Presidents and Parties in the Public Mind 9780226589480

Citation preview

Presidents and Parties in the Public Mind

Presidents and Parties in the Public Mind

GARY C. JACOBSON

The University of Chicago Press Chicago and London

The University of Chicago Press, Chicago 60637 The University of Chicago Press, Ltd., London © 2019 by The University of Chicago All rights reserved. No part of this book may be used or reproduced in any manner whatsoever without written permission, except in the case of brief quotations in critical articles and reviews. For more information, contact the University of Chicago Press, 1427 E. 60th St., Chicago, IL 60637. Published 2019 Printed in the United States of America 28 27 26 25 24 23 22 21 20 19

1 2 3 4 5

ISBN-13: 978-0-226-58920-6 (cloth) ISBN-13: 978-0-226-58934-3 (paper) ISBN-13: 978-0-226-58948-0 (e-book) DOI: https://doi.org/10.7208/chicago/9780226589480.001.0001 Library of Congress Cataloging-in-Publication Data Names: Jacobson, Gary C., author. Title: Presidents and parties in the public mind / Gary C. Jacobson. Description: Chicago : The University of Chicago Press, 2019. | Includes bibliographical references and index. Identifiers: LCCN 2018037061 | ISBN 9780226589206 (cloth : alk. paper) | ISBN 9780226589343 (pbk. : alk. paper) | ISBN 9780226589480 (e-book) Subjects: LCSH: Presidents—United States—Public opinion. | Political parties—United States—Public opinion. | Public opinion—United States. Classification: LCC JK2265 .J33 2019 | DDC 324.273—dc23 LC record available at https://lccn.loc.gov/2018037061 ♾ This paper meets the requirements of ANSI/NISO Z39.48–1992 (Permanence of Paper).

For Martin and Barbara Shapiro

CONTENTS

List of Tables / ix List of Figures / xiii Preface / xix ONE

T WO

/ Introduction / 1

/ The Coevolution of Affect toward Presidents and Parties / 6 THREE

/ Evaluations of Parties and Party Leaders / 32

FOUR

/ Assessments of Party Competence / 63

FIVE

SIX

SEVEN

/ Cognitive Views of Parties / 99

/ Party Identification I: Partisan Change / 122

/ Party Identification II: Generational Imprinting / 144 EIGHT

/ Elections / 163

/ Polarized Parties, the 2016 Elections, and the Early Trump Presidency / 191

NINE

TEN

/ Conclusion / 223

Appendix: Data Sources / 227 References / 231 Index / 243

TA B L E S

2.1a Change in Democratic Party affect as a function of change in Democratic candidate affect / 25 2.1b Change in Republican Party affect as a function of change in Republican candidate affect / 26 2.2 Coefficients from regressions of party affect on candidate affect over the 2008 election year / 30 3.1 Approval of the president’s job performance and favorable views of president’s party / 36 3.2 Positive views of the president and positive views of president’s party / 36 3.3 Consistent positive or negative opinions of the president and his party (percent) / 40 3.4 Consistent approval/disapproval of president’s performance and favorable/ unfavorable opinions of his party (percent) / 40 3.5 Favorable opinion of opposition party as a function of presidential approval / 43 3.6 Positive views of the president and the opposition party / 44 3.7 Approval of the president’s job performance and approval / favorable opinion of the president’s party or party leaders in Congress / 50 3.8 Ratings of the president and president’s congressional parties performance as “excellent” or “pretty good” (Harris Polls) / 51 3.9 Approval of the president’s job performance and opinion of opposition party or party leaders in Congress / 53 3.10 Ratings of the president and opposition congressional party’s performance as “excellent” or “very good” (Harris Polls) / 53 3.11 Approval of the president’s job performance and approval of parties or party leaders in Congress, controlling for congressional approval / 54

x / Tables 3.12 Ratings of the president and congressional parties performance as “excellent” or “pretty good,” controlling for quarterly congressional approval (Harris Polls) / 54 3.13 Approval of the president’s job performance and approval of parties or party leaders in Congress, controlling for congressional approval, by control of the House, G. W. Bush and Obama administrations / 55 3.14 Effects of president’s thermometer ratings on party thermometer ratings, 1978– 2016 (regression coefficients) / 56 3.15 Relationships between party and president thermometer ratings over time / 59 3.16 Party thermometer ratings as a function of the president’s thermometer scores (Democracy Corps Polls) / 60 4.1 Presidential approval and perceptions of relative party competence / 67 4.2 Estimated effects of presidential approval on individual respondents’ perceptions of relative party superiority in handling the most important problem / 69 4.3 Presidential approval and the party more trusted to cope with the nation’s problems / 70 4.4 Presidential approval and preference for president or rival congressional party in handling national problems / 72 4.5 Approval of the president’s handling of the economy and party advantage on managing the economy / 80 4.6 President’s party’s advantage over rival party in Congress in trust on handling the economy as a function of approval of president’s performance on the economy / 83 4.7 The effects of presidential approval on education on relative party reputation for handling education / 88 4.8 Approval of the president’s handling of foreign policy and party advantage on managing foreign policy, Reagan through Obama administrations / 91 4.9 Approval of G. W. Bush’s handling of terrorism and the Republican Party advantage on dealing with it / 94 4.10 Approval of president’s job performance and opinion on which party is better for / cares more about “people like you” or “the middle class” / 98 5.1 ANES questions on ideology and policy positions from the cumulative data file / 101 5.2 Regression of mean locations of parties on locations of presidents / presidential candidates across election years / 107 5.3 Regression of mean locations of parties on locations of presidents / presidential candidates across election years, by party identification / 108 5.4 Effects of president/candidate location on party location on the 7-point liberal–conservative scale (regression coefficients, individual surveys) / 109 5.5 Effects of president/candidate location on party location on the 7-point issue scales (regression coefficients, individual surveys) / 111

Tables / xi 5.6 Changes in party location on ideology and issues scales as a function of changes in the president’s location (with controls) / 113 5.7 Coefficients from regressing party locations on presidents / presidential candidates and House candidates locations on the 7-point ideology and issue scales (with controls) / 116 5.8 Coefficients from regressing placement of Democratic Party on placement of Barack Obama and Democratic congressional candidates on the 7-point liberal–conservative scale, 2010–2014 / 117 5.9 Coefficients from regressing placement of the Republican Party on placement of Mitt Romney on the 7-point liberal–conservative scale, 2012 / 118 5.10 Coefficients from regressing placements of the parties on placement of the presidential candidates, Obama, and the House and Senate candidates on the 7-point liberal–conservative scale in 2016 (with controls) / 119 6.1 The effects of presidential approval on changes in individual party identification / 130 6.2 The effects of presidential approval on changes in individual party identification (with lagged presidential approval) / 131 6.3 Presidential approval and change in party identification between panel waves (weighted averages) / 132 6.4 Change in party identification as function of change in candidate affect, 2008–2009 / 134 6.5 Presidential approval and macropartisanship / 139 7.1 Party identification of whites, 1952–2016 (percentages) / 162 8.1 Ticket splitting and party loyalty in presidential and state elections (percentages) / 173 8.2 Net generic vote for president’s party’s candidate as a function of presidential approval / 176 8.3 Net preference for president’s party’s control of congress as a function of presidential approval / 176 8.4 Net number of House seats won or lost by the president’s party, 1946–2016 / 181 8.5 Presidential approval and partisan voting in midterm elections, 1974–2014 (percent defecting to other party’s House candidate) / 185 9.1 Polarized views of parties as a function of polarized views of the president / 193 9.2 Age, education, and the presidential vote in 2016 (percent voting for Trump) / 202 9.3 Opinions of groups and policies in 2016 / 204 9.4 Favorable opinion of the Republican Party as a function of Trump’s job approval (January 2017–April 2018) / 215

