margin of the economic order, subsequently
affecting those citizens’ attitudes towards democracy as well. Using the Dutch
Parliamentary Election Studies we test hypotheses 4 and 5, the former testing
the link between socio-economic status and issue positions, and the latter
testing the link between issue positions and attitudes towards democracy. We
expect that over time, those at the margins of the economic order are increas-
ingly found in quadrant A of the socio-economic and cultural issue dimen-
sions as sketched in Table 11.1, and that those who are in quadrant A of the
socio-economic and cultural issue dimensions, feel increasingly less repre-
sented by political parties and are less satisfied with democracy.
We set out by examining the trends of the main variables. Figures 11.4 and
11.5 show the trends in the indicators of socio-economic status used here: level
of completed education and social class self-image.16 In accordance with the
expectation, the percentage of respondents who only completed lower or
medium education dropped from over 90 per cent in 1971 to just over 30 per
cent in 2010. This means that, through a combined process of increased partici-
pation in higher education and population replacement, persons with lower or
medium education only have become a small minority of the electorate in the
2000s. Similarly, as Figure 11.4 shows, the percentage of respondents who con-
sider themselves to be working class or upper working class has over the same
period decreased from almost half of the electorate to circa 30 per cent.17
How are these indicators of socio-economic status related to attitudes
towards democracy over the years? In Figures 11.6 and 11.7, the trends
in external political efficacy (taken as an indicator of feelings of being

217
Kees Aarts, Jacques Thomassen, and Carolien van Ham

100 lower/medium education

90

80

70

60
percentage

50

40

30

20

10

0
1970 1975 1980 1985 1990 1995 2000 2005 2010
year
Source: Dutch National Election Studies, weighted data

Figure 11.4.  Level of completed education, the Netherlands, 1971–2010

100 (upper) working class

90

80

70

60
percentage

50

40

30

20

10

0
1970 1975 1980 1985 1990 1995 2000 2005 2010
year
Source: Dutch National Election Studies, weighted data

Figure 11.5.  Social class self-image, 1971–2010

218
 Globalization, Representation, and Attitudes towards Democracy

represented) and in satisfaction with democracy are depicted according to

level of education and social class, respectively.
In Figure 11.6 external efficacy and satisfaction with democracy are broken
down to lower/medium and higher levels of completed education. Several
observations should be made. First, the sense of external efficacy (the belief

High External Efficacy

100 lower/medium education
higher education

80

60
percentage

40

20

0
1970 1980 1990 2000 2010
year

High Satisfaction With Democracy

100
lower/medium education
higher education

80

60
percentage

40

20

0
1970 1980 1990 2000 2010
year

Figure 11.6.  Level of education and attitudes towards politics

219
Kees Aarts, Jacques Thomassen, and Carolien van Ham

High External Efficacy

100 upper (middle) class
middle class
(upper) working class

80

60
percentage

40

20

0
1970 1980 1990 2000 2010
year

High Satisfaction With Democracy

100 upper (middle) class
middle class
(upper) working class

80

60
percentage

40

20

0
1970 1980 1990 2000 2010
year

Figure 11.7.  Social class self-image and attitudes towards politics

220
 Globalization, Representation, and Attitudes towards Democracy

that politics is responsive to citizens like oneself) and satisfaction with democ-
racy do not show clear upward or downward trends in the time intervals for
which data are available. It is also clear that the lower- and medium-educated
respondents have a consistently lower sense of external efficacy than the
higher-educated respondents. While this does demonstrate that the negative
effects hypothesis holds in the Netherlands when considering static differ-
ences between social groups, confirming hypothesis 1 for this specific coun-
try, it also indicates that the gap between the socio-economic advantaged and
disadvantaged does not seem to be widening, disconfirming hypothesis 2 in
the case of the Netherlands. Rather, Figure 11.6 seems to show class differ-
ences in attitudes towards democracy that appear to be rather stable over
time.18 The only period in which the gap between socio-economic groups
briefly narrows is around the 2002 election (in which Pim Fortuyn’s LPF par-
ticipated with much success). The time series for satisfaction with democracy
is shorter than that for the other variables in Figure 11.6, as the question was
only included in the DPES from 1998 onwards. It is still remarkable that in
1998 and 2002, the first two time points for this variable, the difference
between lower/medium-educated and higher-educated citizens is barely
noticeable. Only since 2003, differences in satisfaction with democracy are
observed for people with different educational levels, with higher-educated
people being relatively more often satisfied.
Similar patterns as observed in Figure 11.6 appear when we consider the
relationship between attitudes towards democracy and social class self-image.
Figure 11.7 shows the trends for three levels of social class: upper (middle),
middle, and (upper) working class. Almost without exception, people with a
higher self-image of their social class have a higher sense of external efficacy.
Again, social class appears to have a much weaker impact on people’s satisfac-
tion with democracy, at least until 2003. After 2003, social class is associated
with different levels of satisfaction with democracy as well, with the higher
classes being the most satisfied. Taken together, Figures 11.4–11.7 have made
it clear that in the case of the Netherlands, which serves here as a pars pro toto,
ever fewer citizens have a lower or medium level of education and ever fewer
citizens consider themselves as working class. At the same time, education
and social class are clearly associated with attitudes towards democracy (exter-
nal efficacy and satisfaction with democracy), with higher education and
class leading to higher efficacy and higher satisfaction with democracy.19

11.6.4  Issue Positions

The next step is to assess the impact of the trends just outlined on people’s
positions on political issues—income differences and European integration—
and the relationship of socio-economic status and positions on these issues as

221
Kees Aarts, Jacques Thomassen, and Carolien van Ham

well as the relation between these issue positions and attitudes towards
democracy. The goal of this analysis is to test the causal mechanism by which
globalization might affect the relation between socio-economic status and
attitudes towards democracy, as outlined in Figure 11.2. We focus first on the
relationships between socio-economic status (education and social class) and
these issue positions over the 1994–2010 period, testing hypothesis 4, and
then consider the relation between issue positions and attitudes towards
democracy, testing hypothesis 5.
Table 11.4 shows the results for level of education. As a reading example,
consider the penultimate column in Table 11.4. This column shows the pro-
portion of respondents with issue positions in quadrant A as outlined in
Table  11.1, i.e. those respondents who were in favour of demarcation on
both the socio-economic and cultural dimensions. These issue positions are
measured here as being both in favour of smaller income differences and less
European integration. Of respondents with a lower–medium education in
1994, 13 per cent placed themselves in this quadrant. By 2010, this propor-
tion had increased to 39 per cent (and it was even higher in 2006). Among
the respondents with higher education, the percentage of respondents plac-
ing themselves in quadrant A also increased, but these percentages are much
lower (11 to 23 per cent respectively), indicating a clear effect of education
on these issue positions. Several other aspects of the results in Table 11.4 are
worth noting. First, consider the relative sizes of the different groups of
respondents over the five election years. The first column of Table 11.4 shows
the group of respondents that take middle or don’t know positions on one or
both issues. This group appears to have shrunk, from 55.5 per cent in 1994
to 35.2 per cent in 2006, and then went up again somewhat in 2010. The
underlying reason for this trend is that fewer respondents ended up in the
middle or the ‘don’t know’ category of the European unification issue. Euro-
pean unification provoked more people to take a for- or against-position in
these years. In practice, this development favoured the camp against further
integration. Again, the largest change can be found in quadrant A. Table 11.4
shows that the group of persons who want less integration and also smaller
income differences grew from 13.0 per cent in 1994, to 39.0 per cent in 2006,
and then decreased again to 29.8 per cent in 2010.20 Citizens chose sides on
European unification, and those who opposed larger income differences also
tended to oppose further European integration. The other three groups show
some changes in relative size, but much less.
A similar analysis has been conducted for social class self-image and the
issue typology, and the results are shown in Table 11.5. The table can be read
in the same way as Table 11.4. Again, the largest change appears to have
occurred in the group of respondents in quadrant A, i.e. those who support
smaller income differences and less European integration: the higher one’s

222
 Globalization, Representation, and Attitudes towards Democracy

Table 11.4. Level of education, and position on income differences and European

integration

Neutral Larger Smaller Larger Smaller N

or don’t differences, differences, differences differences,
know more more less less
integration D integration C integration B integration A

1994 All 55.5 8.9 15.3 7.2 13.0 1,522

Lower/medium 58.0 7.6 14.0 7.1 13.3 1,227
 education
Higher 45.4 14.6 20.7 7.8 11.5 295
 education
1998 All 47.6 7.1 19.7 4.9 20.6 1,811
Lower/medium 44.4 7.0 17.7 5.7 25.2 968
 education
Higher 51.2 7.2 22.1 4.0 15.4 843
 education
2002 All 48.2 5.5 18.7 7.1 20.5 1,562
Lower/medium 47.9 4.0 16.7 6.6 24.8 580
 education
Higher 48.4 6.3 20.0 7.4 17.9 982
 education
2006 All 35.2 3.9 15.8 6.0 39.0 2,040
Lower/medium 31.4 4.8 18.7 5.5 49.8 656
 education
Higher 38.2 4.8 18.7 6.4 31.9 1,384
 education
2010 All 41.0 4.8 18.8 5.6 29.8 1,927
Lower/medium 38.3 3.0 14.8 4.4 39.5 608
 education
Higher 43.4 5.8 21.0 6.2 23.7 1,319
 education

subjective class, the smaller this group is and vice versa. Among respondents
who identify themselves as working class the proportion supporting issue
positions in this quadrant in 1994 is 15 per cent and increases over time to
reach 41 per cent in 2010. Tables 11.4 and 11.5 thus show that the socio-
economic position of citizens matters for how they think about political
issues, confirming hypothesis 4. How are people’s positions on these issues
related to attitudes towards democracy?
Table 11.6 reports the percentages of respondents with a high level of
external efficacy and satisfaction with democracy among the various
groups defined on the basis of their position on the issues of income differ-
ences and European integration. For example, of the persons in quadrant
A, i.e. those who preferred smaller income differences and less European
integration in 1994, 37.7 per cent had a high score on the index of external
efficacy. This compares with 59.8 per cent of persons with high external
efficacy among those who prefer smaller income differences and more
integration.

223
Kees Aarts, Jacques Thomassen, and Carolien van Ham

Table 11.5. Social class self-image, and position on income differences and European
integration

Neutral Larger Smaller Larger Smaller N

or don’t differences, differences, differences differences,
know more more less less
integration D integration C integration B integration A

1994 All 55.6 9.1 15.3 7.1 12.8 1,480

(Upper) 57.3 3.8 17.2 5.9 15.8 443
  working class
Middle class 56.1 8.9 14.8 7.4 12.7 795
Upper (middle) 50.8 19.4 13.6 8.3 7.9 242
 class
1998 All 47.4 7.1 19.7 5.0 20.8 1,786
(Upper) 37.7 4.5 18.1 3.9 35.8 509
  working class
Middle class 51.5 7.4 20.0 5.0 16.1 999
Upper 50.7 10.8 21.2 6.8 10.4 278
  (middle) class
2002 All 47.2 5.4 19.1 7.3 20.9 1,505
(Upper) 45.9 2.1 18.2 3.8 30.1 292
  working class
Middle class 46.5 5.9 18.3 8.3 21.0 867
Upper 50.0 7.2 22.0 7.8 13.0 346
  (middle) class
2006 All 35.0 3.8 16.0 6.1 39.1 2,123
(Upper) 27.8 1.4 13.0 4.5 53.3 733
  working class
Middle class 37.0 3.5 15.4 6.9 37.1 980
Upper 42.7 9.0 22.7 7.1 18.5 410
  (middle) class
2010 All 41.0 4.8 18.8 5.6 29.8 1,927
(Upper) 34.8 1.4 16.2 5.8 41.7 635
  working class
Middle class 41.2 4.7 19.6 5.4 29.2 1,008
Upper 48.4 10.2 22.7 6.8 12.0 384
  (middle) class

The patterns in Table 11.6 are clear. Citizens who think that European inte-
gration has already gone too far tend to feel less well represented, as measured
by external efficacy, than the persons belonging to the other groups (with a
single exception in 2002). This is regardless of whether they favour larger or
smaller income differences. As for satisfaction with democracy, in 1998 it
appears to be still uniformly high among all groups (with a low of 75.5 per
cent among those who want smaller differences and less integration). In the
more recent years (2002–10), however, it appears that satisfaction with
democracy is markedly lower among persons who oppose further European
integration. With some delay, satisfaction with democracy thus shows the
same pattern as external efficacy. External efficacy and satisfaction with

224
 Globalization, Representation, and Attitudes towards Democracy

Table 11.6 Position on income differences and European integration, and attitudes

towards democracy

Neutral Larger Smaller Larger Smaller

or differences, differences, differences differences,
don’t more more less less
know integration D integration C integration B integration A

High external 1994 50.5 67.2 59.8 38.5 37.7

 efficacy 1998 54.6 49.6 69.5 47.2 41.7
2002 49.6 46.5 64.5 33.3 33.3
2006 59.5 79.3 66.6 48.1 44.8
2010 60.5 64.0 65.0 44.9 37.2
High satisfaction 1998 87.0 90.2 89.9 90.9 75.5
  with democracy 2002 51.4 56.0 54.6 28.3 34.5
2006 81.6 86.0 86.1 67.7 67.3
2010 77.9 84.5 82.9 60.2 62.8

democracy are highest among those who want further integration. While this
partly confirms hypothesis 5, i.e. those with issue positions in quadrant A
have less positive attitudes towards democracy, the relationship is weakened
by the fact that citizens with issue positions in quadrant B share these
attitudes.
So far we have been able to reconstruct in part the process that leads to
lower satisfaction with democracy. It involves socio-economic changes,
which affect issue positions on both socio-economic and cultural dimen-
sions, which subsequently affect attitudes towards democracy. It appears to
be especially the position on the cultural dimension, exemplified here by
European integration, that is related to attitudes towards democracy.

11.7 Conclusion

This chapter aimed at assessing the validity of the claim that a new structural
conflict dimension in the politics of industrialized democracies is gradually
affecting the legitimacy of democracy among specific social-economic groups
in these democracies. We have analysed this claim via three sets of analyses:
cross-nationally using the Comparative Study of Electoral Systems, cross-
nationally and longitudinally using the Eurobarometer, and longitudinally
using the Dutch parliamentary election studies.
In the cross-national analysis based on CSES data we have assessed to
what extent disadvantaged socio-economic positions go together with the
feeling that one is not represented by any political party, and with dissatis-
faction with democracy. With some notable exceptions, a large majority of

225
Kees Aarts, Jacques Thomassen, and Carolien van Ham

the percentage differences for the feeling of being represented in Table 11.2
point in the expected direction, confirming the negative effects hypothesis.
Hence, in the majority of democracies in our sample, the positive effects
hypothesis is refuted. Instead, it is those citizens on the margin of the eco-
nomic order that feel less well represented and are less satisfied with democ-
racy. However, it must be noted that the differences are generally not very
large, indicating that socio-economic status is most likely only one of mul-
tiple factors explaining attitudes towards democracy. Moreover, we did not
find evidence for a larger gap between citizens of different socio-economic
backgrounds in majoritarian democracies than in consensus democracies.
Furthermore, in the longitudinal analysis based on Eurobarometer data, we
have shown that the gap between different social groups has not increased,
as the globalization argument proposes (nor has it decreased, as the mod-
ernization thesis posits). Rather, class differences in attitudes towards
democracy appear to follow similar fluctuations over time, and the gap
between groups appears to fluctuate over time, rather than consistently
increase or decrease.
In the analysis based on DPES data from the Netherlands we attempted to
unpack the causal mechanism connecting globalization processes to attitudes
towards democracy. We found that in the case of the Netherlands the level of
education has clearly risen and self-identification as working class has become
ever more rare. Higher education and social class are also associated with
higher external efficacy, and—after 2002—also with satisfaction with democ-
racy. This is in line with the negative effects hypothesis. The case of the Neth-
erlands thus bears out the central expectations formulated in the first section
of the chapter: those who are at the margins of the economic order: the less
educated, the less skilled, and those with lower incomes have lower external
efficacy and are relatively dissatisfied with the functioning of democracy. As
regards the causal mechanism implied by the negative effects hypothesis and
the globalization argument, the analysis of the Dutch case sought to demon-
strate how socio-economic characteristics affect citizens’ positions on old and
new political issues, which subsequently affect attitudes towards democracy.
Here, we found that indeed those at the margin of the economic order tend
to increasingly favour ‘demarcation’ positions on both the socio-economic
and the cultural dimension. Moreover, those issue positions appear to be also
related to more negative attitudes towards democracy.
Finally, though we discussed a variety of relationships that seem to confirm
our expectations, it must be noted that practically all of those relationships
were weak to modest. We are observing tendencies, not molar causation (to
borrow a term from Shadish, Cook and Campbell 2002: 10). Future analyses
should seek to further examine the causal mechanisms connecting processes
of globalization to attitudes towards democracy.

226
 Globalization, Representation, and Attitudes towards Democracy

Appendix—A.11.1: Indicators and sample used

A. CSES
Table A.11.1.  Construction of CSES social class indicators

Name Employment status Occupation Group

(B2010) (B2011) size

Self-employed farmers 1 ≤3 61, 62, 92 33.66%

Other self-employed, non-professional 2 ≤3 94
Semi- or unskilled workers 3 ≤3 81–84, 91, 93
Skilled workers, foremen 4 ≤3 51–53, 71–74
Routine non-manual workers (white collar) 5 ≤3 41–43
Managers and other professionals 6 ≤3 1, 11–13 21.30%
  in socio-administrative occupation
Professional with technical expertise 7 ≤3 21, 22, 24, 31,
  32, 34
Social-cultural specialists 8 ≤3 23, 33
Not in labour force 9 > 3 &< 11 – 45.04%

Note: * For the purpose of the analyses in Tables 2 and 3, these more fine-grained indicators of social class were
regrouped into lower social class (1–5), higher social class (6–8), and not in the labour force (9).

Table A.11.2.  Construction of CSES socio-economic status indicator

Name Education (B2003) Income (B2020) Group size

Low status 1 Secondary, or lower Lowest two quintiles 32.89%

Medium status 2 (all other respondents) 48.03%
High status 3 Post-secondary, or higher Highest two quintiles 19.07%

Table A.11.3.  CSES Module 2 elections included in the analyses

Country Year Country Year

High income economies Middle and low income economies

Consensus democracies Consensus democracies

Belgium 2003 Brazil 2002
Czech Republic 2002 Romania 2004
Denmark 2001
Finland 2003 Majoritarian democracies
Israel 2003 Albania 2005
Italy 2006 Bulgaria 2001
Japan 2004 Chile 2005
Netherlands 2002 Mexico 2003
Norway 2001 Peru 2006
Poland 2001 Philippines 2004
Slovenia 2004 Russia 2004
Sweden 2002
Switzerland 2003

(Continued)

227
Kees Aarts, Jacques Thomassen, and Carolien van Ham

Table A.11.3.  (Continued)

Country Year Country Year

High income economies Middle and low income economies

Majoritarian democracies
Australia 2004
Canada 2004
France 2002
Germany 2002
Great Britain 2005
Hungary 2002
Iceland 2003
Ireland 2002
New Zealand 2002
Portugal 2002
Portugal 2005
South Korea 2004
Spain 2004
Taiwan 2001
Taiwan 2004
United States 2004
Not classified
Hong Kong 2004

Source: Two-dimensional factor solution Bernauer et al., executive–parties dimension. Scores below 0 were coded as
majoritarian and scores above 0 as consensus. Classification of economies into high income economies and middle
and lower income economies based on the World Bank classification by GNI per capita 2011 (see: <http://data.world-
bank.org/about/country-classifications>).

