French Presidential Elections (French Politics, Society and Culture)

French Politics, Society and Culture Series General Editor: Jocelyn Evans, Professor of Politics, University of Salford, UK. France has always fascinated outside observers. Now, the country is undergoing a period of profound transformation. France is faced with a rapidly changing international and European environment and it is having to rethink some of its most basic social, political and economic orthodoxies. As elsewhere, there is pressure to conform. And yet, while France is responding in ways that are no doubt familiar to people in other European countries, it is also managing to maintain elements of its long-standing distinctiveness. Overall, it remains a place that is not exactly comme les autres. This new series examines all aspects of French politics, society and culture. In so doing it focuses on the changing nature of the French system as well as the established patterns of political, social and cultural life. Contributors to the series are encouraged to present new and innovative arguments so that the informed reader can learn and understand more about one of the most beguiling and compelling of all European countries. Titles include: Gill Allwood and Khursheed Wadia GENDER AND POLICY IN FRANCE Sylvain Brouard, Andrew M. Appleton, and Amy G. Mazur (editors) THE FRENCH FIFTH REPUBLIC AT FIFTY Beyond Stereotypes June Burnham POLITICIANS, BUREAUCRATS AND LEADERSHIP IN ORGANIZATIONS Lessons from Regional Planning in France Tony Chafer and Emmanuel Godin (editors) THE END OF THE FRENCH EXCEPTION? Decline and Revival of the ‘French Model’ Jean K. Chalaby THE DE GAULLE PRESIDENCY AND THE MEDIA Statism and Public Communications Pepper D. Culpepper, Bruno Palier, and Peter A. Hall (editors) CHANGING FRANCE The Politics that Markets Make Gordon D. Cumming FRENCH NGOs IN THE GLOBAL ERA France’s International Development Role David Drake FRENCH INTELLECTUALS AND POLITICS FROM THE DREYFUS AFFAIR TO THE OCCUPATION David Drake INTELLECTUALS AND POLITICS IN POST-WAR FRANCE John Gaffney POLITICAL LEADERSHIP IN FRANCE From Charles de Gaulle to Nicolas Sarkozy Graeme Hayes ENVIRONMENTAL PROTEST AND THE STATE IN FRANCE

9780230_300088_01_prexx.indd i

7/7/2011 6:51:40 PM

David J. Howarth THE FRENCH ROAD TO EUROPEAN MONETARY UNION Andrew Knapp PARTIES AND THE PARTY SYSTEM IN FRANCE A Disconnected Democracy? Michael S. Lewis-Beck (editor) THE FRENCH VOTER Before and After the 2002 Elections Michael S. Lewis-Beck, Richard Nadeau, and Éric Bélanger FRENCH PRESIDENTIAL ELECTIONS John Loughlin SUBNATIONAL GOVERNMENT The French Experience Mairi Maclean and Joseph Szarka FRANCE ON THE WORLD STAGE Mairi Maclean, Charles Harvey and Jon Press BUSINESS ELITES AND CORPORATE GOVERNANCE IN FRANCE AND THE UK Susan Milner and Nick Parsons (editors) REINVENTING FRANCE State and Society in the Twenty-First Century Rainbow Murray PARTIES, GENDER QUOTAS AND CANDIDATE SELECTION IN FRANCE Gino G. Raymond THE FRENCH COMMUNIST PARTY DURING THE FIFTH REPUBLIC A Crisis of Leadership and Ideology Paul Smith THE SENATE OF THE FIFTH FRENCH REPUBLIC Francesca Vassallo FRANCE, SOCIAL CAPITAL AND POLITICAL ACTIVISM Sarah Waters SOCIAL MOVEMENTS IN FRANCE Towards a New Citizenship Reuben Y. Wong THE EUROPEANIZATION OF FRENCH FOREIGN POLICY France and the EU in East Asia French Politics, Society and Culture Series Standing Order ISBN 978–0–333–80440–7 hardcover Series Standing Order ISBN 978–0–333–80441–4 paperback (outside North America only) You can receive future titles in this series as they are published by placing a standing order. Please contact your bookseller or, in case of difficulty, write to us at the address below with your name and address, the title of the series and the ISBNs quoted above. Customer Services Department, Macmillan Distribution Ltd, Houndmills, Basingstoke, Hampshire RG21 6XS, England

9780230_300088_01_prexx.indd ii

7/7/2011 6:51:40 PM

French Presidential Elections Michael S. Lewis-Beck F. Wendell Miller Distinguished Professor of Political Science, University of Iowa, USA

Richard Nadeau Professor of Political Science, University of Montreal, Canada

Éric Bélanger Associate Professor of Political Science, McGill University, Canada

9780230_300088_01_prexx.indd iii

7/7/2011 6:51:40 PM

© Michael S. Lewis-Beck, Richard Nadeau and Éric Bélanger 2011 Foreword © Harold D. Clarke 2011 All rights reserved. No reproduction, copy or transmission of this publication may be made without written permission. No portion of this publication may be reproduced, copied or transmitted save with written permission or in accordance with the provisions of the Copyright, Designs and Patents Act 1988, or under the terms of any licence permitting 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 publication may be liable to criminal prosecution and civil claims for damages. The authors have asserted their rights to be identified as the authors of this work in accordance with the Copyright, Designs and Patents Act 1988. First published 2011 by PALGRAVE 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 companies and 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 978–0–230–30008–8 hardback This book is printed on paper suitable for recycling and made from fully managed and sustained forest sources. Logging, pulping and manufacturing processes are expected to conform to the environmental regulations of the country of origin. A catalogue record for this book is available from the British Library. Library of Congress Cataloging-in-Publication Data Lewis-Beck, Michael S. French presidential elections / Michael S. Lewis-Beck, Richard Nadeau, Éric Bélanger. p. cm. Includes bibliographical references and index. ISBN 978–0–230–30008–8 (alk. paper) 1. Presidents – France – Election. I. Nadeau, Richard, 1956– II. Bélanger, Éric. III. Title. JN2959.L48 2011 324.944—dc22

2011016930

10 9 8 7 6 5 4 3 2 1 20 19 18 17 16 15 14 13 12 11 Printed and bound in Great Britain by CPI Antony Rowe, Chippenham and Eastbourne

9780230_300088_01_prexx.indd iv

7/7/2011 6:51:40 PM

To Philip E. Converse and Maurice Duverger

9780230_300088_01_prexx.indd v

7/7/2011 6:51:40 PM

9780230_300088_01_prexx.indd vi

7/7/2011 6:51:40 PM

Contents List of Illustrations

x

Foreword

xiv

Preface

xviii

1 Introduction 1.1 1.2 1.3 1.4

1

French party competition French electoral institutions and the presidency Theories of French elections Our theoretical approach

2 Social Structure

2 5 9 12 15

2.1 An analytical method of explaining vote choice in France 2.2 Socio-demographic determinants of vote choice in France 2.3 A bivariate analysis of socio-demographic determinants of the presidential vote in France 2.4 A multivariate analysis of socio-demographic determinants of the presidential vote in France 2.5 Conclusion 3 Class and Patrimony

16 17 20 31 38 41

3.1 Occupation, economic sector, and the vote 3.2 Occupation, economic sector, and the vote: contemporary bivariate results 3.3 Income, patrimony, and the vote 3.4 A multivariate analysis of the impact of socioeconomic variables on voting: cumulative results 3.5 A multivariate analysis of the impact of socioeconomic variables on voting: detailed results 3.6 Conclusion

44 48 53 55 62 66

vii

9780230_300088_01_prexx.indd vii

7/7/2011 6:51:40 PM

viii

Contents

4 Left-Right Ideology 4.1 The meaning of left-right ideology 4.2 Ideological identification vs. party identification 4.3 Ideology and the vote: bivariate results 4.4 Ideology and the vote: multivariate results 4.5 Conclusion 5 The Role of Issues

68 69 73 75 81 85 87

5.1 5.2 5.3 5.4 5.5

The meaning of issue voting 88 Economics and the French voter 95 Economics and the French voter: bivariate relationships 98 Immigration and the French voter 100 Immigration and the French voter: bivariate relationships 102 5.6 Economics and immigration: multivariate relationships 104 5.7 Adding Europe to the equation 107 5.8 Conclusion 111 6 Leader Image 6.1 6.2 6.3 6.4

Leader image in French electoral studies Components of leader image Candidate image in French presidential elections General and specific evaluations of presidential candidates 6.5 The impact of candidate image on French presidential voting 6.6 Conclusion 7 Campaign Influence 7.1 7.2 7.3 7.4

Campaign attentiveness Time of vote decision The impact of television Conclusion

8 Conclusion 8.1 First round versus second round: the institutional context

9780230_300088_01_prexx.indd viii

113 114 116 120 125 135 139 141 144 151 155 160 163 164

7/7/2011 6:51:40 PM

Contents ix

8.2 Change and continuity in French presidential elections 8.3 Is the French electoral system “stalled”? 8.4 Short-term vs. long-term factors 8.5 Is a theory of French presidential elections possible? 8.6 The French voter in perspective

165 169 170 173 174

Appendix

177

Notes

213

Bibliography

215

Index

227

9780230_300088_01_prexx.indd ix

7/7/2011 6:51:41 PM

List of Illustrations

Tables 1.1 1.2 2.1

2.2

2.3

2.4

2.5

2.6

3.1 3.2

3.3 3.4

French political parties, first-round vote share, presidential elections 1988–2007 French presidential election results, second ballot, Fifth Republic, 1965–2007 Percentage support according to religion in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007) Percentage support according to gender in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007) Percentage support according to level of education in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007) Percentage support according to age in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007) Mean support score for each category of voters for four French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007) Change in vote probability in French presidential elections, going from the minimum to the maximum category of each of the socio-demographic variables Occupation and vote in French presidential elections (1988, 1995, 2002, and 2007, first and second rounds) Employment sector and vote in French presidential elections (1988, 1995, 2002, and 2007, first and second rounds) The distribution of patrimonial assets in France, 2002 Patrimony and vote in French presidential elections (1988, 1995, 2002, and 2007, first and second rounds)

3 6

22

24

25

26

27

33 49

52 55 56

x

9780230_300088_01_prexx.indd x

7/7/2011 6:51:41 PM

List of Illustrations xi

3.5 The variation explained by the different models (McFadden pseudo-R-squared) 3.6 The signs and significance of the model coefficients 3.7 Change in vote probability in French presidential elections, from the minimum to the maximum category, for the occupation, sector, income, and assets variables 4.1 Bivariate relationship between ideology and first-round presidential vote 4.2 Bivariate relationship between ideology and second-round presidential vote 5.1 Evaluating the importance of problems before the 1995 presidential ballot, first round 5.2 The ranking of issues in presidential election campaigns (1995, 2002, 2007) 5.3 The relationship between economic perception and the 1995 presidential vote, first round (%) 5.4 Relationship between attitude toward immigrants and 2007 presidential vote, first round (%) 5.5 Perception of the construction of Europe as an important problem 5.6 Distribution of opinion on “France in the European Union” 6.1 Thermometer scores for leaders in the 1988, 1995, 2002, and 2007 French presidential elections 6.2 Opinions on leaders in the French presidential elections of 1988, 1995, 2002, and 2007 6.3 Regression models of thermometer scores of candidates, 1988, 1995, 2002, and 2007 (ordinary least squares method) 6.4 Bivariate correlations of personality traits with thermometer measures, 1988, 1995, 2002, and 2007 6.5 Logistic regression coefficients of thermometer variables for the French presidential elections of 1988, 1995, 2002, and 2007 6.6 Effect of leader image on voting in French presidential elections of 1988, 1995, 2002, and 2007 7.1 Campaign attentiveness 7.2 Who is more attentive to campaigns?

9780230_300088_01_prexx.indd xi

58 59

63 77 82 91 92 99 103 109 110 121 123

128 135

137 138 145 146

7/7/2011 6:51:41 PM

xii List of Illustrations

7.3 Campaign attentiveness and first-round voting 149 7.4 Time of vote decision during the first round 152 7.5 Time of vote decision and first-round voting 153 7.6 Change in first-round candidates’ image after having seen them on television 156 7.7 Relationship between change in a candidate’s image and vote choice for that candidate in the first round 158 7.8 Televised debate and second-round voting 160 A.1 Reported vote for the extreme left (Lajoinie, Juquin, or Laguiller), French presidential election, 1988 178 A.2 Reported vote for the left (Mitterrand), French presidential election, 1988 180 A.3 Reported vote for the center (Barre), French presidential election, 1988 181 A.4 Reported vote for the right (Chirac), French presidential election, 1988 182 A.5 Reported vote for the extreme right (Le Pen), French presidential election, 1988 184 A.6 Reported vote for the right (Chirac), second round of the French presidential election, 1988 185 A.7 Reported vote for the extreme left (Laguiller or Hue), French presidential election, 1995 186 A.8 Reported vote for the left (Jospin), French presidential election, 1995 188 A.9 Reported vote for the center (Balladur), French presidential election, 1995 189 A.10 Reported vote for the right (Chirac), French presidential election, 1995 190 A.11 Reported vote for the extreme right (Le Pen), French presidential election, 1995 192 A.12 Reported vote for the right (Chirac), second round of the French presidential election, 1995 193 A.13 Reported vote for the extreme left (Laguiller, Hue, or Besancenot), French presidential election, 2002 194 A.14 Reported vote for the left (Jospin), French presidential election, 2002 196 A.15 Reported vote for the center (Bayrou), French presidential election, 2002 197

9780230_300088_01_prexx.indd xii

7/7/2011 6:51:41 PM

List of Illustrations xiii

A.16 Reported vote for the right (Chirac), French presidential election, 2002 A.17 Reported vote for the extreme right (Le Pen), French presidential election, 2002 A.18 Voting intention for the right (Chirac), second round of the French presidential election, 2002 A.19 Reported vote for the extreme left (Laguiller, Buffet, or Besancenot), French presidential election, 2007 A.20 Reported vote for the left (Royal), French presidential election, 2007 A.21 Reported vote for the center (Bayrou), French presidential election, 2007 A.22 Reported vote for the right (Sarkozy), French presidential election, 2007 A.23 Reported vote for the extreme right (Le Pen), French presidential election, 2007 A.24 Reported vote for the right (Sarkozy), second round of the French presidential election, 2007

198 200 201 202 204 205 206 208 209

Figure 4.1

Left-right ideology distribution in the French electorate, 1988–2007

9780230_300088_01_prexx.indd xiii

71

7/7/2011 6:51:41 PM

Foreword French Presidential Elections makes an important contribution to research on voting and elections in mature democracies. Scientific studies of electoral choice in such polities were massively invigorated some six decades ago by the trailblazing inquiries conducted by Angus Campbell and his colleagues at the University of Michigan (e.g., Campbell et al., 1960, 1966). Since that time, the “Michigan” election studies have made foundational theoretical and methodological contributions to our understanding of factors affecting voting behavior and election outcomes. In French Presidential Elections, Lewis-Beck, Nadeau, and Bélanger employ these theoretical and methodological tools with an admirable combination of skill and insight to explain voting behavior in a major Western democracy. The theoretical approach adopted by the authors has its roots in what may be called “the puzzle of ’52.” As readers of “a certain age” will recall, this puzzle involved the inability of the once-dominant sociological model of voting behavior, which emphasized the stabilizing effects of ponderous social forces, to provide a satisfactory account of the landslide victory by Dwight Eisenhower, the candidate of the seemingly perennial minority Republican Party, in the 1952 (and 1956) US presidential elections. The sociological model could explain stability – but not change. Large-scale swings in parties’ vote shares across the relatively brief four-year periods separating successive American presidential elections were beyond the model’s theoretical reach. The Michigan model provided an attractive solution to the puzzle. By translating social forces into social psychology via the concept of party identification, which normally demonstrated remarkable durability, the model allowed for long-term factors to influence the vote. Equally important, by explicitly recognizing the significance of possibly highly mutable candidate and issue orientations that operated much nearer to the tip of the famed “funnel of causality,” the model could explain short-term and medium-term political change. The solution to the puzzle of ‘52 thus involved recognition that the

xiv

9780230_300088_01_prexx.indd xiv

7/7/2011 6:51:41 PM

Foreword

xv

Republican victory was a consequence of a complex of issue concerns and candidate images quite different from those at work in the New and Fair Deal eras of the 1930s and 1940s. Lewis-Beck, Nadeau, and Bélanger’s adaptation of the Michigan model to the French case is carefully informed by controversies springing from previous electoral research in France and elsewhere. Perhaps most noteworthy is how they handle the long-standing debate regarding the relative importance of party identification and left-right ideological orientation as long-term influences on the vote. Estimating a model that allows for simultaneous causal linkages between partisanship and ideology leads them to conclude that ideology dominates (but does not obliterate) partisanship as the long-term influence on voting behavior in France. This methodology acknowledges the limitations of existing French survey data while avoiding an unfortunate tendency among some political scientists to try to settle data-impoverished theoretical debates either by fiat or by “hand-waving” commentary about endogeneity problems in models they oppose. Following in the Michigan tradition, the authors of French Presidential Elections also pay close attention to short-term influences on the vote. Particularly noteworthy is the finding that the images of presidential candidates regularly have sizable effects. This finding accords very well not only with American research, but with studies in countries such as Canada and Great Britain, where multivariate statistical analyses have repeatedly demonstrated that public reactions to competing party leaders are highly significant for understanding electoral choice (e.g., Clarke, 2004, 2009a, 2009b). Lewis-Beck, Nadeau, and Bélanger rightly note that their finding of strong candidate image hearkens back to very early research by Converse and Dupeux (1966) on voters’ images of Eisenhower and de Gaulle. In French Presidential Elections we learn that presidential candidates need not be such towering historical figures to exert consequential impact on how the electorate behaves. There is more – much more – for the reader to appreciate in French Presidential Elections. Due attention is paid to important perennial and contemporary issues (such as the economy and immigration), election campaigns, and various sociological variables that have tried to lay claim to explanatory pride of place. And, as befits a work

9780230_300088_01_prexx.indd xv

7/7/2011 6:51:41 PM

xvi Foreword

with strong intellectual roots in Ann Arbor, the analyses are informed by the idea of a structure of causality flowing from society and economy to political beliefs, attitudes, and opinions, to voting behavior, and, ultimately, to election outcomes. At every stage in their analyses, the authors assess the power of competing explanations of the vote via appropriate multivariate statistical methods. Here, they demand that the claims of rival theoretical models must be adjudicated by consistently strong in-sample statistical performance across data sets gathered in successive national elections. Models that do not perform well are rightly designated as demonstrably inferior to those that can. Like all good scientific works, French Presidential Elections answers important existing questions and poses new ones. As is often the case, a key to answering the latter will involve gathering new and better data. Perhaps particularly valuable will be survey data that will enable scholars to study the short-term dynamics of party and candidate support during French election campaigns and how these are affected by salient events and conditions as reported in the mass media. More generally, there is an evident need for high-quality multiwave panel data that will permit sophisticated multivariate analyses of the individual-level dynamics of key variables in the kinds of models studied in French Presidential Elections. Lewis-Beck, Nadeau, and Bélanger have given us a most useful roadmap that will guide such future inquiries. The result of their efforts is an outstanding example of the contemporary science of electoral choice. Harold D. Clarke Ashbel Smith Professor, School of Economic, Political and Policy Sciences, University of Texas at Dallas, USA

References Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes (1960) The American Voter (New York: John Wiley & Sons). Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes (eds.) (1966) Elections and the Political Order (New York: John Wiley & Sons). Clarke, Harold D., David Sanders, Marianne C. Stewart, and Paul Whiteley (2004) Political Choice in Britain (Oxford: Oxford University Press).

9780230_300088_01_prexx.indd xvi

7/7/2011 6:51:41 PM

Foreword

xvii

Clarke, Harold D., Allan Kornberg, and Thomas Scotto (2009a) Making Political Choices: Canada and the United States (Toronto: University of Toronto Press). Clarke, Harold D., David Sanders, Marianne C. Stewart, and Paul Whiteley (2009b) Performance Politics and the British Voter (Cambridge: Cambridge University Press). Converse, Philip E. and Georges Dupeux (1966) “De Gaulle and Eisenhower: The Public Image of the Victorious General” in Angus Campbell, Philip E. Converse, Warren E. Miller, and Donald E. Stokes (eds.) Elections and the Political Order (New York: John Wiley & Sons).

9780230_300088_01_prexx.indd xvii

7/7/2011 6:51:41 PM

Preface Our study of French politics grows out of intellectual curiosity and deep affection. For two of us, France is much more than simply a foreign country – it is the land of our ancestors, and as such it can be considered as a motherland in the deepest sense of the word. For the third, France is truly a pays d’adoption. In pursuit of explaining, even predicting, French elections, we have spent considerable time in the country itself. Formally, we have participated in conferences, exchanges, residencies, studying, teaching, and research in Paris and other centers of French scholarly life. Informally, we have been “on the ground” a good deal, talking to French citizens, consuming the food and wine, and traveling over the captivating countryside. (Among us, in fact, we have visited all the departments of metropolitan France). As students of comparative politics, we believe that clear understanding of a nation’s electoral ways begins with thorough fieldwork, including immersion in the language and culture. French elections have been under our microscope for a long time. All together, the three of us have published, in English and in French, over 65 works on the subject. While the book at hand draws on findings from many of these studies, it goes well beyond them. French Presidential Elections contains much useful description, and factual information, about voters and the institutions they interact with. However, it extends itself further, into the field of election theory. That is, the presidential vote choice of the French citizen is submitted to a general inquiry. Why do French presidential voters act the way they do? We explain this simple act of political behavior, in a way that is applicable across Fifth Republic contests. We see that the French vote, while subject to issues of the moment, is heavily determined by deep, enduring sociological and ideological forces. This said, to paraphrase the title of a famous book on elections in France – France de gauche, vote à droite – we have tried to demonstrate how the moderate right has been successful on many occasions by putting forward leaders perceived to have the right stuff to be president, by running efficient campaigns, and by being on the “right” side of the issues. Understanding what drives the individual French xviii

9780230_300088_01_prexx.indd xviii

7/7/2011 6:51:41 PM

Preface

xix

voter, we understand also much about the national vote. Electoral change reveals itself as difficult, but not impossible. France, in this way, looks much like other advanced democracies. The lack of French “exceptionalism” here, however, does not diminish its politics. Instead, it simply shows that French voters are citizens, de facto, in the larger family of democratic nations. There are many individuals and institutions in France, Canada, Quebec, and the United States that have contributed, directly or indirectly, to the development of this volume. We attempt to list them all, but undoubtedly will miss some. With respect to institutions, we would like to recognize the following: the United States National Science Foundation (six different grants); the Rockefeller Foundation (one conference), the German Marshall Fund (fellowship); the Camargo Foundation (two fellowships); the Fonds québécois de recherche sur la société et la culture (FQRSC, research funding); Centre d’Étude de la Vie Politique Française (CEVIPOF, as visiting scholars); Université de Paris I (as visiting faculty); Ecole d’été de Lille (as faculty); Institut d’Études Politiques de Paris (as student and faculty); Université de Paris II (as visiting faculty); Université de Nantes (as visiting scholar). With respect to individual scholars of French politics, we would like to recognize the following: Andrew Appleton, Frank Baumgartner, Daniel Boy, Bruno Cautrès, Jean Chiche, Bernard Dolez, Robert Elgie, Christine Fauvelle-Aymar, Martial Foucault, Jacques Gerstlé, Bernard Grofman, Gérard Grunberg, Bruno Jérôme, Véronique JérômeSpeziari, Alain Lancelot, Jean-Dominique Lafay, Annie Laurent, Bertrand Lemennicier, Philippe Maarek, Nonna Mayer, Amy Mazur, Rainbow Murray, Pascal Perrineau, Roy Pierce, William Safran, Martin Schain, Alan Spitzer, Vincent Tiberj, and Marie-France Toinet. Besides general contributors, there are those who made essential contributions to the physical production of the volume itself. Here we would like to acknowledge the fine work of Thomas Didier, Angelo Elias, and Chris Chhim. Finally, we want to thank the very helpful people at Palgrave Macmillan, especially our editor, Liz Blackmore, and our series editor, Jocelyn Evans. Last, but not least, we wish to extend a special thanks to our families, who have patiently supported us through endless (so it seemed) work sessions away from them. We could not have done it without their love and support.

9780230_300088_01_prexx.indd xix

7/7/2011 6:51:41 PM

9780230_300088_01_prexx.indd xx

7/7/2011 6:51:41 PM

1 Introduction

French elections appear to be a labyrinth of parties, institutions, behaviors. But France’s presidential contests serve as a guide through that labyrinth. Over the last 30 years, France has known three presidents of the Republic: Socialist François Mitterrand (1981–1995), Gaullist Jacques Chirac (1995–2007), and the current leader from the right, Nicolas Sarkozy (2007–). Having only three national leaders across this length of time gives the impression that the country has had great continuity of government, drawing on a narrow, traditional political class to replenish itself. In some ways, this impression of political inbreeding rings true. (Take the example of Chirac. He was prime minister for Mitterrand in 1986, and ran against him in 1988 and against Jospin in 1995 and in 2002.) However, this apparent continuity mask the complexity of French electoral politics. When we look at the multitude of parties involved over this period, together with the institutional changes, the notion of seamless power transitions vanishes. By way of introduction, we sketch the contemporary changes in French parties and institutions. Against this backdrop, the behavior of the French voter fits. We explain that fit through an analysis of the dominant, ordering election in the system – presidential selection. We propose a theory of the presidential vote in France, and test that theory across the four most recent contests – 1988, 1995, 2002, and 2007. What this allows us to do is focus, as never before, on continuities in French electoral choice at the mass level. Furthermore, it is in the nature of this approach that deviations from these continuities can be highlighted. Thus, we can carefully examine the 1

9780230_300088_02_ch01.indd 1

6/28/2011 8:22:00 PM

2

French Presidential Elections

stability of the French system, while at the same time documenting changes in it. This ability to examine the dynamics of French presidential elections is extremely important because of what has happened to the parties and institutions surrounding the election of the president. Below, we analyze this shifting background, examining in turn French party competition, presidential electoral institutions, and election theories. Finally, we offer our own theoretical approach to the study of presidential elections in France, and describe how that approach unfolds in this book.

1.1

French party competition

The French multi-party system has rightly earned its reputation for rowdiness. As one distinguished observer, Jacques Fauvet, commented some time ago: “France contains two fundamental temperaments – that of the left and that of the right … three principal tendencies, if one adds the center; six spiritual families; ten parties, large or small, traversed by multiple currents: fourteen parliamentary groups without much discipline; and forty million opinions” (see Ehrmann, 1983, p. 211). These remarks from the past resonate in the present, with only slight modification, as France experiences the sixth decade of the Fifth Republic. Under its constitution, executive authority was intended to flow from the president. One reason for this presidential strength is the force of the personalities occupying the office. De Gaulle stands out, but he is not alone. In addition, the rules of governance were designed to make the office strong, even stronger after the 1962 referendum on direct election by popular majority vote. The power of the presidential office, some would contend, has made for a less complex and more disciplined system of parties, a system accountable to the dictates of government. In Table 1.1, we observe the array of leading French political parties that have contested the four most recent elections. We must emphasize that this table displays a necessary simplification of the French system. In fact, there are many political parties, groups, or tendances in France, and they may not resemble the reader’s idea of a typical political party. Take, for example, the 1997 National Assembly contests, where no fewer than 48 “partisan groups” registered candidates (LewisBeck, 2000, p. 4). Some of these groups stood for a single issue, a

9780230_300088_02_ch01.indd 2

6/28/2011 8:22:00 PM

Introduction 3

Table 1.1 French political parties, first-round vote share, presidential elections 1988–2007

Extreme Left Communists Socialists/Left Radicals Ecologist/Green UDF/MODEM RPR/UMP National Front

1988

1995

2002

2007

4.5 6.8 34.1 3.8 16.5 19.9 14.0

5.3 8.6 23.3 3.3 18.6 20.8 15.0

10.0 3.4 18.5 5.2 6.8 19.9 16.9

5.4 1.9 25.9 1.6 18.6 31.2 10.4

Notes: Entries are percentages of votes cast (suffrages exprimés). Columns do not add up to 100% because not all candidates were included for purposes of this table. The extreme left candidates are Boussel, Laguiller, and Juquin in 1988; Laguiller in 1995; and Laguiller and Besançenot in 2002 and 2007. Communist candidates are Lajoinie in 1988; Hue in 1995 and 2002; and Buffet in 2007. Socialist/left radical candidates are Mitterr and in 1988; Jospin in 1995; Jospin and Taubira in 2002; and Royal in 2007. Ecologist/Green candidates are Waecher in 1988, Voynet in 1995, Mamère in 2002, and Voynet in 2007. UDF/MODEM candidates are Barre in 1988, Balladur in 1995, and Bayrou in 2002 and 2007. RPR/UMP candidates are Chirac from 1988 to 2002 and Sarkozy in 2007. The National Front’s candidate is Le Pen for all elections. Source: Conseil constitutionnel français (1988, 1995, 2002, 2007).

splintered faction, an individual, or a generic type. Nevertheless, it makes for a crowded political playing field. At the presidential level the party playing field is simpler, but not by as much as might be expected. Here we examine only the major party groupings. There are still seven of them, but they do manage to account for over 90 percent of the votes cast. They are listed on a continuum from left to right, as Fauvet suggested above, accompanied by their percentage share of the first-round presidential vote. The parties are several, and not frozen in time. The biggest party of the Fourth Republic, the Christian Democratic Mouvement Républicain Populaire (MRP), no longer exists. The Gaullists, with various party label incarnations, most notably the Rassemblement pour la République (RPR), were a strong presence until 2002, under the direction of their founder, Chirac. Since then, the traditional right has tried to represent itself as more of a catch-all party, in the guise of the Union pour un Mouvement Populaire (UMP), formed during Chirac’s last term and nurtured by his successor, Sarkozy (who still occasionally draws on the Gaullist symbol for political

9780230_300088_02_ch01.indd 3

6/28/2011 8:22:00 PM

4

French Presidential Elections

purposes). The traditional left continues to express itself through the long-standing Socialist party. Some analysts, in considering these leading traditional partisan forces, impose a view of France as a twoparty coalition system, with Government facing Opposition. One factor that would sustain this model would be rotation in office, from one coalition to the other. This test implies a successful switch from right to left, as happened with the Popular Front of 1936. Such alternance actually happened for the Fifth Republic when Mitterrand wrested the presidency from Giscard d’Estaing in 1981. After that, the Socialists regained the presidency with the victory of Mitterrand in 1988, establishing that the presence of a left executive in the Fifth Republic was not to be a unique event. What is more, the 1988 contest saw the first strong showing of the National Front in presidential elections, in the person of Jean-Marie Le Pen. Contrasting with this powerful surge on the extreme right, the Communists began their long-term electoral decline. Today the Communist party still exists, although in a much reduced, even old-fashioned, or passé, form. This does not mean that the extreme left lacks vitality. On the contrary, some splinter parties are capable of capturing a reasonable proportion of the vote, as did the Trotskyite Workers’ Struggle (Lutte Ouvrière, LO) in 1995. Indeed, the 1995 presidential contest provides a vivid example of the many varieties of meaningful partisan struggle in French presidential elections. Here are the first-ballot returns for the 1995 presidential vote in some detail. Altogether, eight candidates received 3 percent or more of the total vote: Jospin (Socialist), 23.3 percent; Chirac (Gaullist), 20.8 percent; Balladur (Gaullist), 18.6 percent; Le Pen (FN), 15.0 percent; Hue (Communist), 8.6 percent; Laguiller (Workers’ Struggle), 5.3 percent; De Villiers (Mouvement pour la France), 4.7 percent; and Voynet (Greens), 3.3 percent. Clearly, the French presidential voter has considerable choice, in terms of ideological and policy options. This is perhaps one reason why voter turnout in these contests has averaged over 80 percent. (We return to the implications of these turnout numbers below.) On the extreme right, the National Front enjoyed a surge from the late 1980s. Its electoral weight is clearly seen in the vote totals for Jean-Marie Le Pen, which reached a peak in 2002 when he actually got through to the second, decisive vote round. The seemingly impossible occurred. In the first round, he beat out Socialist Jospin

9780230_300088_02_ch01.indd 4

6/28/2011 8:22:00 PM

Introduction 5

for second place by a whisker (16.9 percent to 16.2 percent). No one, politician or pundit or proletarian, had foreseen this outcome. In response, leaders of the mainstream left and right rallied together, calling for a triumph of traditional republican virtues against the xenophobic National Front champion. President Chirac roundly thrashed Le Pen at the second round with 82.2 percent of the vote. But the foundations of the democracy were shaken. The fortunes of the center have fluctuated, depending partly on the presence of an appealing candidate, such as Bayrou in 2007. He was offered an unusual opportunity in that contest because, unlike in the past contests (of 1988, 1995, and 2007), neither a prime minister nor a president was running. Both Socialist Royal and UMP Sarkozy were, of course, major party leaders. Nevertheless, they lacked the executive stamp of presidential candidates in the recent past. This situation made for a bracing ideological contest, waged by personalities lacking the armor of office. The absence of these traditional constraints helped to make the centre position more attractive, and Bayrou took full advantage. Indeed, at different points in the campaign, polls actually showed him the “Condorcet winner,” capable of besting all comers in a series of straight, paired fights (Abramson, 2007). How Sarkozy, in the end, took the prize is an interesting story, which we will tell in the chapters to come.

1.2

French electoral institutions and the presidency

French presidential elections, like elections in all modern democracies, are governed by rules. But France possibly has more rules, and certainly the rules change more than in most other democracies. The Fifth Republic has a written constitution. However, since the French Revolution the country has had 16 written constitutions, and more than a dozen national election laws have been used since 1870 and the start of the Third Republic. Today, the election rules for the different levels of office (president, National Assembly, Senate, European Parliament, regional councils, departmental councils, and municipal councils) vary: they use either proportional representation or a majority, and have one or two rounds, which are held at the same or different times. As a sage observer of the French scene famously declared, political institutions in France have “been treated as a weapon in the struggle between different political camps and between different

9780230_300088_02_ch01.indd 5

6/28/2011 8:22:00 PM

6

French Presidential Elections

political forces for the control of State and society” (Campbell, 1965, p. 17). The contemporary set of institutions, formed under the 1958 Constitution of the Fifth Republic, establishes a hybrid presidentialparliamentary system with the dual executive of a president and a prime minister. Currently, the president acts as head of state, may declare a state of emergency, and holds office for a fixed five-year term, with the opportunity for reelection. He or she formulates general policy before the National Assembly, validates that body’s legislation, and designates the prime minister to lead the government in parliament. The president may dissolve parliament and call for new legislative elections. With Charles de Gaulle, presidential power was extensive, unchecked by the prime minister, who was, in the words of Macridis (1975, p. 28) “the President’s man.” Subsequently, only one woman has served as prime minister: Edith Cresson, under Mitterrand in 1991. Although no women have held presidential office, one was a serious presidential candidate: Ségolène Royal, the Socialist opponent to Nicolas Sarkozy in the second round of the 2007 contest. Table 1.2 shows the results from the second ballot of all the French presidential elections during the Fifth Republic. As noted, the many parties of the Fifth Republic are often largely represented within two broad ideological divisions, the left and the right. Until the Socialist alternance of 1981, the presidency of the Fifth Republic had been in the hands of the traditional right. Table 1.2 French presidential election results, second ballot, Fifth Republic, 1965–2007 Incumbent candidate 1965 1969 1974 1981 1988 1995 2002 2007

De Gaulle Pompidou Giscard d’Estaing Giscard d’Estaing Mitterrand Jospin Chirac Sarkozy

Vote (%) Opposition candidate Vote (%) 55.2 58.2 50.8 48.2 54.0 47.4 82.2 53.1

Mitterrand Poher Mitterrand Mitterrand Chirac Chirac Le Pen Royal

44.8 41.8 49.2 51.8 46.0 52.6 17.8 46.9

Note: Entries are percentages of votes cast (suffrages exprimés). Source: Conseil constitutionnel français (1965–2007).

9780230_300088_02_ch01.indd 6

6/28/2011 8:22:00 PM

Introduction 7

Executive authority itself, emanating from the presidency, was undivided until 1986, when a new legislative election was required. Once Mitterrand triumphed over Valéry Giscard d’Estaing in 1981, he went on, as president, to dissolve the National Assembly and declare elections. The legislative election rules had a two-round ballot, with a majority winning on the first and a plurality winning on the second. In 1981, these rules gave the Socialists an absolute majority of seats for the first time. However, going into the 1986 legislative election, the Socialists feared the loss of their majority; therefore, they changed the election rules to a one-round proportional representation system. Nevertheless, the Socialists lost their majority, and the traditional right, led by the Gaullist RPR, assumed control of the National Assembly (whereupon they reinstated the old election rules). One immediate consequence was the capture of a substantial number of seats by the radical right, the National Front. Another immediate consequence of this situation was a divided executive, with the president heading one party coalition and the prime minister heading a rival party coalition. For example, in 1986, Socialist president Mitterrand (on the traditional left) opposed RPR prime minister Jacques Chirac (on the traditional right). This situation of a divided executive, with one side having the president and the other side having the prime minister, the French call cohabitation, a condition which has existed for three periods during the Fifth Republic: 1986–1988 (with Mitterrand as president and Chirac as prime minister), 1993–1995 (with Mitterrand as president and Edouard Balladur, another member of the RPR, as prime minister), and 1997–2002 (with Chirac as president and Socialist Lionel Jospin as prime minister). When cohabitation was put into practice, it soon became clear that under it the prime minister was supreme in terms of directing domestic policy, because he or she controlled the majority of votes in the National Assembly. However, the president managed to maintain some authority over foreign policy, if only because of the president’s constitutional role as head of state. Before the 2002 national elections, the Socialists, who were governing in parliament, changed the order of the electoral calendar, placing the presidential election prior to the National Assembly elections. In part, this was done to end their cohabitation with the right. The Socialist reasoning projected a repeat of its 1981 success, that is, that popular Socialist PM Jospin would win the presidential election

9780230_300088_02_ch01.indd 7

6/28/2011 8:22:00 PM

8

French Presidential Elections

and this win would provide the basis for a Socialist National Assembly majority. As it turned out, the Socialists cut their vote losses, but did not manage an outright win. (For a discussion of this reordering, and its electoral consequences, see Jérôme, Jérôme, and Lewis-Beck, 2003.) Instead, the right, under Chirac’s direction, captured the presidency and the parliament. Cohabitation was at an end, but now perhaps for ever. The 2002 electoral reform of the National Assembly, which required the presidential election to occur before the legislative one, dovetailed with another piece of electoral reform – the reduction of the presidential term from seven years to five years. This réforme du quinquennat made it virtually certain that the presidential and parliamentary elections would be held in tandem, the latter quickly following the former. Under these circumstances, cohabitation would be highly unlikely, since it would mean the newly elected president and his or her majority being repudiated just a few weeks after election. This new configuration of electoral rules has consequences. As Pascal Perrineau (2001, p. 389) reminds us, “The presidential election and the legislative elections form a system.” Thus, changing one can influence the other. The above reforms have increased the “presidentialization” of national elections. Contests for the National Assembly are perceived, more and more, to be tools to further presidential ambitions. Legislative elections are taking on the character of “second-order” events, and this fall in status is reflected in greater voter abstention. For example, in 2002, the abstention rate was at a high for the Fifth Republic: 35.6 percent in the first round; 39.7 percent in the second round. Indeed, turnout for legislative elections, in general, has been adversely affected by earlier presidential manipulations. (See the discussion of these adverse effects of reform on turnout in Fauvelle-Aymar, Lewis-Beck, and Nadeau, 2011.) The electoral calendar shifts of 1962, 1981, and 1988, in addition to 2002, each contributed in its own way to the reduction in legislative voter turnout. In contrast, presidential election turnout has remained steady, and high. Focus on presidential election abstention results in the “decisive” second round: for the initial four elections (1965, 1969, 1974, and 1981) the average was 18.40 percent; for the following four elections (1988, 1995, 2002, and 2007) the average was 18.04 percent. The dramatic effect of “presidentialization” was most obvious in the second round of the 2007 elections, where the

9780230_300088_02_ch01.indd 8

6/28/2011 8:22:00 PM

Introduction 9

abstention rate in the presidential race was only 16 percent, in contrast to the abstention rate of 40 percent in the legislative race. In other words, presidential elections have fully maintained their firstorder status in the face of the plummeting importance of National Assembly elections. Clearly, now more than ever, vote choice in presidential elections merits careful study.

1.3

Theories of French elections

With the great number of parties, and electoral rules, all in a state of flux, the task of finding a general explanation for French election outcomes at times seems hopeless. But certain election patterns can be discerned and accounted for. Let us start with the macro-level, examining two approaches – one old and one new. The former refers to the venerable tradition of electoral geography; the latter refers to new work on partisan cycles. The electoral geography approach, pioneered by André Siegfried (1913), was continued by François Goguel (1951, 1970, 1974) and his followers, chief among them Alain Lancelot (1986). Its essential argument is that disparate regions of the country have their own particular voting patterns, and these patterns persist across time. Thus, for example, we might expect the leftist vote share of a region in the 1988 presidential election to be largely determined by the leftist vote share in the same region in the previous national election. This is in fact the case, for example, with the 1986 legislative vote share of the left (at the regional level of the department) correlating highly with the 1988 presidential vote share of the left in the first round (again by department), with a correlation coefficient, r, of 0.95 (Lewis-Beck, 1993, p. 10). Therefore, in an ecological sense, where you live might “explain,” or at least account for, your presidential vote. The electoral geography tradition continues to influence French political studies. In a recent example, applied to presidential elections for 1981–2002, Laurent (2004, p. 20) finds the following department-level correlations by party vote: RPR, r = 0.85; Communist, r = 0.83; Socialist, r = 0.54. (The Socialist link, as the weakest, suggests they experienced the most electoral geography change over the period.) Clearly, the geographic argument still has sway. However, we set it aside here to pursue a new and different macroapproach, that of partisan cycles. These have been little studied

9780230_300088_02_ch01.indd 9

6/28/2011 8:22:00 PM

10 French Presidential Elections

anywhere. An early US investigation was the first to look at them, in an exploration of equilibrium conditions (Stokes and Iversen, 1966). More recently, Norpoth (2002, p. 128–129) has mapped equilibrium characteristics of the Democratic vote for the US president over a long time series (1828–2000). He finds a stable pattern of fluctuation, invariably within a range of 35–65 percent. Also, the system appears to model “perfect competition,” with the Democrats rhythmically crossing above and below a 50 percent share, the winning benchmark. Overall, the Democratic voting average stands at 49.6 percent, implying that contests are generally close and competitive. In fact, rarely does a party hold the office for more than two terms. Interestingly, the results for Great Britain are not dissimilar (Lebo and Norpoth, 2007; Norpoth, 2004). Over almost 200 years, the Conservative party has never had a winning margin of more than 20 points. In addition, the elections show competition, with an average Conservative difference of only 2.2 percentage points. In practice, then, the races are often close, and 12 out of the 44 elections actually saw a change of government in parliament. No systematic work on partisan cycles has been carried out for France, with the exception of our own (Bélanger and Lewis-Beck, 2010). In part, this can be attributed to the complications of the French political system, with its many parties, movements, and tendencies. Nevertheless, this dazzling partisan array can be ordered using the left-right continuum. In particular, there is a party core on the left that can be tracked consistently through several republics. The voter support for two central parties on the left – the Communists and the Socialists – can be recorded over a relatively long time series (N = 21). (Consult here the excellent work of Knapp, 2004.) Suppose this left vote share is charted for the period 1924–2002, for firstround National Assembly elections. (See the relevant figures in Bélanger and Lewis-Beck, 2010.) There are massive fluctuations in support for the left, from a low of 26.9 percent in 1993 to a high of 52.2 percent in 1981. Moreover, these extremes are not mere outliers. Instead, we see persistent and substantial variation in the left vote across this 78-year period. Finally, unlike the British and American series, the French series is, statistically speaking, not stationary, which implies, among other things, an inherent trend against regular rotation in office. The direction of that trend favors the right. Since 1981, when the left came to power for the first time in the Fifth

9780230_300088_02_ch01.indd 10

6/28/2011 8:22:00 PM

Introduction 11

Republic, they have regained parliamentary power only twice: in 1988 and 1997. Indeed, for the 12 national legislative elections held between 1958 and 2002, the total left vote share, inclusive (Communists + Socialists + extreme left + moderate left + Greens), amounts to just 46.3 percent. This percentage does not come close to the election-winning benchmark of 50 percentage points. Hence, in contrast to the UK and US cycles, the French partisan cycle seems short on stability, competitiveness, and equilibrium. The left, then, has not had complete good fortune at the French ballot box. Furthermore, the changes in this fortune follow a rough pattern, described as a “double hump,” with something of a generational periodicity with peaks in the 1940s and 1980s. What accounts, substantively, for these systematic macro-partisan fluctuations? Put another way, what are voters doing to bring about this pattern? A proper understanding here depends on considerations of partisanship at the micro-level. In France, a debate has long existed over whether party identification or ideological identification weighs more heavily for vote choice (Converse and Pierce, 1986; Boy and Mayer, 1993; Evans, 2004; Fleury and Lewis-Beck, 1993; Lewis-Beck and Chlarson, 2002; Pierce, 1995). In this book, we will engage in that debate. For now, we simply take ideological identification as our principal marker of partisan feeling in the mass electorate. Ideological identification, at the aggregate, national level, can be measured from French public-opinion surveys from 1962 to 2002. The macro-fluctuations of left identification, when plotted, appear similar to the fluctuations of the left vote share reported above. In fact, the left identification measure (scored for election year) has a high correlation with left vote share (r = 0.64; Bélanger and LewisBeck, 2010). Thus, the macro-partisan pattern might have underlying ideological shifts in electoral opinion as a source. Of course, the inference is that individual voters change ideologically, leading finally to the aggregate outcome. However, we have not been directly examining the connections of voters to their ideology. For that examination, we need to turn to micro-approaches to electoral investigation. Regardless of perspective, these micro-investigations draw on publicopinion or election surveys, because in these individual voters are interrogated personally about the motives and factors behind their vote. The leading data source for this is the French National Election

9780230_300088_02_ch01.indd 11

6/28/2011 8:22:01 PM

12

French Presidential Elections

Surveys (FNESs). The history of these surveys is rather different from that of, for example, the US and the UK national election surveys, which have relied mainly on government research grants. In France, funding has been more often commercial than academic, and surveys have been conducted more erratically. Nevertheless, a series of FNESs are now available, beginning with the monumental 1978 legislative election study (Capdevielle et al., 1981). For presidential elections, the particular subject of this book, the first major study, for 1988, was administered, as have been subsequent FNESs, by the Centre d’Etude de la Vie Politique Française (CEVIPOF). That study had a national sample of 4,032, and took almost an hour, on average, to carry out face-to-face (Boy and Mayer, 1990). Each presidential election since then has had its own major survey. (See, respectively, Boy and Mayer, 1997, for 1995; Cautrès and Mayer, 2004, for 2002; Cautrès and Muxel, 2009, for 2007.) In these and other survey-based French election studies, different theoretical emphases have been used in the attempt to understand the determinants of individual vote choice. The emphasis may be on the role of party or ideology (introduced above) or it may be issues, social cleavages, or institutions. (For recent examples of each of these perspectives in English see, respectively, Evans, 2004; Mayer and Tiberj, 2004; Cautrès, 2004; and Elgie, 2004.) The approach of this book, expanded on below, attempts to encompass these different perspectives.

1.4

Our theoretical approach

The study of vote choice represents a vigorous branch of political science. In this vast literature on democratic elections, no choice is generally considered more important than the selection of the president. Different theories, some already enumerated, compete to explain this decision – rational choice, sociological, psychological, historicalinstitutional, communications, geographic, and economic. (For a contemporary review of these approaches, see Lewis-Beck, Jacoby, Norpoth, and Weisberg, 2008, ch. 2.) Our approach is straightforward, drawing on the founding “Michigan model” of political behavior (Campbell, Converse, Miller, and Stokes, 1960), or at least the French variant, “la variante française du paradigme de Michigan” (Mayer and Perrineau, 1992, p. 82). The presidential vote, an act of political

9780230_300088_02_ch01.indd 12

6/28/2011 8:22:01 PM

Introduction 13

behavior, is held to be a function of long-term and short-term forces. These forces can be organized theoretically in terms of a “funnel of causality.” (See the original discussion in Campbell, Converse, Miller, and Stokes, 1960, ch. 2; see also an updated sketch of the funnel of causality theory in Lewis-Beck, Jacoby, Norpoth, and Weisberg, 2008, ch. 2.) Imagine the vote itself occurs at the tip of the funnel, and that the causes of that vote are set ever-further back in the funnel as their origin becomes more remote in time. For example, at one extreme (just before the tip of the funnel), the voter may have just heard some very bad economic news on the way to the voting booth. At the other extreme (the far open end of the funnel), the voter may recall the religious education of his or her childhood. The first variable, economic perception, represents a short-term force operating on the vote decision. The second variable, religious upbringing, represents a long-term force acting on the vote decision. Such short- and longterm forces push and pull at the voter, ultimately triggering a vote for a particular candidate. Our book divides itself equally into three chapters on each set of forces. The long-term forces are of two types: sociological and psychological. The sociological variables are several, including key aspects of social structure and social class, in particular the variables lourdes – the “heavy variables” – of religion and occupation (see Chapters 2 and 3). The psychological variables, pivotal to the Michigan model we employ, include party identification and left-right ideological identification (see Chapter 4). The short-term forces are also of two types: issues and candidates. With respect to issues, the paramount concerns of immigration, economics, and Europe are given special attention (see Chapter 5). Candidate effects are treated in two different ways. First, they are assessed in terms of leadership qualities, a neglected field of presidential election studies in France (see Chapter 6). Second, the role of candidates is assessed dynamically, through examination of campaigns (Chapter 7). In a systematic series of logistic regression analyses, the weight of each of these sets of explanatory variables is assessed, as is the weight of the overall model. The equations are specified and estimated progressively, in accordance with the funnel of causality. That is to say, the most remote, stable, or steady causes are assessed first, beginning with an assessment of the impact of social structure. Progressively,

9780230_300088_02_ch01.indd 13

6/28/2011 8:22:01 PM

14

French Presidential Elections

chapter by chapter, the other elements of explanation are added to the model, as the analysis flows through the funnel from beginning to end. Once the tip of the funnel is reached, the theoretical journey is complete, and a fully specified model of French presidential vote choice, tested across four elections, can be observed. As will be seen, each set of variables makes its own important contribution to our understanding. Finally, a general explanation for French presidential vote choice is rendered. This general framework can be expected to provide a backdrop for assessing differences in effects and outcomes in the study of future elections. Clearly, the behavior of the French presidential voter can be explained. French voters are reasoning, even reasonable, political actors, able to perceive and act on their political interests, given their own background and issue preferences.

9780230_300088_02_ch01.indd 14

6/28/2011 8:22:01 PM

2 Social Structure

Recent work has suggested that the traditional socio-demographic differences among French electoral clienteles are crumbling. In a volume edited by Pascal Perrineau (2007) for instance, claims are made about the “embourgeoisement” of socialist voters (Rey, 2007), as well as the gradual disappearance of gender and generational gaps among voters who support different political standpoints in France (MossuzLavau, 2007; Muxel, 2007). Age, gender, and level of education are no longer reliable indicators of electoral behavior. Only religion, though declining in French society, remains a decisive cleavage distinguishing left- and right-wing electorates (Haegel and Sauger, 2007). At the heart of this chapter is a portrait of a socio-demographically evolving electorate in France. Four aspects of the profile of French electoral clienteles will be examined in detail – age, gender, education, and religion. These choices are easily justified based on the extant academic literature. Age, or more precisely generational replacement, often constitutes an important vector of change in political values, attitudes, and behavior (see Blais et al., 2004; Inglehart, 1990, 1997; Martin, 2000; Miller and Shanks, 1996). The existence of a gender gap, characterized by more support among women for leftist parties and candidates, has been documented in numerous countries (see Inglehart and Norris, 2003; Le Hay and Mossuz-Lavau, 2010; Mossuz-Lavau, 2007). The link between education and electoral behavior, and more generally, the effect of increasing education on the political dynamic in Western countries, has been the subject of much research in France and elsewhere (see Blais et al., 2004; Franklin et al., 1992; Muxel, 2009; Nevitte, 1996; Perrineau, 1997, 15

9780230_300088_03_ch02.indd 15

6/28/2011 8:21:02 PM

16

French Presidential Elections

2009; Rey, 2007). Finally, religious membership and practice constitute one of the more important factors structuring political behavior in numerous countries, notably Canada (see Blais et al., 2002), the United States (see Lewis-Beck et al., 2008; Miller and Shanks, 1996; Niemi and Stanley, 2006), and France (see Haegel and Sauger, 2007; Mayer and Perrineau, 1992; Michelat and Simon, 1977). Thus, our analysis of partisan coalitions in France will begin with an examination of their socio-demographic roots as gleaned from the factors of age, gender, education, and religion. The method used will examine the link between these variables and the presidential vote – first, in a descriptive manner, then through the use of multivariate models. This exercise will continue over the following chapters which factor in the effect of class and patrimony (Chapter 3), ideology (Chapter 4), issues (Chapter 5), leader image (Chapter 6), and electoral campaigns (Chapter 7).

2.1 An analytical method of explaining vote choice in France The dependent variable selected for explanation in this volume is presidential vote during the first and second rounds of the 1988, 1995, 2002, and 2007 elections. As such, we will maintain a comparative perspective throughout the analysis. The data from five largescale election studies – those led by CEVIPOF (1988, 1995, 2002, and 2007) and a study carried out under the guidance of Roy Pierce in 1988 will be analyzed in order to compare the factors that have influenced French vote choice between 1988 and 2007. For reasons of data availability, the focus will be on the votes for the main candidates from each of the five large political families – extreme left, moderate left (Mitterrand, Jospin, and Royal), center (Barre, Balladur, and Bayrou), moderate right (Chirac and Sarkozy), and extreme right (Le Pen). (A brief comment on the extreme-left candidates is in order. The most important candidates for this category are the Communist candidates (in each election – Lajoinie, Hue, or Buffet – average support of 5.2 percent) and Trotskyist candidates (Laguiller for Force ouvrière – in each election, average support of 3.6 percent; Juquin 1988 and Besancenot 2002 and 2007 for the Ligue communiste ouvrière – average support of 3.5 percent).) What explains the votes for these candidates during the first and second rounds of the presidential election? This

9780230_300088_03_ch02.indd 16

6/28/2011 8:21:02 PM

Social Structure 17

approach turns out to be comprehensive, allowing us to account for the electoral behavior of nine out of ten first-round voters and virtually all second-round voters. For all cases, save one, the selected dependent variable will be vote choice as reported by the respondent. (In the case of the second round of the 2002 election, we instead study vote intention in the hypothetical situation of a second round of voting between Chirac and Jospin, thus allowing the choice always to be along the left-right divide.) The order in which the vote determinants will be analyzed is based on the causal proximity between factors to be studied and vote choice; after first examining more remote factors, we will then examine more immediate factors affecting vote choice. As noted, the general model that will guide us throughout this exercise is that of the Michigan School, which distinguishes between long-term factors (such as social belonging and partisan identification) and short-term factors (such as leader image and issues), both of which are held to influence voter preferences (for a classic exposé of this model, see Campbell et al., 1960; for a recent update of this approach, see LewisBeck et al., 2008). The technique chosen is to use a block recursive model (see Bélanger and Nadeau, 2009; Blais et al., 2002; Miller and Shanks, 1996) which involves progressively introducing variables into an explanatory model, starting with long-term factors in voting behavior and ending with short-term factors. The first part of the volume (Chapters 2–4) discusses the long-term determinants suspected of influencing partisan choice in France (socio-demographic characteristics, social class and patrimony, and ideology), while the second part of the volume (Chapters 5–7) will delve into the influence of more time-sensitive factors (issues, leaders, and events).

2.2 Socio-demographic determinants of vote choice in France The first block of variables to be studied includes age, gender, education, and religion. Previous studies of France and other countries allow us to establish certain expectations regarding the effect of these factors on support for different candidates during presidential elections. We will examine these expectations for the four variables under study before undertaking the statistical analysis of their impact on presidential voting from 1988 to 2007.

