Personalization of Politics and Electoral Change

This file is to be used only for a purpose specified by Palgrave Macmillan, such as checking proofs, preparing an index, reviewing,endorsing or planning coursework/other institutional needs. You may store and print the file and share it with others helpingyou with the specified purpose, but under no circumstances may the file be distributed or otherwise made accessible to any otherthird parties without the express prior permission of Palgrave Macmillan.Please contact [email protected] if you have any queries regarding use of the file.

October 25, 2013 14:53 MAC/PERSON Page-i 9781137270221_01_prex

PROOF

Palgrave Studies in Political Psychology series

The Palgrave Studies in Political Psychology book series profiles a range ofinnovative contributions that investigate the leading political issues and per-spectives of our time. The academic field of political psychology has beendeveloping for almost 50 years and is now a well-established subfield ofenquiry in the North American academy. In the context of new global forcesof political challenge and change as well as rapidly evolving political prac-tices and political identities, Palgrave Studies in Political Psychology buildsupon the North American foundations through profiling studies from Europeand the broader global context. From a theoretical perspective, the seriesincorporates constructionist, historical, (post)structuralist, and postcolonialanalyses. Methodologically, the series is open to a range of approaches to polit-ical psychology. Psychoanalytic approaches, critical social psychology, criticaldiscourse analysis, Social Identity Theory, rhetorical analysis, social represen-tations, and a range of quantitative and qualitative methodologies exemplifythe range of approaches to the empirical world welcomed in the series. Theseries integrates approaches to political psychology that address matters ofurgency and concern from a global perspective, including theories and per-spectives on world politics and a range of international issues: the rise of socialprotest movements for democratic change, notably in the global South andthe Middle East; the Israeli–Palestinian conflict and its broader implications;patterns of global migration and associated challenges of integration and reli-gious accommodation; the formation and deformation of political, economic,and strategic transnational entities such as the European Union; conflicts andviolence resulting from local and regional nationalisms; emerging politicalmovements of the new left and the new right; ethnic violence; legacies of warand colonization; and class conflict.

Series editors:

Tereza Capelos is Senior Lecturer at the University of Surrey, UK, Vice-President of the International Society of Political Psychology, Co-Chair of theECPR Political Psychology Standing Group, and Director of the InternationalSociety of Political Psychology Summer Academy (ISPP-SA).

Henk Dekker is Professor of Political Socialization and Integration at theGraduate School of Social and Behavioural Sciences, Associate Professor ofPolitical Science at the Institute of Political Science, and Vice-Dean of theFaculty of Social and Behavioural Sciences of Leiden University. He is a recip-ient of the Nevitt Sanford Award from the International Society of PoliticalPsychology.

Catarina Kinnvall is Professor at the Department of Political Science, LundUniversity, Sweden and former vice-president of the International Society ofPolitical Psychology (ISPP).

October 25, 2013 14:53 MAC/PERSON Page-ii 9781137270221_01_prex

PROOF

Paul Nesbitt-Larking is Professor of Political Science at Huron UniversityCollege, Canada and Visiting Professor, School of Health and Human Sci-ences, University of Huddersfield, UK. He is currently President-elect of theInternational Society of Political Psychology.

Titles include:

Nicolas Demertzis (editor)EMOTIONS IN POLITICSThe Affect Dimension in Political Tension

Diego GarziaPERSONALIZATION OF POLITICS AND ELECTORAL CHANGE

Lisa StrombomISRAELI IDENTITY, THICK RECOGNITION AND CONFLICTTRANSFORMATION

Palgrave Studies in Political Psychology seriesSeries Standing Order ISBNs 978–1137–03466–3 (hardback) and978–1137–03467–0 (paperback)(outside North America only)

You can receive future titles in this series as they are published by placing astanding order. Please contact your bookseller or, in case of difficulty, write tous at the address below with your name and address, the title of the series andthe ISBNs quoted above.

Customer Services Department, Macmillan Distribution Ltd, Houndmills,Basingstoke, Hampshire RG21 6XS, England

October 25, 2013 14:53 MAC/PERSON Page-iii 9781137270221_01_prex

PROOF

Personalization of Politicsand Electoral ChangeDiego GarziaRobert Schuman Centre for Advanced Studies European UniversityInstitute, Fiesole, Italy

October 25, 2013 14:53 MAC/PERSON Page-iv 9781137270221_01_prex

PROOF

© Diego Garzia 2014

All rights reserved. No reproduction, copy or transmission of thispublication may be made without written permission.

No portion of this publication may be reproduced, copied or transmittedsave with written permission or in accordance with the provisions of theCopyright, Designs and Patents Act 1988, or under the terms of any licencepermitting limited copying issued by the Copyright Licensing Agency,Saffron House, 6–10 Kirby Street, London EC1N 8TS.

Any person who does any unauthorized act in relation to this publicationmay be liable to criminal prosecution and civil claims for damages.

The author has asserted his right to be identified as the author of thiswork in accordance with the Copyright, Designs and Patents Act 1988.

First published 2014 byPALGRAVE MACMILLAN

Palgrave Macmillan in the UK is an imprint of Macmillan Publishers Limited,registered in England, company number 785998, of Houndmills, Basingstoke,Hampshire RG21 6XS.

Palgrave Macmillan in the US is a division of St Martin’s Press LLC,175 Fifth Avenue, New York, NY 10010.

Palgrave Macmillan is the global academic imprint of the above companiesand has companies and representatives throughout the world.

Palgrave® and Macmillan® are registered trademarks in the United States,the United Kingdom, Europe and other countries.

ISBN 978–1–137–27022–1

This book is printed on paper suitable for recycling and made from fullymanaged and sustained forest sources. Logging, pulping and manufacturingprocesses are expected to conform to the environmental regulations of thecountry of origin.

A catalogue record for this book is available from the British Library.

A catalog record for this book is available from the Library of Congress.

October 25, 2013 14:53 MAC/PERSON Page-v 9781137270221_01_prex

PROOF

Contents

List of Tables and Figures vi

Acknowledgments ix

Introduction 1

1 The Personalization of Politics 5

2 Attitudinal Consequences 23

3 Behavioral Consequences 42

4 Electoral Consequences 57

5 Normative and Empirical Implications 78

Appendix A: Data Sources and Analytical Methods 90

Appendix B: Detailed model estimation procedures 106

Appendix C: Parties and party leaders in Britain, Germany, andthe Netherlands (1980–2010) 118

Notes 120

References 124

Index 136

v

October 25, 2013 14:53 MAC/PERSON Page-vi 9781137270221_01_prex

PROOF

Tables and Figures

Tables

I.1 Case selection 32.1 Percentage of voters close to the main two parties

among all partisans, by decade 282.2 Social structure and partisanship in three countries 312.3 The attitudinal drivers of partisanship in three

countries 353.1 Leader effects on voting in three countries:

Single-equation estimation 473.2 Summary diagnostics of instrumental variables 503.3 Leader effects on voting in three countries:

Instrumental variable estimation 514.1 Mean thermometer score of party leaders, Britain 644.2 Mean thermometer score of party leaders, Germany 654.3 Mean thermometer score of party leaders, the

Netherlands 664.4 The overall effect of party leaders on electoral

outcomes, Britain 734.5 The overall effect of party leaders on electoral

outcomes, Germany 744.6 The overall effect of party leaders on electoral

outcomes, the Netherlands 75B.1 Social structure and partisanship in Britain – by party

family 106B.2 Social structure and partisanship in Germany – by

party family 106B.3 Social structure and partisanship in the Netherlands –

by party family 107B.4 The attitudinal drivers of partisanship in Britain – by

party family 108B.5 The attitudinal drivers of partisanship in Germany –

by party family 109

vi

October 25, 2013 14:53 MAC/PERSON Page-vii 9781137270221_01_prex

PROOFList of Tables and Figures vii

B.6 The attitudinal drivers of partisanship in theNetherlands – by party family 110

B.7 Estimation procedure for instrumental variables 110B.8 Instrumental variable estimation, British parties 111B.9 Instrumental variable estimation, German parties 113

B.10 Instrumental variable estimation, Dutch parties 114B.11 Leader effects by educational level, Britain 115B.12 Leader effects by educational level, Germany 115B.13 Leader effects by educational level, the Netherlands 116

Figures

1.1 The driving forces of the personalization of politics 71.2 The consequences of the personalization of politics:

Plan of the book 212.1 Social structure and partisanship in Britain – by party

family 322.2 Social structure and partisanship in Germany – by

party family 332.3 Social structure and partisanship in the Netherlands –

by party family 332.4 Social vs. attitudinal drivers of partisanship in three

countries: A comparison 362.5 Leader evaluations and partisanship in Britain – by

party family 372.6 Leader evaluations and partisanship in Germany – by

party family 382.7 Leader evaluations and partisanship in the

Netherlands – by party family 393.1 The “funnel” of causality 443.2 Leader effects in Britain: Single-equation vs.

IV-estimation 523.3 Leader effects in Germany: Single-equation vs.

IV-estimation 523.4 Leader effects in the Netherlands: Single-equation vs.

IV-estimation 533.5 Leader effects in Britain – by party family

(IV-estimation) 54

October 25, 2013 14:53 MAC/PERSON Page-viii 9781137270221_01_prex

PROOFviii List of Tables and Figures

3.6 Leader effects in Germany – by party family(IV-estimation) 55

3.7 Leader effects in the Netherlands – by party family(IV-estimation) 56

4.1 Vote share (%) for the main three parties in eachcountry 62

4.2 Comparative evaluation of main parties’ leaders,Britain 68

4.3 Comparative evaluation of main parties’ leaders,Germany 69

4.4 Comparative evaluation of main parties’ leaders, theNetherlands 69

4.5 Net gains/losses for main parties, Britain 714.6 Net gains/losses for main parties, Germany 714.7 Net gains/losses for main parties, the Netherlands 725.1 Leader effects and voters’ educational level, Britain 835.2 Leader effects and voters’ educational level, Germany 845.3 Leader effects and voters’ educational level, the

Netherlands 84A.1 Stacking the data matrix 101

October 25, 2013 14:53 MAC/PERSON Page-ix 9781137270221_01_prex

PROOF

Acknowledgments

This book owes so much to so many people and places that I willhardly be able to list them all. Yet, some of them must be singledout as their help has been simply essential for the completion ofthis project. Some of the key hints came to my mind while I wasan MA student at Leiden University in 2008. I would like to thank allof those who made my years there special, but given the interminablelength of this list I will limit myself to acknowledge my thesis super-visor, Joop van Holsteyn, for his patience in engaging with my early(and at that time extremely confused) thoughts on voters, leaders,and the outcome of democratic elections. The core part of this bookis based on the research I conducted during my PhD at the Univer-sity of Siena. The Centre for the Study of Political Change (CIRCaP)provided me with the most favourable conditions to draft my doc-toral dissertation. My gratitude goes in particular to my supervisor,Paolo Bellucci, for the time and patience he spent not only in dealingwith my research but also (and mostly) with myself. This project alsoprofited from a few periods I spent as visiting researcher at OxfordUniversity and GESIS, Cologne. The dissertation – by then entitledCauses and Electoral Consequences of the Personalization of Politics inWestern Democracies – was successfully defended in December 2011.It has been short-listed for the ECPR Jean Blondel PhD Prize and wonthe XIV Celso Ghini Prize awarded by the Italian Society for ElectoralStudies (SISE) for the best PhD dissertation defended in the biennium2011–2012.

It took me more than a year to turn my dissertation into anactual book. I was lucky enough to spend this year as Jean Monnetpost-doctoral fellow at the Robert Schuman Centre of the EuropeanUniversity Institute, Fiesole. I will remember this period as one of thehappiest of my whole life. Out of the almost infinite list of peoplethat should be acknowledged for this, I am particularly grateful toAlex Trechsel for his matchless mentorship – both on and off the aca-demic pitch. Being part of IUE Calcio and Marii & The Machine Heads

ix

October 25, 2013 14:53 MAC/PERSON Page-x 9781137270221_01_prex

PROOFx Acknowledgments

also contributed in making that year of my life unforgettable – thankyou all guys, it was Spettacolo Puro!

Of the extensive list of people who supported me throughoutthe various stages of development of this work, I would like toacknowledge in particular Mauro Barisione, Jean Blondel, MaurizioCotta, Andrea De Angelis, Michael Lewis-Beck, Ferdinand Muller-Rommel, Martin Rosema, Hermann Schmitt, Paolo Segatti, and LucaVerzichelli. Finally, I would like to thank Amber Stone-Galilee,Andrew Baird, and the production team at Palgrave Macmillan fortheir patience and assistance in guiding this manuscript through thefinal production phase.

I have greatly benefited from comments and criticism receivedat a variety of conferences and seminars. These include the 3rdECPR Graduate Conference (Dublin, 2010), the PSA Graduate Con-ference (Oxford 2010), the 3rd SISP Graduate Conference (Turin2011), the 34th ISPP Annual Meeting (Istanbul 2011), the 6thECPR General Conference (Reykjavik 2011), the 1st European Con-ference on Comparative Electoral Research (Sofia 2011), the 11thPoliticologenetmaal (Amsterdam 2012), the EUI’s Political BehaviourColloquium (Florence 2012), the WPSA Annual Meeting (Hollywood,CA, 2013), the 3rd EPSA General Conference (Barcelona, 2013), anda guest lecture at the Centre for the Study for Democracy/ZDEMO(Luneburg, June 2013).

Early versions of some of the analyses included here have beenpublished elsewhere. Preliminary versions of parts of Chapter 1appeared in ‘The Personalization of Politics in Western Democra-cies: Causes and Consequences on Leader-Follower Relationships’,The Leadership Quarterly, 22 (2011), and in ‘Changing Parties, Chang-ing Partisans: The Personalization of Partisan Attachments in WesternEurope’, Political Psychology, 34 (2013). An earlier version of Chapter 2appeared as ‘The Rise of Party/Leader Identification in WesternEurope’, Political Research Quarterly, 66 (2013). A preliminary versionof the analysis presented in Chapter 3 appeared in ‘Party and LeaderEffects in Parliamentary Elections: Towards a Reassessment’, Politics,32 (2012).


October 24, 2013 11:36 MAC/PERSON Page-1 9781137270221_02_int01

PROOF

Introduction

This book is about the determinants of voter choice in establishedparliamentary democracies. Unlike many existing studies, however,it will concentrate on the psychological drivers of voting. Thedecline of traditional social and ideological cleavages has, in fact,rendered progressively inadequate a purely sociological understand-ing of voting behavior (Blondel and Thiébault, 2010). Moreover,the long-term trends of secularization and enfranchisement of theworking class in the Western world (Franklin, Mackie, and Valen,1992) have been paralleled by the pervasive mediatization of thepolitical scene and the resulting tendency to portray politics in anincreasingly “personal” – rather than “partisan” – fashion (Garzia,2011; McAllister, 2007). Taken together, these occurrences have allcontributed to shaping the conventional wisdom that “election out-comes are now, more than at any time in the past, determined by vot-ers’ assessments of party leaders” (Hayes and McAllister, 1997: p. 3).

Although widely shared by journalists, politicians, andcommentators – as well as by ordinary citizens – such an argumenthas been fiercely contested by social and political scientists (for areview, see: King, 2002a; 2002b). Traditional interpretations of voting(on which the wide majority of academic research is grounded)emphasize the role of macro-social factors such as class, region,and religion and/or the resulting long-term social-psychological alle-giances with parties (i.e., party identification). In either case, voters’evaluations of the personality of individual politicians stand as asort of residual category, as they appear “strongly mediated by suchsituational factors as the strength as well as the direction of partisanaffiliation” (Brettschneider and Gabriel, 2002: p. 153). Virtually all

1


PROOF2 Personalization of Politics and Electoral Change

the available comparative evidence points, indeed, to the key roleplayed by party identification in orienting voters’ short-term atti-tudes and vote choices in turn (e.g., King, 2002a; 2002b; Curtice andHolmberg, 2005; Karvonen, 2010; Holmberg and Oscarsson, 2011).

This study challenges the “conventional wisdom of electoralresearch” (Midtbø, 1997: p. 143), moving from the assumption thatindividuals’ relationship with political parties depends largely onthe types of parties that are predominant in the party system ateach relevant point in time. Arguably, the profound transformationsundergone by traditional cleavage parties in the last decades shouldhave exerted an effect on the dynamics of party identification at theindividual level. Mass-based parties were characterized by a tight linkwith their respective social milieu (Lipset and Rokkan, 1967). In thecase of contemporary catch-all parties, the nature of this link can beexpected to differ substantially (Lobo, 2008). By showing that voters’party identification is no longer based on prior social and ideologicalidentities, but, rather, on individual attitudes towards more visiblepartisan objects and in primis party leaders, this study highlightsan alternative perspective on voters’ behavior in parliamentary elec-tions. If the growing role of party leaders as drivers of partisanship istaken into account (and properly modeled within the voting equa-tion) then their electoral effect emerges as much stronger than it hasusually appeared. Party leaders can gain (or lose) votes due to theway in which their personality profile is perceived by voters – andthis independently of the electoral effect exerted by voters’ long-termidentifications, ideological orientations, and retrospective economicevaluations. More often than not, voters’ evaluation of party leaderscan be decisive for the election outcome.

In order to strike a balance between needs for comparison andattention to national differences, this study will focus on threeestablished parliamentary democracies in Western Europe: Britain,Germany, and the Netherlands. The choice of these three countries –connoted by sharp differences in terms of electoral system, size ofthe party system, and structure of political competition – highlightsmany of the crucial variations in the structure of democratic politicsand allows a more broadly based assessment of the major researchhypotheses (see Table I.1).

The time frame under analysis spans the last five decades – that is,between 1961 (when the first national election study was conducted


PROOFIntroduction 3

Table I.1 Case selection

Britain Germany The Netherlands

Population 63,181,775 81,799,600 16,751,323Size (km2) 243,610 357,021 41,543GDP per capita(US$)

38,591 41,168 50,355

System ofgovernment

Unitaryparliamentaryconstitutionalmonarchy

Federal parliamentaryconstitutional republic

Unitaryparliamentaryrepresentativedemocracy underconstitutionalmonarchy

Universalsuffrage

1928 1918 1917

Electoral system First Past thePost

Mixed MemberProportionalRepresentation

Proportional

Main politicalparties (year offoundation)

ConservativeParty (1834)

ChristlichDemokratische Un.(1945)

ChristenDemocratisch Appèl(1977)

Labour Party(1900)

Christlich-SozialeUnion (1945)

ChristenUnie (2001)

LiberalDemocrats(1988)

Freie DemokratischePartei (1948)

Democraten 66(1966)

Die Grunen (1979) GroenLinks (1990)Die Linkspartei.PDS(1990)

Partij van de Arbeid(1946)

SozialdemokratischePartei (1875)

Partij voor deVrijheid (2006)StaatkundigGereformeerde Partij(1918)Socialistische Partij(1971)Volkspartij voorVrijheid enDemocratie (1948)

in Germany) and the most recent election for which national elec-tion study data is available at the time of writing (i.e., British andDutch elections of 2010). National election studies provide undoubt-edly the largest body of data upon which scholars have so far based



their analyses of leader evaluations (Bittner, 2011). The choice to stickto public opinion data has been proven to have several advantages(Dalton, 2008). Most notably, focusing on micro-level data allowsthe employment of a top-down approach based on the primacy ofthe political supply (Bellucci and Segatti, 2011), through which thechanging dynamics of voting behavior at the individual level can beinterpreted as a function of the structural transformations occurringat the party level.

The book’s focus on the role of political leaders as drivers ofvote choice in parliamentary elections fits well within the growingbody of literature on the personalization of politics. This is, indeed,where the coordinates of this study should be located.1 A studyof the psychological dynamics of vote choice has a double rele-vance. From the empirical point of view of political psychology, itallows the researcher to apply and test theories of human behaviorwithin the context of large-scale simultaneous decision-making pro-cesses. During elections, millions of people are called to the polls toexpress their political preferences as determined individually – andyet collectively – throughout the campaign. In turn, the outcome ofa collective process of this kind represents the most relevant implica-tion for democracy. Through elections, the psychological dynamicsunderlying voter choice at the individual level affect the balanceof power at the national level in ways that no other social behav-ior can even approach. In times of growing electoral instability, it isthus crucial to place under scrutiny the determinants of change atthe level at which change occurs: individual voters. It is no doubttrue that “huge empirical and statistical obstacles [must] be van-quished” (Midtbø, 1997: p. 152) in order to conclusively disentanglethe role of leaders from that of their parties as a determinant of vot-ers’ choice. As with every other endeavor of this sort, this book doesnot pretend to provide an indisputable answer. If anything, its meritis that of providing voting behavior scholars with an alternative –and hopefully more sound – way to look at leader effects in the ageof personalization.


October 24, 2013 11:56 MAC/PERSON Page-5 9781137270221_03_cha01

PROOF

1The Personalization of Politics

1.1 The causes of the personalization of politics inparliamentary democracies

Whether or not the dynamics of parliamentary elections havebecome more presidential (Mughan, 1993; 2000; Poguntke and Webb,2005; 2012), it is hard to deny that democratic politics is now morepersonalized than ever before. Indeed, the “popular focus on leadersnow appears commonplace across almost all of the major parlia-mentary systems, where parties once occupied the center stage”(McAllister, 2007: p. 572). Impressionistic evidence of this trendincludes the substitution of leader images for party symbols duringelection campaigns (Bowler and Farrell, 1992; McAllister, 1996;Swanson and Mancini, 1996; Farrell and Webb, 2000) and the media’sincreasing propensity to mention leading candidates rather than theparties they belong to (Dalton, McAllister, and Wattenberg, 2000;Mughan, 2000; Langer, 2007; Karvonen, 2010; Ohr, 2011). Executivesthemselves are portrayed in a personalized fashion, as they are rou-tinely labeled after the names of their leaders (Bean and Mughan,1989), and a stronger correlation over time between prime ministe-rial popularity and the executive’s public rating has been documentedin several parliamentary democracies (Lanoue and Headrick, 1994;McAllister, 2003; Campus and Pasquino, 2006).

A growing number of academic studies have concentrated on theincreasingly tighter relationship between political leaders’ personal-ity and the functioning of representative democracy, and in particularon the process of personalization of politics (Kaase, 1994; King, 2002a;Caprara and Zimbardo, 2004; McAllister, 2007; Rahat and Sheafer,2007; Adam and Maier, 2010; Blondel and Thièbault, 2010; Campus,

5



2010; Karvonen, 2010; Garzia, 2011; Kriesi, 2012). According to Rahatand Sheafer (2007), the personalization of politics should be seenas a process, in which “the political weight of the individual actorin the political process increases over time, while the centrality ofthe political group (i.e., political party) declines” (Rahat and Sheafer,2007: p. 65; italics in original). Similarly, Karvonen (2010) puts atthe core of his personalization hypothesis the notion that “individ-ual political actors have become more prominent at the expense ofparties and collective identities” (p. 4). In other words, the notion ofpersonalization does not only imply that individual politicians mat-ter in the political process – they are also assumed to matter morethroughout time.

In broad terms, the personalization of politics can be conceivedas the cumulative effect of the changes occurring in the reciprocalrelationships between the main actors of contemporary democraticpolitics: voters, parties, and the media. The progressive decline ofcleavage politics in established democracies and the resulting changesin voting behavior have, in fact, forced parties to adapt to the shiftingdemands of voters and restructure their electoral profile accordingly.The whole process is fostered by the changing structure of politicalcommunication due to the emergence of television and its inherent“media logic”. Indeed, the personalization of political news can beseen as a consequence of both media technologies and the changingstrategies of parties. In turn, such changes affected voters’ under-standing of politics and informational demands, further reinforcingthe personalization of political competition and the way this is por-trayed by the media. This complex web of inter-relationships isgraphically summarized in Figure 1.1.

Media and citizens. Few would cast doubts over the crucial roleexerted by electronic media, and television in particular, in thepersonalization of contemporary politics. Indeed, the empirical stateof research provides relatively ample support for the personaliza-tion hypothesis when it comes to political communication (fora review, see: Adam and Maier, 2010). The changing structure ofmass communications has been central in emphasizing the role ofpolitical leaders at the expense of parties, making the latter “moredependent in their communications with voters on the essentiallyvisual and personality-based medium of television” (Mughan, 2000:p. 129). A clear example of personalization comes from the increasing


PROOFThe Personalization of Politics 7

Voters

Parties

Media

Figure 1.1 The driving forces of the personalization of politics

importance gained by televised leaders’ debates during national elec-tion campaigns (Adam and Maier, 2010). Begun as a peculiar featureof presidential campaigns in the US in the 1960s, the leaders’ debatespread quickly all around Western parliamentary systems. Of 45democracies surveyed in the mid-1990s, only four did not feature aleaders’ debate during their last campaign (LeDuc, Niemi, and Norris,1996). As of today, a similar exercise would hardly find a singleexception to this rule.

On these grounds, personalization has been defined as “the moregeneral, pervasive, and fundamental element in the process ofchange of electoral campaigns” (Swanson and Mancini, 1996: p. 204).Television-based campaigning accentuates personality factors at theexpense of substantive programmatic goals (Campus, 2010). Becauseof its power to present images, it is easier for television to commu-nicate political information through physical objects such as leadersrather than through more abstract entities like political parties, man-ifestos, or ideologies (Owen, 1991; Graber, 2001; 2006; Hayes, 2009).At the same time, it has been noted that

[a]verage news consumers prefer to read about other people, notabout abstract groups or remote bureaucracies and governmentagencies. To cater to these preferences, news stories, especiallythose that appear on television, are routinely framed from thepoint of view of central actors. News consumers see an individual



in action and are given information about his or her feelings andreactions. Personal motives and mistakes are analyzed. Inevitably,stories about groups are transformed into stories about groupleaders.

(Davis, 1990: p. 169)

In turn, such a way of presenting politics has progressively affectedcitizens’ conception of the political process (Asp and Esaiasson,1996). One of the major consequences of the personalization of poli-tics would seem to lie in the role gained by political leaders as anchorsof voters’ interpretations and evaluations (Holtz-Bacha, Lessinger,and Hettesheimer, 1998). Under the hypothesis that changes in infor-mational input affect citizens’ political belief system (Ohr, 2011),empirical research has witnessed the widespread diffusion of leader-based political schemata among Western electorates, in both pres-idential (Miller, Wattenberg, and Malanchuk, 1986; Sullivan et al.,1990) and parliamentary systems (Pierce, 1993). Personalities doperform a relevant heuristic function in the opinion-making processof voters (Mondak, 1993; Baldassarri, 2013). In an increasingly com-plex political world, political leaders represent a relatively easy objectto understand, as they can be effortlessly evaluated using inferen-tial strategies of person perception that are constantly employed ineveryday life (Kinder, 1986; Rahn et al., 1990). Television consump-tion supports this tendency, as, unlike consumers of other media,“television viewers have access to visual imagery and nonverbal cuesthat often play an important role in shaping personality evaluationsof others” (Druckman, 2003: p. 561). Television’s ability to create a“false intimacy” between politicians and viewers (Hart, 1999) encour-ages the latter to evaluate the former “on the same terms a man greetsany new acquaintance” (Gould, 1972: p. 21).

Political parties and voters. On these premises, it is tempting tosee television as the prime mover behind the personalization ofpolitics (McAllister, 2007). Yet, the key role played by political par-ties themselves in the process must not be overlooked. In the lastdecades, parties have undergone deep transformations which are atonce cause and consequence of personalization. According to Blondeland Thiébault (2010),

[w]hile the process of “modernisation” continued throughout thetwentieth century, its effect was no longer to reinforce the social



character of the relationship between citizens and the politicalsystem, but, on the contrary, to reduce the weight of the socialstructure on the population as a whole: with the spread of educa-tion reactions of citizens were becoming more independent fromthe social group to which they have been attached.

(Blondel and Thiébault, 2010: pp. 1–2)

The widespread erosion of social and partisan attachments docu-mented in almost every advanced industrial democracy (Franklin,Mackie, and Valen, 1992; Dalton and Wattenberg, 2000) has madeit necessary for traditional cleavage parties to adjust their elec-toral strategies in order to extend their appeal beyond the socio-ideological cleavages to which they usually referred (Mair, Muller,and Plasser, 2004). As a result, West European party systems havewitnessed an unequivocal pattern of transformation on the behalf offormer class-mass (i.e., socialist) and denominational (i.e., Christian-democratic) parties, with both by and large converging on thecatch-all typology (Kirchheimer, 1966). This pluralistic ideal-type iscommonly distinguished by a “superficial and vague ideology, andoverwhelmingly electoral orientation” and, most notably, by the“prominent leadership and electoral roles of the party’s top-rankednational-level candidates” (Gunther and Diamond, 2003: p. 185; seealso Farrell and Webb, 2000). This process of transformation, pre-viewed by Downs (1957) and documented by Kirchheimer (1966),found its symbolical culmination in the fall of the Berlin Wall in1989 – a turning point after which parties could not be thoughtof as representing “bodies of particular principle”, but, rather, as“vote maximizing agents without any real ideologies of their own”(Daalder, 2002: p. 52). This development, in turn, made individualpoliticians increasingly relevant in patterns of political competi-tion as well as within their own party structures (Poguntke andWebb, 2005). A careful analysis of country-specific dynamics of partytransformation further highlights the unequivocal direction of thesedevelopments.

1.2 Patterns of party transformation in threeparliamentary democracies

This study concentrates on three established parliamentary democra-cies in Western Europe: Britain, Germany, and the Netherlands. Their



respective patterns of party transformation underline remarkablecommonalities that would seem to point to an unequivocal develop-mental trajectory. Among these three countries, the British case repre-sents possibly the best archetype of a historically class-based politicalsystem. In their pioneering study of British elections, Butler andStokes (1969) depicted the British electorate as two large and stableblocs, with working-class voters on the Labour side, and middle-class on the Conservative one. To the progressive class dealignmentof the 1970s (Crewe, Sarlvik, and Alt, 1977), both parties respondedby employing brand-new catch-all electoral strategies. The increasingvisibility of leaders within parties’ communication with voters hasbecome apparent since the Thatcher years (Blondel and Thiébault,2010). Yet, this development found its most notable example in themid-1990s with the crafting of the New Labour party, which, underTony Blair, cast off much of its ideological baggage – including thehistoric commitment to the public ownership of major industries –and transformed itself into an “exemplar of the modern electoral-professional competitor” (Webb, 2004: p. 44). From an organizationalpoint of view, all the major British parties (including the LiberalDemocrats) are now characterized by extremely high levels of leader-ship autonomy and a thoroughly professionalized approach to polit-ical marketing that results in ever more leader-centered campaignstrategies (Denver, 2007).