Table B.11.1.  Eurobarometer education indicator

Name Education in years Proportion of respondents

Low education 1 Up to 14 years 24.93%

Medium education 2 15–19 years 44.79%
High education 3 20 or more years, still studying 30.28%

B. Eurobarometer
Table B.11.2.  Eurobarometer sample: countries and time period included

Country Time period

France 1973–2010
Belgium 1973–2010
The Netherlands 1973–2010
Germany (West & East) 1973–2010
Italy 1973–2010
Luxembourg 1973–2010
Denmark 1973–2010
Ireland 1973–2010
United Kingdom (Great Britain & Northern Ireland) 1973–2010

228
 Globalization, Representation, and Attitudes towards Democracy

C. DPES
Table C.11.1  Indicator for ‘feelings of representation’ in DPES

Feelings of representation have been measured in DPES by the external efficacy index, an index
consisting of three survey questions. The external efficacy score is calculated by a count of the
number of ‘not true’ answers to the following questions:

• Members of parliament do not care about the opinions of people like me.
• Political parties are only interested in my vote and not in my opinion.
• People like me have no influence on governmental policy

Citizens with scores of 2 or higher are considered to have high external efficacy.

Notes

1. See also Kitschelt (1994) and Flanagan and Lee (2003).

2. See for the same argument van der Brug (2004) and Van der Brug and Van Spanje
(2009).
3. A third reason may be that the composition of electorates in advanced industrial
democracies has changed, and the lower-educated now constitute a much smaller
proportion of the population. This might further increase their sense of
marginalization.
4. What speaks against this expectation, though, is that one of the defining charac-
teristics of populist parties is that they are anti-system parties agitating against the
established political parties and therefore encourage rather than accommodate
negative feelings toward the political institutions and its main actors. As a conse-
quence even in those countries where populist parties have been successful in
gaining representation the losers of globalization might still feel estranged from
the system of representative democracy.
5. Note that in order to make plausible that it is globalization processes that affect
political support, the gap between high and low educated should be increasing
over time; if the gap is static it reflects class differences in political support that are
stable over time.
6. Moreover, note that the causal link between issue positions and feelings of repre-
sentation might run in the opposite direction as well. For example, feelings of not
being well represented may drive demarcation issue positions such as Euroscepti-
cism and anti-immigrant positions.
7. The final, 2007 version of CSES Module 2 is used, which can be obtained from
<http://www.cses.org>.
8. An integrated data set for the 1971–2006 Dutch elections is available from Data
Archiving and Networking Services (DANS—<http://www.dans.knaw.nl>), and
documented in Todosijevic, Aarts and van der Kaap (2010). The 2010 DPES data
are available from DANS as well. Data have been weighted with a structurally
identical model over the various studies, including voting behaviour.

229
Kees Aarts, Jacques Thomassen, and Carolien van Ham

9. ISCO stands for International Standard Classification of Occupations and was

developed by the International Labor Organization ILO. Refer to <http://www.ilo.
org/public/english/bureau/stat/isco/index.htm>, accessed 4 November 2012.
10. In order to measure the vertical dimension of social class, a third variable con-
tained in the CSES, socio-economic status, could be used, as it distinguishes
between white- and blue-collar workers, farmers and self-employed persons. How-
ever, since the combination of occupation and social class resulted in many miss-
ing cases (as not all relevant information is available for all countries), we rely only
on the main occupation of the respondent and do not use socio-economic status.
This is acceptable since the horizontal distinctions of social class are likely to be
more relevant to test the effects of globalization.
11. Social class, education, and income are multidimensional; each indicator con-
tains distinct information about a person’s socio-economic status. To be sure,
there is a medium-to-strong positive correlation between education and income,
but the correlation of both education and income with the respondent’s social
class (which here refers to his status on the labour market) is low or even absent.
This means that social class, education, and income cannot simply be added to
form a single composite measure of socio-economic status. Instead, we use the
two distinct measures of socio-economic status separately: one that is based on
the respondent’s type of employment, and one that is based on the combined
measurement of education and income.
12. This data set is described in Hakhverdian et al., forthcoming.
13. Only in the election studies of 1989 and 1994, this basic dichotomy is deceptive,
since in those studies a different coding scheme has been used. Therefore, in
describing the trend in educational level in the Netherlands we decided to drop
1989 and 1994. For explanatory purposes, the deviant coding schemes in 1989
and 1994 do not pose serious problems, and the 1989 and 1994 studies have con-
sequently been adopted in bivariate analyses.
14. More precisely, considering social class, we find positive differences in 25 out of 28
democracies, i.e. those with higher social class are more likely to feel represented
by a political party. However, the differences are small and only in seven out of 28
democracies was the bivariate regression of social class on feeling represented
positive and significant. Considering the combined education/income indicator,
we find positive differences in 31 out of 38 democracies, i.e. those with higher
education and income are more likely to feel represented by a political party. The
differences appear larger than with social status, and in 17 of 28 democracies the
effect of combined education/income is positive and significant in a bivariate
regression predicting feelings of representation.
15. More precisely, considering social class, we find positive differences in 21 out of 29
democracies, i.e. higher social status respondents were more satisfied with democ-
racy in those countries. In ten of 29 cases, a bivariate regression of social status on
satisfaction with democracy demonstrated a positive and significant coefficient.
However, there were also eight countries with negative differences (most notably,
the USA), and some of those were also significant. Turning to education, we find
positive differences in 18 out of 28 democracies, 16 of which are significant.

230
 Globalization, Representation, and Attitudes towards Democracy

16. The Dutch election studies data cover varying time periods. For some variables,
the full 1971–2010 period is available, for other variables the period is shorter.
17. Although the data have been weighted before the analysis, the relatively low per-
centage of (upper) working class respondents in 2002, 2003, and perhaps even
earlier, may at least partly be attributed to selective non-response. The response
rates in the Dutch election studies had markedly deteriorated after the mid-1980s,
to recover only in 2006 as a result of a set of measures aimed at improving response
quality.
18. Note that the changing composition of the Dutch electorate in terms of socio-
economic groups as demonstrated in Figures 11.4 and 11.5 might nevertheless
affect the overall level of attitudes towards democracy. While group differences
are rather stable, the size of the group with lower socio-economic status has
declined over time, contributing to an aggregate slight increase in levels of politi-
cal support.
19. Note that the findings for satisfaction with democracy need to be interpreted with
some caution because the time series is relatively short. It seems that education
and social class had a hardly noticeable impact on people’s satisfaction with
democracy between 1998 and 2002. The Eurobarometer data for the Netherlands
appear to indicate that the narrow gap between 1998 and 2002 was temporary,
however. After 2002, indeed, satisfaction with democracy appears to follow a simi-
lar pattern as external efficacy: the well-educated and the higher classes tend to be
more often satisfied with democracy than the less well educated and the lower
classes.
20. The peak in 2006 is probably due to the heated discussion around the referendum
on the European Constitutional Treaty in 2005 when 60 per cent of the Dutch
voters voted against the treaty.

231
12

Consenting to Lose or Expecting to Win?

Inter-temporal Changes in Voters’
Winner–Loser Status and Satisfaction
with Democracy
Eric Chang, Yun-han Chu, and Wen-chin Wu

12.1 Introduction

This chapter investigates how inter-temporal changes in the electoral ­winner–

loser status affect voters’ beliefs in the legitimacy of young democracies. More
specifically, we focus on ‘losers’ of the most recent election and ask whether
voters with experience of winning in the past have different levels of satisfac-
tion with democracy than those voters who have lost two consecutive elec-
tions. In addition, we ask whether past electoral losers have their faith in
democracy restored after they become ‘winners’ in the most current election.
These questions are theoretically important because they simultaneously
engage two interrelated branches of literature. First, these questions advance
our understanding of the winner–loser gap thesis pioneered by Anderson and
Guillory (1997). In a series of seminal works, Anderson and his co-authors
compellingly demonstrate that electoral winners are more satisfied with dem-
ocratic practices than losers. Also important, they discover that the gap
between winners and losers is larger in countries with single-member district
(SMD) systems than in countries with proportional representation systems
(PR) (Anderson and Guillory 1997; Anderson et al. 2005). Subsequent research
further demonstrates that the winner–loser gap of satisfaction with democ-
racy exists in emerging democracies in Africa (Cho and Bratton 2006), Latin
America (Booth and Seligson 2009), and post-communist Eastern European
countries (Ruiz-Rufino 2013). Furthermore, similar winner–loser gaps can be

232
 Consenting to Lose or Expecting to Win?

found in citizens’ evaluations of government performance (Anderson and

Tverdova 2001) and political trust in government (Anderson and LoTempio
2002).
Despite the theoretical sophistication and empirical soundness of these
seminal works, they distinguish electoral winners and losers solely on the
basis of a single election and treat the interaction between winners and losers
as a one-shot game. However, since elections in democracies are held rou-
tinely, winners can become losers when electoral alternation occurs. Given
the repetitive nature of elections and the possibility of electoral alternation,
we suggest modelling winner–loser status as a repeated game. Viewing the
electoral winners and losers from this new vantage point, our findings chal-
lenge the conventional wisdom by showing that current losers are not neces-
sarily less satisfied with democracy once we take into account their past
experience of winning. Similarly, after factoring in the expectation of win-
ning in the current election, we show that the gap in support for democracy
between past winners and losers is not as sharp as the extant literature
suggests.
Second, this chapter contributes to research on Huntington’s two-turnover
test in the comparative democratization literature. According to Huntington
(1991: 266–7), a nascent democracy is considered consolidated if it has expe-
rienced two peaceful electoral alternations. At its core, the two-turnover test
evaluates the level of democratic commitment among political elites, since
passing this test suggests that all political forces have agreed to challenge for
office via elections rather than by force or other illegitimate means. Despite
the popularity and utility of Huntington’s test in determining whether a new
democracy has matured, the literature overemphasizes the importance of
elites’ democratic commitment, and consequently neglects the importance
of ordinary citizens’ democratic attitudes during electoral alternations in
young democracies. This analytical imbalance is unfortunate considering
that democracy thrives on popular support and withers in its absence (Easton
1965). Our chapter redresses this insufficiency and complements the two-
turnover test literature with a mass-level analysis by demonstrating how citi-
zens’ democratic support varies as their winner–loser status changes during
repeated elections.
To investigate impacts of inter-temporal change of winner–loser status on
citizens’ political attitudes toward democracy, we utilize the data collected by
the project of Comparative Study of Electoral Systems Module 2 (CSES 2
henceforth). The CSES 2 data set is ideal for our study as it systematically
tracks citizens’ voting history in two consecutive elections. We differentiate
respondents into four groups to document the inter-temporal change of vot-
ers’ winner–loser status depending on whether or not an individual voted for
the ruling party or candidate in the past two elections: previous losers who

233
Eric Chang, Yun-han Chu, and Wen-chin Wu

win the current election (LW), previous winners who lose the current election
(WL), two-time winners (WW), and two-time losers (LL).1
Given our fundamental concerns on democratization, we first narrow our
analytical focus to the third-wave democracies that experience governmental
turnover, including Brazil, Bulgaria, Chile, Hungary, Poland, and Romania.2
Our results show that satisfaction with democracy among the LW, WL, and
WW groups is not significantly different between these groups, while the LL
group is less satisfied than the other three groups. In other words, our results
demonstrate that both the experience of winning and the expectation of elec-
toral alternation are important in sustaining citizens’ satisfaction with
democracy. Then we expand our sample by including both developing and
established democracies. According to Lipset (1959) and Easton (1965), sup-
port for democratic principles, or the diffuse support for democratic systems
more generally, can only incubate slowly in a mature democratic system. This
‘reservoir of good will’ developed over time in advanced democracies would
then help sustain their citizens’ faith in democracy even when they find
themselves sailing against a tough but temporary economic or political tide
(Thomassen and Van der Kolk 2009). In other words, one can expect this
dynamic winner–loser gap to be smaller in developed democracies than
young democracies, and our empirical analysis shows that it is indeed the
case.
The rest of this chapter proceeds as follows. Section 12.2 discusses the con-
tributions and deficiencies of the extant literature on the winner–loser gap of
satisfaction with democracy and on the two-turnover test. Section 12.3 elabo-
rates our dynamic winner–loser thesis under repeated elections. Section 12.4
tests our hypothesis about the dynamic winner–loser status and presents the
empirical results. Section 12.5 discusses the implications of the results for
democratic consolidation and provides our conclusions.

12.2  The Literature

Most democratic theorists would agree that little is as important as popular

support for democratic principles and practices for the success and the
advancement of democratization. Some even assert that mass support
for  democracy is the attitudinal foundation of democratic consolidation
­(Schedler 2001). Indeed, as Linz and Stepan (1996a, 1996b) forcefully argue,
the fundamental prerequisite for consolidation of a new democracy is the
presence of a majority of citizens who believe in and support democracy.
Put differently, democratic consolidation is unlikely to occur if the very
notion of democracy is not accepted as the fundamental mode of legitimi-
zation (­ Diamond 1999).

234
 Consenting to Lose or Expecting to Win?

Indeed, a great amount of scholarly effort has been devoted to identifying

the factors that might induce citizens to develop favourable orientations
toward democracy. One particularly insightful approach pioneered by Ander-
son and his co-authors highlights the difference between electoral winners
and losers in their attitudes toward democracy. According to this perspective,
citizens who voted for the winning parties tend to be more satisfied with
democracy than those who voted for the losing parties (Anderson and Guil-
lory 1997; Anderson and LoTempio 2002; Anderson and Tverdova 2003).
This proposition is built upon the simple intuition that the political system is
more favourable to people whose supported party is in power, and that indi-
viduals tend to favour rules that make them win whereas losers may attribute
their defeat to the rules of the electoral game. In addition, Anderson and his
co-authors argue that the level of satisfaction with democracy in both win-
ners and losers is conditioned by electoral systems. On the one hand, winners
in majoritarian systems are more satisfied than winners under consensus sys-
tems because the winner-take-all nature of majoritarian systems enables their
preferred party to monopolize government resources such as budgetary dis-
cretion and ministerial positions. Losers in majoritarian systems, on the other
hand, are completely deprived of the political pie, while losers’ interests in
consensus systems may still be somewhat represented (Lijphart 1999). It fol-
lows logically that the gap between winners and losers is larger in majoritar-
ian than in consensus systems.
Formally, if we use ‘W’ and ‘L’ respectively to represent winners’ and losers’
satisfaction with democracy, and denote majoritarian systems with subscripts
‘m’ and consensus systems with ‘c’, we can formulate the winner–loser thesis
with following inequalities:

Wm > L m ; Wc > L c (1)

Wm > Wc ; L c > L m (2)

Wm – L m > Wc – L c (3)

This winner–loser gap thesis greatly advances our understanding of how

political institutions mitigate or magnify the effect of winning elections on
popular support for democracy; unfortunately, its analytical power is limited
because it does not account for the role of voting history on individual politi-
cal attitudes toward the political system. Specifically, the conventional wis-
dom defines the electoral winners and losers based on the result of a single
election, and it treats the interaction between winners and losers as a one-
shot game. However, elections in democratic systems should be considered as

235
Eric Chang, Yun-han Chu, and Wen-chin Wu

a set of repeated games in which political actors compete with each other
under certain levels of uncertainty of winning in each round.
Extending from a one-shot game into a repeated-game framework is theo-
retically consequential. In classical game theory, it is well known that new
equilibriums will emerge if the game is played repeatedly because players can
condition their actions in the current round with information available to
them in the previous round (Fudenberg and Tirole 1991). Take the famous
Prisoner’s Dilemma (PD), for example. In a one-shot PD game, each player’s
dominant strategy is to defect, despite the fact that the optimal outcome can
be achieved if players cooperate with each other. As a result, all players are
stuck in a unique Nash equilibrium in which everyone defects and suffers
from the worst outcome. On the contrary, if the PD game is repeatedly played
for an infinite number of times, players can develop some strategies to induce
others’ cooperation instead of defection (Axelrod 1984).3 The important les-
son here is that a player’s action in the current round of a repeated game is
determined by their actions in the past as well as their expectation for the
future.

12.3  A Dynamic Winner–Loser Framework

Viewing elections in democratic transitions via the lens of repeated games

generates important implications. Elections, when held repeatedly, serve as
a self-reinforcing institution that structures the strategic interactions between
winners and losers by shaping each actor’s expectations about the other
actor’s political behaviours and democratic attitudes. In more concrete
terms, Przeworski (1986, 1991) argues that democratic rules create uncer-
tainty of deciding winners and losers. By this logic, democracy can be sus-
tained when relevant political forces are willing to constrain themselves
from subverting the democratic regime and accept their current defeat with
the expectation that they could win future elections. Additionally, Fearon
(2006) argues that holding elections routinely sustains the democratic regime
in two ways. On the one hand, it reveals how fellow citizens evaluate the
performance of the incumbent to individual voters. On the other hand, it
regularly coordinates opposition forces that are regarded as a potential threat
of rebellion to prevent the incumbent from being a dictator. Lindberg (2009)
takes this logic an additional step by arguing that repeated elections, in and
of themselves, create incentives to facilitate democratic behaviour by increas-
ing the cost of repression between both the incumbent and the opposition.
He suggests that repeated elections diffuse democratic qualities to society in
a self-fulfilling manner because they help citizens form an expectation that
other individuals will follow the same democratic rules. Meanwhile, other

236
 Consenting to Lose or Expecting to Win?

related democratic institutions, such as the judicial system and the mass
media, would not only protect and improve the civil rights and liberties of
the people but would also reinforce individuals’ expectation of others’ coop-
erative behaviours. In this sense, Lindberg emphasizes the role of repeated
elections in democratization and extends Rustow’s (1970) thesis that demo-
cratic behaviour contributes to democratic values.
Parallel to Lindberg, this chapter highlights a particular mechanism
through which repeated elections contribute to democratic consolidation:
the possibility of electoral alternation. We argue that when repeated elections
are institutionalized in nascent democracies, the possibility of electoral alter-
nation provides both critical information and necessary enforcement mecha-
nisms to both current winners and losers. On the one hand, it reminds the
current winners that the electoral victory is merely temporary and that they
are likely to be sitting on the other side of the table after the next election. On
the other hand, it also helps lessen the pain of electoral loss by encouraging
current losers to look to the future. In other words, the possibility of electoral
alternation forges an expectation that repeated elections can alternate not
only the composition of the government but also the status of winners and
losers among individuals. As a result, it becomes common knowledge for all
actors that the winner–loser status is not permanently fixed but transforma-
ble in different rounds of elections. Under such circumstances, electoral alter-
nation creates uncertainty about the future, thereby moderating the current
political attitudes of both types of actors and consequently contributing to
more democratic and optimal outcomes.
Of course, there are myriad ways that an individual forms her expectations.
An extremely pessimistic person, for instance, might think that she would
never stand a chance of winning. While a comprehensive study of how an
individual forms her expectation is beyond the scope of this chapter, we
argue that one common denominator among various individuals is their past
experience. Simply put, we argue that an individual is more likely to believe
that she can be a winner if she has won before. Importantly, we hypothesize
that those who have the experience of being winners are more likely to devel-
op a favourable attitude towards democracy than those who have never expe-
rienced winning before.
Specifically, if we focus on the past two consecutive elections, we could
classify voters into four categories on the basis of their voting history: previ-
ous losers who win the current election (LW), previous winners who lose the
current election (WL), two-time winners (WW), and two-time losers (LL). The
main hypothesis in this chapter, then, is that:

WW = WL = LW > LL (4)