9780230_300088_03_ch02.indd 17

6/28/2011 8:21:03 PM

18 French Presidential Elections

Age In general, younger voters tend to support parties that are more to the left (Blais et al., 2002, and Clarke et al., 2004, have observed similar results in Canada and in Great Britain, respectively). However, this effect is not consistent and depends on contextual factors. For example, young American voters were attracted to the Republican Party during the 1980s, before returning to the Democrats and supporting Barack Obama in 2008 (see Abramson et al., 1990; Lewis-Beck et al., 2008; and Norpoth, 1987). The behavior of young French voters also seems to have evolved over time; their traditional support for the left seems to have crumbled over the past few elections (Muxel, 2007). To that, add a peculiar attraction to “new” political parties and to candidates on the extremes of the political spectrum, notably Jean-Marie Le Pen (see Perrineau, 1997; this same author still notes a recent influx in young voter support for Le Pen in 2007). It also seems that younger voters have been susceptible to the “Bayrou temptation” (Strudel, 2009). The portrait of young French voters is rather complex; it suggests that they are tempted by both extremist candidates and new political experiences (e.g., environmentalism and Bayrouism). Gender The women’s vote in France seems to reveal certain unique characteristics that have led Le Hay and Mossuz-Lavau (2010) to ask the question in their title: Genre et politique: y a-t-il une exception française? Having entered electoral politics later, French women in the twentieth century had a tendency to support the right, until after the legislative elections of 1986 they aligned themselves more with the left (see Mossuz-Lavau, 2007). This more left-wing orientation among French women is shown in a lukewarm attitude toward the extreme right (Perrineau, 2009) and a strong attraction to candidates of the moderate right (see Capdevielle et al., 1981, Table 14, p. 240; Cautrès and Mayer, 2004). In a recent book about the influence of gender in French presidential elections, Mariette Sineau (2010) adds another layer of complexity to this question by suggesting that a “gender-generation gap” may have emerged in 2007, with young women more likely to support Royal and older women more likely to support Sarkozy. All these findings point to the question: is it possible that the gender gap observed in various democracies is less pronounced in France?

9780230_300088_03_ch02.indd 18

6/28/2011 8:21:03 PM

Social Structure 19

Education The effect of education on French electoral behavior has also changed. Yet we can see over time that the relationship between level of education and political orientation on a left-right axis has never been linear. In fact, each large political families is characterized by having the support of both more educated and less educated voters. The profiles of voters who supported different political standpoints during the legislative elections of 1978 illustrate this phenomenon (see Capdevielle et al., 1981, Table 16, p. 242). During this election, the more educated voters were those who supported the extreme left, notably those of Michel Rocard’s Parti socialiste unifié (PSU). That group counted among its ranks close to one in two people (46 percent) who had completed postsecondary or university-level education. Continuing along the left-right scale, the Communist electorate was the least educated of all, with an electorate having completed postsecondary or university-level education at half the share (23 percent) of the PSU. The proportion of educated voters climbs back up to 31 percent among Socialist voters, a percentage that is lower than that observed for supporters of the centrist UDF (35 percent), but similar to that of RPR supporters (31 percent). The changes in the French political game over the past few decades have altered the link between education and partisan preferences without modifying their nonlinear quality. The collapse of the PCF, the gentrification of the socialist clienteles, and the rise of the National Front have contributed to reinforcing the education distinction between two large political families on the left. That is to say that nowadays, the voters of the moderate left (notably those of the PSU) are more educated than are those of the extreme left (Rey, 2007). On the right, the centrists make up the most educated group, followed by supporters of the moderate right, and, far behind, those of the extreme right (see Haegel and Sauger, 2007; Perrineau, 1997, 2007, 2009). This jagged relationship between education and ideological position leads one to believe that the link between this variable and the vote could be significantly stronger during the first round of an election, as both less educated and more educated voters from the left and right rally around the candidate who shares their political attitude.

9780230_300088_03_ch02.indd 19

6/28/2011 8:21:03 PM

20

French Presidential Elections

Religion Religion is still fundamental to French political choice and attitudes (Boy and Mayer, 1997; Michelat and Simon, 1977; Cautrès and Mayer, 2004). There are two dimensions of religion in this regard: religious belonging (either Catholic or non-Catholic) and religious practice (either frequent or irregular). The four combinations this model recognizes are practicing Catholics, nonpracticing Catholics, those belonging to other religions, and those who belong to no religious group. The work that has been done on the link between religion and political behavior in France has shed light on the existence of a nonlinear relationship between religion and voting. The voters who are furthest away from practicing Catholicism are those on the extreme left, followed by those on the moderate left. This link changes on the right, where the centrists generally seem to be more influenced by religion than are those who support the moderate right and the extreme right (see Haegel and Sauger, 2007; Perrineau, 2007).

2.3 A bivariate analysis of socio-demographic determinants of the presidential vote in France Tables 2.1 to 2.4 present a first glimpse at the link between these socio-demographic variables and vote choice in French presidential elections. The data in these tables show the percentages of support obtained by the candidates grouped by large political families (extreme left, moderate left, center, moderate right, and extreme right). The comparisons presented are those between men and women, between voters above and below 40 years of age, between those who completed their baccalauréat – a certificate of school completion required to enter higher education – and those who did not, and finally among different categories of religious practice. To facilitate comparison, the mean values of support for candidates by category of voter for the four elections under study are presented in Table 2.5. Completing the portrait are measures of bivariate association – Cramer’s V, Pearson’s chi-square, and Kendall’s tau-b. These measures allow us to establish the strength, direction (positive or negative), and degree of statistical significance of the relations observed among the variables. The marked differences between Cramer’s V and tau-b allow us to discern the presence of a nonlinear relationship between the effect of one variable on the ideological

9780230_300088_03_ch02.indd 20

6/28/2011 8:21:03 PM

Social Structure 21

positioning of the respondents. Such a situation could appear with respect to age because of a certain attraction among young voters for political options situated on the two extremes of the political spectrum (the extreme left and the extreme right). Some conclusions can be drawn from the examination of these tables. The first speaks to religion as an important general difference among voters who support different political families in France. This relationship is presented in Table 2.1 and in columns 7–10 of Table 2.5. For example, in the second round, the candidates on the right collected on average three times more votes among practicing Catholics (74 percent) than among those who reported not belonging to a particular religion (26 percent) (see Table 2.5). The different measures of bivariate association presented in Table 2.1 also emphasize the strong and consistent effect of religion on voting (note for example the value of Cramer’s V, which fluctuates within a narrow range from 0.18 to 0.21). That being said, it is important to note that the impact of religion on political choice is not a simple one such that degree of religiosity equates with degree of conservatism. If the voters on the extreme left seem to be more removed from religion than voters on the moderate left, the portrait is quite different for the right. The factor of religion is less tied with support for the extreme right than with support for the center and the moderate right. In this last group, even if the moderate right benefits from the religious vote, some important variations can be observed from one election to another. In 1995, support for centrist candidate Edouard Balladur was very prominent among practicing Catholics (38 percent) and almost nonexistent among those “without religion” (5 percent), while in 2007 François Bayrou received almost identical levels of support (19 percent for practicing versus 23 percent for nonpracticing) in those same groups (see Table 2.1). The data in Table 2.2 show that the link between gender and political choice is less pronounced and is oriented differently in France than in other Western democracies, notably the United States, where women recently have preferred candidates who are “more to the left” of the Democratic Party. On the contrary, women in France seem to favor right-wing candidates more than men do. During the second round of the past four presidential elections, the right has received, on average, 49 percent of the female vote versus 46 percent of the

9780230_300088_03_ch02.indd 21

6/28/2011 8:21:03 PM

Nonpracticing Catholics

Extreme 27 14 9 left Left 53 41 44 Center 7 13 16 Right 5 20 19 Extreme 8 11 12 right N 373 123 2114 Cramer’s V .19 Pearson’s χ2 .00 (P) Gamma .36 (.02) Kendall’s .22 (.02) tau-b

No religion

1988

545 .20 .00

499

.24 (.02) .16 (.02)

41 5 15 13

24 23 38 10

Practicing Catholics 26

No religion

4

1995

Other

Nonpracticing Catholics 30 16 26 16

12 16 38 35 8

4

303 .21 .00

36 5 12 12

36

No religion

.29 (.04) .20 (.02)

Practicing Catholics

131 1927 390

41 14 21 15

9

Other 60

38 8 25 13

15

Nonpracticing Catholics 655

26 10 29 16

19

Practicing Catholics 112

11 28 42 8

12

No religion .35 (.02) .23 (.01)

1000 .18 .00

38 23 18 5

16

2007

25 21 39 7

7

Nonpracticing Catholics

16 19 57 4

4

207 1860 267

46 27 18 2

7

Other

2002

Practicing Catholics

Table 2.1 Percentage support according to religion in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007)

1st Round

Other

9780230_300088_03_ch02.indd 22

6/28/2011 8:21:03 PM

9780230_300088_03_ch02.indd 23

6/28/2011 8:21:03 PM

Left 86 64 61 34 72 57 46 20 72 63 49 26 68 70 40 24 Right 14 36 39 66 28 43 54 80 28 37 51 74 32 30 60 76 N 388 132 2268 513 599 150 2043 432 720 184 1735 311 1018 203 1931 275 Cramer’s V .28 .30 .27 .30 .00 .00 .00 .00 Pearson’s χ2 (P) Gamma .52 (.03) .52 (.03) .46 (.03) .50 (.02) Kendall’s .26 (.02) .26 (.02) .25 (.02) .28 (.02) tau-b

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the value of P is indicated for Pearson’s χ2. The standard errors are indicated in parentheses for Gamma and Kendall’s tau-b. N = total number of respondents in the relevant category. Tau is generally the leading measure of association for ordinal variables (Lewis-Beck, 1995). A tau of 1.0 (–1.0) indicates a perfect positive (negative) monotonic relationship. A tau of .00 indicates no monotonic relationship. It is superior to a rival ordinal measure of association, gamma, in that it takes into account tied pairs, and is therefore not inflated (as is gamma). There are two tau measures in common use: tau-b and tau-c. If the table is square (same number of rows as columns), then tau-b is preferred. If it is rectangular (different number of rows and columns), then tau-c is technically preferred over tau-b (because it is impossible here for the latter to theoretically reach 1.0). Substantively, the differences between the two are small. That is to say, if one runs a tau-b on a rectangular table it will generally yield a number very similar to tau-c. Thus, throughout this book tau-b is reported. The consistency in use of tau-b also facilitates comparison across tables. Source: CEVIPOF (1988, 1995, 2002, 2007).

2nd Round

24 French Presidential Elections

Table 2.2 Percentage support according to gender in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007) 1988 M 1st Extreme 12 Round left Left 39 Center 15 Right 20 Extreme 14 right N 1491 Cramer’s V .08 Pearson’s .00 χ2 (P) Gamma .04 (.03) Kendall’s .03 tau-b (.02) 2nd Left 60 Round Right 40 N 1562 Cramer’s V .01 Pearson’s .57 χ2 (P) Gamma .02 (.04) Kendall’s .01 tau-b (.02)

1995

2002

2007

F

M

F

M

F

M

F

9

15

12

22

24

10

10

44 17 21 9

32 14 22 18

30 19 27 11

31 8 24 16

24 12 27 13

29 22 31 8

30 21 34 5

1629 1446 1584 .12 .00

549 .11 .01

585

–.00 (.03) –.00 (.02)

.01 (.04) .01 (.03)

61 39 1748

51

46

56

1577 1764 .07 .00 .02 (.03) .01 (.02)

50

50

48

49 54 44 50 1723 1543 1409 1556 –.06 –.05 .00 .00

50 52 1793 1640 –.02 .22

–.11 (.04) –.06 (.02)

–.04 (.03) –.02 (.02)

–.11 (.04) –.05 (.02)

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the value of P is indicated for Pearson’s χ2. The standard errors are indicated in parentheses for gamma and Kendall’s tau-b. N = total number of respondents in the relevant category. Source: CEVIPOF (1988, 1995, 2002, 2007).

9780230_300088_03_ch02.indd 24

6/28/2011 8:21:04 PM

9780230_300088_03_ch02.indd 25

6/28/2011 8:21:04 PM

Left Right N Cramer’s V Pearson’s χ2 (P) Gamma Kendall’s tau-b

2nd Round

61 39 2448 .01 .43 .03 (.04) .01 (.02)

No 10 43 14 20 13 2335 .10 .00 –.02 (.03) –.01 (.02) 59 41 862

Yes 12 37 21 21 8 785

47 53 2092 –.03 .05 –.07 (.04) –.03 (.02)

No 13 29 16 24 17 1983 .11 .00 –.11 (.03) –.07 (.02) 51 49 1174

Yes 13 35 17 26 9 1047

Baccalauréat*

Baccalauréat*

51 49 1735 –.04 .02 –.09 (.04) –.04 (.02)

No 24 24 7 27 19 624 .20 .00 –.14 (.04) –.09 (.03) 55 45 1224

Yes 22 32 14 23 8 510

Baccalauréat*

2002

45 55 2334 –.10 .00 –.22 (.04) –.10 (.02)

No 10 28 18 35 8 2267 .16 .00 –.12 (.03) –.07 (.02)

56 44 1099

Yes 8 33 29 28 3 1074

Baccalauréat*

2007

Source: CEVIPOF (1988, 1995, 2002, 2007).

* A baccalauréat in France is a certificate of school completion, needed for access to higher education.

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the value of P is indicated for Pearson’s χ2. The standard errors are indicated in parentheses for Gamma and Kendall’s tau-b. N = total number of respondents in the relevant category.

Extreme left Left Center Right Extreme right N Cramer’s V Pearson’s χ2 (P) Gamma Kendall’s tau-b

1st Round

1995

1988

Table 2.3 Percentage support according to level of education in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007)

9780230_300088_03_ch02.indd 26

6/28/2011 8:21:04 PM

.13 (.02)

Kendall’s tau-b

55 45 1856

8 42 16 20 11 3120

40 +

.08 (.02)

53 47 1369 .08 .00 .16 (.04)

16 31 11 24 18 1245 .15 .00 .00 (.03) .00 (.02)

18–39

1995

45 55 1897

11 31 20 25 12 1785

40 +

.05 (.02)

56 44 1068 .05 .01 .10 (.04)

28 28 10 20 13 429 .12 .00 .15 (.04) .09 (.03)

18–39

2002

51 49 1897

20 27 10 29 15 705

40 +

.09 (.02)

55 45 1253 .09 .00 .19 (.04)

10 29 28 28 6 1211 .12 .00 .08 (.03) .05 (.02)

18–39

2007

45 55 2062

9 30 18 36 7 2020

40 +

Source: CEVIPOF (1988, 1995, 2002, 2007).

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the value of P is indicated for Pearson’s χ2. The standard errors are indicated in parentheses for gamma and Kendall’s tau-b. N = total number of respondents in the relevant category.

67 33 1454 .13 .00 .26 (.30)

Left Right N Cramer’s V Pearson’s χ2 (P) Gamma

2nd Round

13 45 13 16 12 1358 .13 .00 .16 (.03) .09 (.02)

Extreme left Left Center Right Extreme right N Cramer’s V Pearson’s χ2 (P) Gamma Kendall’s tau-b

1st Round

18–39

1988

Table 2.4 Percentage support according to age in French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007)

9780230_300088_03_ch02.indd 27

6/28/2011 8:21:04 PM

58

42

51

49

33 16 28 11

33 16 22 12

Left

12

17

Right

Extreme left Left Center Right Extreme right

46

54

33 15 24 14

15

49

51

32 17 27 10

14

49

51

31 14 27 14

14

No Baccalauréat*

Religion

45

55

34 20 25 7

14

26

75

42 10 13 10

26

37

64

42 16 21 10

11

Source: CEVIPOF (1988, 1995, 2002, 2007).

*A baccalauréat in France is a certificate of school completion, needed for access to higher education.

51

49

31 16 28 13

12

74

26

17 27 43 8

6

No Nonpracticing Practicing Baccalauréat* religion Other Catholics Catholics

Education

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error.

2nd Round

1st Round

F

Gender

18–39 40+ M

Age

47

53

32 16 26 12

14

Total

Table 2.5 Mean support score for each category of voters for four French presidential elections (first and second rounds for 1988, 1995, 2002, and 2007)

28

French Presidential Elections

male vote (see Table 2.2). This result seems to be the product of two opposing tendencies coming together in the French case. In the second round, the more masculine vote for the extreme right combines with the more feminine support for candidates of the moderate right and the center. The final effect of this rally is a classic gender gap, but one characterized by more support among women for the right. The attraction of the female electorate to candidates of the moderate right, a unique aspect of French electoral sociology, seems to explain why increasing female support for the left in other democracies is not yet reflected in France. On the whole, the relationship between gender and electoral behavior remains mixed, as suggested by the measures of association in Table 2.2, which are generally weak and sometimes statistically insignificant (especially in the second rounds in 1988 and 2007). The effect of education on presidential voting, shown in Table 2.3, presents some elements of both continuity and change. The support of educated voters for centrist candidates and those of the moderate left has continued over the course of the four elections under study. The persistent trouble that the extreme right has had in breaking into this group is also notable; the average support received by Jean-Marie Le Pen among more educated voters is half that of support from less educated respondents (7 percent versus 14 percent, see Table 2.3). Also, two cleavages have become more prominent over time. On the left, support for the socialists seems to have consolidated among more educated voters, while the extreme left has been successful among voters who are less educated (especially since 2002). The opposite is true for the right. While the more moderate candidates in this camp – centrists Barre, Balladur, and Bayrou – have maintained and consolidated their support among the more educated, the candidates of the moderate right have obtained the most votes among the less educated, especially since 2002. The culmination of these tendencies is clear in the second round. Over time, the combined rallying of the less educated voters for the extreme right since 1988, and for the moderate right since 2002, seems to have contributed to reversing the relationship between education and vote choice. While in 1988, Jacques Chirac received slightly higher support among educated voters (41 percent versus 39 percent), the opposite is clearly observed in the second round of the 2007

9780230_300088_03_ch02.indd 28

6/28/2011 8:21:05 PM

Social Structure 29

election. In this case, the gap between the more educated and the less educated was the highest; Nicolas Sarkozy obtained 55 percent support among voters without a college degree versus 44 percent among those with a college degree. The measures of association in Table 2.3 confirm these interpretations. While Cramer’s V, a measure that does not suppose that the relationship between two variables is linear, is statistically significant for all the elections in the first round, this is not the case for Kendall’s tau-b. The nonlinear relationship between education and ideological orientation is clear. The weakness of the link between education and vote choice in the second round is also apparent. The measures of association in this case are largely weak and generally not significant, with the exception of the 2007 election, in which the tendency of less educated voters to support the right seemed to consolidate itself. The relationship between a respondent’s age and vote, presented in Table 2.4, has also seen a change since 1988. On the left, we are witnessing a graying of the socialist clienteles and, inversely, a rejuvenation of support for the extreme left. On the right, the reverse is true. While the clientele of more moderate candidates in the center seems to be rejuvenated, most of the support of the moderate right still falls among older voters. The tendency of young voters to support the extreme right seems to be less substantiated across elections. The cumulative result of these different changes does not seem to have changed in any significant way the generational equilibrium between the electorate for the left and the electorate for the right in the second round. On these occasions, right-wing candidates (Jacques Chirac in 1988, 1995, and 2002 and Nicolas Sarkozy in 2007) seem to attract consistently higher support among those over the age of 40, as compared to among those ages 18–39. As seen in Table 2.4, in the second round of each election right-wing candidates captured between 5 and 12 percent more of the vote of the older demographic compared to the younger demographic. The measures of bivariate association confirm this overall portrait. The diverging results of Cramer’s V and tau-b strongly suggest that the relationship between age and vote choice does not follow in any linear manner the leftright ideological positions of respondents. The results also confirm that the complex game of rallies in the second round of elections contributes to the emerging, yet clear, tendency of older voters to support the right during the second round.

9780230_300088_03_ch02.indd 29

6/28/2011 8:21:05 PM

30 French Presidential Elections

The portrait of electoral clienteles that emerges from this first examination of the link between the socio-demographic variables of age, gender, education, and religion is nuanced. Right-wing candidates, winners in the 1995, 2002, and 2007 elections, have benefited from their higher support among practicing Catholics, women, and older, less educated voters. This overall picture is the result of a mixing of clienteles that are sometimes very different. The case of gender, presented in Table 2.2, is the most clear. During the second round, the rally of a more masculine electorate from the National Front and a more numerous and feminine group of supporters from the moderate right and center produced a final result that shows a gender gap; women support the right more than they do the left, but a right that is quite moderate. Overall, though, the disparity between men and women in their vote choice still remains limited. A similar dynamic can be observed in the case of education, in Table 2.3. In this case, it is the addition of a large number of less educated voters for the National Front and for the moderate right, to the more educated supporters of the center, that explains the electoral coalition of the right-wing candidates in the second round. That is, the voters for the right-wing candidates are a little less educated than are their left-wing counterparts. Such dynamics are also seen on the left, where in the second round, younger and less educated voters for the extreme left are mixed in with older and more educated voters for the moderate left. Only religion escapes this trend of contradictory movements in the second round, since this variable works more or less in mirror fashion on the left and the right. These observations, which will be reexamined in the next section with the help of multivariate models, indicate that French political clienteles still show some rather typical profiles, which are particularly noticeable during the first round. That being said, there are no fixed dividing lines between groups of voters who support different political standpoints. On the left, the gentrification of the socialist electorate, increasingly older and more educated, is drawing the socio-demographic center of gravity of this group away from the extreme left and bringing it closer to the profile of a centrist electorate. From this perspective, the reconfiguration of support for centrist candidate François Bayrou in 2007 represents one of the most significant transformations of an electoral clientele in any of the elections under study. Until 2002, support for Barre, Balladur, and Bayrou came

9780230_300088_03_ch02.indd 30

6/28/2011 8:21:05 PM

Social Structure 31

from voters who fit the traditional profile of centrist partisans – more feminine, older, more educated, and more religious than are the rest of the electorate. The 2007 election marks an important shift in the makeup of centrist supporters. In that election François Bayrou’s supporters still consisted of voters who were more educated, but he won a particularly striking increase of support from men, those “without religion,” and especially young voters. The profile of voters who supported Bayrou in 2007 is not only very different from that of those who supported previous centrist candidates (notably Edouard Balladur in 1995) but is in marked contrast to that of Nicolas Sarkozy supporters (proportionally older, more female, less educated, and more religious than are supporters of Bayrou). The future will tell whether the tentation bayrouiste (Bayrou temptation) marks a long-term change in French electoral clienteles and whether the long-standing conclusion about the homogeneity of the “conservative electorate” (Lancelot, 1981; Sauger and Haegel, 2007) needs to be revisited.

2.4 A multivariate analysis of socio-demographic determinants of the presidential vote in France Through the use of multiple regression, this section will examine the link between respondents’ socio-demographic characteristics (age, gender, education, and religion) and their tendency to support candidates from the extreme left, left, center, right, or extreme right during the first and second rounds of French presidential elections from 1988 to 2007. (The Greens, or Ecologists, are not included here, for reasons of comparability and complexity.1) For each of these models, the dependent variable takes a value of 1 when the respondent reports having voted for the candidate from one of these large political families (for example, Jean-Marie Le Pen) and 0 otherwise. In the case of the second round, the dependent variable takes a value of 1 when the respondent expresses support for the right-wing candidate and 0 in the opposite case. Therefore for each election, there are six models, of the following specification: Presidential vote = age + gender + education + religion. The preferred method of estimation when the dependent variable is dichotomous, the case here, is logistic regression (Woolridge 2006).

9780230_300088_03_ch02.indd 31

6/28/2011 8:21:05 PM

32 French Presidential Elections

The complete results of the analyses are presented in the Appendix, as a series of logistic regression coefficients for each of the elections studied. These coefficients measure the effect of one variable, all other things being equal, on the probability of supporting a candidate. The sign of the coefficient reflects the direction of the observed relationship, while its value indicates the strength of the association in question. The use of asterisks marks the presence of a statistically significant effect. For example, the coefficient of 1.16 for the variable “age” in the 1988 election means that there is a relatively strong, positive, and statistically significant relationship between the respondent’s age and the probability that the respondent will support the moderate right (in this case, Jacques Chirac) at the ballot box. Direct interpretation of logistic regression coefficients is difficult. In order to present this information in a more intuitive manner, these coefficients were transformed into “changes in probabilities.” This information is presented in Table 2.6. The figures in the table indicate the change in the probability of voting for a candidate when one passes from the “minimum” category to the “maximum” category on an independent variable. In the case of gender, which takes a value of 1 for men and 0 for women, the value of the coefficient indicates the difference between men and women in the probability of voting for a candidate, all other things being equal. In the case of religion, measured with the help of four categories, scaled from 0 to 1, where 0 corresponds to “without religion” and 1 corresponds to “practicing Catholics,” the coefficient presented in Table 2.6 represents the increase or decrease in the probability of voting for a candidate among individuals belonging to these extreme categories. The interpretation is the same for the variables of age and education. For example, in the second round of the 1988 election, the coefficients of 0.24, 0.02, 0.14, and 0.47 represent the increase in the probability of supporting right-wing candidate Jacques Chirac among older, male, educated, and practicing Catholic voters versus those having the opposite characteristics. A word must be said about the performance of this socio-demographic vote model. Pseudo-R2 is a measure that provides a useful indicator of this subject. (There are multiple pseudo-R-squared measures for binomial logistic regression models. We prefer, and report, the McFadden pseudo-R-squared, because it is most analogous to the classical regression R-squared (Menard, 1995).) The value of this indicator will always

9780230_300088_03_ch02.indd 32

6/28/2011 8:21:05 PM

Social Structure 33

Table 2.6 Change in vote probability in French presidential elections, going from the minimum to the maximum category of each of the socio-demographic variables Age

Male

Education

Religion

1988

Extreme left Left Center Right Extreme right 2nd round

–.02 .02 .12 .17 –.02 .24

.01 –.04 –.01 .00 .04 .02

.00 –.09 .11 .06 –.05 .14

–.17 –.15 .14 .27 .05 .47

1995

Extreme left Left Center Right Extreme right 2nd round

–.09 .16 .24 –.02 –.24 .03

.01 –.00 –.03 –.03 .07 –.03

–.05 .06 .08 .02 –.19 .01

–.18 –.21 .18 .16 .02 .45

2002

Extreme left Left Center Right Extreme right 2nd round

–.08 .18 .02 .25 –.06 .03

–.03 .02 –.02 –.04 .02 –.03

–.08 .05 .07 –.05 –.19 –.04

–.14 –.15 .08 .17 .03 .39

2007

Extreme left Left Center Right Extreme right

–.04 .09 –.09 .17 –.07

–.01 –.03 –.01 –.01 .03

–.06 .03 .13 –.00 –.09

–.11 –.21 –.00 .32 .03

.10

.02

–.08

.41

2nd round

Notes: Entries are changes in voting probabilities generated by Clarify (Tomz, Wittenberg, and King 2003) for a variation in age, gender, education, and religion, when all other variables are held constant. Probability changes are based on regression model 1 (column 1 in the Appendix). Source: CEVIPOF (1988, 1995, 2002, 2007).

9780230_300088_03_ch02.indd 33

6/28/2011 8:21:05 PM

34

French Presidential Elections

fall between 0 and 1; it will approach 0 if the capacity of the model to explain vote choice is limited and will approach 1 (only in exceptional cases) if the model has a very precise predictive power. Given these baselines of comparison, the information contained in the Appendix shows that these socio-demographic variables only provide a limited explanation of French electoral behavior. The mean pseudo-R 2 for the second round of the presidential election is 0.06. The same measure for each of the political families in the first round is 0.04 for the extreme left, 0.02 for the moderate left, 0.05 for the center, 0.06 for the right, and 0.04 for the extreme right. These results are comparable to what has been observed in other countries, notably Canada, Great Britain, and the United States (see Blais et al., 2002; Clarke et al., 2004; Lewis-Beck et al., 2008), and they show that a full explanation of vote choice must include additional factors such as the social status of respondents; their ideological orientation; and their opinions on issues, on leaders, and campaign events themselves. The contribution of these additional factors to explaining the vote will be examined over the course of the rest of the volume. The slightly higher pseudo-R 2 value in the vote models for right-wing candidates suggests that the socio-demographic profile of this voter is slightly more structured than that of voters for left-wing candidates. Finally, the relatively constant level of these measures from one election to another (for the second round, 0.07, 0.06, 0.06, and 0.07, respectively) suggests there is no decline in their impact on the French vote over the past 20 years. Further, our regression analyses show that religion is the variable that has the most important and consistent impact on the vote, more so in the first round than in the second round. All the coefficients linked to this variable are of the expected sign that is to say, negative for left-wing candidates and positive for those of the center and the right. Also, they are all statistically significant with the exception of the extreme right in 1995 and in 2002 (see the Appendix). The persistent effect of religion on support for left-wing and right-wing candidates explains the important and consistent impact of this variable during the second round. The four coefficients of the religion variable measure the effect of religion on the tendency of voters to support a right-wing candidate (Chirac in 1988, 1995, and 2002; Sarkozy in 2007). These coefficients, all positive and statistically significant, indicate the steady and marked effect of religion on voting for the

9780230_300088_03_ch02.indd 34

6/28/2011 8:21:05 PM

Social Structure 35

elections under study. The data in Table 2.6 confirm this interpretation of the results. The effect of the variable of religion on the probability of supporting right-wing candidates in the four elections under study is not only substantial – the mean effect is 43 percentage points – but also consistent, as it varies within a narrow range (between 39 and 47 points). The variable of religion has different effects according to large political families and the round of voting. The negative effect of religion on one’s disposition to support extreme-left candidates is consistent and of a rather sizeable magnitude. All of the coefficients tied to this variable for left-wing candidates are both negative and statistically significant (see the Appendix). The effect of this factor on the probability of supporting candidates from these two left-wing families is similar, with those frequently practicing Catholicism reducing the average probability of supporting the moderate left and the extreme left by 18 and 15 points during the first round, respectively. For the right, there is a completely different picture. First, the impact of religion is consistently weaker in explaining vote choice for the extreme-right and the moderate-right. Two of the four coefficients linking religious sentiment and support for the extreme right are not statistically significant (those that correspond to the 1995 and 2002 elections; see the Appendix) and the average effect of this variable on the probability of voting for the National Front candidate is small (4 percentage points on average; see Table 2.6). The strength of the relationship between this variable and support for candidates of the center and the right has varied considerably from one election to another. The contrast between religion’s significant effect in the case of Edouard Balladur in 1995 and its absent effect on support for François Bayrou in 2007 is particularly striking. The moderate right largely benefits from the significance of religious sentiment in consolidating its support. However, the magnitude of this advantage has varied. The impact of religion on support for the moderate right was comparatively low in 1995 and 2002 (an effect of 16 and 17 points), high in 1988 (effect of 27 points), and highest in 2007 with the candidacy of Nicolas Sarkozy (effect of 32 points). If religion has a significant and consistent impact on the French vote, this does not seem to be the case for gender. For the second round, only one coefficient out of four is statistically significant, and that just barely (in 1995). The effect of this variable is limited (about

9780230_300088_03_ch02.indd 35

6/28/2011 8:21:06 PM

36 French Presidential Elections

3 percentage points) and unstable; it is sometimes slightly positive (in 1988 and 2007, which reflects a slight tendency for men to vote for the right) and sometimes negative (as in 1995 and 2002). The impact of gender on voting is more noticeable in the first round, with the vote for the extreme right being the clearest case. The positive coefficients in all cases (three of which are statistically significant) indicate that men were more inclined than were women to support Jean-Marie Le Pen. The other apparent trend is the attraction of women for centrist and moderate-right candidates, particularly in the 1995 and 2002 elections. On this occasion, the four coefficients that link back to gender are both negative (indicating a less marked disposition among men to support the center and moderate right in the first round) and statistically significant. The opposite trends for the right, with the attraction of men for the extreme right and that of women for the moderate right, nonetheless indicate an impact on the vote that is overall quite limited (see Table 2.6); during the second round, any impact cancels itself out when the more male electorate of the National Front combines with the more female electorate of the classic right. The effect of a respondent’s level of education on vote choice has changed over the course of the last four presidential elections. This change is particularly noticeable during the decisive second round of the election. In 1988, the probability of supporting right-wing candidate Jacques Chirac was 14 percentage points higher for more educated respondents. In 2007, the effect of education acted in the opposite way. The probability of supporting Nicolas Sarkozy decreased 8 percentage points among more educated voters2 (see the last row of the third column in Table 2.6). An examination of the relationship between education and vote choice in the first round allows us to understand the change in the link between education and partisan preferences. For example, in 1988 support for François Mitterrand was inversely related to the respondent’s level of education (the statistically significant coefficient of this variable is –0.42, and the probability of supporting a socialist candidate in the first round is 9 percentage points lower among more educated respondents). This trend reverses itself in the following election. For the next three elections, the link between education and support for the moderate left switches. The coefficients associated with this variable are now positive and statistically

9780230_300088_03_ch02.indd 36

6/28/2011 8:21:06 PM

Social Structure 37

significant in two out of three cases. The probability of voting for Lionel Jospin is 5 and 6 percentage points higher among more educated voters, which is a reversal of about 15 percentage points in relation to the 1988 election (see the Appendix and Table 2.6). However, for the extreme left, the effect of education takes the opposite direction, starting out positive in 1988 but switching to negative in the subsequent elections. The net effect of these different currents appears in the second round, in which we see that the effect of education on support for the right changes from positive to negative from 2002 on. This conclusion comes out of the mixing of a more numerous and progressively more educated electorate for the moderate left, with a less numerous and less educated electorate for the extreme left. The evolution of the relationship between education and voting has found its counterpart on the right. On this side, the tendency of more educated voters to support centrist candidates has been maintained across time. The negative link between education and support for the extreme right has come into play since 1995. A similar change has manifested itself in the sources of support for the moderate right. In 1988 there was a positive (and statistically significant) relationship between education and support for the moderate right (see the logistic regression coefficient of 0.46 for this election; a higher level of education translates to an increase of 6 percentage points in the probability of supporting Jacques Chirac in the first round). However, this relationship receded with the next election. All of these changes – the increasing education among socialist voters and the rise in support for the moderate right and the extreme right among less educated voters – have contributed to reversing over time the relation between support for left-wing and right-wing candidates in the second round of French presidential elections. The relationship between age and voting also presents a set of complicated issues. The marginal tendency of older voters to support extremist candidates, either from the left or the right, stands out. All the coefficients linked to this variable suggest a negative connection – seven of these eight coefficients are statistically significant – between a respondent’s age and support for extremist candidates. In this regard, older voters’ reserve toward Jean-Marie Le Pen’s candidature is the most consistent. The graying of the moderate-left electorate relative to that of the extreme right since 1995 is equally notable. The

9780230_300088_03_ch02.indd 37

6/28/2011 8:21:06 PM

38

French Presidential Elections

logistic regression coefficients reveal the existence of a positive link between age and socialist vote and a negative association between the same variable and support for extreme-right candidates, since 1995. The attraction of older people for the moderate right is notable over the whole period under study, with the exception of the 1995 election, in which Jacques Chirac seems to have been able to rally support among a higher proportion of younger voters. Older voters who were largely more inclined to favor Edouard Balladur in 1995 (the positive effect observed is 24 percentage points) seemed to be more hesitant than did younger voters to support François Bayrou in 2007 (the negative effect observed is 9 percentage points; see Table 2.6). If the rallying of younger voters, perhaps attracted by Chirac’s discourse on “la fracture sociale” (the social divide) in 1995, has led to a relative graying of the centrist electorate in relation to that of the moderate right, then the tentation bayrouiste has clearly produced the inverse effect 12 years later.

2.5 Conclusion This chapter examined the socio-demographic profile of the main electoral clienteles in France. The overall contribution of the four variables studied – age, gender, education, and religion – to explaining French electoral behavior remains quite modest. In this sense, the conclusions of this chapter are in line with the findings from large-scale electoral studies carried out in the United States, Canada, Great Britain, and Europe in general. That being said, the results presented should still be of interest. They show significant changes in the socio-demographic structure of partisan groups in France. They also reveal several dynamics that seem to be unique to the French case. For these two reasons, the study of these variables remains important for understanding French electoral behavior. The first characteristic in the French case is the importance of religious sentiment in voters’ electoral choices. Catholic faith and the frequency with which such faith is observed strongly contribute to structuring partisan choices in France, more so in the first round of the election than in the second round. In fact, even if religion does not always have a conservatizing influence on vote choice, with voters on the extreme right being less sensitive to this question than are those on the classic right, the effect of religion clearly divides left-wing

9780230_300088_03_ch02.indd 38

6/28/2011 8:21:06 PM

Social Structure 39

and right-wing partisans in the second round. The important and consistent effect of religion on voting remains a defining facet of French electoral sociology. The observations of Lewis-Beck’s (1984, 438) investigation of the impact of religion on the French vote, carried out over 25 years ago, still hold: “across the years of the Fifth Republic, the allegiance of the faithful to the parties on the Right has been unswerving.” Of course, it remains to be seen if this variable will continue to exercise such an impact, as French society continues to secularize and diversify itself. The effect of gender on voting is also peculiar in France. If the more marked support of men for the extreme right is not surprising, then the attraction of French women to the classic right would seem to be less in line with the behavior of this electorate in other democracies, where the “traditional” gender gap has given way to a “modern” gender gap, characterized by growing support among women for the left. In an interesting text, Janine Mossuz-Lavau (2007) outlines three stages of political change among French women since obtaining the right to vote in 1944 – learning (l’apprentissage), implementation (le décollage), and autonomy (l’autonomie) – during which women become progressively more free from their loyalty to the right. In light of the data contained in this chapter, one can conclude that this evolution still remains incomplete. The evolution of the impact of the variables of age and education on presidential vote reveals some interesting dynamics, perhaps changing the profile of French electoral clienteles in a lasting manner. In this regard, the most clear-cut effects seem to have touched the important contingents of the moderate left and the moderate right. The electorate that supported Ségolène Royal in 2007 is older and more educated than that which supported François Mitterrand in 1988 (although it is younger than for Jospin in 1995 and 2002). The average Jacques Chirac supporter was proportionally more educated in 1988, while the inverse was true for those who supported Nicolas Sarkozy 20 years later. Education has long served as a line of demarcation between leftwing voters, who were less educated, and right-wing voters, who were more educated. Now that imbalance has shifted and changed. Today in the second round of presidential elections it is the right-wing electorate that possesses a lower level of education. In fact, education now not only separates left-wing voters from right-wing voters, but

9780230_300088_03_ch02.indd 39

6/28/2011 8:21:06 PM

40 French Presidential Elections

also separates voters within the left, with the more educated voting more for moderate candidates in both cases. One of the most surprising results coming out of this chapter is the reversing of the effect of age in the 2007 election. The traditional draw of older voters to centrist candidates such as François Bayrou has given way to a marked infatuation among youth for the same. In this same election, religious sentiment, common to the two electorates of the classic right, was not tied to the support given to Bayrou. It is still too early to say if the election of 2007 will be as defining for the right as the 1995 election was for the left – an election during which support for the left, in particular that of the Socialist Party, lost significantly (others would say definitively) its character of a vote populaire (popular choice). The future will tell if henceforth it will be possible to clearly distinguish between two electorates among the classic right in France.

9780230_300088_03_ch02.indd 40

6/28/2011 8:21:06 PM

3 Class and Patrimony

For Karl Marx, France was a prime testing ground for his theories of class struggle, in particular the experiences of the 1848 revolution. (See especially his Class Struggle in France, 1848 to 1850; various editions dating from 1850.) As Theodore Zeldin (1973, p. 477), a leading British historian of France, remarked, “The importance of 1848 is that the masses were drawn into politics, and won over to the movement” of republicanism and electoral politics. The class stakes waged in the 1936 Popular Front elections were the heaviest to that time, with abounding talk of a “popular vote” and “working class” victory. Seymour Martin Lipset (1960, pp. 223–4), noteworthy political sociologist of the twentieth century, summed up the dominant modern hypothesis on social class and voting: “The most important single fact about political party support is that … the lower income groups vote mainly for the parties of the Left, while the higher income groups vote mainly for the parties of the Right.” This chapter concerns itself with the impact of such class-related variables on French presidential elections over the past two decades. Do income, occupation, and employment sector (private vs. public) have significant influences on vote choice? This inquiry rests on a dual comparison. First, we will see if these variables indeed have less influence today than they have had in the past. Then, we will examine whether these forces continue to play a larger role in France as compared to other established democracies, such as Great Britain and the United States. Existing literature, both in France and abroad, has already provided some tentative answers to these questions. Electoral analysis carried out in the United States (Abramson et al., 41

9780230_300088_04_ch03.indd 41

6/28/2011 8:16:43 PM

42

French Presidential Elections

2003; Flanagan and Zingale, 2006), Canada (Blais et al., 2002), the United Kingdom (Clarke et al., 2004; Evans and Norris, 1997), and Europe in general (Franklin et al., 1992) has generally concluded that people’s socioeconomic status (or SES) is exercising less and less influence on their electoral behavior. In France itself, class has been held to have a major impact on vote choice in the Fifth Republic. As experts on French electoral behavior Daniel Boy and Nonna Mayer have observed, social class constitutes one of the “heavy” variables (les variables lourdes) that “continue to sketch the contours of the French electoral landscape” (1993, p. 174). (In support of social class as a “heavy variable,” see also Boy and Mayer, 1997; Cautrès, 2004; Mayer and Perrineau, 1992.) How “heavy” is it now? The contrast is striking between the attention paid to these variables in early books, such as France de gauche, vote à droite (Capdevielle et al., 1981), compared to contemporary works, such as Cautrès and Muxel (2009) on the 2007 election. Capdevielle and his coauthors (1981) attach much importance to the influence of long-term sociological factors on elections, including the developing concept of patrimony (effet patrimoine). However, Cautrès and Muxel (2009), writing 30 years later, place more emphasis on short-term factors and the mobility of vote choice. This slackening interest on the part of French scholars with respect to sociological variables has been linked to a gradual decline in the strength of these variables for French electoral behavior (Cautrès and Heath, 1996). Studies published under the direction of Pascal Perrineau (2007) in Atlas electoral: qui vote quoi, où, comment? are revealing. François Gougou has drawn conclusions over the “splitting up of the popular vote” and speaks about the “end of the working-class vote,” whose scattering appears to have especially benefited Jean-Marie Le Pen (Platone, idem, p. 47; Perrineau, idem, p. 65). Elisabeth Dupoirier has taken note of an ebb in the socialist vote among the working middle class (idem, p. 90) which, when combined with its concomitant decline among the working class, points to the gentrification (embourgoisement) (Rey, idem, p. 51) of the socialist electorate. The demise of the traditional sociological profile of the left-wing voter implies that the right-wing voter is becoming more “popular” and more heterogeneous. The unity of the camp conservateur (Lancelot, 1981, p. 10) led Haegel and Sauger (idem, p. 58) to conclude that “from 1981 to 2002, it was difficult to distinguish the

9780230_300088_04_ch03.indd 42

6/28/2011 8:16:43 PM

Class and Patrimony

43

two electorates of the moderate right.” Today, that unity seems to have fallen apart in light of certain clear differences between supporters of Nicolas Sarkozy and those who have succumbed to the tentation bayrouiste (Bayrou temptation) in 2007 (Strudel, 2009). Are these seemingly authoritative conclusions unquestionable, and do they explain the loss of interest in studying the effect of heavy variables on the French electorate? Three arguments go against these conclusions. The results of the last chapter, notably those regarding the persistent effect of religion and the changing effect of age and education on the presidential vote, bring to our attention the enduring importance of socio-demographic variables on vote choice. Two recent American studies, notably those of Stonecash (2000; see also Nadeau et al., 2004) on income and those of Bartels (2008) on inequality of wealth distribution, have brought into question the belief that SES factors are becoming less influential for voting. They emphasize the possibility that the impact a person’s material situation has on his or her political behavior may have been underestimated. These works call out to what remains one of the most fruitful innovations in French political science: the effect of patrimony. Examination of this question is virtually absent in electoral studies outside of France. The patrimonial idea supposes that one’s material situation is not summarized just by job or income; rather, the possession of financial assets (such as a house or investments) is a reflection of one’s attitude toward risk, which can have a clear influence on vote choice. However, in spite of its innovative character, this idea has received very limited attention in the past few years. This neglect does not seem to be justified, considering the important continued impact this variable has had on legislative elections (Nadeau et al., 2011) and French presidential elections (Foucault et al., 2011; Nadeau et al., 2010; see Lewis-Beck and Nadeau, 2011, for similar conclusions in the American case) over the past few decades. The aforementioned suggests that a systematic examination of class-related determinants to voting still has a place in the French context. We will first approach this question by focusing on the effect of one’s occupation and economic sector on French presidential election choice from 1988 to 2007. Next, we will examine the impact of two direct measurements of one’s material situation: income and financial assets. Finally, we draw some general conclusions from the results.

9780230_300088_04_ch03.indd 43

6/28/2011 8:16:43 PM

44

French Presidential Elections

3.1 Occupation, economic sector, and the vote Social class has been measured in different ways, but the primary determinant remains occupation. This emphasis on occupation stems from historic nineteenth-century roots, when during the Industrial Revolution Marx emphasized the class struggle between those who own the means of production (the bourgeoisie) and those who do not (the proletariat). In modern times, this struggle was often reduced to one between capital and labor. However, Marx himself believed that this simple dichotomy was inappropriate, at least in the earlier stages of capitalist evolution. Besides factory owners, another occupation group in particular that he identified as potentially having interests different from the workers were peasants. Well into the twentieth century, peasants (or farmers) composed a large portion of the electorate in Western democracies. In France, peasants have had a pivotal role in the political development of the nation. Commenting in the 1970s, historian Theodore Zeldin (1977, p. 45) declared that “for the past two centuries, progress for the French has meant eliminating the peasants.” Size is a major reason the French peasantry has been so crucial electorally. After World War II, 36 percent of the work force was agricultural, a figure much higher than for most other advanced Western nations. Even by 1977, the figure was still relatively high, at 9 percent, a number surpassed in Western Europe only by Italy. (For a review of these statistics, see Lewis-Beck (1981).) Thus, at least traditionally, peasants constitute a formidable voting block. Moreover, they have been viewed as strategic actors in the political economic system of the nation. President Giscard d’Estaing underlined this vital international role in a key speech on trade policy, “L’agriculture doit être notre pétrole.” (Le Figaro, December 17–18, 1977, p. 2). For our purposes, the question is how this farm force votes, compared to other occupations. There are opposing hypotheses. The dominant view, classically, comes out of the electoral geography perspective, contending that peasants have no distinctive political preferences left or right (Dupeux, 1959, pp. 159–62; Goguel, 1966, p. 242; Kaltzmann, 1958, pp. 39–40). A few scholars challenge this view, noting that peasants are notoriously susceptible to the appeals of Communism (Ehrmann, 1952, pp. 39–42; Fauvet, 1955, pp. 175–7). Another body of researchers, however, affirm that the peasantry

9780230_300088_04_ch03.indd 44

6/28/2011 8:16:43 PM

Class and Patrimony

45

stands uniquely on the right (Aron, 1955, pp. 246–348; Duverger, 1955, p. 26; Lewis-Beck, 1977, p. 450). This last hypothesis receives strong support in an examination of vote intentions from nine legislative elections of the Fourth and Fifth Republics (1946 to 1978). Lewis-Beck (1981, p. 528) shows that, according to these national election surveys, peasant support for left parties always falls behind the voting population in general, and the workers in particular. Indeed, for this postwar period, left vote support from workers exceeds left vote support from peasants by, on average, 25 percentage points. Data from 1978, the last election in the series, illustrate well the conservatism of the farm voter. In a 1978 legislative election survey (N = 7643), four occupation groups were identified, along with their support for parties on the left (communists + socialists and left radicals + splinter left), as follows: professionals and business (36 percent); clerks and bureaucrats (55 percent); workers (72 percent); peasants (38 percent). Indeed, peasant support for the ruling majorité (RPR, UDF, splinter right), at 61 percent, exceeded even that of the professional and business group, at 58 percent. Clearly, peasants are not allies of the workers. On the contrary, they appear to be allied with the bourgeoisie. In a more refined analysis of the occupational strata, Lewis-Beck (1981, p. 530) reports that, for these above nine elections, peasant vote support for the right “always exceeds that of the lower-level and the middle-level whitecollar occupations, e.g., clerks and fonctionnaires, and sometimes (1962, 1973, 1978) actually surpasses that of the upper-level whitecollar occupations, e.g., professionals and business executives.” Such behavior led Duverger (1960, p. xxii) to declare peasants a “supporting class” (classe de soutien) of the bourgeoisie. Why do farmers behave this way when, in objective terms such as amount of education, income, and manual labor, they might reasonably be expected to join with the workers? The answer lies with more careful consideration of Marxist theory (Tavernier, 1975–6, pp. 34–33). Most peasants (88 percent in 1970) are owners of land they farm (Lewis-Beck, 1981, p. 533). As owners of this means of production, their class identity is formed as owners, and with it their attachment to right-wing politics. (It is worth noting that this pattern in politics occurs also in other Western European nations, such as Germany, Italy, Belgium, and the Netherlands (Lewis-Beck, 1977).

9780230_300088_04_ch03.indd 45

6/28/2011 8:16:43 PM

46

French Presidential Elections

As fascinating as the peasant class distinction is, it has come to be increasingly irrelevant in contemporary times, as farm populations in France, and elsewhere, have tumbled. Indeed, it has now become virtually impossible to treat farmers as a separate occupational group in current surveys, because the numbers are too small for statistical reliability. For example, in assessing the 2000 and 2004 US presidential elections, Lewis-Beck et al. (2008, pp. 360–1) found very few farmers (e.g., only 13 in one table). They concluded that “agrarian political behavior, as an object of study in the NES, is a thing of the past, unless the group is systematically oversampled in future surveys” (Lewis-Beck et al., 2008, p. 361). A comparable data situation holds in France. Therefore, we refrain from breaking out the farm population in the surveys at hand. However, the above results still have value, making clear that, when categorizing occupations into the common general dichotomy, middle versus working class, farmers are more properly coded as middle class. (Furthermore, this conclusion sustains itself when the four surveys are experimentally pooled and the peasant vote is analyzed separately.1) For purposes of cross-national analysis, Robert Alford (1963) simplified the class structure into two occupational categories, middle and working, and the vote into two categories, right and left. Then, he created a class voting index, by subtracting the percentage of the middle class who voted left from the percentage of the working class who voted left. He applied this to election surveys in several Western nations and found that it ranged from a low of +16 in the United States to a high of +40 in Great Britain. What does the Alford Index look like for France? Bruno Cautrès (2004, p. 78) calculated it for French legislative (L) and presidential (P) election surveys (first round), with the following results: 1967 (L) = 20; 1978 (L) = 18; 1988 (P) = 21; 1995 (P) = 10; 1997 (L) = 14; 2002 (P) = 1; 2002 (L) = 4. By these measures, class voting in France appears rather modest, both absolutely and comparatively, and to be perhaps in decline. Of course, that picture comes in part from how occupations are grouped into class categories. Take the same elections, but look at the left vote differences between two class extremes – the left vote percentage of the highly salaried subtracted from the unskilled workers: 1967 (L) = 28; 1978 (L) = 32; 1988 (P) = 26; 1995 (P) = 5; 1997 (L) = 7; 2002 (P) = –3; 2002 (L) = 14 (Cautrès, 2004, p. 79). We see that there is more class voting between the extremes, as would

9780230_300088_04_ch03.indd 46

6/28/2011 8:16:44 PM

Class and Patrimony

47

be expected. Still, even here, it exceeds 30 only in one election, 1978. Further, the downward trend in class voting, measured this way, remains. Another way to consider the class-vote relationship is to look at the simple correlation between the two variables. With respect to contemporary presidential elections, which is our special interest, Boy and Mayer (1993) report on survey results from the first round of the 1988 presidential contest. They find that working-class attributes are associated at 0.24 (gamma) with the vote for Mitterrand (Boy and Mayer, 1993, Table 1, p. 175). Lewis-Beck (1998, p. 4) finds a comparable result from 1995 while looking at the second round of presidential voting. Relating the variables as two dichotomies – class (worker vs. middle) and vote (Jospin vs. Chirac) – he calculates an Alford Index of 16 (i.e., 61 percent of workers vote Jospin – 45 percent of middle class votes Jospin = 16). This translates into a correlation (Pearson’s r) of 0.14. (The relationship does not gain in strength if a subjective measure of class, based on self-placement on a 1 to 10 scale, is used. For the same contest, that subjective class-vote correlation is r = 0.08 (Lewis-Beck, 1998, p. 5). Clearly, social class, measured in classic blue-collar/white-collar terms, does not yield much impact on the vote in current times. Perhaps the relationship strengthens if other measures are used. A newer approach looks at the nature of the work itself, in terms of whether it is in the public or private sector of the economy. The idea is that those who are more dependent on the state for their livelihood are more likely to vote for the left. We do see that this seems to be the case, at least for 1995. At the one extreme, public employees voted 61 percent in favor of Jospin in the second round, in contrast to those who were private employers, at only 21 percent (Lewis-Beck, 1998, p. 6). Suppose the private-public nature of the job is ordered as follows: private employer, self-employed, private employee, civil servant, public employee. Relating this variable to second-round vote choice, Chirac-Jospin, yields an r = 0.21 (LewisBeck, 1998, p. 6). For the 1995 contest, this is the class-vote association of greatest magnitude. It suggests the continued relevance of class, especially when measured in terms of the sectorial origins of work, public versus private. Below, we combine these different approaches to measuring social class in France, in order to more fully explore its contemporary impact.

9780230_300088_04_ch03.indd 47

6/28/2011 8:16:44 PM

48

French Presidential Elections

3.2 Occupation, economic sector, and the vote: contemporary bivariate results Any analysis of the impact of socioeconomic variables on voting should take into account the changing relative weight of different groups that make up the electorate. In the previous chapter, we have seen that France has experienced rather marked transformations over the past few decades. Those who elected Nicolas Sarkozy were older, more educated, and less religious than were those who brought François Mitterrand into power. Over time, France has become more diverse, and more attention should be paid to this phenomenon over the next few years (see Tiberj, 2007). The preceding remarks apply a fortiori to the transformation of the French labor market. The political import of some of these changes – such as rising job insecurity and the rise of self-employment – has yet to be seen. The transformation of the French economy to one that is based on tertiary industry, along with the rise of the laboring middle class, to the detriment of small, independent businesses, remains a dominant feature of this transformation, offering an important benchmark for more fully evaluating the role of one’s occupation in his or her political choices. Specifically, two variables are examined to study the contemporary link between occupation and presidential vote from 1988 to 2007. The first is the respondent’s present occupation (or past, in the case of retirement). The second is the place of that occupation in the economic sector, be it private or public. Previous work on electoral behavior in France amply justifies the choice of these variables (see preceding section); the opposition between the working-class vote and the conservative vote of executives, professionals, and small entrepreneurs (including farmers) has long structured political choices in France. The fight for suffrage among the emerging middle classes has gained political importance over time. Furthermore, around 30 percent of voters in France are part of the public sector, and the support of these civil servants has long composed an electoral stronghold for the left. The data in Table 3.1 offer a first glimpse at the relationship between respondents’ occupation and their vote choice in the presidential elections from 1988 to 2007. The trends observed are generally the same for the first and second rounds. While the right

9780230_300088_04_ch03.indd 48

6/28/2011 8:16:44 PM

9780230_300088_04_ch03.indd 49

6/28/2011 8:16:44 PM

1st round

Table 3.1

N

885

11

18

.25 (.03) .18 (.02)

Gamma

Kendall’s tau-b

.00

668

Extreme right

Pearson’s χ2

11

Right

16

45

.20

33

Center

10

White-collar employees

Cramer’s V

31 19

Left

6

Managers

Extreme left

Blue-collar workers 707

13

11

9

51

15

Managers .11 (.02)

.15 (.03)

581

12

35

23

25

5

.00

.19

773

17

24

17

28

14

White-collar employees Blue-collar workers 637

19

18

10

34

19

Managers .10 (.03)

.15 (.05)

278

13

31

18

25

14

2002

.00

.18

220

14

29

8

21

29

White-collar employees

1995 Blue-collar workers 332

18

20

6

26

30

.10 (.02)

.14 (.03)

580

4

48

21

21

5

2007

697

10

28

18

30

15

Continued

.00

.14

904

7

34

20

28

10

White-collar employees

1988

Managers

Occupation and vote in French presidential elections (1988, 1995, 2002, and 2007, first and second rounds)

Blue-collar workers

9780230_300088_04_ch03.indd 50

6/28/2011 8:16:44 PM

Continued

74

Blue-collar workers 742

26

31

Managers .22 (.02)

.38 (.03)

658

69

47

.00

.24

821

53

61

Blue-collar workers 659

39

41

Managers .15 (.02)

.27 (.04)

795

59

.00

.16

529

56

44

Blue-collar workers 880

41

59

2007

.10 (.02)

.17 (.03)

599

64

36

Managers

47

.00

.11

906

53

725

56

44

Source: CEVIPOF (1988, 1995, 2002, 2007).

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b.