Similarly to Britain, the Netherlands has historically representedan example of cleavage-based polity. Dutch civil society has longbeen founded on pillars, and virtually all areas of social life, includ-ing politics, were organized along the principles of class and religion(Andeweg and Irwin, 2003). Accordingly, the voters’ relationshipwith parties was based on their belonging to the pillars, thus leav-ing little room for politicians’ personalities to affect their politicalattitudes and behavior (Irwin and van Holsteyn, 1989). However,the erosion of pillars and the resulting deterioration of traditionalbonds between parties and voters have led Dutch parties to reshapetheir appeal to increasingly volatile voters by highlighting “the qual-ities of individual politicians”, and most notably the “managerialskills of their prime ministerial candidates” (Fiers and Krouwel, 2005:p. 151). A trend towards increasing personalization in general, andan increasing concentration of the public attention on a limited setof personalities, was already noticeable in the 1990s (Kriesi, 2012:



p. 841). However, the critical step towards the personalization ofDutch politics is in all likelihood represented by the 2002 elec-tion, which heralded Pim Fortuyn’s entrance into the political scene.His flamboyant rhetoric gained him unprecedented attention in themedia (Kleinnijenhuis et al., 2003), in spite of the fact that he was anoutsider with no formal party or political position (de Graaf, 2010).Fortuyn was credited with changing the Dutch political landscape(Pellikaan, de Lange, and van der Meer, 2007), and in particular theway that politics was presented to the public (Koopmans and Muis,2009). Nowadays, it is common for Dutch campaigns to be depictedas horse races between the main parties’ leaders (Fiers and Krouwel,2005). Correspondingly, party structures have eventually convergedaround a small group of party leaders as key decision-makers withinthe party (Andeweg, 2000).

In terms of institutional features – most notably, the electoral sys-tem and the role of the executive in the constellation of politicaldecision-making – the case of Germany can be seen as a “middle way”between the two aforementioned cases (Wagner and Weßels, 2012).Here, the crucial role of political parties in the constitutional settingis, in fact, paralleled by the dominant figure of the chancellor in thesystem of governance (Saalfeld, 2000). After a long period of balancebetween the two, the last decades have witnessed a marked declinein the public image of political parties (Arzheimer, 2006) and a corre-spondingly growing exposure of the chancellors (as well as of individ-ual candidates to the chancellorship) at the expense of their parties,especially during electoral campaigns. Although German campaignshave always been centered on candidates to some extent, it was onlyin the 1990s that the notion of personalization was, for the firsttime, discussed at length (Brettschneider and Gabriel, 2002). Thecharismatic figure of Helmut Kohl has been crucial in this respect,and his persistence as a figure on the political scene eventually ledhis contenders and successors to follow suit. This was particularlyevident in the 2002 campaign, which large parties focused “almostexclusively” on their chancellor-candidates (Poguntke, 2005). Theincreasingly central role of the personality features of political leadersin their parties’ appeal is corroborated by analyses of party structures,which testify to an unambiguous adoption of leader-centered elec-toral strategies on behalf of the main German parties (Gunther andDiamond, 2003).



1.3 Electoral consequences of the personalization ofpolitics: Empirical evidence so far

The previous sections have highlighted, as a common denominator,the intertwined effect of (i) technological innovations in the realmof political communication, and (ii) the structural transformationsin the electoral market as drivers of party transformation.1 One ofthe consequences of this process would seem to lie in the increas-ing centrality of individual party leaders within party structures andpatterns of electoral competition altogether (Poguntke and Webb,2005). Hypothetically, this should, in turn, have made more relevant“the role of individual politicians and of politicians as individualsin determining how people view politics and how they express theirpolitical preferences” (Karvonen, 2010: pp. 1–2). According to thisinterpretation, contemporary voters are thought to “vote differentlyfrom one election to another, depending on the particular personscompeting . . . Voters tend increasingly to vote for a person and nolonger for a party or a platform” (Manin, 1997: p. 219). Others go asfar as contending that “election outcomes are now, more than at anytime in the past, determined by voters’ assessments of party leaders”(Hayes and McAllister, 1997: p. 3).

Clearly, how far voters will base their voting decision on the per-sonal profile of the contenders depends heavily on the political andinstitutional structure in which an election is fought. Presidentialelections encourage focus on personalities to a greater degree than doparliamentary ones. A number of comparative analyses corroboratethis contention, and highlight the differentiated electoral impact ofleaders in presidential systems as compared with parliamentary ones(McAllister, 1996; 2007; Curtice and Hunjan, 2011; Holmberg andOscarsson, 2011; Tverdova, 2011). Such disparity can be explainedby the fact that the executive authority in presidential systems“resides with an individual who is [directly] elected to the position”(McAllister, 2007: p. 575). Evidently, the electoral system employedfosters personalization, as voters in majoritarian systems are askedto vote for a person. A survey by Sabato showed that 92 per cent ofAmericans were in agreement with the sentence “I always vote forthe person who I think is best, regardless of what party they belongto” (Dalton and Wattenberg, 1993: p. 203). What matters most, asargued by Wattenberg (1991), is that American voters “have not only



increasingly said that they vote for the man rather than the party;they have actually done so with great frequency” (p. 158).

As opposed to presidential settings, parliamentary systems arebased on the responsible party government model (Schattschneider,1942; APSA, 1950). In such a model, political leaders are not supposedto play much of a role. Parties compete on the basis of a number ofpolicies they promise to enact if they win elections. It is party plat-forms, then, rather than party leaders, that drive votes (McAllister,1996). Such an interpretation of voting does not leave much roomfor party leaders – as well as other short-term forces, such as issueproximity, and retrospective performance evaluations – to exert anindependent effect on individual vote choices. The electoral impactof leaders is further constrained by the legislative nature of parlia-mentary elections, in which voters face “a structural situation wherethe crucial choice is between parties rather than the personal standsand qualities of prime ministerial candidates” (Dalton, McAllister,and Wattenberg, 2000: p. 51). After all, parliamentary electionsremain nation-wide contests between parties (Mughan, 2012). Eventhough the identity of the potential candidates to the premiershipis well known to the public, a number of problems, both theoreticaland methodological, make it extremely difficult to isolate a vote fora leader from a vote for his party (Barisione, 2009).

In order to elaborate the theoretical argument on which the fol-lowing empirical analyses are based, it is worth taking stock of thestate of the literature on leader effects in the three parliamentarydemocracies under study, its common underlying assumptions aswell as its most notable shortcomings. In terms of case selection, ourthree countries provide a fair amount of variation when it comes tothe (potential) impact of leaders on voters’ electoral choice. Amongthese three countries, Britain would seem the most likely to witnessstrong leader effects on voting. For one thing, the historical domi-nance of prime ministers in Westminster systems as compared withmultiparty governments is generally thought to advantage leadereffects within the former institutional setting (Blais, 2011). At thesame time, the majoritarian electoral law currently in place andthe resulting size of the British party system have been shown toboost leader effects as compared with more “crowded” party systems(Curtice and Holmberg, 2005; Bittner, 2011; Curtice and Hunjan,2011). Academic interest in the effect of leaders’ personalities on



British voting behavior dates back to the pioneering research byButler and Stokes (1969; 1974). Although they did not exclude inprinciple the possibility for leaders’ personal characteristics to affectindividual voting behavior, their research highlights that in the early1970s “attitudes towards the parties were a better guide to votingbehavior than were attitudes towards the leaders” (Butler and Stokes,1974: p. 367). Two decades later, the seminal article by Bean andMughan (1989) did not substantially amend Butler and Stokes’ con-clusions. Albeit showing that in the British election of 1983 leaders’personality traits did have a discernible impact on vote choice,Bean and Mughan observe that, once voters’ party identification isincluded in the regression model, “leadership qualities can be seen tocontribute between four and five percentage points to the explainedvariance” (Bean and Mughan, 1989: p. 1172; similar conclusions arereached by Graetz and McAllister, 1987).

A study by Stewart and Clarke (1992) provides more convincingsupport for the personalization hypothesis, concluding that in 1987“leader images, net of pre-campaign party identification, had largeeffects” (Stewart and Clarke, 1992: p. 467). The same view is sharedby Mughan (2000) with respect to the Conservatives’ narrow victoryof 1992, when, according to the author’s estimation, widely posi-tive evaluations of John Major made the difference “between theformation of a majority Conservative government and a hung Par-liament” (Mughan, 2000: p. 114). A seemingly emerging consensusover the nontrivial importance of party leaders in British electionswas, however, not paralleled by systematic efforts to grasp the growingrelevance of leaders. Mughan’s (2000) analysis stands as an exception,but his conclusions are widely disputed in the few other longitudinalanalyses available (Crewe and King, 1994a; 1994b; Bartle and Crewe,2002).

In the last decade, the literature on valence politics (Clarke et al.,2004; 2009a; 2011) has somehow magnified the role of party leadersas drivers of British voter behavior. According to the valence politicsmodel, vote choices in contemporary British elections are based on“people’s judgment of the overall competence of the rival politicalparties. The judgments, in turn, are arrived at through two principaland related short-cuts: leadership evaluation and party identifica-tion” (Clarke et al., 2004: p. 9). In their analysis of the 2001 election,Clarke and colleagues conclude that Labour won not by virtue of



their leader’s appeal to the electorate, but, rather, because of their“commanding lead in party identifiers” (ibid., p. 123). All in all,British party leaders would seem to matter, albeit with two majorcaveats: they do not matter more than before, and they do not appearto matter that much once the seemingly imposing role of party iden-tification is taken into account within the voting model. The stateof the literature on leader effects in Britain is aptly summarized byBartle and Crewe (2002), who conclude that

[g]iven our information about a voter’s social background, partisanpredispositions, policy preferences, and evaluations of nationalconditions, their assessment of the attributes of party leadersadded very little to our ability to predict how they voted.

(Bartle and Crewe, 2002: p. 93)

The same conclusions would seem to fit well also in the case ofGermany. Indeed, the relative importance of party identificationsvis-à-vis leader evaluations for individuals’ vote choice appears evenmore accentuated in the German context. Both Pappi and Shikano(2001) and Schoen (2007) found relatively strong evidence of leadereffects in some elections but not in others. Comparable findingsare reported in Kaase’s (1994) analysis of the eight federal electionsheld between 1961 and 1987. As the author concludes, “[n]ot onlyis the short-term (issue and candidate) component of the votingintention in comparison to partisanship insubstantial (in terms ofexplained variance); it also does not systematically increase in impor-tance over time, as a personalization concept would require” (Kaase,1994: p. 226).

In light of the parallel decline of religious voters (the traditionalelectoral base of the Christian Democratic Union) and of manualworkers (Social Democratic Party’s historical supporters) witnessedduring the 1970s, some analysts of German politics hypothesized thatpolitical influences would have eventually become “more importantin future elections” (Pappi and Mnich, 1992: p. 208). Among thesepolitical influences, the potential role of party leaders has been high-lighted as a consequence of parties’ increasingly common strategyto personalize political communication during election campaigns(Lass, 1995). However, more recent assessments of leader effects inGerman elections do not seem to alter the picture drawn by Kaase



nearly 20 years ago. In their longitudinal analysis of the periodbetween 1961 and 2005, Brettschneider, Neller, and Anderson (2006)conclude that “the evaluations of the candidates for chancellor playonly a small role on the behavior of voters” (p. 495). According toHolmberg and Oscarsson (2011), the effect of chancellor candidateson the individual voting calculus of German voters can be quanti-fied as circa one-third of that exerted by party evaluations. As in theBritish case, the reason for such a weak effect is to be found in thedominant role of party identification in the cognitive structures ofvoters. As Brettschneider and Gabriel (2002) put it, “the influence ofcompeting party leaders is strongly mediated by such situational fac-tors as the strength as well as the direction of partisan affiliation”(p. 153).

Contrary to Britain and Germany, the Netherlands is hardly a casein point when it comes to personalization of voting behavior. Thepure proportionality of the electoral system, the resulting size of theparty system and the practice of coalition governments have all con-tributed in making less “plausible” an interpretation of the vote basedon leader evaluations – as further testified by the relatively scarceamount of available literature on the topic. Indeed, Dutch votingbehavior has long been interpreted as an expressive act of politicalparticipation, with voters grounding their electoral choice on theparty that belongs to their pillar (Andeweg and Irwin, 2003). Theprogressive decline of pillars in the last decades moved the focus tothe likely growth of other factors, such as party leader evaluations,within voters’ decision-making. Van Wijnen’s (2000) longitudinalstudy determined that overall evaluations of party leaders and cit-izens’ faith in their performance as prime ministers have had anincreased impact on the vote since 1986. However, a reanalysis ofthe same data by Aarts (2001) showed that candidate evaluations inthe Dutch context are heavily influenced by party perceptions. Onceparty evaluations are taken into account, no substantial increase incandidate-centered voting can be discerned.

The entrance of Pim Fortuyn on the political scene and the unex-pected success of his party list in the 2002 election fostered the debateon leader effects in Dutch parliamentary elections (van Holsteyn andIrwin, 2003), albeit with inconclusive evidence. According to vanHolsteyn, Irwin, and den Ridder (2003), the results of the May 2002



election should be interpreted in light of voters’ proximity to theissues that Fortuyn brought to the forefront of the Dutch politicalagenda, rather than by personality evaluations. More recent studiesdo not alter the conclusion that party leaders play only a secondaryrole in shaping Dutch voting behavior, which appears to be bestexplained by voters’ evaluation of the parties they belong to (vanHolsteyn and Andeweg, 2010).

The large amount of empirical evidence stemming from theaforementioned case studies can be summarized into two majorconclusions: (i) political leaders do not matter that much in thedecision-making of process voters, and (ii) they have not come tomatter more throughout time. These conclusions find further sup-port in the few available comparative analyses of leader effects indemocratic elections. As to the latter conclusion, a meta-analysis ofthe available voting literature by Karvonen (2010) finds hardly anyevidence of a personalization trend at work in established Westerndemocracies. As Curtice and Holmberg (2005) bluntly put it in theiranalysis of six parliamentary democracies, “voters’ evaluations ofparty leaders appear to be as important or unimportant now as theywere when they were first measured . . . Nothing much seems to havechanged” (pp. 250–251). An empirical reassessment of leader effectsby Holmberg and Oscarsson (2011), taking into account a wider num-ber of countries and a longer time span, do not seriously affect suchconclusions. The authors even resort to the “myth” metaphor withregard to the alleged rise of leader effects on voting in Western parlia-mentary democracies: “[t]he argument that leaders, over time, havebecome a greater influence on the vote is simply not substantiated”(Holmberg and Oscarsson, 2011: p. 50).

Not only is the literature skeptical about the increasing relevanceof party leaders to voting throughout time. As a matter of fact, themajority of available works would seem to cast serious doubts overthe very existence of significant leader effects, as summarized byKing’s (2002b) contention that

the almost universal belief that leaders’ and candidates’ personali-ties are almost invariably hugely important factors in determiningthe outcomes of elections is simply wrong.

(King, 2002b: p. 216)2



1.4 Linking party change with changes in votingbehavior: Theoretical framework

Against the common wisdom that sees popular party leaders as afundamental asset for their parties, the response of the scientificcommunity appears almost unanimous in downplaying the electoraleffects of leader images in parliamentary elections. As the abovereview of the literature suggests, such a unanimous response can beimputed to the theoretical and methodological framework employedin these studies.

As a common denominator, the present literature relies, in fact, onthe classic social-psychological interpretation of voting behavior setforth nearly half a century ago in The American Voter (Campbell et al.,1960). In such a model, vote choices are conceived as a function of“the cumulative consequences of temporally ordered sets of factors”(Miller and Shanks, 1996, p. 192). At the heart of this model lies thenotion of party identification – a long-term affective orientation to apolitical party, which is rooted in early socialization and based on anobjective location in the social structure (Campbell et al., 1960). Dueto its social-psychological nature, party identification is conceived asan unmoved mover: that is, a pre-political attitude that is nonethe-less able to shape the individual’s political world-view in a way thataccords with their partisan orientation (Johnston, 2006). On thesebases, partisanship is thought to be a cause (but not a consequence)of less stable attitudes and opinions about issues and candidates.In other words, both the act of voting and its more proximate influ-ences (i.e., leader evaluations) are subject to explanation in terms oftemporally and causally prior partisan identifications.

An “orthodox” understanding of this theoretical approach some-how forces researchers to investigate leader effects adopting theso-called “improved-prediction strategy”. This strategy consists insequentially entering variables in the model according to the sup-posed causal impact of voting determinants, from long-term toshort-term ones (King, 2002b; see also the various country chaptersin King, 2002a). The operational choice to enter the leader variableat the very last stage of the empirical analysis (as is often the casein the available voting literature) has clear implications for the actualassessment of the impact of leaders on the vote. However, such choicecan be justified only as long as partisan identifications are effectively



rooted in the realm of society (e.g., early socialization, placementin the socio-economic structure) and thus relatively immune to theeffect of short-term political forces (i.e., party leader evaluations).

The present study brings forward the idea that the transformationsat the party level have ignited a related change in the nature of theirbonds with the electorate. Individuals’ relationship with political par-ties depends largely on the types of parties that are predominantin the party system at each relevant point in time (Crewe, 1976;Gunther and Montero, 2001; Gunther, 2005; Lobo, 2008). Indeed,previous empirical studies have shown how specific party characteris-tics contribute to distinctive types of partisanship (Richardson, 1991;Garzia and Viotti, 2011; Garzia, 2013a; 2013b). Mass-based partieswere characterized by a tight link with their respective social milieu(Lipset and Rokkan, 1967; Butler and Stokes, 1974; Thomassen,1976). In the case of catch-all parties, the nature of this link can beexpected to differ substantially. More specifically, it can be hypothe-sized that, as a result of the process of party transformation, partisanloyalties have shifted accordingly from a mere reflection of previ-ous socio-ideological identities to the result of individual attitudestowards more visible partisan objects – and, in particular, their lead-ers. This contention is based on a number of related occurrences, suchas the widespread practice of candidate-centered campaigning on thebehalf of catch-all parties (Swanson and Mancini, 1996; Farrell andWebb, 2000), the resulting increased influence of leaders in shapingthe appeal of their own party (Curtice and Holmberg, 2005; Barisione,2009; Curtice and Hunjan, 2011), but also the growing tendencyamong voters to evaluate politics in “personal” rather than “parti-san” terms (Miller, Wattenberg, and Malanchuk, 1986; Rahn et al.,1990; Sullivan et al., 1990; Pierce, 1993; Campus, 2000).

Taken together, all these occurrences provide reasons to believethat partisan attachments have become increasingly connected tovoters’ attitudes towards party leaders. In such a context, it may wellbe that parties’ appeal to voters has come to be increasingly shapedby their own leaders’ image (Blondel and Thièbault, 2010). The evermore evident overlap between the image of the party and the imageof the leader has eventually blurred the contour between the two.Nowadays, leaders can be thought not only to represent their partiesin the public sphere – it can be argued, rather, that political leadershave become important in their own right “by personifying the policy



platforms of their respective parties” (McAllister 2007, p. 574; italicsmine). A better understanding of the extent to which this is actu-ally the case is crucial for any unbiased, systematic analysis of leadereffects on the vote throughout time and countries. After all,

[i]f leaders become the prism through which the message of theirparty is communicated . . . then we might fail to ascertain whetherleader evaluations have had a greater influence simply becausepeople’s evaluations of the parties have become increasingly influ-enced by their evaluation of the leaders. Such a trend would meanthere is increasingly little left for leader evaluations to explain.

(Curtice and Holmberg, 2005: pp. 248–249)

Against this background, the aim of the present study is to out-line an alternative framework for the analysis of voting behavior inparliamentary democracies in light of the ongoing process of person-alization of politics. This framework employs a top-down approachbased on the “primacy of the political supply” (Bellucci and Segatti,2011: p. 10) through which the changing dynamics of voting behav-ior at the individual level can be interpreted as a function of thestructural transformations occurring at the party level. The assump-tion, on which the whole study is based, is that such changesin the political supply (i.e., increasingly candidate-centered) musthave exerted an effect on the dynamics of individual voting behav-ior. Most notably, this study will expand on the notion of leadereffects by going beyond their “residual” influence and looking at theincreasing relevance of party leader evaluations as drivers of partisanattachments at the individual level – what I shall call the attitudinalconsequences of the personalization of politics (see Figure 1.2).

The first, and probably most notable, implication of the changingdynamics of partisan attachments at the individual level pertains tothe relative place of partisanship and leader evaluations in the elec-toral calculus of voters. To the extent that attitudes towards partyleaders have become one of the strongest determinants of feeling ofcloseness to the parties, the analyses presented in the chapters thatfollow highlight the possibility for party leaders to bear a strong effecton voters’ behavior indirectly, that is, through partisanship.

As said, the profound changes in the electoral supply have ren-dered traditional social-psychological interpretations of the vote



Attitudinal consequences

Dependent variable

Partisanship

_____________

Chapter 2

Behavioralconsequences

Dependent variable

Vote choice

_____________

Chapter 3

Electoral consequences

Dependent variableElection outcome

_____________

Chapter 4

Figure 1.2 The consequences of the personalization of politics: Plan of thebook

progressively less appropriate, insofar as the inherent endogeneitybetween partisanship and leader evaluations is not taken intoaccount. On these bases, placing parties and leaders at the same stepof the causal sequence is not enough to appraise their relative con-tribution to the final voting decision. Therefore, the third chapterof this book is devoted to a reassessment of the relative effect ofpartisanship and leader evaluations within voters’ electoral calculus.As reciprocal causation is at work, single-equation models of votingare apt to provide seriously biased estimates. Such specification ofthe statistical model “would understate the final impact of leaders’images by misattributing to party identification . . . a portion of lead-ership’s direct effects” (Dinas, 2008: p. 508). As a result, the “effectsof partisanship on the vote are likely to be exaggerated” (Marks,1993: p. 143). By employing the proper econometric procedures (i.e.,two-stage estimation and instrumental variables) new light will beshed on the behavioral consequences of the personalization of poli-tics. If endogeneity is taken into account, then the electoral effect ofleader evaluations at the individual level appears not only strongerthan habitually observed but also on the rise throughout time – asthe personalization hypothesis implies.

After having assessed the relative weight of partisanship andleader evaluations within the individual voting calculus, the fourthand last empirical chapter of the book will concentrate on oneof the crucial issues in voting behavior research – namely, the



effect of party leaders’ personality on the outcome of democraticelections. The methodology employed is the counterfactual thoughtexperiment (Graetz and McAllister, 1987; Bean and Mughan, 1989;Crewe and King, 1994a; 1994b; Jones and Hudson, 1996; Bartels,2002; Johnston, 2002; Marsh, 2007; Dinas, 2008; van Holsteyn andAndeweg, 2010; Aarts and Blais, 2011; Bittner, 2011). By puttingemphasis on explicit “What if?” questions, this strategy magnifiesthe electoral effect of a specific party leader by forecasting the extentto which the electoral outcome would have changed had his or herpersonality been perceived differently by voters. The time span con-sidered covers the last three decades and a total of 20 elections.According to the results, 10 out of the 20 electoral contests underanalysis might have been decided by voters’ evaluation of the partyleaders – and this independently of the electoral effect exerted byvoters’ partisan identifications, ideological orientations, and retro-spective economic evaluations. When it comes to aggregate effects onthe election outcome, their impact appears central as well. In turn,this occurrence raises serious concerns from a normative point ofview and leads to the question of whether candidate-centered votingis good for democracy. To this issue is devoted part of the fifth andlast chapter of this book, where competing conceptions of candidate-centered voting are presented and assessed against the availableempirical evidence.

In light of the noteworthy implications of personalization onthe functioning of contemporary parliamentary democracies, theremainder of Chapter 5 will discuss a number of methodological chal-lenges for forthcoming scholarship on the personalization of politicsand its electoral consequences. The role that political psychology canplay in furthering our understanding of voting behavior in the ageof personalization will be repeatedly underlined as a memento forfuture research in the field.



PROOF

2Attitudinal Consequences

2.1 From party identification to partisanship

Classic studies of political attitudes and behavior have beendominated by approaches that emphasize the role of macro-socialfactors such as class, religion, and territory (Lazarsfeld, Berelson, andGaudet, 1944; Berelson, Lazarsfeld, and McPhee, 1954; Campbellet al., 1954; Lipset and Rokkan, 1967). Yet, the decline of traditionalcleavage structures and their ability to shape the political competi-tion (Franklin, Mackie, and Valen, 1992) has rendered sociologicalinterpretations of the relationship between the parties and their vot-ers increasingly problematic. Social modernization and technologicalinnovations, in particular, have led to new forms of interest repre-sentation and political communication that affected to a substantialextent the role of parties in the democratic process, thus altering theirrelationship with citizens.

In spite of the profound changes undergone during the last decade,political parties have remained the vital link between various ele-ments of the democratic process: they create identities, frame elec-toral choices, and determine the outputs of government. In thissense, democracy without political parties is still “unthinkable”(Schattschneider, 1942; Dalton and Wattenberg, 2000). Therefore, aslong as party-based democracies are around, “people’s different rela-tionships with the major actors – the parties – must be conceptualizedand measured” (Holmberg, 2007: p. 566).

Against this background, the enormous scholarly attentiondevoted to the concept of party identification comes as no surprise

23



(Campbell et al., 1960: pp. 120–167). At the heart of this enduringinterest lies the fundamental observation that voters have some kindof generalized predisposition to support a particular party over time(Miller, 1991). Although virtually all scholars agree on the need toaccount for these predispositions, there is widespread disagreementabout its causes and how these should be interpreted and measured(Bartle and Bellucci, 2009a).

In its classical formulation, party identification was conceived as“the individual’s affective orientation to an important group objectin his environment” (Campbell et al., 1960: p. 121). According tothe social-psychological reading, such orientation is rooted in earlysocialization and based on primary group memberships (e.g., race,religion, social class). Among its crucial features, party identificationwas said to be stable – that is, virtually immune from the effect ofshort-term forces – and it was thus considered to be a cause (butnot a consequence) of less stable attitudes and opinions about, forexample, candidates and issues (Johnston, 2006). As explained by theauthors of The American Voter, “the influence of party identificationon perceptions of political objects is so great that only rarely will theindividual develop a set of attitude forces that conflicts with this alle-giance” (Campbell et al., 1960: p. 141). To put the case more sharply,the social-psychological approach conceives party identification as“an exogenous variable affecting politics but not being affected bypolitics” (Holmberg, 2007: p. 563).

However, it did not take much time before severe criticisms arosewith respect to the supposed stability of party identification. Mak-ing use of richer datasets and increasingly sophisticated statisticaltechniques, later analyses showed that partisan ties at the individ-ual level were much more unstable than originally thought, and,indeed, strongly responsive to those short-term forces that they werethought to cause (Page and Jones, 1979; Fiorina, 1981; Franklin andJackson, 1983). Moreover, sources of scholarly disagreement were notlimited to the debate between Michigan scholars and the “revision-ists” (Fiorina, 2002). Another serious matter of dispute was relatedto the applicability of the concept outside the US. In fact, the veryexistence of partisan identifications in European multi-party systemswas at the core of many critical chapters included in Party Identifica-tion and Beyond (Budge, Crewe, and Farlie, 1976). The cross-nationalapplicability of the concept was especially contested in Thomassen’s


PROOFAttitudinal Consequences 25

(1976) most celebrated chapter (but see also: Crewe, 1976; Inglehartand Klingemann, 1976).

As a result of the joint endeavor of the US and European schol-ars, the debate has switched the attention from party identificationto partisanship more generally. Loosely defined as “the tendency tosupport one party rather than another” (Bartle and Bellucci, 2009a:p. 1), partisanship has remained at the core of electoral researchon both sides of the Atlantic in the last decades (Richardson, 1991;Holmberg, 1994; Schmitt and Holmberg, 1995; Miller and Shanks,1996; Dalton and Wattenberg, 2000; Fiorina, 2002; Berglund et al.,2005; Johnston, 2006; Holmberg, 2007; Bellucci and Bartle, 2009a;2009b; Clarke et al., 2009b; Schmitt, 2009).

Many routes can lead voters to think of themselves as “partisans”.However, the great majority of the recent literature on partisanshipseems to largely converge on an understanding of the concept basedon modern attitude theory (Bartle and Bellucci, 2009a). Accordingto this perspective, partisanship is best interpreted as a “psycholog-ical tendency that is expressed by evaluating a particular entity [theparty] with some degree of favor or disfavor” (Eagly and Chaiken,1993: p. 1). Such an attitudinal interpretation of partisanship isespecially useful insofar as it entails the possibility for voters tosimultaneously develop attitudes towards more than one party, thusfavoring its applicability to European multiparty systems (Pappi,1996).

The aim of this chapter is to shed light on the (changing) determi-nants of partisanship in Britain, Germany, and the Netherlands acrossthe last four decades, focusing in particular on the role played byparty leader evaluations. By means of multivariate statistical tests, itwill show the increasingly stronger relationship between party leaderevaluations and feelings of closeness to parties at the individual level.The results are presented and discussed after a review of the relevantliterature and the description of the various operational measuresincluded in the analyses.

2.2 The attitudinal drivers of partisanshipin Western Europe

Among the possible sources of favorable attitudes towards the parties,the literature assigns a crucial place to issue preferences. The standard



model of rational decision-making based on issues, as applied tothe study of voting behavior, is the spatial model developed byAnthony Downs (1957). Voters and parties are placed on a left–rightcontinuum – a “super-issue which summarizes the programmes ofopposing groups” (Inglehart and Klingemann, 1976: p. 244). In sucha model, issue proximity is responsible for the promotion of posi-tive (or negative) attitudes towards each of the parties (Budge, Crewe,and Farlie, 1976) and, in turn, strongly influences voters’ choices(Dalton, 2008). Another important source of attitudes towards partiesare valence issues – where there is a wide consensus over what goalsare desirable, but there is a conflict over which party is most able todeliver them (Stokes, 1963; 1992). Attitudes can derive in this casefrom either retrospective evaluations of party performance (Fiorina,1981) or prospective competence assessments (Bellucci, 2006). Aggre-gate partisanship rates have also been shown to respond to thestyle of electoral competition in a country and the politicization ofthe respective electorates (Holmberg, 1994; Schmitt, and Holmberg,1995; Berglund et al., 2005; Schmitt, 2009).

As opposed to issue-related considerations, the literature hasdevoted relatively little attention to the role played by personalityassessments in the dynamics of partisanship at the individual level.Early research bears witness to the possibility that leader evaluationscan shape (or at least affect) voters’ party identification. Already in1968, V. O. Key anticipated a later, cognitive view of partisanship,contending that “[l]ike or dislike of a political personality . . . bringshifts in party identification” (Key, 1968; quoted in Clarke et al., 2004:p. 27). In their seminal contribution, Page and Jones (1979) pro-vide empirical evidence that party loyalties “do not function purelyas fixed determinants of the vote; those loyalties can themselves beaffected by attitudes toward the current candidates” (Page and Jones,1979: p. 1088).