FIGURES

2.1 Correlations between party identification and candidate/president affect, 1956–2004 / 12 2.2 Correlations between party identification and president / presidential candidate thermometers, 1968–2016 / 14 2.3a Correlations between party identification and favorable views of Bush, Gore, and Clinton, December 1999–January 2001 / 15 2.3b Correlations between party identification and favorable views of Bush and Kerry, October 2003–November 2004 / 15 2.3c Correlations between party identification and favorable views of Obama, McCain, H. Clinton, and Bush, December 2007–November 2008 / 16 2.4 Favorability of partisans toward Obama and McCain, December 2007 through November 2008 / 17 2.5 Correlations between party identification and thermometer ratings of Obama, H. Clinton, Bush, and McCain, 2008–2009 / 17 2.6a Effect of presidential or presidential candidate affect on party affect, controlling for party identification, Republicans / 19 2.6b Effect of presidential or presidential candidate affect on party affect, controlling for party identification, Democrats / 20 2.7 Effect of ratings of candidates/presidents on ratings of their parties, by status and party identification / 20 2.8 Effect of presidents’ / presidential candidates’ thermometer ratings on party thermometer ratings, controlling for party identification, 1978–2016 / 21 2.9 Correlations between candidate and party affect, by party identification, 2008 / 22 2.10 Relative impact of party identification and candidate affect on party affect / 23 2.11 Relative impact of candidate/president affect on party affect, controlling for party identification, January–October 2008 / 24

xiv / Figures 2.12 Relative effect of Eisenhower and Nixon affect on Republican Party affect, 1956– 1968, controlling for party identification / 27 2.13 Relative effect of Republican leaders’ affect on party affect, controlling for party identification / 28 2.14 Relative effect of Democratic president/candidate thermometer on Democratic Party thermometer, controlling for party identification / 28 2.15 Relative effect of Republican president/candidate thermometers on Republican Party thermometer, controlling for party identification / 29 3.1a Presidential approval and favorable opinions of the president’s party, Clinton, G. W. Bush, and Obama administrations / 35 3.1b Positive opinions of the president and his party, Clinton, G. W. Bush, and Obama administrations / 35 3.2 Approval of G. W. Bush’s performance and favorable opinions of the Republican Party across time / 37 3.3a Approval of Bill Clinton’s performance and favorable opinion of the Democratic Party, by party identification / 38 3.3b Approval of George W. Bush’s performance and favorable opinions of the Republican Party, by party identification / 38 3.3c Approval of Barack Obama’s performance and favorable opinions of the Democratic Party, by party identification / 39 3.4 Strength of opinions of president and favorable opinions of his party / 41 3.5 Approval of the president and Congress 1993–2016 (quarterly averages) / 46 3.6 Approval of the president’s job performance and approval of his party or party leaders in Congress / 48 3.7a Approval of Clinton and Democrats in Congress, by party identification / 48 3.7b Approval of G. W. Bush and Republicans in Congress, by party identification / 49 3.7c Approval of Obama and Democrats in Congress, by party identification / 49 3.8 Rating of president and president’s party in congress as doing an “excellent” or “pretty good” job (Harris Polls) / 50 3.9 Mean thermometer scores for the president and his party, by party identification, 1978–2016 / 58 4.1 Party superiority on handling the most important problem, Truman through Obama administrations / 64 4.2 Partisan opinions of party competence, Truman through Obama administrations / 65 4.3 Presidential approval and relative party preference on most important problem / 66 4.4 Approval of Clinton’s performance, by domain (quarterly averages) / 74

Figures / xv 4.5 Approval of G. W. Bush’s performance, by domain (quarterly averages) / 75 4.6 Approval of Obama’s performance, by domain (quarterly averages) / 76 4.7 Party advantage on handling the economy, 1981–2017 / 78 4.8 Approval on economy and party advantage on the economy / 79 4.9 Average annual Democratic advantage in dealing with health care issues / 84 4.10 Opinions on the ACA, Obama’s handling of health care, and Democratic Party superiority on health care / 86 4.11 Democratic Party advantage on handling education / 87 4.12 Party advantage on handling foreign policy, G. H. W. Bush to Obama administrations / 90 4.13 Presidential approval on foreign policy and president’s party’s advantage on foreign policy / 90 4.14 Party advantage on handling terrorism, 2001–2016 / 93 4.15 Approval of G. W. Bush’s handling of terrorism and the Republican Party’s advantage on the issue / 93 4.16 Party better able to prevent World War III / 95 4.17 Presidential approval and opinion on which party cares more about “people like you” or “the middle class” / 97 5.1 Ideological location of parties and presidential candidates, by partisans / 104 5.2 Average locations of the parties and presidents / presidential candidates on the liberal–conservative scale, 1972–2016 / 105 5.3 Location of presidents/candidates and parties on 7-point scales, 1972–2016 / 106 6.1 Party identification, 1993–2016 (leaners excluded) / 135 6.2 Party identification, 1993–2016 (leaners included) / 136 6.3 Democrats’ share of party identifiers, 1993–2016 / 137 6.4 Presidential approval and macropartisanship, Clinton, G. W. Bush, and Obama administrations (monthly averages of Gallup Polls) / 138 7.1 Party identification, by year respondent turned 20 (leaners included) / 145 7.2 Party identification, by year respondent turned 20 (leaners excluded) / 146 7.3 Party identification by president at age 20 (leaners included) / 147 7.4 Party identification by president at age 20 (no leaners) / 148 7.5 Democratic share of party identifiers by race/ethnicity and year respondent turned 20 (leaners included) / 149 7.6 Racial/ethnic composition of the respondent population by year respondent turned 20 / 150 7.7 Approval of Barack Obama’s job performance, 2009–2016 / 151

xvi / Figures 7.8 Approval of Obama’s job performance, by age group / 152 7.9 Approval of G. W. Bush’s job performance, by age group / 152 7.10 Approval of Obama’s performance by party and year respondent turned 20, 2014–2015 / 153 7.11 Ideology by year respondent turned 20 / 154 7.12 Ideology of Republicans by year respondent turned 20 / 155 7.13 Ideology of Democrats by year respondent turned 20 / 156 7.14 Ideology of pure independents by year respondent turned 20 / 157 7.15 Summary issue positions, by party ID and age group, 2014 / 157 7.16 Republicans’ opinions and media use, by age group / 158 7.17 Approval of Obama’s performance, by race/ethnicity / 159 8.1 Effect of the presidential vote on the House Vote 1952–2016 (regression and correlation coefficients) / 166 8.2 Effects of the presidential vote on the Senate Vote, 1948–2016 (regression and correlation coefficients) / 167 8.3 States and districts delivering split outcomes in presidential and congressional contests, 1952–2016 / 168 8.4 Ticket splitting in national elections, 1952–2016 / 169 8.5 Party loyalty and defection in presidential and House elections, 1952–2016 / 170 8.6 Party loyalty and defection in presidential and Senate elections, 1952–2016 / 170 8.7 Correlations between the vote for president, state assembly, US House, and party registration in California, 1968–2016 / 172 8.8 Presidential approval and generic US House vote / 175 8.9 Presidential approval and preference for party control of Congress / 175 8.10 Midterm House and Senate votes congruent with presidential approval, 1946–2014 (percent) / 187 8.11 Effects of presidential approval on the House and Senate vote, controlling for incumbency and party identification, midterm elections / 188 8.12 Effects of presidential approval on the House and Senate vote, controlling for incumbency and party identification, presidential election years / 188 8.13 Is your vote for Congress a vote for or against the president? / 189 9.1 Partisan differences in presidential approval, Eisenhower to Trump (quarterly averages) / 192 9.2 Partisan thermometer ratings of presidential candidates, 1968–2016 / 200 9.3 Correlations between racial resentment and reactions to parties and candidates / 203