.23 (.02)

Kendall’s tau-b

.00 .40 (.03)

Gamma

Pearson’s χ2

954 .25

704

N

36

64

White-collar employees

2002 White-collar employees

1995 White-collar employees

1988 White-collar employees

Cramer’s V

56

44

Managers

Right

2nd round Left

Table 3.1

Blue-collar workers

Class and Patrimony

51

(including the center) has maintained and even somewhat reaffirmed its support among its natural base – managers, entrepreneurs, and private-sector employees – the left has lagged among its natural base – notably blue-collar workers. A comparison of the evolution of the vote within the first round of presidential elections is especially revealing. In 1988, right-wing candidates (Barre, Chirac, and Le Pen) earned the most support among the manager category (with 63 percent of the vote), followed by the white-collar employees (with 45 percent). Clearly distinguished from this bourgeois group are blue-collar workers (of whom only 33 percent voted for the right). By 2007, support for the right consolidated itself among the managers (a gain of 10 points with support at 73 percent), but also increased quite sensibly among employees and workers (to 61 percent and 56 percent, respectively). The evolution of the blue-collar working-class vote is particularly remarkable. For example, in the second round of the 1988 elections, Jacques Chirac garnered 26 percent of the working-class vote, while Nicolas Sarkozy received 56 percent of this same group’s support 20 years later. The measures of association across these four elections conveniently summarize the declining impact of class, as measured by occupation. For example, here are the Cramer’s V (second-round): 1988, 0.25; 1995, 0.24; 2002, 0.16; 2007, 0.11. In Table 3.2, we turn to another relevant occupational situation: whether the respondent works in the public or private sector, and its relationship to presidential vote. Observe, first, that public-sector employees are more likely to vote left than are private-sector employees. Clearly, working for the state engenders some attachment to government benefits, and their personal protection, if not expansion. Also, observe that this relationship is fairly stable across the period. For these four elections (first-round), the association has held modest, within a very narrow range (Cramer’s V goes from 0.13 to 0.15). Examination of second-round results reveals, within this stability, an interesting internal shift. Beginning with 1988, a minority of private-sector employees voted right in the second round; that share became a majority in 1995 and in subsequent second-round elections. Obviously, the right-left divide has changed over the period, with ground ceded to the right. These preliminary results, which must be confirmed through multivariate analysis (see below), are clear. The right seems to have done

9780230_300088_04_ch03.indd 51

6/28/2011 8:16:44 PM

9780230_300088_04_ch03.indd 52

6/28/2011 8:16:44 PM

Left Right N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

2nd round

58 42 1977 .10 .00 .22 (.04) .10 (.02)

9 40 16 23 12 1864 .13 .00 .19 (.03) .11 (.02) 69 31 964

15 47 15 13 11 906

Public

26 74 1976 .16 .00 .34 (.04) .16 (.02)

4 21 24 37 13 1835 .15 .00 .18 (.03) .11 (.02)

Private

61 39 883

15 40 13 18 13 823

Public

1995

48 52 1796 .13 .00 .26 (.04) .13 (.02)

20 24 11 28 16 655 .15 .00 .20 (.04) .12 (.03)

Private

61 39 1169

27 33 7 21 12 479

Public

2002

44 56 1898 .10 .00 .21 (.04) .10 (.02)

8 26 22 37 6 1833 .13 .00 .16 (.03) .09 (.02)

Private

54 46 988

12 34 21 26 7 985

Public

2007

Source: CEVIPOF (1988, 1995, 2002, 2007).

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b.

Extreme left Left Center Right Extreme right N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

Private

1988

Employment sector and vote in French presidential elections (1988, 1995, 2002, and 2007, first and second

1st round

Table 3.2 rounds)

Class and Patrimony

53

a better job than has the left at maintaining its support among its traditional base. This change, noticeable since 1995, has contributed to assuring the right’s hold on the Élysée over the past 15 years. Further, it has helped soften the distinctions between the electoral clienteles of political parties on the left and right.

3.3

Income, patrimony, and the vote

The variable traditionally used in electoral studies to measure the material comfort of individuals is income (personal or household). This measure provides precious, yet incomplete, information regarding both the absolute and relative wealth of the electorate. Additionally, the effect of income on vote choice seems to be in decline in many countries (see Lewis-Beck and Nadeau, 2011, for the American case). The seemingly mild impact of income on voting has led political scientists to pursue two different approaches. The first consists of simply excluding income from explanatory models of voting (see Blais et al., 2002, for the Canadian case and Clarke et al., 2004, for the UK case). The second option is to explore other measures of a person’s economic situation. (For an example, see the work of Bartels (2008) on the political impact of unequal distribution of income in the United States). One of the most fruitful attempts of the latter approach has come from French political science. In a pioneering study carried out 30 years ago, Capdevielle et al. (1981, chp. 4) brought to light an effet patrimoine, which helped explain French electoral behavior during the legislative elections of 1978. The general conclusion of these authors is that ownership of tangible assets is becoming a significant determinant for whether or not voters support a right-wing party. The contributions of this study are evident on multiple levels. On a theoretical level, the authors of this study highlight the importance of keeping in mind both income and assets, in order to adequately understand the material situation of households. With regard to political analysis, the results show that the possessions of a diversified patrimony exercise an independent impact on voting beyond the “heavy variables” – occupation, income, religion, and political ideology (Boy and Mayer, 1997; Cautrès 2004; Cautrès and Muxel, 2009; Mayer and Perrineau, 1992). Finally, this innovation is particularly striking because of its predictive aspect, which gives it the character of a scientific

9780230_300088_04_ch03.indd 53

6/28/2011 8:16:44 PM

54

French Presidential Elections

breakthrough. The size and diversity of households’ asset portfolios has quickly grown in developed economies over the past few decades (see Piketty, 2003; Piketty and Saez, 2003). The evaluation of such indicators and their effects on political behavior is therefore more relevant than it has ever been (see Lewis-Beck and Nadeau, 2011, for a similar argument in the case of the United States). In spite of its weight, study of the effect of patrimony has not incited much recent interest. While the 1978 French electoral study included nine indicators of patrimony, the 2007 study only included four, of which only one – holding of assets – allows us to measure the possession of high-risk assets. This decline in scholarly interest in the effect of patrimony is even more paradoxical as the wealth of French families increases and becomes more complex. However, contemporary work (see Foucault et al., 2011; Nadeau et al., 2010a, 2010b) concludes that the effect of patrimony still constitutes an important dimension of electoral behavior in France today. These studies make use of a battery of six questions in CEVIPOF’s national election studies that inquire about financial assets. These questions ask respondents if they or a member of their family owns (1) their home or apartment, (2) a country house, (3) a savings account, (4) a business, a piece of land, or a farm, (5) a rental property, or (6) stocks. Nadeau and his coauthors (2010a, 2010b) proposed that these assets can be regrouped into two general categories based on the relationship between the anticipated benefit and risk incurred by the investors. “Low-risk” assets, referring to a home or apartment, a country house, or a savings account, require less vigilance on the part of the holder and entail fewer risks, but provide fewer benefits. On the other hand, possessing “risky” assets, such as a business, stocks, or some sort of rental property, requires more attention from the investor, as this investment strategy leaves them vulnerable to more risks, but also open to larger rewards. The work of Nadeau and his colleagues is based on the relationship between an attitude – being risk-averse – and a series of behaviors, notably the propensity to hold risky assets; it rests on a double validation of these scales. Table 3.3 allows us to examine the distribution of assets that the French possessed in 2002, the last year for which these three components of both categories of assets have been measured (the possession of three types of low-risk assets was measured in 2007; however, only one question in that investigation, that of the possession of stocks,

9780230_300088_04_ch03.indd 54

6/28/2011 8:16:44 PM

Class and Patrimony

Table 3.3

55

The distribution of patrimonial assets in France, 2002

Number of elements 0 1 2 3

low-risk assets

High-risk assets

17 39 37 6

68 23 7 2

Source: CEVIPOF (2002).

is related to risky assets). As one can see, the holding of a non-risky asset is more widespread, since the French are divided into two relatively equal groups in this chapter: those who possess no more than one type of asset (56 percent) and those who possess more than one. The holding of risky assets, while on the rise (the percentage of respondents in CEVIPOF studies that indicated they possessed such assets grew from 7 percent in 1978 to 26 percent in 2002) is still quite limited. In 2002, more than two out of three respondents (68 percent) reported possession of none of these elements of patrimony. The appropriate question now is how this assets variable relates to presidential vote choice. In Table 3.4, the entire four-category variable is related to vote choice, in all four elections. Begin by looking at the data from the most recent election, 2007; the pattern is clear. As riskiness of assets increases, support for the traditional right steadily increases (from 29 to 46 percent). Similarly, support for the traditional left steadily decreases (from 32 to 21 percent). Propensity for risky investment brings, it seems, a tendency to vote pro-right, antileft. This phenomenon is seen most sharply when the second-round votes are examined, across all four contests. In each case, right-wing support is markedly more at the high-risk end of the spectrum, compared to the low-risk end. A key issue is whether this apparent patrimonial effect sustains itself, once multivariate controls are applied. We address this concern below.

3.4 A multivariate analysis of the impact of socioeconomic variables on voting: cumulative results The voting models that will now be examined will combine the variables from the previous chapter – age, gender, education, and religion – with the socioeconomic variables that we have discussed in

9780230_300088_04_ch03.indd 55

6/28/2011 8:16:45 PM

9780230_300088_04_ch03.indd 56

6/28/2011 8:16:45 PM

Kendall’s tau-b

.23 (.02)

.41 (.03)

.00

Pearson’s χ2

Gamma

33 1440 .26

45 13 17 12 1350 .16 .00 .28 (.02) .20 (.02) 77

13

Low-risk (0, 1)

Right N Cramer’s V

Left Center Right Extreme right N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b Left

Extreme left

Low-risk (2, 3) 15

11

High-risk (0) High-risk (1+) 71

46

45 29 54 1813 1807 1450

55

39 49 42 18 11 15 23 14 20 11 11 12 1712 1690 1381

8

Low-risk (0, 1) 11

High-risk (0) 16

10

25

High-risk (1+) Low-risk (0, 1)

2002

21

26

.17 (.02)

.28 (.03)

.00

48 1460 .19

55 44 62 1746 1795 1394

.11 (.02)

.18 (.03)

.00

46 1594 .15

42

25 15 31 13 381

16

48 42 58 1336 1957 954

30 32 34 26 28 28 29 13 20 13 22 8 12 8 24 26 20 30 25 25 22 17 12 16 12 14 13 14 1366 1606 1701 1255 589 531 737 .12 .11 .00 .00 .09 (.02) .11 (.03) .07 (.02) .08 (.02) 52 45 56 38 54 52 58

15

Low-risk (2, 3)

1995 Low-risk (2, 3) High-risk (0) High-risk (1+)

1988 Low-risk (0, 1) 8

11

5

.13 (.02)

.21 (.03)

.00

47 1874 .14

56 48 62 1559 2683 750

32 27 32 21 21 23 21 23 29 38 30 46 6 5 6 4 1832 1509 2608 733 .11 .00 .15 (.02) .11 (.01) 53 44 52 38

12

2007 Low-risk (2, 3)

Patrimony and vote in French presidential elections (1988, 1995, 2002, and 2007, first and second rounds)

High-risk (0)

Source: CEVIPOF (1988, 1995, 2002, 2007).

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b.

2nd round

1st round

Table 3.4

High-risk (1+)

Class and Patrimony

57

this chapter. These include occupation (managers, white-collar employees, blue-collar workers), economic sector (private or public), and the possession of risky or non-risky assets (from zero to three elements). This model can be represented in a summary way as follows: Vote = f (age, gender, education, religion, occupation, economic sector, assets). The complete results for the logistic regression analysis are presented in the Appendix (column 2). Remember that these results come from models in which the explanatory variable takes the value of 1 in the case of a vote for a candidate who belongs to an ideological family (extreme left) or for a candidate associated with such (for example, François Mitterrand for the moderate left) and 0 otherwise (see the previous chapter for more details). In the second round, this variable takes the value of 1 when the respondent supports a right-wing candidate (be it Chirac in 1988, 1995, or 2002 or Sarkozy in 2007). In all cases, the respondents who refused to respond, either expressing their indecision or showing their intention to abstain, are excluded from the analysis. The data in Table 3.5 offer a first look into the full results. The table shows the statistical performance of the complete model in terms of goodness-of-fit, with McFadden’s pseudo-R-squared, whose value varies between 0 (no explanatory power) and 1 (perfect explanatory power). This highly useful table allows us to examine the global performance of the primary candidates from the five large political families (extreme left, left, center, right, and extreme right) in France during both rounds of the four elections under study. Two main findings arise from this table. The first, and perhaps most important, is that the total explanatory power of socio-demographic variables in the model (and notably socioeconomic variables) is still quite limited. The mean pseudo-R-squared value for the candidates in the first round of voting is only 0.06. The second finding is that the model’s performance, while not spectacular, is better in the second round; the mean pseudo-R-squared reaches 0.10, a level comparable with that observed by Clarke and his collaborators (2005, table 3, p. 98) during the 2001 general election in Great Britain.

9780230_300088_04_ch03.indd 57

6/28/2011 8:16:45 PM

58 French Presidential Elections

Table 3.5 The variation explained by the different models (McFadden pseudo-R-squared) 1st round (principal candidates)

1988

1995

2002

2007 Average

(1) Age, gender, education, religion

.04

.04

.03

.04

.04

(2) + Occupation, sector, income, patrimony

.09

.07

.06

.06

.07

(1) Age, gender, education, religion

.01

.02

.02

.02

.02

(2) + Occupation, sector, income, patrimony

.03

.04

.04

.04

.04

(1) Age, gender, education, religion

.04

.08

.06

.02

.05

(2) + Occupation, sector, income, patrimony

.07

.10

.08

.02

.07

(1) Age, gender, education, religion

.06

.02

.06

.06

.05

(2) + Occupation, sector, income, patrimony

.10

.04

.06

.09

.07

(1) Age, gender, education, religion

.01

.04

.04

.05

.04

(2) + Occupation, sector, income, patrimony

.03

.06

.05

.07

.05

Extreme left

Left

Center

Right

Extreme right

2nd round (support for the rightist candidate) 1988

1995

2002

2007 Average

(1) Age, gender, education, religion

.07

.06

.05

.07

.06

(2) + Occupation, sector, income, patrimony

.12

.11

.08

.09

.10

Right

Source: CEVIPOF (1988, 1995, 2002, 2007).

9780230_300088_04_ch03.indd 58

6/28/2011 8:16:45 PM

Class and Patrimony

59

Another way of evaluating these results consists of examining the signs of the coefficients of the explanatory variables, then comparing their observed and expected values. In the present case, the analysis will be limited to the variables that touch upon the occupation of the respondents, their employment sector, their income, and the holding of a risky asset. For example, for a right-wing candidate, the expected sign for four of these variables is positive (manager, private-sector employment, income, and risky assets), with the expected sign being negative for the variable of blue-collar working class. In the case of a left-wing candidate, the inverse pattern is expected – that is, four negative signs and a positive sign for the variable of working class. The results of this exercise are presented in Table 3.6. For each election and each candidate, the table compiles the number of coefficients for which the observed sign corresponds to the expected sign (labeled ES), as well as the number of statistically significant coefficients (labeled SS). (All the statistically significant coefficients carry the expected sign, with three exceptions.) The data show that the five variables examined largely act in line with expectations. Roughly three coefficients out of four (74 out of 100; 5 candidates × 5 coefficients × 4 elections) have the expected sign during the first round. Table 3.6

The signs and significance of the model coefficients

1st round

Extreme left Left Center Right Extreme right Total

1988

1995

2002

2007

Total

ES 4 4 4 5 4 21

SS 3 3 3 3 1 13

ES 5 4 5 5 2 21

SS 3 3 3 3 1 13

ES 4 4 4 3 2 17

SS 3 3 2 0 0 8

ES 5 4 2 4 0 15

SS 3 3 1 4 1 12

ES 18 16 15 17 8 74

SS 12 12 9 10 3 46

ES 5

SS 4

ES 5

SS 5

ES 4

SS 4

ES 4

SS 4

ES

SS

18

17

2nd round

Right

Note: Entries are the number of regression coefficients that have the expected sign (ES) and statistical significance (SS). Source: CEVIPOF (1988, 1995, 2002, 2007).

9780230_300088_04_ch03.indd 59

6/28/2011 8:16:45 PM

60

French Presidential Elections

Furthermore, close to half of them (46 out of 100) are statistically significant. However, the results do vary from one candidate to another. For example, with extreme-left candidates, 18 out of 20 coefficients are of the expected sign, with 12 significant. At the other extreme of the political spectrum, the picture is very different. For the extreme right, perhaps because of the socially heterogeneous character of its supporters, only 8 coefficients out of 20 are of the expected sign. Furthermore, only one of them is statistically significant. (Two other coefficients are significant, but are of the opposite, unexpected sign.) The results observed between these two extremes conform more fully to expectations. The coefficients for moderate-left party supporters exhibit 16 expected sign coefficients (with 12 significant), making them slightly more in line with expectations than those for the extreme-left voters. The proximity of the centrist and moderateright electorates stand out, reflecting the long unity of the conservative camp. (The two voting groups have roughly the same number of coefficients of expected sign and significance: for the center, 15 and 9, respectively; for the moderate right, 17 and 10, respectively.) These results reveal some important aspects of the French political dynamic. The lack of continuity in the profile of the extreme-left and extreme-right electorates is particularly striking. While the results going from the extreme left to the moderate right are characterized by a logical progression in terms of the number of expected sign coefficients – higher for the extreme left (18) than for the left (16), higher for the moderate right (17) than for the center (15) and the left (16), and higher for the moderate right (17) when compared to the center (15) – the number of expected sign coefficients for the extreme right falls to 8. (These findings suggest that the exploitation of the immigration issue by the National Front has somewhat reconfigured the political game in France by facilitating the emergence of a group of voters whose location on the left-right spectrum cannot be solely explained in terms of conflicting views about the role of the state. The emergence of the National Front electorate does not necessarily reflect a decline of the left-right opposition but rather suggests the transformation of some of its foundations. This transformation may explain the diversity in Jean-Marie Le Pen’s supporters as well as a certain depolarization between the social status of voters and their ideological self-placement on the left-right scale).

9780230_300088_04_ch03.indd 60

6/28/2011 8:16:46 PM

Class and Patrimony

61

The peculiar position of the centrist electorate in France also stands out. The results for the elections of 1988, 1995, and 2002 show the similarity between the profiles of those who voted for the centrist “three Bs” (Barre, Balladur, and Bayrou) and those partisans of the rightist Jacques Chirac (13 expected sign coefficients in both cases). This similarity almost completely disappears in 2007. While the candidacy of Sarkozy reaffirmed the profile of right-wing partisans (four significant sign coefficients), that of Bayrou contributed, on the contrary, to imprint upon him a more “centrist” character (just two expected coefficient signs, of which only one was significant). The change in support for extreme right and centrist candidates seems to have contributed to the weakening of the profiles of the electoral clienteles. In the case of partisans of the National Front, this change is not new. While one can still find some traces of a right-wing clientele among the supporters of Jean-Marie Le Pen in 1988 (the working-class variable carries a negative sign and the variable of income carries a statistically significant positive sign), the five coefficients associated with his model in 2007 are opposite of what is expected, of which one (that of income) is statistically significant. The gradual rise in popular support for the extreme right, combined with the diversification of the centrist clientele in 2007, explains a good part of the decline of socioeconomic vote determinants in France. While 21 out of 25 coefficients carried the expected sign in 1988 (8 of 10 from the clienteles of Raymond Barre and Jean-Marie Le Pen), only 15 had the expected sign in 2007 (2 out of 10 for the supporters of François Bayrou and Jean-Marie Le Pen). This brings to light a significant softening of ideological polarization between social groups in France over the past few decades, notably due to the tossing (tassement) to the right side of an extreme-right and centrist clientele, as well as the gentrification of the socialist electorate. Still, it remains that this ideological depolarization of partisans is less marked in the second round, as the relative stability of expected statistically significant sign coefficients shows. (Over the four elections, the pattern of expected coefficients is as follows: 5, 5, 4, and 4; the pattern for statistical significance, respectively, is as follows: 4, 5, 4, and 4.) These results suggest that the electoral scattering of groups during the first round – the tentation bayrouiste, the ecologist vote, support for the extreme right, and so on – does not prevent voters from regrouping themselves into two large coalitions in the

9780230_300088_04_ch03.indd 61

6/28/2011 8:16:46 PM

62 French Presidential Elections

second round, coalitions that still exhibit the traditional elements of left- and right-wing electorates.

3.5 A multivariate analysis of the impact of socioeconomic variables on voting: detailed results Beyond these larger trends, what can we say about the particular effect of different variables on vote choice? The complete results allow a more nuanced examination. We will first investigate the variables of occupation and economic sector, before continuing on to the effects of income and assets. In each of these cases, our analysis will restrict itself to the statistically significant coefficients tied to these variables, since in fact it is for those that we have reasonable confidence regarding the effect. In addition, Table 3.7 presents the changes in voting probabilities associated with each of these variables’ effects on the vote, in both rounds of the presidential elections. The most interesting result relating to occupation is the diverging behavior of managers relative to blue-collar workers. In the 1988 elections, the “working-class variable” was dominant among all the occupational categories, as indicated by its associated coefficients during the first round (coefficients of 0.44, –0.59, and –0.62 for, respectively, Mitterrand, Barre, and Chirac) as well as the second round (coefficient of –0.67 for Chirac). The contrast between the impact of this “working-class” variable and the “manager” variable is particularly striking in the second round. While the latter is not significant (zero impact), the act of identifying oneself as “blue-collar worker” substantially reduces the probability of voting for Jacques Chirac (by 14 percentage points; see the previous section regarding the logic behind this calculation). The results for the 2007 election, which already reflect a noticeable change from 1995, are entirely different. Now it is the “manager” variable that dominates the first round (coefficients of –0.53 and 0.69 for Royal and Sarkozy). During the second round, a situation opposite to that of the 1988 election takes place. The “working-class” variable is no longer significant (zero impact), while being a manager significantly heightens the probability of voting for Nicolas Sarkozy (by 13 percentage points). The multivariate analysis shows the breakup of the working-class vote, most prominently with regard to socialist candidates. It also

9780230_300088_04_ch03.indd 62

6/28/2011 8:16:46 PM

Class and Patrimony

63

Table 3.7 Change in vote probability in French presidential elections, from the minimum to the maximum category, for the occupation, sector, income, and assets variables Managers

Bluecollar

Private sector

Income

High-risk assets

1988 Extreme left Left Center Right Extreme right 2nd round

–.03 .00 –.00 .03 –.01 .04

.02 .10 –.05 –.06 –.00 –.14

–.05 –.07 .01 .06 .01 .11

–.00 –.02 .10 .05 .10 .26

–.07 –.17 .11 .10 .04 .31

1995 Extreme left Left Center Right Extreme right 2nd round

–.08 –.03 .00 .08 –.03 .11

.04 .06 –.04 –.03 –.03 –.13

–.01 –.10 .04 .06 –.00 .14

–.08 .03 .05 .20 –.18 .15

–.02 –.12 .07 .05 .00 .24

2002 Extreme left Left Center Right Extreme right 2nd round

.00 –.04 .03 –.01 .01 .06

.07 .03 –.01 –.03 .01 –.07

–.06 –.04 .02 .03 .02 .09

–.07 .14 –.00 –.06 .00 –.03

–.11 –.10 .03 .07 –.00 .27

2007 Extreme left Left Center Right Extreme right

–.02 –.09 –.03 .15 –.00

.03 –.03 –.05 .04 .02

–.02 –.04 .02 .07 –.01

–.06 –.02 –.02 .21 –.03

–.01 –.06 –.00 .10 –.01

.13

.01

.07

.15

.07

2nd round

Note: Entries are changes in voting probabilities generated by Clarify (Tomz, Wittenberg, and King 2003) for a variation in each occupation and economic variable, when all other variables are held constant. Probability changes are based on regression model 2 (column 2 in the Appendix). Source: CEVIPOF (1988, 1995, 2002, 2007).

9780230_300088_04_ch03.indd 63

6/28/2011 8:16:46 PM

64

French Presidential Elections

shows the ability of the right to reaffirm support among its natural base (managers). With the right’s consolidation being of a smaller magnitude than that of the left’s demobilization, ideological polarization among social groups has become weaker, a tendency which is slightly reinforced by the growing support of workers for right-wing candidates. The change in strength of the “economic sector” (private vs. public) variable provides us with an overview of these changes. The relatively stable effect of this variable (12 statistically significant coefficients out of a possible 20 for the first round, 4 out of 4 for the second round) shows that there exist distinctions between left and right electorates in France. That being said, the relative weakening of its impact over time – belonging to the private sector raises the probability of voting for the right in the second round by 11 percentage points in 1988, but only by 7 percentage points in 2007 – also points to a certain depolarization of electoral behavior among social groups in France over the past few decades. What can be said about the effect of variables that more directly measure the material situation of respondents and their families, be it income or possession of both risky and non-risky financial assets? First, the possession of non-risky assets has no systematic effect on vote choice. None of the coefficients linked to this variable are statistically significant for the first round (and only one is significant for the second round). How can we explain these results? Perhaps by the ambiguous meaning attached to owning financial assets, notably owning a house and a savings account. These two cases of “popular” assets are accessible, and their acquisition reflects a concern over financial security; furthermore, owning one’s house can be seen as a form of social right. These characteristics are probably jumbled together with other connotations linked to the acquisition of assets that are tied to a strategy of asset accumulation and maintenance. Mixing up these distinct motivations can perhaps explain the fact that the possession of non-risky assets does not seem to push their holders toward any particular political orientation. The key explanatory variables here are income and possession of risky assets. Two findings are particularly interesting with regard to income. First is the reversal of the relationship between income and support for Jean-Marie Le Pen. While in 1988, those with a higher income voted for the far-right candidate in higher numbers

9780230_300088_04_ch03.indd 64

6/28/2011 8:16:47 PM

Class and Patrimony

65

(coefficient of 1.02), we observe the reverse trend in 1995 and later in 2007 (coefficients of –1.64 and –0.85). The other interesting finding is a reconfiguration of the effect of income on the ranks of both the left and the right. This is evident during the first round. For the left, income has clearly separated supporters of the extreme left (negative coefficients of –0.78 in 1995 and –0.92 in 2007) from supporters of the moderate left (see below), particularly since 1995. The same pattern is observed on the right, where the more wealthy individuals support the moderate right and center more than they do the extreme right (note the contrast between the coefficients for the right and extreme right; 1.24 versus –1.64 in 1995, and 0.97 versus –0.85 in 2007). This dynamic can explain the diminishing impact of income on vote choice and is even more visible in the second round (the coefficient for the variable of income has been nearly halved between 1988 and 2007, from 1.14 to 0.60). In this instance, the coalition of more affluent and less well-to-do voters on both sides of the ideological spectrum (the socialist voters with the extreme-left voters; the backers of right parties with supporters of the far right) contribute to dilute the impact of income on presidential vote choice. The other variable of interest is possession of risky assets. A comparison of these results with the 2007 data is more difficult because ownership of risky assets is measured with only one item (stocks). That being said, it is interesting to note that the variables corresponding to possessing the two types of assets are significant in the second round; further, the effect of possessing risky assets is significant in the first round of the voting model where the choice is Ségolène Royal or Nicolas Sarkozy. These results show that there is still a link between people’s material situation and their political preferences, but also that it is necessary to look for the cause not so much in a traditional variable such as income, but in a more novel indicator of a person’s wealth, such as assets. These results find solid confirmation in elections for which all the indicators of level of risky assets held are available (1988, 1995, and 2002). Five of the six coefficients in the models of left-wing candidates during the first round are statistically significant. Even more revealing is the fact that three coefficients for socialist candidates are significant (coefficients of –0.84, –0.66, and –0.76), while the coefficients corresponding to the income variable for these same candidates are either not significant or, as in the case of 2002, of the

9780230_300088_04_ch03.indd 65

6/28/2011 8:16:47 PM

66

French Presidential Elections

opposite sign of what was expected (a coefficient of 0.93 in this case signals a positive relationship between income and the probability of voting for Lionel Jospin). These results seem to suggest that the line of separation between moderate candidates from both the left and right is not income, but rather the possession of risky assets. The results of the second round for the 1988, 1995, and 2002 elections seem to confirm this. For these three elections for which we have sufficient measures of risky assets, this variable is one that has the most systematic and pronounced impact on vote choice (the coefficients of 1.29, 1.00, and 1.12 correspond to a higher probability of voting for Jacques Chirac (by 31, 24, and 27 percentage points). Among all SES variables examined in this chapter and the last, only the religion variable proves itself to be slightly more important than assets for structuring French electoral behavior. In light of these results, we must push ahead in our interest in studying the effet patrimoine.

3.6 Conclusion The study of class-related determinants of the vote (occupation, economic sector, income, and the possession of financial assets) within the context of French presidential elections over the past two decades has brought to light some important dynamics. First, the influence of these variables on French vote choice is declining (but not ended). These variables have slowly come down to the same level of influence as in other advanced democracies, such as the United States and the United Kingdom. Second, support for Jean-Marie Le Pen indicates, in part, a break with the notion that political choice in France is solely centered around the role of the state, to be expressed on a single left-right axis. From a certain perspective, the left-right vote in France deriving from the role of state intervention starts with the extreme left and ends with the moderate right. That is, while leftright ideology is still very much influential, as we will see in Chapter 4, certain issues such as immigration can directly account for much of the support going to the extreme right. The decrease in ideological polarization among social groups can be explained by two dynamics – one is older and can be seen in the popular support of the extreme right. The other is more recent and can be seen in the increased support for centrist candidate François

9780230_300088_04_ch03.indd 66

6/28/2011 8:16:47 PM

Class and Patrimony

67

Bayrou in 2007. These two dynamics have weakened the conservative camp and led to a gentrification of the socialist camp. The weakening of social cleavages for presidential vote choice mainly stems from the decline in the “popular” leftist vote. One of the most marked displays of this phenomenon is the fall in support among blue-collar workers for left-wing candidates, notably those from the Socialist Party (PS). This ebb in support among the left’s traditional supporters, combined with the consolidation of the rightwing vote in groups traditionally favorable to the right (executives, professionals, farmers, and private-sector employees), has allowed candidates from the right to diversify their support and to maintain control over the Élysée since 1995. The gains of the right among the traditional left camp have offset their losses among their own traditional electorate, while the opposite is true for the left. These patterns have contributed both to the depolarization of social groups and to the success of right-wing candidates in presidential elections since 1995. The results also show that social cleavages remain more salient than previous studies on electoral behavior in France and elsewhere would suggest. This persistence is especially apparent during the second round of presidential elections. The electoral scattering and diversions that are characteristic of the first round of voting give way to the reformation of two large coalitions, largely exhibiting the usual traits of a left-wing and a right-wing electorate, something that the French polity has shown for a long time. Finally, the results show the importance of painting a more accurate portrait of the voter’s material situation. Holding risky financial assets has an important impact on vote choice. In particular, it is a more important distinction between left- and right-wing voters (notably for those who support moderate candidates) than is income. In sum, the results of this chapter draw attention to the need for further study of class-related determinants of the vote, especially the effects of patrimony.

9780230_300088_04_ch03.indd 67

6/28/2011 8:16:47 PM

4 Left-Right Ideology

The notion of left versus right pervades the discussion of politics the world over. This use of the terms “left” and “right” began in France itself, during the Revolution. In the 1790s, the deputies of the national legislative assembly regularly grouped themselves physically from left to right across the space of the chamber’s half-circle of seats. These factions – the Jacobins, the Plaine, the Girondins – held different points of view about the proper course of revolutionary action. In the National Convention (1792–3), the pivotal issue was disposition of King Louis XVI (Lewis-Beck, Hildreth, and Spitzer, 1988). During the relevant votes – the six Appels Nominaux – the king’s fate was decided. Take for example the third Appel, addressing his punishment. A majority (361) were for death, a large minority (286) were for imprisonment, and the rest (102) were for some middle position (or did not vote). In this historic ballot on regicide, the deputies declared a policy stance, indicating their position on the issue by where they located themselves in the hall. Since that time, political positions in France have generally been articulated along a leftcenter-right continuum. Modern France, with a political party system and a voting public structured ideologically from left to right, perhaps began around 1870 with, in the words of historian Gordon Wright (1981, vi), the “rooting of the Republican System.” Early voting studies, pioneered by André Siegfried (1913) and continued by Goguel (1951), linked left-right voting patterns to specific geographic regions of France. Certain continuities were remarked upon, such as the persistent conservatism of Brittany and the enduring socialism of the Midi. As 68

9780230_300088_05_ch04.indd 68

6/28/2011 8:15:48 PM

Left-Right Ideology

69

noted in Chapter 1, electoral geography investigations remain a strong French tradition. Also, we observed that national cycles of partisan voting over time, rather than space, offer another way to appreciate the left-right dimension in French politics. Overall, substantial variation in left or right support reveals itself across national elections of the Fourth and Fifth Republics (see especially Bélanger and Lewis-Beck, 2010, as well as Chapter 1). The fortunes of the left, for example, are by no means constant. The stunning Socialist victory in 1981, to take the most dramatic illustration, implies the coming together of a set of issues very favorable to the traditional left. But this inference, made from aggregate results, really concerns individual behavior. Such an inference commits an ecological fallacy, as does any inference about individual voting behavior drawn from the aforementioned patterns of electoral geography. Valuable as examination of these different political aggregates can be, they do not reveal how left-right ideological thinking actually impinges on the voter’s decision-making process. To understand that, we have to study individual voters themselves, in a scientific way. For this purpose, we focus on the four presidential election surveys at hand.

4.1

The meaning of left-right ideology

In French politics, the terms “left” and “right” provide the foundational discourse. Candidates, party officials, journalists, TV commentators, and people in the street eagerly talk politics, and these ideologically charged words always come up. But what does it mean to be “on the left” or “on the right?” For one thing, it means you stand at a point in the political space that increases or decreases your distance from other voters. By saying you are a “leftist” (or a “rightist”), you declare your political views. Such self-labeling is more than a mere abstraction – as a psychological identification with a particular frame of reference, it has certain effects. Being aware of whether you are on the right or left puts you in a general relationship with political objects that will attract or repel, depending on the object. Election surveys in France routinely ask citizens questions of the following type: In political matters people talk of “the left” and “the right.” How would you place your views on this scale? (Respondent shown scale).

9780230_300088_05_ch04.indd 69

6/28/2011 8:15:48 PM

70 French Presidential Elections

If left-right discourse has meaning for most French voters, we would expect them to readily respond to such an item. This is in fact the case. For example, from 1964 to 1988, the polling firm SOFRES asked such a question 11 times. The response rate in that series ranged from 90 to 94 percent (Michelat, 1990). In the 1995 CEVIPOF study, when the modal, seven-point self-placement scale was used, over 98 percent succeeded in placing themselves (Lewis-Beck and Chlarson, 2002, p. 494). In other words, respondents seem to have no trouble understanding the question, then scaling themselves in their answers. What is the distribution of opinion on this scale? Is it lopsided – that is, overwhelmingly leftist or rightist? That would seem unlikely, given the aggregate pattern of left vote support already reported. What about a center? Is the distribution bimodal, with almost no one in the center, so making for a polarized electorate warring between two camps? Or, are there a substantial number of moderates, placing themselves neither left nor right, offering a more balanced electorate? According to a Fourth Republic survey from 1946 (the oldest we could locate), when respondents were asked to place themselves into one of three categories, 35 percent selected left, 39 percent selected center, and 26 percent selected right. We see that those on the left exceed those on the right, with the center the largest category. With such a large center, polarization from unmitigated left-right conflicts could be to some extent avoided. Indeed, this pattern, approximating a normal distribution, characterized ideological opinion in France from the 1960s to the 1980s (Lewis-Beck, 1984, p. 440). The picture is brought into sharp focus when contemporary data are examined. It appears that starting around the 1980s, the French electorate has become rather more polarized in its political thinking. What we see is a diminished center, with more electors taking on a true ideology – that is, left or right. This pattern is confirmed in Figure 4.1, which makes use of our four presidential survey data sets, 1988 to 2007. This figure sketches a polygon for a five-category ideology variable. For all four of these presidential contests, ideological opinion tends to a bipolar distribution. While French voters, then, appear generally responsive to the leftright notion, the idea itself remains an abstraction unless we can define its contents. In terms of the issues, what does it mean to be on the left or on the right? Note, first, that the terms definitely do have an issue content. That is, they are not simply empty phrases that citizens

9780230_300088_05_ch04.indd 70

6/28/2011 8:15:48 PM

Left-Right Ideology

71

50

In percentages

40 1988 1995 2002 2007

30

20

10

0 Extremeleft

Left

Center

Right

Extremeright

Score on left-right scale Figure 4.1 Left-right ideology distribution in the French electorate, 1988–2007 Notes: For 1988, 1995, and 2002, a seven-point scale recalibrated to five points was used: 1 = extreme left; 2–3 = moderate left; 4 = center; 5–6, = moderate right; 7 = extreme right. For 2007, an 11-point scale was also recalibrated to five points: 0–1 = extreme left; 2–4 = moderate left; 5 = center; 6–8 = moderate right; 9–10 = extreme right. Source: CEVIPOF (1988, 1995, 2002, 2007 (third wave of the French Electoral Panel)).

throw around, bereft of substance. Take as an example the 1995 French presidential election study, with its unusually large national sample (N = 4078) and unusually extensive battery of 13 issue items. Respondents expressed their opinions on the following policy topics: authority, Catholicism, competition, equality, feminism, Islam, nationalization, privatization, private schools, profit, reform, solidarity, and unions. On each, they scored their opinion from “very positive” (4) to “very negative” (1). For instance, someone who was very pro-union would give a score of 4, while someone very antiunion would give a score of 1. These item scores, added together, form an issue index (with a maximum score of 52 at the extreme left, a minimum 13 at the extreme right). We would expect, then, that the index would tend to predict ideology, if left-right self-placement takes its

9780230_300088_05_ch04.indd 71

6/28/2011 8:15:48 PM

72 French Presidential Elections

shape from issue preferences on political topics. Indeed it does. The issue index has a strong multiple correlation, R = 0.50, with ideology (Lewis-Beck and Chlarson, 2002, p. 496; see also Bélanger et al., 2006, p. 506). Of course, the relationship lacks perfection, as the data are self-reported. Nevertheless, we see clearly that knowing a citizen’s ideology tells us a lot about his or her issue preferences in different areas of policy. One question that needs addressing is the issue awareness of those in the center of the ideological spectrum. In a classic study, Deutsch, Lindon, and Weill (1966) argued that respondents picking the middle value of the left-right scale tended to know nothing about politics and hold no coherent political opinions. Conceptually, in other words, they composed a “swamp” (marais). Different studies suggest this notion of a politically “mindless middle” are exaggerated. Tiberj (2004, p. 273) concludes, from his examination of this issue with 1995 survey data, that “self-placement in the center seems much more motivated by an opposition toward, rather than by a lack of knowledge about, the Left and the Right” (our translation). In addition, those who place themselves in the center appear to express about the same level of political interest as other ideological groups. For example, in 1988, those declaring they were “not at all” interested in politics break down as follows: 28 percent, left; 39 percent, center; 32 percent, right (Ranger, 1990). Further, that “middle” placement on the ideology scale predicts vote intention as well as the “left” and “right” placements do. For instance, if the assumption of left-center-right ordinality imposes itself on ideological and vote intention, the two scales have been correlated (gamma) at 0.62; dropping that ordinality assumption for ideological self-placement yields a comparable correlation (eta) = 0.64 (Fleury and Lewis-Beck, 1993, p. 1102). This suggests that, for moderates, their ideological stance is electorally meaningful, as it is for leftists and rightists. In some sense, then, a voter’s issue score can be taken as a summary of preferences, or a “super-issue” measure (Inglehart, 1984, p. 37). However, this is not to say that the issue content of ideology is static, unchanging. Recall our introductory example, from the revolutionary assembly. The meaning of left-right in those days was largely defined by one’s attitude toward treatment of the king. A burning issue through most of the nineteenth century, the question

9780230_300088_05_ch04.indd 72

6/28/2011 8:15:48 PM

Left-Right Ideology

73

of monarchy is a dead letter today. Now left-right ideology is defined through other issues, some new – immigration, feminism – and some old – unions, private schools. The fact that the issue contents of leftright changes does not make left-right of less value as a political organizing principle in the French electorate. On the contrary, it demonstrates its resilience as a driving political signal to voters.

4.2

Ideological identification vs. party identification

In any democratic electorate, most voters are anchored to a political place. So anchored, they find it very hard to move away. Attitudes, feelings, and preferences distill into a general point of view they hold, ordering their political universe, telling them which candidates, policies, and parties to favor. That point of view takes the form of a psychological attachment enduring through time and experience. In the United States, that anchoring attachment is primarily to a political party (Campbell et al., 1960; Lewis-Beck et al., 2008). In France, by way of contrast, it is primarily to an ideology. In neither country, of course, does it mean the anchoring variable is exclusively one or the other. It is a question of balance and, in the French case, it tips overwhelmingly to ideology. This judgment, on the dominance of ideology over party, comes with some controversy. As Evans (2004, p. 54) observed, this question has composed “the major debate” in recent literature on French electoral behavior. Thus, our position needs a strong defense. There are three aspects to consider in resolving the issue: scope, structure, and strength. Scope refers to how many electors have each identification, structure refers to how the identifications influence each other, and strength refers to their relative impact on the vote. As shall be seen, when these different aspects are taken into account, in France ideology overwhelms party in importance. Much of the debate has been fueled by measurement questions. Two basic question formats have been employed to ascertain a respondent’s partisan identification: open-ended and closed-ended. Converse and Pierce (1986, p. 127), in their pioneering work, established the preferred approach for both: open-ended for party and closed-ended for ideology. Take party first. The question takes the following form: “What is the party you feel closest to?” Analyzing the responses to this item in the 1967 French National Election Survey, they came up with

9780230_300088_05_ch04.indd 73

6/28/2011 8:15:48 PM

74 French Presidential Elections

an estimate of 59.5 percent party identifiers. Now take ideology (Converse and Pierce, 1986, p. 75). Their question takes the form shown above, in which respondents are asked to place themselves on a left-right scale. Analyzing these responses, Converse and Pierce find 77 percent report an ideological identification. However, they challenge the magnitude of that estimate because the ideology of “middle of the road” may mean empty of contents, as discussed earlier (Converse and Pierce, 1986, pp. 127–9). The most thorough study, after Converse and Pierce (1986), is that of Lewis-Beck and Chlarson (2002), on the 1995 CEVIPOF survey. The item there reads, “Generally speaking, is there a party or political movement which you feel close to?” Taking all respondents, 56.3 percent gave a party or movement name. However, when these “names” were matched against real political groups, the percentage drops to 46.9 percent of the total. In contrast, with respect to ideology, 98.1 percent were able to place themselves on a left-right (seven-point) scale. These initial results suggest ideological identification has a much wider spread. But, as with certain party names offered, some of these ideological responses are not real. In particular, Lewis-Beck and Chlarson (2002, p. 496) demonstrate that about 29 percent of the respondents do not hold a meaningful ideology – that is, their ideology score does not predict their positions on issues. It seems reasonable, then, to label this group “non-ideologues.” Subtracting them from the sample yields an estimate of 69 percent ideological identifiers. Interestingly, this is close to the final number of Converse and Pierce (1986) when they conclude “about 65 per cent understand what these terms mean.” On the basis of this study, then, it seems fair to say that, conservatively, there are 47 percent party identifiers versus 69 percent ideology identifiers. In other words, there are about two-thirds as many party identifiers as ideology identifiers, giving the latter a clear edge with regard to scope in the electorate. Still, advantage in scope does not necessarily translate into advantage in structure. Among partisans, which variable is the prime mover: ideology or party? Does a citizen’s ideological preference determine his or her party preference? For example, if I am a traditional rightist, do I therefore identify with UMP? Or, does it work the other way? For example, if I am attached to UMP, do I therefore see myself more generally as on the right? Or, is it a little bit of both, in a reciprocal relationship in which my ideology influences my party choice, and my party choice influences my ideology?

9780230_300088_05_ch04.indd 74

6/28/2011 8:15:49 PM

Left-Right Ideology

75

To answer such questions requires formulation and estimation of simultaneous equation models of the vote. Fleury and Lewis-Beck (1993, p. 1104) carry out such an exercise with two-stage least squares on 1967 legislative election data, while Evans (2004, pp. 62–3) follows a similar strategy with 1993 legislative election data, in a maximum likelihood estimation. In carefully specified causal models, both find significant reciprocal links between party identification and ideological identification. However, they disagree over which indirect influence on the vote is more important, ideology via party, or party via ideology. Fortunately, Bélanger et al. (2006), in a still more statistically sophisticated attack, address the controversy using simultaneous equation estimation of panel data from the 2002 legislative election. This work clearly shows that party and ideology influence each other, with a slight edge going to the link from ideology to party. Analysis of scope and structure leads to the final component: strength itself. Among partisans, what happens to the vote when ideology and party change a comparable amount? In their clearest example, Bélanger et al. (2006, p. 509) demonstrate, from a twostage ordered logit, that the direct effect of ideology is about twice that of party. This heavy direct advantage of ideology, coupled with its indirect advantage just discussed, combine to give it much greater weight. Taken together with the much wider scope in the electorate, ideology deserves the role of the primary social-psychological anchoring variable in the French electorate. It is this driving attachment to ideology that seeks new parties, as new issues emerge.

4.3 Ideology and the vote: bivariate results In general, left-right ideological thinking occupies pride of place in the French electoral psyche. While our discussion has focused on national elections, it could apply to local ones as well (Lewis-Beck and Jérôme, 1999). But what, specifically, of our main concern: the recent presidential contests? Given their crown position in the hierarchy of elections in France, their importance reigns supreme. Viewing the 1988, 1995, 2002, and 2007 races, we have emphasized determinants of choice for the five leading candidates in each category: the extreme left, moderate left, center, moderate right, and extreme right. These choices form the categories of our dependent variable, on the first-round presidential ballot.

9780230_300088_05_ch04.indd 75

6/28/2011 8:15:49 PM

76 French Presidential Elections

For instance, with 2007, we look at extreme-left candidates (Laguiller, Besancenot, Buffet), Royal, Bayrou, Sarkozy, and Le Pen. While this simplifies the field, it nevertheless is comprehensive, accounting for about 94 percent of the voters. This figure is close to the average of 90 percent across all four elections and thus helps justify its choice as representative. With respect to the independent variable of ideology, we treat it as an ordinal scale of five, highlighting the parallel to the five candidate choices. Thus, in other words, across all four elections, we examine the five ideological categories of extreme left, moderate left, center, moderate right, and extreme right. We have stressed the importance of ideology for the French voter. Still, there is recurring talk, among scholars and journalists, of the “end of ideology,” of the notion that the French electorate is adrift politically, the left-right idea having lost its power to pull in voters. Do we see such impotence in the most recent French presidential elections? Table 4.1 relates ideological identification to vote choice, in these four first-round contests. Many things are worth noting. First, ideology seems far from dead. To illustrate, take the most recent example, that of 2007. Consider the modal candidate choices of those on the moderate left (Royal at 74 percent) and the moderate right (Sarkozy at 81 percent). Overwhelmingly, for these traditional voters, ideology structures their vote. The ideological values of the other choices also prove a compelling draw, as the modal values again show: extreme left (Laguiller, Besancenot, or Buffet at 54 percent); extreme right (Le Pen at 58 percent). The center also acts as expected, with the modal response of Bayrou (at 45 percent), and all the remaining responses balanced evenly for right-wing candidates (27 percent) or left-wing candidates (28 percent). In terms of percentage differences, this vote distribution provides a textbook example of the workings of ideological identification in the French national electorate. (As we shall see, the 2002 election departs some from the textbook model). Another way of appreciating the power of ideology is to consider how many voters actually “cross the line” to vote for a candidate of the opposing ideology – that is, how many voters who identify with the left vote for a candidate from the right, and vice versa. In 2007, of those who placed themselves on the left (“extreme” or “moderate”), only 4 percent said they voted for a right-wing candidate. Similarly, of those who placed themselves on the right (“extreme” or

9780230_300088_05_ch04.indd 76

6/28/2011 8:15:49 PM

Left-Right Ideology

77

Table 4.1 Bivariate relationship between ideology and first-round presidential vote Extreme left 1988

Extreme left Mitterrand Barre Chirac Le Pen N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

70 29 0 0 1 80 .44

Left

Center

Right

Extreme right

N

18 75 2 1 3 1277

4 34 28 19 14 769

1 4 28 50 18 861

0 1 9 28 62 87

327 1281 491 622 353 3074

.00 .80 (.01) .64 (.01)

1995

Extreme left Jospin Balladur Chirac Le Pen N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

42 48 1 2 7 182 .40 .00 .65 (.01) .53 (.01)

24 66 1 3 6 974

10 25 21 26 19 642

2 4 30 47 16 1037

1 2 21 36 40 165

399 926 500 746 429 3000

2002

Extreme left Jospin Bayrou Chirac Le Pen N

43 41 1 5 10 122

27 46 10 8 9 398

22 19 14 32 13 218

11 6 13 51 19 241

0 4 9 49 38 55

236 291 107 257 143 1034

Cramer’s V

.30

Pearson’s χ2 Gamma

.00 .50 (.03) (Continued)

9780230_300088_05_ch04.indd 77

6/28/2011 8:15:49 PM

78

French Presidential Elections

Table 4.1 Continued Extreme left

2007

Kendall’s tau-b

.39 (.02)

Extreme left Royal Bayrou Sarkozy Le Pen N Cramer’s V Pearson’s χ2 Gamma

54 44 1 1 1 131 .49 .00 .80 (.01)

Kendall’s tau-b

.65 (.01)

Left

Center

12 74 9 2 2 940

10 18 45 20 7 1198

Right

Extreme right

1 2 9 81 6 987

0 0 0 42 58 72

N

321 989 723 1093 202 3328

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. For 1988, 1995, and 2002, a seven-point scale recalibrated to five points is used (see Figure 4.1). For 2007, a five-point scale is also used (see Appendix). Source: CEVIPOF (1988, 1995, 2002, 2007).

“moderate”), a still smaller share (a little over 1 percent) said they voted for a left-wing candidate. The rarity of a deviation from one’s ideological political family emphasizes the weight of the left-right continuum. Ideological defection is a step not taken lightly. Put in practical terms, if you know a voter’s left-right ideological identification, you can almost perfectly predict the camp – left or right – to which that vote will go. With respect to those in the ideological center, they too behave according to a “middle of the road” script. If they do not pick the centrist candidate, they jump about evenly to a right-wing candidate or a left-wing candidate. The presence of this general left-right pattern, however, neglects the nuances of choice revealed in the rest of the table and the rest of the elections. To summarize the entire relationship between ideological identification and vote choice, we need to calculate the overall

9780230_300088_05_ch04.indd 78

6/28/2011 8:15:49 PM

Left-Right Ideology

79

measure of association between the two variables. Given that the two variables are ordinal, and the table is symmetric, tau-b becomes the preferred measure. Tau-b assesses the degree of monotonicity in the relationship and generally renders a numeric value just a bit below the familiar Pearson’s r, making it especially useful for baseline comparisons. In Table 4.1, tau-b is provided for each contest. For example, for 2007, tau-b = 0.65. This association easily achieves statistical significance (p < .01). Also, it demonstrates the strong relationship between ideology and the vote. Finally, the measure provides a good way of summarizing the story told in our elaboration of the percentage differences. What of the relationship in the earlier elections? These are also reported in Table 4.1, along with the appropriate significance tests and measures of association. Note that, regardless of the election, the relationship is always highly significant statistically. As well, observe that it is always hard for voters to “cross the line” and vote for a candidate of the opposing political family, emphasizing the enduring importance of the left-right divide. The overall association itself, as measured by tau-b, is everywhere robust. However, its strength varies from contest to contest: 1988, tau-b = 0.64; 1995, tau-b = 0.53; 2002, tau-b = 0.39; 2007, tau-b = 0.65. As it turns out, its strength is greatest in the 1988 and 2007 elections, and weaker in 1995 and 2002. Indeed, it is easy to believe that the heightened force of ideology in the Royal-Sarkozy battle was, at least in part, a reaction to the ideological disarray of the earlier, Jospin-Chirac competition. The first election in the series, that of 1988 (tau-b = 0.64), actually appears closest in ideological form to that of 2007. This is not surprising since the traditional left and the traditional right were each unified around one candidate – Mitterrand on the left and Chirac on the right – while Barre was the main appeal for the center (much as was Bayrou in 2007). Thus far, we have confined ourselves to the role ideology plays in deciding the first-ballot choice. However, a full understanding of its workings requires attention to the second ballot as well. On the first ballot, the French voter struggles over multiple party choices. Certainly, a voter’s ideology almost invariably anchors him or her to the left party camp or the right party camp. A leftist will eye a left party, a rightist a right party. As tight as that general link is, it still does not reveal which particular party label will be picked. For

9780230_300088_05_ch04.indd 79

6/28/2011 8:15:49 PM

80

French Presidential Elections

example, an extreme leftist may go for the mainstream Socialist Party, perhaps because he or she likes the Socialist Party leader or simply does not want to “waste” the vote on an extreme candidate who has no chance of winning. These nonideological forces, and the accompanying uncertainty, reduce somewhat the influence of ideology in the first round. However, in the second round these uncertainties are removed; by the rules of the game, only two candidates are then in contention. In general, one of these candidates represents the left, the other represents the right – for example, Royal versus Sarkozy in 2007. For our series of elections, the exception comes in 2002, when extremeright candidate Le Pen faced traditional-right candidate Chirac. As already noted, the failure of Socialist leader Lionel Jospin to reach the second round surprised everyone – pollsters, politicians, scholars, and citizens alike. In a real sense, then, this candidate duo represents an aberrant, random pairing. Because we wish to generalize about mass electoral behavior across all four elections, we consider 2002 vote intention before the second round, rather than reported vote behavior after. Of course, the actual vote behavior held paramount interest, with its resounding defeat of Le Pen. Still, as important as this political outcome was, it tells us precious little about the role of ideology. Why? Because almost everyone – moderate right, left, or center – opposed Le Pen. Under the usual circumstances of a presidential election campaign before the second round, ideology operates freer of competing forces. For most voters, their preferred party is no longer in the race. More generally, voter strategy shifts to defeating the candidate farthest away from their home ideology. (See the discussion in Lewis-Beck and Chlarson, 2002, p. 498; Tiberj, 2004). Hence, ideology plays an even greater role than in the first round. In Table 4.2, we see the relationship between ideology and presidential vote in the second round of these four elections. Observe that voter solidarity with the political family appears strong, even stronger than in the first round. For example, in 2007 those on the right (“extreme” or “moderate”) voted almost exclusively for Sarkozy (at 98 percent), while those on the left (“extreme” or “moderate”) voted almost exclusively for Royal (at 96 percent). This pattern of political family solidarity continues back through the other contests. Here are the percentages, respectively, linking total (“extreme” plus “moderate”) ideology to vote: 1988 (left,

9780230_300088_05_ch04.indd 80

6/28/2011 8:15:50 PM

Left-Right Ideology

81

98 percent; right, 92 percent), 1995 (left, 95 percent; right, 95 percent), 2002 (left, 84 percent; right, 85 percent), 2007 (left, 96 percent; right, 98 percent). The great strength of the relationship continues to reflect itself in the measures of association, which include those in the ideological center, as well as the different left and right categories. Cramer’s V is consistently over 0.70, with the exception of 2002 (where it is .63). Again, the somewhat weaker relationship in 2002 mirrors the ideological disarray of the contest, with the “protest attitude” of the extreme left, the unexpected move to the center by Jospin, and the last-minute surge by Le Pen. While many bemoaned the final outcome of that race, it has value in illustrating that, for a candidate, ideology need not be destiny. Put another way, ideology, strong as it may be, is not the only force driving the electorate. Indeed, it is time to consider other forces, which we do below.

4.4

Ideology and the vote: multivariate results

In considering the democratic vote choice, a single variable, no matter how powerful, cannot stand as the sole determinant. While ideology certainly plays an important role in shaping candidate preference in French presidential elections, there are other long-term forces at work on the electorate. These must be taken into account, otherwise we exaggerate the impact of left-right ideology. In previous chapters, we in fact explored other variables of this type, when we examined the impact of age, gender, education, religion, occupation, income, and patrimony. We can group these variables together, under the “social factors” rubric. Considering these alongside ideology gives us the following model of the French presidential election voter: Presidential vote = social factors + ideological identification. Thus, we have elaborated our model, which is now composed of two kinds of long-term forces – that is to say, forces that are more or less fixed, enduring, and stable in the electorate. Obviously, there are short-term forces, such as issues and candidate quality, that we have yet to consider. Still, it is highly useful to see how well we can predict vote choice with these more exogenous variables.

9780230_300088_05_ch04.indd 81

6/28/2011 8:15:50 PM

82 French Presidential Elections

Table 4.2 Bivariate relationship between ideology and second-round presidential vote Extreme left 1988

Mitterrand Chirac N Cramer’s V Pearson’s χ2

99 1 81 .78

Left

Center Right

Extreme right

N

98 2 1391

53 47 829

8 92 886

8 92 75

1961 1301 3262

Gamma Kendall’s tau-b

.00 .95 (.01) .72 (.01)

1995

Jospin Chirac N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

97 3 190 .79 .00 .94 (.01) .71 (.01)

94 6 1064

42 58 711

5 95 1110

7 93 162

1557 1680 3237

2002

Jospin Chirac N Cramer’s V Pearson’s χ2 gamma Kendall’s tau-b

91 9 245 .63 .00 .81 (.01) .56 (.01)

83 17 1018

39 61 527

15 85 747

14 86 111

1399 1249 2648

2007

Royal Sarkozy N Cramer’s V Pearson’s χ2 Gamma

99 1 144 .76 .00 .97 (.00)

95 5 981

49 51 1175

2 98 1052

0 100 69

1671 1750 3421

Kendall’s tau-b

.70 (.01)

Notes: Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. Ideology scales are identical to Table 4.1. Source: CEVIPOF (1988, 1995, 2002, 2007).