The lack of further assessments of the role of party leaders asdrivers of partisanship in more recent decades is all the more surpris-ing in light of the progressive personalization of politics in Westerndemocracies, whose beginnings are traced right back to the early1980s (Bean and Mughan, 1989; McAllister, 1996). At the core ofthe personalization hypothesis lies the notion that “individual polit-ical actors have become more prominent at the expense of partiesand collective identities” (Karvonen, 2010: p. 4). The idea of anincreased prominence of individual politicians at the expense of



collective identities – on which traditional partisan identificationsare supposedly based – has clear theoretical implications for ourunderstanding of partisanship, and it would seem to link well withestablished theories of party–voter relationships. Building on previ-ous lines of research, it can be assumed that individuals’ relationshipwith political parties depends largely on the types of parties thatare predominant in the party system at a given point in time(Richardson, 1991; Gunther and Montero, 2001; Gunther, 2005;Lobo, 2008; Garzia and Viotti, 2011; Garzia, 2013a; 2013b).

Voters’ identification with European mass-based parties wasstrongly mediated by the former’s belonging to separate social milieusand subcultures (Lipset and Rokkan, 1967; Butler and Stokes, 1969;Thomassen, 1976). This contention, however, does not seem to holdfor contemporary catch-all parties. Indeed, the paramount relevanceacquired by party leaders within patterns of political communica-tion and electoral competition altogether has led some scholars tocontend that contemporary political leaders do not only lead theirparties: to a certain extent, they personify them (McAllister, 2007;Barisione 2009; Blondel and Thièbault, 2010; Garzia, 2011). On thebasis of the assumption postulating partisanship as a function ofparty characteristics, it seems plausible to envisage a strong associa-tion between individuals’ partisanship and their assessment of partyleaders. Indeed, this relationship can be hypothesized to have grownstronger throughout time – as the personalization hypothesis wouldimply.

Personalization has not only affected parties. From a political psy-chology perspective, one of its crucial consequences lies in the pivotalrole achieved by political leaders within voters’ cognitive frameworks(Baldassarri, 2013). Empirical research shows that the most diffusepolitical schema among contemporary voters is that based on leaders(Miller, Wattenberg, and Malanchuk, 1986; Sullivan et al., 1990). Thereason is clear: ideologies, issues, and performance assessments areinherently political, and thus require more sophistication to imple-ment (Shively, 1979; Pierce, 1993). Party leaders, on the contrary, canbe easily evaluated using inferential strategies of person perceptionthat are constantly employed in everyday life (Kinder, 1986; Rahnet al., 1990). Relying on implicit personality assessments, individualsare thus able to determine new judgments based on an overall char-acter appraisal when more concrete cognitions are required (Greene,2001). Accordingly, it can be hypothesized that, among all possible



sources of attitudes towards parties (i.e., leader evaluations, issueproximity, performance assessments), those related to their leadershave by and large gained prevalence.

2.3 Data and methods

The main data sources employed throughout the volume are theseries of national elections studies in our three countries, pooledby country and for each decade (for detailed study description anddataset construction, see: Appendix A).

Ever since The American Voter, empirical analyses of partisanshiphave by and large resorted to the “classic” seven-point measurementscale (Campbell et al., 1960; Fiorina, 1981; Bartle and Bellucci, 2009a).In order to make this operational measure applicable to Europeanmulti-party systems, however, one would be forced to narrow downthe analysis to the main two parties in each country. As the percent-age of identifiers with these parties has tended to decline over time(albeit with the partial exception of British Labour; see: Table 2.1),the “middle” category would be artificially conflated by featuring notonly true independents, but also respondents identifying with minorparties – an occurrence that is likely to engender serious bias in thestatistical estimates.

Table 2.1 Percentage of voters close to the main twoparties among all partisans, by decade

1970s 1980s 1990s 2000s

BritainLabour 44.2 35.3 42.6 48.5Conservatives 41.5 42.9 37.3 28.0

GermanySPD 54.5 47.1 40.7 38.6CDU/CSU 39.7 43.1 40.4 37.1

The NetherlandsPvdA 34.1 34.8 29.7 25.2CDA 34.5 31.4 28.2 28.0

Note: Cell entry for CDA in the 1970s represents the sum of par-tisans for the three parties (KVP, ARP, CHU) that converged intoCDA after the 1977 election.



Against this methodological background, the analysis that followswill employ the so-called “stacked data matrices” in order to obtain adata structure defined at the level stemming from the interaction ofindividuals and parties (van der Eijk, 2002; van der Eijk et al., 2006).The choice to stack the data allows one to overcome the drawbacksof discrete choice models and, at the same time, permits one to focusthe analysis on all the available alternatives in each political system(van der Brug, Franklin, and Toka, 2008). Following the logic of thestacked data matrix, the unit of analysis is represented by respon-dent*party combinations (for a detailed explanation of the stackingprocedure, see: Appendix A). The dependent variable, partisanship,is measured through the usual combination of survey questions tap-ping both the directional and the strength component.1 Respondentsare thus assigned a value ranging from “0” (not identified with theparty in the specific combination) to “3” (strongly identified withthat party). The resulting partisanship variable in the stacked datamatrix no longer refers to a specific party, but to parties in general.

Two sets of independent variables will be subsequently includedin the analysis. The first set consists in those items that are sup-posed to tap the cleavage-based nature of party identification (Bartleand Bellucci, 2009a). Respondents’ religiousness is measured throughtheir frequency of church attendance, whereas two different indi-cators are included as proxies for one’s placement in the socio-economic structure: trade union membership and subjective socialclass assessment.2

The second set of predictors features items related to individu-als’ attitudes towards relevant partisan objects, as identified by therelevant literature. Voters’ ideological proximity to parties is opera-tionalized through the respondents’ self-placement on the left–rightscale – an easily comparable and widely available measure of thedistance between voters and parties on the left–right “super-issue”throughout countries and time. As to competence assessments, theset of attitudinal variables includes voters’ retrospective assessment ofthe state of the economy. Following the economic voting literature,positive evaluations of the economic situation are expected to fosterpositive attitudes towards parties in the government, with negativeevaluations exerting the opposed effect (Lewis-Beck, 1988). Finally,voters’ attitudes towards party leaders are measured through the stan-dard thermometer score on a ten-point scale. The thermometer is



a very general measure of voters’ perception of party leaders and,admittedly, “it may not be the most reliable source of information”(Bittner, 2011: p. 18; italics mine). At the same time, and most impor-tantly for our purposes, it has also been noted that “the thermometeris the most frequently included type of question about leaders inelection studies” (ibid., p. 16). As such, it is the preferred measureto be employed in a longitudinal cross-national analysis like thepresent one.

2.4 A reassessment of the Michigan model

As a preliminary step, the analysis must rule out a possible criticisminherent in the Michigan model itself. In its original conception,party identification acts as a powerful perceptual screen. Because ofsuch a psychological sense of identification, the individual “tends tosee what is favorable to his partisan orientation” (Campbell et al.,1960: p. 133). Accordingly, partisans are thought to “like a partyleader, irrespective of their personal qualities, if that leader were theleader of their own party, and to dislike them if they were leading adifferent party” (Curtice and Blais, 2001: p. 5). This argument, how-ever, holds only as long as partisan identifications are effectively fixedin time as a result of voters’ placement in the social structure, andthus immune to the effect of short-term forces (i.e., party leader eval-uations). If this were really the case, then our research hypotheseswould be seriously flawed from the outset.

Testing this model is relatively easy. As the Michigan concep-tion postulates party identification as by and large mediated byvoters’ placement in the socio-economic structure, the statisticalmodel includes only identity items (i.e., religiousness, social class,union membership) and socio-demographic controls (i.e., age, gen-der, educational level; coefficients not shown) as covariates. Becausethe dependent variable, partisanship, is not measured on an equal-interval scale, an ordered maximum likelihood estimation techniquesuch as ordinal probit is preferred to linear regression (on this point,see the useful discussion in Fiorina, 1981: pp. 103–105).

The results presented in Table 2.2 would seem to offer almost nosupport for the enduring validity of an identity-based explanationof partisanship. Admittedly, all estimates are statistically significantand signed as expected. However, an unequivocal decline of the



Table 2.2 Social structure and partisanship in three countries

1970s 1980s 1990s 2000s

BritainSocial class 0.28 (0.01)∗∗ 0.26 (0.01)∗∗ 0.26 (0.01)∗∗ 0.15 (0.01)∗∗Union membership 0.17 (0.01)∗∗ 0.12 (0.01)∗∗ 0.12 (0.01)∗∗ 0.05 (0.01)∗∗Nagelkerke R2 0.114 0.081 0.091 0.033McFadden R2 0.056 0.040 0.045 0.016N 18240 22869 20940 22722

GermanyReligiousness 0.26 (0.01)∗∗ 0.20 (0.01)∗∗ 0.13 (0.01)∗∗ 0.16 (0.01)∗∗Union membership 0.11 (0.01)∗∗ 0.09 (0.01)∗∗ 0.09 (0.01)∗∗ 0.08 (0.01)∗∗Nagelkerke R2 0.091 0.048 0.032 0.042McFadden R2 0.062 0.034 0.021 0.027N 24890 27575 19800 27905

The NetherlandsReligiousness 0.35 (0.01)∗∗ 0.35 (0.01)∗∗ 0.33 (0.01)∗∗ 0.26 (0.01)∗∗Social class 0.30 (0.01)∗∗ 0.24 (0.01)∗∗ 0.19 (0.01)∗∗ 0.18 (0.01)∗∗Union membership 0.07 (0.01)∗∗ 0.10 (0.01)∗∗ 0.10 (0.01)∗∗ 0.07 (0.01)∗∗Nagelkerke R2 0.259 0.157 0.118 0.087McFadden R2 0.201 0.118 0.088 0.065N 23814 52470 29177 53388

Note: Dependent variable: Partisanship (4 cat.) on a stacked data matrix. Cell entries arestandardized ordered probit estimates. Standard error estimates (in parentheses) are clus-tered robust at the individual level. ∗∗p < 0.01, ∗p < 0.05. Intercepts and controls (age,gender, educational level) included, coefficients not shown.

coefficients’ magnitude throughout time must also be noted, whichsignals a progressive dealignment between voters’ placement in thesocial structure and their feelings of partisanship. Moving fromreligiousness, the decline is only moderate in the Dutch case, whereasit appears more substantial in Germany. Also, the social class identityvariable highlights a widespread decline. In fact, an almost two-folddiminution of the coefficient throughout the four decades underanalysis can be observed in both Britain and the Netherlands.

Further evidence for the progressive inability of an identity-basedmodel to “explain” voters’ party identification comes from an obser-vation of the various model-fit statistics.3 Based on these measures,the overall fit of the model to the data at hand declines in anastonishingly monotonic fashion, regardless of the country underanalysis and the measure under observation. Yet, steadiness of thedecline is not uniform across countries. In Germany, the explanatory



power of the model reports a two-fold diminution across the fourdecades under analysis. The decline is even more accentuated inBritain and the Netherlands. In the Dutch case, the model-fit goesdown by a 3:1 ratio, while in the British case the decrease is as highas four-fold.

These developments are further detailed in Figures 2.1–2.3, whichpresent the results of a party-by-party analysis of social structure andpartisanship. Focusing on the three major parties in each country,an almost unequivocal decline of the relationship between voters’placement in the social structure and feelings of partisanship canbe observed for each of the parties under observation. The down-ward trend is particularly evident in Britain and the Netherlands.In the former country, a two-fold diminution in the strength ofthe association between social structure and partisanship can beobserved for both the Conservative and the Labour party. As to theLiberal Democrats, no decline can be witnessed – the strength ofthe association is just too weak since the beginning of our timeseries. The Dutch case provides substantially comparable figures, withan unmistakable decline for each and every party under analysis.

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

0.16

1970s 1980s 1990s 2000s

Nag

elke

rke'

s R

2

Conservatives Labour Liberal Democrats

Figure 2.1 Social structure and partisanship in Britain – by party familyNote: Full estimation procedure is available in Appendix B.1


PROOF33

0

0.02

0.04

0.06

0.08

0.1

0.12

0.14

1970s 1980s 1990s 2000s

Nag

elke

rke'

s R

2

CDU/CSU SPD FDP

Figure 2.2 Social structure and partisanship in Germany – by party familyNote: Full estimation procedure is available in Appendix B.2

0

0.05

0.1

0.15

0.2

0.25

0.3

0.35

0.4

1970s 1980s 1990s 2000s

Nag

elke

rke'

s R

2

CDA PvdA VVD

Figure 2.3 Social structure and partisanship in the Netherlands – by partyfamilyNote: Full estimation procedure is available in Appendix B.3



The Christian-Democratic Appeal (CDA) appears the more stronglyrooted party in the social structure. Yet, even in this case the strengthof the relationship goes down by a 2:1 ratio throughout the fourdecades under analysis.

Results from the analysis of the German case provide mixedfindings. The decline is especially marked in the case of the Social-Democratic (SPD) party. As to the Liberals (FDP), the slight upwardtrend witnessed in the period 1970–1990 is counterbalanced by apowerful decrease in the last decade that leads the strength rela-tionship between social structure and FDP partisanship to levelscomparable to those of the SPD. As an exception, the case of theChristian-Democratic Union (CDU) provides evidence for an actualupward trend throughout the most recent decades. Possible explana-tions for such a result will be advanced in the following section, oncethe effect of political forces on partisanship is taken into account.

Overall, the results presented in this section would seem to dis-confirm the enduring validity of a Michigan-style interpretation ofpartisanship – at least for the cases at hand and with respect to themost recent decades. As this analysis suggests, the roots of contempo-rary Europeans’ partisanship have steadily moved away from society.In turn, this occurrence enhances the likelihood for an attitude-basedinterpretation to provide a more solid account of the dynamics ofpartisanship in our three countries.

2.5 Testing the attitudinal model of partisanship

The previous section should have settled the theoretical concern overthe potential spuriousness in the association between partisanshipand attitudes towards partisan objects (as driven by the simul-taneous effect of socio-structural forces). The analysis can thusmove towards an assessment of the relative ability of various atti-tude forces in predicting voters’ partisanship. A structurally simplemodel of attitudinal partisanship can be specified as a function ofvoters’ attitudes towards the most relevant partisan objects iden-tified by the literature: leaders, ideology, and performance-relatedconsiderations.4

As the dependent variable is the same one employed in theprevious analysis, estimation takes place once again through an ordi-nal maximum likelihood technique. For reasons of cross-country



Table 2.3 The attitudinal drivers of partisanship in three countries

1980s 1990s 2000s

BritainLeader evaluations 0.31 (0.01)∗∗ 0.60 (0.02)∗∗ 0.85 (0.02)∗∗

Ideological proximity 0.36 (0.01)∗∗ 0.41 (0.01)∗∗ 0.58 (0.02)∗∗

Economic assessment 0.23 (0.01)∗∗ 0.06 (0.01)∗∗ 0.16 (0.02)∗∗

Nagelkerke R2 0.233 0.318 0.361McFadden R2 0.123 0.173 0.207N 10338 11598 13568

GermanyLeader evaluations 0.85 (0.03)∗∗ 0.90 (0.02)∗∗ 0.96 (0.03)∗∗


Economic assessment 0.01 (0.01) 0.13 (0.02)∗∗ 0.06 (0.01)∗∗


The NetherlandsLeader evaluations 0.49 (0.02)∗∗ 0.63 (0.02)∗∗ 0.70 (0.02)∗∗


Economic assessment 0.04 (0.01)∗∗ −0.01 (0.01) 0.06 (0.01)∗∗


Note: Dependent variable: Partisanship (4 cat.) on a stacked data matrix. Cell entries arestandardized ordered probit estimates. Standard error estimates (in parentheses) are clus-tered robust at the individual level. ∗∗p < 0.01, ∗p < 0.05. Intercepts and controls (age,gender, educational level) included, coefficients not shown.

comparability the models are estimated only with respect to the threemost recent decades.5

The results presented in Table 2.3 provide substantial confirma-tion of the main research hypotheses. An assessment of the model-fitstatistics highlights, in fact, a significant growth in the explanatorypower of the attitudinal model of partisanship as compared with theidentity-based one. Figure 2.4 presents the relative explanatory powerof the two models with respect to each country and decade.

In line with the previous findings, the upward trend is mostnoticeable in the British case, and especially in the most recentdecade. As to the Dutch case, the trend line is not as steep as inBritain. Nonetheless, the increasing appropriateness of an attitudinal



0.00

2.00

4.00

6.00

8.00

10.00

12.00

1980s 1990s 2000s

Ind

ivid

ual

att

itu

des

/So

cial

str

uct

ure

(Rat

io)


Figure 2.4 Social vs. attitudinal drivers of partisanship in three countries:A comparisonNote: Figure entries represent the ratio between attitudinal and socio-structural models’pseudo R2 (Nagelkerke’s) as taken from Tables 2.2 and 2.3. Values higher than 1 indicatethe dominance of attitudes vis-à-vis identities.

conception of partisanship in the Netherlands appears uniformthroughout time. Looking at the value relative to the most recentdecades, an attitudinal model of Dutch partisanship outperforms theidentity-based one on a 3:1 ratio.

Once again, the German case stands as an exception. Here the ini-tially upward trend line witnesses an inversion of the tendency in thelast decade. Social structure’s ability to account for individual partisanties appears to have become slightly more important in recent years.Yet, this result, in all likelihood driven by the growing relevance ofsocial structure for CDU partisanship, does not really alter the con-clusion that individual attitudes are a much better explanation ofGerman partisanship as compared with identity-based explanations.Indeed, the attitudinal model outperforms the latter on an impressive10:1 ratio in both the 1990s and the 2000s.

As to the role of leader evaluations in the various models, pro-bit coefficients are always significantly related to partisanship and,consistently with the personalization hypothesis, their magnitude



highlights an unequivocal increase throughout time. When it comesto the relative effect of leader evaluation vis-à-vis other attitudinalforces considered, their hypothesized dominance is confirmed too.Indeed, retrospective economic assessment seems to play hardly anyrole. Ideological proximity, on the contrary, starts the time series asa force almost paralleling that of leader evaluations. Looking at thevalues presented in Table 2.3 from left to right, however, one notesthat the massive increase in the magnitude of the leaders’ coeffi-cients is not paralleled by those relative to issue proximity, whoseimpact increases only slightly throughout the three decades underanalysis. A comparative assessment of the results further highlightsthe differentiated impact of leader evaluations on partisanship acrosscountries. As expected, leaders play a bigger role in majoritarian andmixed systems (Britain and Germany) than in purely proportionalones (the Netherlands).

A reanalysis of the same data, broken down by party, helps illus-trate the dynamics of the development across countries. Figures 2.5–2.7 present the effect of leader evaluations on partisanship for the

0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

0.9

1

1980s 1990s 2000s

Sta

nd

ard

ized

pro

bit

co

effi

cien

t


Figure 2.5 Leader evaluations and partisanship in Britain – by party familyNote: Full estimation procedure is available in Appendix B.4



0

0.2

0.4

0.6

0.8

1

1.2

1.4

1980s 1990s 2000s

Sta

nd

ard

ized

pro

bit

co

effi

cien

t

CDU/CSU SPD FDP

Figure 2.6 Leader evaluations and partisanship in Germany – by party familyNote: Full estimation procedure is available in Appendix B.5

three major parties in each country in the last three decades. Oncemore, the most unequivocal evidence of increasing party leadereffects on partisanship comes from the British case, where leadersappear to matter more in the dynamics of partisan attachment witheach of the parties under consideration. Interestingly, the increase isrelatively steeper for the Labour party than for the Conservatives inthe decades marked by the withdrawal of Margaret Thatcher frompolitics and the subsequent ascent of Tony Blair as leader of the NewLabour party. Findings from Germany and the Netherlands are lessclear cut. In the latter country, both the Labour (PvdA) and the Liberal(VVD) parties show a slight decline in the association between leaderevaluations and partisanship in the last decade. As to the Germancase, the CDU stands again as the exception, with a declining impactof leaders in the most recent years.

Taken together, these occurrences allow a preliminary inferencebased on the governing/opposition role of parties and their lead-ers. In fact, the relationship between party leader evaluations andpartisanship appears stronger (and on the rise) in times when the



0

0.1

0.2

0.3

0.4

0.5

0.6

0.7

0.8

1980s 1990s 2000s

Sta

nd

ard

ized

pro

bit

co

effi

cien

t

CDA PvdA VVD

Figure 2.7 Leader evaluations and partisanship in the Netherlands – by partyfamilyNote: Full estimation procedure is available in Appendix B.6

leader of the party is also head of government. In British Labour’scase, the growing impact of leader evaluations parallels the ascentof Tony Blair as prime minister. The same goes for the DutchCDA under the leadership of Jan-Peter Balkenende (2002–2010) andGerard Schroeder’s German SPD (1998–2005). On the other hand,when parties lose their role in government (along with the foremostvisibility of their leaders in the cabinet) the relationship somehowweakens, as is the case with the German CDU since the 1990s andthe Dutch PvdA and VVD in the 2000s. In turn, this can also accountfor those instances in which, contrary to expectations, the relevanceof social structure appears to be on the rise (i.e., German CDU afterthe 1990s). To the extent that leader evaluations become less impor-tant, there may be more room for other factors. However, this doesnot deny the overall message conveyed by the analysis, which con-firms by and large the crucial importance acquired throughout timeby leader evaluations as drivers of voters’ closeness to political partiesin our three parliamentary systems.



2.6 Changing parties, changing partisans: The rise of“party/leader identification” in Western Europe

In recent decades, political leaders have become increasingly visi-ble to mass publics due to the ongoing process of personalization ofpolitics common to all established parliamentary democracies. Thisdevelopment has not only affected political communication. Partyleaders have been found to exert a stronger effect over time in theexecutive branch of parliamentary democracies as well as withintheir own parties’ structures. Some have gone as far as contendingthat nowadays political leaders personify the policy platforms of therespective parties. Against this background, the intuition that voters’party loyalties should be interpreted (also) as a function of their eval-uation of the leaders has been recurrently advanced – and yet neverput to systematic test. Indeed, empirical research on partisanship hasbeen surprisingly reluctant in addressing this debate.

This chapter has taken up the task of reassessing the cross-nationalmeaning of partisanship in three Western European parliamentarysystems in light of the progressive personalization of politics thatcharacterizes them. It has shown that the roots of partisanship havesteadily moved away from society (e.g., early socialization, placementin the socio-economic structure) towards the realm of individualattitudes. What was once conceptualized as a mere reflection of long-term allegiances has nowadays turned into one of the crucial driversof partisanship itself.

With respect to the weakening part played by socio-structuralforces in shaping voters’ partisanship, these findings link well withtraditional interpretations of social change based on the cleavagedealignment thesis. As has been repeatedly argued, social cues maystill represent a potent source of political attitudes for people “whoare integrated into traditional class or religious networks . . . but todaythere are fewer people who fit within such clear social categories”(Dalton and Wattenberg, 1993: p. 201). Nowadays the political rele-vance of traditional cleavage structures is markedly less than it waswhen the concept of party identification was conceived (Oskarson,2005). Yet, as Berglund et al. (2005) argue, “party identificationshould not necessarily decline in the slipstream of the decline ofthe relationship between social structure and party system” (p. 107).Indeed, empirical research documents that a substantial – albeit



declining – proportion of citizens in established Western democra-cies still declare themselves to feel close to one of the parties (Dalton,2008). In this respect, an attitudinal interpretation of partisanshipbecomes especially useful for our understanding of the nature ofthis bond.

As to the relative importance of attitude forces, this chapter pro-vides unequivocal confirmation of the “personalization hypothesis”.According to the empirical evidence presented here, individual politi-cians have, in fact, gained prominence at the expense of both tradi-tional socio-economic groups and classic party features such as issuesand ideology. When it comes to partisanship, voters’ evaluation ofparty leaders appears to have become the most powerful predictorof partisan alignments at the individual level. As expected, leadersappear slightly more important in majoritarian and mixed countriesas opposed to purely proportional ones. The governing/oppositionstatus of parties also helps to assess the actual importance of leaderevaluations in the partisanship calculus. Party leaders in charge ofthe country’s executive are more visible to the public, more powerfulwithin their own party structures, and, in turn, more relevant to theirown party supporters.

Overall, the findings presented in this chapter link well withthe notion of candidate-centered politics (Wattenberg, 1991), wherebyvoters’ attention is thought to shift from political parties and ide-ologies to individual politicians and their personal characteristics.The evidence presented here supports this notion, and elaborates onone of its crucial implications: namely, that different ways of think-ing about politics can lead to different ways of relating to politics.If individuals’ feelings of partisanship are actually shaped by theirevaluation of party leaders’ personality, then the possibility for lead-ers to bear a strong(er) effect on voters’ behavior through partisanshipcan be envisaged. The extent to which this is actually the case will beassessed in the chapter that follows.



PROOF

3Behavioral Consequences

3.1 Reassessing leader effects in parliamentary elections

Throughout the last decades, political parties have undergone deeptransformations that are at once cause and consequence of the per-sonalization of politics. As shown in the previous chapter, parties’strategy of focusing increasingly on the image of their leaders – asa result of the changes undergone in both political communicationand the electoral market – has affected dramatically the dynamics ofpartisan attachment at the individual level. Voters’ attitudes towardsparty leaders have indeed become a crucial driver of feelings ofpartisanship. Yet, the personalization of politics is supposed to haveignited an even more relevant change at the behavioral level. Indeed,one of the most crucial consequences of personalization lies – or, atleast, should lie – in the increasing centrality of leaders’ personalityin the individual voting calculus.

However, empirical research on leader effects in democratic elec-tions has not reached a consensus on the actual contours of the leadereffect, even concerning the very existence of a tangible impact ofparty leader evaluations on voting itself. The prevailing reading onsuch an effect in parliamentary democracies has been rather skep-tical, typically no more than a residual influence once traditionalvoting determinants (i.e., partisanship and ideology) have enteredthe explanatory equation.

The enduring validity of what could be rightly called the “com-mon wisdom” in electoral research (Midtbø, 1997) rests on the strongassumption that party identification is relatively fixed in time and

42


PROOFBehavioral Consequences 43

thus immune to short-term political forces. However, a variety ofempirical works have already shown that the assumed exogene-ity of partisanship is, at best, doubtful (for a review, see: Marks,1993). Early analyses (see, most notably: Fiorina, 1981) focused onthe role played by issue preferences and performance assessmentsas drivers of partisanship instability. Yet, as shown in the previouschapter, party leader evaluations also play a role – and, indeed, agrowing one. This occurrence has noticeable implications when itcomes to empirical analyses of leader effects using ordinary regres-sion designs. In such a context, trying to estimate the magnitudeof leadership effects by controlling for party identification “wouldunderstate the final impact of leaders’ images by misattributing toparty identification . . . a portion of leadership’s direct effects” (Dinas,2008: p. 508).

This aim of this chapter is to provide a reassessment of leader effectson voting behavior in Britain, Germany, and the Netherlands inthe last three decades, taking into account the endogenous relation-ship between partisanship and party leader evaluations. This issuewill be addressed by employing the standard econometric procedures(i.e., two-stage estimation and instrumental variables) to overcomethe problem of reciprocal causation between predictor variables.Through consistent statistical estimation, the empirical analysis willshed light on the relative effect of parties and leaders on the indi-vidual voting calculus throughout time and countries. The results arepresented after a review of the (few) studies that addressed – or atleast recognized – the problem of endogeneity in the study of leadereffects.

3.2 Partisanship and leader evaluations as votedeterminants: An endogenous relationship

As a common denominator, the present literature rests on the clas-sic social-psychological framework set forth in The American Voter,in which short-term influences on voting behavior are themselvessubject to explanation in terms of temporally and causally priorforces (Campbell et al., 1960: pp. 24–37; Thomassen, 2005: pp. 7–17).At the heart of this model lies the notion of party identification,a long-term feeling of attachment to a political party rooted inearly socialization and primary group membership. According to



the social-psychological reading, party identification is conceived asan unmoved mover – an “exogenous variable affecting politics butnot being affected by politics” (Holmberg, 2007: p. 563). In sucha framework, short-term political influences (i.e., issue preferences,performance assessments, party leader evaluations) stand as a sortof residual category, as they appear “strongly mediated by such sit-uational factors as the strength as well as the direction of partisanaffiliation” (Brettschneider and Gabriel, 2002: p. 153). This theoret-ical model is depicted in Figure 3.1 through the well-known funnelanalogy.

The enduring validity of such an interpretation of voters’ behav-ior rests, however, on the strong assumption that party identificationis relatively fixed and immune to short-term influences exerted byvoters’ attitudes towards, for example, party leaders. In their seminalcontribution, Page and Jones (1979) demonstrate that party loyal-ties “do not function purely as fixed determinants of the vote; thoseloyalties can themselves be affected by attitudes toward the currentcandidates. Even short of major realignments, party affiliations areeffects as well as causes in the electoral process” (Page and Jones,

Historical patterns

Issueopinions

Candidateimage

Vo

te

Group loyalty

Group loyalty

Value orientations

Influence of friends

Media influence

Social divisions

Economic structure

Economic conditions

Political conditions

Government actions

Campaign activity

Party attachment

Figure 3.1 The “funnel” of causality



1979: p. 1088). Voters might well like a party leader just because heis the leader of the party with which they identify. Yet, the reversemight be true as well – voters could declare themselves partisans sim-ply because of the appeal of the party’s leader. The results presentedin the previous chapter support the latter possibility, and highlightthe increasing likelihood for voters’ feeling of attachment to partiesto be shaped by their evaluations of the respective leaders.

Regrettably, the endogenous relationship between partisanship andleader evaluations has been seldom addressed in the available votingliterature. Some works limit themselves to recognizing the prob-lem of endogeneity (Crewe and King, 1994a; 1994b; Evans andAndersen, 2005; Dinas, 2008), while in only a bunch of empiricalcase studies is the two-way causal link between party identifica-tion and leader/candidate evaluations addressed empirically (Archer,1987; Marks, 1993). This is an unfortunate occurrence, for, if there isreciprocal causation between party identification and leader evalua-tions, then these variables become effectively endogenous and theirestimated effects biased. In particular, “the effects of partisanship onthe vote are likely to be exaggerated” (Marks, 1993: p. 143), withleader effects substantially downsized as a result.

In light of these methodological considerations, previous anal-yses’ failure to detect a substantial impact of leaders on votingcan be ascribed to the ways in which leader effects have beeninvestigated – that is, through single-equation models (see, mostnotably: Bartels, 2002; Bartle and Crewe, 2002; Brettschneiderand Gabriel, 2002; Curtice and Holmberg, 2005; Karvonen, 2010;Holmberg and Oscarsson, 2011). As pointed out, a reciprocal relation-ship exists between voters’ evaluation of leaders and their feelingsof closeness to the respective parties. Such reciprocal causation,each independent variable endogenous to the other, generates seri-ous difficulties when the usual regression estimation techniquesare applied, namely, the problem of simultaneous equation bias(Wooldridge, 2006).

Previous scholars’ choice to model leader effects in single-equationdesigns controlling for (endogenous) party identification rests on theidea that “on the one hand, if the party . . . factor is not controlledfor, party leader effects will be seriously overestimated. On the otherhand, if we control for the party variable, leader effects might besomewhat underestimated” (Holmberg and Oscarsson, 2011: p. 37).



This solution, however, does not solve the inherent endogeneitybetween the two predictor variables. Bartle and Crewe (2002) clearlyacknowledge this, and in their analysis of the British case concludethat “the precise relationship between leadership and party imagescannot be fully determined given the limited data available . . . Ourbloc recursive model assumes that party and leader images are locatedat the same stage within our model” (p. 81).

