The Culture Wars & Political Polarization In

THE CULTURE WARS & POLITICAL POLARIZATION IN PERSPECTIVE: WHY POLARIZATION AND ITS PERTURBATIONS ARE A PERSISTANT PUZZLE IN POLITICAL SCIENCE

______

A Dissertation

presented to

the Faculty of the Graduate School

at the University of Missouri-Columbia

______

In Partial Fulfillment

of the Requirements for the Degree

Doctor of Philosophy

______

DONALD MICHAEL GOOCH

Dr. John R. Petrocik, Dissertation Supervisor

DECEMBER 2009

The undersigned, appointed by the dean of the Graduate School, have examined the [thesis or dissertation] entitled

THE CULTURE WARS & POLITICAL POLARIZATION IN PERSPECTIVE: WHY POLARIZATION AND ITS PERTURBATIONS ARE A PERSISTANT PUZZLE IN POLITICAL SCIENCE presented by Donald Michael Gooch, a candidate for the degree of doctor of philosophy,

and hereby certify that, in their opinion, it is worthy of acceptance.

Professor John R. Petrocik

Professor James Endersby

Professor Marvin Overby

Professor Jay Dow

Professor Paul Speckman

To my

Mom & Dad

Everything I am I owe to you.

ACKNOWLEDGEMENTS

I would first like to thank my dissertation advisor, Professor Petrocik, for his tireless efforts on my behalf and the innumerable helpful suggestions and patient edits that have made this dissertation what it is today. I would like to thank Professor James Endersby for his service as co-chair of my dissertation committee. I would further like to extend my gratitude to the entire dissertation committee: Dr. Marvin Overby, Dr. Jay Dow, and Dr. Paul Speckman. This dissertation would have been impossible without them. I would also like to thank Mr. Ray

Bacon and Ms. Marie Concannon for their above-and-beyond aid in navigating the Byzantine labyrinth that is SAS code and the variety of data sources I needed to access for this dissertation.

Their patience and expertise was essential for completing this project. Finally, I would like to thank William Gooch and Sally Gooch who, in addition to serving in the roles of father and mother, doubled as editors, research assistants, and counselors in seeing me through to the end. I cannot express in words my appreciation for all of your efforts.

TABLE OF CONTENTS ACKNOWLEDGEMENTS ...... II LIST OF FIGURES...... VIII LIST OF TABLES ...... XI ABSTRACT ...... VI CHAPTER 1: THE CONVENTIONAL WISDOM ON POLITICAL POLARIZATION AND THE CULTURE WARS ...... 1 Discovering the Culture War ...... 5 Political Geography: A War of Red vs. Blue or a Nation of Olive Gardens? ...... 10

Political Polarization:Trends in the Body Politic vs. Regional and Reference Group Trends ...... 10

Polarized Elections & Nat. Issue Trends: Local Landslides vs the Continental Consensus 13

Godless Highbrow Bluebloods vs. Devout Country Rednecks: Income, Region & Issues ...... 14

The Conventional Wisdom on Political Polarization ...... 18

2: THE CULTURE WAR MYTHOLOGY: CRITICS OF THE CONVENTIONAL WISDOM ...... 20 Nothing to See Here! Fiorina and the Culture War Myth ...... 20 Red vs. Blue? Polarization as an Electoral Phenomenon ...... 23 Mass Polarization: Distribution Trends in the Electorate’s Attitudes & Valence vs. Policy .... 24 Dude, Where’s my Polarization? The DiMaggio Study on Political Polarization ...... 26 Wagging the Dog: Elites hold the Median Voter’s Leash and do all the Barking ...... 28 What’s the Matter with Thomas Frank? ...... 32 Blowback: Critics take on Fiorina’s Criticism of Polarization ...... 38 Skewering the Conventional Wisdom ...... 40 3: RIGHTS & WRONGS & CULTURE WARS: TOWARDS AN UNCONVENTIONAL WISDOM ON POLARIZATION ...... 41

Section I: A Pox on All Their Houses: What is Wrong with the Polarization Literature ...... 41 Purple Politics: Getting Beyond Red vs. Blue ...... 41 The 50:50 Nation & Polarizing “Close” Elections: Being Right for the Wrong Reasons .. 44 Size [Density] Matters: Dispersion across a Distribution vs. Average Location ...... 46 Sorting v. Polarization: Squares v. Rectangles ...... 47 Elites & Masses: The Paradox of Elite Polarization ...... 49

iii

Section II: Polarization as a Conceptual Problem: Formal & Empirical Foundations...... 52 Polarization Nuts & Bolts: Necessary and Sufficient Components ...... 53 Political Polarization: From Consensus to Conflict ...... 54 Political Polarization Requires Partisan Conflict ...... 61 Section III: Measures of Political Polarization ...... 64 Dynamic and the Static: Polarization vs. Polarized ...... 69 Measures of Political Polarization: An Empirical Assessment...... 69 4: EMPIRICAL MEASURES OF POLARIZATION ...... 76

Section1: Measures of Consensus, Conflict, Bimodality, and Dispersion ...... 77 Method: Principles, Measures, & Expectations ...... 77 Mean, Dispersion, and Bimodality Measures of Polarization ...... 81 Issue Dimension Measures for Party Likes and Dislikes & Mass Perceived National Problems ...... 87

Section 2: Measures of Group Polarization...... 91 Models ...... 92 Analysis & Measures ...... 92 5: FROM CONSENSUS TO CONFLICT – GAY RIGHTS AND THE CULTURE WARS ...... 96 The Growing Conflict over Gays and Gay Rights ...... 96 Does Emerging Centrism on Gay Rights Prove the Culture War a Myth ...... 98 Calm Seas – The Commanding Consensus, 1970 – 1988 ...... 99 Sowing the Wind: An Emergent Social Conflict on Gay Rights, 1988 – 1991 ...... 102 Reaping the Whirlwind: The Culture War on Gay Rights, 1991 – Present ...... 103 Data Sources & Variables – Public Attitudes towards Gay Rights ...... 105 Data ...... 105 Variables ...... 108 Analysis – Aggregate Polarization Trends in Public Opinion on Gay Rights ...... 111 Conflict and Consensus Models on Specific Gay Rights Issues, 1971 – 2007 ...... 113 Exploring the Path from Consensus to Conflict on Gay Rights ...... 119 A Comparative Analysis of Gay Rights Attitude Polarization on Select Measures ...... 123 Polarization Trends in Gay Rights Attitudes, GSS 1973 – 2006 ...... 126 Analysis – Partisan Polarization Trends in Public Opinion on Gay Rights ...... 131 Conclusion ...... 141

6: MULTIDIMENSIONAL POLARIZED POLITICS: SOCIAL ISSUES, THE ECONOMY, & FOREIGN POLICY ...... 143 Issue Polarization – ‘You can Have My Gun When You Pry It from My Cold, Dead Hands’ .. 143 Data ...... 145 Variables ...... 146 Social Conflict and Multivariate Political Polarization ...... 148 Political Polarization and Attitudes towards Groups ...... 149 Partisan Groups ...... 152 Economic Groups ...... 152 Defense / Foreign Policy Group – the Military ...... 153 Social & Religious Groups ...... 153 Trends in Mass Public Opinion on Issue Dimensions ...... 154 Jobs and Defense Spending Trend Models ...... 157 Ideological and Government Spending Trend Models ...... 159 Social Issue Trends ...... 160 Conclusion ...... 162 7: DETERMINANTS OF POLARIZATION ...... 166 Causal Factors of Public Opinion Polarization: Partisanship, Political Events, and Public Policy ...... 169 Ideology Dispersion & Bimodality: Consistent Ideological Polarization of the American Public ...... 172 Government Spending Bimodality & Dispersion: Mixed Evidence of Political Polarization . 176 Defense Spending / Foreign Policy Bimodality – A Rational Public Reacts to the Real World ...... 182 Social Issue Trends: Polarized on Minorities, Polarizing on Abortion, Depolarized on Women ...... 209 Women in Society: Depolarization, “The Year of the Woman,” & the Separability of Abortion ...... 211 The Rise of Conflict over Minority Rights: The Polarizing Effect of Affirmative Action .. 215 Conclusion ...... 219 8: POLARIZATION IN ISSUE DIMENSIONS OF PARTISAN AFFECT AND PERCIEVED NATIONAL PROBLEMS ...... 223 Issue Salience and the Culture Wars ...... 223 Analysis: Rising Attention to Social Issues in the Mass Public ...... 226