Figures / xvii 9.4 Approval of Trump’s job performance, January 2017–April 2018 / 207 9.5 Approval of Obama and Trump’s job performances, by demographics (Obama’s last year, Trump’s first year) / 208 9.6 Republicans’ opinions of Trump, the Republican Party, and Republicans in Congress and their leaders / 218

P R E FAC E

The idea for this book originated in research for an earlier book on the public’s reaction to President George W. Bush’s performance and policies, particularly regarding the Iraq War. I suspected that the growing unpopularity of Bush and the war during his second term might have inflicted collateral damage on the Republican Party, and I found evidence that it did. This raised the question whether performance ratings of other presidents affected their party’s public standing, and the answer, after some additional research, was a clear yes. While pursuing that question, I discovered the wide variety of ways in which opinion surveys had, over the postwar period, sought to measure popular reactions to presidents and their parties. These studies have produced a remarkably rich trove of data for examining myriad ways in which modern presidents have influenced their party’s popularity, reputation for competence, assumed policy commitments, appeal as objects of identification, and electoral performance. I reported analyses of some of these data in a series of papers and journal articles but eventually decided that only a book would be adequate to the data and subject. The result is in your hand (or on your screen). I had intended to end the book with Barack Obama’s just-completed presidency, but Donald Trump’s disruptive election and singular early presidency raised intriguing questions about how he might be affecting his party’s public standing, so where possible I have extended my analyses to cover his election and first year as well. The book could not have been written without the help and encouragement of many colleagues and friends. My graduate students and faculty colleagues at UC San Diego were first to hear many of the arguments presented here and have been generous in their comments and critiques; the contributions of Amy Bridges, Zoli Hajnal, Seth Hill, Sam Kernell, Thad Kousser, Rick Kronick, Megumi Naoi, Simeon Nichter, Sam Popkin, and David Weins

xx / Preface

deserve special recognition. Beyond my department, I’ve benefited from comments from the participants in conference panels where some of this material was first presented, with special nods to Jon Bond, Larry Bartels, Hans Noel, John Sides, Steve Rogers, Lynn Vavreck, and John Zaller. My largest collegial debt, however, is to George W. Edwards III, who as editor of Presidential Studies Quarterly encouraged me to submit and agreed to publish some of my initial findings there. I am also grateful to Chuck Myers, Holly Smith, and Ruth Goring at the University of Chicago Press and to my indexer, Meg Wallace, for their diligent and proficient work in seeing this project to its completion. My most profound obligation, however, is to all of those involved in gathering and disseminating the countless surveys that have gone into making this book. The gatherers include a long list of academic and commercial survey researchers whose contributions are cited throughout the following pages. Prominent among them are the venerable Gallup Poll, with useful surveys stretching back to Harry Truman’s presidency; several decades of polls sponsored by the major media firms; and the high-quality surveys conducted in recent years by academic and charitable institutions. The American National Election Studies (both the time series covering every presidential election since 1952 and several specialty studies) and the newer comprehensive academic election studies, notably, since 2006, the Cooperative Congressional Election Study, have also been essential to the project. Far too many people have contributed importantly to these studies for me to name them all here, but I am especially grateful to the ANES’s principal investigators, from Warren Miller and his Michigan colleagues to today’s Vincent Hutchings, Shanto Iyengar, and Ted Brady, as well as to the outstanding roster of scholars who have served on the ANES board of overseers over the years, for keeping the enterprise going so well for so long. I also owe special thanks to talented people behind the CCES, created and led by Steve Ansolabehere and Brian Schaffner and executed by Doug Rivers and an exceptional staff, notably including Ashley Grosse, Samantha Luks, and Liz Salazar. I am also deeply obliged to the people and institutions that have made the data accessible to scholars. Foremost is the Roper Center for Public Opinion Research, formerly at the University of Connecticut under Paul Herrnson’s direction, now at Cornell University under Peter Enns’s leadership. The Roper archives are the source of a large majority of the surveys I examine in the following chapters. The Pew Research Center has also been generous in disseminating its collection of high-quality surveys. I am particularly obliged to Michael Dimock, the Pew Center’s former director of

Preface / xxi

polling and current president, not only for his help in acquiring data but also, as my graduate student many years ago, for introducing me to the graphic capabilities of Excel. I also owe PollingReport.com for my daily fix of marginals on the wide range of political questions I routinely track, including many analyzed here. Last but not at all least, I gratefully acknowledge the contribution of the millions of ordinary citizens who have agreed over the decades to participate in political surveys. Their usually unsung and unrewarded acts of civic engagement have been indispensable to the work of monitoring, explaining, and evaluating modern democracy in America. Finally, I take great pleasure in dedicating this book to Martin and Barbara Shapiro, mentors and friends since my very first undergraduate foray into political science more than fifty years ago. Their inspiration and example put me on a career path that has become only more rewarding over the ensuing years. Some of the research reported here appeared in journals listed below, and I appreciate the permissions they grant to their authors to include in later work material first published in their pages: “The Effects of the George W. Bush Presidency on Partisan Attitudes.” Presidential Studies Quarterly 39:2 (June 2009): 172–209. ©2009 Center for the Study of the Presidency and Congress. “The President’s Effect on Partisan Attitudes.” Presidential Studies Quarterly 42:4 (December 2012): 683–718. ©2012 Center for the Study of the Presidency and Congress. “How Presidents Shape Their Party’s Reputation and Prospects: New Evidence.” Presidential Studies Quarterly 45:1 (March 2015): 1–28. ©2015 Center for the Study of the Presidency and Congress. “The Coevolution of Affect toward Presidents and Their Parties.” Presidential Studies Quarterly 46:2 (June 2016): 1–29. ©2016 Center for the Study of the Presidency and Congress. “The Effects of the Early Trump Presidency on Public Attitudes toward the Republican Party.” Presidential Studies Quarterly, forthcoming. ©2018 Center for the Study of the Presidency and Congress. “The Obama Legacy and the Future of Partisan Conflict: Demographic Change and Generational Imprinting.” Annals of the American Academy of Political and Social Science 667 (September 2016): 72–91. ©by the American Academy of Political and Social Science. “Polarization, Gridlock, and Presidential Campaign Politics in 2016.” Annals of the American Academy of Political and Social Science 667 (September 2016): 226–46. ©by the American Academy of Political and Social Science. “The Triumph of Polarized Partisanship: Donald Trump’s Improbable Victory in 2016.” Political Science Quarterly 132:1 (2017): 1–34. ©2017 by the Academy of Political Science. “Donald Trump, the Public, and Congress: The First Seven Months.” Forum 15:3 (October 2017): 525–45. ISSN (Online) 1540-8884, DOI: https://doi.org/10.1515/for-2017 -0034. ©2017 Walter de Gruyter GmbH Berlin/Boston.