237
Eric Chang, Yun-han Chu, and Wen-chin Wu

Astute readers would notice the conceptual difference between our dynamic
framework and the conventional winner/loser thesis. Specifically, we argue
that it is the experience of being winners, not the current winner/loser status per
se, that enhances voters’ satisfaction with democracy. In other words, the cur-
rent losers may not be necessarily less satisfied than current winners, because
some of them may be previous winners and still hold some support for the
democratic system (i.e. the WL group). Only repeated losers, the LL group,
will be less satisfied with the system than other groups.
One might reasonably suspect that the two-time winners should exhibit
higher levels of satisfaction with democracy than one-time winners. On the
other hand, in the case where electoral alteration occurs, the WW group con-
sists of individuals who switch from the previous incumbent party to the
current winning party. To the best of our knowledge, no one has directly
investigated the changes of democratic attitudes of citizens who shift their
votes between consecutive elections; yet a recent study on vote shifting by
Chang (2010) may provide some useful clues. According to Chang (2010: 77),
vote shifting is ‘a form of rational voting consisting of critical evaluations of
government performance and careful weighing of policy issues’. Importantly,
Chang shows that among previous incumbent supporters in East Asian
democracies, those who have drifted apart from the incumbent on important
policy issues and those who are dissatisfied with government performance
are more likely to withdraw their electoral support from the previous incum-
bent and vote for the current winning party. In the case of government alter-
nation, two-time winners are exactly those vote shifters who supported the
previous incumbent yet chose to change their electoral allegiance to the new
winner. Extending this rationale, it seems reasonable to assume that the sat-
isfaction with democracy of two-time winners during electoral alternation
may be decreased due to their disappointing experience with the past govern-
ment. In sum, given the lack of a clear theoretical foundation, we consider
the difference between two-time winners and one-time winners an empirical
question that can best be answered through data analysis.
Before proceeding, it is worth noting that an emerging literature has also
started to re-examine the winner–loser thesis from a dynamic perspective. In
their subsequent landmark study, Anderson et al. (2005, Chapter 4) investi-
gate how the winner–loser gap of democratic satisfaction evolves inter-tem-
porally in Britain, Germany, and Spain, and they report three major findings.
First, they demonstrate that when a citizen’s winner–loser status changes
because of government alternation, her democratic satisfaction would change
as well. As a result, electoral alternation also yields a substantial shift in the
winner–loser gap. Second, they further demonstrate that the winner–loser
gap persists over time. In other words, losers’ dissatisfaction with democracy
is not just a transient disheartenment but a discontent with the current

238
 Consenting to Lose or Expecting to Win?

government over the entire electoral cycle. Finally, Anderson et al. show that
repeated defeat in the electoral arena gradually undermines citizens’ demo-
cratic satisfaction due to the accumulation of post-election frustration. Spe-
cifically, Anderson et al. (2005: 63) find that in Britain and Germany, losers’
democratic satisfaction starts to decline after being defeated in two elections
in a row.
The finding of Anderson et al. provides a parallel and important insight to
our study. Specifically, their study emphasizes the importance of being able to
change the winner–loser status over time. Because of this possibility, current
losers may accept their defeat with an expectation that they will win in the
future. Importantly, it is exactly this possibility of being able to win in the
future that constitutes what Anderson et al. call ‘the losers’ consent’ that legit-
imizes and continues the practices of democracy. On the other hand, howev-
er, our study differs from Anderson et al.’s in at least three important aspects.
First, at the operationalization level, Anderson et al. define the winner–
loser status based on whether a citizen supports a given party before an elec-
tion, and then they make their inference of dynamic effects with a static
measure. With this operationalization, Anderson et al. make an assumption
that voters never change their voting pattern in a sequence of elections. Then
they aggregate the level of satisfaction with democracy among the same
group of voters and show how their frustration of losing accumulates and
hence leads to their declining democratic satisfaction. As we shall elaborate
below, we relax this strict assumption on voting behaviour and employ a
more realistic measurement to pin down the changes of winner–loser status
during two consecutive elections. We believe that our operationalization of
dynamic winner–loser status can better capture the effects of repeated losing,
since it can differentiate repeated losers from the current losers who are in
fact previous winners but shift their votes to the losing parties in the current
election. Meanwhile, our individual-level research design can further help us
more accurately estimate the effect of changes in the winner–loser status on
one’s democratic attitudes.
Second, at the empirical test level, our study provides a more direct way to
test the dynamic winner–loser gap. Specifically, while Anderson et al. focus
on how repeated loss in elections gradually deteriorates losers’ satisfaction
with democracy over time, in this chapter we focus explicitly on the gap of
democratic satisfaction between those constant losers and those who have
become winners in recent elections. Similarly, while Anderson et al. force-
fully demonstrate a consistent and persistent winner–loser gap throughout
the time period they examine, their study does not rule out the possibility
that winners’ democratic satisfaction may also simultaneously decline with
losers’. In fact, Anderson et al. (2005: 59–60) point out that the ‘euphoria’ of
unifying with their former communist neighbour pleases both winners and

239
Eric Chang, Yun-han Chu, and Wen-chin Wu

losers in West Germany during the late 1980s, but the post-unification ‘hang-
over’ knocks down the democratic satisfaction of both winners and losers in
the early 1990s. They also provide visual evidence that winners’ satisfaction
with democracy also fluctuates over time (see their figures). Hence, one
remains unclear regarding whether the winner–loser gap widens or shrinks
over time.
The final distinction between this chapter and Anderson et al. is the sam-
pling strategy. Since Anderson et al. are mainly interested in the consequenc-
es of repeated losing, their analysis covers both developing and established
democracies in which a certain party or party coalition dominates elections
over a long period of time. On the other hand, we derive our theoretical focus
from Huntington’s two-turnover test. Therefore, we limit our sample within
nascent democracies in which the latest government turnovers reshuffle the
identity of most previous winner and losers.
In addition to Anderson et al. (2005), a recent study by Curini et al. (2012)
has also begun to examine how inter-temporal changes of winner–loser sta-
tus affect individuals’ political attitudes in two consecutive elections. They
show that voters are more satisfied with democracy if they perceive that the
government is closer to their ideal point along the left–right ideological
dimension. On the one hand, they argue that winning for the first time dra-
matically enhances citizens’ satisfaction with democracy. On the other hand,
however, they assert that the marginal increase in satisfaction when winning
again decreases unless the ideological distance between winners and the gov-
ernment decreases.
While we address a similar question about the inter-temporal change of
winner–loser status, we adopt a considerably different perspective from that
of Curini et al. (2012). First, Curini et al. include 31 countries covered by the
CSES 2 project, but we only focus on cases of government alternation in
power in our analysis. This distinction results in dramatic contextual differ-
ence when specifying the dynamic winner–loser status because the occur-
rence (absence) of government alternation makes most previous winners
(losers) become losers (winners) in the current election. Take the US presiden-
tial elections, for example. If we formulate the dynamic winner–loser status
for the 2000 and 2004 presidential elections, the LL group is mainly com-
posed of Democrats. In contrast, if we focus on the 2004 and 2008 presiden-
tial elections, the LL group will be mainly composed of switch voters who
voted for Kerry in 2004 but McCain in 2008. Obviously, these two groups can
have quite different political attitudes and opinions (even though they both
share the same LL label). Without making a conceptual distinction between
the presence and the absence of electoral alternation, we would blindly and
mistakenly pool these two very different groups into the same category,
which would lead to erroneous inferences.

240
 Consenting to Lose or Expecting to Win?

Second, our substantive results and conclusion differ greatly from Curini
et al.’s. They argue that two-time winners (WW) should be the most satisfied
group, followed by previous losers who win the current election (LW), previ-
ous winners who lose the current election (WL), and finally, two-time losers
(LL). By contrast, we find that the level of satisfaction with democracy is sta-
tistically indistinguishable among those who win at least once. Moreover, our
results indicate that the levels of democratic satisfaction of those who experi-
ence winning at least once are significantly higher than those who never
have the chance to celebrate electoral victories. While Curini et al. argue for
the primacy of winning the current election (especially for those who lost the
previous time), we emphasize the importance of winning at least once,
regardless of which election. We applaud their efforts of exploring the mar-
ginal effect of inter-temporal change of winner–loser status, yet we conclude
that their methods are likely to be ineffective and lead to spurious results.
More specifically, to effectively probe the marginal effect of the inter-­temporal
change of winner–loser status, one should trace the inter-temporal change of
the dependent variable; unfortunately, however, this critical information is
not available in the CSES 2 data.4
Finally, while Curini et al. usefully investigate the impact of ideological
proximity between citizens and government, we maintain that it is overly
simplistic to use one single left–right dimension to capture the ideological
proximity across all 31 countries comprising well established and newly
emerging democracies. Instead of ideological proximity, this chapter focuses
on government alternation in young democracies. Since the occurrence of
government alternation sufficiently implies an increase (decrease) of the ide-
ological proximity between current winners (losers) and the new government
along the major political cleavage(s) in a given country, the issue of ideologi-
cal proximity is then embedded within our analysis.

12.4  Empirical Analysis

12.4.1  Data
To investigate inter-temporal change of winner–loser status on citizens’
political attitudes toward democracy, this chapter uses the data compiled by
CSES 2. The CSES 2 data set is ideal for our study in that it systematically
tracks citizens’ voting history in two consecutive national elections. While
the CSES 2 project conducts comparative surveys of citizens’ attitudes
toward politics and democracy in 38 countries, we narrow our analytical
focus to young third-wave democracies where changes of the winner–loser
status caused by government alternation matter more to democratic s­ urvival

241
Eric Chang, Yun-han Chu, and Wen-chin Wu

and consolidation. As a result, our sample consists of respondents surveyed

in the following six countries: Brazil, Bulgaria, Chile, Hungary, Poland, and
Romania.

12.4.2  The Dependent Variables

Our main focus, citizens’ support for democracy, is a multidimensional con-
cept. At the broader level, support for democracy can be regarded as a subset
of political support. According to Easton (1965, 1975), members of the politi-
cal system express their support in a response to how their demands (inputs)
are met by the resultant policies (outputs), and the sustenance of a given
system depends on the balance between inputs and outputs of the system.
Accordingly, citizens’ political support is a function of their evaluation of the
short-term government performance and long-term support for the political
system. Furthermore, Easton (1965) distinguishes between support for the
community, the regime, and the authorities, while other scholars expand the
classification into a fivefold framework: support for the political community;
support for regime principles or democracy as an ideal form of government;
evaluations of the regime’s performance; support for regime institutions; and
support for political actors (Norris 1999b). In this chapter, we focus on two
variables as the proxies of mass political support for democracy: (1) citizens’
satisfaction with democratic practices; and (2) their support for democratic
ideals in favour of other regime type choices. Conceptually, the former inves-
tigates how citizens evaluate the democratic performance in practice, where-
as the latter taps into how much they prioritize democratic principles. In
Easton’s terms, satisfaction with democratic practices involves specific sup-
port for the performance of democratic regimes, while support for democratic
principles refers to the diffuse support for the overall democratic system.

12.4.2.1  SATISFACTION WITH DEMOCRATIC PRACTICES

Citizens’ satisfaction with democratic practices can be regarded as the mass
support of democracy after they evaluate the performance of democratic
regimes (Kornberg and Clarke 1994; Klingemann 1999). The conventional
wisdom suggests that electoral winners may be more satisfied with democ-
racy because they perceive the political system to be more responsive to
them, whereas the electoral losers may feel deprived due to the absence of
their favoured parties in the government. To effectively measure how
respondents evaluate the performance (output) of democratic regimes,
CSES 2 asks respondents: ‘On the whole, are you very satisfied, fairly satis-
fied, not very satisfied, or not at all satisfied with the way democracy works
in your country?’ Table 12.1 summarizes the distribution of respondents’

242
 Consenting to Lose or Expecting to Win?

Table 12.1.  Satisfaction with democracy and support for democracy (%)

Satisfaction with Not satisfied at all Not very Fairly satisfied Very satisfied Total
democracy satisfied
14.84 33.58 41.44 10.14 100(N=17,162)
Support for Strongly disagree Disagree Agree Strongly agree Total
democracy 3.49 8.68 44.24 43.59 100(N=16,918)

Note: Satisfaction with democracy is measured by the following question: ‘On the whole, are you very satisfied, fairly
satisfied, not very satisfied, or not at all satisfied with the way democracy works in your country?’ Support for democ-
racy is measured by the following question: ‘Do you agree strongly, agree, disagree, or disagree strongly with this
statement that “Democracy may have problems but it’s better than any other form of government”?’

answers to this question. As we can see, 48.42 per cent of the respondents
are dissatisfied with the democracy practices in their countries, while the
other 51.58 per cent are satisfied. Thus, we recode this four-category varia-
ble into a binary one.

12.4.2.2  SUPPORT FOR DEMOCRATIC PRINCIPLES

While satisfaction with democratic practices may fluctuate with different
institutional contexts as well as short-term events such as political scandals
or economic shocks (Linde and Ekman 2003), citizens’ support for demo-
cratic principles tends to be more stable over time and hence have greater
implications for democratic consolidation. Thus, we take into account this
important dimension of citizens’ democratic support with respect to the
inter-temporal changes of their winner–loser status. In short, it is democratic
rules that determine who winners and losers are in a political system. Citizens
are likely to doubt the legitimacy of democracy and may turn to other alter-
natives forms of political regimes if they consistently find themselves losing
the electoral game (Przeworski 1986). CSES 2 captures the concept of support
for democratic principles by asking respondents: ‘Do you agree strongly,
agree, disagree, or disagree strongly with this statement that “Democracy
may have problems but it’s better than any other form of government”?’
Table 12.1 reports the survey results. Although most respondents are dissatis-
fied with democratic practices in their countries, there is a supermajority of
respondents who still perceive democracy to be better than other regime
types. Since respondents may just pay lip service by saying that democracy is
better (Inglehart 2003; Bratton 2004), we take a stricter definition of demo-
cratic support by only considering those who ‘strongly’ favour democracy
over other regime types. As such, we focus on democratic ‘true believers’ as
opposed to others who show either tepid support for or even disapproval of
democracy in our model specification.
Before proceeding, we would like to emphasize that the correlation between
two newly created binary variables, satisfaction with democracy and support

243
Eric Chang, Yun-han Chu, and Wen-chin Wu

for democracy, is as small as 0.18. This distinction not only reinforces the
notion that democratic support is a multidimensional concept but also pro-
vides us more opportunities to investigate how the dynamic winner–loser
status influences democratic attitudes among individuals.

12.4.3  The Independent Variables

12.4.3.1  DYNAMIC WINNER–LOSER STATUS
Our key independent variable is individual voter’s winner–loser status in two
consecutive national elections. Depending upon whether or not an individu-
al voted for the ruling party or winning candidate in the past two national
elections, we document the inter-temporal change of voters’ winner–loser
status by differentiating respondents into four groups: previous losers who
win the current election (LW), previous winners who lose the current election
(WL), two-time winners (WW), and two-time losers (LL).5 In our model,
we use three dummy variables to represent citizens’ winner–loser status, and
we use the LL group as the baseline category. According to our hypothesis, we
expect that the coefficients of all three group dummies should be signifi-
cantly larger than that of the baseline LL group.
Table 12.2 reports the distribution of winner–loser status across all six
young democracies experiencing government turnover between two consec-
utive national elections. As shown in Table 12.2, the LW and WL groups
jointly account for 56.54 per cent of respondents, while the LL and WW
groups account for 12.02 per cent and 31.44 per cent, respectively. This result
suggests that the majority of voters consistently vote for the same camp of
parties in both elections and experience the frustration of losing in turn, but
at the same time about 43.46 per cent of voters switch their votes in the sec-
ond election.
We next incorporate other individual-level determinants of support for
democracy suggested in the literature as control variables in our model. In
particular, our control variables include: voters’ strength of partisanship,
evaluation of government performance, and perception of corruption.

Table 12.2.  Distribution of dynamic winner–loser status (%)

Current election

Loser Winner

Previous election Loser 12.02 28.05

Winner 28.49 31.44

Note: N = 5566. Voters are coded as losers if their voted candidates or parties
are excluded from the post-election governments.

244
 Consenting to Lose or Expecting to Win?

12.4.3.2  STRENGTH OF PARTISANSHIP

While the Michigan school contends that party identification significantly
influences people’s voting behaviour (Campbell et al. 1960), many studies
further demonstrate that the strength of partisanship may shape individuals’
political attitudes (Duch et al. 2000). Bartels (2002) reports that American
citizens express a partisan bias while evaluating the economic performance of
the president. More specifically, he shows that strong Republicans tend to
give Bush higher evaluations than do strong Democrats. In the context of this
chapter, a voter with stronger partisanship may express higher satisfaction
(dissatisfaction) with democracy if his party is included (excluded) in the
government, since he may expect his affiliated party to be more responsive.
To partial out this potentially confounding factor, we include the variable
measuring respondents’ strength of partisanship (Partisan).6

12.4.3.3  EVALUATION OF GOVERNMENT PERFORMANCE

Since we conceptualize democratic support as a function of individual per-
ceived responsiveness of the political system, it follows logically that govern-
ment performance, as other scholars have argued (Mattes and Bratton 2007),
may also affect citizens’ attitude toward their democratic regime. As force-
fully summarized by Evans and Whitefield (1995), people support demo­
cracy because it ‘works’. Essentially, this school of scholars regards political
legitimacy as a materialistic form of specific support that is highly contin-
gent on assessments of economic performance by rational voters. Thus, in
our model we control for respondents’ evaluation of government perfor-
mance (Performance).7

12.4.3.4  PERCEPTION OF CORRUPTION

Some studies have demonstrated that corruption may affect citizens’ atti-
tudes toward their political regimes. Simply put, corruption is destructive as
it violates the fundamental principle of democracy. As Dahl (1971) persua-
sively notes in his seminal work, the signature characteristic of a democracy
is the prompt responsiveness of the government to the equally weighed
preferences of its citizens. However, the notion of equality and fairness of
­citizens under democracy is severely undermined by corruption since gov-
ernmental services become available only for those who have paid. Powell
(2004: 99) concurs with this argument by contending that ‘simple corrup-
tion can obstruct policy implementation’. He also concludes that a higher
level of corruption implies less responsiveness of the political system. Other
scholars have documented the corrosive effect of corruption on citizens’
trust in political institutions (Chang and Chu 2006) and support for the
political system (Anderson and Tverdova 2003). Accordingly, we include

245
Eric Chang, Yun-han Chu, and Wen-chin Wu

respondents’ perceptions of corruption in their country (Corrupt) to control

for the effects of perceived corruption on democratic attitudes.8

12.4.4  Model Specification

Since the CSES 2 data used in this chapter is pooled from six countries, we
employ random intercept multilevel analysis in which each respondent is
nested within one of the six countries. Using random intercept multilevel
analysis enables us to estimate the fixed and random effects in the same model
simultaneously so as to more effectively deal with the issue of heterogeneity
among the countries. (Steenbergen and Jones 2002). Formally, our basic
­random intercept two-level model includes the following two equations:

Yij = β0 j +  β1 j LW + β2 jWL + β3 jWW + β4 j Partisanij

+ β5 j Performij + β6 j Corrupt ij + εij (5)

β0 j = γ 00 + δ0 j (6)

where subscripts i ∈ {1,2, ,N } and j ∈ {1,2, , J } represent units in the indi-

vidual and the country levels, respectively.9
Combining equations (5) and (6), we get the following random intercept
model:

Yij = γ 00 + β1 j LW + β2 jWL + β3 jWW + β4 j Partisanij

+ β5 j Performij + β6 j Corrupt ij + δ0 j + εij (7)

12.4.5  Empirical Results

To highlight how the experience of winning elections affects individuals’
democratic attitudes, we first estimate its effects on citizens’ satisfaction with
democratic practices. Recall that we use three dummy variables to represent
the dynamic winner–loser status with the two-time losers as the baseline cat-
egory. As the results in Model 1 of Table 12.3 demonstrate, the coefficients of
all of the three dummy variables are positive and significant, indicating that
the two-time losers have lower levels of democratic satisfaction than any of
the other three groups. We repeat the same exercise by using support for
democratic principles as the dependent variable in Model 2, and we obtain
similar results.10
In Model 1 and Model 2, we observe that the experience of winning makes
voters more satisfied with and supportive of the democratic regime. A pertinent

246
 Consenting to Lose or Expecting to Win?

Table 12.3.  Estimated effects of political support

Variables Models

(1) (2) (3) (4) (5) (6)