9780230_300088_05_ch04.indd 82

6/28/2011 8:15:50 PM

Left-Right Ideology

83

In the Appendix (Model 3), the above model is estimated, in a series of logistic regression equations, on first-ballot choice in each of the four contests. Again, for each election we have the five leading candidates, each regressed on the social factors, plus the ideology variables. It is worth noting that ideology continues to be measured in five categories: extreme left, left, center, right, and extreme right. These categories are represented as “dummy” variables in each equation, with “center” as the baseline category. (That means, effectively, that the score for those in the center is captured in the intercept term of the equation.) A first question is whether the presence of these social factors seriously diminishes the impact of ideology. Theoretically, of course, it is possible that the bivariate results we just examined are illusionary, a product of demographic conditions operating behind the scenes. If that is so, once we control for them in a multivariate context, the impact of ideology would disappear. In other words, it would be rendered spurious, a correlation without causation. An essential test here involves examination of the coefficients for statistical significance. We observe that the left-right ideology variables are almost always significant, at 0.001 or beyond. (The few times they are not is mostly due to severe collinearity problems, which made it necessary to remove a dummy in certain cases.) What is more, ideology structures vote choice in the expected way: broadly speaking, French voters on the left align themselves with candidate from the left, voters on the right support one of the candidates on the right, and the centrists tend to support a center (or a center-right) candidate. What about the impact of the social factors? Do they lose their statistical significance when ideology is controlled for? We see that some, scattered coefficients do now fail to reach a conventional level of significance. The most striking examples concern support for Jospin in 1995, for which only age and the private sector remain significant once the influence of ideology is controlled for, and support for Royal in 2007, for which all six of the social variables found significant in the previous two chapters are not so anymore (although men’s and employees’ support for Royal does become statistically significant now that we control for ideology). Nevertheless, on balance, we see that social factors maintain their impact. Indeed, whereas 101 such coefficients were found to be significant in the previous stage of the vote choice model, as many as 85 of those are still

9780230_300088_05_ch04.indd 83

6/28/2011 8:15:50 PM

84 French Presidential Elections

significant in the current stage. Some blocks of significant social factors do not change at all from one stage to the next: support for Chirac and for Le Pen in 2002 is driven by the exact same social factors as support for the extreme left in 1988. It is thus clear that both sets of long-term factors are important for vote choice and, to a considerable degree, operate independently of each other. Taken together, they fill out the profile of the typical voter. For example, consider what the coefficients say about Sarkozy supporters: older, female, religious, higher income, with high-risk assets, and, most especially, from the moderate right. How well, in general, do the equations perform? Compare the pseudo-R-squared across candidates, within each election. Observe the following range, for each election: 1988, 0.15 to 0.30; 1995, 0.13 to 0.39; 2002, 0.07 to 0.18; 2007, 0.16 to 0.40. First, as expected, each single candidate equation alone leaves a fair amount of variance unexplained. This is what should happen, since the choice, for presentation purposes, simplifies in each equation to one candidate versus all others. What interests us is that the typical ability to predict the vote remains about the same, regardless of candidate, as the narrow range of scores shows. Furthermore, the fit is rather stable regardless of election – that is, the median pseudo-R-squared for each contest is as follows: 1988, 0.21; 1995, 0.22; 2002, 0.10; 2007, 0.19. The possible exception here, as already noted, comes with 2002. Otherwise, ideology operates in a generally dependable way on the French voting public. This result is seen most clearly when the second round of voting is examined, in the Appendix. When the vote choice is reduced to two candidates, as invariably happens during the second round, the structuring power of ideology becomes quite impressive. Here are the pseudo-R-squared for the second-round contests: 1988, 0.58; 1995, 0.56; 2002, 0.34; 2007, 0.57. Compare these to the pseudo-R-squared for the equations without the ideology variable (recall that these are equivalent to the model fits from the last chapter): 1988, 0.12; 1995, 0.11; 2002, 0.08; 2007, 0.09. Only with the introduction of ideology into the models are we generally able to explain more than half the variance in the vote choice. Comparing the two long-term forces acting on the French voter, ideological identification weighs much more heavily than social factors, although both offer critical parts of the explanation.

9780230_300088_05_ch04.indd 84

6/28/2011 8:15:50 PM

Left-Right Ideology

85

A final piece of evidence as to the structuring force of ideology in the second round of presidential elections comes from the regression coefficients themselves. Notice that in all four elections, the propensity of extreme-left and moderate-left voters to support the secondballot left candidate is very high. The likelihood of extreme-right and moderate-right voters to support the candidate from the right is similarly high. Looking at the ideological extremes, it seems that those on the left have a particularly strong inclination to rally themselves behind the left candidate even though the latter typically comes from the more traditional camp. With regards to extremeright voters, they too are more likely to support the second-ballot candidate from the right but not as much as moderate-right voters. The exception to the latter trend is the 2007 election, in which in the second round Sarkozy was able to attract overwhelming support from voters on the extreme right. (Indeed, the extreme-right dummy is dropped from the 2007 second-round vote choice model because it predicts the vote perfectly, a result that is consistent with the findings presented in Table 4.2.)

4.5

Conclusion

Ideology, in the form of political views on the left or the right, has been at the heart of France’s political life ever since the Revolution. As was shown in this chapter, it still acts today as the main psychological anchor among the French electorate. Most French voters are able to identify with an ideological camp, be it left, center, or right, moderate or extreme. Moreover, this identification appears meaningful, in the sense that it is a real attitude and not an opinion devoid of any political content. Ideology constitutes a key variable that helps French voters to orient themselves toward political objects and, more importantly, that structures their choice among a list of presidential election candidates. Looking at the last four presidential contests, 1988–2007, we have seen that left-right ideology predicts to a large extent which candidates voters align themselves with. Even voters located at the center of the ideological spectrum have a greater tendency to support a centrist candidate, and also they tend to split their support almost equally between candidates from the traditional left or right. In general, very few voters actually dare to “cross the line” and support a

9780230_300088_05_ch04.indd 85

6/28/2011 8:15:50 PM

86

French Presidential Elections

candidate from an ideological camp opposite to theirs. The strong structuring effect of ideology reveals itself to be even more important when it comes to the second ballot of a presidential race, when only two candidates remain. Indeed, in the second round voters from the extreme left (or right), whose preferred candidate is not normally on the ballot anymore, tend to rally en masse behind the candidate of the traditional left (or right). The same is true for centrist voters, who now have to support one of the two remaining, left and right candidates. Ideology operates similarly strongly in a multivariate context, once we control for the variables lourdes in the form of social factors. Independently of these other variables, left-right identification helps account for a sizable portion of the variance in vote choice, up to about half of the variance in the case of the second-round vote. The centrality and importance of ideology in explaining French voting behavior thus cannot be overstated.

9780230_300088_05_ch04.indd 86

6/28/2011 8:15:50 PM

5 The Role of Issues

In contemporary democracies, issues stand at the heart of electoral debate. We have not yet talked directly about the impact of issues on the French voter. To be sure, matters that weigh heavily on the public mind have been discussed: social class, religion, and left-right ideology, to name a few. But these long-term forces, powerful as they are, differ from the current preferences of voters on the burning political questions of the moment. While such issues may be short-term events, and hence phenomena more changeable than enduring in the public psyche, they nevertheless can enter very immediately into a presidential Election Day choice. Undeniably, for at least some voters, their feelings on a single topic mean everything to them at the ballot box. Indeed, this argument has been made for electors of the Greens (les Verts) and the National Front (Chiche and Mayer, 1997). Speaking generally, leading scholars of French presidential races have not shied away from an examination of issues. After 1978, a major French national election survey (by CEVIPOF) was not conducted until 1988. However, in that survey, respondents were queried about many topics – for example, nuclear weapons spending, salary inequalities, abortion, homosexuality, regulation of business, trade unions, women’s rights, the death penalty, and minority rights (Boy and Mayer, 1993). In the CEVIPOF volume on the 1995 contest, a pivotal chapter devotes itself to “les enjeux de l’élection,” looking at questions of Europe, the environment, immigration, security, purchasing power, corruption, the homeless, AIDS, and unemployment (Chiche and Mayer, 1997, chp. 6).

87

9780230_300088_06_ch05.indd 87

6/28/2011 8:15:13 PM

88 French Presidential Elections

For the 2002 election, a French electoral panel was carried out, for the first time. Exceptionally, a top issue identified is law and order. Mayer and Tiberj (2004, p. 45) observe: “This is one of the main reasons for Le Pen’s success on 21 April. Immigration still is his key issue. But… It is the focus on criminality during the presidential campaign that benefited the National Front’s leader. It provided him with enough extra votes to qualify himself for the second-round of the election.” With respect to the 2007 race, further panel analysis is conducted, on a host of issues. In an extensive treatment of “The Electoral Agenda and the Issue Vote” Sauger (2009, p. 200) finds “an absence or at the very least a weakness of issue voting in 2007.” He goes further, speculating about “the difficulty of distinguishing a specific impact of issues” on the French voter. In this chapter, we attempt to draw some tentative conclusions about the direct effect of issue voting during French presidential elections, with an analysis that includes, but does not confine itself to, the 2007 struggle ended by the Sarkozy win.

5.1

The meaning of issue voting

While citizens know that they may affect government policy, they also know that government policy may affect them. Therefore, citizens are sensitive to public issues close to their hearts, and they worry about how candidates might change the ways goods on these issues are allocated. An environmentalist worries over a change in the rules about recycling glass. An antiabortionist might want to stop the government subsidy of private clinics that perform the procedure. A nuclear-free-zone advocate could fret about dependency on atomic energy for the nation’s fuel. These are just a few examples of issues in the public domain that might influence voters for or against a candidate, depending on the candidate’s declared position. So, what issues are important to French voters, and how does this play out at the ballot box? In other words, how can issues come to exercise influence in the electoral arena? Such influence, we note by way of general caution, is not easily had. While issues do matter, we shall find that those that do are few in number, and their force may flow as a trickle rather than a torrent. Consider the conditions necessary for an issue to have relevance and thus possibly lead to “issue voting”: (1) the voter must perceive

9780230_300088_06_ch05.indd 88

6/28/2011 8:15:14 PM

The Role of Issues 89

the issue; (2) the voter must have a preference on the issue; (3) the voter must believe the candidates themselves differ on the issue. These conditions are not new. They were first lain down in the classic treatise on political behavior, The American Voter (Campbell et al., 1960, chp. 8). (In addition, see the update and literature review in The American Voter Revisited (Lewis-Beck et al., 2008, chp. 8).) These conditions are not easily met. It is immediately obvious that, for an issue even to be a contending influence on the French voter, many citizens have to care about it, have differences of opinion about it, and perceive that the candidates offer different solutions. Take an illustration at one extreme. In an early Fifth Republic election, a renegade presidential candidate campaigned on the platform of providing voters with inexpensive pear brandy and chocolate. (Historically, it has not been difficult to at least get a name on a presidential ballot.) Clearly, few citizens would believe provision of brandy and chocolate was an important public issue. Undoubtedly, there would be virtually no distribution of preferences on brandy and chocolate provision, since everyone would think this was not in the government realm. Finally, voters would know that no serious candidate would take any sort of position on this issue. In sum, the question of pear brandy and chocolate fails miserably on all three criteria of issue influence. Now take an illustration at the other extreme: that of the economy. Most voters are aware of the economy as an issue and have strong preferences about it – most often, they want prosperity. Furthermore, they often believe that one candidate is more likely than another to deliver that prosperity. Given that the economy appears to fully meet the three necessary conditions, it may be expected that it would influence the vote. That is not necessarily the case, however, because it may be crowded out by other issues. That is to say, French voters, going into a certain presidential election, may feel that the economic issue, real as it is, is not important compared to something else, say immigration. In that scenario, opinions on immigration might trump those on the economy. Obviously, the importance of an issue counts for a great deal. Thus, we begin our examination by considering what issues the French public considers important. To begin, we define an “issue attitude” as a socio-psychological combination of cognition and affect concerning an issue. That is, the citizen mentally notes the issue and forms positive or negative

9780230_300088_06_ch05.indd 89

6/28/2011 8:15:14 PM

90 French Presidential Elections

feelings about it. Further, these feelings can vary in their constancy and their intensity. As survey researchers, we elicit these issue attitudes by posing questions to respondents. Most of the time, respondents cooperate willingly in election surveys. Of course, the researcher’s goal is not just to elicit any answer, no matter what. Instead, the goal is to somehow tap a “true” attitude, not one that is lightly held, even random, merely a thought of the moment. Special care, then, is demanded in questionnaire construction. The items need to be phrased in such a way that they provide a neutral but informational cue about the issue, offering a meaningful framework for evaluation. Successfully executed, such questions can yield thoughtful, accurate responses. Consider the 1995 CEVIPOF survey, which asked the following: “Here are a certain number of problems that present themselves in France these days. With the aid of this scale, can you assign each a grade from 0 to 10, according to the importance that it would have in your vote in the first round of the presidential election?” (Chiche and Mayer, 1997, pp. 221–22). Note that the respondent is encouraged to reflect on a set of issues regarded, at least popularly, as relevant for the campaign. They are asked to then evaluate these issues in a comparative framework. The exercise appears to be a straightforward way of getting a respondent’s serious but objective attention. The results appear in Table 5.1. First, it should be recorded that, almost always, respondents graded the problem. (The average non-response for the 13 items is only about 5 percent). Moreover, they stayed away from the “neutral,” perhaps empty or meaningless, grade of “5.” Indeed, the lowest average score for any item was 6.2 (on European construction), with the median average score at 7.5. Clearly, most of these respondents perceived these issues and had preferences on them. Tellingly, not all issues were of equal importance. For example, unemployment, education, and social protection were rated very important (with scores 8.0 or better), while Europe, the environment, and work-sharing were not (with scores less than 7.0). Voters, then, were discriminating in their issue judgments, on the eve of the 1995 presidential contest. What does the picture look like for later contests? In Table 5.2, issues are ranked in importance, from similar election surveys before the 2002 and 2007 races and on the basis of the data reported in Table 5.1 for the 1995 race. (As can be seen, the issues are

9780230_300088_06_ch05.indd 90

6/28/2011 8:15:14 PM

The Role of Issues 91

Table 5.1 Evaluating the importance of problems before the 1995 presidential ballot, first round Average score Unemployment Education Social protection Homelessness Inflation AIDS Corruption Personal security The environment Work-sharing Immigration France in the world

8.9 8.3 8.0 7.9 7.8 7.5 7.5 7.2 6.8 6.7 6.5 6.4

Construction of Europe

6.2

Note: Entries are average scores on a scale from 0 to 10. Source: CEVIPOF (1995).

labeled in a very general way, usually with only one word. That is in order to group sometimes differently worded survey items under the same label.) Several things about the surveys are worth noting from the table. Always, 11 to 13 “problems,” no more and no less, are evaluated for their “importance” (although the exact wording and exact measures depend on the particular survey, noted at the bottom of the table). Within this common structure, continuities emerge. Some issues, such as immigration, were mentioned in all three surveys. Some issues, such as unemployment, were consistently rated highly important. Some issues, such as personal security, seem to bounce around in their importance. Of special concern are the problems that persist, the ones that are always on the voter’s agenda. These problems themselves would seem most able to move the mass public, as they socialize the French voter from one election cycle to the next. There are six of these: personal security, immigration, the environment, schools, unemployment, and Europe. Each made the list of important problems in

9780230_300088_06_ch05.indd 91

6/28/2011 8:15:14 PM

92

French Presidential Elections

Table 5.2 The ranking of issues in presidential election campaigns (1995, 2002, 2007) Rank in importance

Unemployment Education Social protection Homelessness Inflation AIDS Corruption Personal security Environment Work sharing Immigration France in the world Europe Inequality Retirement Taxes Terrorism Sovereignty Deficits Housing Suburbs

1995

2002

2007

1 2 3 4 5 6 7 8 9 10 11 12 13

1 7

1 4

3 10 2 6

7 6

4

10

11 3 5 8 9 12

12 2 8

5 9 11

Notes: Rankings are based on the “most important problem” perceived by respondents, on a choice of 13 (1995), 12 (2002), or 11 (2007) problems. The ranking for the “Europe” variable in 2007 is drawn from the first wave of the Panel électoral français (13 items in the question), because it is not available in the post-electoral survey. Source: CEVIPOF (1995, 2002, 2007).

every election. Certainly, these problems compose part of the common political discourse in French presidential election campaigns. No one would deny, on the face of it, their potential as vote-getters (or vote-losers). For the sake of consistency and generalization, we wish to examine the impact of the same strong set of issues across

9780230_300088_06_ch05.indd 92

6/28/2011 8:15:14 PM

The Role of Issues 93

all four presidential elections under study. While this is a constraint, it also gives focus to the promising battery of these six issues. Unfortunately, all six cannot be easily examined because, while they are measured in some manner in each election, empirically those measures can be quite diverse. For example, while the item on the environment asks about “the environment” in 1995, it asks about “pollution” in 2002, and about “the environment and global warming” in 2007. However, we do have three issues with measures conceptually and empirically closely equivalent across these elections: unemployment, immigration, and Europe. In addition to exhibiting strong measurement continuity, these items also stand for representative issue types. First, concerning the domestic to international issue spectrum, they can be arranged from unemployment (domestic) to the environment (domestic/international mix) to Europe (international), respectively. Second, they divide into what Carmines and Stimson (1980) call “easy” issues and “hard” issues. Easy issues are more symbolic than technical, concern ends more than means, and have been publicly prominent for a long time. Immigration, with its emphasis on ethnicity, and unemployment, with its emphasis on general economic well-being, could be classified as “easy” issues, whereas Europe, with its emerging concerns over political and economic regulation, is a “hard” issue. Lastly, they nicely represent “valence” versus “positional” issues. Valence issues are those on which the public holds a consensus view (Stokes, 1963). The economy is the valence issue par excellence, because all voters prefer the prosperity that low unemployment brings. Positional issues are those on which the public distributes itself along a preference continuum, say, on immigration, from a closed-borders policy to an open-borders policy position (Downs, 1957). (For a recent comparison of the relative impact of the two issue types in the 2008 US presidential election, see Lewis-Beck and Nadeau (2009); for a similar comparison in Canadian elections, see Bélanger and Meguid (2008).) Surely, with respect to foreign policy, the construction of Europe stands for France as perhaps the leading positional issue of the day. In sum, our three issues – unemployment, immigration, and Europe – while far from exhaustive of the issues landscape, provide a favorable opportunity for issues to exert themselves on the French voter, if they are able.

9780230_300088_06_ch05.indd 93

6/28/2011 8:15:14 PM

94 French Presidential Elections

Thus far, we have ignored the third criterion for issue voting: the candidates must be perceived as offering different solutions on the issue. In American National Election Surveys (ANES), once this criterion is included, the pool of those who are able to engage meaningfully in issue voting is further reduced. For example, in the 2004 ANES, only about half the respondents (54 percent) correctly located the Republican and Democratic party positions on a similar, relevant battery of six issue items (Lewis-Beck et al., 2008, p. 169). Interestingly, we find a comparable result in the 2002 CEVIPOF survey: “On one third of the problems, more than half see ‘no candidate’ capable of tackling the issue, or do not answer.” (Mayer and Tiberj, 2004, pp. 36–37). With respect to the issues under consideration here, in the 2002 survey 49 percent, 30 percent, and 32 percent of respondents perceived no difference between candidates on unemployment, immigration, and Europe, respectively. For voters who fail to see the candidates as different on an issue, it is of course impossible for that issue to influence their candidate choice. Further, issue voting may also be weakened, if not actually explained, by partisan bias among those who do see candidate differences (Mayer and Tiberj, 2004, p. 37). To continue the 2002 example, overwhelmingly the candidate considered “best” on the unemployment issue was Jospin (with 38 percent, compared to Laguiller in second place with 16 percent). An implication is that this majority response, favoring the two leading candidates on the left, is largely a projection of the ideological bias of the voter. The same picture emerges with the issue of immigration, for which the “best” candidate was considered to be Le Pen (with 59 percent, followed by Chirac with 9 percent), choices suggesting again ideological motivation. Finally, on the Europe issue, respondents handed “best” candidate to Jospin (with 30 percent) and second “best” to Bayrou (with 23 percent). Those preferences, too, are not without their partisan flavor. One disturbing implication is that issue voting, when it appears, actually merely fronts for left-right ideological leaning. In other words, the issues, in themselves, may have little if any impact once partisanship is taken into account. Eventually, we fully address this possibility. However, we begin with simple consideration of our prime valence issue – economics – and our prime positional issue – immigration, examining the theory behind these particular “issue votes” in the French context, then offering initial

9780230_300088_06_ch05.indd 94

6/28/2011 8:15:14 PM

The Role of Issues 95

bivariate tests of that theory. Once those results are thoroughly familiar, we move to more rigorous exploration of their effects in a multivariate context, finally giving full inclusion to our third issue: the construction of Europe.

5.2

Economics and the French voter

The classic theory of economic voting is straightforward. (For full discussion of the theory in the context of Western European elections, see Lewis-Beck (1988).) When economic conditions are good, voters tend to vote for the government. When economic conditions are bad, voters tend to vote against the government. This simple reward-punishment theory is based on the notion of attribution of responsibility, the psychological mechanism by which voters link the government to the economy – that is, hold the government responsible for economic performance. A first question concerns what aspects of the economy voters consider. One perspective holds they evaluate the economic past, usually national economic performance over the past year. Called retrospective economic voting, this view contrasts with prospective economic voting, which looks to the economic performance in the future. (On the theoretical sources of these distinctions, see Downs (1957), Fiorina (1981), and Key (1966).) A second question concerns the place of economic activity itself. In particular, do voters look at the collective level – the nation? Or do they look at the individual level – their personal finances? The former refers to sociotropic economic voting, the latter to egotropic (pocketbook) economic voting. (For more about these distinctions, see Kiewiet (1983) as well as Kinder and Kiewiet (1981).) Economic voting has been subjected to much study, in national election surveys from democracies around the world. (See the review in Lewis-Beck and Stegmaier (2007).) France stands as no exception here. In fact, it is one of the most investigated cases, after the United States and Great Britain. First, there is the long tradition of popularity functions, with government support largely driven by macroeconomic forces. This mode of analysis, begun by Lafay (1977, 1991), depends on time series rather than survey data and, while not pursued here, reveals the consistent and significant influences on French executive popularity from growth, unemployment, and inflation. (For a review of that literature, see Lewis-Beck (2001); Lewis-Beck and

9780230_300088_06_ch05.indd 95

6/28/2011 8:15:15 PM

96

French Presidential Elections

Stegmaier (2000).) With regard to survey research, the first study comes from Lewis-Beck (1983), who examined Euro-Barometer data, 1970–1978. He located 55 disparate economic measures and found on average they correlated (r) at 0.20 with left-right vote intention. Moreover, in regression models of the legislative vote, 1973 and 1978, both sociotropic and egotropic effects are statistically significant. In a subsequent investigation of the French Euro-Barometer data, again modeling legislative vote intention, he uncovered statistically significant influences from collective retrospective and prospective economic items, in the face of controls on class, religion, and ideology (Lewis-Beck, 1988, pp. 56, 82). With respect to French national election studies themselves, the first serious battery of (eight) economic items was fielded in 1995. Each of these eight items was either retrospective or prospective, personal or collective (egotropic or sociotropic), and simple or complex. (The last merely refers to whether government was explicitly mentioned in the item.) A pretest of these items provided the following correlations (r) with vote intention: personal, simple, retrospective (0.12); personal, simple, prospective (0.22); personal, complex, retrospective (0.25); personal, complex, prospective (0.28); collective, simple, retrospective (0.36); collective, simple, prospective (0.49); collective, complex, retrospective (0.48); collective, complex, prospective (0.54) (Lewis-Beck, 1996, p. 521). An important thing to note is that the collective items are rather robust and consistently outperform the personal items. Analysis of these items, embedded within a full specification of the presidential vote choice estimated via logistic regression, emphasizes the presence of the economic vote, with a character again more sociotropic than egotropic, and prospective as well as retrospective (Lewis-Beck, 1997a, p. 245). A vivid illustration of the relationship between sociotropic evaluation and the presidential vote comes from the 1995 election survey. As the perception of the future of the national economy deteriorated, so the likelihood of a Jospin vote increased, and strongly so. (Of those who saw the economic future as “worse,” 82 percent expressed a second-round vote for Jospin, compared to only 28 percent among those who saw it as “better” (Lewis-Beck, 1997a, p. 242)). This scenario nicely fits the attribution-of-economic-responsibility pattern for the French voter. The 1995 contest took place under cohabitation (in which the prime minister and the president are from different

9780230_300088_06_ch05.indd 96

6/28/2011 8:15:15 PM

The Role of Issues 97

party coalitions). At that time, voters held the Gaullist PM Balladur chiefly responsible for overseeing the economy, not the Socialist president Mitterrand. For those who perceived a worsening economic situation, they thus tended to favor Socialist Jospin over Gaullist Chirac. The role of cohabitation in conditioning the economic voting coefficient merits further development in the context of presidential elections, the object of the investigation at hand. Lewis-Beck and Nadeau (2000, pp. 177–80) examine the presidential elections of 1988 and 1995, both occurring under cohabitation (1988 with Chirac PM and Mitterrand president, 1995 with Balladur PM and Mitterrand president). They observe that economic voting remains strong in these presidential contests. However, the target is the candidate of the PM’s coalition (not the candidate of the president’s coalition), since it is the PM who holds the role of economic manager. The president can assume this role only under unified government when, à la Général de Gaulle, the PM becomes the right hand of the president. The pre-1988 presidential elections of the Fifth Republic, all under unified government, do follow this pattern of economic voting (Lewis-Beck, 1997a). What about the presidential contest of 2002? Prime Minister Jospin’s government at the time was a four-party coalition (Socialists, Communists, Left Radicals, and Greens), which might be expected to weaken the attribution of economic responsibility. Moreover, crime seemed an especially important issue, leading to the “Le Pen surprise” of his appearance on the second-round ballot. In these conditions, asks Guy Whitten (2004), “could there possibly have been economic voting?” His answer is in the affirmative, finding the incumbent vote (first-round legislative) to be heavily determined by current performance on the particular macroeconomic indicators of growth, inflation, and unemployment. Lewis-Beck and Nadeau (2004) support Whitten’s conclusion, placing it within an institutional context. As with the 1988 and 1995 elections, that of 2002 occurred under cohabitation, resulting in the tactical movement of economic responsibility from the president’s shoulders to the prime minister’s. Of course, the 2007 presidential election did not take place under cohabitation. The expectation, then, would be for Sarkozy and the UMP presidential coalition to be the target of the economic vote. Jérôme and Jérôme-Spéziari (2007), in a presidential voting model

9780230_300088_06_ch05.indd 97

6/28/2011 8:15:15 PM

98 French Presidential Elections

relying heavily on improvements in the unemployment rate going into that contest, predicted the second-round presidential vote for Sarkozy virtually without error. In sum, we observe that the French economic voter appears rather sophisticated, with the ability to reassess who is in charge and to adjust his or her vote choice accordingly. When government is unified, the presidential economic vote takes aim at the presidential coalition. In contrast, when government is divided under cohabitation, the presidential economic vote takes aim at the prime minister’s coalition.

5.3 Economics and the French voter: bivariate relationships Given the sometimes complicated dynamics of the economic vote in France, we should begin our current analysis looking at simple relationships, using only minimal assumptions. We start with the 1995 contest, since that survey offers the greatest array of measures to select from. Take the classic retrospective sociotropic item of the form “Over the last twelve months, do you feel the economic situation of the country has gotten better, gotten worse, or stayed the same?” (Its response distribution is as follows: better, 16 percent; worse, 41 percent; same, 39 percent.) In a logistic regression model predicting the first round, with controls on class, religion, and ideology, it manifests a statistically significant economic effect exceeding that from egotropic measures (Lewis-Beck, 1997a, p. 245). For now, we just show bivariate Table 5.3, relating this economic attitude to vote choice arranged left to right (Hue and Laguiller, Jospin, Balladur, Chirac, Le Pen). There are different ways to summarize the results of this table. One is to look at the percentage of support for Prime Minister Balladur, himself a candidate. Observe that among those who saw the economy as “worse” only 11 percent favored him, but for those who saw the economy as “better,” the percentage rises to 33. Clearly, a group of voters were willing to reward the PM with their vote, in return for his good economic stewardship. But other things are going on, too. The candidates on both the extreme left and extreme right draw an economic vote from the disgruntled, while Balladur’s RPR rival, Chirac, seems also to gain votes after the fashion of Balladur, but less decisively so. Because of the apparent multidimensionality of the relationship, it may be best not to impose any ordinality on the

9780230_300088_06_ch05.indd 98

6/28/2011 8:15:15 PM

The Role of Issues 99

Table 5.3 The relationship between economic perception and the 1995 presidential vote, first round (%)

Hue and Laguiller Jospin Balladur Chirac Le Pen N Cramer’s V Pearson’s χ2

Economy worse

Economy same

16 33 11 22 18 1256 .19

12 34 16 25 13 1136

Gamma

.00 .09 (.02)

Kendall’s tau-b

.06 (.02)

Economy better 8 17 33 33 10 505

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. The economic question reads as follows: “In the last twelve months, do you feel that the economic situation of the country has gotten better, gotten worse, or stayed the same?” Source: CEVIPOF (1995).

measures, either X or Y. Instead, we simply apply a measure of association that summarizes their relationship as two nominal variables, Cramer’s V. This measure bases itself on the chi-square and achieves a score of 1.0 when the relationship is perfect, 0 when it is independent. Here the chi-square for the table registers statistical significance. Also, V = 0.19, indicating a modest relationship between national economic evaluation and first-round presidential vote in 1995. Suppose we run the same such table, relating economic evaluation to vote, for each of our first-round presidential results, each time calculating Cramer’s V. Here are those results: 1988, V = 0.31; 1995, V = 0.19; 2002, V = 0.20; 2007, V = 0.06. What do we observe? Always, there is a modest relationship between economic perception and the vote. Indeed, the average V = 0.19 across these four elections. Observe that this is equivalent to the average correlation that Lewis-Beck (1983) long ago reported from an extended pool of economic items. (Recall V approximates r, if data are assumed ordinal.) Further, there is mild variability, partly due to the fact that the measure of the economy changes across contests. The measure for 1988 and 1995 is

9780230_300088_06_ch05.indd 99

6/28/2011 8:15:15 PM

100 French Presidential Elections

a classic retrospective, sociotropic measure of the national economy. However, such a measure is not available for these 2002 and 2007 surveys, in which the measure is a simple dichotomy, separating out whether the respondent ranked unemployment as “a top problem” or not. Even with these less than ideal measures for 2002 and 2007, statistically significant economic effects still manage to emerge.

5.4

Immigration and the French voter

France once was, and in some ways still pretends to be, a colonial power. With the establishment of the Fifth Republic and the breaking up of empire, the cities of metropolitan France have become home to waves of immigrants. After the Algerian war, President de Gaulle welcomed the arrival of ex-colonial citizens from North Africa, who could provide inexpensive labor for the economic boom the nation enjoyed in the 1960s. The immigrant population swelled and soon became an object of political attention, spurred by the founding of the National Front (FN) in 1972. The anti-immigrant political activities of Jean-Marie Le Pen and his FN followers were little noticed until the 1983 municipal elections in Dreux, when the party list got a surprising 17 percent of the vote (Schain, 1987). Besides Dreux, FN candidates in Marseilles outdid themselves in their tough anti-Arab rhetoric. The party appeared something of a political contagion, quickly taking 11 percent of the vote in the 1984 European parliamentary elections. In the regional elections of 1986, the FN actually captured 135 seats. More seriously, in the legislative elections of that year, FN candidates took 10 percent of the vote and, because of the proportional representation rules at the time, gained 35 seats in the National Assembly. Clearly, the National Front had arrived as a force on the French political scene. Its fuel is what Nonna Mayer (1993) called “ethnocentrism,” a complex of racial beliefs with the immigrant problem at its core. Our purpose here is to detect the influence of these issue attitudes on the French presidential vote. Before the FN strutted onto the national political stage in 1986, the particular importance of the immigration issue had become clear. National survey results from 1984 and 1985 relating issue preferences to legislative vote intention show that, regardless of party preference, at least a quarter of the respondents “agree completely” that there are too many immigrant workers (Lewis-Beck and Mitchell, 1993).

9780230_300088_06_ch05.indd 100

6/28/2011 8:15:15 PM

The Role of Issues 101

Further, in these data there seems to be something of a left-right divide, with those on the left rather more tolerant of immigrants than are those on the right. Most particularly, FN supporters were extreme in their views, almost doubling the anti-immigrant feeling expressed by the traditional right Gaullist (RPR) voters – 88 percent of FN supporters said there were “too many” immigrant workers, compared to 45 percent of those from the RPR. Almost all FN voters, then, had a very strong anti-immigrant attitude. However, that is not all that distinguishes this extreme-right party from the others. We see that, more so than other voting groups, they perceive a failed national economy, a malfunctioning democracy, and a materialist society. In different surveys around the same time, Nonna Mayer (1988, p. 44) discovered that, besides the immigration issue, Le Pen supporters also were deeply concerned about crime and unemployment. Overall, these survey responses paint a picture of a set of French voters gravely troubled with contemporary social currents, in particular that of immigration. The FN party leader, Jean-Marie Le Pen, has made the most of these fears. In his 1988 presidential campaign he stoked them (as quoted in Lewis-Beck and Mitchell, 1993, pp. 112–13). Speaking of the nation, he placed “The French first!” With regard to crime, he charged that present laws favor “attacks on old men more than the disciplining of hoodlums.” On the subject of immigration, he called for preservation of “the national identity” and protection against “the menace” of the “demographic explosion of the Third World.” He claimed the Nazi genocide of the Jews was a mere “detail.” His solution for AIDS (Sida) was “sidatoriums.” As a defense for these positions, he offered that his was the voice of the people, telling “out loud what the French think quietly.” At the presidential ballot box, these appeals have played well. The FN has come to replace the Communists in the “gang of four” parties that dominate the electoral theater. In 1995, for example, here are the firstballot shares for the six leading candidates: Jospin (Socialist), 23.3 percent; Chirac (RPR), 20.8 percent; Balladur (RPR), 18.6 percent; Le Pen (FN), 15.0 percent; Hue (Communist), 8.6 percent; Voynet (Ecologist), 3.3 percent. One observes that Le Pen took fourth place, not far behind Prime Minister Balladur. In the FN’s biggest success to date, Le Pen placed second in the first round of the 2002 contest, with 17 percent of the valid votes, unexpectedly edging out the Socialist prime minister Jospin. As

9780230_300088_06_ch05.indd 101

6/28/2011 8:15:15 PM

102

French Presidential Elections

already noted, much of that vote was driven by immigration fears and surrounding concerns. Here is the explanation offered by a 40-year-old wine trader from Brittany: “Le Pen won because of a configuration of circumstances – low voter turnout, a mobilized Le Pen vote, and people on the left who just thought they would show Jospin they disapproved of him on the first round, then vote for him on the second. A Frenchman who works all day, leaves his shop and sees a car set fire by immigrants, he is not a racist. He just wants something done.” (Lewis-Beck, 2004, p. 2). Such sentiment suggests something larger at work, beyond the immediate issue of immigration itself. Jocelyn Evans (2004) hints that the 2002 Le Pen surge can be attributed to the intensification of party identification among FN supporters. In similar fashion, Robert Elgie (2004) contends that the rise of extreme-right ideology itself made the difference. These speculations of Evans and Elgie reinforce the notion put forward by Lewis-Beck and Mitchell (1993) that commitment to the FN rests on a deep ideological extremism. Certainly, FN voters have exhibited more stability over time in their voting than have other partisans, suggesting the emerging exogeneity of their extreme-right views. Conceivably, votes for the FN could increase, if major parties are unable to solve immigration-related issues. However, the 2007 presidential contest does not bear out this possibility. As Pascal Perrineau (2009, p. 201) observed, “one of the essential elements” of that vote was “the astonishing erosion of the vote in favor of Jean-Marie Le Pen,” which fell from 17 percent of the valid first-round 2002 votes to 10 percent of the valid first-round 2007 votes. Perrineau goes on to comment that Le Pen in 2007 recovered less than 40 percent of his 2002 supporters. This “défidélisation,” as Perrineau calls it, largely benefitted Sarkozy, as we shall see more of below (Perrineau, 2009, p. 203). In a sense, this brings the anti-immigrant movement full circle, reminding us of the strategic crisis in 1983 Dreux, with the traditional right arguing for a secondround alliance with the FN to defeat the left.

5.5 Immigration and the French voter: bivariate relationships Fortunately, attitudes on the immigration issue are comparably and cleanly measured across our four presidential election surveys. For 1988, it is a seven-point scale of unfavorable to favorable opinion,

9780230_300088_06_ch05.indd 102

6/28/2011 8:15:15 PM

The Role of Issues 103

from “it is necessary that immigrants return to their home country” to “it is necessary to integrate the immigrants that currently live in France into French society.” For 1995, 2002, and 2007, it is always a response to a four-point scale of favorable to unfavorable opinion on the statement “There are too many immigrants in France.” (For each election, the scales were rescored 0 to 1.) In Table 5.4 we observe the bivariate relationship between opinion of immigrants and firstround vote choice. We focus on 2007, since it nicely illustrates the draw of the immigration issue for each party candidate and for Le Pen in particular. As expected, those who held a “highly unfavorable” opinion of immigrants were 25 times more likely to vote for Le Pen (i.e., 18/0.7 > 25). However, we see that strong anti-immigrant feeling did not exhaust itself with the Le Pen voters. A Sarkozy vote was clearly more likely when the voter disliked immigrants (percent extremely unfavorable – percent extremely favorable = 44 – 17 = 27 percentage point difference). Table 5.4 Relationship between attitude toward immigrants and 2007 presidential vote, first round (%) Highly favorable

Somewhat favorable

Somewhat unfavorable

Highly unfavorable

Laguiller, Buffet, and Besancenot Royal Bayrou Sarkozy Le Pen N Cramer’s V Pearson’s χ2 Gamma

14

10

8

6

44 25 17 1 949 .24 .00 .39 (.02)

31 29 28 2 831

21 17 46 8 941

20 12 44 18 599

Kendall’s tau-b

.29 (.01)

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. The immigration question is as follows: “There are too many immigrants in France.” 0 = not at all in agreement (highly favorable); .33 = somewhat disagree (somewhat favorable); .67 = somewhat agree (somewhat unfavorable); 1.0 = completely in agreement (highly unfavorable). Source: CEVIPOF (2007).

9780230_300088_06_ch05.indd 103

6/28/2011 8:15:15 PM

104

French Presidential Elections

In contrast, a vote for Royal was clearly more likely when the voter liked immigrants (percent extremely favorable – percent extremely unfavorable = 44 – 20 = 24 percentage points difference). Interestingly, the immigrant effect here, negative versus positive, balances out almost evenly for voters choosing the moderate right versus the moderate left candidate (i.e., the percentage differences of 27 and 24 are quite close). Overall, the relationship between the two variables is moderate, V = 0.24. Furthermore, this relationship barely changes from election to election. Here are the calculated associations, by election year: 1988, V = 0.23; 1995, V = 0.24; 2002, V = 0.31; 2007, V = 0.24. Looking at all four elections, and relating immigration opinion to first-round presidential vote, the average V = 0.26. In general, then, the bivariate relationship is modest and not dissimilar to that uncovered with economics and the vote, discussed earlier. Of course, we have yet to look at these relationships in a multivariate context. This is now easy enough, given the theoretical and empirical background we have covered.

5.6 Economics and immigration: multivariate relationships As we know, presidential vote choice is a function of several forces, both long-term and short-term. In previous chapters, we examined the long-term social factors of age, gender, education, religion, income, occupation, and patrimony, and a long-term psychological anchor: ideological identification. In this chapter, we examine our first set of short-term factors – issues. Of course, because issues do not operate on the voter independently, we must embed them into a multivariate model specification that controls for these long-term factors. Thus, theoretically, we posit the following explanation of the French presidential vote: Vote = social factors + ideological identification + issues. We observe that this model simply adds our current issue variables (economics, immigration, and Europe) to the multivariate model estimated at the end of Chapter 4. These new logistic regression estimations appear in full in the Appendix, for both the first and second rounds.

9780230_300088_06_ch05.indd 104

6/28/2011 8:15:16 PM

The Role of Issues 105

Following our customary strategy, estimation on first-round choice consists of a series of binomial logistic regressions, on the five leading candidates, left to right. For each issue variable, this produces 20 slope coefficients over the four elections. For the issues of economics and immigration, then, this produces a total of 40 coefficients. It is useful to summarize these results before getting into more detail. First, these issues usually have a statistically significant impact, regardless of the candidate. Out of the 40 coefficients, 28 pass conventional significance levels. With respect to one issue versus the other, 12 of the economics coefficients reach significance, while 16 of the immigration ones do. These results are important, establishing that these central issues matter to the French voter, across the political spectrum. Furthermore, they matter even when the powerful controls of social factors and ideology are imposed. In other words, the French voter is not a simple victim of deep social and psychological forces essentially beyond immediate change. Explore the economics coefficients more carefully. One observes that the classic economic voter hypothesis receives broad support. That is, hypothesize the political executive, responsible for managing the national economy, receives punishment from voters when economic times are bad and receives reward from voters when economic times are good. This executive, sometimes called the political economic incumbent (PEI), is either the president (under unified government) or the prime minister (under cohabitation) (LewisBeck, 1997b). Then expectations are met in three of these contests, as follows: 1988, vote for Chirac (PEI) if see good times, b = 0.95; 1995, vote for Balladur (the PEI) if see good times, b = 1.04; 2007, vote for Royal (the PEI opponent) if see bad times, b = -0.27. Given the wide array of candidates presented in the first round of a typical Fifth Republic presidential election, the ability of voters to discern political economic responsibility, and act on that discernment, is impressive. There are other patterns of interest with regard to the economic vote. The fact that there is no significant economic voting for the candidate from the center (save in 1995 when that candidate was actually PM) underscores the valence character of the economic issue. A centrist candidate, or party, typically stands outside of government, and so is unable to lead the economy. A centrist vote to foster economic prosperity, then, simply makes no sense. However,

9780230_300088_06_ch05.indd 105

6/28/2011 8:15:16 PM

106

French Presidential Elections

this is not to say that smaller parties experience no economic voting. Observe that in three of the four elections the extreme-left candidate received significantly more votes from those who perceived economic trouble. The vote for the extreme-right candidate, Le Pen, was likewise significantly influenced in three of the four elections, suggesting the truth of the notion that his supporters are right-wing economic protesters. Accordingly, the Le Pen vote was higher among those who believed that the economy was not doing well (1995) and lower among those who believed that unemployment (a left issue) was a top priority (2002 and 2007). Overall, though, the presence of any statistically significant economic voting for these smaller parties indicates a caution against assuming they are merely one-issue parties. What about the immigration issue? Several points are worth noting. First, to the surprise of no one, it is always a significant factor in the vote for Le Pen. Further, it is no surprise that its impact was strongest in 2002. Compare the magnitude of the slope coefficients across elections: 1988, 3.69; 1995, 3.84; 2002, 5.14; 2007, 2.53. Noteworthy about this series is the drop in the coefficient in 2007. In fact, that 2007 coefficient stands as the lowest in the series, so pointing to the changing role of the immigration issue in French presidential elections. That is to say, voters who believe there are too many immigrants have been increasingly drawn to the traditional right. Observe the pattern of coefficients for the moderate-right candidate (Chirac three times, Sarkozy once): 1988, –0.77 (not significant); 1995, –0.44 (negative and significant); 2002, 0.66 (positive and significant), 2007, 0.65 (positive and significant). In the earlier elections, those fretting over immigration directed their attention away from the moderate right. In the later elections, they have directed their attention toward it. In both 2002 and 2007, anti-immigrant sentiment was an important component of Chirac and Sarkozy support in the first round, not to mention the second. This result emphasizes the contemporary dilemma (opportunity) for the traditional right, and it helps explain the reduced support for the FN itself. What of the second-round results? These are more easily summarized. Seven of the eight coefficients are in the expected direction, and six of eight are statistically significant. Here are the magnitudes of the coefficient for the economy variable: 1988, 1.80; 1995, 0.92; 2002, 0.65; 2007, –0.01. And, for the immigration variable: 1988,

9780230_300088_06_ch05.indd 106

6/28/2011 8:15:16 PM

The Role of Issues 107

0.24; 1995, 1.00; 2002, 1.73; 2007, 1.38. A number of conclusions emerge. First, these issues are almost always significant in the decisive second round of the French presidential election, even in the face of the strong competing forces of social factors and ideology. Again, French voters do look at issues, setting themselves apart from what may be their deeper convictions. Second, as expected, a good economy generally favors the PEI. Third, the immigration issue generally works to the advantage of the moderate-right candidate in this final duel.

5.7

Adding Europe to the equation

We have set aside the question of Europe for special treatment. Since the idea of European integration first surfaced, French political opinion on it has been divided. With the dawn of the Fifth Republic, elite opinion gravitated toward the idea, mostly because the immensely powerful President de Gaulle decided the Common Market would help French agriculture and provide a French forum to counter Soviet and American influence. Across the 1960s and 1970s, with the Gaullists dominating national politics, open European opposition came mostly from the Communists, who would occasionally check themselves in attempts to establish left unity with the Socialists. During this period, into the 1980s, France took on the leadership role for bringing about European integration, culminating in the Single Market Program (1985). When Mitterrand came into the office of president, he undertook domestically to blunt the force of the Communists and, internationally, to assert French preeminence on the continent. His machinations had little public visibility, as most of the work took place within European Union institutions. In particular, via the European Monetary Union (EMU), it aimed to wrest financial direction away from the Germans. (For a useful review of this period, see Ross (2000).) Against this background, Mitterrand called for a French vote on the Maastricht treaty, which would deepen French commitment to the construction of Europe. (See the discussion in Lewis-Beck and Morey (2007).) On September 20, 1992, the French narrowly ratified the treaty, giving it a petit oui (small “yes”) with only 51 percent of the valid votes. It is worth considering how this vote came to pass. First, President Mitterrand was striving to assert the French

9780230_300088_06_ch05.indd 107

6/28/2011 8:15:16 PM

108 French Presidential Elections

European presence, with him as the driving force. Second, he felt that a strong pro-treaty vote would strengthen the Socialist Party in the upcoming legislative elections. The electoral consequence of this political push was that the issue of Europe moved onto the voter’s agenda. Heretofore, Europe had largely been a concern at the elite level, where party leaders there maintained a delicate consensus. This consensus was rent asunder when the issue “went public.” (See the valuable discussion in Appleton (1996).) Divisions within the traditional right were exposed: Charles Pasqua and Phillipe Séguin were hostile, Edouard Balladur was for it, and Jacques Chirac could not decide. The leaders of the UDF, however, were well behind it. On the extreme right, the National Front was staunchly against. On the traditional left, there was discord. JeanPierre Chevènement was much against. The Greens, after an internal quarrel, took no stand. The Communists, true to their colors, were virulently opposed. The heavy partisan coloration given to the idea of Europe in the Maastricht debate spilled over into the 1992 vote and subsequent relevant votes, such as the referendum on the 2005 European Union Constitution. While that referendum did not pass, the balance of forces going into the decision was similar. Until it got on the ballot, the issue of Europe had not been high on the electoral agenda. For instance, in the 2002 presidential election study, European unity ranked eleventh in importance, when compared to other issues (recall Table 5.2). However, with the vigorous pro-Constitution campaign of President Chirac in 2005, it climbed to the top of the agenda. Once again, the public was divided along partisan lines familiar to the pattern of 1992. And, as we shall see below, this pattern tends to repeat itself in the actual presidential elections over the period. Before looking at those data, however, it is useful to provide a brief update on the status of the European issue in France today. Bruno Cautrès and Dominique Reynié (2001, p. 87) write of the “Construction européenne: le mythe du grand projet.” At least as an issue on the voter’s agenda, it might indeed appear as something of a myth. In Table 5.5, we report how many voters ranked the construction of Europe as the first or second “most important problem,” according to the presidential elections of 1995, 2002, and 2007. As can be seen, very few respondents (never more than 14 percent) rank

9780230_300088_06_ch05.indd 108

6/28/2011 8:15:16 PM

The Role of Issues 109

Table 5.5 problem

Perception of the construction of Europe as an important 1995

2002

2007

It’s the no. 1 problem It’s the no. 2 problem

14% 6%

2% 3%

1.5% 3%

N

3827

4107

4004

Notes: For 2002 and 2007: the percentage of respondents mentioning Europe as the most, or the second most, important issue when they voted. For 1995: the percentage of respondents giving Europe the highest or the second-highest score (9 or 10) on a scale going from 0 to 10 measuring the importance of a series of issues at the time of the first-round vote. Source: CEVIPOF (1995, 2002, 2007).

it as the primary problem. Indeed, it would appear to be an issue simply not on most voters’ radar. For those voters, it would fail to meet our first condition of issue voting: the voter must perceive the issue. There are also difficulties with meeting the second condition: the voter must have a preference on the issue. In Table 5.6, we observe (for 2002 and 2007) the percentage who thought the presence of “France in the European Union” was “neither good nor bad.” For 2002, that share registers 27 percent; by 2007, it rises to 39 percent. In other words, many voters simply do not have a preference concerning whether France should or should not be in the EU. Without such a preference, the issue provides no signal for them in the voting booth, even if the candidates differ on the issue, which, in fact, they do. As discussed above, the candidates themselves are sharply divided on the question of Europe. How does all this ultimately play out in the vote? What happens when we examine the effects of the Europe variable within the issues set and alongside the prior variables from our earlier chapters in our logistic regressions? For 1988, we have no measure of the Europe variable. However, for 1995 we measure it by evocation of the word “Europe”; for 2002 and for 2007 the measure is “opinion about France in the EU.” (More details appear in the Appendix.) Based on the contents of party leader pronouncements, as discussed above, we pose the following hypotheses.

9780230_300088_06_ch05.indd 109

6/28/2011 8:15:16 PM

110

French Presidential Elections

Table 5.6 Distribution of opinion on “France in the European Union”

Positive (%) Neutral (%) Negative (%) No answer (%) N

2002

2007

68 27 5 0

51 39 8 1

1822

4006

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Source: CEVIPOF (2002, 2007).

H1: A pro-Europe attitude positively relates to the vote for the moderate-left candidate. H2: A pro-Europe attitude positively relates to the vote for the moderate-right candidate. H3: The positive association of H1 exceeds the positive association of H2. H4: An anti-Europe attitude positively relates to the vote for the extreme-left candidate. H5: An anti-Europe attitude positively relates to the vote for the extreme-right candidate. H6: The positive association of H4 exceeds the positive association of H5. We first test these hypotheses against the 1995 contest, the vote closest in time to the 1992 Maastricht treaty vote. All hypotheses are confirmed. Pro-European thinking significantly benefited both the moderate-left candidate, Jospin, and the moderate-right candidate, Chirac. Furthermore, the benefit for Jospin was greater (i.e., 1.53 > 0.92), as would be expected given the long-term Socialist commitment to Europe. Also, anti-European thinking significantly benefited the extreme-left candidates, Laguiller or Hue, and the extreme-right candidate, Le Pen. As well, the benefit for the extreme left was greater (i.e., 1.64 > 1.09), which is to be expected given the historic Communist commitment against. What about 2002? Again, the hypotheses are confirmed, except for H6. Moderate-left and moderate-right candidates continued to

9780230_300088_06_ch05.indd 110

6/28/2011 8:15:16 PM

The Role of Issues 111

benefit from pro-European sentiment, with the same asymmetry. Extreme-left and extreme-right candidates continued to benefit from anti-European sentiment; however, the Le Pen effect now exceeds that for Laguiller, Hue, or Besancenot (i.e., 1.73 > 1.15). For 2007, we see a continuation of the 2002 pattern of sign and significance. Moreover, the extreme right now benefits more than the extreme left, as a function of anti-European sentiment (1.54 > 1.42, again turning over H6). As well, we see a turning over of H3, since the moderate-right coefficient now exceeds the moderate-left coefficient (0.91 > 0.56). These last changes, with the right (both moderate and extreme) benefitting more from European attitudes than the left (both moderate and extreme) bear remarking. An implication is that the right is coming to “own” the issue of Europe, struggling to take it away from its past owner, the left. Much of this complexity disappears in the second round of presidential voting. First, for two of the three elections, the European issue simply fails to register statistical significance. This is not especially surprising given the expectation that in this contest, normally fought between the leader of the moderate right and the leader of the moderate left, the candidates themselves would not meaningfully differ on the issue of Europe. That is, condition number three for issue voting – the candidates offer differ solutions – may well not be obtained, with both of the candidates sharing in the elite, pro-EU consensus prevailing. However, one does observe that in 2007 attitude toward Europe was significant, with those more pro-Europe more likely to support Sarkozy. This current result implies a shifting tide, with the EU mantle passing from left to right. Will the Socialists “lose” this issue, which they have traditionally owned? The 2012 contest may tell the tale.

5.8 Conclusion Before they can influence the vote, issues must meet a set of rather stringent conditions. We examine many issues for their potential relevance in French presidential election choice. There appear to be exemplary candidates: economics, immigration, and Europe. The first two are important domestic issues, while the last stands as a critical international issue. All are enduring concerns and are regularly measured across this set of elections. Further, they represent

9780230_300088_06_ch05.indd 111

6/28/2011 8:15:16 PM

112

French Presidential Elections

variety in issue types: valence versus positional, easy versus hard. We find that these issues almost always matter in presidential vote choice, even after the imposition of extremely strong statistical controls. French voters, regardless of social circumstance or ideological disposition, tend to punish the government for a bad economy. Further, ceteris paribus, their attitudes on immigration, to the extent they are unfriendly, sway them to the right. With respect to Europe, if they are pro-EU, they incline to the traditional left, independent of other concerns (although the 2007 results suggest this might be changing). While these are generalizations, they are worth making. Of course, the particular mechanics of their operation may vary a bit from one election to another or from one round of voting to the next. Nevertheless, these short-term forces, embedded within those of the long-term, are persistent across electoral cycles, relentlessly shaping in their dependable way the political behavior of the French electorate.

9780230_300088_06_ch05.indd 112

6/28/2011 8:15:17 PM

6 Leader Image

What is the influence of a candidate’s image on voting in French presidential elections? There have been surprisingly few attempts to answer this question over the past few decades. This is despite the fact that historic leaders such as General de Gaulle have left a lasting imprint on French politics (Hoffman, 1974). According to some, the French electoral system seems to be “candidate-centered in ways that create multiple opportunities for aspirants to the office to try to exploit whatever leadership attributes they may possess” (Pierce, 2002, p. 96). Recent institutional reforms, such as modifying the electoral calendar, have brought increasing “presidentialization” of the French political game by diminishing the importance of legislative elections (Fauvelle-Aymar, Lewis-Beck, and Nadeau, 2010). Data collected from each election on issues, and on leaders, seem to suggest that these factors have more weight today than long-term factors such as ideology, social class, and religion. How can we explain the contrast between the little attention paid in France to the effect of a leader’s image on voting, and the abundance of work on this same theme in other democracies? Numerous studies have focused on the United Kingdom (Clarke et al., 2004; Crewe and King, 1994), the United States (Kessel, 2004; Wattenberg, 1998), Portugal (Costa Lobo, 2008), Canada (Blais et al., 2002), and Australia (Bean and Mughan, 1989); more general analyses around the theme of the growing “personalization” of the electoral dynamic have notably been carried out by Curtice and Holmberg (2005) and by Webb and Poguntke (2005).

113

9780230_300088_07_cha06.indd 113

6/29/2011 10:37:09 AM

114

French Presidential Elections

A few explanations can be suggested. The first explanation, although seemingly paradoxical, speaks to the larger-than-life stature of Charles de Gaulle. For many, his retirement from political life marked a return to normalcy, putting to an end the exceptional ascendance of a leader in French political life (Pierce, 2002, p. 110). This has contributed to reinforcing a more sociological orientation in French electoral studies. Furthermore, the very limited number of indicators in the major French electoral studies touching on this subject has accentuated the decline in studies on leader image. (On this question, see Ysmal (1994), Nadeau and Lewis-Beck (2010), and LewisBeck and Nadeau (2010).) This conundrum sets the stage for this chapter. We will first survey the small number of studies that have systematically examined the impact of leader image on vote choice in French presidential elections. We will then try to make use of available indicators to offer a few larger conclusions on this topic. The conclusions from these analyses are clear: leader image is an important determinant of vote choice in French presidential elections. In this context, it is hoped that future large-scale French electoral studies allocate a major place to this question.