Methods & Models ...... 226 Discussion – Increasing Social Issue Salience ...... 227 Conclusion...... 237 9: GROUP POLARIZATION OF PUBLIC ATTITUDES I: PARTISAN POLARIZATION ...... 238 Partisan Polarization on Ideology: Heterogeneous or Homogenous Parties? ...... 240 Data ...... 242 Variables ...... 243 Data Presentation Organization ...... 244 Mass Partisan Polarization on Ideology – Decomposition and Contribution to Group Polarization ...... 247 Partisan Polarization: The Death of Compromise & the Paradox of Failed “Mandates” for Change ...... 258 Partisan Polarization on Abortion ...... 269 An Example of Partisan Polarization on Abortion and Public Policy: The PBA Ban ...... 277 Conclusion...... 281 10: Group Polarization of Public Attitudes II: Class & Religiosity Polarization ...... 283 Social Groups: Organizing and Driving Political Polarization...... 283 Data ...... 287 Variables ...... 288 Data Presentation Organization ...... 289 Are Social Issues the Opiate of the Working Class?...... 290 Class Polarization on Partisanship ...... 290 Class Polarization on the Ideological Dimension ...... 300 Class Conclusions ...... 306 Religiosity, Secularization and the Culture Wars...... 307 Religiosity Polarization on Partisanship ...... 309 Religiosity Polarization on Ideology ...... 317 Conclusion: The Culture Wars are a Crisis of Faith in Christ, Not Marx ...... 324 11: PERCIEVED POLITICAL POLARIZATION: PARTIES, CANDIDATES, AND THE MASS PUBLIC .... 325 Trends in Attitudes on Issues: Perceptions of Distance from Parties and Candidates ...... 325 Data ...... 327 Variables ...... 327

Expectations ...... 329 Measures ...... 330 Ideological Distance Measures ...... 334 Abortion Distance Measures ...... 336 Jobs Distance Measures ...... 339 Mass Partisan Perceptions of the Distance from Elite Parties and Candidates ...... 341 Conclusion ...... 347 12: WHO IS WAGGING WHOM? RESPONSIVE PUBLICS, RECEPTIVE ELITES ...... 350 Models of Polarization ...... 363 Aggregate Polarization ...... 363 Polarization Causes: Mass vs. Elite ...... 364 Polarization Causes: Constituents, Partisans, & Partisan Constituents ...... 364 Measures of Polarization ...... 365 Polarization Hypotheses ...... 366 Data Sources, Variables, & Methods ...... 367 Data Sources ...... 367 Variables ...... 369 Findings & Analysis: Elite Responsiveness to Mass Ideological Movement ...... 374 Ideological Polarization over Time ...... 374 Partisan Polarization over Time ...... 376 Mass Æ Elite vs. Elite Æ Mass Polarization ...... 378 Constituents vs. Representative Ideological Differences over Time ...... 383 Conclusion ...... 385 APPENDIX A: STATES BY PRESIDENTIAL VOTING, MEDIAN HOUSEHOLD INCOME, & GROSS STATE PRODUCT ...... 388 B: GRD – FREQUENCY TABLE OF POLLS ON GAYS AND GAY RIGHTS, 1971-2007 ...... 389 C: GRD – QUESTIONS BY YEAR ...... 392 D: GRD – ISSUE CATEGORY DESCRIPTIONS ...... 393 E: CHAPTER 4 GRAPHED POLL QUESTIONS ...... 395 F: GSS TABLE – PUBLIC OPINION ON HOMOSEXUAL RELATIONS ...... 398 G: ANES CUMULATIVE FILE VARIABLES INCLUDED IN ANALYSIS, 1964-2004 ...... 399

vii

H: COLLAPSED CATEGORIES FOR OPEN-ENDED RESPONSES ...... 409 I: SELECTED CHAPTER 5 ISSUE DISTANCE MEASURES ...... 411 BIBLIOGRAPHY ...... 413 VITA ...... 426

viii

LIST OF FIGURES Figure Page

1.1: Red v. Blue Map for the 2000, 2004, & 2008 U.S. Presidential Elections ...... 2-4 1.1A: 2000 Election Map ...... 2 1.1B: 2004 Election Map ...... 3 1.1C: 2008 Election Map ...... 4 1.2: Frequency of Social Issue Mentions in the New York Times – 1969-2008 ...... 7 1.3: Frequency of Political Polarization Mentions in the New York Times – 1980-2007 ...... 8 1.4: Red & Blue State Mentions in the New York Times – 2000-2007 ...... 12 1.5: Red vs. Blue Communities: 2008 Presidential Election Results by County ...... 15 2.1: Polarized Elites? Are Elites Located Distant from Median Voter and the Mass of Voters? ...... 31 3.1: Red, Blue & Purple – State Coded Composition, by Party, of the 110th U.S. Senate ...... 43 3.2: Hypothetical Policy Consensus ...... 56 3.3: Hypothetical Normal Policy Distribution ...... 57 3.4: Hypothetical Bi-Modal Policy Polarization ...... 58 3.5: Hypothetical Policy Multiple & Uniform Modal Non-Consensus ...... 59 3.6: Consensus to Conflict – Intra-Policy or Inter-Policy Polarization Over Time ...... 62 3.7A: Various Distributions (1-6) of Attribute X, N = 100 ...... 72 3.7B: Various Distributions (7-9) of Attribute X, N = 100 ...... 73 4.2: Creating a Measure of Consensus ...... 79 4.2A: Four Binary Variables ...... 79 4.2B: Variable Categories Relative to 50 percent Centerpoint ...... 79 4.2C: Absolute Value of Category Deviation from 50 ...... 79 4.2D: Combined Consensus Measure ...... 79 5.1: 1977 Harris Survey: What jobs should homosexuals be allowed to hold? ...... 100 5.2: Gay Marriage in the 50 States and United States Territories ...... 112 5.3: Polarized Attitudes on Gay Rights, 2004 ...... 121 5.3A: Gay Marriage Constitutional Amendment ...... 121 5.3B: Gay Adoption ...... 121 5.3C: Legality of Gay Marriage...... 121 5.3D: Gay Rights ...... 121