LW 0.672*** 0.588***
[0.212] [0.199]
WL 1.050*** 0.514***
[0.194] [0.188]
WW 1.034*** 0.681***
[0.185] [0.177]
Wint1 0.027
[0.132]
Wint0 0.238
[0.160]
Winner 0.284 –0.022
[0.177] [0.141]
Young –1.619*** –0.980***
[0.406] [0.357]
Winner × 0.596** 0.603***
Young
[0.245] [0.216]
Partisan –0.003 0.373*** –0.089 0.107 0.155** 0.425***
[0.108] [0.101] [0.130] [0.144] [0.074] [0.067]
Perform 1.166*** 0.398*** 1.215*** 0.882*** 1.094*** 0.223***
[0.088] [0.074] [0.108] [0.111] [0.055] [0.046]
Corrupt –0.176** –0.078 –0.096 –0.123 –0.305*** –0.074
[0.071] [0.069] [0.087] [0.094] [0.049] [0.045]
Constant –3.557*** –2.056*** –2.829*** –2.403*** –1.395*** –0.737**
[0.450] [0.386] [0.520] [0.532] [0.397] [0.333]
N 2103 2005 1392 1211 4802 4722

Note: The dependent variable is democratic satisfaction, except for Model 2 and Model 6, which use democratic sup-
port as the dependent variable. Standard errors in brackets. * p < 0.1, ** p < 0.05, *** p < 0.01. All tests are two-detailed.

question then is whether the two-time winners have more favourable demo-
cratic attitudes than those who win just one time. Eyeballing the coefficients of
three dummy variables in both Model 1 and Model 2 suggests that they are
roughly equivalent. Nevertheless, we systematically examine the group differ-
ences in additional models. Therefore, we create two dummy variables, Wint0
and Wint1, to denote winners in the previous and the current elections, respec-
tively. Model 3 analyses whether losing the current election may result in lower
satisfaction with democracy among previous winners. In other words, we com-
pare the WL group and the WW group and see if the latter shows more faith in
democratic systems. As we can see, the coefficient of Wint1 is statistically insig-
nificant, suggesting that there is no observable difference between the WL and
WW groups in terms of satisfaction with democracy.
Similarly, Model 4 investigates whether the experience of being a loser in
the previous election may influence individual satisfaction with democracy
among current winners. That is, we compare the democratic attitudes between

247
Eric Chang, Yun-han Chu, and Wen-chin Wu

the LW group and the WW group. Again, we find no difference in satisfaction

with democracy between these two groups. As with Model 2, we also use sup-
port for democratic principles as the alternative dependent variable and we
obtain the same conclusion (results not shown in Table 12.3).
Taken together, our results show that LW = WL = WW > LL in terms of citi-
zens’ satisfaction with and support for democracy. Most importantly, our
models demonstrate that the experience of winning at least once matters in
shaping individuals’ support for democratic practices and principles.
Finally, while this chapter proposes a dynamic winner–loser framework in
the context of nascent democracies where peaceful government alternations
are crucial for democratic consolidation, an interesting point worth further
exploring is whether the same mechanism can be found in established
democracies.
Specifically, we investigate whether the gap of democratic attitudes between
winners and losers is smaller in developed or in developing democracies. We
first expand our sample in the CSES 2 data set and include developed democ-
racies that also experience government turnover. This empirical decision adds
Denmark, Finland, Israel, Italy, the Netherlands, and Spain to our analysis. In
addition, we create a dummy variable, Young, to represent the young democra-
cies that are probed in our initial analysis. To ease our analysis, we further
create another dummy variable, Winner, by grouping those respondents who
have won at least once in the two elections (i.e. the WL, WW, and LW groups).11
Most importantly, we expand our initial random intercept two-level model to
a random coefficient model that estimates the influence of the country-level
variable (i.e. Young).12 Formally, our model includes three equations:

Yij = β0 j + β1 jWinnerij + β2 j Partisanij + β3 j Performij + β4 j Corrupt ij + εij (8)

β0 j = γ 00 + γ 01Young + δ0 j (9)

β1 j = γ10 + γ11Young + δ1 j (10)

where εij, δ0j, and δ1j are independently normally distributed disturbance
terms.
Combining equations (8) to (10), we get the following random coefficient
model:

Yij = γ 00 + γ 01Young + ( γ10 + γ11Young )Winnerij + β2 j Partisanij

+ β3 j Performij + β4 j Corrupt ij + εij + δ0 j + δ1 jWinnerij (11)

The estimation results of the random coefficient model specified in equation

(11) are reported in Model 5 and Model 6. Note that the coefficients involving

248
 Consenting to Lose or Expecting to Win?

the dummy variable Winner are associated with the winner–loser gap of politi-
cal attitudes, so we should expect that the interaction term Winner × Young is
significant and positive. Using democratic satisfaction as the dependent varia-
ble, Model 5 demonstrates that the coefficient of Winner is insignificant but the
coefficient Winner × Young is significant and positive as expected. These results
not only indicate that the winner–loser gap is larger in young democracies
than in their developed counterparts, but also implies that the winner–loser
gap in developed democracies is indistinguishable under our conceptualiza-
tion of dynamic winner–loser status. The same conclusion can be drawn from
Model 6 when we use democratic support as the dependent variable.
Figure 12.1 further highlights the main findings of our research. Based on
the estimation of Model 5, we simulated the predicted probabilities with their
95 per cent confidence intervals of being satisfied with democracy for win-
ners and losers between young and developed democracies. Specifically,
­Figure 12.1 illustrates that in young democracies constant electoral losers
(i.e. LL) are less likely to be satisfied with democracy than their cohorts expe-
riencing the winner status at least once (i.e. LW, WL, and WW). A similar case
is also found in developed democracies, as illustrated in Figure 12.1, where
the point estimate of predicted probability for losers is lower than that of

.75
Predicted Probability

.5

.25

0
Loser Winner Loser Winner
Young Democracies Developed Democracies

Figure 12.1. Predicted probabilities with 95 per cent confidence intervals of being

satisfied with democracies for losers and winners between young and developed
democracies

249
Eric Chang, Yun-han Chu, and Wen-chin Wu

winners. This result is consistent with Anderson et al. (2005), who show that
satisfaction with democracy is substantially lower among repeated losers in
three European countries (i.e. Britain, Germany, and Spain). However, as
shown in Figure 12.1, the overlapped 95 per cent confidence intervals of pre-
dicted probabilities between winners and losers in developed democracies
imply an indistinguishable winner–loser gap. In other words, being electoral
winners matters more to citizens in young democracies than to those in
developed democracies, because constant losers may express significantly
lower support for the democratic system after losing limited rounds of elec-
tions. By contrast, as democratic practices are repeated and consolidated,
constant losers may either have experienced winner status or have more good
beliefs of becoming winners in the future.

12.5  Discussion and Conclusion

In this chapter, we conduct a multilevel analysis to investigate how the

dynamic winner–loser status affects individual democratic support in six
emerging democracies. We find that individuals who have ‘lost’ two consecu-
tive elections have substantially lower levels of support for democracy and
lower levels of satisfaction with democracy compared to individuals who
have ‘won’ in at least one election. Our results not only complement the con-
ventional static winner and loser thesis, but also show that electoral losers
may not necessarily have lower democratic satisfaction once one takes into
account how their winner–loser status evolved over time. As we have shown,
once an individual has won, whether in the previous or current election, then
her satisfaction with democracy and democratic support will not only be
boosted to a higher level than that of the two-time losers (i.e. LW = WL > LL),
but will also be less susceptible to the incidence of losing in the future (i.e.
WW = WL). Therefore, the experience of winning in at least one election
appears to be essential for democratic consolidation.
In addition, our study also provides other useful insights to the study of
comparative democratization. As many suggest, a young democratic regime
is consolidated only after all relevant political actors are willing to accept the
democratic rules as ‘the only game in town’. Our findings in this chapter sup-
plement Huntington’s two-turnover test from the perspective of ordinary
citizens. Specifically, while previous studies on the two-turnover test focus on
how elites’ competition and compromise result in democratic consolidation,
we demonstrate why electoral alternation also matters for developing favour-
able democratic attitudes among ordinary citizens. We show that constant
losers exhibit lower levels of democratic satisfaction and support, and we
highlight the potential of democratic breakdown from the bottom when a

250
 Consenting to Lose or Expecting to Win?

certain group of citizens are chronically excluded from sharing the political
pie in democratic systems.
Since this chapter emphasizes the importance of the experience of winning
at least once for the consolidation of democracy, we can also extend this logic
to argue for the importance of the expectation of winning in young democra-
cies. While citizens in established democracies have become more accus-
tomed to electoral alternation as the fortunes of parties shift over time, those
in young democracies may still be new to the electoral game and have not
experienced both winning and losing after the founding election. In this
sense, current winners in new democracies will gradually learn how to win
humbly as they form a reasonable expectation that they may not be holding
their office permanently. Meanwhile, current losers will come to realize that
the recent defeat is merely temporary and that sooner or later there will be
another opportunity to determine who holds political power. In order to
forge a consensus about new rules of the game after the democratic transi-
tion, it is imperative that current losers will not be permanently excluded
from political power but will have chances to contest and win subsequent
elections.
Additionally, our analytical focus on electoral alternations is well situated
with a growing body of literature that investigates the role of alternation in
the democratization process. Bratton (2004) first reports the alternation
effects in Africa in which citizens’ political attitudes toward the democracy
decline with the passage of time since the initial alternation. He further sug-
gests that routine elections restore citizens’ faith in democracy. Meanwhile,
Moehler and Lindberg (2009) find that the winners and losers converge in
their perceptions of institutional legitimacy after government alternation
and argue for a positive effect of electoral alternation on democratic consoli-
dation. Finally, Cho and Logan (2010) argue that electoral alternation also
reduces the gap of perceived democratic durability between winners and los-
ers. According to Cho and Logan (2010: 1), electoral alternations not only
widen the pool of individuals who ‘feel that they have a strong stake in the
system’, but also make them more confident in political elites’ commitment
to democracy after alternations.
Nevertheless, our research suffers from the limitation that we cannot exam-
ine the institutional effects on the winner–loser gap of democratic attitudes
as others have done (Anderson and Guillory 1997; Cho and Bratton 2006),
because we lack institutional variations in our sample of six young democra-
cies that experience government alternations during the survey period of
CSES 2. However, this inadequacy does not necessarily invalidate the main
findings in this chapter because our conceptualization of dynamic winner–
loser status can be regarded as a synthesis of electoral history and political
institutions. More specifically, once we take individual voting history into

251
Eric Chang, Yun-han Chu, and Wen-chin Wu

account, it seems not surprising that the winner–loser gap is larger in SMD
systems than in their PR counterparts. Since SMD systems favour large parties
(Duverger 1962; Taagepera and Shugart 1989), supporters of small parties
have a lower probability of becoming winners in SMD systems. By contrast,
in PR systems, it is rather common to see a coalition government formed by
various small parties after the election (Lijphart 1999). Accordingly, the cur-
rent losers from small parties in PR systems, on average, are not as marginal-
ized as in SMD systems because they either had assumed winner status before
or may have more confidence of being winners in the future. Of course, a
more systematic study that integrates the dynamic winner–loser status with
political institutions awaits further research.

Notes

1. Following the standard practice, we define individuals as winners (losers) if their

supported parties or candidates are included in (excluded from) the government
after elections.
2. As we shall elaborate below, the constituents of these four groups in countries are
different in countries with and without governmental turnover.
3. For instances, the ‘grim trigger’ strategy specifies a strategy ‘to cooperate if others
also cooperate, but to defect forever if someone defects’. Since the benefits of
defection once may be outweighed by the costs of being defected forever, players
will not defect but cooperate with others in repeated games. There are other pos-
sible Nash equilibrium strategies, such as ‘tit-for-tat’ strategy. See Fudenberg and
Tirole (1991: 180–1) and Osborne (2004: 430–1) for detailed discussions.
4. Methodologically, we also disagree with Curini et al.’s approach to model the
impact of winning under the framework of dynamic winner–loser status. They use
two dummy variables with an additional interactive term of both dummies to
represent the dynamic winner–loser status and to estimate the marginal impact of
being winners. We believe their approach is problematic because the coefficient
can only differentiate the WW group from the other three groups. In addition,
without calculating the predicted probability for each group it is difficult to inter-
pret the magnitude as well as the significance level of the interactive terms.
5. We focus on presidential elections in countries using presidential systems (Brazil:
the 1998 and the 2002 presidential elections; Chile: the 2000 and the 2006 presi-
dential elections) or semi-presidential systems (Romania: the 2000 and 2004 pres-
idential elections), whereas we examine the parliamentary elections in countries
with parliamentary systems (Hungary: the 1998 and 2002 general elections). The
only exceptions are Bulgaria and Poland: although these two countries use semi-
presidentialism, we have to use results of parliamentary elections for these two
countries since no information is available on the results of presidential elections
(Bulgaria: the 1997 and 2001 general elections; Poland: the 1997 and 2001 general
elections).

252
 Consenting to Lose or Expecting to Win?

6. In the CSES 2 questionnaire, the strength of partisanship is measured by the fol-

lowing question: ‘Do you feel yourself a little closer to one of the political parties
than the others?’ We regard ‘Yes’ as an indicator of the strength of partisanship.
7. In the CSES 2 questionnaire, the government performance is measured by the fol-
lowing question: ‘Now thinking about the performance of the government in
(capital)/president in general, how good or bad a job do you think the govern-
ment/president in (capital) has done over the past years. Has it/he/she done a very
good job, a good job, a bad job, or a very bad job?’.
8. In the CSES 2 questionnaire, the perception of corruption is measured by the fol-
lowing question: ‘How widespread do you think corruption such as bribe taking is
amongst politicians in (country): very widespread, quite widespread, not very
widespread, it hardly happens at all?’.
9. Yij denotes the dependent variable (i.e. support for democracy), β0j and γ00 the
intercepts; εij and δ0j are independently normally distributed disturbance terms.
10. We also relax our definition of democratic support by incorporating those who
support the regime into those with ‘strong support’. The results are similar to
Model 2, with an exception that the coefficient of LW dummy is not statistically
significant but has an expected positive sign.
11. Therefore, the baseline of comparison is those who have never won (LL).
12. The main difference between the random intercept model and random coefficient
model is as follows. The random intercept models assume that the impacts of vari-
ables at higher levels are constant across different clusters, so the differences
between clusters are captured by the intercept. By contrast, the random coefficient
models allow variables at higher levels to have different impacts across clusters, so
the difference between clusters is captured by the intercept as well as the slope (i.e.
coefficient).

253
References

Aarts, K., Blais, A., and Schmitt, H. (eds) (2011). Political Leaders and Democratic Elec-
tions, Oxford: Oxford University Press.
Aarts, C., and Thomassen, J. (2008a). ‘Satisfaction with Democracy: Do Institutions
Matter?’ Electoral Studies, 27: 5–18.
Aarts, K., and Thomassen, J. (2008b). ‘Dutch Voters and the Changing Party Space’.
Acta Politica, 43 (2/3): 203–34.
Abramson, P. R., and Aldrich, J. H. (1982). ‘The Decline of Electoral Participation in
America.’ American Political Science Review, 76: 502–21.
Achen, C. H. (2005). ‘Two-Step Hierarchical Estimation: Beyond Regression Analy-
sis’. Political Analysis, 13 (4): 447–56.
Almond, G. A., and Verba, S. (1963). The Civic Culture: Political Attitudes and Democ-
racy in Five Nations, Princeton, NJ: Princeton University Press.
Alonso, S., and Ruiz-Rufino, R. (2007). ‘Political Representation and Ethnic Conflict
in New Democracies’. European Journal of Political Research, 46 (2): 237–67.
Anderson, C. J. (2000). ‘Economic Voting and Political Context: A Comparative Per-
spective’. Electoral Studies, 19: 151–70.
Anderson, C. J., Blais, A., Bowler, S., Donovan, T., and Listhaug, O. (2005). Losers’
Consent: Elections and Democratic Legitimacy, New York: Oxford University Press.
Anderson, C. J., and Guillory, C. A. (1997). ‘Political Institutions and Satisfaction
with Democracy: A Cross-National Analysis of Consensus and Majoritarian Sys-
tems’. American Political Science Review, 91 (1): 66–81.
Anderson, C. J., and LoTempio, A. J. (2002). ‘Winning, Losing and Political Trust In
America’. British Journal of Political Science, 32 (2): 335–51.
Anderson, C. J., and Tverdova, Y. V. (2001). ‘Winners, Losers, and Attitudes about
Government in Contemporary Democracies’. International Political Science Review,
22 (4): 321–38.
Anderson, C. J., and Tverdova, Y. V. (2003). ‘Corruption, Political Allegiances, and
Attitudes Toward Government in Contemporary Democracies’. American Journal
of Political Science, 47 (1): 91–109.
APSA (1950). The American Political Science Association: Toward a More Responsible
Two-Party System, New York: Rinehart and Company.
Armingeon, K., and Careja, R. (2004). Comparative Data Set for 28 Post-Communist
Countries, 1989–2004, Berne: Institute of Political Science, University of Berne.
Armingeon, K., Leimgruber, P., Beyeler, M., and Menegale, S. (2006). Comparative
Political Data Set, 1960–2004, Berne: Institute of Political Science, University of
Berne.

255
References

Axelrod, R. M. (1984). The Evolution of Cooperation, New York: Basic Books.

Bale, T., and Van Biezen, I. (2007). ‘Political Data in 2006’. European Journal of Politi-
cal Research, 46 (7/8): 853–66.
Balinski, M., and Young, H. P. (2001). Fair Representation: Meeting the Ideal of One
Man, One Vote, Second Edition, Washington, DC: Brookings Institution Press.
Banducci, S. A., Donovan, T., and Karp, J. A. (1999). ‘Proportional Representation
and Attitudes about Politics: Results from New Zealand’. Electoral Studies, 18:
533–55.
Bartels, L. M. (2002). ‘Beyond the Running Tally: Partisan Bias in Political Percep-
tions’. Political Behavior, 24 (2): 117–50.
Birch, S. (2009). Full Participation: A Comparative Study of Compulsory Voting, Man-
chester: University of Manchester Press.
Blais, A. (2006). ‘What Affects Voter Turnout?’ Annual Review of Political Science, 9:
111–25.
Blais A., and Aarts, K. (2006). ‘Electoral Systems and Turnout’. Acta Politica, 41:
180–96.
Blais, A., and Achen, C. H. (2009). ‘Duty, Preference, and Turnout’. Paper presented
at the ECPR General Conference, Potsdam.
Blais, A., and Carty, R. K. (1991). ‘The Psychological Impact of Electoral Laws: Meas-
uring Duverger’s Elusive Factor’. British Journal of Political Science, 21: 79–93.
Blais, A., Gidengil, E., Dobrzynska, A., Nevitte, N., and Nadeau, R. (2003). ‘Does the
Local Candidate Matter? Candidate Effects in the Canadian Election of 2000’. Ca-
nadian Journal of Political Science, 36: 657–64.
Blais, A., and Loewen, P. J. (2009). ‘The Electoral System and Its Effects’. Western Eu-
ropean Politics, 34: 342–56.
Bogaards, M. (2000). ‘The Uneasy Relationship between Empirical and Normative
Types in Consociational Theory’. Journal of Theoretical Politics, 12 (4): 395–423.
Booth, John A., and Seligson, Mitchell A. (2009). The Legitimacy Puzzle in Latin Amer-
ica: Political Support and Democracy in Eight Nations, Cambridge: Cambridge Univer-
sity Press.
Bovens, M., and Wille, A. (2009). Diploma Democracy: On the Tensions between Meri-
tocracy and Democracy, Utrecht/Leiden: NWO.
Bowler, S., Lanoue, D. J., and Savoie, P. (1994). ‘Electoral Systems, Party Competi-
tion, and Strength of Partisan Attachment: Evidence from Three Countries’. Jour-
nal of Politics, 46: 991–1007.
Brambor, T., Clark, W. R., and Golder, M. (2006). ‘Understanding Interaction Models:
Improving Empirical Analyses’. Political Analysis, 14 (1): 63–82.
Bratton, M. (2004). ‘The “Alternation Effect” in Africa’. Journal of Democracy, 15 (4):
147–58.
Brazil Ministry of Finance (2002). Tax System and Administration in Brazil—An Over-
view. Retrieved 17 March 2010, from <https://www.receita.gov.br/Publico/­
estudotributarios/estatisticas/Ingles/ETO08-TaxSystemandAdministration.pdf>.
Brennan, G., and Lomasky, L. E. (eds) (1993). Politics and Process: New Essays in Dem-
ocratic Thought, Cambridge: Cambridge University Press.