ONE

Introduction

The reciprocal bonds between American presidents and their political parties are nearly as old as the republic. The national parties developed to compete for the presidency—the Democratic Party’s roots go back to the campaign to elect Thomas Jefferson in 1800—and, by fielding slates of pledged electors, reduced the electoral college to an accounting device, democratizing the choice of president and strengthening the executive’s hand in the constitutional system. During the first decades of the nineteenth century, and inspired in good part by pursuit of the presidency, suffrage and national party organization expanded together. Party entrepreneurs, assembled at national conventions, sought to nominate candidates who could excite and mobilize a growing mass electorate. The party-printed ballots of the time inhibited ticket splitting, and the fate of a party’s down-ballot candidates depended heavily on the drawing power of the top of the ticket. Once a president was elected, his chief duty was, except in times of crisis, to dispense the patronage jobs and other spoils of office that helped keep the party in business. Progressive-era reforms at the end of the century led eventually to fundamental changes in party organization (by undermining patronage-based party machines) and electoral processes (by introducing primary elections and the Australian ballot), but pursuit of the White House remained the focus and purpose of national party activity, and success or failure at the top continued to echo down the ticket. Moreover, the dramatic expansion of the executive branch in the twentieth century that gave presidents a vastly greater role in national government also gave them a vastly enlarged public presence, magnifying their role as their party’s preeminent public voice and face. Modern presidents have thus had a profound and pervasive impact on how the public views their parties, or so I argue in this book, which

2 / Chapter One

examines the influence of every postwar president on popular opinions and beliefs about his party. That influence, I find, has been pervasive. The president’s words and actions articulate and define his party’s current principles and objectives. Judgments about the president’s competence in managing domestic and foreign affairs inform assessments of the party’s competence in such matters. The components of a president’s supporting coalition, and the interests he favors while governing, help to define the party’s constituent social base and thus its appeal as an object of individual identification. People’s affective reactions to the president, whatever their source, inevitably color their feelings about the other politicians in his coalition. Every president thus shapes public attitudes toward his party as well as beliefs about who and what it stands for and how well it governs when in office. Insofar as the party label represents a brand name, the president bears prime responsibility for the brand’s current image and status. The most proximate and concrete consequences for the president’s party register when the ballots are counted on election day, but presidents also affect their parties’ fortunes in the longer run, for their successes or failures influence partisan attachments, identities, and images that remain long after they are gone from the scene. The idea that modern presidents (and presidential candidates) have been prime movers in shaping the national parties is of course neither new nor controversial. Scholars have produced a rich literature on how they have constructed or reconstructed the parties’ electoral coalitions, dealt with national and local party organizations, sought to influence their congressional parties, and managed the executive branch to serve partisan goals (for example, Milkis 1993; Galvin 2010; Smith and Seltzer 2015; Skowronek 1997; Kriner and Reeves 2015; not to mention the major accounts of each presidential campaign and presidency that cover such matters). This literature typically treats presidential candidates and presidents as strategic actors pursuing their own career and policy goals within (or battling) the constraints imposed by existing national conditions and other institutional players, including their party coalitions. It focuses on the president’s dealings with party elites in and out of government, with the broader public as a secondary if ultimately crucial audience; its main concern is with parties as institutions. My concern here, in contrast, is with the ways in which presidents affect how that public audience comes to regard those institutions. As we shall see, the president’s influence on popular beliefs and attitudes toward the parties occurs quite independently of, for example, his deliberate partybuilding activities or lack thereof (Galvin 2010) or position in “political time” (Skowronek 1997). It also occurs whether the president intends it or

Introduction / 3

not. All presidents cultivate public support, and if they are successful, their party benefits. But public support and a pleasing image for their party is not always a priority, and even when it is, much of what shapes both is beyond the president’s control. Presidents do not assume office with a blank partisan canvas, of course. Existing partisan biases and party images shape reactions to future presidents as soon as they arrive on the national stage and continue to exert a powerful influence on evaluations of their performance throughout their time in office. For each administration, however, these initial reactions are the starting point, a product of the existing configurations of public attitudes toward and beliefs about parties, revised or reinforced during the president’s successful campaign, that are subsequently updated in response to developments during the president’s years of service. My aim in this book is to flesh out and defend these various claims by examining a vast and diverse set of survey and other data on presidential candidates, presidents, and their parties that have accumulated over the past seven decades. I begin in the next chapter by analyzing how affective reactions to parties and presidents evolve together over the course of electoral and presidential careers—before, during, and after their presidencies. Because analyses of survey data are central to this and all subsequent chapters, I also review theoretical ideas about how individuals respond to surveys to establish a framework for understanding the ubiquitous links between popular thinking about presidents and parties documented in this and the following chapters. The third chapter focuses on how the public’s general assessments of a president’s performance in office shape feelings about their parties generally and their congressional parties and leaders in particular. It documents powerful presidential effects on partisans of all persuasions, across every administration examined, and for every available measure of president and party affect. The chapter also explores the influence of a president’s job evaluations on the rival party’s public standing, which turns out to be very modest except under special conditions. Chapter 4 examines how opinions of the president’s performance have influenced judgments of relative party competence in dealing with the nation’s most important problem in every administration from Harry Truman’s to Barack Obama’s. It also examines how the president’s perceived success or failure in handling specific policy areas has shaped his party’s reputation for managing them. And it evaluates the president’s contribution to and occasional modification of “issue ownership”—a party’s enduring reputation for superiority in a particular policy domain.

4 / Chapter One

In the fifth chapter I explore in detail how perceptions of presidents and presidential candidates’ ideological and policy positions shape perceptions of their parties’ positions on these same dimensions. Their influence is, I find, pervasive and remarkably consistent over time, across dimensions, and among partisan subgroups. I also find that ordinary Democrats and Republicans have developed quite divergent notions of where the Democratic Party and its presidents lie on the left–right spectrum. Regardless of partisanship, however, presidents strongly and consistently influence where people locate their parties in ideological and policy space. The sixth and seventh chapters examine the president’s impact on party identification. In chapter 6 I show that presidential performance ratings over the short term have a systematic impact on individual and mass partisanship, but that when reactions to a president shift individual party identification or the aggregate partisan balance out of its long-term equilibrium, the change is temporary and reversible. Still, the effect is potentially quite consequential for the president’s party depending on its location in the electoral cycle. Chapter 7 looks at presidential effects on mass partisanship over the longer term, assessing the direction and durability of each of the postwar presidents’ influence on party fortunes through a process of generational imprinting. Durable presidential effects appear as systematic variations across generations in aggregate partisanship that reflect the successes or failures of the administrations during which successive age cohorts entered the electorate. I combine data from several hundred Gallup surveys to compose a snapshot of the political generations active political during the Obama administration that reveals the impact of past presidential “shocks” and bears implications for the future distribution of political identities and attitudes. The data document the emergence of a decisive Democratic advantage among cohorts entering the electorate during the presidencies of Clinton, G. W. Bush, and Obama, attributable to both demographic and ideological differences between younger and older citizens. Chapter 8 examines the ultimate bottom line: how popular evaluations of presidents and presidential candidates affect their parties’ electoral prospects as measured between elections by the generic vote intentions as well as, most importantly, their actual fortunes in presidential and midterm election years. I find that American electoral politics has become increasingly nationalized and president-centered over the past several decades, elevating the president’s electoral importance to his party to levels not seen since the heyday of ballot-induced straight-ticket voting in the nineteenth century. Chapter 9 considers recent presidents’ contributions to arguably the

Introduction / 5

most important development in American national politics over the last several decades, the increasing partisan polarization of national politics. It analyzes Barack Obama’s contribution to this process in the context of analyzing the 2016 election and the sources of Donald Trump’s victory. It also speculates about Trump’s potential for amplifying the widening demographic and cultural differences between ordinary Republicans and Democrats, further polarizing the public. It reviews Trump’s first year in office and considers, in light of the findings reported in previous chapters and initial data from his presidency, what public reactions to Trump and his performance portend for the future public standing of his party. The concluding chapter briefly summarizes the book’s central findings.