5.4: Hypothetical “Central” Distribution on Gay Rights ...... 122 5.5: Consensus Distribution on the Unacceptability of Homosexual Relations, 1974 ...... 122 5.6: GSS – Homosexual Relations are Wrong, 1973-2006, by Party Identification...... 133 5.7: GSS – Homosexual Relations are Wrong, 1973-2006, Rep/Dem Difference ...... 133 5.8: GSS – Homosexual Relations are Not Wrong, 1973-2006, Strong Rep/Dem ...... 135 5.9: GSS – Homosexual Relations are Not Wrong, 1973-2006, Weak Rep/Dem ...... 135 5.10: GSS – Homosexual Relations are Not Wrong, 1973-2006, Ind. Leaners...... 136 5.11: GSS – Homosexual Relations are Not Wrong, 1973-2006, Independents ...... 136 5.12: Which Party is Closest to Respondent on Gay Marriage, 2005 ...... 141 7.1: Dispersion Trend in Mass Ideology, 1972-2008 ...... 172 7.2: Bimodality Trend in Mass Ideology, 1972-2008...... 173 7.3: Bimodality Trend in Public Opinion on Government Spending & Services, 1972-2008 ..... 175 7.4: U.S. Government Spending as Percentage of GDP, Fiscal Years 1970-2010 ...... 177 7.5: Bimodality Trend (-Z Score) Actual Spending % GDP (Z-Score), Public Opinion on Spending & Services (Z-Score), 1980-2008 ...... 178 7.6: Bimodality Trend in Mass Opinion on Defense Spending, 1980-2008 ...... 182 7.7: Defense Spending as a Percentage of Gross Domestic Product, 1980-2003 ...... 184 7.8: George W. Bush Job Approval Rating (Gallup / USA Today Poll), February, 2001 – December, 2008 ...... 192 7.9: Bimodality Trend & Defense Spending Trend Expressed as Z Scores, 1980-2008 ...... 194 7.10 Bimodality Trend (-Z Score) & Defense Spending (Z Score) in War & Peace, 1980 – 2008 ...... 194 7.11: Bimodality Trend (-Z Score) Actual Defense Spending % GDP (Z-Score), Public Opinion on Defense Spending (Z-Score), 1980-2008...... 196 7.12: Bimodality Trends in Distrib. of Mass Opinion on Social Issues, 1972-2008 ...... 208 7.13: Bimodality Trend in Mass Opinon on Abortion, 1972-2008 ...... 212 7.14: Dispersion Trend in Mass Opinion on Equality for Women, 1972-2008 ...... 213 7.15: Dispersion Trend in Mass Opinion on Aid to Blacks, 1970-2008 ...... 215 7.16: Public Opinion on Affirmative Action by Race, 1978...... 218 7.17: Public Opinion on Affirmative Action by Race, 2005 ...... 218 8.1: ANES Open-ended Partisan Likes & Dislikes Trend on Social Issues, 1976-1998 ...... 229 8.2: Response Trends on Government, Social, Economic Isssues and Foreign Policy, 1976-1998 ...... 230

9.1: Party ID Group Polarization on Ideology over Time, 1972-2004 ...... 250 9.2: Decomposition of Partisan Group Polarization on Ideo. over Time, 1972-2004 ...... 250 9.3: Unweighted Mean Deviation Trends in Avg. Ideology for Partisan Identifers ...... 255 9.4: Bipartisan Policy Compromise Between Ideologically Moderate Parties ...... 261 9.5: Intractable Policy Conflict between Ideologically Polarized Parties ...... 263 9.6: Partisan Polarization in Congress, 1954-2004 ...... 265 9.7: Trends in Z-Standardized Partisan Polarization at the Mass and Elite Levels, 1972- 2004 ...... 266 9.8: Party ID Group Polarization on Abortion over Time, 1972-2004 ...... 271 9.9: Decomposition of Partisan G.P. on Abortion over Time, 1972-2004 ...... 271 9.10: 2003 ABC News Opinion Poll – Public Opinion on Abortion by Situation ...... 278 10.1: Income Group Polarization on Partisanship over Time, 1970-2004 ...... 292 10.2: Unweighted Mean Deviation Trends in Avg. Party ID for Rich & Poor Citizens ...... 296 10.3: Weighted & Unweighted Mean Dev. of Avg. Party ID for Secular Citizens ...... 313 10.4: Weighted Mean Deviation of Avg. Party ID for Religious & Secular Citizens ...... 313 12.1: Avg. Partisan House & Senate 1st Dimension DW-Nominate Scores, 1954-04 ...... 352 12.2: Partisan Polarization in Congress, 1954-2004 ...... 357 12.3: Fiorina’s Elite Polarization with Centrist Electorate ...... 359 12.4: Rational Polarization Given Shifts in Constituencies ...... 361

LIST OF TABLES Table Page

1.1: Assessing the Problems for Future Presidents ...... 10 2.1: Partisan “Sorting” with No Ideological “Polarization”...... 28 2.2: Income Polarization at the State Level by Presidential Voting 1996-2004 ...... 37 3.1: Illustration of Polarization Not Reflected in Average Positions in a Hypothetical Population...... 46 3.2: Hypothetical Example of Partisan Polarization on Abortion ...... 67 3.3: Group Size Weights for Hypothetical Partisan Groups...... 67 3.4: Hypothetical Partisan Individual Group Polarization Scores ...... 68 4.1: Issue Dimension Variable Definitions Created from ANES Open-Ended Response Sets ...... 89-90 5.1: Gay Rights and Homosexual Political Issues by Question Type 1971-2007...... 110 5.2: Conflict Trend Models for Gay Rights Issues 1971-2007 ...... 114 5.3: Consensus Trend Models for Gay Rights Issues 1971-2007 ...... 118 5.4: Dispersion & Bimodality Statistics for Selected Gay Rights Polls & Hypothetical Distribution ...... 124 5.5: GSS – Public Opinion on Homosexual Relations 1973 – 2006 ...... 127 5.6: Trends in Polarization of Attitudes on Homosexual Relations 1973 – 2006 ...... 129 5.7: Party means & S.D. for Public Attitudes on Homosexual Relations 1973-2006 ...... 138 5.8: Partisan Trends in Public Attitudes on Homosexual Relations 1971-2007...... 139 6.1: Trend Regressions of Public Attitudes on Issue Dimension-Related Groups ( ) ...... 150

6.2: Trend Regressions of Public Attitudes on Issue Dimension-Related Groups (SD)ࢄഥ ...... 151 6.3: Simple Statistics on Means, SD’s, & Kurtosis for Issue Self-Placements for Time Series ...... 155 6.4: Polarization Trend Models for Multiple Issue Dimensions ...... 158 6.5: Polarization Trend Models for the Social Issue Dimension ...... 160 6.6: Dimensional Polarization and Polarization Trends ...... 163 7.1: Deviation Models Regressing Gov’t Spending Levels (%GDP) on Gov’t Spending Public Opinion ...... 180 7.2: Deviation Models Regressing Defense Spending Levels (%GDP) on Def. Spending Public Opinion ...... 199 7.3: Coding Scheme for Foreign Policy Period & Presidential Party Variable ...... 201

xii

7.4: D.M. Regressing Defense Spending & Party of Presidential Admin on D.S. Public Opinion ...... 202 7.5: D.M. Regressing Defense Spending, Party & Foreign Policy Period on D.S. Public Opinion ...... 204 7.6: Regressing Defense Spending, Party & FPP on D.S. Public Opinion w/ Interact...... 206 8.1: Univariate Statistics for ANES Open Responses – Partisan Likes & Dislikes & National Problems ...... 227 8.2: OLS with RSS – Time Trends in ANES Open-Ended Responses – Partisan Likes & Dislikes ...... 232 8.3: Poisson Regression of Time Trends in ANES Partisan Likes & Dislikes ...... 232 8.4: OLS with RSS – Time Trends in ANES Open-Ended Responses – Important National Problems ...... 235 8.3: Poisson Regression of Time Trends in ANES Important National Problems ...... 235 9.1: Party ID – Decomposition of Partisan Group Polarization on Ideology, 1972-04 ...... 248 9.2: Trend Regressions of Party ID Group Polarization on Ideology, 1972-2004 ...... 251 9.3: Percent Contribution & Mean Deviation Weighted Partisan G.P. on Ideology, 1972 - 2004 ...... 253 9.4: Percent Contribution & Mean Deviation Unweighted Partisan G.P. on Ideology, 1972 – 2004 ...... 253 9.5: Weighted & Unweighted Mean Deviation Trend Regressions of PID on Ideology, 1972 – 2004 ...... 256 9.6: Congressional Models of Elite Party Differences & Mass Partisan Polarization, 1972 – 2004 ...... 267 9.7: Party ID – Decomposition of Partisan G.P. on Abortion, 1972-2004 ...... 269 9.8: Trend Regressions of Party ID G.P. on Abortion, 1972-2004 ...... 273 9.9: Percent Contribution & Mean Deviation Weighted Partisan G.P. on Abortion, 1972 - 2004 ...... 274 9.10: Percent Contribution & Mean Deviation Unweighted Partisan G.P. on Abortion, 1972 – 2004 ...... 274 9.11: Weighted & Unweighted Mean Deviation Trend Regressions of PID on Abortion, 1972 – 2004 ...... 277 9.12: 2005 Congressional Votes on Partial Birth Abortion Act ...... 279 10.1: Income Classes – Decomposition of G.P. on Party ID, 1970-2004 ...... 292 10.2: Trend Regressions of Income Group Polarization on Party ID, 1970-2004 ...... 293 10.3: Class – Weighted Percent Contribution & Mean Deviation on GP on Party ID, 1970 – 2004 ...... 295