6.1

Leader image in French electoral studies

General de Gaulle left a deep impact on French political life. Therefore, it is not surprising that the few works done on leader image in France have emphasized the impact of the former leader of la France libre. A pioneering study by Converse and Dupeux (1966) established a suggestive parallel between two victorious generalsturned-politicians: de Gaulle and Eisenhower (see De Gaulle and Eisenhower: The Public Image of the Victorious General). Hoffman (1974) spoke of the era of “heroic leaders” to characterize de Gaulle’s impact. Pierce attributed an exceptional role to de Gaulle when he wrote, “More than any political leader in a Western Democracy, de Gaulle was the epitome of what most people have in mind when they refer to a candidate’s distinctive style of leadership and personal appeal as a source of electoral support” (Pierce, 2002, p. 97). De Gaulle’s impact on researchers’ understanding of the role of leaders is therefore paradoxical. On the one hand, the exceptional rise of this politician seems to be proof that leader image can be a

9780230_300088_07_cha06.indd 114

6/29/2011 10:37:10 AM

Leader Image

115

determinant for many voters; on the other hand, because of its explicitness, the “de Gaulle effect” seems to be an exception, nothing but a peculiar historical occurrence in a political system that remains largely dominated by heavy factors such as social class, religion, and ideology. This situation could explain why studies on the role of leaders have been relatively rare in France until recently. In fact, save for the analyses of particular elections (for example, those of Pierce on the 1995 election) and work done on leaders in the realm of political communication (Cayrol, 1988; Johnston and Gerstlé, 1995; Maarek, 2000), only two studies have systemically examined the impact of leader image on French presidential elections. The first of these studies was carried out by Roy Pierce (2002), looking to see if candidate image in and of itself could have a significant impact on the result of a French presidential election from 1965 to 1995. Pierce’s conclusion is nuanced; the results of two elections, those of 1965 and 1969, were clearly influenced by leader image (that of de Gaulle in 1965, and Pompidou to a lesser degree in 1969). The tight elections of 1974 and 1995 might have possibly been influenced by the relative advantage of right-wing candidates Giscard d’Estaing and Jacques Chirac, as both were perceived as having more of “the right stuff” (l’étoffe présidentielle) to be president than their main leftwing opponents (Mitterrand and Jospin). (On this question, see also Lindon and Weill (1974); Lancelot (1975).) However, this is not the case for the 1981 and 1988 elections, which were, according to Pierce, dominated by heavy factors, such as ideology. Nadeau and Lewis-Beck (2010)’s study is different from Pierce’s on multiple accounts. Their study deals with more recent elections (1988 to 2007), is more interested in the impact of leader image on vote choice, examines this issue for the first and second rounds of voting, and rests on substantive and comparable multivariate analyses for the elections studied. The results show a systematic and important link between voter sentiment toward party leaders – measured with the help of feeling thermometers, explained below – and the vote choice of partisans belonging to the general political families of moderate left, center, moderate right, and extreme right. The study also shows that feelings toward party leaders are explained not only by ideological proximity between a voter and a leader, but also by perceptions of certain attributes such as competence and honesty.

9780230_300088_07_cha06.indd 115

6/29/2011 10:37:10 AM

116

French Presidential Elections

At last, the duel between Nicolas Sarkozy and Ségolène Royal seems to have revived interest among French researchers in the impact of leader image on voting. Making use of a number of indicators, Boy and Chiche (2009, p. 102) conclude that “the fact of positively assessing the qualities of a candidate constitutes a determining factor to voting for that same candidate” and that “the importance of image in electoral choice is more or less equal to, and sometimes greater, than the importance of placement on a left-right axis.” The few studies on the effect of leader image for voting seem to be conclusive. The intersection of existing work indicates that this variable has exercised a more than negligible impact on a good number of elections, including those of 1965, 1969, 1974, 1995, and 2007. These same analyses suggest that the influence of this factor in electoral decision-making is not only significant, but sometimes even dominant. These findings urge us to undertake a more systematic examination of leader image and its impact on recent presidential elections in France.

6.2 Components of leader image The classic work of Campbell and his colleagues (1960) on electoral behavior offers a useful framework for examining the effect of leader image on voting. As explained in Chapter 1, these authors propose a model (labeled the “Michigan model”) that distinguishes between long-term (for example, party identification in the American context) and short-term factors that are likely to influence vote choice. The latter category reflects the specific context of each election. While parties, or large ideological orientations, are constant (or are at least persistent for long periods of time), leaders and issues are more ephemeral. The specific effect of leader image must therefore be established after we control for other long-term factors (for example, ideology, religion, and social class), as well as short-term factors (such as the economy) that exercise an impact on voting. Studies on the effect of leaders on voting have evolved over time. The first studies emphasized the idiosyncrasies of each leader by postulating the “unique” character of their personality (Miller et al., 1986). This approach gradually gave way to the concept of image. That is, the qualities necessary to exercise effective political leadership, and the attributes of a leader, are interrelated components of a

9780230_300088_07_cha06.indd 116

6/29/2011 10:37:10 AM

Leader Image

117

leader’s image. These qualities, called “traits” in Anglo-Saxon writings, have a universal character. The perceived attributes of a good political leader are stable across time and widely dispersed throughout the electorate. In the eyes of voters, what are the qualities of a good leader? Three qualities stand out in many works, especially American and British, devoted to this question (most notably, see the studies of Fiorina (1981), Kinder (1986), Miller et al. (1986), Funk (1997), Clarke et al. (2004), Kessel (2004), and Lewis-Beck et al. (2008)): these qualities are competence, strength, honesty, and empathy. Usage of these dimensions does vary from one study to another. Competence can be associated with experience, knowledge of issues, or the perceived performance of an outgoing incumbent. Empathy can be measured by the capacity to listen, or by the feeling that a candidate is able to understand citizens’ problems. Nonetheless, these three criteria seem to offer a satisfactory image of a good political leader. The studies on voter expectations of candidates rely on perspectives from cognitive psychology. According to this approach, the voter sees if a candidate possesses the required qualities of a leader, be it through an intensive search for information or by relying on impressions or even anecdotal events. (For a discussion of decisional shortcuts used to evaluate candidates, see Popkin (1994).) Another vein of research is interested in the interaction between the emotions candidates evoke in voters and the reasoning that leads these voters to either support or reject these candidates (see, notably, Marcus and MacKuen, 1993; Marcus et al., 2000; Nadeau et al., 1995). According to this approach, the emotions felt – anxiety or enthusiasm – act as triggers to decisional processes informing candidate choice (see also Boy and Chiche, 2009). What can be said about the measures used to assess candidate image in the eyes of voters? Two methods have been used (see LewisBeck and Nadeau, 2010). The first, that of the thermometer, consists of asking voters to gauge the extent to which they like or dislike a candidate on a scale that generally runs from 0 to 100, where 0 corresponds to a strong disliking and 100 corresponds to a strong liking. The other method asks respondents if a statement applies more or less to a particular candidate. These statements can touch upon an aspect of a candidate’s image (“He/she is honest”), as well as the emotions that they evoke (“He/she worries you”).

9780230_300088_07_cha06.indd 117

6/29/2011 10:37:10 AM

118

French Presidential Elections

Have French electoral studies systematically made use of the theoretical approaches and empirical methods that we have just described? The answer to this question is, largely, no. Certain personality traits of candidates have been measured more systematically in some studies than in others. Hence, the data collected have led to proposals that the victories of Giscard d’Estaing in 1974 and Jacques Chirac in 1995 could be explained by the capacity of these candidates to convince critical segments of the electorate that they had “the right stuff to be president of the Republic,” at least more so than their adversaries (see Lindon and Weill, 1974; Pierce, 2002). However, the limited number of indicators that measure candidate image in CEVIPOF studies is striking (see Lewis-Beck and Nadeau, 2010; Ysmal, 1994). The 1988 and 1995 surveys contain no questions related to any aspect of candidate image (competence, strength, honesty, or empathy). However, the 2002 survey can be seen as a small breakthrough. In the first round of elections, it asked respondents whether or not the qualifiers “honest,” “has the right stuff to be president,” and “keeps his/her promises” applied to Chirac and Jospin. In the second round, it asked respondents whether or not the qualifiers “honest,” “has the right stuff to be president of the Republic,” “brings new ideas,” and “is close to the people’s concerns” applied to Chirac and Le Pen. The matchup between Sarkozy and Royal seems to have revived interest in candidate image among French researchers (see the work of Cautrès and Muxel, 2009). These more recent studies use a larger variety of indicators and measure them over a larger number of candidates across multiple panel waves. When the first panel wave was conducted before the first round, the following question was asked: “For each of the following sentences, tell me if it applies very well, fairly well, not very well, or not well at all to [Nicolas Sarkozy/ Ségolène Royal/Jean-Marie Le Pen/François Bayrou]: He/she worries me / has the right stuff to be president of the Republic / is honest / really wants to change things / understands the problems of people like you.” In the second wave, conducted between rounds, the same questions were asked about the two qualifying candidates in the second round: Nicolas Sarkozy and Ségolène Royal. The same battery of questions was applied to the winning candidate after the second round (only the qualifiers “really wants to change things” and “understands the problems of people like you” were measured for

9780230_300088_07_cha06.indd 118

6/29/2011 10:37:10 AM

Leader Image

119

Royal). Finally, the post-electoral questionnaire in the 2007 study contained three questions that determined if voters’ perceived qualities of the two main candidates had any influence on their vote choice or on the result of the election. These questions read as follows: “With regard to the second round, you chose your candidate [Royal or Sarkozy] because (give one answer) you have trust in him/ her / because he is the closest to your ideas / because you want to stop another candidate”; “Would you say that you are somewhat in agreement or somewhat in disagreement with the following sentences regarding the defeat of Ségolène Royal [victory of Nicolas Sarkozy]: Ségolène Royal was not a good candidate / Nicolas Sarkozy was a good candidate.” Is it possible to make use of this sundry information to examine the image of French presidential candidates over time? This task is easier for the studies conducted in 2002 and 2007, as the available indicators for these elections overlap with the majority of the dimensions identified in the literature on this subject. The question of candidates’ étoffe (stuff), an interesting indicator developed by French political scientists, refers to their competence and the strength of their leadership, which are two fundamental criteria voters use to determine if a candidate is “presidential” or not. The relationship between the quality of the candidates and vote choice is also an interesting (albeit more indirect) indicator of these dimensions. The perceived honesty of candidates, a crucial aspect of their image, was measured either directly (“Is the candidate honest?”) or indirectly (“Does the candidate keep his/her promises?”; “Did you choose him/her because you trust him/her?”) in the 2002 and 2007 studies. Empathy, was also measured in the 2002 (Le Pen, Chirac) and 2007 (Royal, Sarkozy, Bayrou, Le Pen) studies, using similar questions (“Is the candidate close to people’s concerns?” (2002); “Does the candidate understand the problems of people like you?” (2007).) As mentioned, unfortunately the 1988 and 1995 CEVIPOF studies give us nowhere near as much data regarding candidate image as the later studies. The measurement of leader image and its impact on voting therefore poses a real challenge in the French context. With no variables available in the 1988 CEVIPOF survey, we turn to Roy Pierce’s 1988 study (see Pierce, 1995). This investigation contains questions on satisfaction with the performances of President Mitterrand and Prime Minister Chirac, a measure that will be used in

9780230_300088_07_cha06.indd 119

6/29/2011 10:37:10 AM

120 French Presidential Elections

the next section as an indicator of a candidate’s competence. The 1995 CEVIPOF study contains few usable questions on leader image, apart from those measuring satisfaction with the performance of outgoing Prime Minister Edouard Balladur. Fortunately, there is a scale measuring feelings toward candidates available for each presidential election from 1988. For 1988, we again refer to Pierce’s study. However, the measurement of this variable has been erratic. Three different scales have been used (0 to 100 in 1988 and 1995, 1 to 10 in 2002, and 0 to 10 in 2007). Nevertheless, this variable, as measured, provides the most complete and comparable information on the largest number of French presidential candidates for the period under study (Lewis-Beck and Nadeau, 2010). The available data will be used in the next sections to paint a picture of candidate image and its influence on presidential elections from 1988 to 2007. The discussion will start with an examination of candidates’ scores on feeling thermometers. The dimensions of candidate image, specifically those of their competence, honesty, and empathy, will then be studied. The candidate scores will next be used in two ways: first as a dependent variable, then as an explanatory variable. In the first case, the objective will be to better understand the meaning of candidates’ scores on these scales. (Furthermore, this exercise will also serve as an antidote of sorts to Fiorina’s remark (1981, p. 154) that “no one knows what thermometer scores measure.”) Then, by introducing these scores into models that contain the whole set of variables examined in the four preceding chapters, it becomes possible to establish the relative weight of candidate image in explaining vote choice. (See Boy and Chiche (2009, table 4, p.112) for a similar exercise.)

6.3

Candidate image in French presidential elections

Table 6.1 offers an initial glance at French perceptions of the main presidential candidates from 1988 to 2007. First, we notice that the percentage of respondents who refused to tender a judgment on candidates from the main political families is very low – an average of less than 2 percent – even for those candidates with limited support. The main presidential candidates seem to be well-known among the electorate. The mean score on a scale from 0 to 100 is 44 (n = 23

9780230_300088_07_cha06.indd 120

6/29/2011 10:37:10 AM

Leader Image

121

Table 6.1 Thermometer scores for leaders in the 1988, 1995, 2002, and 2007 French presidential elections 2007 Besancenot Buffet 45 40

Royal 57

Bayrou 57

Sarkozy 58

Le Pen 21

2002 Laguiller 37

Besancenot Jospin 38 55

Bayrou 46

Chirac 51

Le Pen 13

1995

Laguiller 48

Hue 44

Balladur 50

Chirac 57

Le Pen 22

1988

Lajoinie 29

Chirac 47

Le Pen 20

Jospin 59

Mitterrand Barre 61 52

Note: The table shows the mean scores on the thermometer scales, rescaled from 0 to 100. Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

candidates) and 49 (n = 19 candidates) if we exclude voters’ assessments of Jean-Marie Le Pen. This mean score is slightly lower than what is considered to be a neutral candidate evaluation of 50. (For example, in the 1988 study, a score of 50 is explicitly presented in the following way: “100 means that you have a strong affinity, 0 means that you do not like the candidate at all, 50 means that you have neither an affinity nor a strong dislike, or that you do not know much about these candidates”). This overall average hides significant variation among elections and candidates. The weakest evaluations are observed in 1988 and 2002 (election means of 42 and 40, respectively) and the strongest ones are observed in 1995 and 2007 (means of 47 and 46, respectively). It is interesting to note that evaluations of candidates are in decline in 2002 (tracking a decline in electoral participation), and on the rise in 2007 (following a rebound from vote abstention). The unfavorable evaluations of Jean-Marie Le Pen, especially those from 2002 when he qualified for the second round, clearly stand out. They are not only noticeably lower relative to all the other candidates (mean of 19), but also discernibly more negative than the extreme-left candidates (mean of 40). Even the most unpopular candidate in this latter group, communist André Lajoinie, has an evaluation that is clearly more favorable (28) than the highest score that Le Pen ever had (22 in 1995).

9780230_300088_07_cha06.indd 121

6/29/2011 10:37:10 AM

122 French Presidential Elections

A few other interesting observations – some unexpected, others less so – can be deduced from an examination of the thermometer ratings of leaders. For example, it is hardly surprising to note that opinions of moderate candidates (left, center, and right) are usually more favorable (mean of 54) than those of more radical candidates. However, what is more surprising is that left-wing candidates, though defeated three times at the polls, have been more popular than their right-wing counterparts (means of 58 and 53, respectively). Perhaps surprisingly, the most well-liked candidate among the electorate does not necessarily win. That being said, the most popular candidate won two times – in 1988 and in 2007. Furthermore, the contrast between the political fortunes of two popular leaders, Lionel Jospin and Jacques Chirac, is remarkable. Jospin had the highest score of all the candidates in 1995 and 2002, but he could not be elected in the first case and did not even qualify for the final round in the second case. In contrast, Chirac managed to pull ahead of more popular candidates in the first round in 1988 (against Raymond Barre) and in the second round in 1995 (Lionel Jospin). Chirac, twice victorious at the polls, has never been the most favored candidate among voters in the three elections in which he was a presidential candidate. The preceding remarks show the utility of image scales for uncovering certain perceptions among the electorate regarding political leaders. However, the fact that a more favored candidate might not always be victorious emphasizes the importance of delving deeper into the role of image. Indicators measuring more specific aspects of image can complement this information. The data in Table 6.2 include the available information from French electoral studies regarding main components of a candidate’s image: competence, honesty, and empathy. The table consists of a set of percentages reporting the proportion of respondents who thought that a candidate possessed a desired attribute (for example, the “right stuff” to be president). For the 1988 and 1995 elections, the percentage of voters satisfied with Mitterrand, Chirac, and Balladur is interpreted as a measure of their perceived competence. The relationship between (1) perceptions of candidates’ competence and (2) those candidates’ eventual electoral success is the most relevant result that can be gleaned from Table 6.2. The winning candidate, in the three cases in which the data allow a direct comparison

9780230_300088_07_cha06.indd 122

6/29/2011 10:37:10 AM

Leader Image

123

Table 6.2 Opinions on leaders in the French presidential elections of 1988, 1995, 2002, and 2007 A. Competence and ability Good candidate (2007) Right stuff (2007) Right stuff (2002)

Royal 61 Royal 44 Jospin 51

Satisfaction (1995) Satisfaction (1988)

Bayrou 47

Sarkozy 73 Sarkozy 57 Chirac 74

Le Pen 19 Le Pen 11

Balladur 49 Mitterrand 57

Chirac 43

B. Integrity Honesty (2007) Honesty (2002) Keeps promises (2002)

Royal 58 Jospin 60 Jospin 41

Bayrou 60

Sarkozy 39 Chirac 29 Chirac 24

Le Pen 28 Le Pen 24

C. Empathy Understanding (2007)

Royal 56

Sarkozy 39

Note: The table shows the percentage of “agree” responses to various indicators measuring a candidate’s personality traits. Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

(1988, 2002, and 2007), is the one who was perceived by the largest number of voters to have the competence necessary to take the reins of the French state. It is interesting to note that the advantage that the eventual winner had in terms of projecting an image of competence is major in these three cases. In 1988, satisfaction with François Mitterrand was close to 15 points higher than satisfaction with Jacques Chirac. In 2002, the proportion of French who believed that Chirac had the right stuff to be president was substantially higher than for his opponent, Lionel Jospin (74 percent versus 51 percent). The results for the 2007 election are even more revealing, since neither of the candidates had occupied the presidency before. The

9780230_300088_07_cha06.indd 123

6/29/2011 10:37:10 AM

124 French Presidential Elections

eventual winner, Nicolas Sarkozy, distanced his opponent in the fight over competence – 57 percent of voters believed he had the right stuff to be president, versus 44 percent for Ségolène Royal. It is intriguing to observe, furthermore, that the ability of winners of presidential elections to project an image of competence does not necessarily mean that they will fare better than their opponents in terms of the other leadership qualities. For example, Ségolène Royal holds her ground against Nicolas Sarkozy in terms of listening to others and empathy: 56 percent of respondents agreed with the statement that she is “close to the concerns of the people” whereas 39 percent agreed for Sarkozy. This result is not all that unexpected. In general, left-wing candidates are perceived as being more sensitive to the needs of the population (this tendency has been observed in the United States (Lewis-Beck et al., 2008), in Great Britain (Clarke et al., 2004), and in Canada (Blais et al., 2002)). For example, François Mitterrand in 1974 was seen as being closer to the concerns of the French people (but less apt at occupying the presidency) than his right-wing opponent, Valéry Giscard d’Estaing (see Lancelot, 1975). Here French electoral behavior is on par with that of other large democracies. If the dominance of left-wing candidates on the empathy scale is predictable, the incapacity of right-wing candidates, despite having won the elections, to project an image of honesty and transparency also stands out. In this respect, the advantages of Jospin in 2002 and Royal in 2007 are huge; the percentage of voters who thought that these candidates were honest was clearly higher than that of their main right-wing opponents (60 percent against only 29 percent for Chirac, 58 percent versus 39 percent for Sarkozy; the 2002 results based on a more “harsh” indicator of keeping one’s promises show the same effect with a net advantage to Jospin (41 percent approval) over Chirac (only 24 percent approval).) Despite their limited character, these survey data provide some interesting information on candidate image in presidential elections. In 1988, the high ratings of Mitterrand relative to Chirac heralded a victory for the left. The winning candidates in 2002 and 2007 were perceived to be less close to the people (Sarkozy), less honest (Chirac and Sarkozy), and less reassuring (Sarkozy; 57 percent of respondents stated that the right-wing candidate worried them, versus only 35 percent for Royal). Nonetheless, these candidates were successful in

9780230_300088_07_cha06.indd 124

6/29/2011 10:37:11 AM

Leader Image

125

convincing a larger number of voters that they were more “presidential.” Since measures of competence seem to be more powerful than measures of honesty or empathy, this suggests the existence of a hierarchy of qualities for a good candidate in the eyes of voters. These results also imply the usefulness of an indicator measuring the demeanor (the right stuff/l’étoffe) of candidates. This concept, developed by French political scientists, seems to mix two essential qualities expected of a leader: competence and strength. Voters generally associate the former attribute with efficient management of the economy and the latter attribute with strong defense of national interests on the international scene. (See the analysis of Nadeau (2007) on this question, showing that Sarkozy fared better than Royal on these two dimensions in 2007.) The importance of leader image during presidential elections (Boy and Chiche, 2009; Nadeau and Lewis-Beck, 2010) strongly indicates that the next French election studies should invest in the fruitfulness of older indicators (such as having the right stuff to be president) and, in addition, build on the momentum given to image in 2007 (Lewis-Beck and Nadeau, 2010). That being said, the best available measure to examine the impact of a candidate’s personality on voting since 1988 is still that of the image scale, or thermometer. We will use this indicator in the next section in order to establish the weight of this factor in vote choice in presidential elections. Keeping in mind, however, the ambiguity surrounding this variable, we will first focus on candidates for whom we possess more complete information, in terms of different dimensions of image.

6.4 General and specific evaluations of presidential candidates What are French voters’ overall assessments of presidential candidates? Before proceeding to this analysis, a note on the measures is in order. Here our image measurement is based on a scale that asks respondents to place themselves somewhere between a minimum point (0 or 1, according to the study) representing the expression of strong antipathy toward the candidate and a maximum point (10 or 100, according to the study) representing a very strong sympathy toward the candidate. This question has been posed, with the noted variations, in CEVIPOF studies over the past three presidential

9780230_300088_07_cha06.indd 125

6/29/2011 10:37:11 AM

126

French Presidential Elections

elections. The question is also present in Roy Pierce’s study of the 1988 election and will be used in this section. (The thermometer in his study goes from 0 to 100; it is also the only case that explicitly states that a score of 50 represents a rather neutral evaluation of a candidate.) We believe that the variables that explain voters’ partisan choices will also color their evaluations of candidates. All things being equal, a voter who fits a right-wing profile (high level of income, sustained religiosity, and spontaneous identification with the right) will be more likely to have a favorable image of a candidate from the right, when compared to a voter having the opposite profile. This reasoning argues in favor of including, in a model of image scales, variables previously used to explain vote choice. It will be useful to add to these variables indicators measuring the qualities associated with a good leader: competence, honesty, and empathy. Results covering the determinants of image scale are presented in Table 6.3. The models corresponding to the first column consist of variables used in the previous chapters to explain vote choice, while those in the second column include specific indicators of candidate image. The analysis is limited to candidates for which this second group of variables is available. That being said, the results corresponding to the limited model for the other candidates are very similar to those presented in this section. The results of the first column show the existence of a projection effect between voters’ ideological preferences and their evaluation of candidates. There is a consistent and substantial link between the scores given to the candidates and self-positioning on a left-right scale (29 out of 32 coefficients corresponding to ideology are statistically significant). The use of ordinary least squares allows us to interpret these coefficients more easily. They represent the rise and fall of scores for a candidate in a given group (going from extreme right to extreme left) in relation to centrist voters’ evaluations. For example, the coefficients of 0.16 and –0.25 for François Mitterrand in 1988 mean that his scores are 16 points higher (on a scale from 0 to 100) among voters who classified themselves as on the extreme left and 25 points lower among voters who classified themselves as on the extreme right of the political spectrum (centrists constitute the reference category in these analyses).

9780230_300088_07_cha06.indd 126

6/29/2011 10:37:11 AM

Leader Image

127

The ideological polarization in candidate evaluations, consistent in the entire model (although less pronounced in the case of JeanMarie Le Pen), was particularly pronounced in the 2007 election. The main candidates clearly obtained higher scores among respondents who belonged to their same political family and systematically weaker scores among others. (For example, the coefficient for the moderate left for Royal is 0.21, but for the extreme left it is only 0.09; the coefficient for the moderate right for Sarkozy is 0.23, but for the extreme right it is only 0.16. See the first column, 2007 results, for the respective candidates, for the effects of ideology in Table 6.3.) The other group of influential variables is formed of salient issues: the economy, immigration, and the European Union. Here, the observed effects are consistent (19 statistically significant coefficients out of a total of 22). The impact of the immigration issue on JeanMarie Le Pen’s image is, unsurprisingly, very important (Le Pen’s score from respondents who completely agreed with the statement “there are too many immigrants in France” is, ceteris paribus, 26 percentage points higher than that given him by respondents who completely disagreed with the statement). It is also interesting to note the strong impact of a positive evaluation of the economic situation on the image of outgoing prime ministers Chirac in 1988 and Balladur in 1995. (The coefficients for these variables are 0.15 and 0.13 for these two candidates, a result showing that, at least under cohabitation, the French attribute more responsibility to the prime minister than to the president on this issue. See Chapter 5 as well as the pioneering study by Lewis-Beck (1997b) on this question.) The preceding results are interesting on more than one level. First, they show that the determinants of candidate evaluation are not the same as those for voting itself. For example, it is striking to realize that both socio-demographic variables (including religion) and classrelated variables have a less consistent impact on the evaluation of leaders than on their actual electoral support. These evaluations seem to be the result of a combination of voters’ central predispositions (their ideological orientations) and their more pointed evaluations of a candidate’s position and performance on the issues. Does the addition of more specific indicators of candidate image contribute to the explanation of candidate evaluation in a significant way? The results presented in Table 6.4 provide a preliminary

9780230_300088_07_cha06.indd 127

6/29/2011 10:37:11 AM

9780230_300088_07_cha06.indd 128

6/29/2011 10:37:11 AM

Age Male Education Religion Manager Employed Working-class Income Private-sector Non-risky asset Risky asset Extreme left Left Right Extreme right Immigration

–.03 (.04) –.01 (.02) –.04 (.03) –.01 (.03) –.05 (.02) –.04 (.02) –.03 (.02) –.06 (.05) .02 (.02) _ _ .16 (.05) .19 (.02) –.19 (.02) –.25 (.05) .03 (.03) *** *** *** ***

**

*

Mitterrand sig. –.06 (.03) –.02 (.01) –.04 (.02) –.01 (.02) –.02 (.02) –.04 (.02) –.01 (.02) –.05 (.04) .02 (.02) _ _ .02 (.04) .06 (.02) –.09 (.02) –.18 (.04) –.00 (.02) *** *** ***

**

**

*

Mitterrand sig.

1988

.04 (.04) .01 (.02) –.03 (.03) .05 (.03) .02 (.02) –.02 (.02) –.01 (.03) .02 (.05) –.00 (.02) _ _ –.25 (.06) –.15 (.02) .18 (.02) .07 (.06) –.03 (.03)

Chirac

*** *** ***

sig. .02 (.03) –.01 (.02) –.02 (.02) –.01 (.03) .04 (.02) .01 (.02) .02 (.02) .00 (.04) –.00 (.02) _ _ –.16 (.05) –.09 (.02) .08 (.02) .01 (.05) –.01 (.02)

Chirac

*** *** ***

**

sig. .12 (.02) –.01 (.01) –.03 (.02) .10 (.01) .01 (.01) –.01 (.01) –.01 (.01) –.01 (.03) .01 (.01) .05 (.02) .01 (.02) –.27 (.02) –.18 (.01) .14 (.01) .15 (.02) .05 (.01)

Balladur

*** *** *** *** ***

***

** ***

***

sig.

.09 (.02) –.02 (.01) –.05 (.01) .09 (.01) .02 (.01) –.00 (.01) –.01 (.01) –.02 (.02) –.01 (.01) .03 (.02) –.00 (.02) –.21 (.02) –.14 (.01) .11 (.01) .12 (.02) .04 (.01)

Balladur

1995

*** *** *** *** ***

*

*** * *** ***

sig.

Table 6.3 Regression models of thermometer scores of candidates, 1988, 1995, 2002, and 2007 (ordinary least squares method)

9780230_300088_07_cha06.indd 129

6/29/2011 10:37:11 AM

–.01 (.02) _ _ .68 (.05) .49 607 ***

–.00 (.02) _ .49 (.02) .96 (.04) .70 593 *** ***

.15 (.02) _ _ .40 (.05) .47 607 ***

***

.03 (.02) _ .44 (.03) .73 (.05) .63 583 *** ***

*

.13 (.01) –.11 (.02) _ .39 (.03) .44 2962 ***

*** ***

.07 (.01) –.07 (.01) .20 (.01) .35 (.02) .53 2962

*** *** *** ***

Source: Pierce (1988) and CEVIPOF (1995).

Note: The table shows non-standardized regression coefficients obtained through ordinary least squares estimation. The standard deviation is in parentheses. The thresholds for significance are as follows: * p < .10; ** p < .05; *** p < .01, two-tailed test.

Economy Europe Satisfaction Constant R2 N

9780230_300088_07_cha06.indd 130

6/29/2011 10:37:11 AM

Age Male Education Religion Manager Employed Working-class Income Private-sector Non-risky asset Risky asset Extreme left Left Right Extreme right

.10 (.04) .00 (.01) –.05 (.03) –.03 (.02) –.03 (.02) –.00 (.02) .03 (.02) –.00 (.04) –.02 (.01) .00 (.03) –.05 (.03) .12 (.03) .15 (.02) –.07 (.02) –.13 (.02)

Jospin

*** *** *** ***

*

*

*

**

sig. .05 (.04) –.05 (.01) –.03 (.02) .01 (.02) –.05 (.02) –.02 (.02) .04 (.02) .01 (.04) –.01 (.01) .01 (.03) –.03 (.03) .06 (.02) .08 (.02) –.03 (.02) –.06 (.03)

Jospin

1995

*

** ***

**

***

***

sig. .02 (.04) –.06 (.01) –.03 (.03) .09 (.02) .03 (.02) –.00 (.02) .00 (.02) –.03 (.04) –.01 (.01) –.05 (.03) .02 (.03) –.24 (.03) –.17 (.02) .08 (.02) .07 (.04)

Chirac

*** *** *** **

*

*** *

***

sig. –.03 (.04) –.06 (.01) –.03 (.03) .06 (.02) .03 (.02) –.03 (.02) .02 (.02) .01 (.04) –.01 (.01) –.03 (.03) –.01 (.03) –.17 (.03) –.14 (.02) .04 (.02) .05 (.04)

Chirac

*** *** **

*

**

***

sig. –.14 (.03) .03 (.01) –.02 (.01) –.00 (.02) –.03 (.02) .01 (.02) –.00 (.01) .04 (.03) .01 (.01) –.04 (.02) .01 (.03) .01 (.02) –.02 (.02) .05 (.02) .13 (.03)

Le Pen

2002

*** ***

*

*

*** ***

sig.

–.14 (.03) .01 (.01) –.01 (.02) –.01 (.02) –.02 (.01) .02 (.01) –.01 (.01) .05 (.03) .01 (.01) –.04 (.02) –.01 (.02) .01 (.02) –.00 (.01) .06 (.02) .08 (.03)

Le Pen

***

**

***

sig.

Table 6.3 (continued) Regression models of thermometer scores of candidates, 1988, 1995, 2002, and 2007 (ordinary least squares method)

9780230_300088_07_cha06.indd 131

6/29/2011 10:37:12 AM

1454

–.13 (.02) –.08 (.02) –.18 (.03) _ _ _ .68 (.04) .25

***

*** *** ***

1204

–.09 (.02) –.06 (.02) –.11 (.02) .17 (.02) .09 (.02) .13 (.02) .42 (.04) .47

*** *** *** *** *** *** *** 1455

.13 (.02) –.00 (.02) –.20 (.03) _ _ _ .57 (.04) .29 ***

***

***

1241

.09 (.02) .01 (.01) –.15 (.02) .11 (.02) .17 (.02) .09 (.02) .41 (.04) .44 *** *** *** *** ***

***

1453

.26 (.02) .05 (.01) .20 (.02) _ _ _ –.04 (.03) .29

*** *** ***

1374

.17 (.02) .04 (.01) .12 (.02) .23 (.02) .19 (.01) _ –.04 (.03) .49

*** *** *** *** ***

Source: CEVIPOF (2002).

Note: The table shows non-standardized regression coefficients obtained through ordinary least squares estimation. The standard deviation is in parentheses. The thresholds for significance are as follows: * p < .10; ** p < .05; *** p < .01, two-tailed test.

N

Immigration Economy Europe Right stuff Honesty Keeps promises Constant R2

9780230_300088_07_cha06.indd 132

6/29/2011 10:37:12 AM

Age Male Education Religion Manager Employed Working-class Income Private-sector Non-risky asset Risky asset Extreme left Left Right Extreme right Immigration

.01 (.02) –.01 (.01) –.02 (.02) .01 (.01) –.01 (.01) .01 (.01) .04 (.01) –.02 (.02) –.01 (.01) .01 (.02) –.01 (.01) .09 (.02) .21 (.01) –.17 (.01) –.29 (.03) –.07 (.01)

Royal

*** *** *** *** ***

*

***

sig. .00 (.02) .00 (.01) –.00 (.01) .00 (.01) .00 (.01) .01 (.01) .04 (.01) –.03 (.02) –.01 (.01) .01 (.02) –.01 (.01) .03 (.02) .10 (.01) –.12 (.01) –.25 (.03) –.06 (.01)

Royal

** *** *** ***

*

***

sig.

Sarkozy .12 (.02) .02 (.01) –.03 (.02) .11 (.01) –.03 (.01) .00 (.01) .01 (.01) .00 (.02) .0 (.01) –.01 (.02) .02 (.01) –.35 (.02) –.22 (.01) .23 (.01) .16 (.03) .12 (.01)

2007

** *** *** *** *** ***

*** ** ** *** **

sig. .07 (.02) .00 (.01) –.05 (.02) .08 (.01) –.02 (.01) –.01 (.01) .01 (.01) .00 (.02) –.00 (.01) –.01 (.01) .02 (.01) –.22 (.02) –.11 (.01) .08 (.01) .08 (.03) .07 (.01)

Sarkozy

*** *** *** *** ***

*** ***

**

sig.

Table 6.3 (continued) - Regression models of thermometer scores of candidates, 1988, 1995, 2002, and 2007 (ordinary least squares method)

9780230_300088_07_cha06.indd 133

6/29/2011 10:37:12 AM

–.03 (.01) –.07 (.01) _ _ _ .65 (.02) .36 3253 ***

*** ***

–.02 (.01) –.04 (.01) .12 (.01) .21 (.02) .17 (.01) .52 (.02) .52 2781

*** *** *** *** *** ***

–.01 (.01) –.08 (.01) _ _ _ .48 (.02) .50 3253 ***

***

–.00 (.01) –.04 (.01) .19 (.01) .23 (.01) .21 (.01) .32 (.02) .67 2441 *** *** *** *** ***

Source : CEVIPOF (2007).

Note: The table shows non-standardized regression coefficients obtained through ordinary least squares estimation. The standard deviation is in parentheses. The thresholds for significance are as follows: * p < .10; ** p < .05; *** p < .01, two-tailed test.

Economy Europe Good candidate Trust Proximity Constant R2 N

134 French Presidential Elections

indication of this. The simple correlations presented in the table between the indicators measuring a candidate’s competence, honesty, and empathy and their evaluation are on the whole substantial (the average correlations of each dimension are 0.56, 0.46, and 0.47, respectively). These results confirm in a preliminary way the idea that voter evaluation of a candidate rests upon consideration of different aspects of the candidate’s personality. The findings obtained with the help of multivariate models on image scale (see, for each of the candidates, the results corresponding to column 2 of Table 6.3) show a more convincing link between specific aspects of a candidate’s image and his or her overall evaluation. The percentage of explained variation in image-scale variables rises in a meaningful way when these indicators are integrated into the model, on average from 38 percent to 55 percent. The observed effects are both systematic and important, regardless of the dimension of image analyzed. Satisfaction with Mitterrand and Chirac in 1988 – an indicator of their competence – had a substantial impact on evaluation of these candidates (coefficients of 0.49 and 0.44, and an increase in variation explained of 21 and 16 percentage points, respectively). The impact of this indicator is also notable in the case of Balladur in 1995 (a coefficient of 0.20 and rise in explained variation of 9 percentage points). This last result also shows Balladur’s particular inability to sufficiently capitalize on voter satisfaction, notably regarding the economy, in order to boost his image and increase his support. (Observe that the results in Table 6.4 show a weaker correlation between satisfaction and image for Balladur than for Mitterrand and Chirac in 1988). The results of the 2002 and 2007 elections, which rest on more adequate measures of candidate image, further confirm these observations.1 They clearly show that positive perceptions of specific aspects of a candidate’s image – for example, thinking that he or she has the right stuff to be president, or that he or she is honest or empathic – have an important influence on a candidate’s overall image, beyond a voter’s ideological predispositions and public policy preferences. The coefficients of the whole set of variables measuring these factors, eight for the 2002 election and six for the 2007 election, are of the expected sign and highly statistically significant. (The average effect of these variables corresponds to an approximate rise of 16 percentage points in the evaluation of candidates among

9780230_300088_07_cha06.indd 134

6/29/2011 10:37:12 AM

Leader Image

135

Table 6.4 Bivariate correlations of personality traits with thermometer measures, 1988, 1995, 2002, and 2007 A. Competence and ability Good candidate (2007) Right stuff (2002)

Royal .46 Jospin .51

Satisfaction (1995) Satisfaction (1988)

Sarkozy .59 Chirac .36

Le Pen .47

Balladur .57 Mitterrand .79

Chirac .75

Jospin .42 Jospin .47

Chirac .49 Chirac .39

Royal .51

Sarkozy .43

B. Integrity Honesty (2002) Keeps promises (2002)

Le Pen .52

C. Empathy Understanding (2007)

Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

respondents who believed that the candidates possessed a given quality – competence, honesty, or empathy – versus those who believed otherwise.) The objective of this section was to explore the determinants of candidate image. The results, which establish that these evaluations rest on both voters’ ideological predispositions and their evaluation of a candidate’s policies, performance, and personality, underscore the importance of using the thermometer scale in determining the impact of a candidate’s image on voting since 1988.

6.5 The impact of candidate image on French presidential voting The full results of regression analysis, including thermometer scales of candidates, are presented in the Appendix. The model, in words, has the following theoretical specification, with short-term forces

9780230_300088_07_cha06.indd 135

6/29/2011 10:37:12 AM

136 French Presidential Elections

(issues and leader image) operating alongside long-term forces (social factors and ideology): Vote = social factors + ideology + issues + leadership. First, a word on the overall performance of these models is in order. Adding voter evaluations of candidates to these voting models significantly raises their explanatory power, as seen in the noticeable rise in pseudo-R-squared values, which range between 0.26 and 0.41 for the first round and between 0.55 and 0.67 in the second round (see Appendix). Therefore, feelings toward leaders play a key role in the first round. The average increase in pseudo-R-squared values after their inclusion in vote models is higher (0.15) when compared to our other categories of variables, such as ideology (0.14), issues (0.06), socio-demographic variables (0.04), and class-related variables (0.04). Comparatively, the effect seems less marked in the second round, which is dominated by ideology. (In the case of second-round balloting, the average increase in pseudo-R-squared values when, respectively, ideology and leader image are added, is as follows: 0.41 and 0.12.) Nevertheless, candidate personality has a decisive influence on the deciding round of an election when the difference in the number of votes received by each candidate is small. (Such narrow victories were the case in 1988, 1995, and 2007; the measures of vote intention in the 2002 election strongly suggest that a Jospin-Chirac duel in the second round would have also led to a close finish). An examination of logistic regression coefficients in the different models highlights the general impact of leader image on vote choice. Table 6.5 presents, at a glance, the coefficients for the whole set of candidates in the first round (the initial five rows) and second round (the last row). These coefficients, all statistically significant, underscore the existence of a consistent and substantial effect of a leader’s personality on vote choice. This effect can be observed for all political families (slightly weaker coefficients for the extreme left are explained by the absence of thermometer scales measuring the popularity of all the candidates identifying with this political family) and for all of the elections studied. The relative size of the coefficients associated with the thermometer scales show that leader image leads against all other vote determinants. Furthermore, only variables tied

9780230_300088_07_cha06.indd 136

6/29/2011 10:37:12 AM

Leader Image

137

Table 6.5 Logistic regression coefficients of thermometer variables for the French presidential elections of 1988, 1995, 2002, and 2007 1988

1995

2002

2007

Extreme left Left Center Right Extreme right

4.23 7.29 9.11 5.52 5.78

4.53 5.22 5.26 5.85 5.58

5.20 6.52 8.24 6.21 4.88

5.55 7.14 7.78 7.23 7.44

Second round

10.78

6.07

5.46

8.80

Notes: Regressions contain socio-demographic, class and patrimony, ideological, and issue variables (see Appendix). For 1988, the Europe and patrimony variables are absent. All the thermometer coefficients are significant p <.01, two-tailed test. Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

to ideology remain statistically significant, but with reduced impact, when the thermometer measures are included in the models (see Appendix). The data in Table 6.6 allow us to more easily interpret the logistic regression coefficients and to discuss more intuitively the capacity of candidates to capitalize on their image to increase support. The figures in the table correspond to an increase in the probability of voting for a candidate (in terms of percentage points) when all the other variables are fixed at their mean values and when the popularity of the candidate varies within one standard deviation (about 25 points on their image scale).2 This variation of one standard deviation allows us to obtain less impressive, but more realistic, simulations than those that would consist of evaluating the effect of a variation in a candidate’s score going from its minimum (0) to its maximum value (100). These simulations show that leader image has an important effect on vote choice. This impact is particularly marked during the second round. For example, the probability of voting for Nicolas Sarkozy in 2007 rises 60 percentage points when his popularity varies upward by one standard deviation. The effects for the other elections are less substantial, but nonetheless very important (effects of 33, 40, and 36 points for the elections of 1988, 1995, and 2002). The results are

9780230_300088_07_cha06.indd 137

6/29/2011 10:37:13 AM

138 French Presidential Elections

Table 6.6 Effect of leader image on voting in French presidential elections of 1988, 1995, 2002, and 2007 1988

1995

2002

2007

Extreme left Left Center Right Extreme right

.05 .30 .06 .05 .05

.05 .15 .09 .13 .06

.10 .14 .03 .12 .03

.05 .16 .16 .21 .01

Second round

.33

.40

.36

.60

Notes: The entries are changes in voting probabilities generated by CLARIFY (Tomz et al., 2003) for a variation of one-half standard deviation below and above the mean value of a candidate’s score on his or her thermometer scale. Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

clear. The essential element that tips the balance for many voters and that ultimately determines the result of the election is candidate image. The simplicity and clarity of the choice in the second round, combined with the presence of moderate candidates (with the 2002 election being an exception), explains the decisive impact of a candidate’s personality on vote choice. Likewise, simulations for the first round emphasize the importance of leader image on vote choice. The results, which can seem less spectacular, must be interpreted keeping in mind the large fragmentation of political offerings in the first round. For example, the average observed effect for candidates of the left and of the right – 19 and 13 percentage points – has a different meaning if one remembers that their mean support during the first round was 25 percent and 33 percent, respectively. The average observed effect of candidate image on vote choice, from a one-standard-deviation change, is 10 percentage points. This effect, constant across time, seems to be a little more pronounced for moderate candidates than it is for more extremist candidates. Another source of variation can be linked to the large capacity of candidates to translate their personal popularity into electoral support. Success generally seems to coincide with a larger capacity to convert personal affinity into electoral support, as suggested by the simulations for

9780230_300088_07_cha06.indd 138

6/29/2011 10:37:13 AM

Leader Image

139

Mitterrand in 1988 (effect of 30 points), Sarkozy in 2007 (effect of 21 points), or the extreme-left candidates in 2002 (effect of 10 points). On the contrary, failure seems to be linked to difficulty in turning personal popularity into votes, as suggested by the simulations for Chirac in 1988 and Le Pen in 2007. The case of centrist candidates Balladur and Bayrou is interesting. The former came on the scene in 1995, profiting from positive assessments about the economy (recall the effect of perceptions of the economy on his support). The weak logistic regression coefficient in the Balladur vote equation (5.26) and the limited effect measured by the simulations (9 points) suggest that he did not capitalize on his image and increase his support in any substantial way. On the other hand, the observed results for Bayrou in 2007 (an effect of 19 points for the simulation) suggest that he knew how to play the electoral game by turning his personal popularity into support.

6.6 Conclusion The results of this chapter are unambiguous. Candidate image has had a decisive influence on French vote choice during the past four presidential elections. The effect of a leader’s image on voting is consistent and substantial. This impact is clearly observed when other chief determinants of electoral behavior are controlled for. The results also show that the effect of these images depends both on a certain degree of goodwill felt toward a candidate and on a candidate’s capacity to convert this capital into electoral support. One can conclude, in light of the close races in the first and second rounds, that leader image has had an important influence not only on vote choice, but also on the results of French presidential elections over the past 20 years. These conclusions therefore suggest that French electoral studies should give more consistent attention to candidate image and its impact on vote choice. The growth in the number of variables addressing this theme in the 2007 CEVIPOF survey is a step in the right direction. However, it would be preferable if the selected indicators rested on sharper theoretical considerations and were used in a more coherent manner. For example, it is problematic that the five indicators of image in the 2007 survey are not present in the same waves of the study.

9780230_300088_07_cha06.indd 139

6/29/2011 10:37:13 AM

140

French Presidential Elections

In paying more attention to leader image, French political scientists are beginning to reunite with a tradition that was formerly fruitful. The question regarding presidential demeanor (“the right stuff”) is an interesting contribution. This concept seems to overlap with two important dimensions of leader image – competence and strength – and offer a rich measure that is close to the Anglo-Saxon concept of being “fit for the job.” If the post–de Gaulle period in France has been marked by a retreat in the scholarly interest given to candidate image, we can hope that the increasing présidentialisation of French political life will help jump-start studies on this question both in France and elsewhere.

9780230_300088_07_cha06.indd 140

6/29/2011 10:37:13 AM

7 Campaign Influence

The presidential election campaign constitutes one of the most important moments of modern-day French political life. Each candidate uses the campaign as a means to present their ideas, positions, and policy proposals. In addition to promoting their past achievements, candidates also seek to underline the weaknesses of their adversaries. During this short period of a few weeks leading up to election day, the main objective is to influence the opinion of the electorate. In France as elsewhere, it is widely believed that presidential campaigns matter and that the final outcome of a campaign is largely attributable either to the messages conveyed during this period or to the candidates’ smart moves or faux pas. Yet, is the influence of campaigns on the outcome of an election as important as many political actors and observers would like to think? The first scientific theories about this subject, formulated between the two world wars, stated that public opinion was largely susceptible to influence by propaganda (for example, see Lippmann, 1922, 1925). However, studies conducted after World War II in the United States have given rise to the “minimal effects” school by showing that campaigns only activate or reinforce a voter’s partisan predispositions, which have been established long ago through political socialization (for example, see Berelson et al., 1954; Campbell et al., 1960; Lazarsfeld et al., 1944). According to this model, campaigns reduce the gap between vote choice and latent preferences, be it by consolidating the choice of hesitant voters or by helping confused citizens make a decision that is consistent with their predispositions.

141

9780230_300088_08_ch07.indd 141

6/28/2011 8:12:28 PM

142

French Presidential Elections

However, the recent use of more sophisticated methodological tools has led to a refinement of the study of campaign effects, showing that they are not as marginal as previously thought (see Brady et al., 2006). Campaigns themselves have also become more sophisticated following the introduction of new communication techniques (De Vreese, 2010; Small, 2008; Wattenberg, 1998; see Farrell (1996) for a classic exposé on the evolution of campaigns). Furthermore, partisan attachments have eroded over time, leaving voters more open to changing their choice from one election to another (Wlezien, 2010). Research on media effects has demonstrated the capacity of a candidate to draw journalists’ attention to certain issues (the “agendasetting” effect) or to bring them into adopting certain interpretations of a particular subject (the “framing” effect). These effects can have a notable impact on the vote by changing the relative weight of the criteria that influence vote choice, amounting to a “priming” effect. (For a survey of these effect concepts in the context of electoral campaigns, see Ansolabehere et al. (1993); Norris et al. (1999); Nadeau and Bastien (2003); Nadeau et al. (2010).) Consider an example. During a period of prosperity, an incumbent candidate will have an advantage in basing his or her campaign on the economy and in taking credit for favorable economic conditions, while adversaries should choose the opposite strategy. On the other hand, during a period of recession an incumbent should either convince voters that he or she is not responsible for the situation (by attributing it to global factors, for example) or draw attention instead to issues on which he or she has a good track record. These rules have led Lynn Vavreck (2009) to establish, from American presidential elections, a typology of good and bad campaigns. For instance, Democratic presidential candidate Al Gore’s 2000 campaign, in which Gore chose not to capitalize on the economic success of the Clinton administration, would constitute a bad campaign. The quality of candidates’ campaigns exercises a significant impact on the outcome of elections. These new developments have led researchers to reevaluate the impact of campaigns on voting behavior, in France as well as abroad. In a similar spirit, recent works on electoral forecasting in France (for example, see Auberger, 2010; Jérôme and Jérôme-Speziari, 2010; Lewis-Beck et al., 2008; Nadeau et al., 2010) do not discount the influence of campaigns on vote choice in presidential elections. These forecasting models show that certain fundamental variables,

9780230_300088_08_ch07.indd 142

6/28/2011 8:12:28 PM

Campaign Influences 143

such as the popularity of the executive and the economic situation, allow us to predict a few months in advance the result of an election; however, the gap between this prediction and the final result can often be attributed to the role of campaigns. To the extent that the fundamentals do not favor one camp over another, or that the campaigns of the main candidates are of equal strength, the forecast should be accurate (Campbell, 2000; Bartels, 2006; Gelman and King, 1993). A model’s forecast can therefore be seen as a sort of equilibrium from which the actual result can deviate as a function of the impact of certain campaign events (Holbrook, 1996). In other words, a campaign can be a game-changer. Within the context of the present volume, it is possible to conceptualize the impact of campaigns on vote choice in French presidential elections in two different ways. The first assumes that campaign influence has already been taken into account, in an indirect way, through the study of the different vote determinants presented in the five preceding chapters. Campaigns activate or reinforce voters’ various predispositions. The effect of factors such as social cleavages, ideological identification, issues, and leadership on electoral behavior can be delivered by campaign activity toward the electorate. Campaigns, by allowing candidates to (re)mobilize their social and ideological bases, to emphasize certain issues to the detriment of other candidates, and to the detriment of others, have impacted vote choice through these variables; therefore their influence is already reflected in the empirical results that we examined earlier. Yet, it is still possible that campaigns have a more direct influence on vote choice. Studies have shown that a voter’s attentiveness to a campaign plays a role in how much influence it has; the voters most likely to be influenced by the flood of communication during a campaign are those who pay moderate attention to it. This effect is mostly nonlinear. The less attentive voters are “immunized” against electoral messages due to their ignorance of issues, while the more attentive voters are immunized as well due to their deep political knowledge and strong articulation of their views. (On this classic thesis, see the works of Converse (1962); Zaller (1989, 1992); Nadeau et al. (2008).) Likewise, we can expect that voters who make their choice at the very last minute (“late deciders”) do not have the same profile, and do not base their vote on the same considerations, as those who made their decision either during the campaign (“campaign deciders”) or even before the campaign started (“pre-campaign deciders”). (On these

9780230_300088_08_ch07.indd 143

6/28/2011 8:12:28 PM

144

French Presidential Elections

distinctions, see Chaffee and Rimal (1996); Fournier et al. (2004).) Finally, keeping in mind the centrality of communication media in modern electoral campaigns, it is probable that a candidate’s television appearance has an effect on vote choice among those who watched it (Wattenberg, 1998). It is these three dimensions – campaign attentiveness, time of vote decision, and the impact of television – that we will explore more directly in this chapter, in the context of French presidential elections from 1988 to 2007.1

7.1

Campaign attentiveness

If we presume that electoral campaigns can indeed influence vote choice, the first path to explore is the degree of voter attentiveness to campaigns. We can expect that attentiveness to campaigns is a necessary condition for campaign effects. After all, if voters are not paying any attention to campaigns, then it is unlikely that their vote choice will be influenced by them. To what degree do French voters follow presidential campaigns? Does vote choice differ as a function of the level of interest in a campaign? Do individuals who regularly follow a campaign base their vote preferences on dimensions that are different from those who do not follow campaigns at all? The election surveys examined in this volume allow us to answer these questions, as each survey offers a measure of attentiveness to campaigns. The question asked in 2002 and 2007 read as follows: “Did you follow the presidential campaign?” The question used in Pierce’s 1988 study was as follows: “Personally, are you interested in the presidential election campaign?” In 1995, the only question available was one that measures political interest, a variable that we will use here as a proxy for campaign attentiveness. Several response choices are offered, which can be regrouped into three levels of attentiveness that are identical for the four elections: every day or almost every day (regular attentiveness), from time to time (periodic attentiveness), and not at all (non-attentiveness). (For 1988 and 1995, the categories “very” and “somewhat” were regrouped in order to offer a category of regular attentiveness that is equivalent to those of the other elections.) Note that the questions used refer to the campaign in the first round of the presidential elections.

9780230_300088_08_ch07.indd 144

6/28/2011 8:12:28 PM

Campaign Influences 145

Our measure of campaign attentiveness indicates that it differs considerably from one election to another. The data in Table 7.1 show that the presidential elections of 1988 and 2007 are the ones that generated the most attention amongst the French electorate (59 percent and 63 percent, respectively, reporting regular attentiveness). These are also the elections that have seen the weakest proportion of non-attentiveness over the course of the campaign (11 percent in 1988 and only 3 percent in 2007). Periodic attentiveness stayed about the same for all the elections, with the exception of the 2002 election, in which this proportion was higher than usual (48 percent). What is the profile of voters who pay more attention to presidential elections in France? The multivariate analysis presented in Table 7.2, which compares the three types of attentiveness according to a series of socio-demographic, socioeconomic, and sociopolitical variables, allows us to hit upon a few common tendencies across these four elections. First, it seems that more educated individuals pay more attention to presidential campaigns, a result that confirms the conclusions of Vedel (2009, p. 163) regarding the 2007 election. More wealthy individuals do the same, while blue-collar workers and, in a less consistent manner, women, pay less attention to campaigns. Finally, an ideological cleavage emerges from our results: those who closely follow campaigns are generally those who identify with one of the two moderate ideological camps (left or right), as opposed to voters who claim to be of the center or one of the two ideological extremes. For 2007, Vedel (2009, pp. 161–62) observed that the most ideologically “polarized” voters (but not the most extreme ones) were those who followed the campaign on a more regular basis. According to the results in Table 7.2, it seems that this finding Table 7.1

Campaign attentiveness

Regular attentiveness (%) Periodic attentiveness (%) Non-attentiveness (%) N

1988

1995

2002

59 30 11 1011

49 36 15 4076

37 48 15 4107

2007 63 34 3 4006

Note: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

9780230_300088_08_ch07.indd 145

6/28/2011 8:12:28 PM

9780230_300088_08_ch07.indd 146

6/28/2011 8:12:28 PM

Moderate left

Extreme left

Private-sector

Income

.51 (.70) .07 (.32) .35 (.51) –.54 (.53) –.34 (.50) –.73 (.53) 1.52* (.79) .14 (.36) 1.33 (1.16) 1.11*** (.35)

Periodic 1.11* (.68) .79** (.31) .71 (.49) –1.31*** (.51) –.79 (.48) –1.72*** (.51) 2.94*** (.78) .45 (.35) 2.04* (1.11) 1.65*** (.33)

Regular

1988

Who is more attentive to campaigns?