256
 References

Brooks, C., Nieuwbeerta, P., and Manza, J. (2006). ‘Cleavage-Based Voting Behavior
in Cross-National Perspective: Evidence from Six Postwar Democracies’. Social Sci-
ence Research, 35: 88–128.
Brothén, M., and Holmberg, S. (2010). Folkets Representanter, Göteborg: Statsveten-
skapliga Institutionen.
Budge, I. (1982). ‘Electoral Volatility: Issue Effects and Basic Change in 23 Post-War
Democracies’. Electoral Studies, 1: 147–68.
Cain, B., Ferejohn, J., and Fiorina, M. (1990). The Personal Vote: Constituency Service
and Electoral Independence, Cambridge, MA: Harvard University Press.
Campbell, A., Converse, P. E., Miller, W. E., and Stokes, D. E. (1960). The American
Voter, New York: Wiley.
Canache, D., Mondak, J. J., et al. (2001). ‘Meaning and Measurement in Cross-­
National Research on Satisfaction with Democracy’. Public Opinion Quarterly, 65:
506–28.
Carey, J. M., and Shugart, M. S. (1995). ‘Incentives to Cultivate a Personal Vote:
A Rank Ordering of Electoral Formulas’. Electoral Studies, 14: 417–40.
Carlin, J. B., Galati, J. C., and Royston, P. (2008). ‘A New Framework for Managing
and Analyzing Multiply Imputed Data in Stata’. Stata Journal, 8 (1): 49–67.
Castles, F. G. (1999). ‘Decentralization and the Post-War Political Economy’. Euro-
pean Journal of Political Research, 36 (1): 27–53.
Chang, E. C. C. (2010). ‘Vote Shifting and Democratic Consolidation in East Asian
Democracies: Evidence from Taiwan and South Korea’. Taiwan Journal of Democ-
racy, 6 (2): 75–99.
Chang, E. C. C., Chu, Y.-H. (2006). ‘Corruption and Trust: Exceptionalism in Asian
Democracies?’ Journal of Politics, 68(2): 259–71.
Cho, W., and Bratton, M. (2006). ‘Electoral Institutions, Partisan Status, and Political
Support in Lesotho’. Electoral Studies, 25 (4): 731–50.
Cho, W., and Logan, C. (2010) ‘Looking Toward the Future: Alternations in Power
and Popular Perspectives on Democratic Durability in Africa’, Afrobarometer
Working Paper No. 110.
Clark T., N, and Lipset, S. M. (2001). The Breakdown of Class Politics: A Debate on Post-
Industrial Stratification, Baltimore: Johns Hopkins University Press.
Clarke, H. D., Dutt, N., et al. (1993). ‘The Political Economy of Attitudes toward
Polity and Society in Western European Democracies’. Journal of Politics, 55 (4):
998–1021.
Clarke, H. D., Sanders, D., Stewart, M. C., and Whiteley, P. (2004). Political Choice in
Britain, Oxford: Oxford University Press.
Clarke, H. D., Sanders, D., Stewart, M. C., and Whiteley, P. (2009). Performance Poli-
tics and the British Voter, Cambridge: Cambridge University Press.
Colin, Hay (2007). Why We Hate Politics. New York: Wiley.
Converse, P. E., and Pierce, R. (1986). Political Representation in France, Cambridge,
MA/London: Belknap Press.
Corrigall-Brown, C. (2011). Patterns of Protest: Trajectories of Participation in Social
Movements, Palo Alto, CA: Stanford University Press.

257
References

Costello, A. B., and Osborne, J. W. (2005). ‘Best Practices in Exploratory Factor Anal-
ysis: Four Recommendations for Getting the Most From Your Analysis’. Practical
Assessment, Research and Evaluation, 10 (7): 1–9.
Cox, G. W. (1990). ‘Centripetal and Centrifugal Incentives in Electoral Systems’.
American Journal of Political Science, 34: 903–35.
Cox, G. W. (1997). Making Votes Count: Strategic Coordination in the World’s Electoral
Systems, Cambridge: Cambridge University Press.
Cox, G. W., and Amorim, O. (1997). ‘Electoral Institutions, Cleavage Structures and
the Number of Parties’. American Journal of Political Science, 41: 149–74.
Crozier, M., Huntington, S., and Watanuki, J. (1975). The Crisis of Democracy, New
York: New York University Press.
Cukierman, A., Miller, G. P., and Neyapti, B. (2002). ‘Bank Reform, Liberalization
and Inflation in Transition Economies—An International Perspective’. Journal of
Monetary Economics, 49 (2): 237–64.
Cukierman, A., Webb, S. B., and Neyapti, B. (1992). ‘Measuring the Independence of
Central Banks and its Impact on Policy Outcomes’. The World Bank Economic
Review, 6 (3): 353–98.
Curini, L., Jou, W., and Memoli, V. (2012). ‘Satisfaction with Democracy and the
Winner/Loser Debate: The Role of Policy Preferences and Past Experience’. British
Journal of Political Science, 42 (2): 241–61.
Dahl, R. A. (1971). Polyarchy. Participation and Opposition, New Haven/London: Yale
University Press.
Dahl, R. A. (1989). Democracy and its Critics, New Haven: Yale University Press.
Dahlberg, S. (2009). Voters’ Perceptions of Party Politics—A Multilevel Approach, Goth-
enburg: Department of Political Science, University of Gothenburg.
Dalton, R. J. (1985). ‘Political Parties and Political Representation: Party Supporters
and Party Elites in Nine Nations’. Comparative Political Studies, 18: 267–99.
Dalton, R. J. (2002). ‘Political Cleavages, Issues, and Electoral Change’, in L. LeDuc,
R. G. Niemi, and P. Norris (eds), Comparing Democracies 2: New Challenges in the
Study of Elections and Voting, Thousand Oaks: Sage.
Dalton, R. J. (2004). Democratic Challenges, Democratic Choices: The Erosion of Political
Support in Advanced Industrial Democracies, Oxford: Oxford University Press.
Dalton, R. J. (2008). ‘The Quantity and the Quality of Party Systems: Party System
Polarization, Its Measurement, and Its Consequences’. Comparative Political Studies,
41: 899–920.
Dalton, R. J. (2009). The Good Citizen: How the Young are Changing American Politics,
Revised Edition, Washington, DC: CQ Press.
Dalton, R. J., and Anderson, C. (eds) (2011). Citizens, Context, and Choice, Oxford:
Oxford University Press.
Dalton, R. J., Farrell, D., and McAllister, I. (2011a). Political Parties and Democratic
Linkage, Oxford: Oxford University Press.
Dalton, R. J., Farrell, D., and McAllister, I. (2011b). ‘The Dynamics of Political Repre-
sentation’, in M. Rosema, B. Denters, and K. Aarts (eds), How Democracy Works:
Political Representation and Policy Congruence in Modern Societies, Amsterdam: Am-
sterdam University Press/Pallas Publications.

258
 References

Dalton, R. J., Flanagan, S. C., and Beck, P. A. (eds) (1984). Electoral Change in Ad-
vanced Industrial Democracies: Realignment or Dealignment?, Princeton, NJ: Princeton
University Press.
Dalton, R. J., McAllister, I., and Wattenberg. M. P. (2002). ‘The Consequences of Par-
tisan Dealignment’, in R. J. Dalton and M. P. Wattenberg (eds), Parties without Par-
tisans: Political Change in Advanced Industrial Democracies, New York: Oxford Uni-
versity Press.
Dalton, R. J., Van Sickle, A., and Weldon, S. (2010). ‘The Individual–Institutional
Nexus of Protest Behavior’. British Journal of Political Science, 40 (1): 51–73.
De Vaus, D. (2002). Analyzing Social Science Data: 50 Key Problems in Data Analysis,
Los Angeles: Sage Publications.
De Winter, L. (2005). ‘Theoretical, Conceptual and Methodological Problems in Ap-
plying Lijphart’s Patterns of Democracy to Autonomous Regions in Europe’. ECPR
Joint Sessions, Granada, 14–19 April.
Diamond, L. (1999). Developing Democracy: Toward Consolidation, Baltimore: The
Johns Hopkins University Press.
Dow, J. K. (2011). ‘Party-System Extremism in Majoritarian and Proportional Elec-
toral Systems’. British Journal of Political Science, 41: 341–61.
Downs, A. (1957). An Economic Theory of Democracy, New York: Harper and Row.
Duch, R. M., Harvey, M., Palmer, D., and Anderson, C. J. (2000). ‘Heterogeneity in
Perceptions of National Economic Conditions’. American Journal of Political Science,
44 (4): 635–52.
Duch, R. M., and Stevenson, R. T. (2008). The Economic Vote: How Political Institutions
and Economic Institutions Condition Election Results, Cambridge: Cambridge Univer-
sity Press.
Duverger, M. (1962). Political Parties, Their Organization and Activity in the Modern
State, Second Edition. 2nd edn, London: Methuen.
Easton, D. (1965). A Systems Analysis of Political Life, New York: John Wiley.
Easton, D. (1975). ‘A Re-Assessment of the Concept of Political Support’. British Jour-
nal of Political Science, 5 (4): 435–57.
Efron, B., and Tibshirani, R. J. (1993). An Introduction to the Bootstrap, London: Chap-
man & Hall.
Elkins, Z., and Sides, J. (2007). ‘Can Institutions Build Unity in Multiethnic States?’
American Political Science Review, 101(4): 693–708.
Esaiasson, P., and Heidar, K. (eds) (2000). Beyond Westminster and Congress, Colum-
bus: Ohio State University Press.
Esaiasson, P., and Holmberg, S. (1996). Representation from Above: Members of Parlia-
ment and Representative Democracy in Sweden, Aldershot: Dartmouth.
Evans, G. (2000). ‘The Continued Significance of Class Voting’. Annual Review of Po-
litical Science, 3: 401–17.
Evans, G., and Andersen, R. (2006). ‘The Political Conditioning of Economic Percep-
tions’. The Journal of Politics, 68 (1): 194–207.
Evans, G., and Whitefield, S. (1995). ‘The Politics and Economics of Democratic
Commitment: Support for Democracy in Transition Societies’. British Journal of
Political Science, 25 (4): 485–514.

259
References

Ezrow, L. (2010). Linking Citizens and Parties, Oxford: Oxford University Press.
Fearon, J. D. (2003). ‘Ethnic and Cultural Diversity by Country’. Journal of Economic
Growth, 8: 195–222.
Fearon, J. D. (2006). Self-Enforcing Democracy, 2006 Annual Meetings of the Ameri-
can Political Science Association, Philadelphia, 1–4 September.
Fiorina, M. F. (1981). Retrospective Voting in American National Elections, New Haven:
Yale University Press.
Fiorina, M. P. (1977). ‘An Outline for a Model of Party Choice’. American Journal of
Political Science, 21: 601–25.
Fisher, S. D., and Hobolt, S. B. (2010). ‘Coalition Government and Electoral Account-
ability’. Electoral Studies, 29: 358–69.
Fiske, S. T., and Taylor, S. E. (1991). Social Cognition, New York: McGraw-Hill.
Flanagan, S. C., and Lee, A.-R. (2003). ‘The New Politics, Culture Wars and the
­Authoritarian–Libertarian Value Change in Advanced Industrial Democracies’.
Comparative Political Studies, 36 (3): 235–70.
Flinders, M. (2005). ‘Majoritarian Democracy in Britain: New Labour and the Consti-
tution’. West European Politics, 28 (1): 61–93.
Franklin, M. N. (1996). ‘Electoral Participation’, in L. LeDuc, R. G. Niemi, and P.
Norris (eds), Comparing Democracies: Elections and Voting in Global Perspective, Bev-
erly Hills CA: Sage.
Franklin, M. N. (2009). ‘Epilogue’, in M. N. Franklin, T. T. Mackie, and H. Valen
(eds), Electoral Change: Responses to Evolving Social and Attitudinal Structures in West-
ern Countries. Revised ECPR ‘Classics in Political Science’ Edition, Colchester: ECPR
Press.
Franklin, M. N. (2010). ‘Cleavage Research: A Critical Appraisal’. West European Poli-
tics, 33: 648–58.
Franklin, M. N., Mackie, T. T., and Valen, H. (1992). Electoral Change: Responses to
Evolving Social and Attitudinal Structures in Western Countries, Cambridge: Cam-
bridge University Press.
Franzese, R. J. Jr. (2005). ‘Empirical Strategies for Various Manifestations of Multi-
level Data’. Political Analysis, 13: 430–46.
Freedom House (2010). <http://www.freedomhouse.org>.
Freitag, M., and Vatter, A. (eds) (2008). Die Demokratien der Deutschen Bundesländer.
Eine Vergleichende Analyse Ihrer Politischen Institutionen, Opladen: Barbara Budrich
(UTB).
Fuchs, D., Guidorossi, G., and Svensson, P. (1995). ‘Support for the Democratic
System’, in H.-D. Klingemann, and D. Fuchs (eds), Citizens and the State, Oxford:
Oxford University Press.
Fuchs, D., and Klingemann, H.-D. (1989). ‘The Left–Right Schema’, in M. K. Jen-
nings, and J. W. van Deth (eds), Continuities in Political Action, Berlin/New York:
Walter de Gruyter.
Fudenberg, D., and Tirole, J. (1991). Game Theory, Cambridge, MA: MIT Press.
Gallagher, M. (1991). ‘Proportionality, Disproportionality and Electoral Systems’.
Electoral Studies, 10 (1): 33–51.

260
 References

Gallagher, M., and Mitchell, P. (2008). The Politics of Electoral Systems, Oxford:
Oxford University Press.
Ganghof, S. (2005). ‘Normative Modelle, Institutionelle Typen und Beobachtbare
Verhaltensmuster. Ein Vorschlag zum Vergleich Parlamentarischer Demokratien’.
Politische Vierteljahresschrift, 46 (3): 406–31.
Gerber, A., Green, D. P., and Larimer, C. W. (2008). ‘Social Pressure and Voter Turn-
out: Evidence from a Large-Scale Field Experiment’. American Political Science
Review, 102 (1): 33–48.
Giger, N., Kissau, K., Lutz, G., and Rosset, J. (2009). ‘Explaining the Variance of Sub-
jective and Substantive Representation’, ECPR General Conference, Potsdam.
Gilley, B. (2009). The Right to Rule—How States Win and Lose Legitimacy, New York:
Columbia University Press.
Golder, M. (2005). ‘Democratic Electoral Systems Around the World, 1946–2000’.
Electoral Studies, 24 (1): 103–21.
Golder, M., and Stramski, J. (2010). ‘Ideological Congruence and Electoral Institu-
tions’. American Journal of Political Science, 54 (1): 90–106.
Grofman, B. (2000). ‘Arend Lijphart and the New Institutionalism’, in M. Crepaz, T.
Koelble, and D. Wilsford (eds), Democracy and Institutions: The Life Work of Arend
Lijphart (43–73). Ann Arbor: University of Michigan Press.
Habermas, J. (1975). Legitimation Crisis, Boston: Beacon Press.
Hadenius, A., and Teorell, J. (2005). ‘Assessing Alternative Indices of Democracy’.
Political Concepts. IPSA Committee on Concepts and Methods Working Paper
Series, 6.
Hakhverdian, A., van Elsas, E., van der Brug, W., and Kuhn, T. (2013). ‘Euroscepti-
cism and Education: A Longitudinal Study of Twelve EU Member States, 1973–
2010’. European Union Politics, 14(4): 522–541.
Hibbing, J. R., and Theiss-Morse, E. (1995). Congress as Public Enemy: Public Attitudes
toward Political Institutions, New York: Cambridge University Press.
Hibbing, J. R., and Theiss-Morse, E. (2002). Stealth Democracy: Americans’ Beliefs about
How Government Should Work, Cambridge: Cambridge University Press.
Holmberg, S. (1974). ‘Riksdagen Representerar Svenska Folket’ Empiriska Studier i Repre-
sentative Demokrati, Lund: Studentlitteratur.
Holmberg, S. (1999). ‘Collective Policy Congruence Compared’, in W. Miller et al.
(eds), Policy Representation in Western Democracies, Oxford: Oxford University Press.
Holmberg, S. (2006). ‘Åsiktsrepresenterad’, in H. Bäck, and M. Gilljam (eds), Valets
Mekanismer, Malmö: Liber.
Holmberg, S. (2011). ‘Dynamic Representation from Above’, in M. Rosema, B. Dent-
ers, and K. Aarts (eds), How Democracy Works: Participation and Representation in
Modern Societies, Amsterdam: Amsterdam University Press.
Honaker, J., King, G., and Blackwell, M. (2007). Amelia II: A Program for Missing Data,
Los Angeles, CA: UCLA.
Horowitz, D. L. (1993). ‘Democracy in Divided Societies’. Journal of Democracy, 4:
18–38.
Hox, J. (2002). Multilevel Analysis: Techniques and Analysis, New Jersey: Lawrence Erl-
baurn Associates.

261
References

Huber, J. D. (2011). ‘Measuring Ethnic Voting: The Political Context and the Politici-
zation of Ethnicity’, Working Paper.
Huber, J. D., and Powell, G. B. (1994). ‘Congruence Between Citizens and Policy
Makers in Two Visions of Liberal Democracy’. World Politics, 46 (3): 291–326.
Huddy, L. (2002). ‘Context and Meaning in Social Identity Theory: A Response to
Oakes’. Political Psychology, 23: 825–38.
Hug, S. (2004). ‘Occurrence and Policy Consequences of Referendums: A Theoretical
Model and Empirical Evidence’. Journal of Theoretical Politics, 16 (3): 321–56.
Hug, S., and Tsebelis, G. (2002). ‘Veto Players and Referendums around the World’.
Journal of Theoretical Politics, 14 (4): 465–516.
Huntington, Samuel P. (1991). The Third Wave: Democratization in the Late Twentieth
Century. Norman: University of Oklahoma Press.
Huntington, S. P. (1993). The Third Wave: Democratization in the Late Twentieth Cen-
tury, Norman: University of Oklahoma Press.
Inglehart, R. (1971). ‘The Silent Revolution in Europe: Intergenerational Change in
Post-Industrial Societies’. The American Political Science Review, 65 (4): 991–1017.
Inglehart, R. (1977). The Silent Revolution: Changing Values and Political Styles Among
Western Publics, Princeton: Princeton University Press.
Inglehart, R. (1997). Modernization and Postmodernization: Cultural, Economic, and Po-
litical Change in 43 Societies, Princeton: Princeton University Press.
Inglehart, R. (2003). ‘How Solid Is Mass Support for Democracy—And How Can We
Measure It?’ PS: Political Science and Politics, 36(1): 51–7.
Inglehart, R., and Rabier, J.-R. (1986). ‘Political Realignment in Advanced Industrial
Society: From Class-Based Politics to Quality-of-Life Politics’. Government and Op-
position, 21: 456–79.
Irwin, G. A., and Van Holsteyn, J. (2008). ‘Scientific Progress, Educated Guesses or
Speculation? On Some Old Predictions with Respect to Electoral Behaviour in the
Netherlands’. Acta Politica, 43 (2/3): 180–202.
Ismayr, W. (ed.) (2004). Die Politischen Systeme Osteuropas, Opladen: Leske und
Buderich.
Jackman, R. (1987). ‘Political Institutions and Voter Turnout in Industrial Democra-
cies’. American Political Science Review, 81: 405–23.
Jung, S. (1996). ‘Lijpharts Demokratietypen und die Direkte Demokratie’. Zeitschrift
für Politikwissenschaft, 6 (3): 623–47.
Jung, S. (2001). Die Logik der Direkten Demokratie, Wiesbaden: Westdeutscher Verlag.
Kaase, M., and Newton, K. (eds) (1995). Beliefs in Government, Oxford: Oxford Uni-
versity Press.
Kaiser, A. (1997). ‘Types of Democracy. From Classical to New Institutionalism’. Jour-
nal of Theoretical Politics, 9 (4): 419–44.
Kaiser, A., Lehnert, M., Miller, B., and Sieberer, U. (2002). ‘The Democratic Quality
of Institutional Regimes: A Conceptual Framework’. Political Studies, 50 (2):
313–31.
Karp, J., and Banducci, S. (2008). ‘Political Efficacy and Participation in Twenty-
Seven Democracies: How Electoral Systems Shape Political Behavior’. British Jour-
nal of Political Science, 38 (2): 311–34.