xiii

10.4: Class – Unweighted Percent Contribution & Mean Deviation on GP on Party ID, 1970 – 2004 ...... 295 10.5: Weighted & Unweighted Mean Dev. Trend Regressions of Class GP on Party ID, 1970 – 2004 ...... 297 10.6: Income Classes – Decomposition of G.P. on Ideology, 1972-2004 ...... 299 10.7: Trend Regressions of Income Group Polarization on Ideology, 1972-2004 ...... 301 10.8: Class – Weighted Percent Contribution & Mean Deviation on GP on Ideology, 1972 – 2004...... 303 10.9: Class – Unweighted Percent Contribution & Mean Deviation on GP on Ideology, 1972 – 2004...... 303 10.10: Weighted & Unweighted Mean Dev. Trend Regressions of Class GP on Ideolgy, 1972 – 2004 ...... 305 10.11: Religiosity – Decomposition of Group Polarization on Party ID, 1970-2004 ...... 309 10.12: Trend Regressions of Religiosity Group Polarization on Party ID, 1970-2004 ...... 311 10.13: Religiosity – Weighted % Contribution & Mean Deviation on GP on Party ID, 1970 – 2004...... 303 10.14: Religiosity – Unweighted % Contribution & Mean Deviation on GP on Party ID, 1970 – 2004 ...... 303 10.15: Weighted & Unweighted Mean Dev. Trend Regressions of Religiosity GP on Party ID, 1970 – 2004 ...... 305 10.16: Religiosity – Decomposition of Group Polarization on Ideology, 1972-2004 ...... 318 10.17: Trend Regressions of Religiosity Group Polarization on Ideology, 1972-2004 ...... 319 10.18: Religiosity – Weighted % Contribution & Mean Deviation on GP on Ideology, 1972 – 2004...... 303 10.19: Religiosity – Unweighted % Contribution & Mean Deviation on GP on Ideology, 1972 – 2004 ...... 303 10.20: Weighted & Unweighted Mean Dev. Trend Regressions of Religiosity GP on Ideology, 1972 – 2004 ...... 305 11.1: Trend Models for Perceived Distance Measures on Ideology ...... 333 11.2: Abortion Distance Measures – Difference of Means Tests b/w 1980 & 1992, 1996, & 2000 ...... 337 11.3: Trend Models for Perceived Distance Measures on Jobs ...... 339 11.4: Total Partisan Distance Measures for Party Identifiers ...... 342 11.5: Total Partisan Distance Measures for Party Identifiers ...... 343 11.6: Republican and Democratic Party Distances from Party Identifiers ...... 345 12.1: Selected Means & Standard Deviations ...... 370

xiv

12.2: Ideology, Conservative FT’s, and Nominate Regression Models by Year for Full Sample ...... 375 12.3: Models Regressing Differences b/w Republican & Dem Identifiers by Year ...... 377 12.4: Mass Public Respondents & Elite Legislators – Simple & Autoregressive Models of Ideology ...... 380 12.5: Mass Party Identifiers & Party Elite (Legislators) Simple & Autoregressive Models of Partisan Ideology ...... 382 12.6: Models Regressing Squared Difference b/w Normalized Constituent Ideology and Normalized Legislator Nominate Score by Year ...... 384

THE CULTURE WARS & POLITICAL POLARIZATION IN PERSPECTIVE: WHY POLARIZATION AND ITS PERTURBATIONS ARE A PERSISTANT PUZZLE IN POLITICAL SCIENCE

Donald M. Gooch

Dr. John R. Petrocik, Dissertation Supervisor

ABSTRACT

Political polarization in the American electorate has received a great deal of attention in recent years with most of the research focusing on social issues and their impact on electoral outcomes. However, scant attention has been paid to polarization on other issue dimensions and the mass-elite affects on polarization. I develop several empirical measures of political polarization: variance to assess the spread of opinion and kurtosis as a measure of bimodality. I assess polarization using ANES and

GSS cumulative data from 1970 – 2008 on several prominent social issues such as abortion, women’s equality, affirmative action, and non-social issues such as government jobs programs and defense spending. I examine public opinion polarization as well as the relationship between mass public opinion and elite public opinion using D-W nominate scores as a measure of elite opinion. I find there is significant polarization of social and non-social policy opinion, both in terms of the average public- preferred level of government action in these areas and the conflict over it. Contrary to the consensus in the literature, I find that elite opinion is responsive to mass opinion and that there is a recursive relationship between mass and elite ideology.

xvi

The Culture Wars & Political Polarization in Perspective: Why Polarization and its Perturbations are a Persistent Puzzle in Political Science

CHAPTER 1: THE CONVENTIONAL WISDOM ON POLITICAL POLARIZATION AND THE CULTURE WARS

My friends, this election is about much more than who gets what. It is about who we are. It is about what we believe. It is about what we stand for as Americans. There is a religious war going on in our country for the soul of America. It is a cultural war, as critical to the kind of nation we will one day be as was the Cold War itself. And in that struggle for the soul of America, Clinton & Clinton are on the other side, and George Bush is on our side. And so, we have to come home, and stand beside him.

- Patrick J. Buchanan, 1992 Republican National Convention

Thus spake Patrick Buchanan in the wake of a failed but strong challenge to a sitting

Republican president in the 1992 election. Drawing an uncompromising line-in-the-sand, he sounded a clarion call to rally social conservatives to man the front lines of the culture war

(Buchanan 1992). In Buchanan’s war—or as it as more popularly known, the culture war— partisans square off over divisive social issues such as abortion, gun rights, gay marriage, religion in the public square, and prayer in schools. The partisan dimension of Buchanan’s message is readily apparent in his speech. For Buchanan, this religiosity-centered political conflict is marked by the clear battle lines drawn between Republicans on the side of the angels and

Democrats dealing with the Devil (Buchanan 1992). This struggle for the “soul of America” in a culture war isn’t merely a throwaway line among the many to be found in modern political convention speeches. Pat Buchanan’s gauntlet throwing at the 1992 Republican National

Convention reflects a powerful idea: there is a distinct and deep moral cleavage in American society marked by significant partisan divisions on a host of social issues. An idea that has engaged scholars, journalists, politicians, activists, and citizens in describing and explaining the rise of alternative media, the results of national and local elections, and the tone of political debate from the halls of Congress to the barstools in local drinking establishments. The Red vs.

Figure 1.1: Red v. Blue Map for the 2000, 2004, & 2008 U.S. Presidential Elections

Figure 1.1A: 2000 Election Map

Figure 1.1B: 2004 Election Map

Figure 1.1C: 2008 Election Map

Blue depiction of American politics has resonated since the 2000 election (see Figure 1.1), and it

remains a common reference point in any discussion of American electoral politics and partisan

competition today.1 “Not since the Civil War and post-Reconstruction period has the country

been so divided” says John Kenneth White of Catholic University (O'Keefe 2004). And that has

become the conventional wisdom.