The analyses that follow overcome the limitations stemming fromthe unrealistic restrictions of a single-equation recursive model.A careful assessment of leader effects on voting requires one, infact, to “disentangl[e] the impact of leaders from that of the partiesthey lead” (Gidengil and Blais, 2007: p. 14). In the attempt to esti-mate as accurately as possible the relative effect of partisanship andleader evaluations on the vote, it must be recognized that there is aneed to control for the party factor within the voting model (oth-erwise, leader effects might be “seriously overestimated”). In doingso, however, remedial actions must be undertaken in order not to“underestimate” the actual effect of leaders. The link of recipro-cal causation between the two variables of interest has to be takeninto account, as in such a context these covariates become effec-tively endogenous and their estimated effects potentially biased. Toovercome this problem, it is necessary to properly exogenize theoffending variable (i.e., partisanship) through the construction ofan instrumental variable in a two-stage process. If partisanship iscorrectly exogenized (that is, purified by the influence of leader eval-uations) then it can be safely included in a model of voting withoutthe risk of unjustifiably downsizing the electoral effect of leaderevaluations (Garzia, 2012). Through careful application of structuralequation methods, the analysis that follows sheds new light on thedynamic relationship between partisanship and leader evaluations,as well as on the way in which these forces variously influencevoters.

3.3 Preliminary analysis: Single-equation models

Table 3.1 presents the results of a standard single-equation model ofvoting, pooled by country and decade. The baseline model features allthe “classic” predictors of electoral research, including demographicand socio-structural controls (i.e., religiousness, subjective social


PROOF47

Table 3.1 Leader effects on voting in three countries: Single-equationestimation

1980s 1990s 2000s

BritainSocial class 0.08 (0.03)∗∗ 0.12 (0.03)∗∗ 0.07 (.03)∗

Union membership 0.08 (0.03)∗∗ 0.06 (0.03)∗ 0.09 (.03)∗∗

Partisanship (endogenous) 1.38 (0.03)∗∗ 1.67 (0.04)∗∗ 1.47 (.03)∗∗

Leader evaluations 0.57 (0.03)∗∗ 0.60 (0.04)∗∗ 0.90 (.04)∗∗

Ideological proximity 0.38 (0.03)∗∗ 0.25 (0.04)∗∗ 0.32 (.05)∗∗

Economic assessment 0.17 (0.03)∗∗ 0.07 (0.03)∗ 0.21 (.04)∗∗

Constant −1.51 (0.04)∗∗ −1.61 (0.04)∗∗ −1.40 (.03)∗∗

Nagelkerke R2 0.545 0.603 0.561Cox & Snell R2 0.377 0.410 0.376N 10293 11598 13559

GermanyReligiousness 0.21 (0.04)∗∗ 0.04 (0.03) 0.03 (.03)vUnion membership 0.10 (0.03)∗∗ 0.06 (0.02)∗∗ 0.02 (.03)Partisanship (endogenous) 1.31 (0.04)∗∗ 1.12 (0.03)∗∗ 0.87 (.03)∗∗



Economic assessment −0.03 (0.03) 0.07 (0.04)∗ 0.03 (.03)Constant −2.24 (0.05)∗∗ −2.28 (0.03)∗∗ −2.07 (.04)∗∗


The NetherlandsReligiousness 0.19 (0.03)∗∗ 0.30 (0.03)∗∗ 0.15 (.02)∗∗

Social class 0.13 (0.03)∗∗ 0.20 (0.03)∗∗ 0.19 (.02)∗∗

Union membership 0.03 (0.03) 0.06 (0.02)∗∗ 0.05 (.02)∗∗

Partisanship (endogenous) 1.22 (0.04)∗∗ 1.18 (0.03)∗∗ 0.91 (.02)∗∗



Economic assessment 0.04 (0.02) 0.01 (0.03) 0.09 (.02)∗∗

Constant −2.43 (0.05)∗∗ −2.51 (0.04)∗∗ −2.91 (.03)∗∗


Note: Dependent variable: Vote choice (dummy) on a stacked data matrix. Cell entries arestandardized logistic estimates. Standard error estimates (in parentheses) are clusteredrobust at the individual level. ∗∗p < 0.01, ∗p < 0.05. Controls (age, gender, educationallevel) included, coefficients not shown.



class, trade union membership) and short-term political factors (partyleader evaluations, ideological proximity, economic assessments), aswell as the endogenous partisanship variable. The dependent vari-able, vote choice, is measured as “1” for respondents having votedfor the party in the specific combination of the stacked data matrix,and “0” otherwise. Given the dichotomous operationalization of thedependent variable, binary logistic regression is preferred to ordinaryleast square estimation.

The models presented in Table 3.1 deliver the “conventional”result, with partisanship by and large dominating over leader evalua-tions. Yet, even such a basic model witnesses a relatively strong effectof leaders, which appears to overcome the impact of partisanshipin the last decade in both Germany and (albeit slightly) in theNetherlands. Ideological proximity also seems to play a role, whereassocio-structural items play little, if any, role.

As informative as they might be, these findings cannot be acceptedas such, because the model specification from which they stemis unable to control for the repeatedly highlighted endogeneitybetween predictor variables. No safe conclusion about the relativeeffect of partisanship and leader evaluations on the vote can be drawnwithout taking into account this potential source of bias. For thisreason, it is necessary to resort to instrumental variable estimation.

3.4 Exogenizing partisanship: An instrumentalvariable approach

In econometric terms, the potential source of bias stems from thecorrelation of the ordinary independent variables, partisanship andleader evaluations, with the equation error term (Wood and Park,2004). Such correlation is inevitable, given the reciprocal causal linkbetween the two. The practical solution to this problem is to replacethe offending endogenous variable (i.e., partisanship) with an instru-mental variable that will not be correlated with the error term.Construction of the instrument proceeds in two stages (two-stageleast squares):

In the first stage, the endogenous independent variable, Y, isregressed on proper exogenous variables, thus creating an instru-mental variable Y′. In the second stage, this Y′ is substituted



into the original equation, and reestimation takes place. Themethod “works” because the instrumental variable is effectivelyexogenous, thereby eliminating the source of the difficulty.

(Lewis-Beck, Nadeau, and Elias, 2008: p. 88)

The second stage parameter estimates will have the desirable prop-erty of statistical consistency, and the reciprocal effects can thus becorrectly compared.1

The crucial requirement for creating an instrumental variable isthat the exogenous variables selected are effectively exogenous. If thatcriterion is not met, the procedure will not overcome the bias prob-lem. To hold exogenous status, these variables must be caused byforces outside the system of equations, and must not be correlatedwith the model error terms (on these points, consult the valu-able discussion in Woolridge, 2006: pp. 525–540). In addition, eachexogenous variable must be (i) uncorrelated with the error term inthe explanatory equation, but (ii) correlated with the endogenousvariable they are instrumenting (Kennedy, 2008: Chapter 9). In thecase of partisanship, these standards are readily obtainable withsafely exogenous socio-economic (SES) variables. Most measures ofsocio-economic status conform to this standard, tending to be fixedcharacteristics the respondent brings to the voting booth. Indeed,party identification itself is conceived as the result of an individ-ual’s placement within the social structure (Campbell et al., 1960).At the same time, recent analyses of voting behavior in advancedindustrial democracies have shown the progressive inability of theseindicators to account for individuals’ vote choice (see, most notably,the various country chapters in: Franklin, Mackie, and Valen, 1992).In other words, SES variables meet the necessary conditions forconsistent estimation (Sovey and Green, 2010) as their effect on theoutcome (vote choice) is transmitted solely through the mediatingvariable (partisanship). Our instrumental variables for partisanshipare thus constructed from a number of SES measures available in eachdataset.2

If properly exogenized, the observed correlation between the partyidentification and leader evaluation variables is sharply downsized(see Table 3.2), thus enhancing our confidence in the correctnessof the procedure employed. Further assurance of the quality of theinstruments comes from a Hausman test, which indicates that they



Table 3.2 Summary diagnostics of instrumental variables


Endogenouspartisanship–Exogenouspartisanship

0.25 0.17 0.23

Endogenous partisanship–Leaderevaluations

0.38 0.41 0.27

Exogenous partisanship–Leaderevaluations

0.14 0.14 0.13

Note: Cell entries are Pearson’s r correlation coefficients. All correlations are significant atthe 0.01 level (two-tailed).

are uncorrelated with their respective error terms, a key assumptionfor the proper use of instrumental variables.

3.5 Instrumental variable estimation

Substituting the original partisanship variable for our instrumen-tal variable in the statistical models yields the results presented inTable 3.3. Once the endogenous status of partisanship is taken intoaccount, the effect of leader evaluations would appear to emergemore clearly. On the other hand, if correctly exogenized, partisanshipplays a much weaker role within voters’ electoral calculus thanpreviously appeared.3

A careful examination of the results from instrumental variableestimation highlights the clearer dominance of leader evaluationsover partisanship in terms of impact on vote choice. Indeed, voters’attitudes towards party leaders represent the most relevant factors ineach and every statistical model, overcoming not only partisanshipbut also retrospective assessments of the economy and ideologicalproximity. Socio-structural variables, on the other hand, appear toplay little, if any, role.

The key findings from the single-equation and the instrumen-tal variable estimations are summarized in Figures 3.2–3.4, whichcompare the logistic coefficient of the leader variable from both esti-mation techniques. The data presented in the figures lead to twomajor observations. The first and most obvious is that, once theendogenous status of partisanship is controlled for, leader effects


PROOF51

Table 3.3 Leader effects on voting in three countries: Instrumental variableestimation

1980s 1990s 2000s

BritainSocial class 0.19 (0.03)∗∗ 0.23 (0.03)∗∗ 0.07 (0.04)Union membership 0.15 (0.03)∗∗ 0.17 (0.03)∗∗ 0.13 (0.03)∗∗

Partisanship (exogenous) 0.36 (0.03)∗∗ 0.33 (0.04)∗∗ 0.38 (0.04)∗∗

Leader evaluations 0.67 (0.03)∗∗ 1.00 (0.04)∗∗ 1.50 (0.05)∗∗


Economic assessment 0.28 (0.03)∗∗ 0.05 (0.03) 0.34 (0.04)∗∗

Constant −1.49 (0.03)∗∗ −1.70 (0.04)∗∗ −1.42 (0.03)∗∗


GermanyReligiousness 0.16 (0.03)∗∗ 0.04 (0.03) 0.01 (0.03)Union membership 0.12 (0.02)∗∗ 0.10 (0.02)∗∗ 0.02 (0.03)Partisanship (exogenous) 0.24 (0.03)∗∗ 0.09 (0.03)∗∗ 0.17 (0.04)∗∗



Economic assessment −0.01 (0.02) 0.19 (0.03)∗∗ 0.06 (0.03)*Constant −2.05 (0.05)∗∗ −2.20 (0.03)∗∗ −2.00 (0.04)∗∗


The NetherlandsReligiousness 0.18 (0.03)∗∗ 0.30 (0.03)∗∗ 0.13 (0.02)∗∗

Social class 0.22 (0.03)∗∗ 0.30 (0.03)∗∗ 0.26 (0.02)∗∗

Union membership 0.11 (0.02)∗∗ 0.11 (0.02)∗∗ 0.09 (0.02)∗∗

Partisanship (exogenous) 0.23 (0.03)∗∗ 0.30 (0.04)∗∗ 0.34 (0.02)∗∗



Economic assessment 0.04 (0.02)* 0.03 (0.02) 0.13 (0.02)∗∗

Constant −2.17 (0.05)∗∗ −2.39 (0.04)∗∗ −2.87 (0.03)∗∗


Note: Dependent variable: Vote choice (dummy) on a stacked data matrix. Cell entries arestandardized logistic estimates. Standard error estimates (in parentheses) are clusteredrobust at the individual level. ∗∗p < 0.01, ∗p < 0.05. Controls (age, gender, educationallevel) included, coefficients not shown.


PROOF52

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1980s 1990s 2000s

Sta

nd

ard

ized

log

it c

oef

fici

ent

Single-equation model Instrumental variable estimation

Figure 3.2 Leader effects in Britain: Single-equation vs. IV-estimationNote: Figure entries are standardized logistic coefficients, as from Tables 3.1 and 3.3

0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1.80

2.00

1980s 1990s 2000s

Sta

nd

ard

ized

log

it c

oef

fici

ent


Figure 3.3 Leader effects in Germany: Single-equation vs. IV-estimationNote: Figure entries are standardized logistic coefficients, as from Tables 3.1 and 3.3



0.00

0.20

0.40

0.60

0.80

1.00

1.20

1.40

1.60

1980s 1990s 2000s

Sta

nd

ard

ized

log

it c

oef

fici

ent


Figure 3.4 Leader effects in the Netherlands: Single-equation vs. IV-estimationNote: Figure entries are standardized logistic coefficients, as from Tables 3.1 and 3.3

appear much stronger in magnitude (solid lines representing leadereffects in two-stage models are always above the dotted lines stem-ming from single-equation models). Most important, however, is theobservation that instrumental variable estimation highlights a moreuniform personalization trend on voting. This is especially the case inBritain, where the contribution of party leader evaluations to the vot-ing model appears to have doubled throughout the last three decades.In the Netherlands, the increase is relatively more modest (i.e., about40 per cent in the period 1986–2010). Yet, two-stage estimates of theDutch data uncover a uniform upward trend which was not apparentthrough single-equation regression analysis. As to the German case,both estimation techniques convey comparable findings – leadersappear to matter at least as much today as they did in Helmut Kohl’sdecade. With respect to the 1990s, however, their estimated impacton voting behavior reports a slight increase. The extent to whichthese trend lines will keep pointing upwards is as yet unknown.Nonetheless, these analyses show that as of now voters’ evaluationof party leaders is a crucial element of their voting decision – and,indeed, an increasingly crucial one.



The changing dynamics of voting behavior at the individual levelare further detailed in Figures 3.5–3.7, which plot the effect of partyleader evaluations on the vote (controlling for exogenous partisanship)for the main three parties in each country under analysis. Onceagain, the clearest upward trend can be witnessed in the British case.It is interesting to observe that the impact of leaders on the Labourvote begins the time series as the weakest in magnitude. A massiveincrease of the leader coefficient in the last two decades, however,brings Labour to the top of the table. An interpretation of these find-ings based on the “primacy of the political supply” supports oncemore the contention that an increasing personalization in the parti-san offer (as is clearly the case with the Labour party in the last twodecades) corresponds to a parallel increase in the weight of “persons”within the individual voting calculus. The one and only exceptionto the personalization trend among the cases under analysis (i.e.,German CDU/CSU) does not seriously alter the conclusion that partyleaders are increasingly relevant to the calculus of Western Europeancitizens, whatever party they end up voting for.

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

1980s 1990s 2000s

Sta

nd

ard

ized

log

it c

oef

fici

ent


Figure 3.5 Leader effects in Britain – by party family (IV-estimation)Note: Full estimation procedure is available in Appendix B.8



0

0.5

1

1.5

2

2.5

1980s 1990s 2000s

Sta

nd

ard

ized

log

it c

oef

fici

ent

CDU/CSU SPD FDP

Figure 3.6 Leader effects in Germany – by party family (IV-estimation)Note: Full estimation procedure is available in Appendix B.9

3.6 The primacy of party leader evaluations in theindividual voting calculus

Against the conventional wisdom that sees party leaders as a fun-damental electoral asset for their parties, previous empirical studieshave seldom recognized the importance of leader evaluations in theindividual voting calculus. In line with traditional interpretationsof voting behavior, earlier studies have habitually explained short-term forces such as party leader images in terms of the (assumed)causal prior strength of party identifications. However, the validityof this interpretation of the vote depends heavily on the effec-tively exogenous status of party identification. As has been repeatedlypointed out throughout this study, however, partisanship and leaderevaluations are, indeed, tightly linked in terms of reciprocal causa-tion. In such a context, single-equation models of voting are likelyto provide seriously biased estimates. To overcome this problem, aclassic econometric remedy has been deployed: instrumental vari-able estimation. The results stemming from this alternative analytical



0

0.2

0.4

0.6

0.8

1

1.2

1.4

1980s 1990s 2000s

Sta

nd

ard

ized

log

it c

oef

fici

ent

CDA PvdA VVD

Figure 3.7 Leader effects in the Netherlands – by party family (IV-estimation)Note: Full estimation procedure is available in Appendix B.10

strategy show that, once endogeneity is taken into account, the elec-toral effect of leader evaluations appears much stronger than oftenobserved. Partisanship, on the contrary, would seem to lose its dom-inant role within the voting model. It still matters, but not as muchas voters’ attitudes towards the party leaders.

Obviously, assessing the extent to which party leader personalitycan affect individual vote choices is a relevant endeavor in votingbehavior research. Yet, these findings have clear implications at themacro level too. First, this is because the impact that leaders’ per-sonalities have on voting behavior can deeply influence the waythese leaders campaign. But it goes even further than that. Politicians’beliefs about the importance of leaders’ personalities, as well as their(presumed) ability to attract votes to their parties and win elections,have, in fact, a profound bearing on whom they actually choose asleaders. The (actual) ability of party leaders as persons to have an effecton the outcome of parliamentary elections is the subject of the nextchapter.



PROOF

4Electoral Consequences

4.1 Party leaders’ personality and the outcome ofparliamentary elections

Former class-mass parties have responded to the steady decline ofcleavage alignments in Western Europe by reshaping their appeal toincreasingly de-anchored voters. Political leaders have gained cen-ter stage within their own parties as well as in voters’ reasoning,and this, in turn, has had a strong effect on the individual-leveldynamics of partisan attachment and vote choice. Nowadays, and inlight of these developments, it can be argued that “leaders’ personali-ties and personal characteristics . . . play a large[r] part in determininghow individuals vote in democratic elections” (King, 2002b: p. 4).Indeed, the results presented so far do provide support for such anargument. Nonetheless, however valuable they may be in deepeningour understanding of individual-level dynamics of voting behaviorin Western Europe, these results do not tell us much about one ofthe most crucial aspects of the personalization of politics – namely,the impact of leaders’ personalities on the outcome of democraticelections (King, 2002a).

The available literature points to a number of macro-level factorsthat enhance the role of party leaders’ personalities as determinantsof democratic election outcomes, such as the presence of a domi-nant climate of opinion, a situation of systemic crisis of the politicalsystem (Barisione, 2009), or the closeness of the electoral outcomeitself (King, 2002b). The presence of each of these factors has beenemployed by electoral analysts to interpret the outcome of specific

57



elections, for example, the landslide victory of the British Conserva-tives in 1983 (dominant climate of opinion), the successful “enteringthe field” of Silvio Berlusconi in the Italian election of 1994 (situa-tion of systemic crisis), or the narrow victory of George W. Bush inthe US presidential race of 2000 (closeness of the electoral outcome).

The main focus of this analysis is, however, on the micro-leveldeterminants of the electoral outcome. More specifically, we areinterested in the effect of voters’ evaluation of political leaders.An increasingly employed technique in the study of leader effectsis the so-called counterfactual strategy (Graetz and McAllister, 1987;Bean and Mughan, 1989; Crewe and King, 1994a; 1994b; Jones andHudson, 1996; Bartels, 2002; Johnston, 2002; Marsh, 2007; Dinas,2008; van Holsteyn and Andeweg, 2010; Aarts and Blais, 2011;Bittner, 2011). This strategy emphasizes the asking and answering ofexplicit “What if?” questions (King, 2002b), and it sheds light on theelectoral effect of the personality profile of a specific candidate byforecasting the extent to which the electoral outcome would havechanged had that candidate’s personality been perceived differentlyby voters.

As a common denominator, the present literature relies on casestudies of single national elections or, at best, on within-countrycomparative evidence (among the few exceptions, see Bean andMughan, 1989). The aim of this chapter is, thus, to provide a fullycomparative assessment of leader effects on election outcomes in thethree parliamentary democracies under study. The time span consid-ered covers the last three decades and a total of 20 elections. Theresults are presented after a review of the available studies, focusingin particular on their methodological features. A unified method ofanalysis is then presented and applied to the cases at hand.

4.2 Counterfactual reasoning and leader effects

The seminal contribution to the study of leader effects on the out-come of parliamentary elections is the most celebrated article byBean and Mughan (1989) appearing in the American Political ScienceReview. Their comparative analysis of Britain and Australia is guidedby the question: how would the election outcome have turned out hadthe winning parties’ leaders been perceived to possess the various per-sonal qualities in the same proportions as their unsuccessful counterparts?


PROOFElectoral Consequences 59

Through OLS multiple regression, the authors attempt to estimatethe number of percentage points by which the winning party’s voteshare could have increased/decreased had the same proportion of vot-ers credited its leader with each personal quality (e.g., competence,empathy, integrity, leadership strength, and so on) as had creditedthe losing party’s leader with the same quality. In other words, theirthought experiment involves a scenario in which the electoral effectof, for example, Margaret Thatcher is measured as if her person-ality were perceived exactly as her main opponent’s (i.e., MichaelFoot) actually is. By means of counterfactual reasoning, the authorscould show that voters’ evaluation of party leaders’ personalitiesdid exert a discernible impact on the outcome of the two elec-tions under study net of other potentially relevant factors (i.e., partyidentification).

Further analyses of leader effects have adopted this technique (e.g.,Crewe and King, 1994a; 1994b) in light of its straightforward appli-cability to bipolar election contests. However, the main assumptionon which these studies are based can be (and has been) contested ondifferent grounds. According to Jones and Hudson, “it is unlikely thata political leader would ‘convert’ everyone simultaneously to viewthat his (her) personality has changed” (1996: p. 231). Furthermore,this assumption is even more unlikely to hold in a simulated contextin which one party leader is hypothesized to be perceived exactly ashis major opponent is. After all, if Margaret Thatcher were perceivedexactly as Michael Foot, she would most probably not be MargaretThatcher – if anything, she would be Michael Foot.

Against this background, a number of empirical works have lookedat rather more sensible changes, simulating shifts in party supportas a function of lowering/increasing leader thermometer evaluationsor perceived personality traits. In his analysis of the six US presiden-tial elections held between 1980 and 2000, Bartels (2002) simulatesa counterfactual in which “the competing candidates were viewedequally favorably (or unfavorably) on each potentially relevant traitdimension by a neutral observer” (Bartels, 2002: p. 64). In a similarvein, Bittner (2011) calculated the net effect of presidential can-didates’ traits (as perceived by voters) by computing the “averagecharacter and competence ratings of all candidates . . . thus establish-ing a fictional ‘neutral’ or baseline candidate” (Bittner, 2011: p. 111).The findings presented by these authors are substantially comparable,



and point to the conclusion that, at least on one occasion (i.e., the2000 presidential election), leader effects might have been decisivefor the final outcome of the election.

A current assessment of the literature finds only one study of leadereffects in the European context employing a comparable analyticalstrategy (Dinas, 2008). Even in this case, however, the conclusionsare limited to a single election in a single country (i.e., the Greek elec-tion of 2004). The analysis presented below expands the scope of thiscounterfactual technique to the universe of elections held in the lastthree decades in Britain, Germany, and the Netherlands. One obviousasset of the results presented in this chapter is their full comparabilityacross time and space, for model specification and variable measure-ment are kept constant. In line with the previous chapters, voters’assessment of party leaders is probed through overall thermome-ter scores rather than personality traits battery (as in, e.g., Bartles,2002; Bittner, 2011) precisely in order to enhance cross-countrycomparability.

The thought experiment introduced below involves a comparisonof the actual electoral outcome with the (simulated) outcome of anelection in which the main party leaders are seen equally favorably(i.e., thermometer scores set at the mean value for all leaders) byvoters. Every divergence in party vote shares between the real-worldelection result and the outcome of an imaginary election fought by“average” leaders will be attributed to voters’ (actual) assessment ofthe personality of party leaders. This analytical strategy finds theoret-ical ground in political psychology, and most notably in theories ofcandidate perception – which has been documented as comparativein nature (Rahn et al., 1990; Sullivan et al., 1990). Empirical groundfor such counterfactual is, indeed, provided by real-world evidence:political leaders are perceived differently by voters – as shown in thedescriptive analyses that follow.

4.3 Analytical strategy

The empirical section will concentrate on the 24 elections held inBritain, Germany, and the Netherlands in the period 1980–2010.The choice to stick to this time frame relates to comparability issuesand keeps up with the analytical choices already undertaken in



the preceding chapters. Yet, one crucial difference with the latterconcerns the level of analysis. In the previous chapters, all modelswere estimated by pooling datasets together by decade. Since themajor concern of these analyses was to understand the overall direc-tion of the personalization trend (in both attitude formation andvoting behavior) throughout time, the choice to present model esti-mates for each decade could be justified through the usual “clarityof presentation” argument. The scope of this chapter is, however,more specific, for our interest is in estimating the magnitude ofleader effects on the outcome of each specific election in terms ofvote share percentages. Moving the level of analysis from atheoret-ically specified time frames (i.e., decades) to explicit cross-sections(i.e., national elections) thus requires the analytical strategy to beadjusted accordingly. According to Bittner (2011),

if we wish to assess the extent to which leaders matter in specificelections, we cannot really extrapolate from the larger picture. Inorder to examine the “real” impact of party leaders on elec-tions, then, it makes sense to look at specific leaders in specificelections. (p. 109)

When looking at the effect of party leaders on the electoral out-come, it is, thus, crucial to identify the relative impact of each leaderon the fortunes of his or her party. In order to favor as much aspossible a straightforward presentation of the results, the analysiswill concentrate on the three major parties in each country. Thiscrucial set includes the two main cleavage parties in each country(i.e., British Labour and Conservatives, German SPD and CDU/CSU,Dutch PvdA and CDA) and the liberal parties (i.e., British LiberalDemocrats, German FDP, Dutch VVD). Smaller parties are excludedon methodological grounds as well as on the basis of substantiveelectoral considerations. The former aspect concerns, in primis, theprecarious availability of thermometer measures for smaller parties’leaders throughout the datasets at hand. Moreover, previous researchhas highlighted the difficulty inherent in survey-based regressiondesigns when it comes to estimating the voting equation in the caseof small parties (van der Brug and Mughan, 2007). On the otherhand, estimation of the voting equation is much safer in the case



0

10

20

30

40

50

60

70

80

90

10019

8019

8119

8219

8319

8419

8519

8619

8719

8819

8919

9019

9119

9219

9319

9419

9519

9619

9719

9819

9920

0020

0120

0220

0320

0420

0520

0620

0720

0820

0920

10

Vo

te s

har

e fo

r m

ain

th

ree

par

ties


Figure 4.1 Vote share (%) for the main three parties in each countryNote: Figure entries are percentage of valid votes cast for the main three parties ineach country/election: Britain: Conservatives, Labour, Liberal Democrats; Germany:CDU/CSU, FDP, SPD; the Netherlands: CDA, PvdA, VVD.

of big parties – for which the great majority of votes are usually cast.Percentage of vote shares for the three major parties in each countryand election is presented in Figure 4.1.

As one easily notes, the proportion of votes granted to these par-ties is linked to the proportionality of the electoral system employed,and, therefore, to the number of alternatives available to voters.Under the British First Past the Post (FPTP) system, Labour, Conser-vatives, and Liberal Democrats alone collect about nine votes out often that are cast in each election. On the contrary, Germany’s mixedelectoral system provides more room for minor parties, and this iseven more the case for the Dutch purely proportional system: thepercentage of votes cast in favor of the main parties decreases accord-ingly (i.e., roughly four out in five in Germany, above one out of twoin the Netherlands). A note of caution is in order here: the figureshows, in fact, a significant decline in terms of electoral strength onthe part of the established parties – this being especially noticeable inthe German and the Dutch cases.



At face value, such an occurrence could represent a complicationin the analyses that follow. However, the results will clearly show thatthe (potential) magnitude of leader effect is directly proportional tothe electoral size of the party under consideration. In this sense, theexclusion of smaller parties from the aggregate-level analysis can besafely assumed to have only minor effects on the overall estimationprocedure.

4.4 Descriptive analysis

The following tables present the mean thermometer score assignedto party leaders in each country and election under analysis. Oneinteresting finding concerns the different patterns of party leaders’popularity across time in our three parliamentary democracies. Thelast column of the tables reports the mean value of party leaders’ ther-mometer score with respect to each election year – what we shall callthe overall image of party leaders. This would appear to be in generaldecline in those countries (Britain and Germany) where the politicalcompetition is relatively more polarized. In the Netherlands, how-ever, a seemingly curvilinear pattern would seem to emerge, withmain parties’ leaders being judged equally favorably nowadays as inthe 1980s (and significantly better than in the 1990s). In all proba-bility, this finding is linked to the inherently consensual nature ofDutch politics (Andeweg and Irwin, 2003). However, one should notoverlook the effect exerted by the sudden appearance of extremelydivisive leaders in the last decade (i.e., Pim Fortuyn, Geert Wilders)on voters’ perception of more “moderate” party leaderships (see, inparticular, the value of the standard deviations across time).

The available data also allows us to observe voters’ perception ofleaders according to the party “family” the leaders belong to (meanvalues of the thermometer score for all leaders of a given party areprovided in the last row of each table). In Britain, Liberal Demo-crat leaders are generally perceived better than both their Labour andConservative counterparts. Given that no leader of the Liberal partywas in charge of the premiership in the three decades under consider-ation, one could attribute this finding to the cost of ruling hypothesis.However, such an argument does not seem to stand comparativescrutiny. In fact, in both Germany and the Netherlands, the lead-ers with lower thermometer scores are exactly those from the liberal



parties. On a speculative level, this discrepancy could be explained bytaking into account the widely different ideological locations of theseparties across political systems (e.g., way more on the right in bothGermany and the Netherlands). Yet, an explanation of the relation-ship between party typology and voters’ assessment of the respectivepolitical leaders lies beyond the scope of this chapter. What we arereally interested in here is the relationship between voters’ evaluationof party leaders at a given election and their effect on the outcome ofsuch an election.

A purely descriptive analysis, such as the one reported inTables 4.1–4.3, would seem to disconfirm the simplistic idea thatthe most popular leader always wins the elections (King, 2002b).Indeed, this is not so often the case. Of the seven British electionsconsidered, only in three instances did the most popular leader win.One such instance is represented by the 1983 general election, whenMichael Foot led the Labour party to its worst electoral defeat since1931 (McAllister, 2011: pp. 71–72). Following that landslide victory,

Table 4.1 Mean thermometer score of party leaders, Britain

Conservatives Labour Lib-Dem Mean

1983 8.31 3.24 6.87 6.14(2.69) (3.07) (2.50) (2.75)

1987 5.27 6.19 6.41 5.96(2.87) (3.27) (3.07) (3.07)

1992 6.79 5.38 7.05 6.41(3.31) (3.22) (3.20) (3.24)

1997 4.52 6.44 5.66 5.54(3.11) (2.81) (2.48) (2.80)

2001 3.98 5.67 5.71 5.12(2.56) (2.78) (1.95) (2.43)

2005 4.40 4.78 5.64 4.94(2.35) (2.80) (2.03) (2.39)

2010 5.21 4.45 5.00 4.89(2.56) (2.70) (2.23) (2.50)

Mean 5.47 5.41 6.02 5.63(2.99) (2.92) (2.65) (2.85)

Note: Cell entries are mean thermometer scores. Standard error estimatesin parentheses. The names of the parties’ leaders for each election areprovided in Appendix C. Election winners in bold.