T WO

The Coevolution of Affect toward Presidents and Parties

The central concern of this book is the diverse ways in which presidents have shaped the public standing, reputation, and electoral success of their parties during the postwar era. But attitudes toward prospective presidents and their parties are already closely entwined by the time they take office. Most Americans hold preformed opinions of the parties that guide reactions to presidential candidates, among them all future presidents, as soon as they have arrived on the national stage tagged with a party label. And in the course of winning nominations and presenting themselves and their ideas to the public, prospective presidents have already begun to influence how people feel and think about their parties. In this chapter I examine the ways in which attitudes toward presidents and their parties come to be linked in the public mind before they are elected, during their time in office, and after they are no longer on the ballot and new candidates emerge to take up their party’s banner. Constrained by the available data, my focus here is on the affective component of political evaluations: how partisanship influences expressed feelings about presidential candidates and presidents, and how feelings about these leaders in turn influence feelings about their parties, over the course of electoral careers. To anticipate, I find clear and consistent evidence that links between party and candidate affect are comparatively modest at first, increase sharply during the initial campaign for the White House for both candidates, tighten further when the winner seeks reelection, and then atrophy once the president leaves office. In this sequence, the causal arrow runs first from party affect (manifest in party identification) to candidate affect but gradually shifts direction to run from candidate or president affect to party affect (and party identification). I also find that the relative impact on party affect of the president and prospective or actual successors within the same party varies

The Coevolution of Affect toward Presidents and Parties / 7

over time in a way that underlines the president’s leading role in shaping popular feelings about his party.

Theoretical Considerations The links between people’s expressed opinions and beliefs about parties, presidential candidates, and presidents explored in this chapter and throughout this book are measured by how they respond to questions posed by survey researchers. It is thus essential to consider the mental processes that generate individuals’ survey responses. The constructivist prospective provides clearest framework for interpreting the patterns that emerge from the data. According to Zaller and Feldman’s well-known model, most citizens do not hold preformed attitudes that they simply reveal when questioned in a survey; “rather they carry around in their heads a mix of only partially consistent ideas and considerations. When questioned, they call to mind a sample of these ideas, including an oversample of ideas made salient by the questionnaire and other recent events, and use them to choose among the options offered. But their choices do not, in most cases, reflect anything that can be described as true attitudes; rather, they reflect the thoughts that are most accessible in memory at the moment of response” (1992, 579–80). Lodge and Taber accept this process with the proviso that the stored considerations that go into constructing expressed opinions are inherently affect-laden: “all political leaders, groups, issues, symbols, and ideas thought about and evaluated in the past become affectively tagged—positively, negatively or both” (2013, 43). The “affective tags are attached when an object (person, group, place, event, issue, or abstract concept) is first evaluated and strengthened with each replication. . . . After but one or two evaluations, a concept is ‘hot,’ affectively charged” (53). A running tally, “representing an automatic integration of all prior evaluations of the object, is then restored to long-term memory where it is readily available for subsequent evaluations” (51). Thus “the evaluations a citizen might report in an opinion poll . . . reflect the integration of thoughts and feelings associated with . . . conscious and unconscious political evaluations. Immediately and without intentional control, a perceived candidate, issue, group, or idea is classified as either good or bad” (44). In this process, political objects and their associated cognitions are inevitably subject to motivated reasoning (to Lodge and Taber, “all reasoning is motivated” [emphasis theirs, 2013, 150]). “Affective experience feeds back on the appraisal process itself, routinely promoting a selective search of long-term memory for belief-confirming evidence and thereby informing

8 / Chapter Two

the conscious expression of opinion, choice, and action in affectivelycongruent ways” (207). Motivated reasoning occurs only with motivation, of course, which is readily supplied by being questioned in a survey that, for example, primes partisan political considerations: “priming is the key mechanism guiding the activation process as these associations [between thoughts, feelings, and beliefs]—weighted by one’s experience—become the considerations that spontaneously inform both conscious and unconscious judgment and choice” (207). This perspective provides a basis for understanding a host of surveybased findings reported throughout this book. For the present chapter, it suggests several hypotheses regarding how the relationships between opinions about presidential candidates and presidents and opinions about their parties should develop and vary over time. First, expressed opinions of political figures will be conditioned by partisan priors. Party identification, the single most consistent organizer of political attitudes and behavior, should be especially potent if it is indeed a central component of an individual’s social and personal identity, as the classic Michigan model and its more recent defenders posit (Campbell et al. 1960; Green, Palmquist, and Schickler 2002). Insofar as people identify with a party—which a large majority of American do1—they will reflexively evaluate politicians with a known party affiliation from a biased perspective (Miller and Shanks 1996). (I will have much more to say about party identification in chapters 6 and 7.) Second, however, the link between opinions of parties and politicians should vary in strength and direction with salience, repetition, and context. The relationship should be weaker for the more obscure and peripheral political figures than for those who become presidential candidates or presidents, because the simultaneous activation of affectively tagged considerations pertaining to the political figures and parties will be less common. For the same reason, the links should also be relatively weaker for future presidential candidates in the years before they win the nomination. Moreover, the direction of influence should be primarily from party identification, which draws on comparatively dense networks of affectively tagged memories, to affect toward the political figure, thinly if at all embedded in long-term memory. The relationship should be much stronger for presidential nominees and presidents, and stronger still in the campaign context, which primes and reinforces partisan sentiments across the board (Erikson and Wlezien 2012). 1. The attitudes and behavior of people who declare themselves independent but lean toward a party clearly identify them as closet partisans (Keith et al, 1992).

The Coevolution of Affect toward Presidents and Parties / 9

Once well on the road to nomination, presidential candidates not only will be evaluated in more explicitly partisan terms but will also begin to have a reciprocal effect on how people feel about their parties. The party convention obviously puts a big official partisan stamp on the nominee, but by its choice (as well as in its formal platform and convention rhetoric) the party also begins to redefine itself, occasionally rather sharply, as did, for example, the Republicans with Barry Goldwater in 1964 and Donald Trump in 2016 or the Democrats with George McGovern in 1972 and Bill Clinton in 1992. The nominee becomes the party’s dominant public face; what people come to think of him or her during the course of the campaign will inevitably color their feelings about the party that did the nominating. Campaigns continually reinforce partisan thinking. As Lodge and Taber put it, first abstractly and then concretely, “with repeated evaluations an affective tag is linked to a concept and springs to mind spontaneously upon mere exposure to the associated object, thereby signaling the concept’s affective coloration. By election eve, most citizens will have formed impressions of the major candidates, parties, and issues and these feelings will be inescapably activated on their mere mention” (2013, 43). Moreover, “affective reacting . . . systematically guides the encoding, search, retrieval, interpretation, and evaluation of information in ways that promote affectively congruent rationalization effects” (2013, 206). Feelings about candidates and parties should thus reach their maximum convergence nearest to the election season. Third, the relationship between affect toward a party and affect toward a political leader should weaken as the leader’s political centrality fades. The linkage should remain strong for successful presidential candidates throughout most of their time in office (the research findings reported in subsequent chapters thoroughly document this reality). As a president’s second term draws to a close, however, the focus of partisan evaluations will begin to shift to the party’s new nominee, and once presidents are out of office, the affectively loaded political memories people draw on when evaluating them and their parties will surface together less frequently. The same holds to an even greater degree for unsuccessful former presidential candidates, now largely out of the public eye. New political figures and events supplant the old in “top-of-the-head” considerations people use to construct their evaluations of parties in response to pollsters’ inquiries.