DISCOVERING THE CULTURE WAR

Scholars and observers of politics have long talked about divisions and divisiveness in the American body politic. Recently, however, the consensus among some scholars, journalists, activists, and elite observers of politics has coalesced around a particular vision of political competition in the United States. Beginning with Hunter’s exposition on the coming culture war in America, scholars have suggested there is a growing schism in the American body politic

(Hunter, 1991, 1995; Wilcox, 1996). It is characterized by an increasingly polarized and yet evenly divided public with social and cultural issues at the fore. While Hunter did not invent the idea of a “culture war” in America, he is responsible for the first serious scholarly treatment of the subject in its current form. Hunter argues that cultural “issues” (abortion, gay rights, funding for the arts, women’s rights, etc) are the surface indicators of a divide that is “really about something deeper and more significant” (Hunter 1991). Hunter suggests that the “new lines of conflict” in America are reflected in an electorate polarized between the “orthodox” and the “progressive” (for a contra argument see Wolfe, 2006).

Hunter’s departure from previous works is in arguing that the culture wars of the present are of a fundamentally different character than those of the past which were characterized by sectarian battles, with some religious denominations taking on one partisan

1 The 2000, 2004, and 2008 electoral maps were generated using the GIS (Geographic Information Systems) program based on U.S. Census Bureau data on these elections.

affiliation and other denominations taking on the other. Hence religion or religiosity was not a factor in and of itself in informing the political divide. Not the case today, according to Hunter.

The outward indicators of the culture war are the policy battles waged over hot-button political issues such as abortion, homosexual rights, public values in education, funding for the arts, affirmative action, child care and the other cultural or social issues. In Hunter’s vision of the near-future, the traditionalists square off on one side with the secular humanists on the other.

This ensures a political landscape marked by intransigent and wholly incompatible viewpoints.

But why is this conflict more intractable than previous incarnations of polarized politics? Hunter posits that the issue positions of traditionalists differ by being deeply rooted in their moral sense. Cultural conflict is “political and social hostility rooted in different systems of moral understanding…They are not merely attitudes that can change on a whim, but basic commitments and beliefs that provide a source of identity, purpose, and togetherness for people who live by them” (Hunter 1991). It is an uncompromising politics that brooks little dissent because it hinges on moral authority rather than the stuff of political compromise

(Hunter, 1991, 1995).

Hunter’s warning of an increasingly divisive and uncompromising conflict over cultural issues resonated with the political classes attempting to understand increasingly close elections.

Political polarization has become a hot topic of discussion and a primary explanatory model for political behavior and electoral outcomes. Hunter’s thesis has had a profound effect on

American political discourse. As can be seen in Figure 1.3, there has been a marked increase in the mention of social issues in the news media (using the NY Times as a proxy) since the

1990’s…climbing from under 20 mentions in a year in 1991 to over 160 mentions in 2005.

Furthermore, whether the culture wars are reality or myth, political polarization itself is an increasingly important and discussed topic in American politics. The number of mentions of

Figure 1.2: Frequency of Social Issue Mentions in the New York Times – 1969-2008

2000

1800

1600

1400

1200

1000

# of mentions # of 800

600

400

200

0 1969 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007

*Search Terms: same-sex marriage OR gay marriage OR abortion OR school prayer OR prayer in school

Figure 1.3: Frequency of Political Polarization Mentions in the New York Times – 1980-2007

180

160

140

120

100

80 # OF MENTIONS OF # 60

*Search Terms: political polarization OR culture war OR religious war OR deep partisan divide

‘political polarization’ in the paper of record (NY Times) has gone from nearly zero in 1980, to a steady increase coincident to Hunter’s “Culture Wars,” to eventually peaking at 158 mentions in

2004 (see Figure 1.4). With the conflict rooted in completely different worldviews, ‘war’ seems an apt descriptor for the competition over policy between these distinct and sharply divided groups within the American electorate (Wilcox, 1996). And indeed, Hunter does not merely contemplate the possibility of the culture “war” precipitating an all too real and bloody cultural conflict within America; he predicts it in the absence of some effort to change the environment of public discourse, an outcome he views as unrealistic. Hunter suggests we are on the path towards violence and, if our stars remain unchanged, violence is the ultimate end. Though acknowledging that the United States is not Beirut and unlikely to become so in the near future,

Hunter warns, “…some escalation in the violence is possible, particularly if the aims of particular political actors seem to them to be interminably frustrated” (Hunter 1991).

Evans strikes a more optimistic note in his work suggesting moral politics in the U.S. isn’t qualitatively different than other more traditional dimensions of political conflict, while Wolfe argues that Americans are a mix of traditionalistic instincts and modern lifestyles that defy a simplistic dichotomy (Evans 1996, 1997; Wolfe and Kohut 2006). However, both scholars concur with the basic fundamentals set out by Hunter: there is a growing cultural conflict that increasingly underlies political divisions in the United States. The American people appear to agree with Hunter. In a poll taken in 2004, 72 percent of Americans assessed the country’s division along fundamentally different views on gay marriage, abortion, and guns as an important or serious problem requiring moderate or major changes by presidents in the future

(see Table 1.1). Only 25 percent of Americans identified the culture war as a small or moderate problem.

Table 1.1: Assessing the Problems for Future Presidents: “The country divided into two Americas where people hold fundamentally different values about gay marriage, abortion and guns” 2

Very Serious / Major Changes 31%

Very Serious / Moderate Changes 13

Important / Major Changes 8

Important / Moderate Changes 20

A Small Prolbem 17

Not a Problem 8

POLITICAL GEOGRAPHY: A WAR OF RED VS. BLUE OR A NATION OF OLIVE GARDENS?

To you, it's sushi. To me, it's bait. – President George W. Bush

Political Polarization: Trends in the Body Politic versus Regional and Reference Group Trends

In Fox’s hit medical drama House, the main character—not coincidentally named

House—asks, “You think it's remotely possible they [two colleagues of House] had sex?” His friend and colleague on the show, Wilson, responds, “They're both single. It's still legal in the

Blue states” (Glatter 2007). The story of an American electorate deeply and sharply divided along geographic boundaries where homogenous communities share similar views on social issues and partisan affiliations has become an indelible part of the popular and journalistic accounts of electoral politics. As the above anecdote indicates, it is now such an ingrained part of the political lexicon that it has become a ubiquitous cultural reference. The increasing

2 Toward a Bold Politics Survey. Survey by Public Interest Project. Methodology: Conducted by Greenberg Quinlan Rosner Research, April 5-April 8, 2004 and based on telephone interviews with a national registered likely voters (see note) sample of 1,000. National registered likely voters are registered voters who voted in the 2000 Presidential election/were ineligible/too young to vote or who didn't vote in the 2000 election but did vote in the 2002 Congressional election. Respondents were asked to assess a list of issues presidents might face in the future, including the one reported here: “The country divided into two Americas where people hold fundamentally different values about gay marriage, abortion and guns.”

frequency with which the Red vs. Blue dichotomy has become a part of our political lexicon is apparent. Furthermore, it not only demonstrates that the terms of the culture wars have become commonplace in our casual discussions of politics, but that the culture wars thesis itself—that we are divided on social issues such as sex outside of marriage—is accepted as a rock-solid characteristic of American politics in the general public. And that’s an important development. As the originator of the culture wars thesis, James Hunter, notes, “the power of culture is first and foremost symbolic. It's the power to name things; it's to define reality, to frame debate“ (Hunter 1991). Popular treatments of the culture war can be as influential on the perceptions of citizens as much as the reality of the political conflict itself. The language that characterizes our public discourse and how it does or does not foster consensus on morality is subject to the conventional wisdom of cultural and political elites. The belief in the Culture

War can affect politics independent of whether a true cultural divide exists. Furthermore, these popular references are indicative of intrinsic knowledge. So ingrained in the meta-conversation of society that they need merely be referenced, not explained. Thus they are evidence of how deeply this concept has penetrated in to the average American’s understanding of politics and thus shaping their own perspectives on politics and political beliefs irrespective of the reality behind the phenomenon. The “Red vs. Blue” dichotomy is a political conventional wisdom.