Blue-collar

Employed

Manager

Education

Male

Age

Table 7.2

–.58** (.29) .25** (.13) .41* (.24) –.37* (.21) –.51*** (.19) –.67*** (.20) 1.04*** (.35) .08 (.14) .68** (.29) .76*** (.14)

Periodic .27 (.29) .85*** (.13) 1.53*** (.24) –.37* (.21) –.46** (.19) –1.04*** (.20) 1.69*** (.35) .05 (.14) 1.80*** (.28) 1.42*** (.15)

Regular

1995

.78** (.31) .06 (.13) .42* (.22) .03 (.16) –.30** (.15) –.26* (.13) .66** (.31) –.39*** (.13) .71*** (.25) .39*** (.14)

Periodic 2.44*** (.32) .48*** (.13) .95*** (.23) .09 (.16) –.25 (.16) –.39*** (.14) 1.29*** (.33) –.29** (.14) 1.15*** (.26) .56*** (.15)

Regular

2002

–.52 (.60) –.09 (.24) .80* (.46) .35 (.43) .06 (.32) .09 (.32) –.23 (.58) –.05 (.22) –.21 (.55) .21 (.26)

Periodic

.79 (.60) .19 (.24) 1.46*** (.45) .82* (.42) .02 (.31) –.16 (.31) .89*** (.57) –.20 (.22) .73 (.53) .94*** (.25)

Regular

2007

9780230_300088_08_ch07.indd 147

6/28/2011 8:12:29 PM

1.05*** (.39) .83 (1.19) –.56 (.80) .08 685

1.53*** (.38) 1.71 (1.11) –1.52* (.78)

.66*** (.14) .91*** (.35) .37 (.32) .08 3284

1.16*** (.14) 2.14*** (.34) –1.15*** (.33)

.40*** (.15) .16 (.31) .60** (.27) .05 3154

.47*** (.16) .73** (.31) –1.25*** (.29)

.95*** (.35) .05 (.76) 2.15*** (.52) .08 3284

1.86*** (.35) .98 (.74) 1.11** (.52)

Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

Notes: Entries are multinomial logistic regression coefficients with standard errors in parentheses. Reference category = non-attentiveness. * p < .10; ** p < .05; *** p < .01, two-tailed test.

Pseudo-R 2 N

Constant

Extreme right

Moderate right

148

French Presidential Elections

is generalizable across all of the presidential elections examined. As for the other variables included in these models, they do not allow us to disentangle any clear or generalizable trends regarding different types of attentiveness. It is possible that voters who pay more attention to campaigns are influenced in their vote choice because of the fact that they have been more exposed to campaigns and their content. A preliminary way to verify this possibility is to relate attentiveness to campaign to first-round vote choice. The results appear in Table 7.3. The association between the two variables is rather modest, but stable across time. Here are the respective Cramer’s V: 1988, 0.09; 1995, 0.08; 2002, 0.07; 2007, 0.09. Some candidates obtained more support among the more attentive, as was the case for extreme-left candidates and Chirac in 1988, for Jospin in 2002, and for Sarkozy in 2007. On the other hand, others seemed to lose support among the more attentive, as did Le Pen in 1988, 1995, and 2007, as well as Mitterrand and Barre in 1988. A more stringent test would be to add the attentiveness variable into the final first-round vote models presented in the last chapter, in order to see if campaign attentiveness significantly influences vote choice, when all other vote determinants are controlled for. The results of this test (not presented due to space issues) indicate that the variable of attentiveness has a significant effect in only 4 of the 20 logistic regressions – the vote favoring Mitterrand and Barre in 1988, Le Pen in 1995, and the extreme left in 2007. For each of these cases, the impact of attentiveness is negative: support diminishes as campaign attentiveness becomes more regular. It is possible that the effect of attentiveness on vote choice may not be strictly linear (see the preceding section). To verify this, we have reestimated the 20 first-round vote regressions for each of the three types of attentiveness separately. There is no clear-cut trend that emerges from these additional analyses; that is to say, it is impossible to conclude that vote choice is better predicted by one form of attentiveness over another. However, we can note that the variables of issues and leaders seem to play a more important role in the choice of individual voters who follow the campaign more regularly. Among the 40 regression coefficients associated with the economy and immigration issues, for instance, only 3 are statistically significant in the non-attentive group, while 8 are statistically significant

9780230_300088_08_ch07.indd 148

6/28/2011 8:12:29 PM

Campaign Influences 149

Table 7.3

Campaign attentiveness and first-round voting NonAttentiveness

1988

1995

2002

2007

Extreme left (%) Mitterrand (%) Barre (%) Chirac (%) Le Pen (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b Extreme left (%) Jospin (%) Balladur (%) Chirac (%) Le Pen (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b Extreme left (%) Jospin (%) Bayrou (%) Chirac (%) Le Pen (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b Extreme left (%) Royal (%) Bayrou (%) Sarkozy (%)

10 43 23 9 15 81 .09 .12 –.03 (.05) –.02 (.03) 11 27 16 24 22 386 .08 .00 –.09 (.02) –.06 (.02) 31 20 10 25 14 97 .07 .14 .02 (.04) .01 (.03) 19 36 17 20

Periodic Regular Attentiveness Attentiveness 11 42 17 18 12 233

16 37 17 19 11 495

13 29 20 24 15 1048

14 33 15 26 12 1595

24 25 11 26 14 497

20 31 9 25 14 540

13 29 22 27

8 30 21 36 (Continued)

9780230_300088_08_ch07.indd 149

6/28/2011 8:12:29 PM

150 French Presidential Elections

Table 7.3

Continued NonAttentiveness Le Pen (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

Periodic Regular Attentiveness Attentiveness

9 81 .09 .00 .10 (.03) .06 (.02)

8 1063

5 2197

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. Source: Pierce (1988) and CEVIPOF (1995, 2002, 2007).

among the periodically attentive, and no less than 17 are statistically significant among voters who regularly follow campaigns. Regarding the feeling thermometer for leaders, out of 20 regression coefficients, half of them are statistically significant among the non-attentive while almost all of them are statistically significant in the two other groups (19 for periodic attentiveness and 20 for regular attentiveness). Overall, we can conclude that the effect of campaign attentiveness on voting has generally been very limited, if not almost null, as suggested by Vedel (2009) with reference to the 2007 election. Voters distinguish themselves very little in their electoral behavior by their attentiveness to campaigns. At best, we can observe a slightly more pronounced influence of opinions about issues and leaders on the vote, among those who follow campaigns at least periodically. In the majority of cases, the gains or losses in support for certain candidates are better explained by the factors discussed in previous chapters. In fact, campaign attentiveness reveals a surprising paradox, as already documented by Cautrès and Jadot (2009, p. 151) as well as Dupoirier and Frognier (2009, p. 527): whereas one should expect that voters who are more attentive to campaigns decide for whom they will cast their vote later than others, the opposite is observed – a phenomenon that is confirmed in our data from the 1995, 2002, and 2007 elections (for the 1988 election, the variable of time of vote

9780230_300088_08_ch07.indd 150

6/28/2011 8:12:29 PM

Campaign Influences 151

decision was not available). Vedel (2009, p. 180) explains this apparent paradox with the idea that individuals who follow the campaign closely are those who look to confirm a vote choice that was established long ago. Whatever it may be, our examination of campaign attentiveness seems to indicate that we would better turn to the time of vote decision, if we wish to unearth the possible effects of campaigns.

7.2

Time of vote decision

Over the past few decades, most Western democracies have entered a period in which voters are deciding for whom they will vote later than they used to (Dalton and Wattenberg, 2000; Fournier et al., 2004; McAllister, 2002). This tendency seems to be linked to societal modernization, which can manifest itself in a larger interest in politics and an even larger decline in partisan identification. McAllister (2002) argues that these “late deciders” are more susceptible to influence by campaigns and therefore show more complex electoral calculations (see also Chaffee and Rimal, 1996; Fournier et al., 2004). In the French case, recent studies have effectively shown, as elsewhere, a diminishing proportion of voters who made their vote choice before the start of the presidential campaign (Cautrès and Jadot, 2007; Dupoirier and Frognier, 2009). This greater delay of the vote decision can be observed in the data presented in Table 7.4. We have at our disposal a measure of timing of the vote decision for three of the four elections analyzed (all save 1988). For these available elections, the question reads as follows: “With regard to your choice of candidate for the first round of the presidential election, would you say that: your decision was made a long time ago; your decision was made over the course of the campaign; you waited until the last moment to make your decision?” The question allows us to create a variable with three categories: decision made a long time ago (pre-campaign deciders), decision made during the campaign (campaign deciders), and decision made at the last moment (late deciders). Table 7.4 shows only a tiny decline in the number of early deciders for the period from 1995 to 2007. The results of the multivariate analysis (not shown because of space limitations) allow us to compare the profile of voters according

9780230_300088_08_ch07.indd 151

6/28/2011 8:12:29 PM

152 French Presidential Elections

Table 7.4

Time of vote decision during the first round

Pre-Campaign Deciders (%) Campaign Deciders (%) Late Deciders (%) No Response (%) N

1995

2002

52 21 20 7 4078

51 20 26 3 4107

2007 50 28 22 0 4006

Note: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Source: CEVIPOF (1995, 2002, 2007).

to the different times of vote decision. Young voters stand out as individuals who make their vote decision during the campaign, or even at the very last minute. Also, women seem to make up their minds later than men, but this trend is less marked than for age. These results conform to those reported by Cautrès and Jadot (2009, p. 137) regarding the 2007 presidential election. One can also note that for none of the elections examined do variables of class or income differentiate groups of deciders in any statistically significant manner. Also, we find in this same analysis that different types of deciders have a precise ideological profile. Most notably, centrists tend to decide their vote choice later than those from other ideological groups. These results suggest that centrist candidates, most notably Bayrou, may have benefited from their electoral campaigns. At the very least, the results confirm the conclusions of Cautrès and Jadot (2009), who found that the late deciders of 2007 showed a weaker ideological attachment than did other types of deciders (see also Dupoirier and Frognier, 2009). It is possible to support these intuitions via bivariate analysis, for which the results appear in Table 7.5. The relationship between time of vote decision and vote choice in the first round of elections reveals that this association has intensified from one election to the other, if we look again at the Cramer’s V statistic: 1995, 0.09; 2002, 0.19; 2007, 0.27. Three other observations can be reported. First, support for moderate candidates from either the left or right is almost always higher among early deciders than for the other groups. Conversely, the vote for a centrist candidate is higher among those who decide

9780230_300088_08_ch07.indd 152

6/28/2011 8:12:29 PM

Campaign Influences 153

Table 7.5

1995

Time of vote decision and first-round voting

Extreme left (%) Jospin (%) Balladur (%) Chirac (%) Le Pen (%) N Cramer’s V Pearson’s χ2

Pre-campaign deciders

Campaign deciders

Late deciders

11 31 17 25 16 1757 .09

14 35 15 26 10 671

20 26 17 22 15 598

Gamma Kendall’s tau-b

.00 –.09 (.02) –.06 (.02)

2002

Extreme left (%) Jospin (%) Bayrou (%) Chirac (%) Le Pen (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

21 32 4 27 15 657 .19 .00 –.04 (04) –.03 (.03)

28 18 15 22 16 218

23 24 22 22 10 250

2007

Extreme left (%) Royal (%) Bayrou (%) Sarkozy (%) Le Pen (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

8 34 9 41 9 1707 .27 .00 –.19 (.02) –.13 (.02)

8 22 37 31 2 960

16 30 33 16 5 674

Notes: After totaling the percentages, certain columns may not add up to 100% because of rounding error. Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. Source: CEVIPOF (1995, 2002, 2007).

9780230_300088_08_ch07.indd 153

6/28/2011 8:12:29 PM

154

French Presidential Elections

during or at the end of the campaign – except in 1995 when it did not vary much as a function of time of vote decision. Finally, we note that the largest gains accrued to Jean-Marie Le Pen in the first round of the 2002 election came from those who decided during the campaign rather than at the end, suggesting that Le Pen’s election breakthrough here had more to do with campaign quality than with some last-minute jolt within the electorate. Over the past three presidential elections, the variable of time of vote decision thus seems to have influenced French voters. However, this bivariate relation could prove to be spurious, in that the effects of the campaign might be explained by other variables, discussed earlier in this volume. In order to delve further into this question, we have done as in the previous section and reestimated the first-round vote models presented in Chapter 6 by adding the variable of time of vote choice, coded ordinally in three categories going from a precampaign decision to a late decision (see last column in the Appendix). In words, here is the theoretical specification for that equation: Vote = social factors + ideology + issues + leadership + campaign timing. These complementary analyses indicate that the time of vote decision was significant for two candidates in 1995 (higher support among late deciders for the extreme-left candidates and a decrease in support for Jospin) and two others in 2002 (higher support among late deciders for centrist Bayrou and lower support for Chirac). In 2007, the time of vote decision variable presents a statistically significant relationship in no fewer than four out of five candidacies: a positive relationship in the cases of Bayrou and extreme-left candidates and a negative relationship in the cases of Royal and Sarkozy. Overall, we observe that for each election, it is the moderate left or moderate right candidates who are at a significant disadvantage among voters who do not decide early. On the contrary, nontraditional candidates (that is, extreme-left or center options) seem to make significant gains during the campaign, once other determinants of the vote are controlled for. These results reflect the conclusions of McAllister (2002), who found that it is mostly smaller parties that benefit from the support of this “new” group of late deciders.

9780230_300088_08_ch07.indd 154

6/28/2011 8:12:30 PM

Campaign Influences 155

One final way to account for the influence of time of vote decision is to reestimate the vote models for each of the three groups of deciders separately (results not presented due to space concerns). First, we observe that the vote models’ fit declines as one approaches making a late decision. For the 15 logistic regressions, the median pseudo-Rsquared is 0.43 for the pre-campaign deciders, 0.42 for the campaign deciders, and 0.26 for the late deciders. Another indicator of this erosion in the models’ explanatory power is the fact that while 94 regression coefficients are statistically significant for the group that made their choice long ago, this number drops to 65 for the campaign deciders, then to 51 for the late deciders. These results provide additional evidence in favor of the idea that late deciders may be not strategic, but rather confused, about their electoral choice (Cautrès and Jadot, 2007). According to Dupoirier and Frognier (2009), it is the median group – that is, those individuals who make their decision during the campaign, rather than before the campaign or at the very last minute – who exhibit a truly rational and sophisticated behavior vis-à-vis the election campaign and all the informational elements that it entails.

7.3

The impact of television

Even with the recent advances in communication technologies made possible by the Internet, television remains a preferred vehicle of information for both politicians and voters during electoral campaigns. This is certainly true in France (Gerstlé, 2008). In fact, a recent study by Tiberj and Vedel (2007) has found that, according to the data from the 2006 Baromètre politique français (French Political Barometer), television remains the most useful way to stay informed about politics: television receives 58 percent of the mentions versus 17 percent for radio, and even less for other means of communication. In France, the role of television in presidential campaigns has for several decades caught the attention of researchers (for example, see Blumler et al., 1978; Gerstlé, 1997). A first question that arises about the impact of television on French voter opinion relates to the evolution of the candidates’ image. As Boy and Chiche (2007) have noted, candidate images tend to change over the course of an electoral campaign. Some candidates are able to improve their image among the electorate during the campaign, while

9780230_300088_08_ch07.indd 155

6/28/2011 8:12:30 PM

156

French Presidential Elections

others see their image among the citizenry eroding away into negative territory. Keeping in mind the strong médiatisation of campaigns along with the regular presence of candidates on television (be it in the news or in other programs), there is place to wonder if a candidate’s appearance on television affects his or her image among voters. The 1995 and 2002 election surveys each had one question asking respondents if seeing a candidate on television changed their opinion of him or her. The responses given to this question are summarized in Table 7.6. Note that the question refers only to the first round of the presidential election; the attention given here to this round is amply justified by the fact that the qualities of the candidates usually play a much larger role in the first round than in the second one (Boy and Chiche, 2007, p. 337; see also Chapter 6). Also note that Arlette Laguiller, alone, represents the extreme left, for the purpose of simplifying the current analysis. First, looking at all the candidates from all the elections, the table reveals that the proportion of individuals who stated that their opinion did not change at all varies from 55 to 76 percent. Therefore, Table 7.6 Change in first-round candidates’ image after having seen them on television For the worse (%)

None (%)

For the No response better (%) (%)

1995

Laguiller Jospin Balladur Chirac Le Pen

2 5 15 6 18

76 55 67 64 71

19 38 16 28 8

2 2 2 2 3

2002

Laguiller Jospin Bayrou Chirac Le Pen

7 13 7 12 14

70 72 67 73 74

19 12 19 11 8

4 3 7 3 4

Note: After totaling the percentages, certain rows may not add up to 100% because of rounding error. The question asked was the following: “Did seeing the following candidates on television change the opinion that you had of them for the better, for the worse, or not at all?” Source: CEVIPOF (1995, 2002).

9780230_300088_08_ch07.indd 156

6/28/2011 8:12:30 PM

Campaign Influences 157

this leaves more or less one-third of respondents who changed their opinion for better or for worse after seeing the candidates on television. At this point, one can observe that Laguiller and Bayrou are two candidates whose image tended to improve during the campaign. Conversely, Jean-Marie Le Pen saw his image negatively affected in these two elections. The figures in the table also reflect the difficulties that Balladur’s 1995 and Jospin’s 2002 campaign faced: for these two candidates, the proportion of those who said their opinion of the candidate changed for the better is almost entirely counterbalanced by the proportion of voters who said that their opinion changed for the worse. The same applies for Chirac’s image in 2002. Of course, one can wonder to what degree these opinions of candidates following their television appearances affected their support at the ballot box. This question is difficult to answer with the data at hand, since the changes in opinion measured in the preceding table risk having been partially influenced by the vote itself. Nonetheless, for each of the candidates, we can put together the evolution of their image with their level of support during the first round of the election. Table 7.7 presents this relationship: for each candidate, we report the percentage of votes received among survey respondents whose opinion changed for the worse, survey respondents whose opinion did not change at all, and survey respondents whose opinion changed for the better. Without much surprise, one can see that the relationship is positive and linear in each case. In other words, all the candidates saw their support grow with increasingly favorable public opinion toward them. That being said, is there a candidate who seems to have benefited the most from television appearances? It seems to be Jean-Marie Le Pen, who in 1995 obtained 48 percentage points more among those whose opinion of him changed for the better than among those whose opinion of him changed for the worse; for the 2002 election, Le Pen’s net gain was 41 percentage points. However, one must remember that his image changed more negatively than positively during these elections (recall Table 7.6). Also, we must reiterate that the results presented here should be interpreted with a bit of caution, keeping in mind a possible voter projection effect on candidate image. A second line of inquiry related to the influence of television on presidential campaigns and on French voting behavior concerns the

9780230_300088_08_ch07.indd 157

6/28/2011 8:12:30 PM

158 French Presidential Elections

Table 7.7 Relationship between change in a candidate’s image and vote choice for that candidate in the first round For the worse (%) None (%)

For the better (%) N (%)

1995 Laguiller Jospin Balladur Chirac Le Pen

2 3 2 2 2

4 24 13 19 12

9 38 36 37 50

162 926 499 742 427

2002 Laguiller Jospin Bayrou Chirac Le Pen

1 5 2 5 1

6 19 5 18 9

14 43 19 42 42

111 309 112 285 158

Source: CEVIPOF (1995, 2002).

traditional televised debate between the two candidates that have qualified for the second round of the election. The first of these debates was held for the 1974 presidential election, between Valéry Giscard d’Estaing and François Mitterrand. Since then, such a debate has been organized between the two rounds of each election, with the exception of 2002, when Jacques Chirac flouted tradition in order to not provide a forum for his extremist opponent Jean-Marie Le Pen. The first two debates, those of 1974 and 1981, are generally considered to be high points, as they contrasted the track record of successive Gaullist governments against the possible “risk” of a more politically and economically socialist government. These two debates were quite intense and were watched by a wide audience (Bréchon and Denni, 2008, pp. 34, 39). The 1988 televised debate was also very important because it pitted, for the first time, the two cohabitants of the executive against one another. With regard to the 1995 and 2007 debates, they did not seem to be as noteworthy, shedding little light on the differences in the candidates’ positions. Work on the impact of televised debates on vote choice has long concluded that they have but a very small influence, as televised debates merely reinforce long-established preferences (see, e.g., Abramowitz, 1978; Hagner and Rieselbach, 1980). More recent studies

9780230_300088_08_ch07.indd 158

6/28/2011 8:12:30 PM

Campaign Influences 159

have nonetheless questioned these conclusions by showing that televised debates can reinforce opinions for some and modify perceptions for others, particularly among undecided voters or voters who are little interested in politics (see, e.g., Druckman, 2003; Geer, 1988; Hillygus and Jackman, 2003). Canadian research on this question has notably allowed refinement of the empirical study of the effects of televised debates on leader image and on voting behavior (Blais and Boyer, 1996; Blais et al., 2003; Giasson et al., 2005; Lanoue, 1991). Two of the election surveys analyzed in this volume enable us to examine the potential impact of French televised debates. The 1988 study asked the following question: “Did you watch the April 28 televised debate between Mitterrand and Chirac?” Approximately 71 percent stated that they had watched the debate, of whom about one-third indicated that they had watched the whole debate. In the 2007 study, the question was “Would you say that you had made your vote choice before or after the televised debate between Nicolas Sarkozy and Ségolène Royal?” About 21 percent of respondents said that they had made their decision after the debate. The 2007 question is less direct than that of 1988, but keeping in mind that several studies have shown that televised debates can also influence the decision of voters who did not watch them (Blais and Boyer, 1996; Blais et al., 2003; Senior, 2008), it is still relevant. These two measures of attentiveness to the debates can be put in relation to second-round voting. Table 7.8 presents the results of this bivariate analysis, which indicates that the debate seems to have had an influence on 1988 vote choice, along with 2007 vote choice, albeit to a significantly smaller degree. In the first case, we observe that Mitterrand’s lead over Chirac was two times smaller among those who watched the debate between the two candidates (lead of 16 points versus 35 points among those who did not watch the debate). In the second case, the gap between Sarkozy and Royal was almost zero among those who reported having decided before the debate, while we note a slight advantage of 6 points in favor of the rightwing candidate among those who made up their minds after the debate. Measures of statistical significance nonetheless indicate that the relationship is weak, significant only for the 1988 election. It is rather difficult to determine with any certainty if this difference in results between the two elections is attributable to the fact that the question used to measure attention to debate was not the same in

9780230_300088_08_ch07.indd 159

6/28/2011 8:12:31 PM

160

French Presidential Elections

Table 7.8

1988

2007

Televised debate and second-round voting Did not see the debate

Saw the debate

67 33 217 .08

58 42 650

Mitterrand (%) Chirac (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

.02 .19 (.08) .08 (.03)

Royal (%) Sarkozy (%) N Cramer’s V Pearson’s χ2 Gamma Kendall’s tau-b

Decided before the debate 49 51 2825 .02 .29 .05 (.05) .02 (.02)

Decided after the debate 47 53 601

Notes: Only the P value is indicated for Pearson’s χ2. Asymptotic standard errors are indicated in parentheses for gamma and Kendall’s tau-b. Source: Pierce (1988) and CEVIPOF (2007).

both surveys. We can also note that in the multivariate analysis, when the debate variable is added to the final models from the preceding chapter, its effect on second-round vote choice is significant for neither of the two elections studied.

7.4

Conclusion

Important advances have been made since the beginning of the twenty-first century in the empirical study of campaign effects in France. The recent use of panel surveys, for 2002 and 2007, has allowed better discernment of the role that campaigns play in French presidential vote choice (Cautrès and Mayer, 2004; Cautrès and Muxel, 2009). For the present volume, we limited ourselves to the post-electoral surveys carried out after the last four presidential elections. This more traditional approach enabled us to uncover some new paths

9780230_300088_08_ch07.indd 160

6/28/2011 8:12:31 PM

Campaign Influences 161

that deserve further exploration, while at the same time confirming some of the conclusions made in previous work. Beyond the diffuse (or indirect) influence of campaigns through various vote determinants already examined in previous chapters, it seems clear that, since the end of the 1980s, people in France are deciding their vote later in the campaign and this new dynamic has important consequences for French electoral behavior. Young voters, women, and especially people with no strong ideological anchor seem to be more likely than others to decide their vote at the last minute. It also seems that the main reason behind a late vote decision is more likely bewilderment in the face of perceived complexity, rather than any elaborate strategic calculation attempting to account for all the information conveyed during the campaign. On the whole, candidates of the moderate left and moderate right seem to lose the most support, at least for the time being, due to French voters prolonging their moment of decision. Our analysis also indicates that campaign attentiveness has a limited impact on voting behavior. Certain factors, such as issue opinions and leader affect, seem to have greater weight on the decision-making process of voters who regularly follow the presidential campaign. However, the direct effect of attentiveness on firstround voting seems to be altogether negligible. The same conclusion holds true for the impact of television. Of course, candidates’ appearance on television during the campaign affects the opinions of them that voters form. However, it is difficult to establish with certainty a causal link between change in opinion and vote choice. All the same, televised debates between the first and second rounds affect the choices of those who watch them, but this effect does not seem nearly as important as other vote determinants. Unearthing either the presence or absence of campaign effects remains an important challenge for researchers. With new methodological and empirical tools, our theoretical understanding of the role and influence of electoral campaigns can only be enhanced and advanced. At this juncture, France is particularly well positioned. In addition to the numerous preexisting post-electoral studies that we have used in this volume, there are, increasingly, new sources of data that will enrich our understanding of the direct influence of electoral campaigns on the French voter.

9780230_300088_08_ch07.indd 161

6/28/2011 8:12:31 PM

162

French Presidential Elections

Hence, it will be possible in the future to further push the study of campaign effects in French presidential elections. An important example here involves taking into account the “quality” of the campaign of different candidates (Vavreck, 2009). We are witnessing a fortunate change already on display in the work of Cautrès and Muxel (2009). This is salutary because the proportion of French voters waiting until later to decide for whom they will vote is consistently, if mildly, on the rise. This issue also takes on increased importance within the electorate in this era of “presidentialization” of the French political system. The study of campaign effects therefore should more than ever be on the agenda of French electoral studies.

9780230_300088_08_ch07.indd 162

6/28/2011 8:12:31 PM

8 Conclusion

Contemporary France, as is often said, is quickly changing. So too is the character of French presidential elections. With the reordering of the election calendar and the possible end of cohabitation, and with both of these evolutions leading to a growing “presidentialization” of French politics, presidential contests have certainly taken on greater importance for understanding contemporary political dynamics in France. Add to this the fact that over the past few decades an increasing number of French voters have been supporting nontraditional presidential candidates, and it seems accurate to say that French politics, and especially French presidential voting behavior, is not what it used to be. These developments have led a number of analysts of contemporary French politics to turn their attention to the exceptional character of each individual presidential contest and to emphasize short-term dimensions in their assessment of electoral outcomes – not without good reason. But how much change and how much continuity can actually be observed over two decades of French presidential elections? Is the French voter mostly moved nowadays by short-term forces? Or, is his or her vote choice still influenced by some of the old cleavages? The analytical framework adopted in this book aims precisely at answering these core questions. First, it combines both long-term and shortterm determinants of electoral behavior in order to provide a comprehensive account of the various factors that affect presidential vote choice in France. Second, the analysis is longitudinal in perspective, assessing together the last four presidential contests so as to draw some solid and clear conclusions about the elements of change 163

9780230_300088_09_ch08.indd 163

6/28/2011 8:11:55 PM

164

French Presidential Elections

and continuity that shape the French voter’s presidential choice. Below, we provide a summary of the major empirical findings presented in this volume. For context, we begin with a discussion of the condition effects of the double-ballot system on key patterns of electoral behavior. Then, we offer a general organization of our findings, in terms of whether they offer evidence of change or continuity in contemporary French politics. This provides prelude to examination of the perennial question of whether the French electorate is “stalled.” We then assess the relative contribution of short-term and long-term determinants of French voting behavior in the context of presidential elections. Such considerations lead us to the weighty issue of theory-building in the study of French elections. We end this concluding chapter with a brief discussion of French “exceptionalism,” exploring the extent to which electoral behavior in France is distinct from that observed in other mature democracies. 8.1

First round versus second round: the institutional context

Before we get into the details of determinants of voting behavior, it is worth highlighting the role of the double-ballot system in shaping these presidential choices. Consider, as an example, how it plays out for Le Pen supporters. The FN clientele seems on many counts (e.g., profile, motivations) to be “outside” the traditional left-right order. A key feature of these presidential contests, 1988–2007, is the emergence of groups of voters (i.e., Greens and, more importantly, the FN) whose location on the political spectrum cannot be solely explained traditionally, in terms of conflicting views about the role of the state. Still, the appearance of these groups does not reflect a decline of the leftright oppositional framework. Rather, it suggests the transformation of its foundations. That is to say, other layers are being added to the traditional state intervention versus laissez-faire explanation of a respondent’s ideological self-identification. Ecology, ethnocentrism, and law and order are new dimensions helping to explain why people define themselves as being on the left or the right. Despite the growing complexities of the foundations of the left and right coalitions, the left-versus-right binary still controls in French politics. For instance, Sarkozy’s ability to attract FN voters in the first round of the 2007 election reminds us that Le Pen’s supporters are right-wing voters above all.

9780230_300088_09_ch08.indd 164

6/28/2011 8:11:55 PM

Conclusion

165

The double-ballot system thus induces two sorts of coalition strategies, in exploitation of the left-right divide. The first round provides the “diversity moment” for different groups to express themselves, the second round the “unity moment.” The coalition of quite diverse groups in the second round, behind one major candidate, is revealing on three accounts. First, it shows the diversity of the right and the left coalitions (on the left: educated and aging socialists, educated and young ecologists, and less educated and young voters of the extreme left; on the right: educated centrist voters, less educated and more feminine voters from the moderate right, and the less educated and more masculine clientele of the extreme right). Second, it shows the strength of left-right ideology in the second round as a rallying principle for such groups. Third, our systematic comparison of the first and second rounds shows very clearly the candidates’ different positions. In the first round they act as “niche-holders,” trying to mobilize a particular clientele. In the second round they are truly trying to keep together the multiple components of a coalition. This point finds illustration in the pre-election discussion about the potential candidates for the Socialists in 2012. Some people thought that Martine Aubry would be successful at winning the PS nomination, but not at winning the election. Others thought, at least for a time, that Dominique Strauss-Kahn would be a formidable opponent facing Sarkozy, but would have difficulty winning the PS nomination. (Of course, we see here that, because of the double-ballot system, French voters are driven by two sometimes conflicting motivations: conformity (to vote for the candidate closer to their views) and electability (to vote for the candidate that has the better chance of winning the decisive ballot.)

8.2 Change and continuity in French presidential elections One way in which French voting behavior is evolving, and one that has been much commented upon in the recent literature, is the declining influence of the sociological variables on vote choice. The results presented in this book offer a nuanced interpretation of this decline. Some socio-demographic variables still appear to exert a significant and stable influence on the French voter, at least during the time period covered (1988–2007). Paramount among them is religion. Religion has a strong influence on French vote choice, and consistently so. Religious

9780230_300088_09_ch08.indd 165

6/28/2011 8:11:55 PM

166

French Presidential Elections

Catholics are much more likely to support candidates from the moderate right. The effects of gender are also stable over time, though more modest than that of religion. Nevertheless, they are there. Women are attracted to the moderate right, and men support the extreme right in greater numbers than women. As for the other socio-demographic variables that were examined, age and education, their influence has not eroded but rather shifted over time. That is, it was found that young French voters since 1988 have slowly moved away from supporting the moderate left to supporting the moderate right, while having been particularly attracted to centrist candidate Bayrou in the 2007 election. A similar pattern was observed with the less educated individuals, who have shifted their support away from the moderate left toward the moderate right over the course of this 20-year period. The same type of nuanced conclusion applies with regard to the class and patrimony variables examined in the volume. We did observe an important decline over time of the traditional class vote in France, particularly in the first round of the presidential elections. Of particular note is what appears to have been a gradual “embourgeoisement” of the Socialist electorate. In particular, France has witnessed a fall in the traditional support of the left from blue-collar workers. Moreover, this fall in support is clearly more pronounced than the decrease in the vote for the right among its traditional supporters (e.g., managers, private-sector employees, holders of high-risk patrimony). Further, this erosion of the class vote was facilitated in part by the rise in support for the extreme right. And, the greater support obtained by centrist candidates also contributes to this dynamic since the centrist electorate was shown to be quite diversified sociologically. These elements of change that relate to the class variables contrast, however, with the continued influence of patrimony on French vote choice. As was shown, the owning of risky assets significantly discriminates between the moderate electorates of the left and right. The role of patrimony appears to be relatively stable over time and thus worthy of reconsideration in studies of French voting behavior. Our analysis of the role of social factors points to two other important conclusions. First, right-wing advances within the moderateleft camp have canceled out declines in the moderate right. However, for the left wing, the contrary to this pattern holds. These conflicting patterns have led to both a depolarization of social groups and the victory of right-wing politicians, at least in presidential contests since

9780230_300088_09_ch08.indd 166

6/28/2011 8:11:55 PM

Conclusion

167

1995. Second, if la tentation bayrouiste of 2007 persists into the future, it will mark a profound change in the relationship between certain key variables (age and religion) and support for centrist candidates. The almost undistinguishable difference between the centrists and the supporters of the moderate right in the past (from 1988 to 2002) may now become sharper. Perhaps a real centrist electorate is on the verge of emerging in the French political landscape. In France, scholars have commonly considered the dominant organizing principle of the electorate to be ideological identification along a left-right continuum, as opposed to party identification, as in some other democracies like the United States. In our interpretation of the long-term determinants that structure French vote choice, we sided with this view and provided strong evidence for the stable influence of left-right ideology on the vote, in both rounds of the presidential elections. Ideological identification thus remains an important psychological tool that helps French voters orient themselves in their political system. That being said, it was shown that the distribution of left-right ideology among the French public has fluctuated over time. From the normal distribution observed in previous studies of the 1960s and 1970s, ideology has adopted a more polarized distribution since at least 1988, with the traditional left group opposing a similarly large traditional right group. This trend of an ideologically polarized French electorate seems to have continued into the present. Focusing next on the short-term forces that influence presidential voting behavior in France, we found that the impact of economic issues on support for the “political economic incumbent” remained fairly stable during the time period examined. However, the target of the economic vote varies depending on the institutional context. Under unified government, in which the political executive responsible for managing the national economy is the president, French voters cast an economic vote for or against the candidate of the president’s party. Under cohabitation, it is the candidate of the prime minister’s party (who is often the prime minister him- or herself) who becomes the main target of economic voting. The two other issues considered in our analysis, immigration and Europe, displayed less stability in their impact on vote choice. Although the comparative literature has shown that issue voting has been on the rise in recent decades, no such tendency is apparent for France, at least in our findings. Indeed,

9780230_300088_09_ch08.indd 167

6/28/2011 8:11:55 PM

168

French Presidential Elections

what we observed is that the issues of immigration and Europe, as vote determinants, have tended to come and go from one presidential election to the next. Anti-immigration views lead one to support the moderate or extreme right, but their impact was strongest in 2002 and weakest in 2007. However, interestingly, that weakness may have been due to Sarkozy’s ability to convince supporters he was for more immigrant regulation. In other words, his UMP campaign managed to steal some of the “thunder” on that issue from the FN. With respect to European Union issues, their influence on the vote has been uneven and rather weak, with the traditional right seemingly coming to own this issue in recent years. As for the role of leader image in French vote choice, the main conclusion from our findings is that it has been substantial as well as consistent throughout the 20-year period covered by our study. The three different components of leadership – competence, integrity, and empathy – are all important. Right-wing leaders, in particular, seem to have been able to establish dominance on the dimensions of competence, which amount to the notion of l’étoffe d’un président (“the ‘right stuff’ for president”). This presence of “the right stuff” in right-wing leaders’ credibility may help explain in part why they have triumphed in three of the last four presidential elections. Since our empirical analyses do not go further back than 1988, it is difficult to tell if this major influence of leadership on vote choice has been greater in recent decades due to the presidentialization of French political life. But we think that the evidence presented in this book suggests that the major impact of de Gaulle’s image on vote choice during the 1960s may not have been particularly exceptional; instead, it appears rather characteristic of the general dynamics of presidential voting behavior in France. We can finally consider the elements of change and continuity in the influence of campaigns on the presidential choice of French voters. On one hand, it seems clear that the time period covered by this study has seen an increase in the number of people making their vote decision either during the campaign or at the last minute. This evolution suggests that today more French voters are susceptible to being influenced by the campaign in their decision. In particular, it was found that late deciders were more likely to support candidates from the center or the extremes, perhaps out of bewilderment or disillusion in the face of the traditional political fare. On the other

9780230_300088_09_ch08.indd 168

6/28/2011 8:11:55 PM

Conclusion

169

hand, we have also seen that presidential campaigns have had minimal direct effects. Campaign attentiveness tends to fluctuate from one presidential contest to the other and does not have much of an effect on the vote. The leaders’ campaign on television does affect their respective support, but only at the margins. This general lack of direct effects does not mean that campaigns have no influence at all, of course, since one of their chief roles is to help candidates mobilize or remobilize their respective constituencies and enable them to make their positions on issues and their personal competencies known to the public. The influence of presidential campaigns is thus captured indirectly through the effects of the long-term and shortterm variables on the vote, a phenomenon that certainly constitutes an element of continuity in French presidential elections. To summarize, important changes were observed between 1988 and 2007 in the relationship between presidential vote choice and the variables of age, education, and class, as well as time of vote decision. The overall distribution of left-right ideological identification in the mass electorate also showed signs of greater polarization than before. Variables that displayed no change over time in their significant influence on the vote were religion, gender, patrimony, ideology, the economy, and leader image. As for the remaining independent variables examined in this book, the issues of immigration and Europe, their impact on the presidential vote was found to be inconsistent.

8.3

Is the French electoral system “stalled”?

In 1984, Lewis-Beck (1984, p. 443) argued, “Under the Fifth Republic, the attitudes and behavior of the French electorate appear to have changed little, at least with regard to the critical variables studied here…. class, religion, region, and ideology have maintained their substantial influence on party preference. …the 1981 election results do not undercut the conclusion that the French electorate is ‘stalled’ in past habits.” Writing 20 years later, regarding the 2002 elections, LewisBeck (2004, p. 11) held to this conclusion, observing that “the continuities are much more impressive than the changes.” It is worth addressing this claim, in the light of the extensive results presented here. Different approaches to this hypothesis can be taken. A problem with the French case, as always, is seeing the forest for the trees. A useful simplification comes from focusing on the second round of the

9780230_300088_09_ch08.indd 169

6/28/2011 8:11:56 PM

170 French Presidential Elections

presidential contest, and the influence of variables that regularly appear in the literature. Such variables fall into the category we have labeled “long-term,” namely social factors and ideological identification. In any discussion of the determinants of French electoral behavior, these long-term forces stand front and center. An early attempt was made to assess the impact of these variables over time, for a collection of Western democracies, by examining the variance they explained in the vote choice (Franklin et al., 1992). The French results for that exercise are suggestive (Lewis-Beck and Skalaban, 1992, pp. 172, 176). For the two pivotal elections under study (1968 and 1981), social factors (i.e., occupation, income, gender, education, community size, religion) accounted for 8.3 percent and 7.7 percent of the variance, respectively. When leftright ideology was added to the regression equation, these R-squared rose to 33 percent and 49 percent, respectively. Suppose we follow a similar strategy for the second-round presidential elections at hand. We observe, first, the pseudo-R-squared for the social factors, election by election: 1988, .12; 1995, .11, 2002, .08; 2007, .09 (see the Appendix, Model 2, column 2, second round). These social factors explain about the same amount of variance, regardless of election. (Moreover, that amount is quite close to the amount explained in the previous elections noted above.) We observe, second, the pseudo-R-squared for the model that includes social factors and left-right ideology: 1988, .58; 1995, .56; 2002, .34; 2007, .57 (see the Appendix, Model 3, column 3, second round). This long-term forces model explains over one-half the variation in these presidential votes (with one exception, in which it explains over onethird). The power and stability of the explanation persists across these four contests, despite the inevitable vagaries of political life. To conclude that the French electorate remains “stalled,” with voters driven by the same old forces, would not seem far-fetched.

8.4

Short-term vs. long-term factors

While the French electorate may be “stalled” with respect to the influence of long-term forces, the picture becomes more nuanced when the short-term forces acting on the voter are considered. What is the relative weight of short-term and long-term determinants of the vote in the French voter’s decision? Is there one set of factors that seems to matter more than the other? Is the pattern different from one

9780230_300088_09_ch08.indd 170

6/28/2011 8:11:56 PM

Conclusion

171

round of election to the other? Is this consistent across the whole set of presidential elections examined in this study? Are there exceptions worth reporting? We briefly address these questions below. A useful tool for shedding light on these issues involves further comparison of the model fits, as measured by the regression equation coefficients, the pseudo-R-squared (see the Appendix). More precisely, we can take the pseudo-R-squared obtained from the model with all the long-term determinants and compare it to the model that adds to it all the short-term factors. That is, compare Model 3 to Model 6 (both in the Appendix, columns 3 and 6, respectively). The difference between the two gives us an indication of the short-term determinants’ contribution to the overall explanation of vote choice, as opposed to an explanation that rests strictly on long-term variables. (Furthermore, while this comparison is sequential, in that the short-term variables are added later, it is not statistically biased, given that the assumptions of our block recursive estimation are met. In particular, the assumption that the variables of a prior block are effectively exogenous vis-à-vis a later block is fully plausible. This assumption conforms well to the funnel of causality theory, which has guided the model specification throughout. That is, the variables formed at a more remote time, such as social class, are at the mouth of the funnel, while those formed more immediately, such as economic opinion, are at the tip of the funnel.) The first conclusion that emerges from examining the difference in the pseudo-R-squareds is that long-term factors seem to contribute more than short-term ones to an explanation of the French presidential vote decision. This is particularly true for the second-round vote. As noted, the pseudo-R-squareds from Model 3 (long-term determinants) are as follows: 1988, 0.58; 1995, 0.56; 2002, 0.34; 2007, 0.57. The pseudo-R-squared from Model 6 (long-term plus short-term determinants) are as follows: 1988, 0.77; 1995, 0.67; 2002, 0.52; 2007, 0.73. Here now is the added explanatory value due to the addition of short-term variables (Model 6 – Model 3): 1988, 0.19; 1995, 0.11; 2002, 0.18; 2007, 0.16. To sum up, long-term determinants can account for about half of the total variance in the second-round presidential vote, whereas short-term factors can add no more than an additional one-fifth to this total variance. Even in the case of the 2002 presidential election, in which long-term variables do not perform as well as for the other three contests, they still manage to

9780230_300088_09_ch08.indd 171

6/28/2011 8:11:56 PM

172

French Presidential Elections

account for nearly twice as much variance in vote choice as shortterm factors do. Upon first glance, the same basic conclusion applies to the firstround vote: in general, long-term determinants account for more of the variance in candidate support than short-term variables during the first round of presidential elections. However, there are two very important exceptions to this trend. The first involves the 2002 election, for which the pattern is reversed: the short-term variables account for much more of the variance in the first-round presidential vote than do the long-term variables. Leader evaluations, in particular, exerted a powerful influence on the vote decision for all five main candidates in that election, as we see the pseudo-R-squared statistics jumping with the introduction of this fifth block of variables. In addition to leadership, issues weighed heavily in the choice to support Jean-Marie Le Pen in the first round of this contest. Thus, the 2002 presidential election turns out to have been an exceptional contest precisely because both the issue environment and the leaders’ images played a much more prominent role in the French voters’ decision than the usual sociological and psychological cleavages. The second exception to the pattern is that short-term determinants appear to be more important for explaining first-round support for the extreme right (Le Pen) as well as for centrist candidates. This is something that can be observed for all four presidential elections examined in this book (except for the case of centrist Edouard Balladur in 1995). Put differently, our findings indicate that in the first round, long-term determinants have more of an impact on the vote for candidates of the moderate left and right, as well as candidates of the extreme left, than short-term considerations do. Hence, support for nontraditional options seems to be more attributable to issues, leader image, and the campaign. This tendency disappears in the second round (with the obvious exception of 2002) since the two remaining candidates usually come from the two traditional political camps. These observations suggest that France’s electoral rules, which allow a wide-ranging political menu on the first ballot, enable shortterm considerations to play a more important role in the vote decision on the first, as opposed to the second, ballot. In the event that the political context and the issue environment were again to favor support for nontraditional alternatives, as in 2002, the scenario

9780230_300088_09_ch08.indd 172

6/28/2011 8:11:56 PM

Conclusion

173

might be a second-round candidate from one of the extremes, or even from the center.

8.5

Is a theory of French presidential elections possible?

There are many studies of national elections in France, as the bibliography drawn upon here makes clear. Election surveys, in particular, have flourished since the 1970s. While the scholarly publications from these surveys have been plentiful and potent, they have pursued a research design that is intrinsically cross-sectional. That is, each presidential (or legislative) election finds itself treated as a unique event, to be understood largely on its own terms. It is thoroughly analyzed, but the analysis tends to be sui generis. Certainly, basic variables are routinely measured, but they are not always measured in the same way, and, more critically, they are not girded by a general theory of the vote. One practical consequence, then, is that these publications tend to feel independent of each other, with little knowledge accumulation. (An exception here would be the study of les variables lourdes, the “heavy” variables.) Our argument rests on the assumption that a theory of the presidential vote can be applied, to good effect, across a series of elections. The vote, we hold, is a political behavior, and one that reliably responds to pretty much the same stimuli, regardless of democratic setting. In other words, voters, in making their choices, are subject to discoverable “laws” in functioning democracies. Furthermore, these laws are composed of long-standing social conditions and psychological propensities, coupled with immediate issues and opportunities (or obstacles). We have attempted to formulate, or synthesize, these laws into a French presidential voting model applicable across time, as tested against the last four elections. How well does this model work? Take Model 6 (column 6, the Appendix) as the fullest statement of the model. (Technically speaking, this equation estimates the direct effect of each of the independent variables identified, controlling for the influence of all the other variables.) That model manages to account for over one-half to over three-quarters of the variation in presidential vote choice. In other words, it offers a theoretical explanation for why most French citizens voted the way they voted in these presidential elections.

9780230_300088_09_ch08.indd 173

6/28/2011 8:11:56 PM

174

French Presidential Elections

French voters, then, are neither renegade nor rogue. Their actions are largely explainable, as are voters’ in other leading Western democracies. Indeed, an interesting parallel comes from comparing the single-equation logistic regression model of Sarkozy’s election in 2007 to a similarly specified model for Obama’s in 2008. For the former, the pseudo-R-squared equals .73; for the latter, the pseudo-R-squared equals .69 (Lewis-Beck et al., 2009). Presidential voting behavior in these two democracies, different as they are, still are equally amenable to explanation.

8.6

The French voter in perspective

Compared to the electorates of other mature democracies like the United States and the United Kingdom, it appears that the French voter, very much like the American or the British voter, is continually influenced by a number of heavy variables that structure his or her electoral decision. These long-term influences act in a similar way across these polities. The case, then, for French exceptionalism, at least with regard to national elections, appears weak. However, that does not mean the individual causal factors are necessarily all of the same nature or importance. In France, religion is a particularly important variable, the gender gap in voting appears to be reversed, and left-right ideology trumps party as the core psychological anchor of the vote. Like voters in these other major democracies, the electorate of France is also subject to important short-term forces when making a presidential choice. These considerations play a substantive role in the first round of the election. When the context makes their influence weightier than usual, they have the potential of overshadowing long-term sociological and ideological cleavages in the second round of the contest. In other words, short-term determinants of the vote – such as economic perceptions, opinions on the immigration issue, leader image, and campaign-specific variables – act similarly in France as in other electoral democracies; however, these variables’ influence can be either downplayed or enhanced due to the political context and the French electoral rules (especially the double ballot). The notion that the French voter’s behavior is more influenced by long-term than by short-term considerations is one of the key observations that derive from our application of a Michigan-style explanatory

9780230_300088_09_ch08.indd 174

6/28/2011 8:11:56 PM

Conclusion

175

model to the French case and from the longitudinal perspective that we adopted at the outset of this study. While after each presidential election it is crucial to try to understand voter movements and volatility, it is equally important to assess how each new election outcome fits into the continuity of French electoral politics. Only by doing so can we appreciate the extent to which each presidential contest may or may not be exceptional. In that respect, it increasingly appears that the 2007 election may not have been a “rupture” of tendencies, such as the rise of protest voting, so much as a return to normal. Still, the normalcy argument cannot stand without qualification. The fact that ideology loomed so large in 2007 is what might be expected, in this unusual case in which no incumbent (parliamentary or presidential) was running. Voters in such a circumstance have to rely on traditional shortcuts (e.g., ideology) rather than on assessments of the incumbent’s record, in order to make a choice. Also, it must be borne in mind that shortterm factors (the appeal of Bayrou, the ability of Sarkozy to attract FN voters) also played a vital role in this election. Finally, some of the changes that occurred – the emergence of a new centrist electorate and the ability of the moderate-right candidate to attract FN supporters – may perhaps become more enduring features of French politics in contests to come. What next? First, there is the question of the design of presidential election studies. Second, there is the question of the outcome of the elections themselves. With respect to the first question, a general and a specific point should be made. French presidential elections, like elections in other leading Western democracies, are shaped by long-term and short-term forces acting on individual voter behavior. Indeed, we have shown that it is possible to model this behavior, by testing a theory of elections against a series of election data sets. The success of such efforts in the future depends on surveys that contain full and consistent complements of items measuring these long-term and short-term forces. This brings us to the specific point: improvement in the measurement of issues (especially the economy), leader traits, and patrimony are much needed, both in terms of scope and continuity. Let us hope they are forthcoming. With respect to the second question, what does this study tell us about the looming 2012 presidential contest? Of course, at this distance, we can only speculate. However, what we know is that the right

9780230_300088_09_ch08.indd 175

6/28/2011 8:11:56 PM

176

French Presidential Elections

has been successful since 1995. Why is the left losing? Perhaps it is because the right is succeeding in coming to own certain issues (e.g., immigration and the EU) and, at the same time, has candidates who exhibit more appealing leadership characteristics, candidates with “the right stuff.” Of course, these areas – issues and leadership – are short-term forces only, and they ignore the fundamental attachments voters have. Nevertheless, it is the short-term forces that are usually most amenable to change through wise decision-making.

9780230_300088_09_ch08.indd 176

6/28/2011 8:11:56 PM

Appendix Below are the estimated binomial logistic regression equations forming the basis of the multivariate discussion in each chapter. It is important to understand how to read them. The equations are keyed sequentially to the independent variables introduced into the analysis chapter after chapter. For example, look at Table A.7, “Reported vote for the extreme left (Laguiller or Hue), French presidential election, 1995, first round.” Column (1) considers the impact of the socio-demographic variables of Chapter 2 – age, gender, education, and religion; column (2) adds to the equation the social-structure variables of class and patrimony of Chapter 3; column (3) adds the ideology variables of Chapter 4; column (4) adds the issue variables of Chapter 5; column (5) adds the feeling thermometer variables of Chapter 6; column (6) adds the campaign variables of Chapter 7. This is the general strategy followed for each table. Note, furthermore, that there is a table for each of the possible first-round vote choices – extreme-left candidate, moderate-left candidate, center candidate, moderate-right candidate, extreme-right candidate – and for the second-round vote choices –right candidate or not. Thus, vote choice in each election is represented by six tables – that is, tables A.1 – A.6 cover 1988, tables A.7–A.12 cover 1995, and so on. The title of each table identifies the dependent variable and the election year. By way of illustration, here is the specification for the second-round vote equation: Vote = socio-demographics + social structure + ideology + issues + leadership + the campaign. A word about the data sets is in order. For 1988, two data sets were used, that of the Centre d’étude de la vie politique française (CEVIPOF) for models in columns 1 to 3 and that of Pierce (1988) for models in columns 4 and 5. (Two data sets were necessary for 1988, because all the necessary items were not in one survey). The CEVIPOF survey was face-to-face and post-electoral, carried out by

177

9780230_300088_10_app.indd 177

6/28/2011 8:01:09 PM

178 French Presidential Elections

Société française d’enquêtes par sondages (SOFRES) between May 9 and May 20, 1988, N = 4032. The Pierce (1988) survey was face-toface and post-electoral, carried out by SOFRES in May, N = 1013. (In the tables for 1988, the first three columns of estimates come from the CEVIPOF data; the last two columns of estimates come from Pierce (1988). With respect to the other elections, CEVIPOF surveys were always used. For 1995, the survey was face-to-face and postelectoral, carried out by SOFRES between May 8 and May 23, N = 4078. For 2002, we used respondents from the first two waves of a three-wave panel. The first wave was pre-electoral and done faceto-face between April 8 and April 20, N = 4107. The second wave was by telephone between May 15 and May 31, between the two rounds of voting, N = 1822 (panelists over the two rounds). The voting questions we used were asked in the second wave; the firstround vote is reported, the second-round vote is intention. For 2007, the survey was face-to-face and post-electoral, carried out by l’Institut français d’opinion publique (IFOP) between May 9 and May 23, N = 4006. Table A.1 Reported vote for the extreme left (Lajoinie, Juquin, or Laguiller), French presidential election, 1988 (1)

(2)

(3)

(4)

(5)

Managers

–.32 (.30) .20* (.12) .05 (.19) –1.76*** (.17) _

Employees

_

.08 (.35) .14 (.14) .20 (.27) –1.73*** (.21) –.56** (.24) –.39* (.20)

.18 (.39) .11 (.15) .20 (.29) –1.01*** (.23) –.59** (.25) –.48** (.22)

–1.20 (.76) –.09 (.34) –1.33** (.54) –1.49*** (.47) .18 (.48) .26 (.46)

–.86 (.82) –.04 (.35) –.99* (.59) –1.33** (.52) .36 (.50) .45 (.49)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 178

6/28/2011 8:01:09 PM

Appendix 179

Table A.1 Continued (1)

(2)

(3)

(4)

(5)

Workers

_

Income

_

Private-sector

_

.21 (.21) .00 (.43) –.61*** (.14)

.08 (.22) .27 (.47) –.42*** (.16)

.31 (.46) .90 (1.05) –.42 (.34)

.28 (.50) 1.36 (1.12) –.21 (.36)

Low-risk assets

_

_

_

_

_

Extreme-left

_

Left

_

_

Right

_

_

Extreme-right Immigration

_ _

_ _

–.36 (.31) –.84** (.39) 3.52*** (.32) 1.46*** (.22) –1.27*** (.46) +Dropped _

_

High-risk assets

–.31 (.29) –1.35*** (.37) _

Economy

_

_

_

Thermometer

_

_

_

2.23** (.92) 1.42*** (.51) –1.00 (.86) +Dropped –.87* (.52) –1.74*** (.50) _

–1.39*** (.17) .04

–.65* (.35) .09

–2.24*** (.43) .21

–.37 (1.08) .28

1.51 (1.03) .60 (.55) –.72 (.87) +Dropped –.76 (.56) –1.22** (.55) 4.23*** (.78) –2.65** (1.19) .37

3694

2993

2869

511

505

Constant Pseudo-R 2 N

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. The Assets variables are not included in columns 5 and 6 because they were not measured in Pierce (1988). The Last Minute Decision variable is not in the table because it was not measured in the 1988 surveys. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (1988) and Pierce (1988).

9780230_300088_10_app.indd 179

6/28/2011 8:01:10 PM

180

French Presidential Elections

Table A.2 Reported vote for the left (Mitterrand), French presidential election, 1988 (1)

(2)

(3)

(4)

(5)

Managers

.06 (.17) –.16** (.07) –.42*** (.12) –.67*** (.13) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

1.79*** (.65) –.49* (.29) 1.14** (.46) .55 (.44) –.53 (.39) –.22 (.39) .11 (.41) –1.02 (.85) .13 (.29) _

_

_

_

_

Left

_

_

Right

_

_

Extreme-right

_

_

.69 (.81) 1.96*** (.30) –2.41*** (.48) +Dropped

–.31 (.92) .90*** (.34) –1.37*** (.51) +Dropped

Immigration

_

_

.65** (.26) –.24** (.10) –.38** (.19) .44** (.17) .20 (.16) .21 (.15) .30* (.15) .11 (.30) –.12 (.11) .30 (.21) –.37* (.22) –.25 (.30) 1.65*** (.10) –2.34*** (.22) –3.25*** (1.01) _

1.56*** (.56) –.31 (.26) .81** (.40) .49 (.39) –.72** (.34) –.34 (.35) .02 (.36) –.88 (.75) .24 (.26) _

High-risk assets Extreme-left

.41* (.21) –.21** (.09) –.18 (.16) –.52*** (.15) .00 (.14) .13 (.13) .44*** (.13) –.07 (.25) –.31*** (.09) .22 (.17) –.84*** (.19) _

Economy

_

_

_

Thermometer

_

_

_

.50 (.39) .61* (.34) _

.52 (.44) .71* (.39) 7.29*** (.93)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 180

6/28/2011 8:01:10 PM

Appendix 181

Table A.2 Continued (1)

(2)

(3)

(4)

(5)

Pseudo-R 2

–.02 (.12) .01

–.14 (.22) .03

–1.54*** (.27) .26

–2.13** (.78) .30

–6.99*** (1.09) .43

N

3694

2993

2940

511

508

Constant

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. The Assets variables are not included in columns 5 and 6 because they were not measured in Pierce (1988). The Last Minute Decision variable is not in the table because it was not measured in the 1988 surveys. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (1988) and Pierce (1988).