The Data I draw data to test these conjectures from four sets of surveys that collectively (though not individually) span the years 1956 through 2016:

10 / Chapter Two Gallup Polls. At irregular intervals beginning in the 1950s, the Gallup Poll asked respondents to rate the two political parties and various political figures using what Gallup called a “scalometer.” This was the question format: Here is an interesting experiment (HAND RESPONDENT SCALOMETER). You notice that the 10 boxes on this card go from the HIGHEST POSITION OF PLUS 5—or someone you like very much—all the way down to the LOWEST POSITION OF MINUS 5—or someone you dislike very much. Please tell me how far up or how far down the scale you would rate the following: [PARTY OR LEADER’S NAME]. Responses were then recoded to take values from 1 (highest rating) to 10 (lowest rating), with 11 representing the “don’t knows” (omitted from the analyses reported here.)2 Scalometer ratings of both parties and of varying sets of leaders were asked in nineteen surveys between 1956 and 1991; all presidential and vice presidential candidates between 1956 and 1984 received at least some coverage, as did a variety of potential nominees, former presidents, and former candidates. Another forty-four surveys gathered scalometer ratings of one or more political figures but omitted the parties; the coverage goes up to 2004 but is much thinner after 1990. These surveys also included Gallup’s standard party identification question. For consistency, I treat it as a 3-point scale, with independents who lean toward a party still considered independents; the leanings of independents were not queried in many of the earlier surveys and are thus unavailable for analysis.3 ANES Time Series. The American National Election Studies (ANES) surveys from 1978 through 2016 (omitting midterms after 1998) have asked respondents to rate the two parties, presidents, presidential candidates, and other political figures on a “feeling thermometer.” This is the question format: I’d like to get your feelings toward some of our political leaders and other people who are in the news these days. I’ll read the name of a person and I’d like you to rate that person using something we call the feeling thermometer. Ratings between 50 degrees and 100 degrees mean that you feel favorable and warm toward the person. Ratings between 0 degrees and 50 degrees mean that you don’t feel favorable toward the person and that you don’t care too much for that person.

2. The proportion of “don’t knows” regarding the parties, presidents, and presidential candidates in these data is usually below 10 percent. 3. The Gallup surveys are all from the Roper Center’s archive; when individual surveys are cited, they are tagged with the date of the last day the survey was in the field.

The Coevolution of Affect toward Presidents and Parties / 11 You would rate the person at the 50 degree mark if you don’t feel particularly warm or cold toward the person. The ANES surveys also include the battery of questions that produce the standard 7-point party identification measure used in analyses of these data.4 ANES Panel Study 2008–2009. The ANES conducted a separate panel study of the 2008 election and its aftermath. In six of the panel’s waves (January, February, June, September. and October 2008 and July 2009), respondents were asked a set of questions prefaced by “We’d like to ask how much you like or dislike various people and groups who have been in the news recently.” The questions took a branching format, first asking whether the respondent liked or disliked the party or politician, then asking whether they liked [disliked] them “a great deal,” “a moderate amount, or “a little.” The responses generated a 7-point scale ranging from liking a great deal to disliking a great deal. The standard ANES party ID question was also ask in the January, September, and October 2008 and July 2009 waves. The 2000, 2004, and 2008 NAES Surveys. The National Annenberg Election Survey (NAES) conducted rolling cross-sectional surveys on a nearly daily basis over the course of the 2000, 2004, and 2008 election years. These surveys did not ask about the parties, but they elicited favorability ratings of candidates and other political figures and also asked the questions that produce the standard 7-point party identification scale, so they offer insight into how the relationship between party ID and affect toward political figures changes over the course of a campaign. In the 2000 survey, affect was measured by a scale ranging from 0 to 100 (from least to most favorable); in 2004 and 2008, the favorability scale ranged from 0 to 10. For the analyses reported here, the rolling cross-sections were aggregated by month.5

Party Identification and Affect toward Presidents and Candidates over Time How do partisan attitudes influence affective evaluations of eventual presidential candidates and presidents from the time they become prominent enough to attract the attention of pollsters through their electoral careers and beyond? According to the first hypothesis proposed in the previous section, partisan differences should appear the very first time polls ask voters to evaluate a prospective presidential candidate, and they should grow as 4. See the appendix for full citations of the ANES time series and panel surveys. 5. See the appendix for full citations for the NAES studies.

12 / Chapter Two

Figure 2.1. Correlations between party identification and candidate/president affect, 1956–2004 Note: Number of correlations averaged for the two columns is in parentheses. Source: Gallup Polls.

the candidates become more widely known (and hence more accessible to and associated with the party in working memory). The correlation between party identification and affect toward candidates should thus increase over the course of campaigns. Evidence from all four data sources supports these conjectures. Figure 2.1 displays the average correlations between party identification and candidate rating on Gallup’s scalometer for twenty presidential candidates and presidents between 1956 and 2004 at various points in their electoral careers. Candidates are distinguished by whether they were relative newcomers6 or veterans (presidents, former candidates and vice presidents).7 Those appearing in the surveys after they were no longer candidates or in office were also distinguished by whether they were defeated (including defeated presidents) or not.8 The sequential patterns are straightforward. Correlations between partisanship and affect for prospective candidates newly on the scene are posi-

6. At one time, John F. Kennedy, Lyndon Johnson, Goldwater, McGovern, Ronald Reagan, Jimmy Carter, G. H. W. Bush, Michael Dukakis, Clinton, G. W. Bush, and John Kerry were newcomers. 7. Dwight Eisenhower, Adlai Stevenson, Richard Nixon, Johnson, Hubert Humphrey, Gerald Ford, Walter Mondale, Carter, Reagan, G. H. W. Bush, B. Clinton, and G. W. Bush. 8. The former include Herbert Hoover, Tom Dewey, Stevenson, Goldwater, Humphrey, McGovern, Ford, Carter, and G. H. W. Bush; the latter include Harry Truman, Eisenhower, Nixon, Johnson, and B. Clinton.

The Coevolution of Affect toward Presidents and Parties / 13

tive but comparatively modest prior to the election year, when they begin to increase; after Labor Day, they are indistinguishable from those of veteran candidates.9 Correlations begin considerably higher for veterans and remain, on average, stable until the post–Labor Day period, when they jump to their highest average levels. The average correlation for sitting presidents in years when they are not involved in pursuing reelection is somewhat lower, and it is lower still once they have left office. For losing candidates, including losing presidents, the postelection correlations are lower still. These patterns hold for most candidates individually; for example, for thirteen of the seventeen, the correlation reaches its highest level in the survey taken closest to the election day.10 A similar pattern appears in data from the ANES election-year time series (figure 2.2). Correlations between party identification and candidate/president feeling thermometer ratings are lowest in surveys taken prior to the election year in which the candidate first wins the nomination, especially among newcomers,11 rise to a high point during the initial election contest, fall a bit in years when the president is not on the ballot, and rise again to a high when the president runs for reelection. Correlations are smaller in later election years when the leader is no longer a candidate, especially following defeats. Again, these patterns are generally repeated for individual leaders; for every one of the ten who appear as a candidate in only one election contest, the correlation is highest in the survey for that year; for the seven who appear twice, it is highest in the second survey for six.12 It is important to note that the correlations between party identification and candidate thermometers have increased steadily over the period cov-

9. Party ID is coded 1 if the respondent identifies with the candidate/president’s party, 0 if independent, and –1 if with the rival party; thus coefficients for both party’s politicians are expected to be positive. 10. Hoover, Truman and Dewey were in surveys only after they had been candidates or presidents. I also computed average correlations for several presidential prospects who were never nominated but appeared on multiple surveys. For George Romney, it was .16 (N=6); for Nelson Rockefeller, .13 (N=8); for Robert Kennedy, it was .31 (N=6), and for his brother Ted Kennedy, it was .33 (N=12). Thus party ID influenced affect toward the first two as if they were newcomers (which of course they were), whereas it influenced affect toward the Kennedys as if they were veterans, results that make intuitive sense from the constructivist perspective outlined in section I. 11. In these data, newcomers are McGovern, Reagan, Mondale, G. W. Bush, Kerry, and John McCain in years before the initial candidacies; the veterans are Nixon, Ford, Carter, Mondale, G. H. W. Bush, B. Clinton, Bob Dole, and G. W. Bush as presidents or vice presidents; Hillary Clinton is also treated as a veteran after 1992. Unlike in the Gallup data, the correlations for newcomers and veterans are indistinguishable in their initial contests for the presidency. 12. The exception is G. H. W. Bush, for which the 1988 correlation was .57, the 1992 correlation .54.