Though the “Culture Wars” thesis was popularized at least a decade before, the geographic component to the cultural conflict in the American electorate was born in the presidential election of 2000, where states that went for George W. Bush were depicted in the color red on national broadcasts while the states that went for Al Gore, Jr. were colored blue. The “Red

State” and “Blue State” terms are believed to have been coined by Tim Russert in analysis shortly following the election (Zeller 2004). This construction of the culture war has become an integral part of the popular explanations of political conflict, as is evident from the trend in

Figure 2.1. Geographic polarization is relevant both as a phenomena and a potential cause of

polarization. Sociologists have long argued that the composition of a person’s reference

group(s) can have an impact on the extremism of their views on issues as well as their

perception of opposing groups (Tajfel and Turner 1979; Taylor and Moghaddam 1994). As Tip

O’Neil famously observed, all politics is local. If localities have become insular and homogenous,

then our local politics is all of one flavor. To the extent that cities, states, and other

geographically bounded communities have become stratified politically, then the implication on

polarization and cultural conflict is clear. Just as a homogeneous social circle contributed to

Pauline Kael’s infamous reaction to Richard Nixon’s landslide electoral victory in 1972, "How can

that be? No one I know voted for Nixon!" …an electorate that consists of each half of the

ideological spectrum only living and interacting with people “like them” could lead to increasing

political polarization on the issues (as the center of ‘reasonable’ politics shifts in either direction)

and to a more negative and distorted view of the opposition.

Figure 1.4: Red & Blue State Mentions in the New York Times – 2000-2007

100

# of mentions 40

0 2000 2001 2002 2003 2004 2005 2006 2007

*Search Terms: Red State OR Blue State AND Politics

Why would we expect our neighbors and friends to shape and influence our political views? A long-line of research on human interaction suggest that in-group bias leads people to become intolerant towards views incongruent with the group consensus (Taylor and

Moghaddam 1994; Tajfel and Turner 1979). Experiments in group settings have shown that people will ‘reward’ one another with higher payoffs even when the characteristics they share are trivial, such as the same birth date (Tajfel 1970, 1982; Brewer 1979; Maass, Corvino, and

Arcuri 1994; Lord, Ross, and Lepper 1979). Studies on the psychological and sociological effects on conflict and consensus among groups of likeminded individuals versus those with diverse viewpoints show that likeminded groups trend towards the polar extremes rather than the center or more moderate positions (Lord, Ross, and Lepper 1979; Sunstein, Schkade, and Ellman

2006). The group think of politically relevant homogenous groups tends towards polarization on issue dimensions. Cass Sunstein finds that even among judges—expert professionals tasked with a job that, in its ideal, requires neutral and objective application of standards rather than personal belief-systems—the tendency to move towards extreme positions when grouped with likeminded individuals (e.g. a court of appeals with mostly conservative Republican appointees tends towards even more extreme conservatism than would be expected given their individual viewpoints) is real and pronounced (Sunstein, Schkade, and Ellman 2006).

Polarized Elections & National Issue Trends: Local Landslides versus the Continental Consensus

Given the limitations of the medium, most reporters have addressed the polarization question from a relatively superficial level. Yet there have been some significant efforts by journalists to get to the bottom of the political polarization puzzle. These focus primarily on illustrating the increasing homogeneity within communities juxtaposed with increasing heterogeneity between communities (See Figure 1. 5). Brousk, suggests, in analyzing electoral returns for the battleground state of Wisconsin, “[t]o put it simply, we are so divided” (Borusk 2004). Though

Wisconsin was seen as a “battleground” state and the 2004 presidential election was closely decided, at the county level it was not close. He notes that while the overall election was decided by just 5,708 votes (0.02%), 52 out of the 72 counties were decided by 5 percentage points or more (Borusk 2004). Ben Bishop of the American-Statesman (in concert with a statistician) conducted a series of journalistic investigations into the polarization of American politics (Bishop 2004). They concluded that American elections are increasingly characterized by nationally close results that hide the fact that locally these elections reflect landsides, or what they call the Big Sort. “For almost half of all voters, the close 2004 presidential election wasn't close at all. It was actually a series of local landslides, as Americans continued a decades-long process of sorting themselves geographically into like-minded communities… voters on average are less likely to live among neighbors who supported a different candidate for president.

Communities are more homogenous, more single-minded” (Bishop 2004).

Godless Highbrow Bluebloods vs. Devout Country Rednecks: Income, Region & Issues

“I never said all Democrats are saloon-keepers. What I said is that all saloon-keepers are Democrats.” – Horace Greely, 1860.

The New Deal realignment produced an issue dimension which cleaved the parties on welfare issues (Petrocik 1981). Economic issues have long been viewed as one of the dominant factors in the political calculus. Social and cultural issues, and their minimal impact on electoral politics, bring up the rear (Edsall 1986; Kinder and Kiewiet 1984; Kiewiet 1983; Geer 1992). The importance of the economy, one’s own financial situation, and interrelated issues condition both vote choice and the perception of parties (Tufte 1975; Fiorina 1978; Nixon 1962). The major works in political behavior of the 1950’s and 1960’s naturally focused on the influence that attitudes on economic issues has on voting behavior (Eldersveld 1952; Lazarsfeld, Berelson, and Gaudet 1948; Berelson, Lazarsfeld, and McPhee 1954; Campbell et al. 1960; Campbell and

Cooper 1956; Campbell, Gurin, and Miller 1954; Campbell and Stokes 1959; Key 1955;

Figure 1.5 Red vs. Blue Communities: 2008 Presidential Election Results by County

http://right-mind.us/blogs/blog_0/archive/2008/11/15/64322.asp

Schlesinger 1958; McClosky, Hoffmann, and O'Hara 1960; Wilson 1962; Lipset and Rokkan 1967;

Lipset 1960). Some scholars believe that the dominance of economic and welfare issues in

American politics reflects a class division in political behavior. Lipset in particular argued theexistence of class voting in a series of case studies on democratic countries with capitalist systems conducted in the 1960’s. The United States was no exception, with the Republican and

Democratic parties representing “the interests of different classes” (Lipset 1960). However, while economic issues are predominant on the public agenda, the relatively small income differences between Republicans and Democrats (when controlling for region of the country) suggests that class hasn’t been essential in explaining partisan differences, or at least, other social demographic factors are just as relevant.

To the extent that other cleavages have become politically salient and resulted in significant changes in the political coalitions of the parties, economics suffers as a dividing line for the parties (Manza, Hout, and Brooks 1995). The emergence of new cleavages, such as the racial or civil rights cleavage identified in the post-New Deal realignment, is one hypothesis explaining the decline of the relevance of class (Petrocik 1981). Huckfeldt and Kohfeld argue that the Democratic coalition was splintered due to racial hostility. “The politics of race disrupts class politics because…the social and political isolation of blacks benefits advantaged groups…by fracturing the political vehicle of lower-class interests: party competition structured along class lines” (Huckfeldt and Kohfeld 1989). This argument is consistent with Petrocik’s finding that the realignment of the 1970’s was determined by racial conflict (Petrocik 1981).

According to the Culture Wars thesis, the social/cultural dimension figures into the cross-cutting cleavages that have supposedly knocked the government welfare issue dimension off of its politically-aligning pedestal. Hunter suggests that the new cleavage upon which the

post-New Deal realignment hinges is the secular versus religious divide. The 2000 election gave this thesis a geographic component, depicting a cultural conflict between Red States and Blue

States. Some observers of politics, while implicitly endorsing the reality of the emergent

“Culture Wars,” object to it on normative rather than empirical grounds. Thomas Frank takes umbrage at the hijacking of the political agenda in favor of social issues in his widely read

What’s the Matter with Kansas? (Frank 2004). He quotes favorably a plaintive query, “’How can anyone who has ever worked for someone else vote Republican?,’ she asked. How could so many people get it wrong? Her question is…the preeminent question of our times. People getting their fundamental interests wrong is what American political life is all about. This species of derangement is the bedrock of our civic order; it is the foundation on which it all rests” (Frank 2004). Frank’s view is that the only relevant partisan and behavioral-determining factor for the rational voter is class.