262
 References

Karp, J. A., and Bowler, S. (2001). ‘Coalition Politics and Satisfaction with Democ-
racy: Explaining New Zealand’s Reaction to Proportional Representation’. European
Journal of Political Research, 40: 57–79.
Karp, J. A., and Brockington, D. (2005). ‘Social Desirability and Response Validity:
A Comparative Analysis of Overreporting Voter Turnout in Five Countries’. Journal
of Politics, 67: 825–40.
Katz, R. S. (1980). A Theory of Parties and Electoral Systems, Baltimore, MD: The Johns
Hopkins University Press.
Katz, R. S. (2003). ‘Political Data in 2002’. European Journal of Political Research, 42
(7/8): 873–9.
Katz, R. S., and Koole, R. (1999). ‘Political Data in 1998’. European Journal of Political
Research, 36 (3/4): 307–15.
Katz, R. S., and Koole, R. (2002). ‘Political Data in 2001’. European Journal of Political
Research, 41 (7/8): 885–96.
Kaufmann, D., Kraay, A., and Mastruzzi, M. (2006). Governance Matters, Vol. V: Aggregate
and Individual Governance Indicators for 1996–2006, Washington, DC: The World Bank.
Keman, H. (2000). ‘Federalism and Policy Performance’, in U. Wachendorfer-
Schmidt (ed.), Federalism and Political Performance. London/New York: Routledge.
Ken, Newton (1999). ‘Mass Media Effects: Mobilization or Media Malaise?’ British
Journal of Political Science 29: 4, 577–99.
Ken, Newton (2006). ‘Political Support: Social Capital, Civil Society and Political and
Economic Performance’. Political Studies, 54: 4, 846–64.
Key, V. O. Jr. (1966). The Responsible Electorate: Rationality in Presidential Voting, 1936–
1960, Harvard, MA: Belknap Press.
Kim, J.-O., and Mueller, C. W. (1978). Factor Analysis: Statistical Methods and Practical
Issues [Sage University Papers Series on Quantitative Applications in the Social Sci-
ences No. 14]. Beverly Hills: Sage Publications.
King, A. (1975). ‘Overload: Problems of Governing in the 1970s’. Political Studies, 23:
284–96.
Kitschelt, H. P. (1986). ‘Political Opportunity Structures and Political Protest: Anti-
Nuclear Movements in Four Democracies’. British Journal of Political Science, 16 (1):
57–85.
Kitschelt, H. P. (1988). ‘Left-Libertarian Parties: Explaining Innovation in Competi-
tive Party Systems’. World Politics, 40: 194–234.
Kitschelt, H. P. (1989). The Logics of Party Formation: Ecological Politics in Belgium and
West Germany, Ithaca, NY: Cornell University Press.
Kitschelt, H. P. (1994). The Transformation of European Social Democracy, Cambridge:
Cambridge University Press.
Kitschelt, H. P. (2010). ‘The Comparative Analysis of Electoral and Partisan Politics:
A Comment on a Special Issue of West European Politics’. West European Politics,
33: 659–72.
Kitschelt, H., and Hellemans, S. (1990). ‘The Left–Right Semantics and the New Poli-
tics Cleavage’. Comparative Political Studies, 23: 210–38.
Kitschelt, H. P., and McGann, A. J. (1995). The Radical Right in Western Europe:
A Comparative Analysis, Ann Arbor: The University of Michigan Press.

263
References

Kittilson, M. C., and Anderson, C. J. (2011). ‘Electoral Supply and Voter Turnout’, in
R. Dalton, and C. Anderson (eds), Citizens, Context and Choice: How Context Shapes
Citizens’ Electoral Choices, Oxford: Oxford University Press.
Klingemann, H.-D. (1999). ‘Mapping Political Support in the 1990s: A Global
­Analysis’, in P. Norris (ed.), Critical Citizens: Global Support for Democratic
­Government, New York: Oxford University Press.
Klingemann, H.-D. (ed.) (2009). The Comparative Study of Electoral Systems, Oxford:
Oxford University Press.
Klingemann, H.-D., and Fuchs, D. (1995). ‘Citizens and the State: A Changing Rela-
tionship?’, in H.-D. Klingemann, and D. Fuchs (eds), Citizens and the State, Oxford:
Oxford University Press.
Klingemann, H.-D., and Fuchs, D. (eds) (1995). Citizens and the State, Oxford: Oxford
University Press.
Klingemann, H.-D., Hofferbert, R. I., and Budge, I. (1994). Parties, Policies, and
­Democracy, Boulder, CO: Westview Press.
Klingemann, H.-D., et al. (2006). Mapping Policy Preferences, Vol. II: Estimates for Par-
ties, Electors, and Governments in Eastern Europe, European Union and OECD 1990–
2003, Oxford: Oxford University Press.
Knutsen, O. (1997). ‘The Partisan and the Value-based Component of Left–Right
Self-placement: A Comparative Study’. International Political Science Review, 18:
191–225.
Knutsen, O. (2004). Social Structure and Party Choice in Western Europe: A Comparative
Longitudinal Study, Houndsmills: Palgrave Macmillan.
Knutsen, O., and Kumlin, S. (2005). ‘Value Orientations and Party Choice’, in J.
Thomassen (ed.), The European Voter: A Comparative Study of Modern Democracies,
Oxford: Oxford University Press.
Koole, R., and Katz, R. S. (1998). ‘Political Data in 1997’. European Journal of Political
Research, 34 (3/4): 329–38.
Koole, R., and Katz, R. S. (2000). ‘Political Data in 1999’. European Journal of Political
Research, 38 (3/4): 303–556.
Koole, R., and Katz, R. S. (2001). ‘Political Data in 2000’. European Journal of Political
Research, 40 (3/4): 223–32.
Korea Ministry of Strategy and Finance (2009). Government Financial Statistics,
­Retrieved 17 March 2010, from <http://english.mosf.go.kr/public/gfs_list.
php?sect=pubs_gfs>.
Korea National Tax Service (2009). Ratio of National and Local Taxes, Retrieved 17
March 2010, from <http://nts.go.kr/eng/resources/resour_31.asp?minfoKey=M
INF7520080211223206#>.
Kornberg, A., and Clarke, H. D. (1992). Citizens and Community: Political Support in a
Representative Democracy, Cambridge: Cambridge University Press.
Kornberg, A., and Clarke, H. D. (1994). ‘Beliefs about Democracy and Satisfaction
with Democratic Government: The Canadian Case’. Political Research Quarterly, 47
(3): 537–63.
Kraatz, S., and Von Steinsdorff, S. (eds) (2002). Parlamente und Systemtransformation
im Postsozialistischen Europa, Opladen: Leske and Budrich.

264
 References

Kreft, I., and De Leeuw, J. (1998). Introducing Multilevel Modeling, London: Sage.
Kriesi, H. (1998). ‘The Transformation of Cleavage Politics: The 1997 Stein Rokkan
Lecture’. European Journal of Political Research, 33: 165–85.
Kriesi, H., Grande, E., Lachat, R., Dolezal, M., Bornschier, S., and Frey, T. (2008). West
European Politics in the Age of Globalization, Cambridge: Cambridge University
Press.
Kurian, G. T. (ed.) (1997). World Encyclopedia of Parliaments and Legislatures, New
York: Congressional Quarterly Press.
Laakso, M., and Taagepera, R. (1979). ‘“Effective” Number of Parties: A Measure with
Application to West Europe’. Comparative Political Studies, 12 (1): 3–27.
Lachat, R. (2008). ‘The Impact of Party Polarization on Ideological Voting’. Electoral
Studies, 27 (4): 687–98.
Lau, R., and Redlawsk, D. (1997). ‘Voting Correctly’. American Political Science Review,
91: 585–98.
Le Duc, L. (2003). The Politics of Direct Democracy: Referendums in Global Perspective,
Peterborough: Broadview Press.
Leighley, J. E. (1995). ‘Attitudes, Opportunities, and Incentives: A Field Essay on
­Political Participation’. Political Research Quarterly, 48 (1): 181–209.
Lijphart, A. (1975). The Politics of Accommodation: Pluralism and Democracy in the
Netherlands, Berkeley: University of California Press.
Lijphart, A. (1984). Democracies, Patterns of Majoritarian and Consensus Government in
Twenty-one Countries, New Haven: Yale University Press.
Lijphart, A. (1994). Electoral Systems and Party Systems: A Study of Twenty-seven Democ-
racies, 1945–1990, Oxford: Oxford University Press.
Lijphart, A. (1998). ‘Consensus and Consensus Democracy: Cultural, Structural,
Functional and Rational-Choice Explanations’. Scandinavian Political Studies, 21
(2): 99–108.
Lijphart, A. (1999). Patterns of Democracy: Government Forms and Performance in
Thirty-Six Countries, New Haven/London: Yale University Press.
Lijphart, A. (2001). ‘Democracy in the 21st Century: Can We be Optimistic?’. Euro-
pean Review, 9: 169–84.
Lijphart, A. (2003). ‘Measurement Validity and Institutional Engineering: Reflections
on Rein Taagepera’s Meta-Study’. Political Studies, 51 (1): 20–5.
Lijphart, A. (2012). Patterns of Democracy Second Edition, New Haven/London: Yale
University Press.
Lindberg, S. I. (2006). ‘The Surprising Significance of African Elections’. Journal of
Democracy, 17 (1): 139–51.
Lindberg, S. I. (2009). ‘Democratization by Elections: A Mixed Record?’ Journal of
Democracy, 20(3): 86–92.
Linde, J., and Ekman, J. (2003). ‘Satisfaction with Democracy: A Note on a Fre-
quently Used Indicator in Comparative Politics’. European Journal of Political Re-
search, 42 (3): 391–408.
Linz, J. J., and Stepan, A. C. (1996a). Problems of Democratic Transition and Consolida-
tion: Southern Europe, South America, and Post-Communist Europe, Baltimore: Johns
Hopkins University Press.

265
References

Linz, J. J., and Stepan, A. C. (1996b). ‘Toward Consolidated Democracies’. Journal of

Democracy, 7(2): 14–33.
Lipset, S. M. (1959). ‘Some Social Requisites of Democracy: Economic Development
and Political Legitimacy’. American Political Science Review, 53 (3): 69–105.
Lipset, S. M. (1966). Political Man, London: Mercury Books.
Lipset, S. M., and Rokkan, S. (1967). ‘Cleavage Structures, Party Systems, and Voter
Alignments: An Introduction’, in S. M. Lipset, and S. Rokkan (eds), Party Systems
and Voter Alignments: Cross-National Perspectives, New York: Free Press.
Listhaug, O., and Wiberg, M. (1995). ‘Confidence in Political and Private Institu-
tions’, in H.-D. Klingemann, and D. Fuchs (eds), Citizens and the State, Oxford:
Oxford University Press.
Llanto, G. M. (2009). ‘Fiscal Decentralization and Local Finance Reforms in the Phil-
ippines’ Philippine Institute for Development Studies, Discussion Paper Series No.
2009–10.
Long Jusko, K., and Shively, W. P. (2005). ‘Applying a Two-Step Strategy to the Anal-
ysis of Cross-National Public Opinion Data’. Political Analysis, 13: 327–44.
Lord, C., and Pollak, J. (2010). ‘Representation and Accountability: Communicating
Tubes?’ West European Politics, 33: 968–88.
Lundell, K. (2011). ‘Accountability and Patterns of Alternation in Pluralitarian, Ma-
joritarian and Consensus Democracies’. Government and Opposition, 46: 145–67.
Lundell, K., and Karvonen, L. (2003). ‘A Comparative Data Set on Political Institu-
tions’. Abo Akademi, Department of Political Science, Finland.
Lupia, A., and McCubbins, M. D. (1998). The Democratic Dilemma: Can Citizens Learn
What They Really Need to Know?, Cambridge: Cambridge University Press.
Lupia, A., McCubbins, M. D., and Popkin, S. (eds) (2000). Elements of Reason: Cogni-
tion, Choice and the Bounds of Rationality, New York: Cambridge University Press.
Mainwaring, S., and Torcal, M. (2006). ‘Party System Institutionalization and Party
System Theory after the Third Wave of Democratization’, in R. S. Katz, and W. J.
Crotty (eds), Handbook of Party Politics, London: Thousand Oaks.
Mainwaring, S., and Zoco, E. (2007). ‘Political Sequences and the Stabilization of
Interparty Competition’. Party Politics, 13: 155–78.
Manin, B., Przeworski, A., and Stokes, S. C. (1999). ‘Elections and Representation’, in
A. Przeworski, S. C. Stokes and B. Manin (eds), Democracy, Accountability, and Repre-
sentation. Cambridge: Cambridge University Press.
Mansbridge, J. (2003). ‘Rethinking Representation’. American Political Science Review,
97: 515–28.
Manza, J., and Brooks, C. (1998). ‘The Gender Gap in US Presidential Elections:
When? Why? Implications?’ American Journal of Sociology, 103: 1235–66.
Massicotte, L., and Blais, A. (1999). ‘Mixed Electoral Systems: A Conceptual and Em-
pirical Survey’. Electoral Studies, 18: 341–66.
Mattes, R., and Bratton, M. (2007). ‘Learning about Democracy in Africa: Awareness,
Performance, and Experience’. American Journal of Political Science, 51 (1):
192–217.
McAllister, I. (2005). ‘Accountability, Representation and Satisfaction with Democ-
racy’. International Journal of Public Opinion Research, 17 (3): 371–79.

266
 References

Merkel, W. (2004). ‘Embedded and Defective Democracies’. Democratization, 11 (5):

33–58.
Miller, W. E., et al. (1999). Policy Representation in Western Democracies, Oxford:
Oxford University Press.
Miller, W. E., and Stokes, D. E. (1963). ‘Constituency Influence in Congress’. Ameri-
can Political Science Review, 57 (1): 45–56.
Mishler, W., and Rose, R. (2001). ‘What are the Origins of Political Trust? Testing
Institutional and Cultural Theories in Post-Communist Societies’. Comparative
­Political Studies, 34 (1): 30–62.
Moehler, D. C., and Lindberg, S. I. (2009). ‘Narrowing the Legitimacy Gap: Turno-
vers as a Cause of Democratic Consolidation’. Journal of Politics, 71 (4): 1448–66.
Morgenstern, S., and Castagnola, A. (2009). ‘Party Nationalization and Institutions’.
The Journal of Politics, 71: 1322–41.
Morgenstern, S., and Swindle, S. M. (2005). ‘Are Politics Local? An Analysis of Voting
Patterns in 23 Democracies’. Comparative Political Studies, 38: 143–70.
Neidhart, L. (1970). Plebiszit und Pluralitäre Demokratie. Eine Analyse der Funktion des
Schweizerischen Gesetzesreferendums, Berne: Francke.
Neto, O. A., and Cox, G. W. (1997). ‘Electoral Institutions, Cleavage Structures, and
the Number of Parties’. American Journal of Political Science, 41: 149–74.
Nieuwbeerta, P., and De Graaf, N. D. (1999). ‘Traditional Class Voting in Twenty
Postwar Societies’, in G. Evans (ed.), The End of Class Politics? Class Voting in Com-
parative Context, Oxford: Oxford University Press.
Norris, P. (ed.) (1999a). Critical Citizens: Global Support for Democratic Government,
New York: Oxford University Press.
Norris, P. (1999b). ‘Introduction: The Growth of Critical Citizens?’, in P. Norris (ed.),
Critical Citizens: Global Support for Democratic Government, New York: Oxford Uni-
versity Press.
Norris, P. (2002). Democratic Phoenix: Reinventing Political Activism, New York: Cam-
bridge University Press.
Norris, P. (2004). Electoral Engineering: Voting Rules and Political Behavior, New York:
Cambridge University Press.
Norris, P. (2011). Democratic Deficit: Critical Citizens Revisited, Cambridge: Cambridge
University Press.
Norton, P., and Olson, D. M. (eds) (2007). ‘Post-Communist and Post-Soviet Legisla-
tures: Beyond Transition’. Special Issue of the Journal of Legislative Studies, 13 (1):
1–196.
Nye, J. S. Jr., Zelikow, P. D., and King, D. C. (eds) (1997). Why People Don’t Trust Gov-
ernment, Cambridge, MA: Harvard University Press.
Oakes, P. (2002). ‘Psychological Groups and Political Psychology: A Response to
Huddy’s “Critical Examination of Social Identity Theory”’. Political Psychology, 23:
809–24.
Ober, J. (2008). Democracy and Knowledge. Innovation and Learning in Classical Athens,
Princeton: Princeton University Press.
Ordeshook, P. C., and Shvetsova, O. V. (1994). ‘Ethnic Heterogeneity, District Magni-
tude, and the Number of Parties’. American Journal of Political Science, 38: 100–23.

267
References

Osborne, M. J. (2004). An Introduction to Game Theory, New York: Oxford University

Press.
Oskarson, M. (2005). ‘Social Structure and Party Choice’, in J. J. A. Thomassen (ed.),
The European Voter: A Comparative Study of Modern Democracies, Oxford: Oxford
University Press.
Parry, G., Moyser, G., and Day, N. (1992). Political Participation and Democracy in Brit-
ain, Cambridge: Cambridge University Press.
Pedersen, M. N. (1979). ‘The Dynamics of European Party Systems: Changing Pat-
terns of Electoral Volatility’. European Journal of Political Research, 7: 1–26.
Peffley, M., and Hurwitz, J. (2010). Justice in America: The Separate Realities of Blacks
and Whites, Cambridge: Cambridge University Press.
Pellikaan, H., De Lange, S., and Van der Meer, T. (2007). ‘Fortuyn’s Legacy: Party
System Change in the Netherlands’. Comparative European Politics, 5: 282–302.
Peter, Mair (2009). ‘Representative versus Responsible Government’. MPIfG Working
Paper 09/8, Cologne: Max Planck Institute for the Study of Societies.
Pierce, R. (1999). ‘Mass-Elite Issue Linkages and the Responsible Party Model of Rep-
resentation’, in W. E. Miller, R. Pierce, J. Thomassen, R. Herrera, S. Holmberg, P.
Esaiasson, and B. Wessels (eds), Policy Representation in Western Democracies,
Oxford: Oxford University Press.
Pitkin, H. F. (1967). The Concept of Representation, Berkeley: University of California
Press.
Pizzorno, A. (1981). ‘Interests and Parties in Pluralism’, in S. Berger (ed.), Organizing
Interests in Western Europe, Cambridge: Cambridge University Press.
Popkin, S. L. (1991). The Reasoning Voter: Communication and Persuasion in Presidential
Campaigns, Chicago: Chicago University Press.
Powell, G. B. (1982). Contemporary Democracies: Participation, Stability and Violence,
Cambridge, MA: Harvard University Press.
Powell, G. B. (2000). Elections as Instruments of Democracy: Majoritarian and Propor-
tional Visions, New Haven, CT: Yale University Press.
Powell, G. B. (2004). ‘The Chain of Responsiveness’. Journal of Democracy, 15 (4):
91–105.
Powell, G. B. (2009). ‘The Ideological Congruence Controversy: The Impact of Alter-
native Measures, Data, and Time Periods on the Effects of Election Rules’. Com-
parative Political Studies, 42 (12): 1475–97.
Powell, G. B., and Whitten, G. D. (1993). ‘A Cross-National Analysis of Economic
Voting: Taking Account of the Political Context’. American Journal of Political Sci-
ence, 37: 391–414.
Przeworski, A. (1986). ‘Some Problems in the Study of the Transition to Democ-
racy’, in G. O’Donnell, P. C. Schmitter, and L. Whitehead (eds), Transitions from
Authoritarian Rule: Prospects for Democracy, Baltimore: Johns Hopkins University
Press.
Przeworski, A. (1991). Democracy and the Market: Political and Economic Reforms in
Eastern Europe and Latin America, New York: Cambridge University Press.
Przeworski, A., Stokes, S. C, and Manin, B. (eds) (1999). Democracy, Accountability,
and Representation, Cambridge: Cambridge University Press.