Table 4.2 Mean thermometer score of party leaders, Germany

CDU/CSU FDP SPD Mean

1980 5.98 7.55 7.67 7.06(3.00) (1.99) (2.52) (2.50)

1983 6.84 5.13 6.12 6.03(2.81) (2.84) (2.53) (2.73)

1987 5.83 4.76 6.04 5.54(3.30) (2.30) (2.74) (2.78)

1990 7.20 5.86 5.93 6.33(2.67) (2.54) (2.86) (2.69)

1994 5.34 3.94 5.17 4.82(3.29) (2.48) (2.67) (2.81)

1998 4.91 3.63 5.99 4.84(3.11) (2.37) (2.83) (2.77)

2002 4.85 n/a 5.74 5.30(3.02) (2.93) (2.98)

2005 4.75 4.04 5.26 4.68(3.00) (2.92) (3.20) (3.04)

2009 5.75 4.18 4.93 4.95(2.78) (2.78) (2.57) (2.71)

Mean 5.77 5.15 5.98 5.63(3.14) (2.91) (2.92) (2.99)

Note: Cell entries are mean thermometer scores. Standard error estimatesin parentheses. The names of the parties’ leaders for each election areprovided in Appendix C. Election winners in bold

however, Margaret Thatcher’s exceptional popularity saw a notice-able decline. Indeed, in both 1987 and 1992, the “Iron Lady” wonthe election regardless of the fact that she was no longer the mostpopular party leader in the country (in both instances, this wasLiberal party leader David Steel). And, while in 1997 Tony Blair’spersonal ascendancy was clearly a key (but not the only) explana-tory factor of the grand return of Labour, their following victories in2001 and 2005 were not linked to a clear supremacy of their lead-ers’ image. In both elections, Labour leaders’ (again Blair in 2001,and then Gordon Brown in 2005) thermometer score was lower thanthat assigned by British voters to the leader of the Liberal Democrats,Charles Kennedy.

In the cases of Germany and the Netherlands, the emerging patterndiffers noticeably from what one could expect on the basis of the



Table 4.3 Mean thermometer score of party leaders, theNetherlands

CDA PvdA VVD Mean

1986 6.57 5.41 4.20 5.39(2.66) (2.79) (2.48) (2.65)

1989 6.97 6.23 4.32 5.84(2.42) (2.34) (2.28) (2.35)

1994 6.66 6.35 4.70 5.90(2.36) (2.20) (2.26) (2.27)

1998 5.15 7.15 4.95 5.75(2.04) (1.86) (2.36) (2.09)

2002 5.82 4.52 5.66 5.33(2.09) (2.14) (1.83) (2.02)

2003 5.76 6.45 4.85 5.69(2.57) (2.15) (2.23) (2.32)

2006 6.38 5.92 5.57 5.95(2.07) (1.99) (1.73) (1.93)

2010 5.39 6.30 6.03 5.91(2.33) (2.00) (1.73) (2.02)

Mean 6.06 6.05 5.13 5.74(2.38) (2.29) (2.19) (2.28)

Note: Cell entries are mean thermometer scores. Standard error estimatesin parentheses. The name of the parties’ leader for each election isprovided in Appendix C. Election winners in bold.

somewhat less central role of party leader evaluations within voters’calculus in these countries. Among the nine German elections underscrutiny, in only two instances (i.e., 1987 and 2005) did the mostpopular leader not become chancellor. Not dissimilar findings emergefrom the Dutch case, with only two elections (i.e., 1994 and 2003)out of the eight held in the period 1986–2010 witnessing a deviationfrom the pattern.

These preliminary results thus fuel the apparent contradictionbetween individual-level findings and their aggregate-level coun-terparts. Indeed, it is in the relatively more personalized Britainthat the connection between party leader popularity among vot-ers and election outcome would appear weaker. Yet, one also notesthat a descriptive analysis like the one performed above is notable to capture the effect exerted by relevant intervening vari-ables at both micro (e.g., voters’ placement in the social structure,



their long-term political identification and ideological orientation aswell as other short-term factors such as the economy) and macrolevel (e.g., margins of victory in each election). In order to solvethis puzzle, it is necessary to resort to multivariate statistical tech-niques.

4.5 Leader effects and election outcomes: Frommicro- to macro-level analysis

The analysis that follows will attempt to answer a relatively straight-forward and yet crucial question: What is the effect of voters’ per-ception of political leaders on the outcome of each of the electionsunder analysis? In order to answer this question, I will make use ofcounterfactual reasoning and ask: how would the elections under analy-sis have turned out had all party leaders been perceived neutrally by voters?By holding every other factor constant, the counterfactual experi-ment sheds light on the independent effect of party leaders as votegetters for their own parties.

From an operational point of view, the reliance on thermometerscores does not allow (as is the case, for instance, in the existingworks on the American case) simulation of manipulations of indi-vidual trait characteristics. In order to simulate the outcome of anelection fought by average leaders, vote probabilities for each of theparties will be re-estimated as if the mean thermometer score for eachof the party leaders was equal to the average thermometer score forall leaders in that election. Any discrepancy arising between the realelection outcome and the simulated outcome of an election in whichall leaders are perceived “neutrally” will thus be attributed to voters’evaluation of the actual leaders.

Take as an example the British election of 1983. In that year,Margaret Thatcher’s mean thermometer score is equal to 8.31. Thethermometer for the other main parties’ leaders, Michael Foot andDavid Steel, is 3.24 and 6.87, respectively. The average value of theleaders’ thermometer is thus 6.14 – which represents the score thatwill be assigned to the fictional average leader. Accordingly, Thatcher’selectoral effect will be assessed on the basis of her 2.17 points ofadvantage vis-à-vis the average leader (8.31 – 6.14 = 2.17). Deviationsfrom the mean value (i.e., comparative evaluation of main parties’



–3

–2

–1

0

1

2

3

1983 1987 1992 1997 2001 2005 2010

Lea

der

s' c

om

par

ativ

e ev

alu

atio

n


Figure 4.2 Comparative evaluation of main parties’ leaders, Britain

leaders) for each leader, election, and country under analysis arepresented in Figures 4.2–4.4.

With respect to the estimation of aggregate leader effects on theirown parties’ vote shares, the procedure is as follows. Changes in voteshare for each party are calculated by comparing the proportion ofvoters in the sample who actually cast a vote for a given party withthe estimated proportion of voters who – keeping all other factorsconstant – would have voted for that party had they assigned to itsleader a thermometer score equal to the mathematically deductedscore of the fictional average leader.

The statistical analysis is performed using the same specificationof the empirical model as in the previous chapter. With the aimof isolating the independent effect of voters’ evaluation of partyleaders, the model controls for the impact exerted by voters’ socio-demographic characteristics (i.e., age, gender, education), placementin the social structure (subjective social class, trade union member-ship, church attendance), ideological proximity (as tapped by theself-placement on the left–right scale), and retrospective economicevaluations. Controls also include a measure to tap respondents’ feel-ings of long-term loyalty to political parties. However, in light of the


PROOF69

–3

–2

–1

0

1

2

3

1983 1987 1990 1994 1998 2002 2005 2009

Lea

der

s' c

om

par

ativ

e ev

alu

atio

n

CDU/CSU FDP SPD

Figure 4.3 Comparative evaluation of main parties’ leaders, Germany

−3

−2

−1

0

1

2

3

1986 1989 1994 1998 2002 2003 2006 2010

Lea

der

’s c

om

par

ativ

e ev

alu

atio

n

CDA PvdA VVD

Figure 4.4 Comparative evaluation of main parties’ leaders, the Netherlands



manifest endogeneity between partisanship and party leaders’ ther-mometer evaluations, the exogenous partisanship measure employedin the previous chapter will be included as control. By keepingconstant model specification across time and space, this analysis pro-vides an unprecedentedly comprehensive assessment of the aggregateeffect of party leaders in parliamentary elections. In order to keepstrict comparability of the results, a few elections are not included inthe analysis due to the lack of relevant measures in the dataset.1

Although the measures employed are exactly the same as thoseemployed in previous models, two operational choices are different.The first has to do with the necessity to analyze the impact of leadersin each election. For this reason, models are estimated with respectto every single election held in the respective country. The seconddifference relates to our interest in the effect of each party leader.Therefore, for all elections three different logit models (one per party)will be estimated. In every case, the dependent variable is coded “1”for voters declaring to have voted for the party under analysis, whilea value of “0” is assigned to all others (i.e., voters for other parties aswell as abstainers).

The results of this counterfactual analysis are presented inFigures 4.5–4.7. The values plotted in the figures represent an approx-imation of the number of percentage points by which the party underanalysis would have increased/decreased its vote share had all votersperceived its leader as an average leader. In other words, it is a mea-sure of the party leaders’ net worth in votes to their own parties. Notethat positive sign indicates in every instance an electoral asset for thewinning party’s leader.

A preliminary assessment of these data provides a number ofinteresting observations. In the British case, the two highest peakscorrespond to the two most crucial elections among the ones underanalysis: the landslide victory of the Conservative party in 1983 andthe return to power of the (New) Labour party in 1997. In bothinstances, an enormous effect on the part of party leaders can be wit-nessed. According to the results stemming from our simulation, hadBritish voters perceived Thatcher in 1983 as an “average” party leader,the Conservatives’ vote share would have been some 13 percentagepoints lower than it actually was. Similarly, voters’ evaluation of TonyBlair in 1997 would seem to have augmented Labour’s vote share inthe election by almost eight percentage points. Curiously, the effect


PROOF71

−15

−10

−5

0

5

10

15

1983 1987 1992 1997 2001 2005

Net

ele

cto

ral g

ain

/loss


Figure 4.5 Net gains/losses for main parties, Britain

−6

−4

−2

0

2

4

6

8

10

12

14

1983 1987 1990 1994 1998 2005 2009

Net

ele

cto

ral g

ain

/loss

CDU/CSU FDP SPD

Figure 4.6 Net gains/losses for main parties, Germany



−10

−8

−6

−4

−2

0

2

4

6

8

10

1986 1989 1994 1998 2002 2006 2010

Lea

der

’s c

om

par

ativ

e ev

alu

atio

n

CDA PvdA VVD

Figure 4.7 Net gains/losses for main parties, the Netherlands

of Liberal Democrat party leaders is always positive – yet relativelyrestricted in magnitude and never exceeding a few percentage points.

In the German case, the most relevant peaks are three. In both1983 and 1990, they magnify the extremely strong effect exertedby Helmut Kohl, whose personality (as perceived by German voters)appears to have brought to the CDU/CSU alliance over ten percent-age points in each of these elections. The end of the long period ofChristian-Democratic government led by Kohl corresponds, indeed,to the best performance reported by an SPD leader – Gerard Schroederin 1998 – whose effect reaches almost ten percentage points of thevote shares. Contrary to the British case (yet not surprisingly, inlight of the aforementioned comparatively lower personal appeal),the leaders of FDP appear a constant weakness for the fortune of theirparties. Consistently with the British case, however, the magnitude oftheir impact is, admittedly, limited in size.

As for the Dutch case, the stronger effects exerted by individualparty leaders on election outcomes are to be found during the elec-tions of 1989 and 1998. In the former instance, voters’ evaluationof incumbent prime minister Ruud Lubbers would seem to havebrought an advantage of about seven percentage points to the CDA.



In the latter case, it was incumbent prime minister and PvdA leaderWim Kok who won eight percentage points for his party. Contraryto the descriptive analysis presented above (but fully in line withthe individual-level findings from Chapter 3), the results of the mul-tivariate analysis show that leader effects are generally stronger inmagnitude in those countries employing FPTP (Britain) or mixedsystems (Germany) as compared with purely proportional ones (theNetherlands).

We now turn to the final, and possibly more relevant, section of theanalysis – namely, the impact of voters’ evaluation of party leaderson aggregate electoral outcomes. Tables 4.4–4.6 compare the actualelection outcome with the simulated outcome of the election hadall leaders been assigned an identical thermometer score (i.e., theaverage leader score) by all voters. The counterfactual is calculated bysimply subtracting the estimated value of the leader effect for each

Table 4.4 The overall effect of party leaders on electoral outcomes, Britain

Actual voteshare

Size ofleader effect

Simulatedoutcomewith noleader effect

Conservatives 42.4 12.6 29.81983 Labour 27.6 −10.3 37.9

Liberal Democrats 25.4 3.2 22.2Conservatives 42.3 −6.6 48.9

1987 Labour 30.8 1.1 29.7Liberal Democrats 22.6 1.7 20.9Conservatives 41.9 3.3 38.6

1992 Labour 34.4 −6.6 41.0Liberal Democrats 17.8 1.7 16.1Conservatives 30.6 −4.3 34.9

1997 Labour 43.2 7.7 35.5Liberal Democrats 16.7 0.4 16.3Conservatives 31.7 −5.4 37.1

2001 Labour 40.7 6.3 34.4Liberal-Democrats 18.3 2.5 15.8Conservatives 32.3 −3.6 35.9

2005 Labour 35.2 −1.6 36.8Liberal-Democrats 22.0 3.4 18.6

Note: Election winners in bold.



Table 4.5 The overall effect of party leaders on electoral outcomes, Germany

Actualvote share



CDU/CSU 38.2 11.5 26.71983 FDP 7.0 −1.2 8.2

SPD 38.2 1.3 36.9CDU/CSU 34.5 2.6 31.9

1987 FDP 9.1 −0.9 10.0SPD 37.0 4.5 32.5CDU/CSU 36.7 10.6 26.1

1990 FDP 11.0 −1.2 12.2SPD 33.5 −3.6 37.1CDU/CSU 34.2 2.7 31.5

1994 FDP 6.9 −0.6 7.5SPD 36.4 2.8 33.6CDU/CSU 28.4 0.4 28.0

1998 FDP 6.2 −1.7 7.9SPD 40.9 9.0 31.9CDU/CSU 27.8 0.4 27.4

2005 FDP 9.8 −1.5 11.3SPD 34.2 4.4 29.8CDU/CSU 27.3 4.6 22.7

2009 FDP 14.6 −2.3 16.9SPD 23.0 0.0 23.0


party leader and election from the real-world vote share awarded tothe respective party. The result is presented in the last column ofeach table (i.e., simulated outcome of an election fought by averageleaders).

As correctly pointed out by King (2002c), “[a]ny attempt to assessthe role that personality factors play in determining who won in anygiven election needs to address beforehand the question of what ‘vic-tory’ means for these purposes” (p. 219). Indeed, one of the reasonsfor the widespread skepticism to be found in the existing literatureover the role of leaders as determinants of election outcomes has todo with the rather neat conceptual understanding of such a role. Theskeptics’ argument can be boiled down to the idea that leaders do



Table 4.6 The overall effect of party leaders on electoral outcomes, theNetherlands

Actualvote share



CDA 34.6 5.6 29.01986 PvdA 33.3 0.2 33.1

VVD 17.4 −2.3 19.7CDA 35.3 7.0 28.3

1989 PvdA 31.9 3.1 28.8VVD 14.6 −3.1 17.7CDA 22.2 1.4 20.8

1994 PvdA 24.0 2.3 21.7VVD 19.9 −8.4 28.3CDA 18.4 −2.0 20.4

1998 PvdA 29.0 7.7 21.3VVD 24.7 −4.5 29.2CDA 27.9 3.0 24.9

2002 PvdA 15.1 −3.8 18.9VVD 15.4 0.9 14.5CDA 26.5 3.1 23.4

2006 PvdA 21.2 −0.2 21.4VVD 14.6 −1.1 15.7CDA 13.7 −1.4 15.1

2010 PvdA 19.6 1.4 18.2VVD 20.4 0.4 20.0


not matter, for they are often unable to “swing the outcome of a par-ticular election” (Mughan, 2005: p. 2). Precisely in order to tacklethis issue, let us focus on those instances in which the most votedparty in the real election does not match with the “winner” of thecounterfactual simulation. We will refer to these as elections thatmight have been decided by voters’ evaluation of party leaders.2

Of 20 elections under analysis, exactly half of them would seem tofall into this category. Britain is at a perfect average, with three elec-tions out of the six under analysis being possibly decided by leadereffects. In both 1983 and 1992, voters’ favorable evaluation of Con-servative party leaders could have made the difference and turned theelection outcome in favor of the Tories. The same goes for the victory



of Tony Blair in 2001. In Germany, elections potentially decided byleader effects are three out of seven. In this case, however, the balanceis constantly in favor of the CDU/CSU, whose victories of 1983, 1990,and 2009 all appear to be due to the crucial effect exerted by partyleaders’ popularity. In the Dutch case, the pattern is time-centeredrather than party-centered, with the four elections being decided byleader effects all concentrated in the 1986–1998 period.

4.6 The conventional wisdom of electoralresearch revisited

In light of the findings presented in this chapter, it seems plausibleto conclude that leaders’ personalities do matter for the outcomeof democratic elections. Politicians can gain (or lose) votes due tothe way in which their personality profile is perceived by votersindependently of the electoral effect exerted by voters’ long-termidentifications, ideological orientations, and retrospective economicevaluations. At times, leader effects can even make the differencebetween victory and defeat.

Of the 20 elections under analysis, ten do indeed witness a poten-tially decisive effect on the part of political leaders (as perceived andevaluated by voters). These findings sharply contradict those fromavailable studies in which the impact of leaders emerges as merelyresidual. Take, once again, as an example the British general elec-tion of 1983. According to King (2002c), Conservative Party’s leaderMargaret Thatcher was “held in higher esteem than either of hersuccessive Labour rivals . . . but, given the then state of the BritishLabour Party, the Conservatives would have won those electionsunder almost anyone” (King, 2002c: p. 215, italics mine). There ismuch to agree with in King’s argument, in the sense that almost anyleader would have been able to overcome Michael Foot’s extremelylow level of popularity in that year and thus bring an electoral advan-tage to the Conservatives. Yet, the results presented above underlinethe crucial importance of voters’ perception of party leaders for thefinal outcome of that election. Had, in fact, voters perceived Thatcherand Foot as equally likeable, the Tories would have not ended upbeing the party most voted for.

The lesson to be drawn from this example is simple but crucial:the effect of leaders on the electoral fortunes of their parties depends



on voters’ comparative assessment of the available alternatives. Inother words, the ability of leaders to win votes for their parties doesnot depend on whether they are good leaders – it depends, rather,on whether voters perceive them as comparatively better leaders. Thehigher the advantage of a specific leader vis-à-vis his (or her) polit-ical counterparts, the higher his impact on the election outcome:an impact that is decisive, indeed, much more often than usuallyobserved by electoral researchers.



PROOF

5Normative and EmpiricalImplications

5.1 Summary of findings and their relationshipwith the existingliterature

The increasingly crucial role of party leaders in the political processcan hardly be contested. The growing personalization of the politicalsphere throughout time has resulted in a parallel change in the polit-ical supply. In an attempt to adapt to the changing environment,contemporary catch-all parties’ electoral strategies and organizationalstructures have, in fact, become heavily leader-centered. On the basisof the profound changes occurring in the political supply, it hasbeen hypothesized that individual politicians have become moreprominent vis-à-vis parties and collective identities in the mind ofvoters. However, empirical research has fallen short of a consensus onwhether party leaders have actually increased their impact on indi-vidual voting behavior, and, in turn, aggregate election outcomes.As repeatedly argued in the previous chapters, the reason for suchuncertainty within the available literature is to be found in the waysin which leader effects have been conceptualized and measured –that is, mainly as a function of temporally and causally prior partisanallegiances.

This study has putforward an alternative framework for the anal-ysis of voting behavior in parliamentary democracies in light ofthe progressive personalization of democratic politics. It did so byemploying a top-down approach that links the changes in the polit-ical supply to the changing dynamics of voting behavior at theindividual level. The assumption, on which the whole study was

78


PROOFNormative and Empirical Implications 79

based, is that such changes in the political supply must have exertedan effect on the dynamics of individual vote choice. The empiri-cal results presented broadly corroborate this assumption, and showthat the changing structure of political parties and party systems inparliamentary democracies did indeed had noticeable effects on thedynamics of partisan attachments and vote preferences at the indi-vidual level. Most notably, it has been demonstrated that the rootsof partisanship have steadily moved away from society (e.g., earlysocialization, placement in the social structure) towards the realmof individual attitudes. What was once conceptualized as a merereflection of long-term allegiances (i.e., party leader evaluations) hasnowadays become the crucial determinant of partisan attachmentsthemselves.

The rise of such party/leader identification (Garzia, 2013b) has note-worthy implications for the relative place of partisanship and leaderevaluations in the voting calculus of individual voters. To the extentthat leader evaluations have increasingly become endogenous to par-tisan identifications, then simply looking at their residual effect islikely to lead to a substantial underestimation of their actual elec-toral impact. For this reason, the analysis has resorted to two-stageestimation and instrumental variables in order to account for thereciprocal causal link between partisanship and leader evaluations.If endogeneity is taken into account, then the effect of leaders on thevote appears just stronger than that exerted by party identification.

These findings should not necessarily be seen in antithesis to thosestemming from funnel-of-causality-based analyses of voting. In linewith Campbell et al. (1960), this study highlights that voters’ choiceis still based to a substantial extent to their feelings of closeness toparties. What has changed throughout the last five decades, how-ever, are the drivers of these feelings. By moving away from society,parties have stopped shaping the political struggle along the lines ofthe class conflict. By converging towards the catch-all typology, theyhave also moved away from the left–right scheme. Party competi-tion would now seem to conform to a valence politics model (Clarkeet al., 2004) in which electoral choices are supposedly driven by vot-ers’ judgment of the overall “competence” of the rival parties. Sincevoters’ evaluation of leaders is a crucial determinant of parties’ per-ceived competence, this model magnifies the role of party leadersin the electoral competition, and supports the notion that parties’



appeal is increasingly based on the personalities of their leaders. Mostimportantly, the findings presented in this volume demonstrate thatvoters’ feelings of closeness to the parties are ever more based on theirevaluation of party leaders.

In theoretical terms, this book also speaks to the predominantapproach to the study of leader effects brought forward by King(2002a). His seminal distinction between direct and indirect leadereffects paved the way for a strand of scholarship (e.g., Curtice andHolmberg, 2005; Karvonen, 2010; Aarts, Blais, and Schmitt, 2011;Bittner, 2011) that added a comparative perspective to the avail-able literature, but kept its focus narrowed on personality-trait-baseddirect effects and broadly disregarded the indirect effect exerted by aleader “as a result of things that he or she does”. According to King,the leader who succeeds in changing his party’s image “is exertinginfluence in this indirect sense” (King, 2002: pp. 4–5; italics in origi-nal). The findings presented in this book can be conceived as a wayto resolve this tension. To the extent that parties’ image is shapedby that of the party leader, and as far as voters’ choice is based onthis perceived image, strong indirect effects are likely to take place.The growing effect of party leader evaluations on partisanship is aclear example of such an indirect effect. Yet, these effects should notbe simply disregarded under the “indirect” influence heading, but,rather, included in a more comprehensive theoretical model able toaccount for the various sources of leader effects on voters. Previousscholars’ failure to grasp the growing relevance of leaders can, indeed,be attributed to the lack of attention to the indirect sources of leaders’influence. If anything, this book contributes to the debate by show-ing that the indirect effect exerted by leaders through partisanshipshould be taken into account in further theorization of leadereffects.

When it comes to the effect of leaders’ personalities on the out-come of parliamentary elections, this study provides strong confir-mation of the “common wisdom”. Party leaders can gain (or lose)votes due to the way in which their personality profile is perceivedby voters – and this independently of the electoral effect exerted byvoters’ long-term identifications, ideological orientations, and retro-spective economic evaluations. In 10 out of 20 elections that havebeen taken into account, voters’ evaluation of party leaders appearedpossibly decisive for the election outcome.



5.2 Normative implications of personality-basedelectoral change

In the most comprehensive study to date of electoral instability inEuropean democracies, Bartolini and Mair (1990; 2007) advance alist of crucial factors to be taken into account in the study of elec-toral change across time. Their framework incorporates factors suchas the extent of cleavage-closure of the electorate, the changing pol-icy distance between voters and parties, and the format of the partysystem in a given country (i.e., a higher number of electoral alterna-tives should correspond to a higher likelihood to switch vote choiceacross two consecutive elections). Further macro-level factors, likechanges in electoral institutions and the level of electoral participa-tion, are also assigned a central position in their theoretical model.The authors further declare themselves aware of a sixth and “largelyresidual” category. According to Bartolini and Mair,

voters will respond not only to the systematic influences whichwe have already mentioned, but also to the emergence of spe-cific salient issues, to the appeal of individual candidates, whetherseen retrospectively or prospectively . . . in other words, voters willrespond to short-term factors.

(2007: p. 44)

Their understanding of short-term factors as “residual” drivers of elec-toral change is undoubtedly determined by the historical period theyanalyzed (i.e., until 1985). However, the present study has made itclear that it is exactly in the 1980s that the personalization of politicstends to become an equally systemic factor of the electoral competi-tion in European parliamentary democracies. The inclusion of thepersonality-based factor in theories of electoral change highlights,and to some extent fuels, one of the most crucial questions in thefield of democratic theory: Is electoral change based on party leaderevaluations good for democracy?

In the “classic” view, framing politics exclusively in terms of per-sonality has often been seen as irrational (Converse, 1964; Page,1978), for the popular cynical view of candidates is that “they areaffectively packaged commodities devised by image makers whomanipulate the public’s perceptions by emphasizing traits with



special appeal to the voters” (Dalton and Wattenberg, 1993: p. 208).This understanding, in turn, provides support for the widespreadscholarly argument that “voting choices based on policy concerns aresuperior to those based on party loyalty or candidate images. Onlythe former represent clearly sophisticated behavior” (Carmines andStimson, 1980: p. 79). The normative implication stemming fromthis argument is, therefore, that “voters ought to consider ‘higher’ fac-tors like platforms and issues, rather than base, ‘symbolic’ factors likepersonality characteristics when they go to the ballot box” (Bittner,2011: p. 2).

Against an interpretation in which personality-based voting is seenby and large as a peril to democracy, cognitive psychologists haveputforward a novel approach to the sophistication of contemporarymass publics (McAllister, 1996). According to their interpretation,candidate evaluations are not simply short-term emotional reactionsto the politicians of the day. Due to the increasing complexitiesof democratic decision-making, politics has become, for the vastmajority of voters, hard to observe and difficult to interpret cor-rectly (McCurley and Mondak, 1995). Evaluating politics based onits most visible actors can, thus, be understood as part of a ratio-nal voting strategy (Page, 1978; Bean, 1993; Mondak and Huckfeldt,2006; Bittner, 2011; Ohr and Oscarsson, 2011). Indeed, some schol-ars highlight that “candidate assessments actually concentrate oninstrumental concerns about the manner in which a candidate wouldconduct governmental affairs” (Miller, Wattenberg, and Malanchuk,1986: p. 536). Empirical research shows that voters are especially sen-sitive to candidates’ competence and honesty (Mondak, 1995). As hasbeen pointed out, “if a candidate is too incompetent to carry out pol-icy promises, or too dishonest for those promises to be trusted, itmakes perfect sense for a voter to pay more attention to personalitythan policies” (Dalton and Wattenberg, 1993: p. 209).

Following this argument, some have gone as far as contending that“citizens, voting for leaders that best represent their views . . . makedemocracy work” (Lau and Redlawsk, 2006: p. 3). This apparentlybold contention is grounded, nonetheless, on the condition that vot-ers are able to determine which leader actually best represents theirview – with the more sophisticated voters (i.e., the better-educatedand the most knowledgeable of political matters) being expected tobe better able to fulfill this condition. Available evidence does show



that the most politically sophisticated are at least as likely as (if notmore likely than) the least sophisticated voters to cast their ballotbased on party leader evaluations (Glass, 1985; Sniderman, Brody,and Tetlock,1991; Lachat, 2009; Bittner, 2011; Gidengil, 2011).

Interestingly, the most sophisticated would also seem to pay agrowing attention throughout time to leaders’ personalities whencasting their vote as compared with the least sophisticated segmentsof the electorate (see Figures 5.1–5.3).

This widely positive view of personality-based voting behavior has,however, gone far from uncontested. For one thing, the fact that lead-ers matter more for the most sophisticated voters has been implicitlycriticized by research showing that voters of this type simply considera wider number of factors when casting their vote (Sniderman, Brody,and Tetlock, 1991; Cutler, 2002). Leaders matter more for them,but so do issues, performance considerations, and other explanatoryfactors in the voting equation (Bittner, 2011).

However, a more fundamental critique revolves around the argu-ment whether contemporary voters – regardless of their level ofpolitical sophistication – are actually capable of making an informed

1.4

1.5

1.6

1.7

1.8

1980s 1990s 2000s

Sta

nd

ard

ized

pro

bit

co

effi

cien

t

Low educational level High educational level

Figure 5.1 Leader effects and voters’ educational level, BritainNote: Full estimation procedure is available in Appendix B.11


PROOF84

0.6

0.8

1

1.2

1.4

1.6

1980s 1990s 2000s

Sta

nd

ard

ized

pro

bit

co

effi

cien

t


Figure 5.2 Leader effects and voters’ educational level, GermanyNote: Full estimation procedure is available in Appendix B.12

0.7

0.9

1.1

1.3

1.5

1980s 1990s 2000s

Sta

nd

ard

ized

pro

bit

co

effi

cien

t


Figure 5.3 Leader effects and voters’ educational level, the NetherlandsNote: Full estimation procedure is available in Appendix B.13



electoral choice (Kuklinski and Quirk, 2001). Given the informationalenvironment that surrounds politics in established parliamentarydemocracies, do they have the relevant information to judge politicalleaders accurately?

As highlighted in the previous chapters, the ubiquitous focus onindividual leaders by contemporary mass media (and television inparticular) has moved the public’s attention from their role of politi-cians to that of persons. The power of television to restructure thesocial space is a recurring theme in sociology of communication(Thompson, 1995). Meyrowitz (1985) describes accurately the rela-tionship between television and political leaders, and in particularthe lowering effect of the former on the latter. Through television wesee “too much” of our politicians. As “the camera minimizes thedistance between audience and performer . . . it lowers politicians tothe level of their audience” (Meyrowitz, 1985: p. 271), thus strippingthem of the aura of greatness that characterizes any ideal conceptionof a political leader.