Frank, a journalist, uses his home state of Kansas as the prototypical example of a state that should be a Democratic state, given its cornucopia of blue-collar workers (despite the fact it has been Republican since the 1950’s and well before the ‘culture wars’ supposedly began).

Kansans, according to Frank, are duped by the opiate of abortion, gay marriage, Hollywood and gun control. “This puzzled me when I first read about it, as it puzzles many of the people I know.

For us it is the Democrats that are the party of the workers, of the poor of the weak and the victimized” (Frank 2004). Setting aside Frank’s presumptuous imposition of his view on the

‘proper’ preference ordering for working class Americans and his conspiracy theory that the

Blue/Red narrative is an invention of clever conservatives duping poor, white America into voting against “latte” liberals, Frank clearly asserts that a sharp decline in class-based voting in the American electorate is an empirical reality. Indeed, Frank is convinced that these social issues have dragged traditional left-wing economic voters over the ideological dividing line and

converted them into low-regulation, lower taxes, economic conservatives. “The [working class]

dad would listen to the son spout off about Milton Friedman and the godliness of free-market

capitalism, and he would just shake his head. Someday, kid, you’ll know what a jerk you are. It

was the dad, though, who was eventually converted. These days he votes for the farthest-right

Republicans he can find on the ballot” (Frank 2004). The working class has gone Right despite

the fact that Republicans fail to deliver on culture war issues and their economic policies are

counter to the working class interest. As Nick Kristof plaintively recounts shortly before the

2004 election, Kerry voters “should be feeling wretched about the millions of farmers, factory

workers, and waitresses who ended up voting—utterly against their own interests—for

Republican candidates” (Kristof 2004). Frank is not alone in his view that cultural political

conflict increasingly informs the political decisions of voters and politics in America.

THE CONVENTIONAL WISDOM ON POLITICAL POLARIZATION

Hunter may or may not have been the first scholar to warn of a growing culture war in

America, but he has certainly not been alone in his assessment that the American public is increasingly divided, that neither side of the divide will brook compromise or negotiation with the enemy, and that the prime mover in the divide is between religious and secular citizens.

With close elections producing bitter partisan battles, social issues becoming more salient for voters and elites, and an apparent increase in local community homogeneity paired with increasing heterogeneity between counties, states, and regions, the evidence suggests the great cultural divide in America is getting deeper and more difficult to bridge through normal

American political institutions. Political compromise and centrist politics is one of the first casualties of the culture war. The media and elites have increasingly given attention to and generated memes on the Red vs. Blue state cultural divide as well as the social issues that are

believed to underlie this cultural war. The American public has responded, identifying the culture war as a significant problem in American politics, perhaps requiring future presidential intervention. According to the conventional wisdom, the culture war is real, it is scary, and it is here.

CHAPTER 2: THE CULTURE WAR MYTHOLOGY: CRITICS OF THE CONVENTIONAL WISDOM

Political polarization has received a great deal of attention from scholars and laypeople alike. The most recent scholarship and the ongoing debates over the nature of the Culture Wars and political polarization provide insight on the theoretical implications and empirical puzzles of political polarization. The materialization of an emergent polarization among scholars is apparent in the recent literature. With scholars on one side of the divide suggesting the Culture

Wars thesis is a myth and that the trends in public opinion show the exact opposite: movement towards more moderation, centrism, and tolerance (Fiorina, Abrams, and Pope 2004). Others suggest that polarization in the electorate is real and detectable (Abramowitz and Saunders

2005; Campbell and Cannon 2006; Abramowitz and Jacobson 2006; Jacobson and Edsall 2006).

The implications of a true culture war—a truly polarized public—are apparent. With the center vacated and public opinion coalescing around two distant poles, the potential for compromise would evaporate. With the political system unable to cope with the disposition of policy opinion, other means could be sought to achieve political ends. Yet there has been no eruption of political violence in America. No end to compromise apparent in the near two decades since Hunter made his grave warning. Is it ahead? Furthermore, is it even true that that the American electorate has become more polarized? If not, it bodes well for the continued existence of a peaceful political culture, however contentious at times. Several scholars and observers of politics have challenged the conventional wisdom on political polarization making exactly that case.

NOTHING TO SEE HERE! FIORINA AND THE CULTURE WAR MYTH

Morris Fiorina, a scholar with a penchant for unconventional takes on conventional theories of politics, registered one of the first counter-strikes to the growing consensus on an emerging Culture War and increasing political polarization in his 2005 work: Culture War? The

Myth of a Polarized America (Fiorina, Abrams, and Pope 2004). In this small but influential tome

(later expanded upon in a second edition), Fiorina makes several arguments regarding the causes and nature of polarization among masses and elites, but his fundamental argument is that the rhetoric regarding the polarization of the American electorate over the last ten to fifteen years is apocryphal. That Hunter and those who have accepted his thesis are wrong.

There is no political polarization. “The simple truth is that there is no culture war in the United

States—no battle for the soul of America rages…” (Fiorina, Abrams, and Pope 2004). The trend, according to Fiorina, over the last several decades in the American public is one of growing centralization and moderation of the electorate on social issues (Fiorina, Abrams, and Pope

2004). This shot across the culture wars bow has generated much debate over what empirical evidence, if any, underlies political polarization in American politics.

Fiorina argues that the American electorate is largely centrist on the bulk of political issues (including such hot button issues as abortion and homosexual rights), that they remain relatively ambivalent on the policies related to these issues, and that the trends in public opinion in the American electorate reflect increasing tolerance rather than divergence and conflict (Fiorina, Abrams, and Pope 2004) . Fiorina points to relatively small shifts in the average and moderate opinions on abortion—the sine qua non of salient social issues—in the electorate over the last few decades as evidence that the American public’s social views are stable and centrist. The electorate over the past thirty to forty years has been relatively stable in its distribution of political opinion. “There is little evidence that American’s ideological or policy

positions are more polarized today than they were two or three decades ago, although their choices often seem to be” (Fiorina, Abrams, and Pope 2004). Thus the American mass electorate is largely centrist on a range of issues (including hot-buttons such as abortion and homosexual rights) and not sharply divided as suggested by the polarization thesis.

So, why all the fuss about a Culture War? The myth of politicial polarization along cultural lines is a product of elite activists promoting the perception of a divided nation

(Stephonopolus’s polarization entrepreneurs) and bad interpretation of the available data by a credulous media establishment. “The myth of a culture war rests on misinterpretation of election returns, lack of hard examination of polling data, systematic and self-serving misrepresentations by issue activists, and selective coverage by an uncritical media more concerned with news value than with getting the story right” (Fiorina, Abrams, and Pope 2004).

The Fiorina posits that the only true polarization observable over this time span is among the elites in the two political parties, whom have polarized absent or despite of any signals from the voters.

One of Fiorina’s primary theoretical arguments is that partisan polarization is not the same thing as ideological or issue polarization. He argues that “sorting” is distinct from

“popular” polarization. Fiorina argues that the realignment of the South has created a more conservative Republican party and a more liberal Democratic party, but the ideological disposition of the mass public has remained stable.

Partisan sorting can occur without any substantial shift in the issue opinions and policy positions of the electorate as a whole (Fiorina, Abrams, and Pope 2004). In 2004, 21 percent of voters described themselves as liberals, 34 percent as conservatives, and 45 percent were self- described moderates. These numbers are nearly the same as the average of the ideological disposition of the electorate found in the exit polls over the last 30 years (Galston and Kamarak

2005). So while the parties have become more ideologically consistent at the mass level, the

distribution of conservatives and liberals and the relative position on cultural issues of the mass

public has not changed.

To support his argument, Fiorina looks at two big, defining cultural issues: abortion and

homosexuality. On abortion, as noted previously, he finds “remarkable” stability over the time

series. Furthermore, demographic groups (e.g. gender) either do not differ in their average

positions or their differences aren’t as significant as expected (religious vs. the non-religious).