14 / Chapter Two

Figure 2.2. Correlations between party identification and president/presidential candidate thermometers, 1968–2016 Note: The number of correlations averaged is in parentheses. Source: ANES Cumulative Data File and ANES 2016 Time Series Study (face-to-face sample).

ered by the data, growing from an average of .36 in the 1970s, to .40 in the 1980s, to .46 in the 1990s, to .56 in the 2000s, and to .65 in the 2010s.13 The correlations for every subcategory of candidate also grew significantly over time, except among newcomers prior to their first campaign. This is the first of many findings presented in this book documenting the growing partisan coherence in the American electorate over the years between Nixon and Obama. The data in figures 2.1 and 2.2 confirm that the election context strengthens the links between partisanship and affect toward political leaders, especially for those who are not already familiar to most citizens. A more detailed picture of how this occurs during election years is provided by the NAES rolling cross-sectional studies of the 2000, 2004, and 2008 elections. In 2000 (figure 2.3a), the correlation between party identification and candidate affect rose during the campaign for both Al Gore and George W. Bush; it was lower for Bush, the newcomer, at the beginning but soon caught up, equaling Gore’s correlation in November, the high point for both candidates as well as for the departing president, Bill Clinton. In 2004 (figure 2.3b), the correlation for John Kerry was relatively low until after the Iowa caucuses in January, which made him the clear frontrunner, after which it 13. Decades defined by reapportionment cycle—e.g., the 1990s are 1992–2000; I include 1968 in the 1970s.

Figure 2.3a. Correlations between party identification and favorable views of Bush, Gore, and Clinton, December 1999–January 2001 Source: NAES 2000.

Figure 2.3b. Correlations between party identification and favorable views of Bush and Kerry, October 2003–November 2004 Source: NAES 2004.

16 / Chapter Two

Figure 2.3c. Correlations between party identification and favorable views of Obama, McCain, H. Clinton, and Bush, December 2007–November 2008 Source: NAES 2008.

increased sharply and continued to rise until October, when it reached its highest point. The correlation was high for the incumbent Bush all along but still increased a bit to reach its peak in November. In 2008 (figure 2.3c), favorability toward newcomers Barack Obama and John McCain was not strongly related to partisanship prior to the election year. The relationship was particularly weak for McCain, who was initially viewed relatively favorably by Democrats as well as Republicans (figure 2.4). That changed over the course of the campaign, with the correlations increasing rather steadily for both candidates to reach their highest levels in November. As in 2004, the correlation for Bush began high but still increased a bit during the election year. The correlation for Hillary Clinton, in contrast, was high early on, when she was the frontrunner (Jacobson 2009b) and had been in the national eye for years, but weakened somewhat as Obama moved ahead of her in the polls and became the party’s nominee and thus principal public face. The growing strength of the connection between partisanship and candidate affect reflected partisans’ increasingly polarized opinions of the contestants (figure 2.4). The same was true for 2000 and 2004; presidential campaigns are inevitably polarizing events. The ANES 2008–2009 Panel Study tells the same story (figure 2.5). The correlations between party identification and thermometer ratings of

Figure 2.4. Favorability of partisans toward Obama and McCain, December 2007– November 2008 Source: NAES 2008.

Figure 2.5. Correlations between party identification and thermometer ratings of Obama, H. Clinton, Bush, and McCain, 2008–2009 Source: ANES Panel 2008–2009.

18 / Chapter Two

Obama and McCain began at comparatively low levels but grew similarly over the course of the campaign. The initial correlations for Bush and Clinton began a high level; Bush’s stayed there but Clinton’s diminished slightly once she had lost the nomination.14 This study also included waves taken after the election in May and July 2009. The correlations for ex-president Bush, Obama, and Clinton, who joined the Obama administration as secretary of state, remained high months after the election; but the correlation for McCain had by May 2009 dropped back to its February 2008 level, largely because ordinary Democrats had reverted to a more positive view of the senator once he was no longer the Republican standard-bearer.

Candidates, Presidents, and Affect toward their Parties The data examined so far leave no doubt that the partisan biases arising from identification with a party increasingly shape affective responses to prospective and actual presidential candidates as their involvement in electoral politics deepens; it also appears that party influence wanes when their time in office or as candidates has passed. This section investigates how such variations in affect toward candidates are linked to affective evaluations of their parties. Three of the data sets examined in the previous section also provide measures of affect toward the parties on the same scales used for the political figures. In every instance the party ratings are of course strongly influenced by partisanship, but they are by no means entirely the product of partisan bias. The correlations are usually substantial but never overwhelming; typically (e.g., 93 percent of the time) the correlation between party ID and party affect falls between .45 and .70.15 Plenty of variance in people’s affect toward the parties thus remains to be explained by something other than party identification, including of course affect toward the parties’ titular leaders. As with the influence of party identification on candidate affect—and in the same manner and for the same reasons—the influence of candidate/ president affect on party affect should vary systematically over the course of a political career. I explored this possibility using the Gallup data by regressing ratings of the parties (on the 10-point scale) on the ratings of the po14. These data again show widening partisan differences in evaluations of candidates over the election year. 15. In the Gallup series the mean correlation between party identification and party ratings is .52, with a standard deviation of .07 and range of .33 to .63 (N=38); in the ANES cumulative series the respective figures are .62, .07, and .47 to .74 (N=30); in the ANES 2008–2009 panel they are .66, .03, and .62 to .70 (N=10).

The Coevolution of Affect toward Presidents and Parties / 19

litical figures (on the same scale) in each survey, controlling for party identification. The results are summarized in figures 2.6a and 2.6b, which display the mean coefficients from regressing party affect on politicians’ ratings at varying stages of their careers. Several patterns stand out. First, the average regression coefficient is invariably larger in years when the politician is a candidate for national office, often substantially so. Second, coefficients are smallest prior to the first candidacy, and second smallest after the political career is over. The average coefficient for sitting presidents in nonelection years is lower than when they are on the ballot but still quite substantial. Richard Nixon’s career pattern is particularly revealing. His influence on party affect peaks with his 1960 candidacy, drops off when he is out of office, then once again reaches a high point during his 1972 reelection campaign, before falling off again when he is no longer president. Every career sequence depicted in these charts supports the idea that partisan political figures’ position in the electoral career cycle determines the degree to which citizens’ feelings about them influence feelings about their parties. These patterns are summarized in figure 2.7, which displays the mean regression coefficients for political figures according to stage of electoral career (controlling for party identification) plus mean coefficients for each partisan subgroup separately. The context has a large effect on the extent to which candidate/president affect influences party affect. Its

Figure 2.6a. Effect of presidential or presidential candidate affect on party affect, controlling for party identification, Republicans Note: The number of surveys averaged is in parentheses. Source: Gallup Polls.