According to Rahn et al. (1990), people have come to evaluatepoliticians in the same way they evaluate ordinary people, becauserelying on the personality of a candidate (as opposed to his ide-ology, or issue positioning, and in virtue of its increased visibility)allows individuals to apply inferential strategies that are constantlyemployed in everyday life. In this way, voters are able to arriveat an overall judgment of a certain politician without exerting theeffort to monitor everything he says or does. Similarly, a study bySullivan et al. (1990) concludes that voters “make their comparativecandidate judgments within the context of their perceptions of every-day people” (Sullivan et al., 1990: p. 463). Empirical research showsthat voters develop a mental image of political leaders as personson the basis of a restricted number of categories. The dimensionsused to evaluate political figures are rather limited in number andtend to fall into a few broad categories (Miller and Miller, 1976;Kinder, Abelson, and Fiske, 1979; Kinder, 1986; Miller, Wattenberg,and Malanchuk, 1986; Pancer, Brown, and Barr, 1999). In this respect,leaders’ personality traits have been found to play a crucial role inorganizing knowledge and guiding voters’ processes of leader percep-tion (Pierce, 1993; Funk, 1999). Traits are, in fact, “a basic componentof our images of other persons of all kinds whether family member,acquaintance, or public figure” (Funk, 1996: p. 98).



The possibility for voters to judge “distant” politicians using thesame strategy they employ in everyday life does certainly allow themto remain involved in the political sphere with a relatively low cogni-tive effort, but it comes at a potentially high cost for the democraticprocess. In fact, decades of research in social psychology have widelydemonstrated that personality judgments are extremely resilient, andeven more so when it comes to likes and dislikes (Funder, 1999).Therefore, against a normatively desirable situation in which “pol-itics is a matter of rational deliberation in which the force of thebetter argument triumphs” (Simons, 2000: p. 82), the personalizationof politics bears the non-trivial effect of rendering political choicesindistinguishable from consumer choices, but “in a context in whichchoices are heavily influenced by appeals to emotions and tastes,rather than reflective judgement” (ibid., p. 83).

5.3 Methodological implications and avenues forfurther research

The (potential) danger brought by the personalization of politics toestablished democracies calls electoral researchers to an ever-growingattention and methodological rigor in understanding causes, dynam-ics, and effects of this widespread phenomenon. One central questionto be addressed by future research on the topic relates to the extent towhich the normative implications of personalization can be empir-ically assessed – in other words, can we measure the “quality” ofvoting behavior? Among the possible lines of inquiry, one couldfocus on whether voters’ judgment of political leaders is based onevaluations of the man or the politician (as well as their reciprocalinteraction). The core part of this study has concentrated on disen-tangling the relationship between allegedly long-term explanatoryfactors (i.e., party identification) and voters’ evaluation of the person-ality of political leaders. Yet, there are grounds to believe that a strongcovariance exists between voters’ evaluation of political leaders andtheir reaction to the policies proposed by these leaders (Miller andShanks, 1996). Sorting out in a simultaneous fashion the endogenousstatus of voters’ evaluations of political leaders, their long-term feel-ings of allegiance to political parties, and their closeness to the issueswould have implied methodological efforts that lie beyond the scope



of this longitudinal, cross-country analysis. However, first steps inthis direction have already been undertaken. Among others, a recentstudy by Bellucci, Garzia, and Lewis-Beck (2013) shows that voters’issue preferences shape their assessment of the party leader just asthe likeability of the leaders affects voters’ perceptions of the policystances they bring about.

Reciprocal causation does not only affect independent variables inthe voting equation. A more fundamental question to be addressedrelates to the possible endogeneity between voters’ evaluation ofparty leaders and the dependent variable. In other words: Did I vote forthe party because I like the leader, or do I like the leader because I voted forhis party? In turn, this question raises attention to the ways in whichvoters’ decision-making process is understood in the first place. Themost celebrated funnel of causality relies heavily on classic attitude-behavior theoretical models, which postulate behavior as driven byindividuals’ core attitudes (Fishbein and Ajzen, 1975). The Michiganmodel conceives individual vote choices as a function of “the cumu-lative consequences of temporally ordered sets of factors” (Miller andShanks, 1996: p. 192). In such a framework, it is political attitudesthat drive behavior. Yet, more recent psychological literature showsthat individuals’ behavior can also lead to changes in attitudes (Eaglyand Chaiken, 1993). In the Michigan model, it is assumed that votingbehavior conforms to previously shaped political attitudes in order tomaintain cognitive consistency. By the same token, however, it couldbe argued that voters are actually choosing their core attitudes to con-form to past (voting) behavior precisely in order to avoid cognitivedissonance. A classic example in this sense is represented by earlyworks on party identification in the European context, which foundthe latter “traveling together” with vote choice (Thomassen, 1976;Thomassen and Rosema, 2009). More recent studies of economicvoting support the notion that individuals’ behavior does lead tochanges in attitudes (Wlezien, Franklin, and Twiggs, 1997; Anderson,Mendes, and Tverdova, 2004; Evans and Andersen, 2006; Evans andPickup, 2010).

The issue of reverse causation (i.e., from voting behavior to polit-ical attitudes) has been seldom recognized in empirical assessmentsof party leader effects on voting. The lack of attention to this cru-cial issue has mostly to do with the fact that virtually all comparative



evidence of leader effects relies on cross-sectional data sources (i.e.,post-election surveys). Yet, one of the major shortcomings inher-ent in cross-sectional analytical strategies lies with their inability totake into account the presence of cognitive feedback running fromvoting behavior (i.e., the dependent variable) to political attitudes(i.e., the independent variables). By and large, this limitation can beattributed to the very structure of classic electoral survey research.1

Voting behavior scholars have long been aware of the methodologicalproblems stemming from cross-sectional inference (see, e.g., Lewis-Beck, Nadeau, and Elias, 2008). Yet, cross-national research couldnot help resorting largely to these data sources because of the lackof appropriate comparative panel datasets.2 In doing so, however,it often failed to take properly into account the reciprocal effect ofbehavior on attitudes, tending to exaggerate the cross-sectional evi-dence concerning the importance of the latter. To future research goesthe task of exploiting the growing number of available election pan-els for a more solid assessment of the reciprocal effect of attitudestowards leaders on voting and vice versa.

5.4 Understanding the effect of leaders on voting: Therole of political psychology

Overall, the results of this research point to the crucial role that polit-ical psychological theories, concepts, and methods can play in ourunderstanding of democratic elections’ outcomes. In times of socialand electoral dealignment, one could hardly disagree with Blondeland Thiébault’s (2010) claim that

[t]he decision to study political behaviour on the basis of thepsychological dimension of the relationship between citizens andthe elite constitutes a major move. This move is indeed unavoid-able as soon as one recognizes that citizens and leaders “relate”to each other; but it is also a logical step to take if politicalbehaviour, especially voting behaviour, is not to go into a down-ward spiral of negative conclusions resulting from the fact that theanalysis of political parties, embedded in the “social cleavages”approach, has ceased to be able to provide truly important newinsights.

(Blondel and Thiébault, 2010: p. 8)



Eventually, political psychology will enhance our understanding ofthe mental processes underlying voters’ choice – the often neglectedhow question of electoral research (Houghton, 2009).

Political psychologists have in front of them a wide research agendaon party leaders and voting. For one thing, the consistency and reli-ability of voters’ attitudes towards leaders could be put under closerscrutiny. Are voters basing their choice on well-developed attitudes orsimply shaping these attitudes in line with their past voting behav-ior? The extent to which voting based on personality is a rationalform of behavior is another aspect to which more attention shouldbe devoted. Are voters evaluating their leaders based on what they areor what they say? In turn, inquiries into these aspects can also helpaddress normative questions with clear implications for democraticpolitics, most notably: Is candidate-centered voting good for democracy?There are reasons to believe that the connection between voters,parties, and leaders will be at the core of electoral politics’ researchagenda in the years to come. As hopefully shown in this study, apsychological perspective will most certainly add to our ability tounderstand such topics in political contexts increasingly connotedby partisan dealignment and the personalization of political power.


October 25, 2013 13:34 MAC/PERSON Page-90 9781137270221_08_app01

PROOF

Appendix A: Data Sourcesand Analytical Methods

A.1 Data sources

All the statistical analyses presented in this volume are performed on the ThreeNations Pooled Dataset, assembled in 2012 by the author in collaboration withAndrea De Angelis (European University Institute). The dataset includes allthe available national election studies conducted in Britain, Germany and theNetherlands up to 2010 and detailed as follows:


1964 1961 19711966 1965 19721970 1969 19771974 February 1972 19811974 October 1976 19821979 1980 19861983 1983 19891987 1987 19941992 1990 19981997 1994 20022001 1998 20032005 2002 20062010 2005 2010

2009

The studies conducted in the period 1961–2001 were already transformed intoa by and large comparable format as a result of the European Voter project(Thomassen, 2005). As for the most recent decade, all available studies havebeen added to the original data source by the authors. Detailed study descrip-tions, variable coding and their availability throughout surveys are reportedbelow:

Britain1964–2001 Hans-Dieter Klingemann, Hermann Schmitt, Bernhard

Weßels, and Tanja Binder. The European Voter Dataset.GESIS Cologne, Germany. ZA3911 data file.

90


PROOFAppendix A: Data Sources and Analytical Methods 91

2005 Harold Clarke, David Sanders, Marianne Stewart, and PaulWhitely. British Election Study 2005. National Centre for SocialResearch. P2474 data file.

2010 Harold Clarke, David Sanders, Marianne Stewart, and PaulWhitely. British Election Study 2009–10. <http://www.bes2009-10.org>

Germany

1961–1998 Hans-Dieter Klingemann, Hermann Schmitt, BernhardWeßels, and Tanja Binder. The European Voter Dataset.GESIS Cologne, Germany. ZA3911 data file.

2002 Jurgen Falter, Oscar Gabriel, and Hans Rattinger. PoliticalAttitudes, Political Participation and Voter Conduct in UnitedGermany 2002. GESIS Cologne, Germany. ZA3861 data file.

2005 The Comparative Study of Electoral Systems: Module 3 (secondadvance release, 31 March, 2011 version). <http://www.cses.org>

2009 Hans Rattinger, Sigrid Roßteutscher, Rüdiger Schmitt-Beck,and Bernhard Weßels. German Longitudinal Election Study2009. GESIS Cologne, Germany. ZA5301 (Post-electionCross-section) data file.

The Netherlands

1971–2003 Bojan Todosijevic, Kees Aarts, and Harry van der Kaap. DutchParliamentary Election Studies Integrated File 1970–2006.DANS – Data Archiving and Networked Services. P1816data file.

2006 Kees Aarts, Henk van der Kolk, Martin Rosema, and HansSchmeets. Dutch Parliamentary Election Study 2006. DANS –Data Archiving and Networked Services.

2010 Dutch Parliamentary Election Study 2010 (pre-release).

A.2 Detailed variable coding

Britain

Age in years

Gender (0) male – (1) female

Education scale from (0) lowest to (2) highest

Social Class (–1) working class – (0) no class identification –(1) middle class

Union Membership (0) not member – (1) member


PROOF92 Appendix A: Data Sources and Analytical Methods

Partisanship (0) not close to the party – (1) not very strong – (2)fairly strong – (3) very strong

Leader Evaluations thermometer evaluation of party leaders – scale from(0) strongly dislike to (10) strongly like

IdeologicalProximity

self-placement on the left–right continuum – scalefrom (0) left to (10) right

EconomicAssessment

retrospective, egotropic assessment of the state of theeconomy on a five-point scale: (1) got a lot worse –(2) got a little worse – (3) stayed the same – (4) got alittle better – (5) got a lot better

Goldthorpe Class (1) professional/manual high grade –(2) professional/manual low grade – (3) routinenon-manual – (4) petty bourgeoisie – (5) manualforeman – (6) skilled manual – (7) unskilled manual

Household Income scale from (1) lowest quintile to (5) highest quintile

Region of Residence (1) Scotland – (2) Wales – (3) England

ReligiousDenomination

(1) no religion – (2) Catholic – (3) Protestant – (4) other

Germany

Age in years


Education (1) primary – (2) secondary – (3) higher

Religiousness frequency of church attendance on a six-point scale:(0) no religion – (1) less than once a year – (2) oncea year – (3) several times a year – (4) once a monthor more – (5) once a week or more


Partisanship (0) not identified with the party – (1) weak –(2) strong – (3) very strong

Leader Evaluations thermometer evaluation of party leaders – scale from(0) very negative view to (10) very positive view

Ideological Proximity self-placement on the left–right continuum – scalefrom (0) left to (10) right

Economic Assessment retrospective, sociotropic assessment of the state ofthe economy on a five-point scale: (1) bad – (2) notthat fine – (3) in between – (4) good – (5) very good

Profession (1) self-employed – (2) white collar – (3) publicservant – (4) manual worker – (5) farmer – (6) neverworked/in education

Region of Residence (1) Schleswig-Holstein – (2) Hamburg –(3) Lower-Saxony – (4) Bremen –



(5) North Rhine-Westfalia – (6) Hesse –(7) Rhineland-Palatinate – (8)Baden-Wuerttemberg – (9) Bavaria –(10) Saarland – (11) Berlin – (12) Brandenburg –(13) Mecklenburg-West Pomerania – (14) Saxony –(15) Saxony-Anhalt – (16) Thuringia

Religious Denomination (1) no religion – (2) Catholic – (3) Protestant –(4) other

Unemployment Status (0) employed – (1) unemployed

Urban/Rural population size of the town of residence:(1) below 5k – (2) 5k–20k – (3) 20k–50k –(4) 50k–500k – (5) over 500k

The Netherlands

Age in years


Education (1) elementary – (2) lower vocational –(3) secondary – (4) middle level vocational/higherlevel secondary – (5) university

Church Attendance frequency of church attendance on a six-pointscale: (0) no religion – (1) almost/never –(2) several times a year – (3) once a month –(4) two/three times a month – (5) at least oncea week

Social Class (1) working class – (2) upper working class –(3) middle class – (4) upper middle class –(5) upper class


Partisanship (0) not adherent – (1) not convinced adherent –(2) convinced adherent – (3) very convincedadherent

Leader Evaluations thermometer evaluation of party leaders – scalefrom (0) very unsympathetic to (10) verysympathetic

Ideological Proximity self-placement on the left–right continuum – scalefrom (0) left to (10) right

Economic Assessment retrospective, sociotropic assessment of thestate of the economy on a three-point scale:(1) unfavorable – (2) neither unfavorable norfavorable – (3) favorable

Household Income 12 categories (low to high)



Religious Denomination frequency of church attendance on a six-pointscale: (0) no religion – (1) (almost) never –(2) several times a year – (3) once a month –(4) two/three times a month – (5) at least oncea week

Unemployment Status (0) employed – (1) unemployed

Urban/rural degree of urbanization of the town of residenceon a five-point scale: (1) not urban – (2) hardlyurban – (3) midly urban – (4) sharply urban –(5) very sharply urban


PROOF95

A.3

Cro

ss-t

emp

ora

lav

aila

bil

ity

of

vari

able

s

Bri

tain

1964

1966

1970

1974

1979

1983

1987

1992

1997

2001

2005

2010

Age

XX

XX

XX

XX

XX

XX

Gen

der

XX

XX

XX

XX

XX

XX

Edu

cati

onX

XX

XX

XX

XX

XX

XSo

cial

clas

sX

XX

XX

XX

XX

XX

XU

nio

nm

embe

rsh

ipX

XX

XX

XX

XX

XX

XPa

rtis

ansh

ipX

XX

XX

XX

XX

XX

XLe

ader

eval

uat

ion

sX

XX

XX

XX

XX

XX

XId

eolo

gica

lp

roxi

mit

yX

XX

XX

XX

XX

XEc

onom

icas

sess

men

tX

XX

XX

XX

XX

XG

old

thor

pe

clas

sX

XX

XX

XX

XX

XX

Hou

seh

old

inco

me

XX

XX

XX

XX

XX

XX

Reg

ion

ofre

sid

ence

XX

XX

XX

XX

XX

XX

Rel

igio

us

den

omin

atio

nX

XX

XX

XX

XX

XX

X


PROOF96

Ger

man

y

1961

1965

1969

1972

1976

1980

1983

1987

1990

1994

1998

2002

2005

2009

Age

XX

XX

XX

XX

XX

XX

Gen

der

XX

XX

XX

XX

XX

XX

XX

Edu

cati

onX

XX

XX

XX

XX

XX

XX

XR

elig

iou

snes

sX

XX

XX

XX

XX

XX

XX

Un

ion

mem

bers

hip

XX

XX

XX

XX

XX

XX

XPa

rtis

ansh

ipX

XX

XX

XX

XX

XX

Lead

erev

alu

atio

ns

XX

XX

XX

XX

XX

XX

XId

eolo

gica

lp

roxi

mit

yX

XX

XX

XX

XX

Econ

omic

asse

ssm

ent

XX

XX

XX

XX

XX

Prof

essi

onX

XX

XX

XX

XX

XX

XX

XR

egio

nof

resi

den

ceX

XX

XX

XX

XX

XX

XX

XR

elig

iou

sd

enom

inat

ion

XX

XX

XX

XX

XX

XX

XX

Un

emp

loym

ent

stat

us

XX

XX

XX

XX

XX

XX

XX

Urb

an/r

ura

lX

XX

XX

XX

XX

XX

XX

X


PROOF97

Th

eN

eth

erla

nd

s

1971

1972

1977

1981

1982

1986

1989

1994

1998

2002

2003

2006

2010

Age

XX

XX

XX

XX

XX

XX

XG

end

erX

XX

XX

XX

XX

XX

XX

Edu

cati

onX

XX

XX

XX

XX

XX

XX

Rel

igio

usn

ess

XX

XX

XX

XX

XX

XX

XSo

cial

clas

sX

XX

XX

XX

XX

XX

XX

Un

ion

mem

bers

hip

XX

XX

XX

XX

XX

XPa

rtis

ansh

ipX

XX

XX

XX

XX

XX

XLe

ader

eval

uat

ion

sX

XX

XX

XX

XId

eolo

gica

lp

roxi

mit

yX

XX

XX

XX

XX

XX

Econ

omic

asse

ssm

ent

XX

XX

XX

XH

ouse

hol

din

com

eX

XX

XX

XX

XX

XX

XX

Rel

igio

us

den

omin

atio

nX

XX

XX

XX

XX

XX

XX

Un

emp

loym

ent

stat

us

XX

XX

XX

XX

XX

XX

XU

rban

/ru

ral

XX

XX

XX

XX

XX

XX

X



A.4 Descriptive statistics of categorical andinterval-level variables

Britain

N Minimum Maximum Mean St. Dev.

Age 35347 16 99 47.81 17.69Gender 35550 0 1 0.53 0.50Education 33336 0 2 0.58 0.79Working class

identification35550 0 1 0.54 0.50

Middle classidentification

35550 0 1 0.29 0.46

Union membership 33541 0 1 0.25 0.42Partisanship (strength) 35550 0 3 1.70 0.97

Leader evaluationsConservatives 34482 0 10 5.47 2.99Labour 34590 0 10 5.41 2.92Liberal Democrats 28250 0 10 6.02 2.65

Left–rightself-placement

25860 0 10 5.57 2.74

Economic assessment 31198 1 5 2.66 1.38Goldthorpe class 29804 1 7 4.22 2.11Household income 28240 1 5 2.91 1.31

Germany


Age 21306 16 99 47.49 17.32Gender 24735 0 1 0.53 0.50Education 24547 1 3 1.59 2.09Church attendance 22844 0 5 2.11 1.56Union membership 22788 0 1 0.17 0.37Partisanship (strength) 24749 0 3 0.91 1.05

Leader evaluationsCDU/CSU 19276 0 10 5.77 3.14FDP 15696 0 10 5.15 2.91Grunen 7758 0 10 4.75 2.76PDS 8674 0 10 3.29 3.04SPD 18923 0 10 5.98 2.92



Left–right self-placement 14260 0 10 4.60 2.21Economic assessment 14615 1 5 3.12 0.94Unemployment status 24674 0 1 0.49 0.50Urban/rural 24735 1 5 3.05 1.45

The Netherlands


Age 25553 17 99 45.49 17.14Gender 25624 0 1 0.50 0.50Education 24928 1 5 2.90 1.45Church attendance 25469 0 5 1.67 1.91Social class 24417 1 5 2.59 1.09Union membership 18888 0 1 0.21 0.41Partisanship (strength) 24080 0 3 1.30 1.09

Leader evaluationsCDA 14442 0 10 6.06 2.38ChristenUnie 6630 0 10 5.23 2.16D66 12789 0 10 5.76 1.99GroenLinks 10683 0 10 5.85 2.16LPF 3674 0 10 4.10 2.99PvdA 14775 0 10 6.05 2.29SGP 3008 0 10 4.28 2.30SP 7756 0 10 6.04 2.10VVD 14373 0 10 5.13 2.19

Left–rightself-placement

18737 0 10 5.35 2.32

Economic assessment 13856 1 3 2.22 0.77Household income 22061 1 12 6.61 3.19Unemployment status 25442 0 1 0.46 0.50Urban/rural 25625 1 5 3.10 1.37

A.5 Stacking the data matrix

The major focus of this book is on the determinants of partisanship and votechoice – two nominal variables by definition. Generally, electoral researchersface the problem of the nominal nature of their dependent variable in twoways. A possible manner to deal with the operationalization of, for example,the voting choice is to assign a value of ‘1’ if the individual cast their bal-lot in favor of the incumbent party, and a value of ‘0’ if the voter opted foran opposing party. This approach is fairly common, for instance, in testing



economic voting theories, where the performance of the incumbent is usuallyamong the key predictors, or in two-party systems such as the US. A differ-ent solution, particularly suitable in multiparty political contexts, consistsin making use of discrete-choice models such as multinomial logit (MNL) orprobit (MNP) regression. Yet, this second solution can be problematic for atleast three orders of reasons. First, as these methods are often employed whendealing with extreme multiparty systems, they can only rarely provide reli-able estimates for small parties, whose voting function is extremely skewed(van der Brug and Mughan, 2007). Second, the label “multinomial” includes avariety of discrete-choice models that present different peculiarities and draw-backs. In particular, both MNL and MNP modeling techniques share a similarstructure, with the important difference that the distribution of the error termin the former is assumed to be very simple and tractable (the Type-I ExtremeValues) while for the latter it is assumed to be normal (Long, 1997). Moreover,the MNL allows only the inclusion of explanatory variables varying acrossthe observations and provides a set of coefficients (one for each alternative)whose identification is heavily dependent on the “Independence of IrrelevantAlternatives” assumption, which is unlikely to be satisfied in most politicalsystems.

An alternative analytical perspective consists in analyzing the determinantsof party choice by “stacking” the data matrix in order to obtain a data struc-ture defined at the level stemming from the interaction of individuals andparties (van der Eijk and Franklin, 1996; van der Eijk et al., 2006; van der Brugand Mughan, 2007; van der Brug et al., 2007; van der Brug et al., 2008; vander Eijk and Franklin, 2009). In the transformed data matrix (which is derivedfrom the “normal” data matrix as illustrated in Figure A.1) the unit of analysisis represented by respondent∗party combinations.

Stacking the data switches the level of analysis and the level of con-ceptualization simultaneously. On the one hand, the level of analysisshifts downwards from the individual to the intra-individual level, forcingone to reinterpret the independent variables in terms of (individual∗parties)relationships. On the other hand, the stacked data matrix leads to abroader interpretation of the concept of party preference in cross-nationalresearch. If the dependent variable is reinterpreted in terms of the dyadicindividual∗party relationships, then the object of analysis is no longer a spe-cific party, but a generic one (regardless of the specific characteristics of theparty system).

The resulting size [N] of the stacked data matrix equals [R ∗P], where P isthe number of respondents in each dataset and P is the number of partiesincluded as stacks. The total size of the stacked datasets employed in this bookis as follows:

BritainR = 35550P = 3: Conservatives, Labour, Liberal DemocratsN = 106650



Original data matrix

Stacked data matrix

1

11

1

222333

11

1

222333

95

4

597247

12

3

123123

5959

59

404040222222

02

3

312631

Res

p-id

Res

p-id

Id-o

f-pa

rty

Age

Age

Left/

right

pos

ition

resp

onde

nt

Left/

right

dist

ance

Per

ceiv

ed L

R-

posi

tion

pty

1

Per

ceiv

ed L

R-

posi

tion

pty

2

Per

ceiv

ed L

R-

posi

tion

pty

3

L/R

dis

t. to

par

ty 1

L/R

dis

t. to

par

ty 2

L/R

dis

t. to

par

ty 3

Vot

e-ch

oice

Vot

e-ch

oice

Util

ity p

arty

1

Util

ity

Util

ity p

arty

2

Util

ity p

arty

3

59 469

433

676

788

036

213

321

123

952

594

477

4022

23

Figure A.1 Stacking the data matrix

GermanyR = 24749P = 5: CDU/CSU, FDP, Die Grunen, Linkspartei, SPDN = 123745

The NetherlandsR = 25625P = 9: CDA, ChristenUnie, D66, GroenLinks, LPF, PvdA, SGP, SP, VVDN = 230625



Note that not all predictors included in the analyses are interpretable in termsof respondent∗party combinations. Indeed, only respondents’ partisanship andtheir evaluation of party leaders have a direct counterpart at this peculiarlevel. For all other variables, it was necessary to produce y-hats (that is, pre-dicted values) regressing the dependent variable of the analysis on syntheticindexes of the covariates of interest though OLS, in order to produce a lin-ear projection (at the respondent∗party level) of previously individual variables(for a more detailed discussion of this method, see: van der Brug, Franklinand Toka, 2008: p. 594). Due to the change of the dependent variable of theanalysis throughout chapters, two different sets of stacked data matrices havebeen generated, the first (as employed in Chapter 2) featuring partisanshipas dependent variable, the second (Chapters 3 and 4) replacing it with votechoice. By construction, the resulting y-hats are different across datasets – asdetailed in the descriptive analysis presented below.

A.6 Descriptive statistics of stacked data matricesemployed in Chapter 2 (dependent variable: partisanship)

Britain


Age 106041 −0.25 0.41 0.00 0.08Gender 106650 −0.03 0.04 0.00 0.03Education 100008 −0.31 0.10 0.00 0.08Social class 106650 −0.34 0.40 0.00 0.22Union membership 100623 −0.21 0.27 0.00 0.11Left–right self-placement 77580 −0.81 0.82 0.00 0.33Economic assessment 93594 −0.17 0.16 0.00 0.09Goldthorpe class 89412 −0.43 0.37 0.00 0.20Household income 84720 −0.24 0.22 0.00 0.11Region of residence 104349 −0.24 0.21 0.00 0.08Religious denomination 101796 −0.22 0.25 0.00 0.11

Germany


Age 106530 −0.16 0.26 0.00 0.04Gender 123675 −0.02 0.03 0.00 0.01Education 122735 −0.84 0.71 0.00 0.01Church attendance 114220 −0.23 0.31 0.00 0.08Union membership 113940 −0.14 0.22 0.00 0.05



Left–right self-placement 71300 −0.75 0.88 0.00 0.19Economic assessment 73075 −0.38 0.35 0.00 0.09Profession 122480 −0.30 0.34 0.00 0.05Region of residence 123675 −0.20 0.21 0.00 0.06Religious denomination 122915 −0.17 0.18 0.00 0.08Unemployment status 123370 −0.02 0.02 0.00 0.01Urban/rural 123675 −0.05 0.05 0.00 0.02

The Netherlands


Age 229977 −0.28 0.52 0.00 0.06Gender 230616 −0.01 0.01 0.00 0.01Education 224352 −0.12 0.12 0.00 0.05Church attendance 229221 −0.31 0.62 0.00 0.13Social class 219753 −0.29 0.31 0.00 0.07Union membership 169992 −0.09 0.20 0.00 0.04Left–right self-placement 168633 −0.57 0.66 0.00 0.14Economic assessment 124704 −0.17 0.18 0.00 0.05Household income 198549 −0.18 0.18 0.00 0.04Religious denomination 229752 −0.31 0.28 0.00 0.11Unemployment status 228978 −0.08 0.10 0.00 0.03Urban/rural 230625 −0.13 0.14 0.00 0.03

A.7 Descriptive statistics of stacked data matricesemployed in Chapters 3 and 4 (dependent variable:vote choice)

Britain


Age 106041 −0.09 0.14 0.00 0.03Gender 106650 −0.01 0.01 0.00 0.01Education 100008 −0.11 0.06 0.00 0.03Social class 106650 −0.13 0.16 0.00 0.09Union membership 100623 −0.09 0.12 0.00 0.05Left–right

self-placement77580 −0.36 0.32 0.00 0.14



(Continued)


Economicassessment

93594 −0.07 0.07 0.00 0.04

Goldthorpe class 89412 −0.16 0.14 0.00 0.08Household income 84720 −0.09 0.10 0.00 0.05Region of residence 104349 −0.11 0.09 0.00 0.04Religious

denomination101796 −0.09 0.12 0.00 0.05

Germany


Age 106530 −0.08 0.13 0.00 0.02Gender 123675 −0.01 0.01 0.00 0.00Education 122735 −0.51 0.48 0.00 0.01Church attendance 114220 −0.14 0.19 0.00 0.05Union membership 113940 −0.08 0.11 0.00 0.03Left–right self-placement 71300 −0.40 0.47 0.00 0.11Economic assessment 73075 −0.21 0.19 0.00 0.05Profession 122480 −0.19 0.26 0.00 0.03Region of residence 123675 −0.15 0.13 0.00 0.04Religious denomination 122915 −0.11 0.10 0.00 0.05Unemployment status 123370 −0.01 0.01 0.00 0.01Urban/rural 123675 −0.06 0.06 0.00 0.02

The Netherlands


Age 229977 −0.08 0.16 0.00 0.02Gender 230616 −0.01 0.01 0.00 0.00Education 224352 −0.07 0.08 0.00 0.03Church attendance 229221 −0.14 0.28 0.00 0.06Social class 219753 −0.13 0.17 0.00 0.03Union membership 169992 −0.05 0.09 0.00 0.02Left–right

self-placement168633 −0.31 0.36 0.00 0.08

Economicassessment

124704 −0.10 0.11 0.00 0.03


PROOF105

Household income 198549 −0.11 0.11 0.00 0.02Religious

denomination229752 −0.15 0.15 0.00 0.05

Unemploymentstatus

228978 −0.03 0.03 0.00 0.01

Urban/rural 230625 −0.07 0.08 0.00 0.02



PROOF

Appendix B: Detailed modelestimation procedures

Table B.1 Social structure and partisanship in Britain – by party family

Conservatives 1970s 1980s 1990s 2000s

Social class 0.23(0.01)∗∗ 0.22(0.01)∗∗ 0.25(0.01)∗∗ 0.18(0.02)∗∗

Union membership 0.20(0.02)∗∗ 0.13(0.01)∗∗ 0.15(0.01)∗∗ 0.13(0.02)∗∗

Nagelkerke R2 0.137 0.102 0.107 0.085McFadden R2 0.063 0.046 0.050 0.043N 6080 7623 6980 7574

Labour 1970s 1980s 1990s 2000s

Social class 0.25(0.01)∗∗ 0.25(0.01)∗∗ 0.22(0.01)∗∗ 0.20(0.01)∗∗



Liberal Democrats 1970s 1980s 1990s 2000s

Social class 0.28(0.13)∗ 0.42(0.04)∗∗ 0.46(0.13)∗∗ 0.43(0.16)∗∗

Union membership 2.69(0.82)∗∗ −0.70(0.66) −0.54(0.79) −0.79(0.77)Nagelkerke R2 0.016 0.006 0.020 0.015McFadden R2 0.010 0.003 0.013 0.010N 6080 7623 6980 7574

Note: Dependent variable: Partisanship (4 cat.) on a stacked data matrix (partysubsamples). Cell entries are standardized ordered probit estimates. Standard error esti-mates (in parentheses) are clustered robust at the individual level. ∗∗p < 0.01, ∗p < 0.05.Intercepts and controls (age, gender, educational level) included, coefficients not shown.