Table A.3 Reported vote for the center (Barre), French presidential election, 1988 (1)

(2)

(3)

(4)

(5)

Managers

.98*** (.24) –.11 (.10) .97*** (.16) 1.52*** (.22) _

Employees

_

Workers

_

Income

_

Private sector

_

Low-risk assets

_

.74** (.32) –.07 (.13) .39* (.22) 1.27*** (.25) –.02 (.17) .04 (.17) –.59*** (.21) .94*** (.36) .14 (.13) .46* (.26)

.65** (.33) –.06 (.13) .42* (.23) .36 (.27) –.08 (.18) .09 (.18) –.44** (.21) .78** (.37) –.09 (.14) .50* (.27)

.18 (.65) .42 (.33) .39 (.48) .45 (.63) .04 (.39) –.04 (.43) –.37 (.52) 1.12 (.82) .12 (.34) _

.59 (.79) .75* (.41) .70 (.60) –.55 (.77) –.02 (.46) .29 (.52) –.46 (.62) 1.13 (.100) .09 (.40) _

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 181

6/28/2011 8:01:10 PM

182

French Presidential Elections

Table A.3 Continued (1)

(2)

High-risk assets

_

Extreme-left

_

.87*** (.23) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

–2.63*** (.23) –.05 (.13) –1.14*** (.44) _

Economy

_

_

_

Thermometer

_

_

_

–1.33*** (.49) .08 (.37) _

Constant

–3.55*** (.20)

–3.99*** (.33)

–2.45*** (.34)

–1.78* (.94)

–1.80*** (.60) –.51 (.45) 9.11*** (1.23) –6.91*** (1.38)

Pseudo-R2

.04

.07

.17

.23

.43

3694

2993

2860

502

499

N

(3)

(4)

(5)

.57** _ _ (.25) +Dropped +Dropped +Dropped –3.75*** –3.40*** (.65) (.73) –.30 –.79** (.30) (.38) +Dropped +Dropped

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. The Assets variables are not included in columns 5 and 6 because they were not measured in Pierce (1988). The Last Minute Decision variable is not in the table because it was not measured in the 1988 surveys. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF and Pierce (1988).

Table A.4 Reported vote for the right (Chirac), French presidential election, 1988

Age Gender Education

(1)

(2)

(3)

(4)

(5)

1.16*** (.22) .03 (.09) .46*** (.15)

.87*** (.29) –.08 (.12) .02 (.21)

.70** (.32) –.13 (.13) –.07 (.23)

.28 (.65) –.30 (.34) –.74 (.47)

.00 (.73) –.43 (.37) –.48 (.51) Continued

9780230_300088_10_app.indd 182

6/28/2011 8:01:10 PM

Appendix 183

Table A.4 Continued (1)

(2)

Managers

2.36*** (.22) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

High-risk assets Extreme-left

_ _

2.34*** (.27) .25 (.16) –.11 (.17) –.62*** (.19) .48 (.33) .55*** (.13) .16 (.24) .74*** (.22) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

–2.69*** (.31) 1.47*** (.13) .39 (.32) _

Economy

_

_

_

Thermometer

_

_

_

–3.70*** (1.05) 1.50*** (.33) .71 (.78) –.77 (.50) .95** (.37) _

–3.76*** (.19) .06

–4.25*** (.32) .10

–3.02*** (.36) .30

–1.85** (.94) .35

–3.09*** (1.07) .72* (.38) .12 (.92) –1.18** (.54) .18 (.41) 5.52*** (1.06) –4.27*** (1.13) .42

3694

2993

2860

514

511

Religion

Constant Pseudo-R 2 N

(3)

(4)

(5)

1.19*** .21 .15 (.29) (.67) (.71) .18 .37 .44 (.18) (.39) (.42) –.17 –.31 .00 (.19) (.45) (.48) –.47** –.22 –.04 (.21) (.51) (.54) .09 –.10 –.27 (.37) (.77) (.84) .32** .49 .50 (.14) (.35) .38) .15 _ _ (.26) .23 _ _ (.25) +Dropped +Dropped +Dropped

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. The Assets variables are not included in columns 5 and 6 because they were not measured in Pierce (1988). The Last Minute Decision variable is not in the table because it was not measured in the 1988 surveys. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF and Pierce (1988).

9780230_300088_10_app.indd 183

6/28/2011 8:01:10 PM

184

French Presidential Elections

Table A.5 Reported vote for the extreme right (Le Pen), French presidential election, 1988 (1)

(2)

(3)

(4)

(5)

Managers

–.19 (.28) .50*** (.11) –.48** (.20) .57** (.23) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets Extreme-left

_

–1.29 (.93) .29 (.44) –.02 (.64) .21 (.90) –.07 (.57) .42 (.57) –.12 (.64) .87 (1.04) –.27 (.46) _

_

_

_

+Dropped

+Dropped

Left

_

_

Right

_

_

Extremeright Immigration

_

_

_

_

–.51 (.36) .59*** (.15) –.75*** (.27) .00 (.29) –.25 (.22) –.00 (.22) .14 (.22) .76* (.42) –.05 (.15) –.02 (.29) –.07 (.30) –2.34** (1.02) –1.41*** (.21) .67*** (.16) 2.39*** (.27) _

–1.91** (.82) .28 (.38) –.29 (.55) .49 (.71) .16 (.49) .74 (.48) –.04 (.54) .20 (.91) –.62 (.39) _

_

–.37 (.34) .60*** (.14) –.86*** (.26) .76*** (.26) –.15 (.21) .04 (.20) –.02 (.20) 1.02*** (.40) .15 (.15) –.08 (.28) .38 (.28) _

Economy

_

_

_

Thermometer

_

_

_

–2.00*** (.60) .36 (.37) 2.61*** (.85) 3.69** (.65) –.64 (.42) _

–1.57** (.70) –.53 (.45) –.42 (.96) 1.83** (.75) –.23 (.51) 5.78*** (.85)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 184

6/28/2011 8:01:11 PM

Appendix 185

Table A.5 Continued

Constant Pseudo-R 2 N

(1)

(2)

(3)

(4)

(5)

–2.63*** (.20) .01

–3.17*** (.37) .03

–2.30*** (.39) .15

–3.19*** (1.09) .34

–4.81** (1.30) .50

3694

2993

2940

514

511

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. The Assets variables are not included in columns 5 and 6 because they were not measured in Pierce (1988). The Last Minute Decision variable is not in the table because it was not measured in the 1988 surveys. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF and Pierce (1988).

Table A.6 Reported vote for the right (Chirac), second round of the French presidential election, 1988 (1)

(2)

(3)

(4)

(5)

Managers

1.02*** (.19) .07 (.08) .60*** (.13) 2.32*** (.17) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets

_

.67*** (.24) .05 (.10) –.17 (.18) 2.29*** (.21) .16 (.14) –.10 (.14) –.67*** (.15) 1.14*** (.28) .48*** (.10) .11 (.20) 1.29*** (.19)

.71* (.37) .03 (.15) –.27 (.27) .94*** (.31) –.11 (.22) –.27 (.21) –.72*** (.21) 1.18*** (.43) .18 (.16) .40 (.31) .96*** (.30)

.21 (.78) .16 (.37) .22 (.58) 1.90*** (.73) 1.04** (.49) –.06 (.48) .43 (.58) 1.22 (1.02) .01 (.40) _

–.48 (.96) .01 (.48) .88 (.73) 2.49*** (.92) 1.06* (.61) .56 (.63) .60 (.74) 1.50 (1.19) .40 (.50) _

_

_

Age Gender Education Religion

_

Continued

9780230_300088_10_app.indd 185

6/28/2011 8:01:11 PM

186

French Presidential Elections

Table A.6 Continued (1)

(2)

(3)

(4)

(5)

Extreme-left

_

_

+Dropped

+Dropped

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

–3.61*** (1.02) –3.60*** (.22) 2.58*** (.17) 2.37*** (.48) _

Economy

_

_

_

Thermometer

_

_

_

–3.69*** (.57) 2.87*** (.41) 1.84** (.93) .24 (.58) 1.80*** (.47) _

Constant

–2.53*** (.15)

–3.17*** (.27)

–1.72*** (.40)

–3.90*** (1.18)

–3.33*** (.64) 2.02*** (.50) 2.95 (2.18) .12 (.70) .21 (.63) 10.78*** (1.70) –9.61*** (1.80)

Pseudo-R2

.07

.12

.58

.66

.77

3262

2661

2632

527

524

N

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. The Assets variables are not included in columns 5 and 6 because they were not measured in Pierce (1988). The Last Minute Decision variable is not in the table because it was not measured in the 1988 surveys. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF and Pierce (1988).

Table A.7 Reported vote for the extreme left (Laguiller or Hue), French presidential election, 1995 (1) Age Gender Education Religion

(2)

(3)

–.94*** –.56 –1.06*** (.28) (.34) (.37) .14 .17 .13 (.11) (.13) (.14) –.53*** .12 –.13 (.18) (.25) (.27) –1.46*** –1.36*** –.75*** (.17) (.19) (.21)

(4)

(5)

(6)

–.95** (.40) .16 (.15) –.11 (.30) –.71*** (.22)

–.98** (.43) .20 (.16) –.18 (.32) –.58** (.24)

–.69 (.44) .23 (.16) –.26 (.32) –.59** (.24) Continued

9780230_300088_10_app.indd 186

6/28/2011 8:01:11 PM

Appendix 187

Table A.7 Continued (1)

(2)

(3)

(4)

(5)

(6)

Managers

_

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets Extreme-left

_

_

–.99*** (.25) –.02 (.18) .34* (.19) –.78* (.43) –.09 (.14) .12 (.27) –.24 (.30) _

Left

_

_

Right

_

_

Extremeright Immigration

_

_

_

_

–.72*** (.26) .06 (.20) .10 (.20) –.94** (.47) .07 (.14) .24 (.28) .15 (.31) 1.70*** (.22) 1.09*** (.17) –1.65*** (.31) –2.58** (1.02) _

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

–.74*** (.27) .00 (.20) .02 (.21) –.95* (.50) .08 (.15) .32 (.30) .26 (.33) 1.54*** (.24) 1.10*** (.18) –1.56*** (.31) –2.55** (1.02) –.74*** (.22) –.40* (.21) 1.64*** (.26) _

Last-minute decision Constant

_

_

_

_

–.58** (.28) .14 (.22) .06 (.23) –.90* (.52) .09 (.16) .19 (.32) .08 (.35) 1.00*** (.26) .73*** (.19) –1.14*** (.32) –1.92* (1.03) –.43* (.22) –.41* (.22) 1.47*** (.28) 4.53*** (.42) _

–.72*** (.18) .04

–.69** (.35) .07

–1.23*** (.40) .19

–.60** (.28) .11 (.22) .03 (.23) –.93* (.52) .08 (.16) .11 (.32) .11 (.35) 1.20*** (.27) .83*** (.20) –1.05*** (.33) –1.66 (1.04) –.52** (.24) –.46** (.22) 1.53*** (.28) 4.54*** (.42) .72*** (.18) –4.25*** (.58) .31

3249

2636

2614

Pseudo-R 2

_

N

–1.30*** –3.93*** (.47) (.57) .22 .30 2465

2422

2419

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (1995).

9780230_300088_10_app.indd 187

6/28/2011 8:01:11 PM

188

French Presidential Elections

Table A.8 Reported vote for the left (Jospin), French presidential election, 1995 (1)

(2)

Managers

.78*** (.20) –.01 (.08) .30** (.13) –1.03*** (.13) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

High-risk assets

_

Extreme-left

_

1.22*** (.25) –.07 (.10) .34* (.18) –.89*** (.15) –.18 (.15) –.07 (.13) .30** (.15) .13 (.31) –.46*** (.10) .15 (.20) –.66*** (.21) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

Economy

_

_

Europe

_

_

Age Gender Education Religion

(3)

(4)

(5)

(6)

1.49*** 1.51*** 1.43*** 1.24*** (.30) (.32) (.35) (.36) –.14 –.18 –.13 –.16 (.12) (.12) (.13) (.13) –.06 –.36 –.24 –.20 (.22) (.24) (.26) (.26) –.15 –.04 –.01 –.00 (.18) (.19) (.21) (.21) .18 .22 .28 .29 (.18) (.19) (.21) (.21) .00 .01 –.02 –.01 (.16) (.17) (.18) (.19) –.04 .12 .00 .01 (.17) (.18) (.20) (.20) .21 .12 –.08 –.10 (.38) (.41) (.44) (.44) –.31*** –.37*** –.26* –.27** (.12) (.13) (.14) (.14) .31 .42* .52* .55** (.23) (.25) (.27) (.27) –.27 –.34 –.26 –.27 (.25) (.26) (.29) (.29) 1.01*** 1.04*** .13 .02 (.19) (.21) (.24) (.24) 1.72*** 1.65*** .91*** .86*** (.13) (.14) (.15) (.15) –2.27*** –2.30*** –1.73*** –1.79*** (.22) (.24) (.25) (.25) –3.72*** –3.56*** –2.86*** –2.98*** (1.01) (1.01) (1.04) (1.03) _ –.46** –.53*** –.50** (.19) (.20) (.21) _ –.31* –.32* –.31 (.17) (.19) (.19) _ –1.53*** –1.19*** –1.21*** (.24) (.26) (.26) Continued

9780230_300088_10_app.indd 188

6/28/2011 8:01:12 PM

Appendix 189

Table A.8 Continued (1)

(2)

(3)

(4)

(5)

Thermometer

_

_

_

_

Last-minute decision Constant

_

_

_

_

5.22*** (.37) _

–.77*** (.14)

–.74*** (.26)

–1.67*** (.32)

–.72* (.39)

Pseudo-R2 N

(6)

5.19*** (.37) –.47*** (.16) –4.23*** –3.96*** (.50) (.50)

.02

.04

.30

.32

.41

.42

3249

2636

2614

2465

2452

2449

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (1995).

Table A.9 Reported vote for the center (Balladur), French presidential election, 1995 (1)

(2)

(3)

(4)

(5)

(6)

Managers

1.90*** (.25) –.30*** (.10) .71*** (.16) 2.06*** (.22) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

1.96*** (.32) –.16 (.13) .38* (.22) 1.96*** (.26) .03 (.17) .11 (.17) –.44** (.21) .43 (.37) .45*** (.14) .07 (.25)

2.20*** (.33) –.15 (.13) .52** (.23) 1.32*** (.26) –.08 (.18) .11 (.18) –.15 (.22) .45 (.39) .32** (.15) –.06 (.27)

2.09*** (.35) –.29** (.14) .43* (.25) 1.17*** (.27) –.07 (.19) .16 (.19) –.03 (.23) .36 (.41) .28* (.15) –.11 (.28)

1.39*** (.38) –.19 (.15) .54** (.26) .94*** (.29) –.09 (.21) .28 (.20) .05 (.25) .49 (.44) .21 (.17) –.24 (.31)

1.39*** (.38) –.19 (.15) .53** (.27) .94*** (.29) –.09 (.21) .28 (.20) .05 (.25) .49 (.44) .21 (.17) –.25 (.31)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 189

6/28/2011 8:01:12 PM

190

French Presidential Elections

Table A.9 Continued

High-risk assets Extreme-left Left Right Extreme-right Immigration Economy Europe Thermometer Last-minute decision Constant Pseudo-R 2 N

(1)

(2)

_

.55** (.23) _

3249

2636

(3)

(4)

(5)

(6)

.30 .20 .30 .31 (.24) (.25) (.27) (.27) _ –3.45*** –3.27*** –2.25** –2.24** (1.01) (1.01) (1.03) (1.03) _ _ –3.01*** –2.93*** –2.04*** –2.04*** (.35) (.36) (.37) (.37) _ _ .33** .26* –.32** –.32** (.15) (.15) (.16) (.16) _ _ –.15 –.21 –1.04*** –1.04*** (.25) (.27) (.29) (.30) _ _ _ –.04 –.35 –.34 (.25) (.27) (.27) _ _ _ 1.04*** .65*** .64*** (.17) (.18) (.18) _ _ _ –.13 .17 .17 (.27) (.29) (.29) _ _ _ _ 5.26*** 5.28*** (.43) (.43) _ _ _ _ _ .04 (.17) –3.96*** –4.52*** –3.75*** –3.77*** –6.52*** –6.55*** (.22) (.36) (.39) (.46) (.56) (.58) .08 .10 .22 .23 .32 .32 2614

2465

2454

2451

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (1995).

Table A.10 Reported vote for the right (Chirac), French presidential election, 1995

Age Gender Education

(1)

(2)

(3)

(4)

(5)

(6)

–.14 (.21) –.17** (.09) .10 (.14)

–.49* (.27) –.27** (.11) –.32* (.19)

–.36 (.29) –.31*** (.12) –.06 (.21)

–.33 (.31) –.27** (.12) –.28 (.22)

–.61* (.34) –.25* (.13) –.19 (.24)

1.08*** (.34) –.60*** (.14) .29 (.24) Continued

9780230_300088_10_app.indd 190

6/28/2011 8:01:12 PM

Appendix 191

Table A.10 Continued (1)

(2)

(3)

(4)

Managers

1.03*** (.16) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets Extreme-left

_

_

.98*** (.19) .44*** (.15) .02 (.15) –.20 (.17) 1.24*** (.32) .39*** (.12) –.25 (.21) .30 (.20) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

.23 (.20) .28* (.16) –.09 (.16) .13 (.19) 1.35*** (.34) .20 (.13) –.42* (.23) –.07 (.22) –2.35*** (.52) –2.16*** (.22) .93*** (.12) .59*** (.21) _

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

.22 (.21) .33** (.17) .01 (.16) .22 (.19) 1.32*** (.36) .28** (.13) –.37 (.24) –.16 (.23) –2.35*** (.52) –2.23*** (.23) .99*** (.13) .64*** (.22) –.44** (.21) .00 (.15) –.92*** (.23) _

Last-minute decision Constant

_

_

_

_

–1.75*** (.15) .02

–1.32*** (.28) .04

–1.78*** (.32) .19

–1.20*** (.37) .20

3249

2636

2614

2465

Religion

Pseudo-R2

_

N

(5)

(6)

.13 1.09*** (.23) (.22) .30 .55*** (.19) (.19) .05 .21 (.18) (.18) .35 .53*** (.21) (.20) 1.48*** 1.79*** (.39) (.40) .15 .48*** (.15) (.14) –.30 –.22 (.27) (.26) .09 .10 (.25) (.25) –1.38** –2.02*** (.64) (.46) –1.43*** –2.30*** (.24) (.20) .37** .96*** (.15) (.14) –.28 –.00 (.25) (.24) –.70*** –.47** (.23) (.23) –.27 .89*** (.16) (.17) –.62** –.42* (.25) (.24) 5.85*** 3.54*** (.38) (.29) _ .16 (.15) –5.01*** –4.58*** (.49) (.47) .32 .45 2454

2451

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (1995).

9780230_300088_10_app.indd 191

6/28/2011 8:01:12 PM

192

French Presidential Elections

Table A.11 Reported vote for the extreme right (Le Pen), French presidential election, 1995 (1)

(2)

Age

–1.98*** –2.15*** (.28) (.34) Gender .55*** .72*** (.11) (.13) Education –1.43*** –1.03*** (.19) (.24) Religion .11 .10 (.18) (.20) Managers _ –.30 (.20) Employees _ .10 (.18) Workers _ –.26 (.19) Income _ –1.64*** (.40) Private-sector _ –.00 (.14) Low-risk assets _ –.06 (.26) High-risk assets _ .02 (.27) Extreme-left _ _ Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

Economy

_

_

Europe

_

_

(3)

(4)

(5)

–1.99*** –2.37*** –1.89*** (.35) (.38) (.46) .83*** .91*** .82*** (.14) (.15) (.19) –.78*** –.06 .41 (.25) (.28) (.34) –.32 –.34 –.66** (.22) (.25) (.30) –.52** –.70*** –.37 (.21) (.23) (.27) .09 –.10 .08 (.19) (.20) (.25) –.09 –.47** –.55** (.20) (.22) (.26) –1.81*** –1.63*** –1.89*** (.41) (.46) (.54) –.09 –.05 –.13 (.14) (.16) (.19) –.03 –.22 –.03 (.27) (.29) (.34) –.09 .04 –.08 (.28) (.31) (.37) –1.26*** –1.29*** –1.09** (.33) (.36) (.49) –1.31*** –.97*** –.63*** (.19) (.20) (.24) .13 –.00 –.39* (.15) (.17) (.20) 1.30*** .96*** –.28 (.22) (.24) (.30) _ 3.84*** 1.65*** (.38) (.41) _ –.77*** –.77*** (.20) (.23) _ 1.09*** .40 (.25) (.31)

(6) –1.95*** (.47) .81 (.19) .44 (.34) –.66** (.30) –.37 (.27) .10 (.25) –.55** (.26) –1.92*** (.55) –.11 (.19) –.01 (.34) –.12 (.37) –1.18** (.49) –.68*** (.24) –.44** (.20) –.37 (.31) 1.66*** (.41) –.77*** (.23) .39 (.31) Continued

9780230_300088_10_app.indd 192

6/28/2011 8:01:13 PM

Appendix 193

Table A.11 Continued (1)

(2)

(3)

(4)

(5)

Thermometer

_

_

_

_

Last-minute decision Constant

_

_

_

_

5.58*** (.33) _

–.97*** (.18)

–.19 (.34)

.19 (.36)

–3.15** (.52)

Pseudo-R 2 N

(6)

5.57*** (.33) –.29 (.20) –3.33*** –2.17*** (.59) (.60)

.04

.06

.13

.24

.46

.46

3249

2636

2614

2465

2450

2447

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (1995).

Table A.12 Reported vote for the right (Chirac), second round of the French presidential election, 1995 (1)

(2)

(3)

(4)

(5)

(6)

Managers

.11 (.19) –.12* (.07) .03 (.12) 1.96*** (.13) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

High-risk assets

_

–.33 (.24) –.15 (.09) –.55*** (.17) 1.78*** (.16) .46**** (.14) .01 (.13) –.55*** (.14) .58** (.29) .56*** (.10) .02 (.19) 1.00*** (.19)

.22 (.36) –.28* (.15) –.25 (.26) 1.07*** (.24) .24 (.22) –.13 (.20) –.12 (.22) 1.20*** (.46) .55*** (.15) –.18 (.29) .63** (.30)

.03 (.39) –.34** (.16) –.11 (.28) .93*** (.25) .26 (.23) –.06 (.21) –.17 (.23) 1.06** (.48) .60*** (.16) –.29 (.31) .65** (.32)

.07 (.45) –.41** (.18) –.06 (.32) .73** (.29) .31 (.26) .04 (.23) .09 (.26) 1.05** (.54) .52*** (.18) –.13 (.35) .71** (.36)

.03 (.46) –.39** (.18) .05 (.33) .78*** (.29) .28 (.27) .03 (.24) .08 (.26) 1.11** (.55) .55*** (.18) –.22 (.36) .78** (.37)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 193

6/28/2011 8:01:13 PM

194

French Presidential Elections

Table A.12 Continued (1)

(2)

(3)

(4)

(5)

(6)

Extreme-left

_

_

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

–3.31*** (.43) –2.98*** (.17) 2.55*** (.18) 2.27*** (.36) _

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

–3.13*** (.43) –2.85*** (.18) 2.54*** (.19) 2.04*** (.37) 1.00*** (.25) .92*** (.21) .04 (.28) _

Last-minute decision Constant

_

_

_

_

–2.52*** (.55) –2.35*** (.20) 2.07*** (.21) 1.36*** (.44) .93*** (.28) .70*** (.24) .41 (.33) 6.07*** (.41) _

–1.17*** (.37)

–2.03*** (.46)

–5.76*** (.59)

–2.54*** (.55) –2.39*** (.20) 2.06*** (.22) 1.52*** (.48) .91*** (.29) .66*** (.25) .48 (.33) 6.05*** (.42) .00 (.19) –5.80*** (.62)

Pseudo-R 2 N

–1.05*** –1.43*** (.13) (.25) .06

.11

.56

.57

.67

.67

3210

2589

2578

2432

2423

2355

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (1995).

Table A.13 Reported vote for the extreme left (Laguiller, Hue, or Besancenot), French presidential election, 2002 (1) Age Gender Education Religion

(2)

–.76** –.64 (.36) (.43) –.24* –.07 (.14) (.16) –.69*** –.30 (.23) (.28) –1.05*** –.96*** (.21) (.22)

(3)

(4)

(5)

(6)

–1.05** (.49) –.06 (.17) –.40 (.30) –.77*** (.24)

–.99* (.52) .00 (.18) –.43 (.33) –.66*** (.25)

.01 (.59) .12 (.20) –.75** (.37) –.45 (.28)

–.06 (.61) .09 (.20) –.76** (.37) –.48* (.28) Continued

9780230_300088_10_app.indd 194

6/28/2011 8:01:13 PM

Appendix 195

Table A.13 Continued (1)

(2)

Managers

_

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets Extreme-left

_

_

.00 (.22) .37* (.20) .47*** (.18) –.51 (.44) –.47*** (.16) .41 (.30) –1.19*** (.41) _

Left

_

_

Right

_

_

Extremeright Immigration

_

_

_

_

_

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

–.90*** (.26) –.58*** (.20) 1.15*** (.28) _

Last-minute decision Constant

_

_

_

_

–1.02*** (.30) –.54** (.22) 1.21*** (.32) 5.20*** (.47) _

Pseudo-R 2

–.38* (.22) .03

–.48 (.35) .06

–.62 (.40) .09

–.49 (.46) .13

–3.16*** (.58) .27

–1.03*** (.30) –.57** (.22) 1.22*** (.32) 5.20*** (.47) –.16 (.24) –3.02*** (.59) .28

N

1557

1401

1239

1210

1165

1153

_

(3)

(4)

(5)

(6)

.04 .04 .11 .11 (.23) (.24) (.26) (.27) .38* .47** .41* .40* (.21) (.22) (.24) (.24) .50*** .48** .33 .31 (.19) (.20) (.22) (.22) –.76 –.59 –.53 –.60 (.48) (.50) (.55) (.56) –.39** –.32* –.30 –.27 (.17) (.18) (.19) (.20) .63* .70** .75** .78** (.32) (.34) (.37) (.38) –1.14*** –.95** –.76 –.76 (.43) (.44) (.48) (.48) .99*** 1.02*** .82*** .82*** (.26) (.27) (.31) (.31) .42** .45** .35 .38 (.21) (.22) (.25) (.25) –.37 –.30 –.23 –.22 (.28) (.30) (.32) (.32) +Dropped +Dropped +Dropped +Dropped

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2002).

9780230_300088_10_app.indd 195

6/28/2011 8:01:13 PM

196

French Presidential Elections

Table A.14 Reported vote for the left (Jospin), French presidential election, 2002 (1)

(2)

(3)

Managers

1.08*** (.33) .16 (.13) .35* (.21) –.96*** (.20) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

High-risk assets

_

Extreme-left

_

1.51*** (.39) –.02 (.15) .41 (.25) –.86*** (.21) –.24 (.19) –.19 (.20) .18 (.17) .93** (.43) –.27* (.15) –.22 (.27) –.76** (.34) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

2.02*** (.45) –.11 (.16) .35 (.29) –.35 (.23) –.32 (.20) –.14 (.22) .15 (.19) 1.02** (.47) –.17 (.16) –.42 (.30) –.58 (.37) .74*** (.26) .99*** (.20) –1.41*** (.34) –1.78** (.75) _

Economy

_

_

_

Europe

_

_

_

Age Gender Education Religion

(4)

(5)

(6)

1.98*** 1.52*** 1.36** (.48) (.56) (.58) –.11 –.13 –.16 (.17) (.19) (.20) –.08 –.30 .18 (.31) (.27) (.37) –.24 –.30 –.26 (.24) (.27) (.28) –.28 –.18 –.18 (.21) (.24) (.25) –.11 –.16 –.16 (.23) (.26) (.26) .16 –.12 –.18 (.20) (.22) (.23) .90* .98* 1.00* (.48) (.57) (.58) –.22 –.04 –.05 (.16) (.19) (.19) –.39 –.51 –.55 (.31) (.37) (.38) –.60 –.49 –.43 (.38) (.43) (.44) –.58** –.01 –.04 (.27) (.31) (.32) .82*** .18 .14 (.21) (.25) (.25) –1.48*** –1.28*** –1.30*** (.34) (.38) (.38) –1.63** –1.40*** –1.45 (.75) (.89) (.89) –.66*** –.12 –.14 (.25) (.29) (.29) –.77*** –.45** –.40* (.19) (.22) (.22) –1.58*** –1.04** –1.02** (.34) (.41) (.41) Continued

9780230_300088_10_app.indd 196

6/28/2011 8:01:14 PM

Appendix 197

Table A.14 Continued

Thermometer Last-minute decision Constant Pseudo-R 2 N

(1)

(2)

(3)

(4)

(5)

_

_

_

_

6.52*** (.52) _

1557

1401

1299

1269

1268

(6)

6.52*** (.53) _ _ _ _ –.29 (.23) –1.60*** –1.83*** –2.55*** –1.24*** –5.93*** –5.74*** (.21) (.34) (.41) (.46) (.66) (.68) .02 .04 .14 .17 .36 .36 1256

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (2002).

Table A.15 Reported vote for the center (Bayrou), French presidential election, 2002 (1)

(2)

(3)

(4)

(5)

(6)

Managers

.41 (.50) –.39* (.20) 1.24*** (.33) 1.79*** (.37) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

–.08 (.62) –.52** (.24) .83** (.40) 1.50*** (.40) .54** (.26) –.27 (.31) –.30 (.32) –.08 (.64) .48** (.24) .34 (.45)

.25 (.65) –.58** (.25) .70* (.42) 1.30*** (.42) .56** (.27) –.38 (.34) –.38 (.35) .18 (.67) .40 (.25) .12 (.48)

.17 (.67) –.63** (.25) .37 (.44) 1.36*** (.42) .55** (.27) –.32 (.35) –.40 (.37) –.06 (.68) .42 (.26) .02 (.49)

–.82 (.79) –.51* (.29) .20 (.50) .95** (.46) .52* (.32) –.12 (.40) –.47 (.42) –.34 (.76) .37 (.29) –.17 (.57)

.42 (.87) –.42 (.31) .18 (.53) 1.10** (.48) .32 (.33) –.14 (.42) –.30 (.44) –.34 (.81) .34 (.30) –.17 (.60)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 197

6/28/2011 8:01:14 PM

198

French Presidential Elections

Table A.15 Continued

High-risk assets Extreme-left Left Right Extreme-right Immigration Economy Europe Thermometer Last-minute decision Constant Pseudo-R 2 N

(1)

(2)

(3)

(4)

(5)

(6)

_

.41 (.44) _

.34 (.46) –2.24** (1.03) –.09 (.29) –.18 (.31) –.01 (.53) _

.38 (.47) –2.25** (1.04) –.08 (.31) –.14 (.32) .19 (.55) –.49 (.38) .47 (.29) –1.32** (.59) _

.49 (.52) –2.11* (1.11) .26 (.35) –.13 (.37) .13 (.62) –.78* (.46) .23 (.33) –.65 (.64) 8.24*** (.82) _

.06

.08

.10

.13

.37

.43

1557

1401

1299

1269

1258

1247

.93* (.56) _ –1.83* (1.11) _ _ .16 (.37) _ _ –.14 (.38) _ _ .33 (.64) _ _ –.82* (.48) _ _ _ .23 (.34) _ _ _ –.77 (.69) _ _ _ 8.67*** (.88) _ _ _ _ 2.12*** (.37) –4.29*** –4.28*** –3.95*** –3.46*** –7.64*** –9.52*** (.38) (.57) (.62) (.70) (.96) (1.13)

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (2002).

Table A.16 Reported vote for the right (Chirac), French presidential election, 2002

Age Gender Education

(1)

(2)

(3)

(4)

(5)

(6)

1.46*** (.34) –.25* (.14) –.35 (.22)

1.26*** (.40) –.29* (.16) –.39 (.26)

1.58*** (.46) –.30* (.17) –.26 (.28)

1.45*** (.47) –.39** (.18) –.22 (.30)

1.32** (.55) –.12 (.21) .17 (.34)

1.10** (.56) –.16 (.21) .18 (.35) Continued

9780230_300088_10_app.indd 198

6/28/2011 8:01:14 PM

Appendix 199

Table A.16 Continued (1)

(2)

(3)

(4)

(5)

Managers

1.29*** (.24) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

High-risk assets

_

Extreme-left

_

1.26*** (.25) –.05 (.19) –.20 (.20) –.25 (.19) –.40 (.42) .24 (.16) –.40 (.29) .40 (.32) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

.55* (.28) –.11 (.21) –.22 (.23) –.26 (.21) –.29 (.46) .15 (.18) –.52 (.33) .03 (.36) –2.03*** (.48) –1.54*** (.25) .61*** (.20) .90*** (.31) _

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

.45 (.29) –.08 (.21) –.24 (.24) –.22 (.22) –.15 (.48) .08 (.18) –.67** (.33) .10 (.36) –1.97*** (.49) –1.54*** (.25) .55*** (.20) .90*** (.32) .66** (.27) –.24 (.20) –.80** (.32) _

Last-minute decision Constant

_

_

_

.08 (.33) –.27 (.25) –.23 (.27) –.33 (.24) .19 (.53) .04 (.21) –.57 (.39) .12 (.42) –.73 (.54) –.75*** (.28) .28 (.23) .54 (.40) .10 (.31) –.35 (.23) .14 (.37) 6.21*** (.53) _

Religion

Pseudo-R2 N

–2.44*** –2.00*** –1.41*** (.24) (.35) (.40)

(6)

.14 (.34) –.19 (.25) –.19 (.27) –.34 (.25) .29 (.54) .07 (.21) –.47 (.39) –.04 (.43) –.82 (.55) –.77*** (.29) .27 (.23) .49 (.40) .12 (.31) –.33 (.23) .15 (.38) 6.09*** (.53) _ –.60** (.24) –1.37*** –5.68*** –5.46*** (.46) (.67) (.68)

.06

.06

.18

.19

.37

.38

1557

1401

1299

1269

1268

1256

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (2002).

9780230_300088_10_app.indd 199

6/28/2011 8:01:14 PM

200 French Presidential Elections

Table A.17 Reported vote for the extreme right (Le Pen), French presidential election, 2002 (1)

(2)

(3)

(4)

(5)

(6)

Managers

–.49 (.44) .19 (.17) –1.78*** (.29) .28 (.28) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets Extreme-left

_

_

–.43 (.52) .29 (.20) –1.67*** (.33) .24 (.29) .14 (.25) .18 (.25) .10 (.22) –.00 (.53) .24 (.20) .19 (.36) –.09 (.43) _

Left

_

_

Right

_

_

Extremeright Immigration

_

_

_

_

–.48 (.57) .30 (.21) –1.58*** (.35) .11 (.32) .13 (.27) .18 (.26) .15 (.23) –.16 (.55) .21 (.22) .07 (.39) .02 (.45) –.16 (.38) –.34 (.28) .44 (.27) 1.18*** (.37) _

Economy

_

_

_

Europe

_

_

_

–1.50** (.65) .42 (.26) –.67 (.42) –.00 (.37) .19 (.31) .13 (.31) –.03 (.26) .33 (.66) .22 (.25) .16 (.44) –.30 (.53) .09 (.45) –.00 (.31) .27 (.31) .63 (.42) 5.14*** (.56) 1.08*** (.28) 1.73*** (.33)

–.16 (.81) .43 (.32) –.51 (.50) .14 (.45) .47 (.37) .05 (.39) .04 (.33) –.16 (.82) .20 (.31) .45 (.53) –.18 (.63) –.15 (.55) .22 (.37) –.02 (.37) .08 (.53) 3.23*** (.63) .80** (.33) .77* (.41)

–.32 (.82) .43 (.32) –.48 (.50) .18 (.46) .50 (.37) .05 (.39) .05 (.33) –.16 (.83) .24 (.31) .45 (.53) –.24 (.64) .15 (.55) .21 (.37) –.02 (.37) .04 (.53) 3.22*** (.63) .83** (.33) .77* (.42)

Age Gender Education Religion

_

Continued

9780230_300088_10_app.indd 200

6/28/2011 8:01:15 PM

Appendix 201

Table A.17 Continued (1)

(2)

(3)

(4)

(5)

(6)

Thermometer

_

_

_

_

Last-minute decision Constant

_

_

_

_

4.88*** (.46) _

–1.35*** (.27)

–1.75*** (.44)

–1.69*** (.49)

–6.89*** (.76)

–7.07*** (.89)

4.86*** (.46) –.32 (.35) –6.96*** (.89)

Pseudo-R 2 N

.04

.05

.07

.33

.51

.51

1557

1401

1299

1269

1266

1254

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (2002).

Table A.18 Voting intention for the right (Chirac), second round of the French presidential election, 2002 (1)

(2)

(3)

(4)

(5)

(6)

Managers

.12 (.18) –.12 (.08) –.18 (.12) 1.66*** (.13) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

–.25 (.23) –.10 (.09) –.45*** (.16) 1.59*** (.14) .25** (.11) .09 (.12) –.28*** (.11) –.14 (.25) .37*** (.09) –.29* (.17)

–.07 (.30) –.13 (.12) –.16 (.21) 1.07*** (.18) .28* (.15) .04 (.16) –.11 (.14) –.13 (.32) .22* (.12) –.35 (.22)

–.97** (.47) –.38** (.18) –.00 (.31) .98*** (.27) .35 (.22) –.04 (.22) –.36* (.21) –.11 (.48) .01 (.18) –.37 (.33)

–1.12** (.55) –.09 (.21) .22 (.36) .64** (.30) .07 (.26) –.14 (.26) –.34 (.24) .33 (.56) .04 (.20) –.32 (.39)

–1.12** (.56) –.10 (.21) .22 (.36) .65** (.30) .08 (.26) –.11 (.26) –.30 (.24) .27 (.56) .01 (.20) –.34 (.39)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 201

6/28/2011 8:01:15 PM

202 French Presidential Elections

Table A.18 Continued (1)

(2)

High-risk assets

_

Extreme-left

_

1.12*** (.21) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

Economy

_

_

Europe

_

_

Thermometer

_

_

Last-minute decision Constant

_

_

–.89*** (.13)

–.81*** (.20)

Pseudo-R 2 N

(3)

(4)

(5)

(6)

1.21*** .92** 1.32*** 1.31*** (.27) (.39) (.47) (.47) –2.51*** –2.52*** –1.61*** –1.63*** (.25) (.36) (.42) (.42) –1.99*** –1.96*** –1.36*** –1.37*** (.14) (.20) (.23) (.23) 1.14*** 1.07*** 1.10*** 1.11*** (.15) (.22) (.26) (.26) 1.23*** 1.52*** 1.13* 1.11* (.31) (.51) (.60) (.60) _ 1.73*** 1.20*** 1.22*** (.26) (.30) (.31) _ .65*** .82*** .81*** (.19) (.23) (.23) _ –.40 .44 .38 (.32) (.38) (.38) _ _ 5.46*** 5.45*** (.43) (.43) _ _ _ –.04 (.24) –.10 –.70 –4.31*** –4.23*** (.27) (.45) (.59) (.60)

.05

.08

.34

.37

.52

.52

2944

2577

2335

1152

1152

1142

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. Source: CEVIPOF (2002).

Table A.19 Reported vote for the extreme left (Laguiller, Buffet, or Besancenot), French presidential election, 2007

Age Gender Education

(1)

(2)

(3)

(4)

(5)

(6)

–.58* (.33) –.11 (.12) –.89*** (.22)

–.22 (.41) –.19 (.15) –.28 (.28)

–.63 (.44) –.30* (.16) –.61** (.29)

–.40 (.45) –.27* (.16) –.59** (.30)

–.58 (.50) –.24 (.18) –.66** (.32)

–.38 (.51) –.18 (.18) –.75** (.32) Continued

9780230_300088_10_app.indd 202

6/28/2011 8:01:16 PM

Appendix 203

Table A.19 Continued (1)

(2)

Religion

–1.47*** –1.46*** (.18) (.20) Managers _ –.37 (.25) Employees _ .03 (.20) Workers _ .43** (.21) Income _ –.92** (.38) Private-sector _ –.32** (.14) Low-risk _ –.16 assets (.27) High-risk _ –.21 assets (.20) Extreme-left _ _ Left

_

_

Right

_

_

Extreme-right Immigration

_ _

_ _

Economy

_

_

Europe

_

_

Thermometer

_

_

Last-minute decision Constant

_

_

–1.09*** –.87** (.19) (.34)

Pseudo-R 2 N

(3)

(4)

(5)

(6)

–.91*** –.85*** –.63*** –.58** (.22) (.22) (.24) (.24) –.23 .01 –.10 –.17 (.27) (.22) (.23) (.29) .07 .36 .36 –.11 (.22) (.22) (.24) (.23) .42* –.69* –.55 .37 (.22) (.41) (.44) (.24) –.78* –.20 –.16 –.50 (.40) (.27) (.29) (.44) –.25* –.27* –.15 –.13 (.14) (.15) (.16) (.16) .01 .10 .16 .21 (.28) (.29) (.31) (.31) .03 .10 –.06 –.06 (.21) (.21) (.23) (.23) 2.26*** 2.05*** 1.00*** 1.24*** (.23) (.24) (.27) (.27) .34** .35** –.15 .01 (.16) (.16) (.18) (.18) –1.66*** –1.57*** –.93*** –.77** (.30) (.31) (.32) (.32) +Dropped +Dropped +Dropped +Dropped _ –.57** –.26 –.35 (.22) (.24) (.25) _ .11 .12 .08 (.16) (.17) (.17) _ 1.42*** 1.21*** 1.30*** (.22) (.23) (.24) _ _ 5.55*** 5.63*** (.43) (.44) _ _ _ .82*** (.19) –1.05*** –1.50*** –5.11*** –5.72*** (.36) (.39) (.52) (.55)

.04

.06

.16

.18

.30

.31

3534

2924

2851

2833

2818

2818

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2007).

9780230_300088_10_app.indd 203

6/28/2011 8:01:16 PM

204

French Presidential Elections

Table A.20 Reported vote for the left (Royal), French presidential election, 2007 (1)

(2)

Age

.43** 1.03*** (.21) (.27) Gender –.13* –.02 (.08) (.10) Education .15 .44** (.14) (.18) Religion –1.06*** –1.03*** (.12) (.13) Managers _ –.53*** (.14) Employees _ –.21 (.13) Workers _ –.17 (.13) Income _ –.09 (.24) Private-sector _ –.21** (.09) Low-risk _ –.18 assets (.17) High-risk _ –.35*** assets (.12) Extreme-left _ _

(3)

(4)

(5)

(6)

–.02 .02 .22 .06 (.35) (.35) (.41) (.41) –.22* –.27** –.11 –.15 (.12) (.13) (.14) (.14) –.18 –.31 –.12 –.10 (.22) (.23) (.26) (.26) –.21 –.16 –.23 –.24 (.17) (.18) (.20) (.20) –.18 –.17 –.16 –.16 (.18) (.18) (.20) (.21) –.33** –.25 –.38** –.39** (.16) (.17) (.19) (.19) –.24 –.14 –.47** –.49** (.17) (.17) (.19) (.20) .22 .08 .20 .17 (.31) (.31) (.35) (.35) .01 –.01 .13 .13 (.11) (.12) (.13) (.13) –.03 –.06 –.08 –.09 (.22) (.22) (.25) (.25) –.24 –.27* –.28 –.28 (.15) (.16) (.17) (.17) 1.10*** 1.10*** .76*** .66** (.22) (.22) (.26) (.26) 2.53*** 2.51*** 1.83*** 1.77*** (.12) (.12) (.13) (.14) –2.88*** –2.84*** –2.05*** –2.15*** (.35) (.35) (.36) (.36) +Dropped +Dropped +Dropped +Dropped

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

_

Economy

_

_

_

Europe

_

_

_

–.48*** (.18) –.27** (.12) –.56*** (.18)

–.49** (.20) –.22 (.14) –.60*** (.21)

–.45** (.20) –.22 (.14) –.63*** (.21) Continued

9780230_300088_10_app.indd 204

6/28/2011 8:01:16 PM

Appendix 205

Table A.20 Continued (1)

(2)

(3)

(4)

(5)

(6)

Thermometer

_

_

_

_

Last-minute decision Constant

_

_

_

_

7.14*** (.42) _

–.62*** (.13) .02

–.54** (.23) .04

–1.19*** (.29) .36

–.63** (.31) .37

–5.59*** (.46) .50

7.05*** (.42) –.45*** (.16) –5.21*** (.48) .50

3534

2924

2851

2833

2832

2832

Pseudo-R 2 N

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2007).

Table A.21 Reported vote for the center (Bayrou), French presidential election, 2007 (1) Age

(2)

–.63*** –.64** (.24) (.30) Gender –.05 –.06 (.08) (.10) Education .78*** .64*** (.15) (.19) Religion –.00 –.02 (.13) (.15) Managers _ –.20 (.14) Employees _ –.23* (.14) Workers _ –.30** (.15) Income _ –.15 (.25) Private-sector _ .11 (.10) Low-risk _ .10 assets (.19) High-risk _ –.01 assets (.12)

(3)

(4)

(5)

(6)

.19 (.33) .14 (.12) .96*** (.22) –.23 (.16) –.15 (.16) –.22 (.15) –.49*** (.16) .02 (.28) .16 (.11) .21 (.21) .13 (.14)

.16 (.34) .06 (.12) .82*** (.22) –.19 (.17) –.15 (.16) –.14 (.15) –.38** (.17) –.04 (.28) .15 (.11) .19 (.21) .12 (.14)

.31 (.41) .04 (.14) .66** (.26) –.38* (.19) .01 (.19) –.11 (.18) –.34* (.19) .33 (.34) .18 (.13) .06 (.25) .11 (.16)

.70* (.42) .14 (.14) .67** (.27) –.33* (.20) .03 (.19) –.06 (.18) –.32 (.20) .33 (.35) .20 (.14) .09 (.25) .15 (.16) Continued

9780230_300088_10_app.indd 205

6/28/2011 8:01:16 PM

206 French Presidential Elections

Table A.21 Continued (1)

(2)

Extreme-left

_

_

Left

_

_

Right

_

_

Extreme-right Immigration

_ _

_ _

Economy

_

_

Europe

_

_

Thermometer

_

_

Last-minute decision Constant

_

_

Pseudo-R 2 N

–1.48*** –1.26*** (.15) (.25) .02 .02 3534

2924

(3)

(4)

(5)

(6)

–4.60*** –4.61*** –3.48*** (1.01) (1.01) (1.02) –2.23*** –2.25*** –2.05*** (.15) (.15) (.16) –2.06*** –2.00*** –1.28*** (.14) (.14) (.16) +Dropped +Dropped +Dropped _ –.39** –.23 (.17) (.19) _ .01 –.02 (.12) (.13) _ –.54*** –.32 (.18) (.21) _ _ 7.78*** (.41) _ _ _ –.71*** (.27) .18

–.34 (.30) .19

2851

2833

–3.29*** (1.02) –1.90*** (.16) –1.10*** (.16) +Dropped –.26 (.20) –.04 (.14) –.27 (.21) 7.92*** (.42) 1.17*** (.16) –5.90*** –6.88*** (.47) (.51) .38 .40 2832

2832

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2007).

Table A.22 Reported vote for the right (Sarkozy), French presidential election, 2007

Age Gender Education Religion

(1)

(2)

(3)

(4)

(5)

(6)

.81*** (.21) –.06 (.08) .00 (.14) 1.55*** (.13)

.32 (.27) –.21** (.10) –.62*** (.18) 1.58*** (.15)

1.08*** (.35) –.30** (.13) –.30 (.24) .92*** (.19)

.77** (.36) –.32** (.13) –.23 (.24) .86*** (.19)

.23 (.40) –.41*** (.14) .04 (.27) .48** (.20)

.02 (.41) –.52*** (.15) .04 (.27) .46** (.22) Continued

9780230_300088_10_app.indd 206

6/28/2011 8:01:16 PM

Appendix 207

Table A.22 Continued (1)

(2)

Managers

_

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets High-risk assets Extreme-left

_

_

.69*** (.13) .33** (.13) .17 (.14) .97*** (.23) .34*** (.10) –.07 (.17) .45*** (.11) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

–2.44*** (.27) 2.61*** (.12) 1.02*** (.28) _

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

–2.43*** (.27) 2.56*** (.12) .99*** (.29) .65*** (.18) –.13 (.12) –.91*** (.19) _

Last-minute decision Constant

_

_

_

_

–1.60*** (.29) 1.70*** (.14) .27 (.33) .03 (.20) –.16 (.14) –.71*** (.21) 7.23*** (.46) _

–2.45*** (.34) .41

–6.92*** (.50) .52

–1.71*** (.29) 1.54*** (.14) –.07 (.33) –.01 (.20) –.15 (.14) –.69*** (.22) 7.11*** (.46) –1.02*** (.17) –6.21*** (.51) .53

2783

2782

2782

Pseudo-R 2

_

(3)

3534

2924

(5)

(6)

.36** .33* .48** .46** (.18) (.18) (.20) (.20) .50*** .51*** .36* .34* (.17) (.17) (.19) (.19) .45** .47** .30 .31 (.18) (.19) (.20) (.21) .87*** .75** .85** .86** (.30) (.31) (.34) (.34) .17 .19 .22 .22 (.12) (.13) (.14) (.14) –.52** –.55** –.60** –.58** (.22) (.23) (.25) (.25) .24* .21 .11 .09 (.14) (.14) (.16) (.16) +Dropped +Dropped +Dropped +Dropped

–1.86*** –2.32*** –2.63*** (.14) (.24) (.31) .06 .09 .40

N

(4)

2801

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2007).

9780230_300088_10_app.indd 207

6/28/2011 8:01:17 PM

208 French Presidential Elections

Table A.23 Reported vote for the extreme right (Le Pen), French presidential election, 2007 (1)

(2)

(3)

(4)

(5)

(6)

Age

–1.76*** –2.40*** –2.09*** –2.26*** –1.81** –1.87*** (.40) (.51) (.56) (.58) (.73) (.73) Gender .67*** .95*** .84*** 1.10*** 1.19*** 1.15*** (.15) (.20) (.21) (.23) (.28) (.28) Education –2.06*** –1.65*** –1.58*** –1.04** –.85 –.84 (.30) (.35) (.39) (.43) (.53) (.53) Religion .61** .65** .27 .20 .42 .43 (.24) (.27) (.30) (.31) (.39) (.40) Managers _ –.05 –.07 –.23 –.54 –.53 (.31) (.33) (.35) (.42) (.43) Employees _ .49* .37 .30 .01 .02 (.27) (.29) (.31) (.38) (.38) Workers _ .31 .37 .07 –.27 –.23 (.27) (.29) (.30) (.37) (.37) Income _ –.85* –1.04** –.82 –.19 –.17 (.46) (.49) (.52) (.66) (.66) Private-sector _ –.21 –.38** –.29 –.11 –.12 (.17) (.19) (.20) (.24) (.24) Low-risk assets _ .44 .29 .36 .72 .75 (.33) (.35) (.37) (.46) (.46) High-risk _ –.15 –.42* –.39 –.50 –.51 assets (.23) (.25) (.27) (.33) (.33) Extreme-left _ _ +Dropped +Dropped +Dropped +Dropped Left _ _ –1.35*** –1.23*** –.87** –.90** (.32) (.33) (.44) (.44) Right _ _ .36* .17 –.13 –.24 (.20) (.21) (.27) (.28) Extreme-right _ _ 3.23*** 2.24*** 1.02** .87** (.31) (.34) (.43) (.44) Immigration _ _ _ 2.53*** .82** .80* (.34) (.41) (.41) Economy _ _ _ .37* .14 .15 (.21) (.25) (.26) Europe _ _ _ 1.54*** .89*** .90*** (.27) (.33) (.33) Continued

9780230_300088_10_app.indd 208

6/28/2011 8:01:17 PM

Appendix 209

Table A.23 Continued (1)

(2)

(3)

(4)

(5)

(6)

Thermometer

_

_

_

_

Last-minute decision Constant

_

_

_

_

7.44*** (.54) _

–4.38*** (.57)

–7.01*** (.76)

7.40*** (.54) –.50 (.31) –6.76*** (.78)

–2.08*** –2.19*** –1.85*** (.24) (.44) (.47)

Pseudo-R 2 N

.05

.07

.19

.28

.55

.55

3534

2924

2801

2783

2782

2782

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2007).

Table A.24 Reported vote for the right (Sarkozy), second round of the French presidential election, 2007 (1)

(2)

(3)

(4)

(5)

(6)

Managers

.40** (.20) .07 (.07) –.31** (.13) 1.72*** (.12) _

Employees

_

Workers

_

Income

_

Private-sector

_

Low-risk assets

_

–.41 (.26) –.04 (.09) –.86*** (.17) 1.75*** (.13) .52*** (.13) .13 (.12) .06 (.13) .60*** (.22) .29*** (.09) .28* (.16)

.94** (.40) .01 (.14) –.67** (.26) 1.27*** (.20) –.02 (.21) .12 (.19) .15 (.19) .24 (.35) .04 (.13) .17 (.25)

.62 (.41) .10 (.15) –.38 (.28) 1.19*** (.21) –.07 (.22) .07 (.19) .03 (.20) .24 (.35) .09 (.14) .14 (.26)

–.79 (.54) .00 (.18) –.28 (.35) .60** (.26) .17 (.27) –.10 (.23) –.30 (.25) .31 (.45) .13 (.17) .07 (.33)

–.78 (.55) .03 (.19) –.31 (.35) .63** (.26) .17 (.27) –.10 (.24) –.34 (.25) .26 (.45) .12 (.17) .10 (.33)

Age Gender Education Religion

Continued

9780230_300088_10_app.indd 209

6/28/2011 8:01:17 PM

210

French Presidential Elections

Table A.24 Continued (1)

(2)

_

(3)

(4)

(5)

(6)

High-risk assets Extreme-left

_

.28*** (.11) _

Left

_

_

Right

_

_

Extreme-right

_

_

Immigration

_

_

_

Economy

_

_

_

Europe

_

_

_

Thermometer

_

_

_

1.38*** (.20) –.01 (.14) –.48** (.21) _

Last-minute decision Constant

_

_

_

_

.76*** (.26) .01 (.18) .36 (.28) 8.80*** (.51) _

–.81*** (.13) .07

–.97*** (.22) .09

–.78** (.32) .57

–1.20*** (.36) .58

–5.76*** (.55) .73

.75*** (.26) .01 (.18) .33 (.28) 8.88*** (.51) .05 (.21) –5.83*** (.58) .73

3427

2828

2761

2744

2743

2740

Pseudo-R 2 N

–.16 –.17 –.35 –.35 (.18) (.18) (.22) (.23) –4.66*** –4.44*** –4.45*** –4.48*** (1.01) (1.01) (1.22) (1.24) –3.21*** –3.23*** –2.79*** –2.79*** (.19) (.19) (.23) (.23) 4.25*** 4.09*** 3.21*** 3.22*** (.32) (.32) (.38) (.38) +Dropped +Dropped +Dropped +Dropped

Notes: Entries are logistic regression coefficients with standard errors in parentheses. * p < .10; ** p < .05; *** p < .01, two-tailed test. + Variable dropped by STATA because of a collinearity problem. Source: CEVIPOF (2007).

Variables coding Vote: First-round vote takes the value of (1) when the respondent reported voting for the indicated candidates; (0) otherwise. Secondround vote takes the value of (1) when the respondent reported voting (or intended to vote in 2002) for the right-wing candidate and (0) for the left-wing candidate.

9780230_300088_10_app.indd 210

6/28/2011 8:01:17 PM

Appendix 211

Age: Age rescaled from (0) for youngest to (1) for oldest. Gender: (1) for men; (0) for women. Education: Eight-point scale (no diploma; elementary school diploma; professional aptitudes certificate (CAP); professional teaching certificate (BEP); technical or professional baccalauréat degree; general baccalauréat degree; baccalauréat + 2; postgraduate degree), rescaled from (0) for lowest education to (1) for highest education. Religion: (1) for practicing Catholics (going to church at least once a month); (.67) for nonpracticing Catholics (rarely or never going to church); (.33) for other religions; and (0) for no religion. Managers: (1) for farmers, merchants, craftsmen, company directors, company executives, or professionals; (0) otherwise. White-collar employees: (1) for white-collar employees; (0) otherwise. Blue-collar workers: (1) for qualified, nonqualified, or agriculture workers; (0) otherwise. Income: 11-point (Pierce 1988), 10-point (1988, 1995), 9-point (2002), and 8-point (2007) salary scales, rescaled from (0) for lowest income to (1) for highest income. Private sector: (1) for independent workers, company directors, or private-firm employees; (0) otherwise. Low-risk assets: Four-point scale from (0) to (1) for owning none, one, two, or three of the following assets: home or apartment, country house, or savings. High-risk assets: For 1988, 1995, and 2002, four-point scales from (0) to (1) for owning none, one, two, or three of the following assets: business, stocks, or rental properties; for 2007, (1) for owning stocks, (0) otherwise. Ideology variables: For 1988, 1995, and 2002, from a seven-point leftright self-positioning scale, extreme-left = (1) when response = 1; left = (1) when response = 2 or 3; center = (1) when response = 4; right = (1) when response = 5 or 6; extreme-right = (1) when response = 7; (0) otherwise. For 2007, from a six-point left-right selfassignment scale, extreme-left = (1) when respondent said “very left”; left = (1) when respondent said “left”; center = 1 when respondent said “center or “neither”; right = (1) when respondent said “right”; extremeright = (1) when respondent said “very right”. Respondents coded “center” are always included in the reference category. Immigration: For 2007, 2002, and 1995 (CEVIPOF): Opinion on the statement “There are too many immigrants in France” coded (0)

9780230_300088_10_app.indd 211

6/28/2011 8:01:17 PM

212

French Presidential Elections

(totally disagree); (.33) (somewhat disagree); (.67) (somewhat agree); and (1) (totally agree). For 1988 (Pierce), opinion on a seven-point scale rescaled from (0) to (1) between the statement “immigrants should go back in their country of origin” (0) and “immigrants living in France should be integrated into French society” (1). Economy: For Pierce (1988) and CEVIPOF (1995), response to the question “In the last twelve months, do you have the perception the country’s economic situation worsened (0), stayed the same (.5), or improved (1)? For 2002 and 2007, coded (0) if unemployment was identified as the most important problem in the country; (.5) if it was identified as the second (or third in 2002) most important problem in the country; (1) otherwise. Europe: For 2002 and 2007, opinion on France in the European Union is a good thing (0); a bad thing (1); neither a good nor a bad thing (.5). In 1995, the connotation of the word “Europe” is very positive (0); somewhat positive (.33); somewhat negative (.67); very negative (1); or no response (.5). No equivalent variable is available for 1988. Thermometer: Sympathy scales from 0 to 100 in 1988 and 1995 and from 0 to 11 in 2002 and 2007 were rescaled from (0) for maximum antipathy to (1) for maximum sympathy. In the extreme-left category, the variable is based on the score of Lajoinie in 1988, the average score of Hue and Laguiller in 1995, the average score of Laguiller and Besancenot in 2002, and the average score of Besancenot and Buffet in 2007. Last-minute decision: Coded (0) if the vote choice was made before the campaign, (.5) if the choice was made during the campaign, and (1) if it was made at the last moment. No equivalent variable is available for 1988.