268
 References

Quinn, K. M. (2004). ‘Bayesian Factor Analysis for Mixed Ordinal and Continuous
Responses’. Political Analysis, 12 (4): 338–53.
Qvortrup, M. (2002). A Comparative Study of Referendums, Manchester: Manchester
University Press.
Rabe-Hesketh, S., and Skrondal, A. (2005). Multilevel and Longitudinal Modeling Using
Stata, College Station, TX: Stata Press.
Rabe-Hesketh, S., and Skrondal, A. (2008). Multilevel and Longitudinal Modeling Using
Stata, Second Edition, College Station, TX: Stata Press.
Radcliff, B., and Davis, P. (2000). ‘Labor Organization and Electoral Participation in
Industrial Democracies’. American Journal of Political Science, 44: 132–41.
Reilly, Benjamin (2007a). ‘Political Engineering in the Asia-Pacific’. Journal of Democ-
racy, 18 (1): 58–72.
Reilly, Benjamin (2007b). ‘Democratization and Electoral Reform in the Asia-Pacific
Region: Is there an “Asian Model” of Democracy?’ Comparative Political Studies, 40
(11): 1350–71.
Riker, W. H. (1982). Liberalism versus Populism: A Confrontation between the Theory of
Democracy and the Theory of Social Choice, San Francisco: W. H. Freeman.
Roberts, A. (2006). ‘What Kind of Democracy is Emerging in Eastern Europe?’. Post
Soviet Affairs, 22 (1): 37–64.
Rohrschneider, R. (2002). ‘The Democracy Deficit and Mass Support for an EU-wide
Government’. American Journal of Political Science, 46 (2): 463–75.
Rohrschneider, R. (2005). ‘Institutional Quality and Perceptions of Representation
in Advanced Industrial Democracies’. Comparative Political Studies, 38 (7):
850–74.
Rohrschneider, R., and Loveless, M. (2010). ‘Macro-Salience: How Economic and Po-
litical Contexts Mediate Popular Evaluations of the Democracy Deficit in the Euro-
pean Union’. Journal of Politics, 72 (4): 1029–45.
Rohrschneider, R., and Schmitt-Beck, R. (2002). ‘Trust in Democratic Institutions in
Germany: Theory and Evidence Ten Years after Unification’. German Politics, 11
(3): 76–94.
Rose, R. (1973). ‘Comparability in Electoral Studies’, in R. Rose (ed.), Electoral Behav-
ior: A Comparative Handbook, New York: Free Press.
Rose, R., and Urwin, D. (1969). ‘Social Cohesion, Political Parties and Strains in Re-
gimes’. Comparative Political Studies, 2: 7–67.
Rosenstone, S., and Hansen, J. M. (1993). Mobilization, Participation and Democracy in
America, New York: Macmillan.
Roth, P.-A. (1996). Riket, Valkretsen och Hemkommunen, Gothenburg: Statsvetenskap-
liga institutionen.
Rubin, D. B. (1987). Multiple Inputation for Nonresponse in Surveys, New York: John
Wiley & Sons.
Ruiz-Rufino, Rubén (2013). ‘Satisfaction with Democracy in Multi-ethnic Countries:
The Effect of Representative Political Institutions on Ethnic Minorities.’ Political
Studies, 61 (1): 101–18.
Rustow, D. A. (1970). ‘Transitions to Democracy: Toward a Dynamic Model’. Com-
parative Politics, 2 (3): 337–63.

269
References

Sadeh, T. (2005). ‘Who Can Adjust to the Euro?’ The World Economy, 28(11):
1651–78.
Sadeh, T. (2006). Sustaining European Monetary Union—Confronting the Cost of Diver-
sity, Boulder, CO: Lynne Rienner.
Samuels, D., and Shugart, M. S. (2010). Presidents, Parties, and Prime Ministers: How
the Separation of Powers Affects Party Organization and Behavior, Cambridge: Cam-
bridge University Press.
Sanches-Cuenca, I. (2000). ‘The Political Basis of Support for European Integration’.
European Union Politics, 1 (2): 147–72.
Saward, M. (2010). The Representative Claim, Oxford: Oxford University Press.
Scarrow, S. E. (1996). Parties and Their Members: Organizing for Victory in Britain and
Germany, Oxford: Oxford University Press.
Scarrow, S. E. (2010). ‘Political Parties and Party Systems’, in L. LeDuc, R. G. Niemi,
and P. Norris (eds), Comparing Democracies, Vol. 3: Elections and Voting in the 21st
Century, London: Sage.
Schattschneider, E. E. (1942). Party Government, New York: Farrar and Rinehart, Inc.
Schattschneider, E. E. (1977). Party Government, Westport, CN: Greenwood Press.
Schedler, A. (1998). ‘What is Democratic Consolidation?’ Journal of Democracy, 9(2):
91–107.
Schedler, Andreas (2001). ‘Measuring Democratic Consolidation’. Studies in Compara-
tive and International Development 36 (1): 61–87.
Schmidt, M. G. (2000). Demokratietheorien, Opladen: Leske and Budrich.
Schmitt, H., and Thomassen, J. (eds) (1999). Political Representation and Legitimacy in
the European Union, Oxford: Oxford University Press.
Schmitt, H., and Wessels, B. (2005). ‘Meaningful Choices: Under which Conditions
do General Elections Provide a Meaningful Choice Set, and What Happens if They
Don’t?’ Stimulus paper for the Third Wave of the Comparative Study of Electoral
Systems, to be conducted from 2006 through 2010, Washington, DC.
Schnapp, K.-U., and Harfst, P. (2005). ‘Parlamentarische Informations- und Kon-
trollressourcen in 22 Westlichen Demokratien’. Zeitschrift für Parlamentsfragen, 36
(2): 348–72.
Schumpeter, J. (1976; 1942). Capitalism, Socialism and Democracy, London: George
Allen and Unwin.
Seligson, M. (2002). ‘The Impact of Corruption on Regime Legitimacy: A Compara-
tive Study of Four Latin American Countries’. Journal of Politics, 64 (2): 408–33.
Setälä, M. (2006). ‘On the Problems of Responsibility and Accountability in Referen-
dums’. European Journal of Political Research, 45 (4): 699–721.
Shadish, W. R., Cook, T. D., and Campbell, D. T. (2002). Experimental and Quasi-Ex-
perimental Designs for Generalized Causal Inference, Boston: Houghton Mifflin.
Shepsle, K. A. (1988). ‘Representation and Governance: The Great Legislative Trade-
off’. Political Science Quarterly, 103 (3): 461–84.
Shikano, S. (2006). ‘Bootstrap und Jackknife’, in J. Behnke, T. Gschwend, D. Schin-
dler, and K.-U. Schnapp (eds), Methoden der Politikwissenschaft. Neuere Qualitative
und Quantitative Analyseverfahren, Baden-Baden: Nomos.

270
 References

Siaroff, A. (2003). ‘Varieties of Parliamentarianism in the Advanced Industrial De-

mocracies’. International Political Science Review, 24 (4): 445–64.
Siaroff, A. (2005). Comparing Political Regimes: A Thematic Introduction to Comparative
Politics, Second Edition, Toronto: University of Toronto Press.
Singh, S. P. (2011). ‘How Compelling is Compulsory Voting? A Multilevel Analysis of
Turnout’. Political Behavior, 33: 95–111.
Smith, G. (1976). ‘The Functional Properties of the Referendum’. European Journal of
Political Research, 4 (1): 1–23.
Sonja, Zmerli, and Marc, Hooghe (eds.) (2011). Political Trust: Why Context Matters,
Colchester: ECPR Press.
Soroka, S., and Wlezien, C. (2010). Degrees of Democracy: Politics, Public Opinion, and
Policy, Cambridge: Cambridge University Press.
Steenbergen, M. R., and Jones, B. S. (2002). ‘Modeling Multilevel Data Structures’.
American Journal of Political Science, 46 (1): 218–37.
Stokes, D. E. (1963). ‘Spatial Models of Party Competition’. American Political Science
Review, 57: 368–77.
Stokes, D. E. (1992). ‘Valence Politics’, in D. Kavanagh (ed.), Electoral Politics, Oxford:
Clarendon Press.
Susan, J. Pharr, and Robert, D. Putnam (eds.) (2000). Disaffected Democracies: What’s
Troubling the Trilateral Countries? Princeton: Princeton University Press.
Taagepera, R. (2003). ‘Arend Lijphart’s Dimensions of Democracy: Logical Connec-
tions and Institutional Design’. Political Studies, 51 (1): 1–19.
Taagepera, R., and Shugart, M. S. (1989). Seats and Votes: The Effects and Determinants
of Electoral Systems, New Haven: Yale University Press.
Tabachnick, B. G., and Fidell, L. S. (2007). Using Multivariate Statistics, New York:
Allyn and Rose.
Thomassen, J. J. A. (1994). ‘Empirical Research into Political Representation: Failing
Democracy or Failing Models’, in M. K. Jennings and T. E. Mann (eds), Elections at
Home and Abroad: Essays in Honor of Warren Miller. Ann Arbor: University of Michi-
gan Press.
Thomassen, J. J. A. (2000). ‘From Comparable to Comparative Electoral Research’, in
J. W. van Deth, H. Rattinger and E. Roller (eds), Die Republik auf dem Weg zur Nor-
malität? Wahlverhalten und Politische Einstellungen nach Acht Jahren Einheit. Op-
laden: Leske and Budrich.
Thomassen, J. J. A. (2005a). ‘Modernization or Politics?’, in J. Thomassen (ed.), The
European Voter: A Comparative Study of Modern Democracies, Oxford: Oxford Univer-
sity Press.
Thomassen, J. J. A. (ed.) (2005b). The European Voter: A Comparative Study of Modern
Democracies, Oxford: Oxford University Press.
Thomassen, J. J. A., and Van der Kolk, H. (2009). ‘Effectiveness and Political Support
in Old and New Democracies’, in H.-D. Klingemann (ed.), The Comparative Study of
Electoral Systems, New York: Oxford University Press.
Todosijevic, B., Aarts, K., and Van der Kaap, H. (2010). ‘Dutch Parliamentary Elec-
tion Studies Data Sourcebook 1971–2006’. DANS Data Guide 7, The Hague.

271
References

Toka, G. (1995). ‘Political Support in East Central Europe’, in H.-D. Klingemann, and
D. Fuchs (eds), Citizens and the State, Oxford: Oxford University Press.
Tomz, M., Wittenberg, J., and King, G. (1999). Clarify: Software for Interpreting and
Presenting Statistical Results, Cambridge, MA: Harvard University, Department of
Political Science.
Treier, S., and Jackman, S. (2008). ‘Democracy as a Latent Variable’. American Journal
of Political Science, 52 (1): 201–17.
Tsebelis, G. (2002). Veto Players: How Political Institutions Work, Princeton, NJ: Prince-
ton University Press.
Tyler, T. R. (1988). ‘What is Procedural Justice? Criteria Used by Citizens to Assess
the Fairness of Legal Procedures’. Law and Society Review, 22: 103–35.
Tyler, T. R. (1990). Why People Obey the Law, New Haven, CT: Yale University
Press.
Tyler, T. R., Casper, J. D., et al. (1989). ‘Maintaining Allegiance toward Political Au-
thorities: The Role of Prior Attitudes and the Use of Fair Procedures’. American
Journal of Political Science, 33 (3): 629–52.
Van Biezen, I., and Katz, R. S. (2004). ‘Political Data in 2003’. European Journal
of ­Political Research, 43 (7/8): 919–26.
Van Biezen, I., and Katz, R. S. (2005). ‘Political Data in 2004’. European Journal
of ­Political Research, 44 (7/8): 919–28.
Van Biezen, I., and Katz, R. S. (2006). ‘Political Data in 2005’. European Journal
of ­Political Research, 45 (7/8): 1023–34.
Van der Brug, W. (1997). Where’s the Party? Voters’ Perceptions of Party Positions,
­Amsterdam: Het Spinhuis.
Van der Brug, W. (2008). ‘Een Crisis van de Partijendemocratie?’ Inaugural Address,
Vossiuspers Universiteit van Amsterdam, Amsterdam.
Van der Brug, W. (2010). ‘Structural and Ideological Voting in Age Cohorts’. West
European Politics, 33: 586–607.
Van der Brug, W., and Franklin, M. N. (1996). ‘What Voters Teach Us about Europe-
Wide Elections: What Europe-Wide Elections Teach Us about Voters’. Electoral
Studies, 15 (2): 149–66.
Van der Brug, W., and Franklin, M. N. (2009). Elections and Voters, Houndmills:
­Palgrave Macmillan.
Van der Brug, W., Franklin, M., Popescu, M., and Tóka, G. (2009). ‘Towards a
­European Electorate: One Electorate or Many?’ in J. Thomassen (ed.), The
­Legitimacy of the European Union After Enlargement, Oxford: Oxford University
Press.
Van der Brug, W., and Marsh, M. (2007). ‘Don’t Expect Me to Vote for You Just Be-
cause I Like You, Even if You Do Make Me Feel Warm Inside: A Comparison of the
Validity of Non-Ipsative Measures of Party Support’. Working Paper.
Van der Brug, W., and Van Spanje, J. (2009). ‘Immigration, Europe and the “New”
Cultural Dimension’. European Journal of Political Research, 48 (3): 309–34.
Van der Eijk, C., and Schmitt, H. (2010). Party Manifestos as a Basis of Citizens Left–
Right Perceptions—A Study Across 10 European Countries, Annual Meeting of the
American Political Science Association, Washington, DC, 2–5 September.

272
 References

Van der Eijk, C., Schmitt, H., and Binder, T. (2005). ‘Left–Right Orientations and
Party Choice’, in J. Thomassen (ed.), The European Voter: A Comparative Study of
Modern Democracies, Oxford: Oxford University Press.
Van der Meer, T. (2009). ‘States of Freely Associating Citizens: Cross-national Studies
into the Impact of State Institutions on Social, Civic and Political Participation’.
PhD thesis, Nijmegen: University of Nijmegen.
Van der Meer, Tom, and Erik van Ingen (2009). ‘Schools of Democracy? Disentan-
gling the Relationship between Civic Participation and Political Action in 17 Euro-
pean Countries’. European Journal of Political Research, 48: 281–308. DOI:
10.1111/j.1475-6765.2008.00836.x
Van der Meer, T. W. G., Van Deth, J. W., and Scheepers, P. L. H. (2009). ‘The Politi-
cized Participant: Ideology and Political Action in 20 Democracies’. Comparative
Political Studies, 42: 1426–57.
Vatter, A. (2000). ‘Consensus and Direct Democracy: Conceptual and Empirical
Linkages’. European Journal of Political Research, 38 (1): 171–92.
Vatter, A. (2005). ‘Bicameralism and Policy Performance: The Effects of Cameral
Structure in Comparative Perspective’. Journal of Legislative Studies, 11 (2):
194–205.
Vatter, A. (2009). ‘Lijphart Expanded: Three Dimensions of Democracy in Advanced
OECD Countries?’ European Political Science Review, 1(1): 125–54.
Vatter, A., and Bernauer, J. (2009). ‘The Missing Dimension of Democracy: Institu-
tional Patterns in 25 EU Member States between 1997 and 2006’. European Union
Politics, 10 (3): 335–59.
Vatter, A., and Bernauer, J. (2010a). Consensus Democracy Indicators in 26 Advanced
Democracies: Political Data Set 1997–2006, Berne: Institute of Political Science, Uni-
versity of Berne; Konstanz: Department of Politics and Management, University of
Konstanz.
Vatter, A., and Bernauer, J. (2010b). Consensus Democracy Indicators in 35 Democracies.
Political Data Set 1997–2006, Berne: Institute of Political Science, University of
Berne, Konstanz: Department of Politics and Management, University of
Konstanz.
Vatter, A., and Bernauer, J. (2011). Consensus Democracy Indicators in 35 Democracies,
1997–2006, Berne/Konstanz: Department of Political Science.
Verba, S., and Nie, N. (1972). Participation in America: Political Democracy and Social
Equality, New York: Harper and Row.
Verba, S., Nie, N., and Kim, J.-O. (1978). Participation and Political Equality, Chicago:
University of Chicago Press.
Verba, S., Schlozman, K. L., and Brady, H. (1995). Voice and Equality: Civic Volunta-
rism in American Politics, Cambridge, MA: Harvard University.
Vrablikova, K. (2010). ‘Contextual Determinants of Political Participation in Demo-
cratic Countries’, CRONEM Annual Conference. University of Surrey, England.
Wängnerud, L., Esaiasson, P., Gilljam, M., and Holmberg S. (2010). The Riksdag
Survey 2010, Gothenburg: The Department of Political Science.
Wattenberg, M. P. (1991). The Rise of Candidate-centered Politics: Presidential Elections
of the 1980s, Cambridge, MA: Harvard University Press.

273
References

Weale, A. (2005). Democratic Citizenship and the European Union, Manchester: Man-
chester University Press.
Welzel, C., Inglehart, R. F., and Deutsch, F. (2005). ‘Social Capital, Voluntary Associa-
tions and Collective Action: Which Aspects of Social Capital Have the Greatest
“Civic” Payoff?’ Journal of Civil Society, 1 (1): 121–46. DOI: 10.1080/
17448680500337475
Wessels, B. (1999). ‘System Characteristics Matter: Empirical Evidence from Ten Rep-
resentational Studies’, in W. Miller et al. (eds), Policy Representation in Western De-
mocracies, Oxford: Oxford University Press.
Wessels, B. (2004). Members of Parliament Survey 2003: Study Report, Berlin: Wissen-
schaftszentrum Berlin fur Sozialforschung (WZB).
Wessels, B. (2007). ‘Political Representation and Democracy’, in R. J. Dalton, and
H.-D. Klingemann (eds), The Oxford Handbook of Political Behavior, New York:
Oxford University Press.
Wessels, B., and Schmitt, H. (2006). ‘What Shapes Electoral Choice Options? Institu-
tional Determinants of the Distinctiveness of Choice Options?’ Presented at the
Electoral Systems and Electoral Politics Bangalore CSES Workshop, 2–3 November.
Whiteley, P., Clarke, H., Sanders, D., and Stewart, M. (2010). ‘Do Institutions Really
Influence Political Participation? Contextual Influences on Turnout and Participa-
tion in the World’s Democracies’. International Journal of Market Research, 52 (1):
1–22.
Whiteley, P., and Seyd, P. (2002). High-Intensity Participation: The Dynamics of Party
Activism in Britain, Ann Arbor, MI: University of Michigan Press.
Wiesehomeier, N., and Benoit, K. (2009). ‘Presidents, Parties, and Policy Competi-
tion’. The Journal of Politics, 71: 1435–47.
Zielinski, J. (2002). ‘Translating Social Cleavages into Party Systems: The Significance
of New Democracies’. World Politics, 54: 184–211.