Table B.2 Social structure and partisanship in Germany – by party family

CDU/CSU 1970s 1980s 1990s 2000s

Religiousness 0.17(0.01)∗∗ 0.13(0.01)∗∗ 0.15(0.01)∗∗ 0.18(0.01)∗∗


Nagelkerke R2 0.089 0.056 0.100 0.120

106


PROOF107

McFadden R2 0.049 0.031 0.054 0.061N 4978 5515 3960 5581

SPD 1970s 1980s 1990s 2000s

Religiousness 0.35(0.03)∗∗ 0.18(0.03)∗∗ 0.02(0.03) 0.03(0.02)Union mem-

bership0.07(0.01)∗∗ 0.07(0.01)∗∗ 0.07(0.01)∗∗ 0.07(0.01)


FDP 1970s 1980s 1990s 2000s

Religiousness 1.12(1.67) −1.63(2.08) −5.95(2.04)∗∗ −1.93(1.37)Union mem-

bership0.17(0.25) 0.78(0.38)∗ 0.87(0.43)∗ 0.75(0.28)∗∗



Table B.3 Social structure and partisanship in the Netherlands – by partyfamily

CDA 1970s 1980s 1990s 2000s

Religiousness 0.27(0.01)∗∗ 0.22(0.01)∗∗ 0.23(0.01)∗∗ 0.19(0.01)∗∗

Social class 0.25(0.12)∗ 0.50(0.09)∗∗ 0.35(0.14)∗ 0.29(0.10)Union membership −0.19(0.09)∗ 0.07(0.06) 0.17(0.09) 0.03(0.06)Nagelkerke R2 0.335 0.250 0.271 0.185McFadden R2 0.193 0.141 0.178 0.115N 2646 5830 3242 5930

PvdA 1970s 1980s 1990s 2000s

Religiousness 0.53(0.03)∗∗ 0.41(0.02)∗∗ 0.30(0.03)∗∗ 0.25(0.03)∗∗

Social class 0.15(0.02)∗∗ 0.13(0.01)∗∗ 0.11(0.02)∗∗ 0.06(0.01)∗∗




PROOF108

Table B.3 (Continued)

VVD 1970s 1980s 1990s 2000s

Religiousness 0.49(0.07)∗∗ 0.45(0.05)∗∗ 0.44(0.07)∗∗ 0.52(0.06)∗∗

Social class 0.27(0.02)∗∗ 0.19(0.01)∗∗ 0.17(0.02)∗∗ 0.17(0.01)∗∗




Table B.4 The attitudinal drivers of partisanship in Britain – by party family

Conservatives 1980s 1990s 2000s

Leader evaluations 0.58(0.04)∗∗ 0.70(0.03)∗∗ 0.83(0.03)∗∗

Ideological proximity 0.35(0.02)∗∗ 0.36(0.02)∗∗ 0.66(0.03)∗∗

Economic assessment 0.15(0.02)∗∗ 0.10(0.02)∗∗ 0.17(0.02)∗∗


Labour 1980s 1990s 2000s



Economic assessment 0.18(0.02)∗∗ 0.03(0.02) 0.12(0.02)∗∗


Liberal Democrats 1980s 1990s 2000s


Ideological proximity −1.07(1.80) 0.06(2.17) −10.96(2.96)∗∗

Economic assessment 0.11(0.28) −0.12(0.26) −0.40(0.31)


PROOF109



Table B.5 The attitudinal drivers of partisanship in Germany – by partyfamily

CDU/CSU 1980s 1990s 2000s

Leader evaluations 0.94(0.04)∗∗ 0.81(04)∗∗ 0.84(0.04)∗∗




SPD 1980s 1990s 2000s



Economic assessment −0.01(0.02) 0.10(0.03)∗∗ 0.07(0.03)∗


FDP 1980s 1990s 2000s


Ideological proximity −0.04(0.78) 2.45(1.02)∗ −0.54(1,16)Economic assessment −18.02(21.33) −23.63(37.25) 26.62(24.56)Nagelkerke R2 0.138 0.192 0.321McFadden R2 0.120 0.176 0.271N 3075 3118 1797



PROOF110

Table B.6 The attitudinal drivers of partisanship in the Netherlands – byparty family

CDA 1980s 1990s 2000s





PvdA 1980s 1990s 2000s



Economic assessment 0.22(0.03)∗∗ −0.02(0.02) 0.09(0.02)∗∗


VVD 1980s 1990s 2000s



Economic assessment 0.79(0.16)∗∗ −0.28(0.09)∗∗ −0.08(0.06)Nagelkerke R2 0.327 0.344 0.280McFadden R2 0.228 0.225 0.187N 2574 3189 5536


Table B.7 Estimation procedure for instrumental variables

Britain b (S.E.) B

Age 2.45∗∗ (0.14) 0.07Education 0.60∗∗ (0.11) 0.02Gender 4.89∗∗ (0.56) 0.03Goldthorpe class 2.04∗∗ (0.05) 0.17Income 1.28∗∗ (0.09) 0.06Region of residence 1.82∗∗ (0.10) 0.07Religious denomination 2.01∗∗ (0.08) 0.10Constant 0.56∗∗ (0.00) –


PROOF111

Multiple R 0.253Adj. R-squared 0.064N 68042

Germany b (S0.E0.) B

Age 1.95∗∗ (0.09) 0.07Education 0.10 (0.25) 0.00Gender 0.15 (0.41) 0.00Profession 1.28∗∗ (0.05) 0.07Region of residence 0.74∗∗ (0.05) 0.05Religious denomination 1.19∗∗ (0.04) 0.10Unemployment status −0.79∗∗ (0.28) −0.01Urban/rural 0.13 (0.08) 0.01Constant 0.19∗∗ (0.00) –Multiple R 0.164Adj. R-squared 0.027N 103664

The Netherlands b (S.E.) B

Age 2.17∗∗ (0.07) 0.08Education 1.08∗∗ (0.05) 0.05Gender −0.40 (0.43) −0.00Income 1.29∗∗ (0.05) 0.06Religious denomination 1.73∗∗ (0.02) 0.18Unemployment status −0.52∗∗ (0.10) −0.01Urban/rural 0.81∗∗ (0.06) 0.03Constant 0.13∗∗ (0.00) –Multiple R 0.234Adj. R-squared 0.055N 191006

Note: Dependent variable: Partisanship (4 cat.) on a stacked data matrix. Cell entries areOLS regression estimates. Standard error estimates in parentheses. ∗∗p < 0.01, ∗p < 0.05.

Table B.8 Instrumental variable estimation, British parties


Social class 0.16(0.03)∗∗ 0.16(0.05)∗∗ 0.16(0.06)∗∗

Union membership 0.14(0.04)∗∗ 0.09(0.06) 0.09(0.06)Partisanship (exogenous) 0.27(0.04)∗∗ 0.36(0.06)∗∗ 0.54(0.06)∗∗



PROOF112




Economic assessment 0.20(0.04)∗∗ 0.13(0.05)∗ 0.44(0.06)∗∗

Constant −1.78(0.10)∗∗ −1.52(0.09)∗∗ −1.35(0.08)∗∗


Labour 1980s 1990s 2000s

Social class 0.24(0.05)∗∗ 0.19(0.05)∗∗ 0.22(0.05)∗∗

Union membership 0.13(0.05)∗∗ 0.15(0.04)∗∗ 0.17(0.03)∗∗

Partisanship (exogenous) 0.41(0.05)∗∗ 0.30(0.05)∗∗ 0.13(0.05)∗∗




Constant −1.20(0.08)∗∗ −1.30(0.07)∗∗ −0.89(0.06)∗∗


Liberal Democrats 1980s 1990s 2000s

Social class −0.23(0.22) 0.39(0.30) 0.72(0.34)∗

Union membership −35.56(21.24) −12.85(30.11) 15.08(25.41)Partisanship (exogenous) 0.84(0.15)∗∗ 0.44(0.19)∗ 0.71(0.17)∗∗


Ideological proximity 3.96(3.94) 0.18(5.91) 27.51(7.42)∗∗

Economic assessment 0.83(0.28)∗∗ 0.40(0.35) −0.27(0.34)Constant −2.00(0.08)∗∗ −2.31(0.10)∗∗ −2.14(0.08)∗∗


Note: Dependent variable: Vote choice (dummy) on a stacked data matrix (partysubsamples). Cell entries are standardized logistic estimates. Standard error estimates(in parentheses) are clustered robust at the individual level. ∗∗p<0.01, ∗p<0.05. Controls(age, gender, educational level) included, coefficients not shown.


PROOF113

Table B.9 Instrumental variable estimation, German parties

CDU/CSU 1980s 1990s 2000s

Religiousness 0.18(0.03)∗∗ 0.06(0.03)∗ 0.07(0.03)∗

Union membership 0.14(0.05)∗∗ 0.04(0.04) −0.05(0.05)Partisanship (exogenous) 0.08(0.05) 0.20(0.04)∗∗ 0.14(0.05)∗∗



Economic assessment −0.00(0.03) 0.11(0.03)∗∗ 0.13(0.03)∗∗

Constant −2.03(0.10)∗∗ −1.41(0.07)∗∗ −1.35(0.07)∗∗


SPD 1980s 1990s 2000s

Religiousness 0.34(0.08)∗∗ −0.08(0.05) 0.02(0.07)Union membership 0.07(0.03)∗∗ 0.07(0.02)∗∗ 0.03(0.03)Partisanship (exogenous) 0.20(0.05)∗∗ 0.22(0.05)∗∗ 0.27(0.07)∗∗



Economic assessment 0.07(0.05) 0.15(0.06)∗ −0.12(0.07)Constant −1.02(0.07)∗∗ −1.25(0.06)∗∗ −1.85(0.07)∗∗


FDP 1980s 1990s 2000s

Religiousness −14.55(15.14) 29.08(16.51) 8.73(17.92)Union membership 0.14(0.33) 0.54(0.43) 0.07(0.41)Partisanship (exogenous) 0.81(0.25)∗∗ 0.89(0.28)∗∗ 0.47(0.31)Leader evaluations 1.05(0.13)∗∗ 1.35(0.15)∗∗ 2.23(0.17)∗∗

Ideological proximity 0.42(0.41) 2.02(0.56)∗∗ 1.04(0.62)Economic assessment 2.93(1.52) −1.67(2.64) 0.82(2.34)Constant −3.09(0.11)∗∗ −3.22(0.12)∗∗ −2.54(0.15)∗∗




PROOF114

Table B.10 Instrumental variable estimation, Dutch parties

CDA 1980s 1990s 2000s

Religiousness 0.15(0.03)∗∗ 0.27(0.03)∗∗ 0.07(0.02)∗∗

Social class 0.10(0.20) 0.04(0.026) 0.13(0.16)Union membership 0.08(0.08) 0.09(0.09) −0.03(0.06)Partisanship (exogenous) 0.22(0.04)∗∗ 0.39(0.05)∗∗ 0.27(0.03)∗∗




Constant −2.13(0.12)∗∗ −2.38(0.11)∗∗ −1.98(0.07)∗∗


PvdA 1980s 1990s 2000s

Religiousness 0.22(0.08)∗∗ 0.46(0.08)∗∗ 0.20(0.07)∗∗

Social class 0.17(0.04)∗∗ 0.18(0.04)∗∗ 0.10(0.03)∗∗

Union membership 0.05(0.02)∗ 0.09(0.02)∗∗ 0.07(0.02)∗∗

Partisanship (exogenous) 0.03(0.09) −0.06(0.08) 0.03(0.07)Leader evaluations 1.10(0.08)∗∗ 0.97(0.08)∗∗ 1.21(0.07)∗∗



Constant −1.27(0.09)∗∗ −2.01(0.09)∗∗ −2.27(0.07)∗∗


VVD 1980s 1990s 2000s

Religiousness 1.60(0.17)∗∗ 0.86(0.12)∗∗ 1.06(0.10)∗∗

Social class 0.18(0.05)∗∗ 0.20(0.04)∗∗ 0.21(0.03)∗∗

Union membership 0.11(0.06) 0.16(0.04)∗∗ 0.08(0.03)∗

Partisanship (exogenous) 0.38(0.12) 0.33(0.09)∗∗ 0.22(0.07)∗∗



Economic assessment 1.28(0.40)∗∗ −0.29(0.23) −0.20(0.15)Constant −3.02(0.17)∗∗ −1.96(0.09)∗∗ −3.02(0.10)∗∗




PROOF115

Table B.11 Leader effects by educational level, Britain

Low educational level 1980s 1990s 2000s

Social class 0.20(0.03)∗∗ 0.23(0.04)∗∗ 0.11(0.05)∗






Constant −1.48(0.04)∗∗ −1.71(0.05)∗∗ −1.43(0.05)∗∗


High educational level 1980s 1990s 2000s

Social class 0.12(0.05)∗ 0.22(0.06)∗∗ 0.18(0.06)∗∗

Union membership 0.13(0.05)∗ 0.13(0.06)∗ 0.10(0.04)∗

Partisanship (exogenous) 0.13(0.07)∗ 0.13(0.08) 0.29(0.06)∗∗




Constant −1.59(0.07)∗∗ −1.71(0.09)∗∗ −1.41(0.05)∗∗


Note: Dependent variable: Vote choice (dummy) on a stacked data matrix. Subsamples arestratified by educational level. Cell entries are standardized logistic estimates. Standarderror estimates (in parentheses) are clustered robust at the individual level. ∗∗p < 0.01,∗p < 0.05. Controls (age, gender, educational level) included, coefficients not shown.

Table B.12 Leader effects by educational level, Germany


Religiousness 0.14(0.04)∗∗ 0.05(0.04) −0.08(0.06)Union membership 0.12(0.03)∗∗ 0.14(0.03)∗∗ −0.06(0.06)Partisanship (exogenous) 0.18(0.04)∗∗ 0.27(0.05)∗∗ 0.35(0.07)∗∗



Economic assessment −0.03(0.03) 0.12(0.05)∗ 0.13(0.06)∗

Constant −3.23(0.10)∗∗ −2.83(0.08)∗∗ −2.30(0.08)∗∗

Nagelkerke R2 0.592 0.529 0.444


PROOF116 Appendix B: Detailed model estimation procedures



Cox & Snell R2 0.410 0.304 0.256N 6358 6196 3508


Religiousness 0.20(0.05)∗∗ 0.01(0.03) 0.03(0.03)Union membership 0.07(0.04) 0.04(0.02) 0.04(0.03)Partisanship (exogenous) 0.11(0.06) 0.12(0.03)∗∗ 0.14(0.04)∗∗



Economic assessment −0.02(0.03) 0.15(0.04)∗∗ 0.04(0.03)Constant −2.21(0.08)∗∗ −2.22(0.04)∗∗ −1.94(0.05)∗∗



Table B.13 Leader effects by educational level, the Netherlands


Religiousness 0.11(0.03)∗∗ 0.29(0.04)∗∗ 0.13(0.04)∗∗

Social class 0.30(0.03)∗∗ 0.30(0.03)∗∗ 0.28(0.04)∗∗

Union membership 0.10(0.03)∗∗ 0.11(0.03)∗∗ 0.06(0.03)∗




Economic assessment 0.05(0.03) 0.10(0.03)∗∗ 0.08(0.03)∗∗

Constant −2.22(0.06)∗∗ −2.29(0.05)∗∗ −2.73(0.05)∗∗



PROOFAppendix B: Detailed model estimation procedures 117


Religiousness 0.31(0.05)∗∗ 0.36(0.05)∗∗ 0.12(0.03)∗∗

Social class 0.22(0.05)∗∗ 0.24(0.04)∗∗ 0.27(0.02)∗∗





Economic assessment −0.01(0.04) −0.10(0.04)∗ 0.14(0.02)∗∗

Constant −2.03(0.08)∗∗ −2.59(0.07)∗∗ −2.90(0.04)∗∗





PROOF

Appendix C: Parties and partyleaders in Britain, Germany, andthe Netherlands (1980–2010)

Britain

Conservative party Labour party Liberal Democrats

1983 Margaret Thatcher Michael Foot David Steel1987 Margaret Thatcher Neil Kinnock David Steel1992 John Major Neil Kinnock Paddy Ashdown1997 John Major Tony Blair Paddy Ashdown2001 William Hague Tony Blair Charles Kennedy2005 Michael Howard Tony Blair Charles Kennedy2010 David Cameron Gordon Brown Nick Clegg

Germany

ChristlichDemokratischeUnion/Christlich-Soziale Union inBayern (CDU/CSU)

Freie DemokratischePartei (FDP)

SozialdemokratischeParteiDeutschlands(SPD)

1980 Franz-Josef Strauß Hans-DietrichGenscher

Helmut Schmidt

1983 Helmut Kohl Hans-DietrichGenscher

Hans-Jochen Vogel

1987 Helmut Kohl Martin Bangemann Johannes Rau1990 Helmut Kohl Otto Graf Lambsdorff Oskar Lafontaine1994 Helmut Kohl Klaus Kinkel Rudolf Scharping1998 Helmut Kohl Wolfgang Gerhardt Gerhard Schröder2002 Edmund Stoiber Guido Westerwelle Gerhard Schröder2005 Angela Merkel Guido Westerwelle Gerhard Schröder2009 Angela Merkel Guido Westerwelle Frank-Walter

Steinmeier

118


PROOFAppendix C 119

The Netherlands

Christen-DemocratischAppèl (CDA)

Partij van deArbeid (PvdA)

Volkspartij voorVrijheid enDemocratie (VVD)

1986 Ruud Lubbers Joop den Uyl Ed Nijpels1989 Ruud Lubbers Wim Kok Joris Voorhoeve1994 Elco Brinkman Wim Kok Frits Bolkestein1998 Jaap de Hoop Scheffer Wim Kok Frits Bolkestein2002 Jan Peter Balkenende Ad Melkert Hans Dijkstal2003 Jan Peter Balkenende Wouter Bos Gerrit Zalm2006 Jan Peter Balkenende Wouter Bos Mark Rutte2010 Jan Peter Balkenende Job Cohen Mark Rutte


October 24, 2013 18:46 MAC/PERSON Page-120 9781137270221_11_not01

PROOF

Notes

Introduction

1. The seemingly related concept of presidentialization (Poguntke and Webb,2005) will be touched upon here and there, but this is not what this book isabout. The connection, if any, can be found with one of the three faces ofthe presidentialization – namely, its electoral face, which is rightly thoughtto involve “a shift from partified control to a domination by leaders”(Poguntke and Webb, 2005: p. 10).

1 The Personalization of Politics

1. It must be highlighted that this interpretation of the causal dynamicsunderlying party change does not find unequivocal consensus in the lit-erature. For example, Evans and Tillie’s (2012) analysis of Britain arguesthat the decline of the class–party association in British politics occurredprimarily as a result of ideological convergence between the main partiesrather than by increasing class heterogeneity (e.g., dealignment). Furtherinterpretative accounts focus on the links between parties and cleavage-related social organizations as main explanatory factors (Bellucci andHeath, 2012).

2. Rather more optimistic conclusions are reached in a recently availablecomparative study by Bittner (2011). However, her book-length analysisof 35 election studies from seven countries is mostly concerned with vot-ers’ perception of leaders’ personality traits and the conditional factorsmediating the effect of these perceptions on the vote. Only one chapter(Chapter 6) is devoted to the effect of party leaders on the individual vot-ing calculus. Furthermore, the analysis of the impact of leader perceptionson the electoral outcomes is confined to just one presidential system (i.e.,the US).

2 Attitudinal Consequences

1. Question wording reads as follows: Britain: Generally speaking, do you thinkof yourself as Labour, Conservative, Liberal Democrat, or what? Germany: Manypeople in the Federal Republic lean toward a particular party for a long time,although they may vote for a different party. How about you? Do you in gen-eral lean toward a particular party? The Netherlands: Many people think ofthemselves as adherents of a particular party, but there are many other peoplewho do not regard themselves as such. How about you, do you regard yourself as

120


PROOFNotes 121

an adherent of a political party or don’t you? One notes that question word-ing is not fully comparable throughout countries. Nevertheless, nontrivialsemantic similarities can be found between the ways in which respondentsare asked about their attachment to parties in the three sets of nationalelection studies under analysis. As Dalton (2010) argues, these conceptionssacrifice “the notion of long-term partisan identity for a feeling of close-ness to a party”, but at the same time tap “affinity to a party separatefrom the vote, and it can be used in systems with diverse party tradi-tions” (Dalton, 2010: p. 159). Furthermore, question wording has beenkept constant in each national survey, thus allowing comparison over timewithin countries.

2. These measures are simultaneously included only in the analyses of theDutch case. Respondents’ subjective social class is in fact not availablein the German datasets, while religiousness is deliberately excluded fromthe analysis of British data (for a discussion of Britain’s unidimensionalcleavage structure, see: Oskarson, 2005).

3. Although discrete choice models do not offer a straightforward counter-part to the R-squared in Ordinary Least Squares (OLS) regression, measuresof fit based on the overall model chi-squared (such as McFadden’s pseudoR-squared, or its adjusted Nagelkerke’s version) provide a satisfactory alter-native (Greene and Hensher, 2010). Chi-squared measures of fit assess “thefit of the predictions by the model to the observed data, compared to nomodel” (Greene and Hensher, 2010: 126).

4. Checks both on the correlation matrix of the independent variables (allinter-correlations are less than r =. 40) and the variance inflation factors(reported values are all below 2) assure that their simultaneous inclusionin the model is safe from problems of multicollinearity.

5. Dutch studies did not ask respondents to evaluate party leaders on thefeeling thermometer until 1986.

3 Behavioral Consequences

1. For a lucid standard treatment of instrumental variables estimation, seeKmenta (1997); for something more current, see Woolridge (2006). Usefulexamples of the technique, applied to election survey research, appear inFiorina (1981) and Lewis-Beck, Nadeau, and Elias (2008).

2. Based on data availability, the set of exogenous variables employed in theconstruction of the instruments varies slightly across datasets. The vari-ables employed are as follows. Britain: age, education, gender, Goldthorpeclass, annual household income, region of residence, religious denomi-nation. Germany: age, education, gender, profession, region of residence,religious denomination, unemployment status, degree of urbanization ofthe respondent’s town of residence. The Netherlands: age, education, gen-der, annual household income, religious denomination, unemploymentstatus, degree of urbanization of the respondent’s town of residence. Fullestimation procedure is available in Appendix B.7. The instruments arrived


PROOF122 Notes

at are “good”, as their Pearson’s correlation with the original partisanshipvariable ranges between 0.17 in the German case and 0.25 in the Britishcase – these values being substantially comparable to those reported byLewis-Beck, Nadeau, and Elias (2008: p. 91) in a similar exercise.

3. To test the robustness of these findings, a number of jackknife tests havebeen performed. To evaluate the stability of the instruments, I excludedone exogenous variable at a time from the construction of each instru-ment, every time re-estimating the model with the new instrument. Themodel-fit remains in every instance practically unchanged, thus assur-ing that the performance of the models does not rest on the presenceor absence of any specific exogenous variable in the construction of theinstruments.

4 Electoral Consequences

1. The British election of 2010 is excluded due to the lack of ideologicalproximity and economic assessment measures in the dataset. Similarly,the Dutch election of 2003 is excluded because of the lack of economyas well as party identification variables. Two German elections have beenexcluded: namely, those of 1980 (missing variables on ideological proxim-ity and economic assessment) and 2002 (thermometer measures were onlyavailable for the SPD and CDU/CSU leaders).

2. The use of the conditional is necessary due to a number of compli-cations arising, in the first place, from the disjunction that exists innonproportional systems between winning votes and winning seats inparliaments. Especially in FPTP systems like the British one, votes donot translate neatly into seats. A further, and to some extent even moresevere, complication arises from the practice of coalition governments(e.g., Germany, the Netherlands). According to King (2002c), “[c]ountriesin which there is a disjunction between vote-winning at elections and theprocess of government-formation following elections are common . . . onceagain, anyone interested in assessing the political impact of party leaders’personal characteristics needs to decide in advance what ‘victory’ meansin this kind of contexts” (p. 219).

5 Normative and Empirical Implications

1. In post-election surveys, in fact, respondents are asked about their votechoice after the election has taken place. Even assuming that respondents’vote recall is reported sincerely, this may still provide them with a sufficientspan of time to “shape” their attitudes in a way that conforms more closelyto their past behavior. Moreover, the fact that political attitudes are actu-ally being measured after the election provides further ground to believethat, if anything, cognitive feedback is actually running from behavior to


PROOFNotes 123

attitudes (for a better discussion of this point, see: Garzia and De Angelis,2011).

2. While recognizing the potential drawbacks due to the use of suboptimalcross-sectional data sources also in the context of this study, it must alsobe highlighted that a reassessment of leader effects across three countriesand five decades could only take place taking advantage of the availablepost-election datasets.


October 25, 2013 14:48 MAC/PERSON Page-124 9781137270221_12_ref01

PROOF

References

Aarts, K. (2001). The Impact of Leaders on Electoral Choice in theNetherlands – Revisited. Acta Politica 36: 380–401.

Aarts, K.; Blais, A. (2011). Pull or Push? The Relative Impact of Positive andNegative Leader Evaluations on Vote Choice. In K. Aarts, A. Blais andH. Schmitt (eds), Political Leaders and Democratic Elections. Oxford: OxfordUniversity Press.

Aarts, K.; Blais, A.; Schmitt, H. (eds) (2011). Political Leaders and DemocraticElections. Oxford: Oxford University Press.

Adam, S.; Maier, M. (2010). Personalization of Politics: A Critical Reviewand Agenda for Research. In C. Salmon (ed.), Communication Yearbook 34.London: Routledge.

Anderson, C. J.; Mendes, S. M.; Tverdova, Y. V. (2004). Endogenous EconomicVoting: Evidence from the 1997 British Election. Electoral Studies 23(4):683–708.

Andeweg, R. (2000). Fractiocracy? Limits to the Ascendancy of the Parlia-mentary Party Group in Dutch Politics. In R. Koole and K. Heidar (eds),Parliamentary Party Groups in European Democracies. London and New York:Routledge.

Andeweg, R.; Irwin, G. (2003). Governance and Politics in the Netherlands.London: Macmillan.

APSA (1950). Towards a More Responsible Two-Party System. American PoliticalScience Review 44: supplement.

Archer, K. (1987). A Simultaneous Equation Model of Canadian VotingBehaviour. Canadian Journal of Political Science 20: 553–572.

Arzheimer, K. (2006). Dead Men Walking? Party Identification in Germany,1977–2002. Electoral Studies 25: 791–807.

Asp, K.; Esaiasson, P. (1996). The Modernization of Swedish Campaigns: Indi-vidualization, Professionalization, and Medialization. In D. Swanson and P.Mancini (eds), Politics, Media, and Modern Democracy: An International Studyof Innovations in Electoral Campaigning and Their Consequences. Westport, CT:Praeger.

Baldassarri, D. (2013). The Simple Art of Voting: The Cognitive Shortcuts of ItalianCitizens. New York: Oxford University Press.

Barisione, M. (2009). So What Difference Do Leaders Make? Candidates’Images and the “Conditionality” of Leader Effects on Voting. Journal ofElections, Public Opinion and Parties 19: 473–500.

Bartels, L. (2002). The Impact of Candidate Traits in American Presiden-tial Elections. In A. King (ed.), Leaders’ Personalities and the Outcomes ofDemocratic Elections. Oxford: Oxford University Press.

124


PROOFReferences 125

Bartle, J.; Bellucci, P. (2009a). Partisanship, Social Identity and Individual Atti-tudes. In J. Bartle and P. Bellucci (eds), Political Parties and Partisanship. SocialIdentity and Individual Attitudes. London and New York: Routledge.

Bartle, J.; Bellucci, P. (eds) (2009b). Political Parties and Partisanship. SocialIdentity and Individual Attitudes. London and New York: Routledge.

Bartle, J.; Crewe, I. (2002). The Impact of Party Leaders in Britain: StrongAssumptions, Weak Evidence. In A. King (ed.), Leaders’ Personalities and theOutcomes of Democratic Elections. Oxford: Oxford University Press.

Bartolini, S.; Mair, P. (1990). Identity, Competition and Electoral Availability:The Stabilisation of European Electorates, 1885–1985. Cambridge: CambridgeUniversity Press.

Bartolini, S.; Mair, P. (2007). Identity, Competition and Electoral Availability: TheStabilisation of European Electorates, 1885–1985. 2nd Edition. Colchester:ECPR Press.

Bean, C. (1993). The Electoral Influence of Party Leader Images in Australiaand New Zealand. Comparative Political Studies 26: 111–132.

Bean, C.; Mughan, A. (1989). Leadership Effects in Parliamentary Elections inAustralia and Britain. American Political Science Review 83: 1165–1179.

Bellucci, P. (2006). Tracing the Cognitive and Affective Roots of “PartyCompetence”: Italy and Britain, 2001. Electoral Studies 25: 534–568.

Bellucci, P.; Garzia, D.; Lewis-Beck, M. S. (2013). Issues and Leaders as VoteDeterminants: The Case of Italy. Party Politics, article in press.

Bellucci, P.; Heath, O. (2012). The Structure of Party-Organization Linkagesand the Electoral Strength of Cleavages in Italy, 1963–2008. British Journalof Political Science 42: 107–135.

Bellucci, P.; Segatti, P. (eds) (2011). Votare in Italia: 1968–2008.Dall’appartenenza alla scelta. Bologna: Il Mulino.

Berelson, B.; Lazarsfeld, P.; McPhee, W. (1954). Voting. A Study of OpinionFormation in a Presidential Campaign. Chicago: University of Chicago Press.

Berglund, F.; Holmberg, S.; Schmitt, H.; Thomassen, J. (2005). Party Identifi-cation and Party Choice. In J. Thomassen (ed.), The European Voter. Oxford:Oxford University Press.

Bittner, A. (2011). Platform or Personality? The Role of Party Leaders in Elections.Oxford: Oxford University Press.

Blais, A. (2011). Political Leaders and Democratic Elections. In K. Aarts, A. Blaisand H. Schmitt (eds), Political Leaders and Democratic Elections. Oxford:Oxford University Press.

Blondel, J.; Thiébault, J. (2010). Political Leadership, Parties and Citizens.London and New York: Routledge.

Bowler, S.; Farrell, D. (1992). Electoral Strategies and Political Marketing.New York: Macmillan.