Figure 2.6b. Effect of presidential or presidential candidate affect on party affect, controlling for party identification, Democrats Note: The number of surveys averaged is in parentheses. Source: Gallup Polls.

Figure 2.7. Effect of ratings of candidates/presidents on ratings of their parties, by status and party identification Note: The number of Gallup surveys used to compute the average regression coefficient is in parentheses; the coefficients for all respondents are estimated with controls for party identification.

The Coevolution of Affect toward Presidents and Parties / 21

Figure 2.8. Effect of presidents / presidential candidates’ thermometer ratings on party thermometer ratings, controlling for party identification, 1978–2016 Note: The number of coefficients averaged is in parentheses. Source: ANES cumulative data file and ANES 2016 time series (face-to-face sample).

influence grows as the election approaches, peaking after Labor Day. The effect for sitting presidents is about the same as that registered early in the election year, and it shrinks notably when they are no longer in office. The coefficients for partisans of all stripes shift similarly in response to the context, but the effects are almost always substantially larger for opposition partisans than for the candidate/president’s partisans.16 That is, opposition partisans’ feelings toward a party are considerably more sensitive to their feelings about its leaders than are those of the leader’s partisans. The ANES time series offers a second perspective on these relationships. The mean coefficient from regressing party thermometer on candidate thermometer, controlling for party identification, appears in figure 2.8. Again we see that the relationship between candidate/president affect and party affect is lowest in years before candidates win the nomination and after they are no longer running. The coefficients are much larger in years when they are on the ballot and largest when presidents are seeking reelection. In years when the president is in office but not on the ballot, the coefficient is slightly smaller. Partisans again respond similarly to the electoral context, with the party affect among the rival party’s identifiers again being on average more sensitive to variations in candidate/president affect, although 16. This is true of 89 percent (63/71) of the individual comparisons from these surveys.

22 / Chapter Two

Figure 2.9. Correlations between candidate and party affect, by party identification, 2008 Source: ANES 2008–2009 Panel Study.

the differences are smaller than in the Gallup data, particularly when the candidate is on the ballot (results not shown). The ANES 2008–2009 panel study offers a more detailed account of how the link between candidate and party affect tightens as candidates pursue and win the nomination. At the simplest level, figure 2.9 displays the increasing correlation between the two among partisans of all stripes from over the course of the 2008 election year. The increase is especially striking among Democrats, rising from .22 to .62 for Obama and the Democratic Party, and from .28 to .59 (after peaking at .64 in September) for McCain and the Republican Party. Over the same time, the relative impact of party identification and candidate affect on party affect shifted sharply as well. The partial regression coefficients displayed in figure 2.10 indicate that party identification dominated party affect at the beginning of the election year but over the course of the campaign gradually ceded that dominance to presidential candidate affect. Observe also that the new balance remained in place after Obama’s first seven months in office (he of course remained his party’s most prominent figure by far). Even as the relationship between party identification and candidate affect tightens (figure 2.5), then, feelings about the candidates have an increasingly powerful independent effect on feelings about their parties. The ANES panel also documents a shift in the relative impact on party affect of the presidential candidates compared to that of initially more

The Coevolution of Affect toward Presidents and Parties / 23

Figure 2.10. Relative impact of party identification and candidate affect on party affect, January 2008-July 2009 Source: ANES Panel 2008–2009.

prominent leaders in their respective parties (figure 2.11). Of the two major Democratic competitors for the 2008 nomination, Hillary Clinton was much the better known at the beginning of the campaign, and in January, respondents’ feelings about her had a notably larger influence on feelings about the Democratic Party than did feelings about Obama (party identification controlled). By June that was no longer the case, as feelings about Obama had become much more influential. A similar pattern occurred on the Republican side; initially, affect toward President Bush had a much greater influence on affect toward his party, but after McCain had become the party’s choice, he supplanted Bush as the greater influence. The panel structure of this study opens yet another window for observing the influence of candidate affect on party affect by allowing examination of how feelings about the parties changed in response to changes in feelings about the candidates over the course of the campaign. Tables 2.1a and 2.1b report the results of regressing changes in party ratings between pairs of waves on changes in candidate ratings, controlling for lagged candidate and party ratings and party identification.17 In every case, changes in candidate 17. The equations are estimated with robust standard errors and the cross-sectional weights applying to the second of the two waves examined. The affect variables are from the 7-point scale described in section II. Party identification is the standard 7-point scale.

24 / Chapter Two

Figure 2.11. Relative impact of candidate/president affect on party affect, controlling for party identification, January–October 2008 Source: ANES Panel 2008–2009.

affect have a substantial effect on changes in party affect; lagged candidate affect is almost as influential. The middle column in the second row of equations in each table shows the results for net changes between the first (January) and last (October) wave of interviews during the election year.18 The coefficients indicate that change in affect toward Obama had a greater effect on changes in affect toward the Democratic Party than was the case for McCain and his party over the full course of the election year. Estimates from the additional survey wave taken in May 2009 show that changes in feelings about the Democratic Party echoed changes in feelings about Obama after he took office as well as during the campaign. Tables 2.1a and 2.1b also include equations comparing the influence of Obama and Clinton on changes in feelings about the Democratic Party, and McCain and Bush on changes in feelings about the Republican Party, over the election year (the data on Bush run through the October wave; Clinton 18. There was a fair amount of net change in both directions in affect toward the candidates and parties, with more change for the former than the latter. For Obama, 42 percent reported no net change, 26 percent a one-step change, 18 percent a two-step change, and 15 percent a change of three or more steps. The respective figures for McCain were 39 percent, 26 percent, 23 percent, and 12 percent. The respective figures for the Democratic Party were 54 percent, 21 percent, 16 percent, and 9 percent; for the Republican Party, 53 percent, 20 percent, 19 percent, and 8 percent. Thus feelings about the candidates were more volatile than feelings about the parties. For both candidates and both parties, more changes were in the negative than in the positive direction.

The Coevolution of Affect toward Presidents and Parties / 25 Table 2.1a. Change in Democratic Party affect as a function of change in Democratic candidate affect Δ Democratic Party affect January–February 2008 Δ Obama affect Obama affect, t–1 Democratic Party affect, t–1 Party identification, t Constant Adjusted R2 N

June–September 2008

.20*** .16***

.04 .03

.31*** .27***

.04 .04

.31*** .28***

.04 .04

–.46*** .22*** .09* .25 1374

.04 .03 .04

–.55*** .21*** .11*** .33 1229

.04 .03 .03

–.50*** .14*** –.41*** .31 1275

.04 .03 .04

September– October 2008 Δ Obama affect Obama affect, t–1 Democratic Party affect, t–1 Party identification, t Constant Adjusted R2 N

February–June 2008

January–October 2008

October 2008– July 2009

.31*** .23***

.03 .03

.43*** .39***

.04 .04

.39*** .35***

.03 .03

–.54*** .19*** –.51*** .31 2453

.04 .02 .07

–.73*** .22*** –.58*** .48 1367

.04 .04 .11

–.68*** .21*** –.81*** .45 1275

.03 .02 .07

January–September 2008 Δ Obama affect Δ Clinton affect Obama affect, t–1 Clinton affect, t–1 Democratic Party affect, t–1 Party identification, t Constant Adjusted R2 N

.33*** .10** .36*** .09**

.04 .03 .03 .03

–.68*** .19*** .52*** .43 1356

.04 .03 .04

Note: First entry in each column is the regression coefficient; the second entry is the standard error. Source: ANES Panel 2008–2009. *p