9780230_300088_10_app.indd 212

6/28/2011 8:01:17 PM

Notes 2

Social Structure

1.

The Ecologists, or the Greens, as a voice for environmental issues, have had a persistent, but small, presence in French national politics. For example, in the first round of the 1995 presidential contest, Voynet, the green candidate, garnered only about 3 percent of the vote. The small numbers of green voters make analysis of their vote in individual surveys unreliable. Therefore, we pooled the four surveys into one, and examined the green vote (yes or no) in the first round. Exploring simply the four socio-demographic variables of this chapter, we find that green voters are significantly younger, more educated, and less religious than other voters. When full controls are imposed, on all long-term and short-term forces, as is done in our final models (see the Appendix), the effects from the sociodemographic variables of education and age remain significant. In addition, green voters are found to be significantly less likely to be blue-collar, right-wing, or anti-immigrant. All in all, these results conform to the known portrait of green voters. Sarkozy’s ability to appeal to those with lower qualifications may be linked to the fact that his own educational background differed from the traditional French political elite path of Sciences Po/ENA (École nationale d’administration). His pride in asserting his nontraditional trajectory may likely have appealed to voters without elite qualifications. We thank an anonymous referee for offering this interpretation.

2

3 Class and Patrimony 1.

Because the number of peasants in each survey was too small to analyze with any reliability, we pooled the four surveys into one. This allowed us to break out the peasant occupation category for separate analysis. As a rigorous test, we ran the fully specified model eventually applied to all the surveys (following Model 6 in the Appendix, column 6). The results confirm the established voting pattern. In the first round, peasants are significantly less likely to vote for the extreme candidates, either left or right. As well, they are not attracted to centrist candidates. Instead, they favor the moderate right, opposing the moderate left. This significant support for the traditional right sustains itself in the second round. Clearly, peasants have maintained their role as a “supporting class” of the bourgeoisie.

213

9780230_300088_11_not.indd 213

6/28/2011 8:00:41 PM

214

Notes

6 Leader Image 1.

2.

Methodological difficulties arise. Recent CEVIPOF surveys from 2002 and 2007 use a complex design, which includes not only a panel, but also waves of interviews. The case of the 2007 election is the clearest. This investigation encompassed a panel with four waves and a postelection survey administered to different samples of respondents. The difficulty results from the fact that the questions on specific aspects of image were asked to respondents in the panel, while the thermometer scale measuring overall candidate image was in the post-electoral wave of the survey. Unfortunately, it is therefore impossible to analyze the link between the rich indicators of image in the panel and overall evaluations of these candidates. To overcome this problem, three questions in the post-electoral survey were used as approximate measures of competence, honesty, and liking toward Royal and Sarkozy. The questions asked the respondents to say if one or the other was a good candidate, or whether the trust the candidate inspired within them, or the feeling that they were close to the respondents’ concerns, played a significant role in the decision to support them. The choice of indicators and cases in this section were therefore dictated not only by the availability of information, but also by the simultaneous presence of specific and general indicators regarding candidate image in the same wave of interviews. As a result, we hope that the next French electoral study will include a battery of questions regarding perceived traits of candidates, and as well that these questions will be administered to the same respondents. For example, in the case of Ségolène Royal, her mean image score was 57, with a standard deviation of 27. The figures in Table 6.6 therefore represent the increase in probability of voting for her, all other things being equal (that is to say, when the values of the other explanatory variables are fixed at their means) when her thermometer score goes from 43.5 to 70.5 (an average of 57, plus or minus one-half standard deviation). This example gives a good idea about the nature of these simulations for most of the candidates, since the mean standard deviation for all the thermometer scores is exactly 27. This figure varies little over the years (between 25 and 29) and across candidates (from 22 to 31).

7

Campaign Influence

1.

The use of primaries in the Socialist campaign, and the eventual adoption of this idea by the UMP, could potentially have the effect of extending, at least informally, the length of French presidential campaigns. This said, some may argue that the 1988 election was preceded by an uninterrupted two-year campaign between Jacques Chirac, who became prime minister following the 1986 legislative election, and sitting president François Mitterrand. The impact of the introduction of primaries on candidates’ campaigns and on voting behavior remains to be seen. We thank an anonymous reviewer for having raised this interesting conjecture.

9780230_300088_11_not.indd 214

6/28/2011 8:00:41 PM

Bibliography A. I. Abramowitz (1978) “The Impact of a Presidential Debate on Voter Rationality,” American Journal of Political Science, 22, 680–90. P. R. Abramson (2007) “The French Presidential Election of 2007: Was Sarkozy the Condorcet Winner?” French Politics, 5, 287–91. P. R. Abramson, J. H. Aldrich, and D. W. Rohde (1990) Change and Continuity in the 1988 Elections (Washington, DC: Congressional Quarterly Press). P. R. Abramson, J. H. Aldrich, and D. W. Rohde (2003) Change and Continuity in the 2000 Elections (Washington, DC: Congressional Quarterly Press). R. Alford (1963) Party and Society: The Anglo-American Democracies (Chicago, IL: Rand McNally). S. Ansolabehere, R. Behr, and S. Iyengar (1993) The Media Game (New York: Macmillan). A. Appleton (1996) “The Maastricht Referendum and the Party System” in J. T. S. Keller and M. A. Schain (eds.) Chirac’s Challenge: Liberalization, Europeanization, and Malaise in France (New York: St. Martin’s). R. Aron (1955) “Électeurs, partis et élus,” Revue française de science politique, 5, 246–8. A. Auberger (2010) “Popularity and Vote: Forecasting the 2007 French Presidential Election,” Canadian Journal of Political Science, 43, 123–36. L. M. Bartels (2006) “Priming and Persuasion in Presidential Campaigns” in H. E. Brady and R. Johnston (eds.) Capturing Campaign Effects (Ann Arbor, MI: University of Michigan Press). L. M. Bartels (2008) Unequal Democracy: The Political Economy of the New Gilded Age (Princeton, NJ: Princeton University Press). C. Bean and A. Mughan (1989) “Leadership Effects in Parliamentary Elections,” American Political Science Review, 83, 1165–80. É. Bélanger and M. S. Lewis-Beck (2010) “French National Elections: Democratic Disequilibrium and the 2007 Forecasts,” Canadian Journal of Political Science, 43, 107–22. É. Bélanger, M. S. Lewis-Beck, J. Chiche, and V. Tiberj (2006) “Party, Ideology, and Vote Intentions: Dynamics from the 2002 French Electoral Panel,” Political Research Quarterly, 59, 503–15. É. Bélanger and B. M. Meguid (2008) “Issue Salience, Issue Ownership, and Issue-Based Vote Choice,” Electoral Studies, 27, 477–91. É. Bélanger and R. Nadeau (2009) Le comportement électoral des Québécois (Montréal: Presses de l’Université de Montréal). B. B. Berelson, P. F. Lazarsfeld, and W. N. McPhee (1954) Voting (Chicago, IL: University of Chicago Press). A. Blais and M. Boyer (1996) “Assessing the Impact of Televised Debates: The Case of the 1988 Canadian Election,” British Journal of Political Science, 26, 143–64. 215

9780230_300088_12_bib.indd 215

6/28/2011 7:59:57 PM

216

Bibliography

A. Blais, E. Gidengil, R. Nadeau, and N. Nevitte (2002) Anatomy of a Liberal Victory: Making Sense of the Vote in the 2000 Canadian Election (Peterborough, ON: Broadview Press). A. Blais, E. Gidengil, R. Nadeau, and N. Nevitte (2003) “Campaign Dynamics in the 2000 Canadian Election: How the Leader Debates Salvaged the Conservative Party,” PS: Political Science and Politics, 36, 45–50. A. Blais, E. Gidengil, N. Nevitte, and R. Nadeau (2004) “Where Does Turnout Decline Come From?” European Journal of Political Research, 43, 221–36. J. Blumler, R. Cayrol, and G. Thoveron (1978) La télévision fait-elle l’élection? (Paris: Presses de la Fondation Nationale des Sciences Politiques). D. Boy and J. Chiche (2007) “L’image des candidats dans la décision électorale,” Revue française de science politique, 57, 329–42. D. Boy and J. Chiche (2009) “L’influence décisive des images” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leur choix? (Paris: Presses de Sciences Po). D. Boy and N. Mayer (eds.) (1990) L’électeur français en questions (Paris: Presses de la Fondation Nationale des Sciences Politiques). D. Boy and N. Mayer (eds.) (1993) The French Voter Decides (Ann Arbor, MI: University of Michigan Press). D. Boy and N. Mayer (eds.) (1997) L’électeur a ses raisons (Paris: Presses de la Fondation Nationale des Sciences Politiques). H. E. Brady, R. Johnston, and J. Sides (2006) “The Study of Political Campaigns” in H. E. Brady and R. Johnston (eds.) Capturing Campaign Effects (Ann Arbor, MI: University of Michigan Press). P. Bréchon and B. Denni (2008) “L’élection dans son contexte: candidats et forces en présence” in P. Bréchon (ed.) Les élections présidentielles en France: quarante ans d’histoire politique (Paris: La Documentation française). A. Campbell, P. E. Converse, W. E. Miller, and D. E. Stokes (1960) The American Voter (New York: Wiley). J. E. Campbell (2000) The American Campaign (College Station: Texas A&M University Press). P. Campbell (1965) French Electoral Systems since 1789 (Hamden, CT: Archon Books). J. Capdevielle, E. Dupoirier, G. Grunberg, É. Schweisguth, and C. Ysmal (1981) France de gauche, vote à droite (Paris: Presses de la Fondation Nationale des Sciences Politiques). E. G. Carmines and J. A. Stimson (1980) “The Two Faces of Issue Voting,” American Political Science Review, 74, 78–91. B. Cautrès (2004) “Old Wine in New Bottles? New Wine in Old Bottles? Class, Religion and Vote in the French Electorate. The 2002 Elections in Time Perspective” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). B. Cautrès and A. Heath (1996) “Déclin du vote de classe? Une analyse comparative franco-britannique,” Revue internationale de politique comparée, 3, 541–68. B. Cautrès and A. Jadot (2007) “L’(in)décision électorale et la temporalité du vote: le moment du choix pour le premier tour de l’élection présidentielle de 2007,” Revue française de science politique, 57, 293–314.

9780230_300088_12_bib.indd 216

6/28/2011 7:59:57 PM

Bibliography

217

B. Cautrès and A. Jadot (2009) “Quand les électeurs se décident-ils?” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leur choix? (Paris: Presses de Sciences Po). B. Cautrès and N. Mayer (eds.) (2004) Le nouveau désordre électoral: les leçons du 21 avril 2002 (Paris: Presses de Sciences Po). B. Cautrès and A. Muxel (eds.) (2009) Comment les électeurs font-ils leur choix? Le Panel électoral français 2007 (Paris: Presses de Sciences Po). B. Cautrès and D. Reynié (2001) “Construction européenne: le mythe du grand projet mobilisateur” in D. Reynié and B. Cautrès (eds.) L’opinion européenne (Paris: Presses de Sciences Po). R. Cayrol (1988) “The Electoral Campaign and the Decision-Making Process of French Voters” in H. R. Penniman (ed.) France at the Polls, 1981 and 1988: Three National Elections (Durham, NC: Duke University Press). CEVIPOF (1988) “Enquête post-électorale des élections présidentielles de 1988,” Database distributed by the Centre de données socio-politiques (Paris: Sciences Po). CEVIPOF (1995) “Enquête post-électorale des élections présidentielles de 1995,” Database distributed by the Centre de données socio-politiques (Paris: Sciences Po). CEVIPOF (2002) “Panel électoral français des élections présidentielles de 2002,” Database distributed by the Centre de données socio-politiques (Paris: Sciences Po). CEVIPOF (2007) “Enquête post-présidentielle des élections françaises de 2007,” Database distributed by the Centre de données socio-politiques (Paris: Sciences Po). S. H. Chaffee and R. N. Rimal (1996) “Time of Vote Decision and Openness to Persuasion” in D. C. Mutz, P. M. Sniderman, and R. A. Brody (eds.) Political Persuasion and Attitude Change (Ann Arbor, MI: University of Michigan Press). J. Chiche and N. Mayer (1997) “Les enjeux de l’élection” in D. Boy and N. Mayer (eds.) L’électeur a ses raisons (Paris: Presses de Sciences Po). H. D. Clarke, D. Sanders, M. C. Stewart, and P. Whiteley (2004) Political Choice in Britain (Oxford: Oxford University Press). Conseil constitutionnel français (1965, 1969, 1974, 1981, 1988, 1995, 2002, 2007) Proclamation des résultats de l’élection du Président de la République, www.conseil-constitutionnel.fr. Conseil constitutionnel français (1988, 1995, 2002, 2007) Déclaration présidentielle, résultats du premier tour de scrutin, www.conseil-constitutionnel.fr. P. E. Converse (1962) “Information Flow and the Stability of Partisan Attitudes,” Public Opinion Quarterly, 26, 578–99. P. E. Converse and G. Dupeux (1966) “De Gaulle and Eisenhower: The Public Image of the Victorious General” in A. Campbell, P. E. Converse, W. E. Miller, and D. E. Stokes (eds.) Elections and the Political Order (New York: Wiley and Sons). P. E. Converse and R. Pierce (1986) Political Representation in France (Cambridge, MA: Belknap-Harvard University Press).

9780230_300088_12_bib.indd 217

6/28/2011 7:59:57 PM

218 Bibliography

M. Costa Lobo (2008) “Parties and Leader Effects: Impact of Leaders in the Vote for Different Types of Parties,” Party Politics, 14, 281–98. I. Crewe and A. King (1994) “Did Major Win? Did Kinnock Lose? Leadership Effects in the 1992 Election” in A. Heath, R. Jowell, and J. Curtice (eds.) Labour’s Last Chance? The 1992 Election and Beyond (Dartmouth: Aldershot). J. Curtice and S. Holmberg (2005) “Leadership” in J. Thomassen (ed.) The European Voter (Oxford: Oxford University Press). R. J. Dalton and M. P. Wattenberg (eds.) (2000) Parties without Partisans: Political Change in Advanced Industrial Democracies (Oxford: Oxford University Press). E. Deutsch, D. Lindon, and P. Weill (1966) Les familles politiques aujourd’hui en France (Paris: Éditions de Minuit). C. H. De Vreese (2010) “Campaign Communication and Media” in L. LeDuc, R. G. Niemi, and P. Norris (eds.) Comparing Democracies 3: Elections and Voting in the 21st Century (Los Angeles: Sage). A. Downs (1957) An Economic Theory of Democracy (New York: Harper). J. N. Druckman (2003) “The Power of Television Images: The First KennedyNixon Debate Revisited,” Journal of Politics, 65, 559–71. G. Dupeux (1959) Le Front Populaire et les élections de 1936 (Paris: A. Colin). É. Dupoirier (2007) “Les classes moyennes” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). É. Dupoirier and A.-P. Frognier (2009) “Les temporalités de la décision électorale: approche comparée de la France et de la Belgique (Wallonie),” Revue française de science politique, 59, 507–33. M. Duverger (1955) Partis politiques et classes sociales en France (Paris: Fondation Nationale des Sciences Politiques, Armand Colin), Cahier no. 74. M. Duverger (1960) “Introduction” in J. Cotteret, C. Emeri, and P. Lalumière (eds.) Lois électorales et inégalités de représentation en France, 1936–1960 (Paris: Fondation Nationale des Sciences Politiques, Armand Colin), Cahier no. 107. H. W. Ehrmann (1952) “The French Peasant and Communism,” American Political Science Review, 46, 19–43. H. W. Ehrmann (1983) Politics in France, 4th edn (Boston, MA: Little, Brown). R. Elgie (2004) “Institutions and Voters: Structuring Electoral Choice” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). G. Evans and P. Norris (eds.) (1997) Critical Elections: British Parties and Voters in Long-Term Perspective (London: Sage). J. A. J. Evans (2004) “Ideology and Party Identification: A Normalization of French Voting Anchors?” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). D. M. Farrell (1996) “Campaign Strategies and Tactics” in L. LeDuc, R. G. Niemi, and P. Norris (eds.) Comparing Democracies (Thousand Oaks, CA: Sage). C. Fauvelle-Aymar, M. S. Lewis-Beck, and R. Nadeau (2011) “French Electoral Reform and the Abstention Rate,” Parliamentary Affairs, 64, 45–60. J. Fauvet (1955) “La représentation politique du monde paysan” in M. Duverger (ed.) Partis politiques et classes sociales en France (Paris: Fondation Nationale des Sciences Politiques, Armand Colin), Cahier no. 74.

9780230_300088_12_bib.indd 218

6/28/2011 7:59:58 PM

Bibliography

219

M. P. Fiorina (1981) Retrospective Voting in American National Elections (New Haven, CT: Yale University Press). W. H. Flanagan and N. Zingale (2006) Political Behavior of the American Electorate, 11th edn (Washington, DC: Congressional Press Quarterly). C. J. Fleury and M. S. Lewis-Beck (1993) “Anchoring the French Voter: Ideology versus Party,” Journal of Politics, 55, 1100–9. M. Foucault, R. Nadeau, and M. S. Lewis-Beck (2011) “La persistance de l’effet patrimoine lors des élections présidentielles françaises,” Revue française de science politique, 9, 97–119. P. Fournier, R. Nadeau, A. Blais, E. Gidengil, and N. Nevitte (2004) “Time-ofVoting Decision and Susceptibility to Campaign Effects,” Electoral Studies, 23, 661–81. M. N. Franklin, T. T. Mackie, and H. Valen (1992) Electoral Change: Responses to Evolving Social and Attitudinal Structures in Western Nations (Cambridge: Cambridge University Press). C. L. Funk (1997) “Implications of Political Expertise in Candidate Traits Evaluation,” Political Research Quarterly, 50, 675–97. J. Geer (1988) “The Effects of Presidential Debates on the Electorate’s Preferences for Candidates,” American Politics Quarterly, 16, 486–501. A. Gelman and G. King (1993) “Why are American Presidential Election Campaign Polls So Variable When Votes are So Predictable?” British Journal of Political Science, 23, 409–51. J. Gerstlé (1997) “La persuasion de l’actualité télévisée,” Politix, 37, 81–96. J. Gerstlé (2008) “Les campagnes présidentielles depuis 1965” in P. Bréchon (ed.) Les élections présidentielles en France: quarante ans d’histoire politique (Paris: La Documentation française), 67–104. T. Giasson, R. Nadeau, and É. Bélanger (2005) “Débats télévisés et évaluations des candidats: la représentation visuelle des politiciens canadiens agit-elle dans la formation des préférences de l’électorat québécois?” Canadian Journal of Political Science, 38, 867–95. F. Goguel (1951) Géographie des élections françaises de 1870 à 1951 (Paris: Armand Colin). F. Goguel (1966) “L’élection présidentielle française de décembre 1965,” Revue française de science politique, 16, 221–54. F. Goguel (1970) Géographie des élections françaises sous la Troisième et la Quatrième République (Paris: Presses de la Fondation Nationale des Sciences Politiques). F. Goguel (1974) “Du 5 mai 1946 au 19 mai 1974,” Revue française de science politique, 24, 893–910. F. Haegel and N. Sauger (2007) “L’électorat de droite: le rapport des forces UMP-UDF à l’épreuve” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). P. R. Hagner and L. N. Rieselbach (1980) “The Impact of the 1976 Presidential Debates: Conversion or Reinforcement?” in G. F. Bishop, R. G. Meadow, and M. Jackson-Beeck (eds.) The Presidential Debates: Media, Electoral, and Policy Perspectives (New York: Praeger).

9780230_300088_12_bib.indd 219

6/28/2011 7:59:58 PM

220

Bibliography

D. S. Hillygus and S. Jackman (2003) “Voter Decision Making in Election 2000: Campaign Effects, Partisan Activation, and the Clinton Legacy,” American Journal of Political Science, 47, 583–96. S. Hoffman (1974) “The Rulers: Heroic Leadership in Modern France” in S. Hoffman (ed.) Decline or Renewal? France since the 1930s (New York: Viking). T. M. Holbrook (1996) Do Campaigns Matter? (Thousand Oaks, CA: Sage). R. Inglehart (1984) “The Changing Structure of Political Cleavages in Western Society” in R. J. Dalton, S. C. Flanagan, and P. A. Beck (eds.) Electoral Change in Advanced Industrial Societies: Realignment or Dealignment? (Princeton, NJ: Princeton University Press). R. Inglehart (1990) Culture Shift in Advanced Industrial Society (Princeton, NJ: Princeton University Press). R. Inglehart (1997) Modernization and Post-Modernization (Princeton, NJ: Princeton University Press). R. Inglehart and P. Norris (2003) Rising Tide: Gender Equality and Cultural Change around the World (New York: Cambridge University Press). B. Jérôme and V. Jérôme-Speziari (2007) “Présidentielle: ce que prévoit la simulation des économistes,” L’Expansion, January 25. B. Jérôme and V. Jérôme-Speziari (2010) “Fonctions de vote et prévisions électorales: une application à la présidentielle française de 2007,” Canadian Journal of Political Science, 43, 163–86. B. Jérôme, V. Jérôme-Speziari, and M. S. Lewis-Beck (2003) “Reordering the French Election Calendar: Forecasting the Consequences for 2002,” European Journal of Political Research, 42, 425–40. A. Johnston and J. Gertslé (1995) “The Role of Television Broadcasts in Promoting French Presidential Candidates” in L. L. Kaid and C. Holz-Bacha (eds.) Political Advertising in Western Democracies: Parties and Candidates on Television (Thousand Oaks, CA: Sage). J. H. Kessel (2004) “Views of the Voters” in H. F. Weisberg and C. Wilcox (eds.) Models of Voting in Presidential Elections (Stanford, CA: Stanford University Press). V. O. Key Jr. (1966) The Responsible Electorate (Cambridge, MA: Harvard University Press). D. R. Kiewiet (1983) Macroeconomics and Micropolitics: The Electoral Effects of Economic Issues (Chicago, IL: University of Chicago Press). D. R. Kinder (1986) “Presidential Character Revisited” in R. R. Lau and D. O. Sears (eds.) Political Cognition (Hillside, NJ: Earlbaum). D. R. Kinder and D. R. Kiewiet (1981) “Sociotropic Politics: The American Case,” British Journal of Political Science, 11, 129–62. J. Klatzmann (1988) “Comment votent les paysans français,” Revue française de science politique, 38, 13–41. A. Knapp (2004) Parties and the Party System in France: A Disconnected Democracy? (Basingstoke: Palgrave Macmillan). J.-D. Lafay (1977) ‘Les conséquences électorales de la conjoncture économique: essais de prévision chiffrée pour mars 1978’, Vie et sciences économiques, 75, 1–7.

9780230_300088_12_bib.indd 220

6/28/2011 7:59:58 PM

Bibliography

221

J.-D. Lafay (1991) “Political Dyarchy and Popularity Functions: Lessons from the 1986 French Experience” in H. Norpoth, M. S. Lewis-Beck, and J.-D. Lafay (eds.) Economics and Politics: The Calculus of Support (Ann Arbor, MI: University of Michigan Press). A. Lancelot (1975) “Une métaphore économique en sociologie politique: l’analyse de l’offre électorale,” Revue européenne des sciences sociales, 23, 55–75. A. Lancelot (1981) “Préface” in J. Capdevielle, É. Dupoirier, G. Grunberg, É. Schweisguth, and C. Ysmal (eds.) France de gauche, vote à droite (Paris: Presses de Science Po). A. Lancelot (1986) 1981: les élections de l’alternance (Paris: Presses de la Fondation Nationale des Sciences Politiques). D. Lanoue (1991) “Debates That Mattered: Voters’ Reaction to the 1984 Canadian Leadership Debates,” Canadian Journal of Political Science, 24, 51–65. A. Laurent (2004) “France’s 2002 Presidential Elections: Earlier and Later Territorial Fractures” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). P. F. Lazarsfeld, B. B. Berelson, and H. Gaudet (1944) The People’s Choice (New York: Columbia University Press). M. J. Lebo and H. Norpoth (2007) “The PM and the Pendulum: Dynamic Forecasting of British Elections,” British Journal of Political Science, 37, 71–87. V. Le Hay and J. Mossuz-Lavau (2010) “Genre et politique: y a-t-il une exception française?” in D. Boy and B. Cautrès (eds.) Les Français, des Européens comme les autres? (Paris: Presses de Sciences Po). M. S. Lewis-Beck (1977) “Explaining Peasant Conservatism: The Western European Case” British Journal of Political Science, 7, 447–64. M. S. Lewis-Beck (1981) “The Electoral Politics of the French Peasantry: 1946– 1978,” Political Studies, 29, 517–36. M. S. Lewis-Beck (1983) “Economics and the French Voter: A Microanalysis,” Public Opinion Quarterly, 47, 347–60. M. S. Lewis-Beck (1984) “France: The Stalled Electorate” in R. J. Dalton, S. C. Flanagan, and P. A. Beck (eds.) Electoral Change in Advanced Industrial Societies: Realignment or Dealignment? (Princeton, NJ: Princeton University Press). M. S. Lewis-Beck (1988) Economics and Elections: The Major Western Democracies (Ann Arbor, MI: University of Michigan Press). M. S. Lewis-Beck (1993) “The French Voter: Steadfast or Changing?” in D. Boy and N. Mayer (eds.) The French Voter Decides (Ann Arbor, MI: University of Michigan Press). M. S. Lewis-Beck (1996) ‘Cross-National Election Surveys: A French Pre-Test’, Electoral Studies, 15, 513–28. M. S. Lewis-Beck (1997a) “Le vote du ‘porte-monnaie’ en question” in D. Boy and N. Mayer (eds.) L’électeur a ses raisons (Paris: Presses de Sciences Po). M. S. Lewis-Beck (1997b) ‘Who’s the Chef? Economic Voting Under a Dual Executive’, European Journal of Political Research, 31, 315–25.

9780230_300088_12_bib.indd 221

6/28/2011 7:59:58 PM

222

Bibliography

M. S. Lewis-Beck (1998) “Class, Religion, and the French Voter: A ‘Stalled’ Electorate?” French Politics and Society, 16, 43–51. M. S. Lewis-Beck (ed.) (2000) How France Votes (New York: Chatham House). M. S. Lewis-Beck (2001) “Modèles économétriques du vote” in P. Perrineau and D. Reynié (eds.) Dictionnaire du vote (Paris: Presses Universitaires de France). M. S. Lewis-Beck (2004) “French Election Theories and the 2002 Results: An Introduction” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). M. S. Lewis-Beck, É. Bélanger, and C. Fauvelle-Aymar (2008) “Forecasting the 2007 French Presidential Election: Ségolène Royal and the Iowa Model,” French Politics, 6, 106–15. M. S. Lewis-Beck and K. Chlarson (2002) “Party, Ideology, Institutions and the 1995 French Presidential Election,” British Journal of Political Science, 32, 489–512. M. S. Lewis-Beck, A. Hildreth, and A. B. Spitzer (1988) “Was There a Girondist Faction in the National Convention, 1792–1793?,” French Historical Studies, 15, 519–48. M. S. Lewis-Beck, W. G. Jacoby, H. Norpoth, and H. F. Weisberg (2008) The American Voter Revisited (Ann Arbor, MI: University of Michigan Press). M. S. Lewis-Beck and B. Jérôme (1999) “Is Local Politics Local? French Evidence,” European Journal of Political Research, 35, 181–97. M. S. Lewis-Beck and G. Mitchell II (1993) “French Electoral Theory: The National Front Test,” Electoral Studies, 12, 112–27. M. S. Lewis-Beck and D. Morey (2007) “The French ‘Petit Oui:’ The Maastricht Treaty and the French Voting Agenda,” Journal of Inter-Disciplinary History, 38, 65–87. M. S. Lewis-Beck and R. Nadeau (2000) “French Electoral Institutions and the Economic Vote,” Electoral Studies, 19, 171–82. M. S. Lewis-Beck and R. Nadeau (2004) “Dual Governance and the Economic Vote: France and the United States” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). M. S. Lewis-Beck and R. Nadeau (2009) “Obama and the Economy in 2008,” PS: Political Science and Politics, 42, 479–83. M. S. Lewis-Beck and R. Nadeau (2010) ‘Leadership Traits in French National Studies: Measurement Issues’, French Politics, 8, 215–21. M. S. Lewis-Beck and R. Nadeau (2011) “Economic Voting Theory: Testing New Dimensions,” Electoral Studies, 30, 288–94. M. S. Lewis-Beck and M. Stegmaier (2000) “Economic Determinants of Electoral Outcomes,” Annual Review of Political Science, 3, 183–219. M. S. Lewis-Beck and M. Stegmaier (2007) “Economic Models of Voting” in R. Dalton and H.-D. Klingemann (eds.) The Oxford Handbook of Political Behavior (Oxford: Oxford University Press). D. Lindon and P. Weill (1974) “Autopsie d’une campagne: pourquoi M. Giscard d’Estaing a-t-il gagné?” Le Monde, May 22. W. Lippmann (1922) Public Opinion (New York: Macmillan).

9780230_300088_12_bib.indd 222

6/28/2011 7:59:58 PM

Bibliography

223

W. Lippmann (1925) The Phantom Public (New York: Harcourt Brace). S. M. Lipset (1960) Political Man: The Social Bases of Politics (Garden City: Doubleday). P. Maarek (2000) Communication et marketing de l’homme politique (Paris: Litec). R. Macridis (1975) French Politics in Transition: The Years after De Gaulle (Cambridge, MA: Winthrop). G. E. Marcus and M. B. MacKuen (1993) “Anxiety, Enthusiasm, and the Vote: The Emotional Underpinnings of Learning and Involvement during Presidential Campaigns,” American Political Science Review, 87, 672–85. G. E. Marcus, R. W. Neuman, and M. B. MacKuen (2000) Affective Intelligence and Political Judgment (Chicago, IL: University of Chicago Press). P. Martin (2000) Comprendre les évolutions électorales: la théorie des réalignements revisitée (Paris: Presses de Sciences Po). N. Mayer (1988) “L’effet Le Pen s’est nourri de l’effet premier tour” in Le Monde: Dossiers et Documents, L’élection présidentielle, 44, May. N. Mayer (1993) “Ethnocentrism, Racism, and Intolerance” in D. Boy and N. Mayer (eds.) The French Voter Decides (Ann Arbor, MI: University of Michigan Press). N. Mayer and P. Perrineau (1992) Les comportements politiques (Paris: Armand Colin). N. Mayer and V. Tiberj (2004) “Do Issues Matter? Law and Order in the 2002 French Presidential Election” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). I. McAllister (2002) “Calculating or Capricious? The New Politics of Late Deciding Voters” in D. M. Farrell and R. Schmitt-Beck (eds.) Do Political Campaigns Matter? Campaign Effects in Elections and Referendums (London: Routledge). G. Michelat (1990) “À la recherche de la gauche et de la droite” in D. Boy and N. Mayer (eds.) L’électeur français en questions (Paris: Presses de Sciences Po). G. Michelat and M. Simon (1977) Classe, religion et comportement politique (Paris: Presses de la Fondation Nationale de Sciences Politiques). G. Michelat and V. Tiberj (2007) “Gauche, centre, droite et vote: permanence et mutation d’une opposition,” Revue française de science politique, 57, 371–92. A. H. Miller, M. P. Wattenberg, and O. Malanchuk (1986) “Schematic Assessments of Presidential Candidates,” American Political Science Review, 80, 521–40. W. E. Miller and J. M. Shanks (1996) The New American Voter (Cambridge, MA: Harvard University Press). J. Mossuz-Lavau (2007) “Les femmes” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). A. Muxel (2007) “Les jeunes” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). A. Muxel (2009) “Loyautés, mobilités et abstentions” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leur choix? (Paris: Presses de Sciences Po). R. Nadeau (2007) “Ségolène Royal peut-elle encore gagner?” Le Devoir, March 10, A9.

9780230_300088_12_bib.indd 223

6/28/2011 7:59:58 PM

224 Bibliography

R. Nadeau and F. Bastien (2003) “La communication électorale” in A.-M. Gingras (ed.) La communication politique: états des savoirs, enjeux et perspectives (Québec: Presses de l’Université du Québec). R. Nadeau, M. Foucault, and M. S. Lewis-Beck (2010a) “Patrimonial Economic Voting: A Neglected Dimension,” French Politics (under review). R. Nadeau, M. Foucault, and M. S. Lewis-Beck (2010b) “Patrimonial Economic Voting: Legislative Elections in France,” West European Politics, 33, 1261–77. R. Nadeau and M. S. Lewis-Beck (2010) Leader Images and the Presidential Vote in France, paper presented at the Portuguese Political Science Association, Aveiro, Portugal, March 2010. (An earlier version was presented in the ECPR Workshop “Leaders and Democratic Elections,” Lisbon, April 2009). R. Nadeau, M. S. Lewis-Beck, and É. Bélanger (2010) “Electoral Forecasting in France: A Multi-Equation Solution,” International Journal of Forecasting, 26, 11–18. R. Nadeau, N. Nevitte, E. Gidengil, and A. Blais (2008) “Election Campaigns as Information Campaigns: Who Learns What and Does it Matter?” Political Communication, 25, 229–48. R. Nadeau, R. G. Niemi, and T. Amato (1995) “Emotions, Issue Importance and Political Learning,” American Journal of Political Science, 39, 558–74. R. Nadeau, R. G. Niemi, H. W. Stanley, and J.-F. Godbout (2004) “Class, Party and South/Non-South Differences: An Update,” American Politics Quarterly, 32, 52–67. R. Nadeau, F. Pétry, and É. Bélanger (2010) “Issue-Based Strategies in Election Campaigns: The Case of Health Care in the 2000 Canada Federal Election,” Political Communication, 27, 367–88. N. Nevitte (1996) The Decline of Deference (Peterborough, ON: Broadview Press). R. G. Niemi and H. W. Stanley (2006) “Partisanship, Party Coalitions and Group Support, 1952–2004,” Presidential Studies Quarterly, 36, 172–88. P. Norris, J. Curtice, D. Sanders, M. Scammel, and H. Semetko (1999) On Message: Communicating the Campaign (London: Sage). H. Norpoth (1987) “Under Way and Here to Stay: Party Realignments in the 1980s?” Public Opinion Quarterly, 51, 376–91. H. Norpoth (2002) “On a Short Leash: Term Limits and the Economic Voter” in H. Dorussen and M. Taylor (eds.) Economic Voting (London: Routledge). H. Norpoth (2004) “Forecasting British Elections: A Dynamic Perspective,” Electoral Studies, 23, 297–305. P. Perrineau (1997) Le symptôme Le Pen: radiographie des électeurs du Front national (Paris: Fayard). P. Perrineau (2001) “Élections présidentielles” in P. Perrineau and D. Reynié (eds.) Dictionnaire du vote (Paris: Presses Universitaires de France). P. Perrineau (ed.) (2007) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). P. Perrineau (2009) “La “défidélisation” des électeurs de Jean-Marie Le Pen” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leur choix? (Paris: Presses de Sciences Po).

9780230_300088_12_bib.indd 224

6/28/2011 7:59:58 PM

Bibliography

225

R. C. Pierce (1988) “French Presidential Election Survey, 1988,” ICPSR version (Ann Arbor, MI: Roy Pierce, University of Michigan (producer), 1995) (Ann Arbor, MI: Inter-University Consortium for Political and Social Research (distributor), 1996). R. C. Pierce (1995) Choosing the Chief: Presidential Elections in France and the United States (Ann Arbor, MI: University of Michigan Press). R. C. Pierce (2002) “Candidate Evaluations and Presidential Electoral Choices in France” in A. King (ed.) Leaders’ Personalities and the Outcomes of Democratic Elections (Oxford: Oxford University Press). T. Piketty (2003) “Income Inequality in France, 1991–1998,” Journal of Political Economy, 111, 1004–42. T. Piketty and E. Saez (2003) “Income Inequality in the United States, 1913– 1998,” Quarterly Journal of Economics, 118, 1–39. F. Platone (2007) “L’électorat communiste: la peau de chagrin” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). S. L. Popkin (1994) The Reasoning Voter. Communication and Persuasion in Presidential Campaigns, 2nd edn (Chicago, IL: University of Chicago Press). J. Ranger (1990) “Les Français s’intéressent-ils à la politique?” in D. Boy and N. Mayer (eds.) L’électeur français en questions (Paris: Presses de Sciences Po). H. Rey (2007) “L’électorat socialiste: volatilité et inconstance” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). G. Ross (2000) “Europe Becomes French Domestic Politics” in M. S. LewisBeck (ed.) How France Votes (New York: Chatham House). N. Sauger (2009) “Agenda électoral et vote sur enjeux” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leur choix? (Paris: Presses de Sciences Po). M. Schain (1987) “The National Front in France and the Construction of Political Legitimacy,” West European Politics, 10, 227–52. P. Senior (2008) “Electoral Impact of Televised Leaders’ Debates on Australian Federal Elections,” Australian Journal of Political Science, 43, 443–64. A. Siegfried (1913) Tableau politique de la France de l’Ouest sous la Troisième République (Paris: Armand Colin). M. Sineau (2010) La force du nombre: femmes et démocratie présidentielle (La Tour d’Aigues, Éditions de l’Aube, 2nd edn). T. A. Small (2008) “The Facebook Effect? On-Line Campaigning in the 2008 Canadian and US Elections,” Policy Options, 29, 85–87. D. E. Stokes (1963) “Spatial Models of Party Competition,” American Political Science Review, 57, 368–77. D. E. Stokes and G. Iversen (1966) “On the Existence of Forces Restoring Party Competition” in A. Campbell, P. E. Converse, W. E. Miller, and D. E. Stokes (eds.) Elections and the Political Order (New York: Wiley). J. M. Stonecash (2000) Class and Party in American Politics (Boulder, CO: Westview Press). S. Strudel (2009) “La tentation bayrouiste” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leurs choix? (Paris: Presses de Sciences Po). Y. Tavernier (1975–76) Sociologie politique du monde rural et politique agricole de 1914 à nos jours (Paris: Fondation Nationale des Sciences Politiques).

9780230_300088_12_bib.indd 225

6/28/2011 7:59:58 PM

226

Bibliography

V. Tiberj (2004) “Compétence et repérage politiques en France et aux ÉtatsUnis: une contribution au modèle de ‘l’électeur raisonnant,” Revue française de science politique, 54, 261–87. V. Tiberj (2007) “Existe-t-il ‘un vote de l’immigration’? Les nouveaux Français à la veille de 2007” in P. Perrineau (ed.) Atlas électoral: qui vote quoi, où, comment? (Paris: Presses de Sciences Po). V. Tiberj and T. Vedel (2007) “Les effets de l’information télévisée sur les évaluations politiques et les préoccupations des électeurs français,” Les Cahiers du Cevipof, 47, 1–26. M. Tomz, J. Wittenberg, and G. King (2003) CLARIFY: Software for Interpreting and Presenting Statistical Results. L. Vavreck (2009) The Message Matters: The Economy and Presidential Campaigns (Princeton, NJ: Princeton University Press). T. Vedel (2009) “Le suivi de la campagne: politisation, confirmation et polarisation” in B. Cautrès and A. Muxel (eds.) Comment les électeurs font-ils leur choix? (Paris: Presses de Sciences Po). M. P. Wattenberg (1998) The Decline of American Political Parties, 1952–1996 (Cambridge, MA: Harvard University Press). P. Webb and T. Poguntke (2005) The Presidentialization of Parliamentary Democracies (Oxford: Oxford University Press). G. Whitten (2004) “Could There Have Possibly Been Economic Voting?” in M. S. Lewis-Beck (ed.) The French Voter: Before and After the 2002 Elections (Basingstoke: Palgrave Macmillan). C. Wlezien (2010) “Election Campaigns” in L. LeDuc, R. G. Niemi, and P. Norris (eds.) Comparing Democracies 3: Elections and Voting in the 21st Century (Los Angeles: Sage). J. M. Woolridge (2006) Introductory Econometrics: A Modern Approach, 3rd edn (Mason, OH: Thomson/South Western). G. Wright (1981) France in Modern Times, 3rd edn (New York: Norton and Company). C. Ysmal (1994) “The History of Electoral Studies in France,” European Journal of Political Research, 25, 367–85. J. R. Zaller (1989) “Bringing Converse Back In: Modeling Information Flow in Political Campaigns” in J. Stimson (ed.) Political Analysis (Chicago, IL: University of Chicago Press). J. R. Zaller (1992) The Nature and Origins of Mass Opinion (Cambridge: Cambridge University Press). T. Zeldin (1973) France 1848–1945. Vol.1: Ambition, Love and Politics (Oxford: Clarendon Press). T. Zeldin (1977) “The Destruction of the Peasants,” New York Review of Books, 24, 45 (November 24).

9780230_300088_12_bib.indd 226

6/28/2011 7:59:58 PM

Index age, 15, 18, 26, 29, 31–40 agenda-setting effect, 142 Alford, Robert, 46 American National Election Surveys (ANES), 94 Appels Nominaux, 68 Aubry, Martine, 165 Australia, 113

center, 5, 16, 20, 30, 34–6, 58, 60, 61, 70, 72, 76–8, 81, 83, 85 centrists, 105–6, 152–3, 166, 167, 172 CEVIPOF survey, 12, 16, 54–5, 70, 74, 87, 90, 94, 118, 119–20, 125–6, 139, 214n1 Chevènement, Jean-Pierre, 108 Chirac, Jacques, 1, 3, 4, 7, 29, 79, 98, 106, 214n1 election of 1988, 28, 36, 51, 105, 139, 148 election of 1995, 97 election of 2002, 5, 8, 17, 84, 94, 136, 154, 157, 158 image of, 115, 118, 123, 124, 127, 157 issues and, 108 popularity of, 122 satisfaction with, 134 supporters of, 32, 37–9, 61, 66 class-related determinants, 41–67, 166 economic sector, 44–53 occupation, 44–53 Clinton administration, 142 cognitive psychology, 117 cohabitation, 7, 8, 96–7, 167 Common Market, 107 Communist party, 4, 10, 16, 44, 107, 108 competence, 122–5, 134, 135, 140, 168 Conservative party, 10 constitution, 5, 6 Cramer’s V, 29, 51, 81, 99, 148, 152 Cresson, Edith, 6 criminality, 88

Balladur, Edouard, 7, 31, 38, 98, 108, 157 Bayrou, François, 5, 21, 30–1, 35, 38, 40, 66–7, 157 Bayrou temptation, 18, 38, 43, 61, 167 block recursive model, 17 blue-collar workers, 51, 59, 62, 63, 67, 166 Boy, Daniel, 42 campaign attentiveness, 144–51, 169 campaign deciders, 143, 152, 153 Canada, 16, 93, 113 candidates, 13, 94 competence of, 122–3, 125, 134, 135, 140, 168 debates among, 158–60 evaluations of, 125–35 honesty of, 115, 118, 119, 122, 124, 125, 134, 135 image of, 113–40, 168 impact of television on, 155–60 incumbent, 6, 97, 105, 117, 142, 175 personality traits of, 116–18, 134, 135, 136 thermometer ratings of, 115, 117, 120–2, 125, 128–34, 137 voter expectations of, 117 Catholics, 21, 30, 35, 166

debates, 158–60

227

9780230_300088_13_ind.indd 227

6/28/2011 7:59:22 PM

228 Index

de Gaulle, Charles, 2, 6, 97, 100, 107, 113–15, 168 Democratic party, 10, 18, 21 depolarization, 166–7 diversity moment, 164 divided executive, 7, 8, 96–7 double-ballot system, 164–5 Dupoirier, Elisabeth, 42 economic issues, 89, 93, 95–100, 103–7, 127, 167 economic perception, 13 economic sector, 44–53, 57–66 education level, 15–16, 19, 25, 27, 28–9, 30, 31–8, 39–40, 166 egotropic economic voting, 95 1848 revolution, 41 election campaigns, 141–62, 168–9 agenda-setting effect, 142 campaign attentiveness and, 144–51 framing effect, 142 impact of television in, 155–60, 169 priming effect, 142 time of vote decision and, 151–5 election rules, 5–9 election surveys, 11–12 electoral forecasting, 142–3 electoral geography, 9, 44–5, 68–9 electoral institutions, 5–9 electoral reform, 8, 113 electoral rules, 172 embourgoisement, 42, 166 empathy, 124, 134 environment, 91–2 Europe, 42, 91–2, 93, 94, 107–11, 127, 167–8 European Union, 107 executive authority, 7 extreme left, 3, 4, 11, 16, 19, 20, 21, 28–31, 34, 35, 37, 58, 65, 76, 80–1, 83–4, 126 extreme-left candidates, 16, 60, 75, 76, 98, 106, 110–11, 121 extreme-left voters, 60, 65, 85–6

9780230_300088_13_ind.indd 228

extreme right, 4–5, 16, 18–21, 28–9, 34–9, 58–61, 65–6, 75–6, 101, 102, 108, 126–7, 166, 172 extreme-right candidates, 98, 106, 110–11, 139, 148, 154 extreme-right voters, 60, 85 farm vote, 44–6 Fauvet, Jacques, 2 Fifth Republic, 2, 4, 5, 6, 69, 107 financial assets, 43, 53–5, 57–66 first round, 164–5, 172 foreign policy, 7, 93 Fourth Republic, 3, 69 framing effect, 142 French elections, theories of, 9–12 French electoral politics, complexity of, 1 French electorate, as stalled, 169–70 French National Election Surveys (FNESs), 11–12 French presidential elections, dynamics of, 1–2 French voters female, 18, 20, 21, 24, 28, 39, 152, 166 male, 20, 21, 30–2, 36, 39, 166 perspective on, 174–6 funnel of causality, 13 gender, 15, 18, 21, 24, 28, 30, 31–8, 39, 166 generational gaps, 15 Giscard d’Estaing, Valéry, 4, 7, 44, 115, 118, 124, 158 Goguel, François, 9 Gore, Al, 142 government continuity, 1 Great Britain, 10, 42, 57, 66, 113, 124 Greens, 87, 108, 213n1 home ownership, 64 honesty, 115, 118, 119, 122, 124, 125, 134, 135 Hue, Robert, 4, 16, 98, 99, 101, 110, 111, 121, 156–8

6/28/2011 7:59:22 PM

Index

ideological identification, 11, 34, 60–2, 66–70, 167 candidate evaluation and, 126–7 left-right, 68–86 party identification and, 73–5 vote and, 75–85 immigration issue, 60, 88, 91–4, 100–7, 127, 167–8 income, 43, 53–5 incumbents, 6, 97, 105, 117, 142, 167, 175 issue attitude, 89–90 issues, 13, 70–3, 87–112 candidate evaluation and, 94, 127 domestic vs. international, 93 easy vs. hard, 93 economic, 89, 93, 95–100, 103–7, 127, 167 Europe, 107–11, 127, 167–8 immigration, 60, 88, 91–3, 100–7, 127, 167–8 law and order, 88 ranking of, 91–2 valence vs. positional, 93 issue voting, 88–95 Jospin, Lionel, 4–5, 7–8, 37, 66, 80, 94, 101–2, 122, 124, 157 labor market, 48 Laguiller, Arlette, 4, 16, 76, 94, 110, 111, 157 Lancelot, Alain, 9 late deciders, 143, 151, 152, 153 law and order, 88 leader image, 113–40, 168 components of, 116–20, 122 evaluations of, 125–35 in French electoral studies, 114–16 impact of television on, 155–60 impact on voting, 135–9 in presidential elections, 120–5 leadership qualities, 13, 117 left, 2–7, 4, 9–15, 18–19, 28–31, 34, 39–40, 45–8, 55, 58, 64–7, 67 left-right continuum, 10, 16–17, 60, 66–7

9780230_300088_13_ind.indd 229

229

left-right ideology, 20, 68–86, 165, 167 issues and, 70–3 meaning of, 69–73 party identification and, 73–5 vote choice and, 75–85 left-wing candidates, 34–5, 37, 59–60, 67, 166 left-wing voters, 39–40, 42, 51, 60, 67 legislative elections, 7–9, 113 Le Pen, Jean-Marie, 4–5, 18, 28, 36–7, 42, 60–1, 64–6, 80, 88, 100–2, 106, 121, 127, 154, 157, 172 Lipset, Seymour Martin, 41 long-term forces, 13, 17, 170–5 Louis XVI, 68 managers, 62, 63 Marx, Karl, 41, 44 Marxist theory, 45 Mayer, Nonna, 42 men, 20, 21, 30–2, 36, 39, 152, 166 Michigan model, 12–13, 17, 116, 174–5 middle class, 42, 46, 48 Mitterrand, François, 1, 4, 7, 36, 107–8, 123–6, 134, 214n1 moderates, 21, 29–30, 34–7, 60, 65, 72, 166 Mouvement Républican Populaire (MRP), 3 multi-party system, 2–5 National Assembly, 6, 7, 8–9 National Convention, 68 National Front (FN), 4–5, 7, 30, 35, 36, 60, 61, 87, 100–2, 108 Obama, Barack, 18 occupation, 44–53, 57–66 older voters, 29, 30, 38 parliament, 6 partisan cycles, 9–12 partisan groups, 2–3 partisanship, 11

6/28/2011 7:59:22 PM

230 Index

Parti socialiste unifié (PSU), 19 party identification, 11, 73–5, 167 Pasqua, Charles, 108 patrimony, 42, 43, 53–6, 166 PCF, see Socialist Party peasant class, 44–6, 231n1 Perrineau, Pascal, 8, 102 personality traits, 116–18, 134, 135, 136 personal security, 91–2 Pierce, Roy, 115, 126 political campaigns, 13, 141–62, 168–9 campaign attentiveness and, 144–51 impact of television in, 155–60, 169 time of vote decision and, 151–5 political economic incumbent, 105, 167 political ideology, see left-right ideology political institutions, 5–9 political parties, 1 competition among, 2–5 leading, 2, 3 left-right continuum, 10–11 Pompidou, Georges, 115 Popular Front, 4, 41 Portugal, 113 positional issues, 93 pre-campaign deciders, 143–4, 152, 153 president, power of, 2, 6–7 presidential candidates, see candidates presidential elections of 1974, 115, 118, 124 of 1981, 7 of 1988, 28, 47, 51, 62, 64–6, 84, 97, 99–101, 105 of 1995, 4, 35, 36, 66, 96–100, 105, 115, 118, 157 of 2002, 4–5, 7–8, 16–17, 35, 36, 66, 80, 81, 83–4, 94, 97, 100–2, 110–11, 123, 124, 134, 136, 157, 168, 172

9780230_300088_13_ind.indd 230

of 2007, 5–9, 28–31, 35, 62, 80, 97–8, 100, 102–4, 111, 123–4, 127, 134, 166, 168, 175 of 2012, 175–6 campaign effects on, 141–62, 168–9 change and continuity in, 163, 165–9 class-related determinants, 41–67, 166 double-ballot system, 164–5 electoral institutions and, 5–9 impact of television in, 155–60 leader image and, 113–40, 168 left-right ideology and, 75–85 role of issues in, 87–112 second ballot results, 1965-2007, 6 second round of, 67, 81–2, 84, 164–5, 172–3 socio-demographic variables in, 15–40, 165–6 socioeconomic variables in, 41–67 studies of, 175 theory of, 173–4 in US, 10, 18, 93, 142 presidentialization, 8–9, 113, 162, 163 presidential-parliamentary system, 6 primaries, 214n1 prime minister, 6, 7, 97, 167 priming effect, 142 private sector, 63, 64, 166 proportional representation, 7 prospective economic voting, 95 pseudo-R 2, 32, 34, 57, 58, 84, 136, 170, 171 psychological variables, 13 public employees, 47, 48, 51, 166 public-opinion surveys, 11–12 Rassemblement pour la République (RPR), 3, 19, 101 religion, 15, 16, 20–3, 27, 31–9, 165–6, 174

6/28/2011 7:59:22 PM

Index

religious belonging, 20 religious practice, 20 religious upbringing, 13 Republican party, 18 republicanism, 41 retrospective economic voting, 95 right, 10–11, 21, 28, 34, 58, 67 right-wing candidates, 30, 34, 37, 166–7, 168 right-wing voters, 39–40, 42–3, 61, 67 risky assets, 54–5, 57, 59, 64–7 Rocard, Michel, 19 Royal, Ségolène, 6, 65, 118–19, 124, 214n2 Sarkozy, Nicolas, 1, 3–4, 6, 29, 35, 36, 43, 51, 61, 65, 84, 97–8, 111, 118–19, 124, 137, 168, 213n2 schools, 91–2 second round, 67, 81–2, 84, 164–5, 172–3 Séguin, Phillipe, 108 self-employment, 48 short-term forces, 13, 17, 170–3, 174–5 Siegfried, André, 9 Single Market Program, 107 social class, see class-related determinants social divide, 38 social factors, 81, 83–4, 86, 166–7, 170 Socialist party, 4, 7–8, 10, 19, 40, 67, 69 social structure, 15–40 socio-demographics, 15–40, 165–6 age, 18, 26, 29, 31–8, 39, 40 bivariate analysis of, 20–31 education level, 19, 25, 27–40 gender, 18, 21, 24, 28, 30–9 multivariate analysis of, 31–8 religion, 20–3, 27, 31–9 socioeconomic variables, 41–67 bivariate analysis of, 48–53 economic sector, 57–66

9780230_300088_13_ind.indd 231

231

financial assets, 53–5, 57–66 income, 53–5 multivariate analysis of, 55–66 occupation, 57–66 sociological variables, 13 sociotropic economic voting, 95, 96 Strauss-Kahn, Dominique, 165 strength, 125, 140, 168 tau-b, 29, 79 television, impact of, 155–60, 169 thermometer ratings, 115, 117, 120–2, 125, 128–34, 137 time of vote decision, 151–5 transparency, 124 Trotskyist candidates, 16 Trotskyite Workers’ Struggle, 4 two-party coalition system, 4 UDF, 19, 108 UMP, see Union pour un Mouvement Populaire (UMP) unemployment, 91–4 Union pour un Mouvement Populaire (UMP), 3–4 United Kingdom, see Great Britain United States, 41–2, 66, 113, 124, 167 unity moment, 165 US elections, 10, 18, 93, 142 valence issues, 93 Vavreck, Lynn, 142 vote choice age and, 18, 26, 29, 31–8, 39, 40 analysis of, 16–17 bivariate analysis of, 20–31 campaign effects on, 141–62, 168–9 class-related determinants, 166 economic sector and, 44–53, 57–66 education and, 19, 25, 27–40 financial assets and, 53–5, 57–66 gender and, 18, 21, 24, 28, 30–9 income and, 53–5

6/28/2011 7:59:22 PM

232 Index

vote choice – continued leader image and, 113–40, 168 left-right ideology and, 75–85 long-term forces, 170–3 multivariate analysis of, 31–8 occupation and, 44–53, 57–66 patrimony and, 53–6 religion and, 20–3, 27, 31–9 role of issues in, 87–112 short-term forces, 170–3 socio-demographic determinants of, 17–40 socioeconomic variables, 41–67 vote decision, time of, 151–5

9780230_300088_13_ind.indd 232

voter attentiveness, to campaigns, 144–51 voter turnout, 4 in legislative elections, 8–9 in presidential elections, 8–9 wealth distribution, 43 women, 18, 20, 21, 24, 28, 30, 39, 152, 166 working class, 42, 46, 48, 51, 61–4, 67, 166 Wright, Gordon, 68 young voters, 29, 38, 152 Zeldin, Theodore, 41

6/28/2011 7:59:22 PM