Brettschneider, F.; Gabriel, O. (2002). The Nonpersonalization of VotingBehaviour in Germany. In A. King (ed.), Leaders’ Personality and the Outcomesof Democratic Elections. Oxford: Oxford University Press.

Brettschneider, F.; Neller, K; Anderson, C. (2006). Candidate Images in the2005 German National Election. German Politics 15: 481–499.


PROOF126 References

Budge, I.; Crewe, I.; Farlie, D. (eds) (1976). Party Identification and Beyond.Hoboken, NJ: John Wiley and Sons.

Butler, D.; Stokes, D. (1969). Political Change in Britain. London: Macmillan.Butler, D.; Stokes, D. (1974). Political Change in Britain. 2nd Edition. London:

Macmillan.Campbell, A.; Converse, P.; Miller, W.; Stokes, D. (1960). The American Voter.

Chicago and London: The University of Chicago Press.Campbell, A.; Gurin, G.; Miller, W. (1954). The Voter Decides. Evanston, IL:

Row Peterson.Campus, D. (2000). L’elettore pigro. Informazione politica e scelte di voto. Bologna:

Il Mulino.Campus, D. (2010). Mediatization and Personalization of Politics in Italy

and France: the cases of Berlusconi and Sarkozy. International Journal ofPress/Politics 15: 219–235.

Campus, D.; Pasquino, G. (2006). Leadership in Italy: The Changing Role ofLeaders in Elections and in Government. Journal of Contemporary EuropeanStudies 14: 25–40.

Caprara, G.; Zimbardo, P. (2004). Personalizing Politics. American Psychologist59: 581–594.

Carmines, E. ; Stimson, J. (1980). The Two Faces of Issue Voting. AmericanPolitical Science Review, vol. 74, 78–91.

Clarke, H.; Sanders, D.; Stewart, M.; Whiteley, P. (2004). Political Choice inBritain. Oxford: Oxford University Press.

Clarke, H.; Sanders, D.; Stewart, M.; Whiteley, P. (2009a). Performance Politicsand the British Voter. Oxford: Oxford University Press.

Clarke, H.; Sanders, D.; Stewart, M.; Whiteley, P. (2009b). Travel Tips forStudents of Electoral Choice: The Dynamics of Partisanship in Britainand Elsewhere. In J. Bartle and P. Bellucci (eds), Political Parties andPartisanship. Social Identity and Individual Attitudes. London and New York:Routledge.

Clarke, H.; Sanders, D.; Stewart, M.; Whiteley, P. (2011). Valence Politics andElectoral Choice in Britain, 2010. Journal of Elections, Public Opinion & Parties21: 237–253.

Converse, P. (1964). The Nature of Belief Systems in Mass Politics. In D. Apter(ed.), Ideology and Discontent. New York: Free Press.

Crewe, I. (1976). Party Identification Theory and Political Change in Britain.In I. Budge, I. Crewe and D. Farlie (eds), Party Identification and Beyond.Hoboken, NJ: John Wiley and Sons.

Crewe, I.; King, A. (1994a). Did Major Win? Did Kinnock Lose? LeadershipEffects in the 1992 Election. In A. Heath, R. Jowell and. J. Curtice (eds),Labour’s Last Chance? The 1992 Election and Beyond. Aldershot: Dartmouth.

Crewe, I.; King, A. (1994b). Are British Elections Becoming More “Presiden-tial”? In M. Jennings and T. Mann (eds), Elections at Home and Abroad.Essays in Honor of Warren E. Miller. Ann Arbor: University of MichiganPress.

Crewe, I.; Sarlvik, B.; Alt, J. (1977). Partisan Dealignment in Britain, 1964–1974. British Journal of Political Science 7: 129–190.


PROOFReferences 127

Curtice, J.; Blais, A. (2001). Follow My Leader? A Cross-National Analy-sis of Leadership Effects in Parliamentary Democracies. Working Paper 91.Centre for Research into Elections and Social Trends, Department of Sociology,University of Oxford.

Curtice, J.; Holmberg, S. (2005). Party Leaders and Party Choice. InJ. Thomassen (ed.), The European Voter. Oxford: Oxford University Press.

Curtice, J.; Hunjan, S. (2011). Elections as Beauty Contests: Do the Rules Mat-ter? In K. Aarts, A. Blais and H. Schmitt (eds), Political Leaders and DemocraticElections. Oxford: Oxford University Press.

Cutler, F. (2002). The Simplest Shortcut of All: Sociodemographic Characteris-tics and Electoral Choice. Journal of Politics 64(2): 466–490.

Daalder, H. (2002). Parties: Denied, Dismissed, or Redundant? A Critique. InR. Gunther, J. Montero and J. Linz (eds), Political Parties: Old Concepts andNew Challenges. Oxford: Oxford University Press.

Dalton, R. (2008). Citizen Politics. Public Opinion and Political Parties inAdvanced Industrial Democracies. Washington: CQ Press.

Dalton, R. (2010). Ideology, Partisanship and Democratic Development. InL. LeDuc, R. Niemi and P. Norris (eds), Comparing Democracies 3. Electionsand Voting in the 21st Century. New York and London: Sage Publications.

Dalton, R.; McAllister, I.; Wattenberg, M. (2000). The Consequences of Par-tisan Dealignment. In R. Dalton and M. Wattenberg (eds), Parties withoutPartisans: Political Change in Advanced Industrial Democracies. Oxford: OxfordUniversity Press.

Dalton, R.; Wattenberg, M. (1993). The Not So Simple Act of Voting. InA. Finifter (ed.), Political Science. The State of the Discipline II. Washington:The American Political Science Association.

Dalton, R.; Wattenberg, M. (eds) (2000). Parties without Partisans: PoliticalChange in Advanced Industrial Democracies. Oxford: Oxford University Press.

Davis, D. (1990). News and Politics. In D. Swanson and D. Nimmo (eds), NewDirections in Political Communication. Newbury Park, CA: Sage Publications.

De Graaf, J. (2010). The Irresistible Rise of StemWijzer. In L. Cedroni andD. Garzia (eds), Voting Advice Applications in Europe: The State of the Art.Napoli: Civis.

Denver, D. (2007). Elections and Voters in Britain. Basingstoke: PalgraveMacmillan.

Dinas, E. (2008). Big Expectations, Small Outcomes: The Impact of Leaders’Personal Appeal in the 2004 Greek Election. Electoral Studies 27: 505–517.

Downs, A. (1957). An Economic Theory of Democracy. New York: Harper.Druckman, J. (2003). The Power of Television Images: The First Kennedy-

Nixon Debate Revisited. Journal of Politics 65: 559–571.Eagly, A.; Chaiken, S. (1993). The Psychology of Attitudes. Orlando, FL: Harcourt

Brace Jovanovich.Evans, G.; Andersen, R. (2006). The Political Conditioning of Economic

Perceptions. Journal of Politics 68(1): 194–207.Evans, G.; Pickup, M. (2010). Reversing the Causal Arrow: The Political Condi-

tioning of Economic Perceptions in the 2000–2004 US Presidential ElectionCycle. Journal of Politics 72(4): 1236–1251.


PROOF128 References

Evans, G.; Andersen, R. (2005). The Impact of Party Leaders: How Blair LostLabour Votes. Parliamentary Affairs 58: 818–836.

Evans, G.; Tillie, J. (2012). How Parties Shape Class Politics: Explaining theDecline of the Class Basis of Party Support. British Journal of Political Science42, 137–161.

Farrell, D.; Webb, P. (2000). Political Parties as Campaign Organizations.In R. Dalton and M. Wattenberg (eds), Parties without Partisans: Politi-cal Change in Advanced Industrial Democracies. Oxford: Oxford UniversityPress.

Fiers, S.; Krouwel, A. (2005). The Low Countries: From “Prime Minister” toPresident-Minister. In T. Poguntke and P. Webb (eds), The Presidentializationof Politics. A Comparative Study of Modern Democracies. Oxford: OxfordUniversity Press.

Fiorina, M. (1981). Retrospective Voting in American National Elections. NewHaven: Yale University Press.

Fiorina, M. (2002). Parties and Partisanship: A 40-year Retrospective. PoliticalBehavior 24: 93–115.

Fishbein, M; Ajzen, I. (1975). Belief, Attitude, Intention, and Behavior: An Intro-duction in Theory and Research. Reading, MA: Addison-Wesley.

Franklin, C.; Jackson, J. (1983). The Dynamics of Party Identification.American Political Science Review 77: 957–973.

Franklin, M.; Mackie, T.; Valen, H. (eds) (1992). Electoral Change: Responsesto Evolving Social and Attitudinal Structures in Western Countries. Cambridge:Cambridge University Press.

Funder, D. C. (1999). Personality Judgment: A Realistic Approach to PersonPerception. San Diego, CA: Academic Press.

Funk, C. (1996). The Impact of Scandal on Candidate Evaluations: AnExperimental Test of the Role of Candidate Traits. Political Behavior 18: 1–24.

Funk, C. (1999). Bringing the Candidate into Models of Candidate Evaluation.Journal of Politics 61: 700–720.

Garzia, D. (2011). The Personalization of Politics in Western Democracies:Causes and Consequences on Leader-Follower Relationships. The LeadershipQuarterly 22: 697–709.

Garzia, D. (2012). Party and Leader Effects in Parliamentary Elections: Towardsa Reassessment. Politics 32: 175–185.

Garzia, D. (2013a). Changing Parties, Changing Partisans. The Personaliza-tion of Partisan Attachments in Western Europe. Political Psychology 34:67–89.

Garzia, D. (2013b). The Rise of Party/Leader Identification in Western Europe.Political Research Quarterly 66: 533–544.

Garzia, D.; De Angelis, A. (2011). Partisanship, Leader Evaluations, and theVote: Breaking the New Iron Triangle in Electoral Research. Paper Presentedat the 1st European Conference on Comparative Electoral Research, Sofia, 1–3December.

Garzia, D.; Viotti, F. (2011). Leader, identità di partito e voto in Italia, 1990–2008. Rivista Italiana di Scienza Politica 41: 411–432.


PROOFReferences 129

Gidengil, E. (2011). Voter Characteristics and Leader Effects. In K. Aarts,A. Blais and H. Schmitt (eds), Political Leaders and Democratic Elections.Oxford: Oxford University Press.

Gidengil, E.; Blais, A. (2007). Are Party Leaders Becoming More Importantto Vote Choice in Canada? In J. Courtney, H. Michelmann, D. Story andJ. Steeves (eds), Political Leadership and Representation in Canada. Essays inHonour of John C. Courtney. Toronto: University of Toronto Press.

Glass, D. (1985). Evaluating Presidential Candidates: Who Focuses on theirPersonal Attributes? Public Opinion Quarterly 49(4): 517–534.

Gould, J. (1972). The X of the Campaign: TV “Personality”. In J. Fixx (ed.),The Mass Media and Politics. New York: Arno Press.

Graber, D. (2001). Processing Politics. Learning from Television in the Internet Age.Chicago: University of Chicago Press.

Graber, D. (2006). Mass Media and American Politics. Seventh Edition.Washington, DC: CQ Press.

Graetz, B.; McAllister, I. (1987). Political Leadership and Electoral Outcomesin Britain, 1974–1983. Comparative Political Studies 19: 484–507.

Greene, W.; Hensher, D. (2010). Modeling Ordered Choices. Cambridge:Cambridge University Press.

Greene, S. (2001). The Role of Character Assessments in Presidential Approval.American Politics Research, 29: 196–210.

Gunther, R. (2005). Parties and Electoral Behavior in Southern Europe.Comparative Politics 37: 253–275.

Gunther, R.; Diamond, L. (2003). Species of Political Parties: A New Typology.Party Politics 9: 167–199.

Gunther, R.; Montero, J. (2001). The Anchors of Partisanship: A ComparativeAnalysis of Voting Behavior in Four Southern European Democracies. InN. Diamandouros and R. Gunther (eds), Parties, Politics, and Democracy in theNew Southern Europe. Baltimore and London: The Johns Hopkins UniversityPress.

Hart, P. (1999). Seducing America: How Television Charms the Modern Voter.Second Edition. New York: Oxford University Press.

Hayes, B.; McAllister, I. (1997). Gender, Party Leaders, and Election Outcomesin Australia, Britain, and the United States. Comparative Political Studies30: 3–26.

Hayes, D. (2009). Has Television Personalized Voting Behavior? PoliticalBehavior 31: 231–260.

Holmberg, S. (1994). Party Identification Compared across the Atlantic. InK. Jennings and T. Mann (eds), Elections at Home and Abroad. Essays in Honorof Warren Miller. Ann Arbor: University of Michigan Press.

Holmberg, S. (2007). Partisanship Reconsidered. In R. Dalton and H.Klingemann (eds), The Oxford Handbook of Political Behavior. Oxford: OxfordUniversity Press.

Holmberg, S.; Oscarsson, H. (2011). Party Leader Effects on the Vote. InK. Aarts, A. Blais and H. Schmitt (eds), Political Leaders and DemocraticElections. Oxford: Oxford University Press.


PROOF130 References

Holtz-Bacha, C., Lessinger, E., Hettesheimer, M. (1998). Personalisierungals Strategie der Wahlwerbung. In K. Imhof and P. Schulz (eds), DieVeröffentlichung des Privaten – Die Privatisierung des Öffentlichen. Wiesbaden:Westdeutscher Verlag.

Houghton, D. (2009). Political Psychology: Situations, Individuals, and Cases.New York and London: Routledge.

Inglehart, R.; Klingemann, H. (1976). Party Identification, Ideological Pref-erence and the Left-Right Dimensions among Western Mass Publics. InI. Budge, I. Crewe and D. Farlie (eds), Party Identification and Beyond.Hoboken, NJ: John Wiley and Sons.

Irwin, G.; van Holsteyn, J. (1989). Towards a More Open Model of Competi-tion. West European Politics 12: 112–138.

Johnston, R. (2002). Prime Ministerial Contenders in Canada. In A. King (ed.),Leaders’ Personalities and the Outcomes of Democratic Elections. Oxford: OxfordUniversity Press.

Johnston, R. (2006). Party Identification: Unmoved Mover or Sum of Prefer-ences? Annual Review of Political Science 9: 329–351.

Jones, R.; Hudson, J. (1996). The Quality of Political Leadership: A Case Studyof John Major. British Journal of Political Science 26: 229–244.

Kaase, M. (1994). Is There Personalization in Politics? Candidates and VotingBehavior in Germany. International Political Science Review 15: 211–230.

Karvonen, L. (2010). The Personalisation of Politics. A Study of ParliamentaryDemocracies. Colchester: ECPR Press.

Kennedy, P. (2008). A Guide to Econometrics. 6th Edition. Malden, MA:Blackwell Publishing.

Kinder, D. (1986). Presidential Character Revisited. In R. Lau and D. Sears(eds), Political Cognition. Hillsdale, NJ: Lawrence Erlbaum.

Kinder, D.; Abelson, R.; Fiske, S. (1979). Developmental Research on Candi-date Instrumentation: Results and Recommendations. Report Submitted tothe Institute for Social Research, University of Michigan.

King, A. (2002a). Leaders’ Personalities and the Outcomes of Democratic Elections.Oxford: Oxford University Press.

King, A. (2002b). Do Leaders’ Personalities Really Matter? In A. King (ed.),Leaders’ Personalities and the Outcomes of Democratic Elections. Oxford: OxfordUniversity Press.

King, A. (2002c). Conclusions and Implications. In A. King (ed.), Leaders’ Per-sonalities and the Outcomes of Democratic Elections. Oxford: Oxford UniversityPress.

Kirchheimer, O. (1966). The Transformation of the Western European PartySystem. In J. LaPalombara and M. Weiner (eds), Political Parties and PoliticalDevelopment. Princeton: Princeton University Press.

Kleinnijenhuis, J.; Oegema, D.; de Ridder, J.; van Hoof, A.; Vliegenthart,R. (2003). De puinhopen in het niews. De role van de media bij de Tweede-Kamerverkiezingen van 2002. Alphen aan den Rijn: Kluwer.

Kmenta, J. (1997). Elements of Econometrics. Second Edition. Ann Arbor:University of Michigan Press.


PROOFReferences 131

Koopmans, R.; Muis, J. (2009). The Rise of Right-wing Populist Pim Fortuynin the Netherlands: A Discursive Opportunity Approach. European Journal ofPolitical Research 48: 642–664.

Kriesi, H. (2012). Personalization of National Election Campaigns. PartyPolitics 18, 825–844.

Kuklinski, J. H.; Quirk, P. J. (2001). Conceptual Foundations of CitizenCompetence. Political Behavior 23(3): 285–311.

Lachat, R. (2009). Individual and Contextual Variation in the Impact of LeaderEvaluations and Issue Positions. Paper Presented at the ECPR Joint Sessions ofWorkshops, Lisbon, 14–19 April.

Langer, I. (2007). A Historical Exploration of the Personalisation of Politicsin the Print Media: The British Prime Ministers 1945–1999. ParliamentaryAffairs 60: 371–387.

Lanoue, D.; Headrick, B. (1994). Prime Ministers, Parties and the Public. PublicOpinion Quarterly 58: 191–209.

Lass, J. (1995). Vorstellungsbilder über Kanzlerkandidaten. Zur Diskussion um diePersonalisierung von Politik. Wiesbaden: Deutscher Universitätsverlag.

Lau, R. ; Redlawsk, D. (2006). How Voters Decide: Information Processing inElection Campaigns. Cambridge: Cambridge University Press.

Lazarsfeld, P.; Berelson, B.; Gaudet, H. (1944). The People’s Choice: How theVoter Makes Up His Mind in a Presidential Campaign. New York: ColumbiaUniversity Press.

LeDuc, L.; Niemi, R.; Norris, P. (1996). Comparing Democracies: Elections andVoting in Global Perspective. Thousand Oaks, CA: Sage Publications.

Lewis-Beck, M. (1988). Economics & Elections. The Major Western Democracies.Ann Arbor: The University of Michigan Press.

Lewis-Beck, M.; Nadeau, R.; Elias, A. (2008). Economics, Party, and the Vote:Causality Issues and Panel Data. American Journal of Political Science 52:84–95.

Lipset, S.; Rokkan, S. (1967). Party Systems and Voter Alignments. New York: TheFree Press.

Lobo, M. (2008). Parties and Leader Effects. Impact of Leaders in the Vote forDifferent Types of Parties. Party Politics 14: 281–298.

Mair, P.; Muller, W.; Plasser, F. (eds) (2004). Political Parties and Electoral Change:Party Responses to Electoral Markets. London: Sage Publications.

Manin, B. (1997). The Principles of Representative Government. Cambridge:Cambridge University Press.

Marks, G. (1993). Partisanship and the Vote in Australia: Changes over Time1967–1990. Political Behavior 15: 137–166.

Marsh, M. (2007). Candidates or Parties? Objects of Electoral Choice inIreland. Party Politics 13: 500–527.

McAllister, I. (1996). Leaders. In L. LeDuc, R. Niemi and P. Norris (eds), Com-paring Democracies. Elections and Voting in Global Perspective. Thousand Oaks,CA: Sage Publications.

McAllister, I. (2003). Prime Ministers, Opposition Leaders and GovernmentPopularity in Australia. Australian Journal of Political Science 38: 259–277.


PROOF132 References

McAllister, I. (2007). The Personalization of Politics. In R. Dalton andH. Klingemann (eds), The Oxford Handbook of Political Behavior. Oxford:Oxford University Press.

McAllister, I. (2011). Political Leaders in Westminster Systems. In K. Aarts,A. Blais and H. Schmitt (eds), Political Leaders and Democratic Elections.Oxford: Oxford University Press.

McCurley, C.; Mondak, J. (1995). Inspected by #1184063113: The Influence ofIncumbents’ Competence and Integrity in U.S. House Elections. AmericanJournal of Political Science 39: 864–885.

Meyrowitz, J. (1985). No Sense of Place: The Impact of Electronic Media on SocialBehavior. New York: Oxford University Press.

Midtbø, T. (1997). The Electoral Effect of Party Leader Approval in Norway.Scandinavian Political Studies 20: 135–158.

Miller, A.; Miller, W. (1976). Ideology in the 1972 Election: Myth or Reality.A Rejoinder. American Political Science Review 70: 832–849.

Miller, A.; Wattenberg, M.; Malanchuk, O. (1986). Schematic Assessments ofPresidential Candidates. American Political Science Review 80: 521–540.

Miller, W. (1991). Party Identification, Realignment, and Party Voting: Back tothe Basics. American Political Science Review 85: 557–568.

Miller, W.; Shanks, M. (1996). The New American Voter. Cambridge: HarvardUniversity Press.

Mondak, J. (1993). Public Opinion and Heuristic Processing of Source Cues.Political Behavior 15: 167–192.

Mondak, J. (1995). Competence, Integrity, and the Electoral Success ofCongressional Incumbents. Journal of Politics 57: 1043–1069.

Mondak, J.; Huckfeldt, R. (2006). The Accessibility and Utility of CandidateCharacter in Electoral Decision Making. Electoral Studies 25(1): 20–34.

Mughan, A. (1993). Party Leaders and Presidentialization in the 1992 Elec-tion: A Post-War Perspective. In D. Denver, P. Norris, D. Broughtonand C. Rallings (eds), British Elections and Parties Yearbook 1993. HemelHempstead: Harvester Wheatsheaf.

Mughan, A. (2000). Media and the Presidentialization of Parliamentary Elections.London: Palgrave Macmillan.

Mughan, A. (2005). On the Conditionality of Leader Effects. Paper Presented atthe EPOP Annual Meeting, Univeristy of Essex, 9–11 September.

Mughan, A. (2012). Parties, Conditionality and Leader Effects in Parliamen-tary Elections. Party Politics, article in press.

Ohr, D. (2011). Changing Patterns in Political Communication. In K. Aarts,A. Blais and H. Schmitt (eds), Political Leaders and Democratic Elections.Oxford: Oxford University Press.

Ohr, D.; Oscarsson, H. (2011). Leader Traits, Leader Image, and Vote Choice.In K. Aarts, A. Blais and H. Schmitt (eds), Political Leaders and DemocraticElections. Oxford: Oxford University Press.

Oskarson, M. (2005). Social Structure and Party Choice. In J. Thomassen (ed.),The European Voter. Oxford: Oxford University Press.


PROOFReferences 133

Owen, D. (1991). Media Messages in American Presidential Elections. New York:Greenwood Press.

Page, B. (1978). Choices and Echoes in Presidential Elections. Chicago: Universityof Chicago Press.

Page, B.; Jones, C. (1979). Reciprocal Effects of Policy Preferences, PartyLoyalties and the Vote. American Political Science Review 73: 1071–1090.

Pancer, S.; Brown, S.; Barr, C. (1999). Forming Impressions of Political Leaders:A Cross-National Comparison. Political Psychology 20: 345–368.

Pappi, F. (1996). Political Behaviour: Reasoning Voters and Multiparty Sys-tems. In R. Goodin and H. Klingemann (eds), A New Handbook of PoliticalScience. Oxford: Oxford University Press.

Pappi, F.; Mnich, P. (1992). Germany. In M. Franklin, T. Mackie and H. Valen(eds), Electoral Change: Responses to Evolving Social and Attitudinal Structuresin Western Countries. Cambridge: Cambridge University Press.

Pappi, F.; Shikano, S. (2001). Personalisierung der Politik inMehrparteiensystemen am Beispiel deutscher Bundestagswahlen seit1980. Politische Vierteljahresschrift 42: 355–387.

Pellikaan, H.; de Lange, S.; van der Meer, T. (2007). Fortuyn’s Legacy:Party System Change in the Netherlands. Comparative European Politics 5:282–302.

Pierce, P. (1993). Political Sophistication and the Use of Candidate Traits inCandidate Evaluation. Political Psychology 14: 21–36.

Poguntke, T. (2005). A Presidentializing Party State? The Federal Republic ofGermany. In T. Poguntke and P. Webb (eds), The Presidentialization of Politics.A Comparative Study of Modern Democracies. Oxford: Oxford University Press.

Poguntke, T.; Webb, P. (2012). The Presidentialisation of Politics ThesisDefended. Parliamentary Affairs (2013), vol. 66, 646–654.

Poguntke, T.; Webb, P. (eds) (2005). The Presidentialization of Politics. A Com-parative Study of Modern Democracies. Oxford: Oxford University Press.

Rahat, G.; Sheafer, T. (2007). The Personalization(s) of Politics: Israel, 1949–2003. Political Communication 24: 65–80.

Rahn, W.; Aldrich, J.; Borgida, E.; Sullivan, D. (1990). A Social-cognitive Modelof Candidate Appraisal. In J. Ferejohn and J. Kuklinski (eds), Information andDemocratic Processes. Urbana and Chicago: University of Illinois Press.

Richardson, B. (1991). European Party Loyalties Revisited. American PoliticalScience Review 85: 751–775.

Saalfeld, T. (2000). Germany: Stable Parties, Chancellor Democracy, and theArt of Informal Settlement. In W. Müller and K. Strøm (eds), CoalitionGovernments in Western Europe. Oxford: Oxford University Press.

Schattschneider, E. (1942). Party Government. New York: Holt, Rinehart,Winston.

Schmitt, H. (2009). Partisanship in Nine Western Democracies: Causes andConsequences. In J. Bartle and P. Bellucci (eds), Political Parties andPartisanship. Social Identity and Individual Attitudes. London and New York:Routledge.


PROOF134 References

Schmitt, H.; Holmberg, S. (1995). Political Parties in Decline? InH. Klingemann and D. Fuchs (eds), Citizens and the State. Oxford: OxfordUniversity Press.

Schoen, H. (2007). Campaigns, Candidate Evaluations, and Vote Choice:Evidence from German Federal Election Campaigns, 1980–2002. ElectoralStudies 26: 324–337.

Shively, P. (1979). The Development of Party Identification among Adults:Exploration of a Functional Model. American Political Science Review 73:1039–1054.

Simons, J. (2000). Ideology, Imagology, and Critical Thought: TheImpoverishment of Politics. Journal of Political Ideologies 5(1): 81–103.

Sniderman, P.; Brody, R.; Tetlock, P. (1991). Reasoning and Choice. Explorationsin Social Psychology. New York: Cambridge University Press.

Sovey, A.; Green, D. (2010). Instrumental Variables Estimation in PoliticalScience: A Readers’ Guide. American Journal of Political Science 55: 188–200.

Stewart, M.; Clarke, H. (1992). The (un)Importance of Party Leaders: LeaderImages and Party Choice in the 1987 British Election. Journal of Politics 54:447–470.

Stokes, D. (1963). Spatial Models of Party Competition. American PoliticalScience Review 57: 368–377.

Stokes, D. (1992). Valence Politics. In D. Kavanagh (ed.), Electoral Politics.Oxford: Clarendon Press.

Sullivan, J., Aldrich, J., Borgida, E., Rahn, W. (1990). Candidate Appraisal andHuman Nature: Man and Superman in the 1984 Election. Political Psychology11: 459–484.

Swanson, D.; Mancini, P. (eds) (1996). Politics, Media, and Modern Democ-racy: An International Study of Innovations in Electoral Campaigning and TheirConsequences. Westport, CT: Praeger.

Thomassen, J. (1976). Party Identification as a Cross-National Concept: ItsMeaning in the Netherlands. In I. Budge, I. Crewe and D. Farlie (eds), PartyIdentification and Beyond. Hoboken, NJ: John Wiley and Sons.

Thomassen, J. (ed.) (2005). The European Voter. Oxford: Oxford UniversityPress.

Thomassen, J.; Rosema, M. (2009). Party Identification Revisited. In J. Bartleand P. Bellucci (eds), Political Parties and Partisanship. Social Identity andIndividual Attitudes. London and New York: Routledge.

Thompson, J. (1995). The Media and Modernity: A Social Theory of the Media.Stanford: Stanford University Press.

Tverdova, Y. (2011). Follow the Party or Follow the Leader? Candidate Evalu-ations, Party Evaluations and Macropolitical Context. In R. Dalton and C.Anderson (eds), Citizens, Context, and Choice. How Context Shapes Citizens’Electoral Choices. Oxford: Oxford University Press.

van der Brug, W.; Franklin, M.; Toka, G. (2008). One Electorate or Many?Differences in Party Preference Formation between New and EstablishedEuropean Democracies. Electoral Studies 27: 589–600.

van der Brug, W.; Mughan, A. (2007). Charisma, Leader Effects and Supportfor Right-wing Populist Parties. Party Politics 13: 29–51.


PROOFReferences 135

van der Brug, W.; van der Eijk, C.; Franklin, M. (2007). The Economy andthe Vote: Economic Conditions and Elections in Fifteen Countries. New York:Cambridge University Press.

van der Eijk, C. (2002). Design Issues in Electoral Research: Taking Care of(Core) Business. Electoral Studies 21: 189–206.

van der Eijk, C.; Franklin, M. (1996). Choosing Europe? The European Electorateand National Politics in the Face of Union. Ann Arbor: University of MichiganPress.

van der Eijk, C.; Franklin, M. (2009). Elections and Voters. Basingstoke: PalgraveMacmillan.

van der Eijk, C.; van der Brug, W.; Kroh, M.; Franklin, M. (2006). Rethink-ing the Dependent Variable in Voting Behavior: On the Measurement andAnalysis of Electoral Utilities. Electoral Studies 25: 424–447.

van Holsteyn, J.; Andeweg, R. (2010). Demoted Leaders and Exiled Candidates:Disentangling Party and Person in the Voter’s Mind. Electoral Studies 29:628–635.

van Holsteyn, J.; Irwin, G. (2003). Never a Dull Moment: Pim Fortuyn and theDutch Parliamentary Election of 2002. West European Politics 26: 41–66.

van Holsteyn, J.; Irwin, G.; den Ridder J. (2003). In the Eye of the Beholder:The Perception of the List Pim Fortuyn and the Parliamentary Elections ofMay 2002. Acta Politica 38: 69–87.

van Wijnen, P. (2000). Candidates and Voting Behavior. Acta Politica 35:430–458.

Wagner, A.; Weßels, B. (2012). Parties and their Leaders. Does It MatterHow They Match? The German General Elections of 2009 in Comparison.Electoral Studies 31: 77–82.

Wattenberg, M. (1991). The Rise of Candidate-Centered Politics. Cambridge, MA:Harvard University Press.

Webb, P. (2004). The Reaction of British Parties to the Erosion of Voter Loy-alties. In P. Mair, W. Muller and F. Plasser (eds), Political Parties and ElectoralChange: Party Responses to Electoral Markets. London: Sage Publications.

Wlezien, C.; Franklin, M.; Twiggs, D. (1997). Economic Perceptions and VoteChoice: Disentangling the Endogeneity. Political Behavior 19(1): 7–17.

Wood, D.; Park, S. (2004). Instrumental Variable. In M. Lewis-Beck, A. Brymanand T. Liao (eds), The Sage Encyclopedia of Social Science Research Methods.Thousand Oaks, CA: Sage Publications.

Personalization of Politics and Electoral Change

Documents

Transcript of Personalization of Politics and Electoral Change