274
Index
Aarts, K.  3–4, 15, 19, 99, 103, 201, 206, 229, 256
accountability  1–4, 6–7, 17, 60, 62, 101, 135
function of elections  1–4, 6, 17, 101, 135
model of  2–4, 17, 135
Anderson, C. J.  5, 19, 64, 100, 102, 105, 113,
122, 188–9, 232–3, 235, 238–40, 250, 256
Banducci, S. A.  101, 113–14, 127
Bartels, L. M.  245
benefits of participation  113, 117
Bernauer, J.  7, 20, 22, 24, 27, 30, 37, 45, 54–6,
60, 72, 105–6, 114–15, 120, 123, 131, 136,
147, 149, 214, 228
bicameralism  20–1, 26, 29, 35, 62, 105
Blais, A.  10, 99, 103, 113–14, 117, 122, 124,
127, 256
Brady, H.  116
Brambor, T.  53
Bratton, M.  251
Brooks, C.  80, 94–5
Budge, I.  48, 80
Burke, E.  133
Burlacu, D.  8, 60
calculus of voting  8, 45
campaign activity  11, 118–30
central bank independence  22, 27, 29–30, 34, 36
Chang, E. C. C.  16, 232, 238
Cho, W.  251
civic duty  99, 115
Civic Voluntarism Model  116
civil society  12, 116, 128, 205
clarity of
policy positions  8, 42, 45, 48, 50–3, 55–9, 76
political supply  8, 42, 45, 48, 50–3, 57–9
responsibility  3, 6, 13–14, 18–19, 153,
159–62, 164–7, 169–73, 175, 177–80
Clark, T. N.  53, 85
class
or union membership voting  90–1, 93–4, 96
voting  79, 83, 85, 90–7, 230
cleavage politics  80, 86, 97, 204, 241
cognitive mobilization  15, 202
communal action  118
Comparative Study of Electoral Systems
(CSES)  1, 5, 19–21, 24, 33–5, 60, 63, 80,
103, 113, 118–19, 121, 125, 138, 140, 207,
209, 225, 233
consensus
(model of) democracy  1–10, 12–13, 15–27,
29–34, 36, 38–42, 44, 51–6, 58–61,
63–4, 68, 72, 76–8, 83–4, 92–3, 95–6,
100–1, 105–9, 111, 113–16, 120, 122–3,
126–8, 131, 135–6, 139, 147–52, 163–4,
183, 185, 189, 195, 197, 206- 7, 211–15,
226–8, 235, 251, 256
versus majoritarian democracy  6–7, 41, 61,
72, 78, 148–9, 163
constitutional federalism  21, 26, 29, 35, 128,
183
constitutional rigidity  21, 26, 29–31, 33–5
contacting politicians  122, 123, 128
corruption  15, 158, 163–9, 173, 188, 190–1,
193–9, 244–6, 253, 256
crisis of legitimacy  201
Cukierman, A.  27, 36–7
cultural dimension  120, 204–6, 209, 212, 217,
222, 225–6
cultural revolution  203–4, 206
Curini, L.  240–1, 252
Dahl, R. A.  39, 43, 114, 187, 245
measure of polyarchy  114
Dalton, R. J.  3, 11, 19, 80, 83, 85, 113, 116,
131, 201–3, 207
De Montesquieu, C. L.  133
democracy
age of  13, 86, 105–8, 120, 124–5, 131, 135,
148–9, 151–2, 160, 163, 168, 170–1,
173, 175, 190, 192, 195–6, 214
dimensions of  2, 10, 20–2, 28–34, 40–1, 44,
46, 58, 61, 72, 75, 83, 105–8, 114, 120,
123, 197, 206, 209, 212, 217, 222, 225
direct  7–8, 13, 18, 21–4, 27–34, 36, 47, 76,
107, 127, 131, 133, 136, 168, 171, 239
patterns of  4, 7, 20–2, 24, 28, 30, 40, 92,
106, 113, 120, 122–3, 128, 131, 178,
180, 185–6, 192, 196, 221, 224
275
Index
democratic support/support for democracy/
democratic satisfaction  1, 5–6, 9, 11–19,
30, 59, 66, 68, 74–6, 83, 86, 95–6, 114,
117, 120, 122–5, 127, 129, 132–3, 136,
148–9, 153–69, 171–6, 178–9, 181–7, 189,
192–3, 197, 201–2, 205, 207–9, 211–12,
214, 216–17, 219–26, 229–35, 238–50,
253, 256
democratic citizenship  87, 127, 128
democratic malaise  201
democratic performance  3–4, 6–8, 13–14, 34,
38, 63, 74, 153, 157, 160–1, 171, 174–6,
181–4, 186, 189–93, 195–7, 199–201, 233,
236, 238, 242, 244–5, 253
democratic principles, support for  14, 17, 184,
234, 242–3, 246, 248
diffuse support  83, 234, 242
district magnitude  25, 121–2
Duch, R. M.  62, 64–5
Dumitrescu, D.  99, 114
Easton, D.  153, 157, 233–4, 242
effective number of electoral parties index  24,
35, 93, 105, 136
efficacy  103, 106, 108, 110, 114, 117, 123–5,
127, 129, 212, 217, 219–21, 223–6, 229,
231
elections as instrument of democracy  4, 8, 39
electoral accountability  1, 3–4, 7, 17, 60, 62,
101, 135
electoral disproportionality  18, 21, 25, 29, 35,
41, 93, 105, 136, 147, 161–2, 164, 169–71,
175–6
electoral system  1, 3–5, 7, 9, 11–13, 17–21,
23–5, 33–5, 39, 41–2, 45, 60, 63–4, 69, 73,
75, 77, 79–85, 93, 96–8, 100–3, 105, 108,
111, 113–19, 121–5, 127–8, 135–6, 138,
140, 147, 149–50, 153, 158–63, 168, 171,
174–6, 206–9, 225, 232–5, 237–8, 242–3,
250–1, 256
electoral volatility  69, 80, 86
elite-challenging behaviour  117, 124
ethnic voting  81, 83, 88–90, 256
Eurobarometer  154, 156, 174, 207, 209, 211,
214, 225–6, 228, 231
European integration  47, 49, 204, 212, 222–5
Evans, G.  64, 80, 245
executive dominance  9, 25, 29, 35, 37, 41, 45,
60, 62, 93, 97
index of  25, 35, 37, 93
external efficacy  212, 217, 219–21, 223–6,
229, 231
Fearon’s formula  94
federalism  21, 26, 29, 35, 62, 76, 105, 115,
125–6, 128, 167, 183, 190–2, 195–6, 199
276
fiscal decentralization  21, 24, 26, 29
Franklin, M. N.  80, 87–8, 97, 103
Franzese, R. J. Jr.  53
Gallagher, M.  25, 35, 88, 158
index  25, 35, 88
Ganghof, S.  22, 25
gender voting  9, 81, 86, 89–91, 93–8, 103,
106, 108, 124
Giger, N.  7, 20, 111, 114
globalization  15–16, 201–5, 207–9, 214, 217,
222, 226, 229–30
thesis  15, 202–3, 205, 209, 226
Golder, M.  4, 35, 53, 60–1
government performance  2–4, 6–7, 13–14, 34,
38, 40, 62–8, 74–5, 78, 153, 157, 160–9,
171–6, 178–84, 192–3, 195–200, 233, 238,
242, 244–5, 253
evaluation  2, 14, 65–7, 160, 174, 197, 242,
244–5
Grofman, B.  22
Guillory, C. A.  5, 189, 232, 235, 256
Holmberg, S.  13, 38, 132, 138, 142, 152
Honaker, J.  67
Huber, J. D.  2, 12, 38, 81, 83, 88–9, 94
human rights  15, 158, 163, 165, 167–9,
172–4, 191, 193–9
Huntington, S. P.  16, 151, 233, 240, 250
IDEA  2, 14–15, 36, 39, 48, 58, 85, 119–22,
130–1, 133, 140, 172, 181, 187, 197
measure of turnout  120, 122
ideological extremism  117, 123–5, 129
ideological voting  8–9, 61–3, 65–6, 73–7,
116–17, 123–5, 128, 241
inclusiveness  3, 5, 114
Inglehart, R.  98, 116, 202, 206
input institutions  18, 183, 188, 201
institutional reforms  19, 184
judicial review  20–1, 27, 29, 36, 62, 105
Karvonen, L.  26, 35–6
Kitschelt, H.  61, 81, 98, 115, 117, 229
opportunity theory of participation  117
Kittilson, M. C.  105, 113, 122
Knutsen, O.  62, 64, 80
Kriesi, H.  98, 201–4, 206, 210
Kumlin, S.  62
Laakso, M.  24, 35, 105
leader representation  99, 143, 145–9, 152
least-squares index  88

left–right  8–9, 40, 44–59, 61, 63–8, 72–5,
77–8, 80, 86, 88, 102, 116, 123, 128,
149–50, 199, 203, 206, 240–1
dimension  8–9, 44–6, 48, 54, 56, 58, 61, 63,
77–8, 128, 199, 206, 240–1
ideological dimension  8–9, 46, 61, 63, 77,
128, 240–1
positions  8, 40, 44–5, 48–53, 55–9, 61, 66,
72, 75, 102
scale  40, 44–6, 48–50, 52–6, 58–9, 61, 65,
75, 78, 149
self-placement  53, 55, 57, 64–7, 73–4, 78,
86, 149–50
legitimacy (of the political system)  13, 16, 19,
113, 153, 157, 160–1, 174, 181–2, 201,
203, 225, 232, 243, 245, 251, 256
libertarian revolution  203–4
libertarian-authoritarian dimension  45, 93, 105, 122, 136, 147
204–5
Lijphart, A.  2, 4–5, 7–8, 10–13, 15, 19–30,
32–4, 37–41, 44–5, 52, 54–6, 58–60, 62–3,
67, 72, 83, 97, 113–15, 120, 123, 128,
135–6, 163, 165, 167–70, 189–90, 195–7,
199–200
Lindberg, S. I.  236–7, 251
Linz, J. J.  234
Lipset, S. M.  79, 82–3, 85–6, 97, 205–6, 234
Logan, C.  251
Lundell, K.  3, 26, 35–6
Madison, J.  133
Magalhães, P. C.  9, 79
Mainwaring, S.  86–7
majoritarian (model of) democracy/
­majoritarian institutions  2–13, 15–27,
32–6, 38–42, 44, 52, 55, 58–62, 64, 67, 72,
75, 77–8, 82–4, 101, 113–16, 123, 127–8,
135–6, 138–40, 146–52, 163, 173, 183,
189, 197, 206–7, 211–15, 226–8,
235, 256
majoritarian versus consensus ­democracy  6–7,
41, 61, 72, 78, 148
Mandate Model  2, 133–6
Manza, J.  80, 94–5
McAllister, I.  80, 101, 131
meaningful choices  8, 38–40, 42, 51–2, 58–9
meaningful elections  7, 8, 39–40, 42–3, 51,
58–9
median voter  23, 41–2, 84
minority cabinets  24, 29, 35, 83
minority government  62, 64, 128, 174, 185
modernization thesis  15, 202, 226
Moehler, D. C.  251
multipartism 62
multiparty system  3, 9, 20, 23, 35, 42, 60, 69,
73, 75, 102, 175, 206
Index
negative effects thesis  13, 96, 160, 202–3,
206–8, 214, 221, 226
Nie, N.  113, 117–18
Norris, P.  25, 35, 83, 85–6, 88, 113, 117, 119,
123, 178, 201–2, 242
older democracies  13, 77–8, 87, 95, 98, 126,
139, 141–2, 151, 160, 175
output institutions  14–15, 18, 183, 188–94,
197, 201, 242
oversized cabinets  24, 29, 35
oversized coalitions  24, 35, 62
participation
conventional  115, 117, 124, 126, 130
unconventional  117, 124
parties, effective number of  24, 28–9, 35, 41,
partisanship  53, 73, 244–5, 253
strength of  73, 244–5, 253
party
attachment  30, 117, 129
identification  53, 65–7, 100–1, 106–8, 111,
114, 123–4, 126–7, 149, 245
strength of  65–7, 245
Manifesto Project  45–6
platforms  42, 45–6, 48, 50–1, 58, 85, 107, 163
representation  2, 7, 10–11, 16, 18, 21,
39–42, 44–5, 52–4, 58–60, 79, 82, 87,
99–103, 105–11, 114, 121–2, 134, 137,
143, 145–9, 151–2, 159, 164, 166,
168–9, 173, 183, 194–5, 199, 209, 214,
230, 256
system  3, 7, 9–11, 16, 18–19, 21, 23, 25, 35,
40–2, 45–6, 48, 51–3, 58–61, 64, 66, 69,
73, 75, 82–3, 85, 87, 97–8, 100–3, 105–8,
110–11, 116–17, 121–4, 126–8, 144,
146–7, 149–52, 157–9, 161, 163, 165,
168, 173, 183, 208, 235, 238, 245, 256
Pedersen index  69, 70, 75
Peffley, M.  14, 181, 183
performance
-based voting  8, 60–63, 65, 76, 78
evaluations  8, 14, 62–71, 74–5, 78, 153,
157, 161, 174, 181–4, 186, 188–93,
195–200, 233, 238, 242, 245
of the incumbent government  2, 3, 7, 157,
184, 238
-oriented voting  9, 61, 65–6, 75, 78
personal vote  78, 85, 99, 127, 188
personalization of politics  84, 128
plebiscites  23, 27
plurality
electoral systems  13, 18, 100, 138, 140, 153,
159–62, 168, 171, 174–5
system  13, 18, 100, 138, 140, 153, 158–63,
165, 167–8, 171, 173–5, 179
277
Index
polarization  6, 8–10, 18–19, 62–3, 72–3, 76–7,
102, 105, 107, 110–11, 116, 121, 123,
125–6
policy
-based voting  8, 18, 45, 60–3, 65, 72–3,
76–8, 100, 132, 134
congruence  61, 111, 134, 136–8, 142,
149–50
mandate model  2, 133–6
-oriented voting  9, 18, 61, 65–6, 75, 78, 134
representation  1, 2, 6–7, 12–13, 18, 38–40,
42, 44–5, 52–4, 58–60, 82, 100–2, 111,
133–9, 141–3, 145–9, 151–2, 159, 168,
174–5, 186
voting  8–9, 12, 18, 39–40, 42–5, 51–3,
58–63, 65–6, 72–8, 82, 84, 100, 111,
128, 132, 134, 136, 159, 238, 245
political efficacy  103, 106, 108, 114, 117,
123–5, 127, 212, 217, 221, 223, 225–6,
229, 231
political protest  114, 117–21, 124–6, 128, 131
political representation, quality of  6, 12–15,
21, 38–40, 113, 183, 187–8, 190, 192,
194–6, 199
populist parties  203, 206, 229
populist revolution  203, 206
positive effects thesis  13, 96, 158, 160, 202–3,
207–8, 214, 226
Powell, G. B.  2–4, 12, 25, 38, 60–62, 101–2,
136, 189, 245
power
efficacy  117, 124–5, 129, 217
-sharing  5, 22–4, 26, 62, 251
presidential systems  9, 37, 78, 84–5, 94, 96,
116, 123, 125–6, 252
presidentialism  37, 84, 85, 93, 95–7, 252
principal-agent model  100
procedural fairness  14, 182–3, 187–8, 193, 195
proportional (electoral) system  3, 10–11, 16,
18, 20, 25, 35, 38, 40–1, 45, 60–1, 100–2,
105, 107–8, 110, 114, 122, 136, 138–9,
146–50, 159, 171, 173–4, 183, 189,
206, 232
proportional representation  2, 10–11, 16,
18, 20, 38, 40–1, 44–5, 60, 100–2, 105,
107–10, 114, 122, 136, 138–9, 146–9, 159,
171, 173–4, 183, 189, 232
protest activity  117–21, 124–6, 128–9
Proximity Voting  12, 39–42, 44–5, 51–3,
58–9, 241
Przeworski, A.  109, 236
referendum  23, 26–7, 34, 36, 188, 231
regime
performance  13–14, 153, 157, 160–1, 166,
168–9, 171–6, 181, 183–4, 186, 188–90,
195–7, 236, 242, 245, 253
278
support  14–15, 153, 158, 166, 168, 171,
174, 181, 183–4, 186–7, 189, 197,
242–3, 245–6, 250, 253
religiosity-based voting  80, 90, 96
religious voting  9, 79–81, 89–91, 94–8, 130
representation
feelings of  10, 13, 15–16, 100, 102–3,
105–11, 139, 141–2, 144, 147–9, 151,
195, 209, 214, 229–30
from above  7, 10, 44, 103, 133, 166
from below  20, 133, 166, 188, 190
function of elections  1, 2, 6, 13–14, 54, 101,
108, 135, 174, 188
perceptions of  6, 11–15, 18, 42, 45, 52, 58,
99–103, 106–8, 110–11, 141, 145,
149, 164, 174, 182, 187–9, 191,
194–6, 199
sense of  39, 45, 52, 59, 101, 115, 168–9,
172–3, 190, 229
substantive  14, 39, 58, 181, 183, 186–8,
192, 194
representativeness  4, 6, 7, 17, 135
responsible Party Model  2, 3, 40, 42, 50, 134
responsiveness  8–9, 15, 60–1, 76–7, 100–1,
245
retrospective evaluations of government  43,
62, 134
retrospective voting  8, 18, 43, 62, 134
Rohrschneider, R.  14, 181, 183–4, 189, 199
Rokkan, S.  79, 82–3, 86, 97, 204
Rustow, D. A.  237
Samuels, D.  84
Sanction Model  3–4, 133–5
Sanders, D.  13, 153, 182, 189
satisfaction with democracy  1–3, 5–6, 12–14,
16–18, 66, 68, 75, 117, 123–5, 136, 153–4,
156–61, 166–8, 171–6, 179–81, 183–6,
197, 206–9, 211–12, 214, 216–17, 219–21,
223–6, 230–5, 238–43, 245, 247–50, 256
Schlozman, K. L.  116
Schmitt, H.  7, 38, 47–8, 58, 101, 137–8, 184,
199, 256
Schumpeter, J.  1–2
Secularization 85
semi-presidential system  9, 93–4, 96, 252
Shepsle, K. A.  38
Shikano, S.  22
Shugart, M. S.  84, 158
Siaroff, A.  25, 35–6
Singh, S. P.  10, 99, 105, 114
single-member district (SMD) systems  232
social cleavage(s)  19, 79–80, 82–3, 85–87
model 86
theory  80, 86
social modernization  15, 81, 85–6, 98, 202,
207, 226
 Index

social-economic dimension  9, 90, 108, 204–6,
209–10, 212, 214, 217, 222, 225–6, 230
Soroka, S.  139
Stepan, A. C.  234
Stevenson, R. T.  62, 64–5
Stewart, M.  13, 153
Stramski, J.  4, 60–1
structural voting  9, 45, 52, 62, 79–98
Taagepera, R.  22, 24, 27, 30, 33, 35, 105, 158
third-wave democracies  33, 37, 87, 95, 234, 241
Thomassen, J. J. A.  1–4, 15, 19, 38, 40, 42, 98,
135, 138, 201, 206, 256
Toka, G.  8, 154
trust in political leaders  201, 256
turnout  5, 10–12, 99–101, 103–6, 108–11,
113–16, 118–20, 122–4, 126–8, 130–1, 136
two-turnover test  16, 233–4, 240, 250
Van der Eijk, C.  47–8, 58, 87, 98
Van Ham, C.  15, 201
Vanhanen’s index of democracy  114
Vatter, A.  7, 20, 22–4, 27, 30, 35–7, 45, 54–6,
60, 72, 105–6, 114, 131, 136, 147, 149
Verba, S.  113, 116–18, 256
Verba and Nie’s modes of political action 
118
veto players theory  23
Weldon, S.  11, 113, 116
Wessels, B.  7, 19, 38, 101, 138
Whitefield, S.  245
Whiteley, P.  13, 153, 158
Whitten, G. D.  62, 102
winner–loser gap  16, 19, 232–5, 238–40,
248–52
Wlezien, C.  139

valence issues  61 Zoco, E.  86–7

279