On homosexuality, Fiorina identifies a strong liberalizing trend in the American public. Far from

a “major front” in the culture war, he posits that homosexuality is a decreasingly relevant issue

dimension, since the American public is converging on a more tolerant view of homosexuals and

homosexual rights. This increase in tolerance of homosexuals, from Fiorina’s perspective, is

strong evidence against the Culture Wars thesis (Fiorina, Abrams, and Pope 2004).

He concludes his analysis by arguing that class remains a strong predictor / correlate of

partisan and political beliefs in the American electorate, and that, while religiosity (as opposed

to religious denomination) has become an increasingly relevant influence on partisan and

political beliefs, income divisions are even more important than they were decades previous.

“On the contrary, comprehensive research indicates that income divisions are more important

now than in earlier decades” (Fiorina, Abrams, and Pope 2004). He further argues that the only

polarization that has occurred is at the elite level, driven by a “hijacking” of American politics by

interest groups as a consequence of the funding they provide politicians and the parties.

RED VS. BLUE? POLARIZATION AS AN ELECTORAL PHENOMENON

One important contribution of Fiorina’s analysis is his debunking of the popular notion that close elections constitute evidence of political polarization and a “great divide” between the American populace. A close election can result just as easily from a centrist, unimodal

electoral distribution of political beliefs that inform electoral choices as it does from a bifurcated

and sharply divided distribution of political beliefs in the electorate. Likewise, significant

polarization in society can exist despite the fact that an election is a landslide. “An election may

be closely divided without being polarized, as it was in 1960, or deeply polarized without being

closely divided, as it was in 1936, or neither, as seems to have been the case in the famous ‘Era

of Good Feeling’ between the war of 1812 and Andrew Jackson’s arrival on the presidential

stage” (Galston and Nivola 2006). The only significant evidence of political polarization that

Fiorina acknowledges has occurred at the elite level. This has happened despite the stable

views of the mass electorate. Indeed, Fiorina suggests that the problem facing American

democracy is not a war between sharply divided groups in the electorate but rather a run-away

cadre of elites pushing policy outside the mainstream of mass opinion (Fiorina, Abrams, and

Pope 2004; Fiorina and Levendusky 2006). As Mayhew noted, there is clear and rational

connection between elected officials and their constituents, whether or not this reelection

motivation is single-minded (Mayhew 1974).

MASS POLARIZATION: DISTRIBUTION TRENDS IN THE ELECTORATE’S ATTITUDES & VALENCE VS. POLICY

Fiorina argues very little mass polarization has occurred. In Culture War? he argues that

Americans are moderate, tolerant, and relatively ambivalent towards politics and the divisive ideological battles that are so intense at the elite and activist level. “Americans are closely divided, but we are not deeply divided, and we are closely divided because many of us are ambivalent and uncertain, and consequently reluctant to make firm commitments to parties, politicians, or policies. We divide evenly in elections or sit them out entirely because we instinctively seek the center while the parties and candidates hang out on the extremes”

(Fiorina, Abrams, and Pope 2004). Fiorina suggests that most of the American electorate reflects the predominance of cross-cutting ideological beliefs, with individuals holding a mixture

of conservative and liberal views, depending on the issue at hand. Though sorting has occurred at the mass and elite levels, the extent to which they have sorted is as night is to day. Partisan polarization is mostly an elite phenomenon, accordingly. He suggests that geographical polarization is overstated and that social cleavages have become more 'fuzzy' rather than stark dividing lines between intractably disagreeable groups. Economics continues to drive the partisan wheel of politics and has not been displaced by religiosity. In his previous work he suggested partisan gerrymandering may explain increasing congressional polarization, though he has abandoned that argument in more recent takes on the subject (Fiorina, Abrams, and

Pope 2004).

The geographic tale of a polarized public isn’t without its skeptics beyond Fiorina, even in the popular accounts of polarized politics. Robert Samuelson objects to the depiction of a deeply divided Red vs. Blue political electorate (Samuelson 2004). He argues, as others scholars have argued, that opinion polling shows an increasing consensus on cultural issues such as racial and gender equality, traditionally social hot buttons. He takes on the argument that underlies the partisan map that some observers , such as E.J. Dionne, focus on: “The divide in

American politics is about more than the ideological distance between the two parties. Right now, Red Staters and Blue Staters live in two different political universes” (Dionne Jr. 2006).

Samuelson counters by arguing that the technological revolution (the internet and the information revolution) has served to make distant lands in the United States more culturally homogenous (i.e. the Wal-Mart effect) rather than reflecting the growing political dichotomy between “Blue” and “Red” consistent with the Culture Wars thesis. “Texas and New York have more in common now than in 1950 or 1960,” he notes, though for some reason fails to cite the prevalence of Olive Gardens in his litany of similarities (Samuelson 2004). Samuelson asserts that the color-coded partisan map masks fairly close presidential elections in those states. In

2000, the election that birthed the Red vs. Blue political maps, nineteen states were won with

51 percent of the vote or less. Hardly the picture of geographically homogenous states

bifurcated by a deep political division (Samuelson 2004).

DUDE, WHERE’S MY POLARIZATION? THE DIMAGGIO STUDY ON POLITICAL POLARIZATION

DiMaggio and his colleagues produced the most comprehensive time-series on political polarization to date in their 1996 article and subsequent update (DiMaggio, Evans, and Bryson

1996; Evans 2003). Assessing polarization across various social issue measures in the ANES and

GSS, they examine trends in the variance and kurtosis of these opinion indexes over time.

Consistent with Fiorina’s argument, they found little evidence of increased polarization over the period of 1972 through 1994 in both the GSS and NES, excepting abortion. The distance between various groups (age, education, sex, race, region, religion, etc.) on racial and gender issues, crime, sexual morality, and the role of the welfare state have narrowed over their time series (DiMaggio, Evans, and Bryson 1996). “We find no support for the proposition that the

United States has experienced dramatic polarization in public opinion on social issues since the

1970’s” (DiMaggio, Evans, and Bryson 1996). DiMaggio and his colleagues formulate four theoretical conceptualizations of polarization: consolidation, constraint, bimodality, and dispersion. Their empirical measure of dispersion is variance:

Equation 2.1: Dispersion

= ) 1) ଶ ଶ ෍ሺݔെݔҧ ൗሺܰ െ ݏ

This is the “No one I know voted for Nixon” phenomenon or the “separate camps” aspect of polarization as described by DiMaggio et al. The more that the distribution of opinion on an issue reflects two relatively distant poles with a vacated center, the more intractable

political competition becomes. In other words, a tendency towards extremism and a resistance

to compromise is inherent to the formation of separate camps, as similarly situated individuals

all view the issue relatively similarly and those who differ with them on the question are that

much more distant from their views on the subject. DiMaggio uses kurtosis to measure

bimodality. Though it should be noted that objections regarding the extent to which kurtosis

reflects bimodality have been raised in the statistics literature. Most such objections refer to

relatively distinct distributions (such as two modes relatively close to one another, i.e. the

gamma distribution), though there are other issues as well, leading some scholars to dispute

kurtosis as a measure of bimodality (DiMaggio, Evans, and Bryson 1996; Mouw and Sobel 2001).

Equation 2.2 Kurtosis

= [ ) ] ( ) – 3 ସ ସ ത is peakedቅ (large percentage of ܵכܰ Kurtosis is, in theory, ݇positiveቄ෍ whenሺܺെܺ a distributionΤ responses located at one point) irrespective of whether the peak is located in the center or elsewhere in the distribution. As a distribution flattens, kurtosis becomes more negative. As responses vacate the center in favor of the extremes, kurtosis becomes further negative. The constant “3” is subtracted from the kurtosis equation so that the normal distribution is roughly equivalent to zero. Kurtosis is distinct from skewness in that skewness provides a measure of the direction a distribution is lopsided (one way or the other) while kurtosis distinguishes between a dimension heavily weighted towards one side of the distribution and where a distribution is characterized by significant proportions of the responses on both sides of the center (Mouw and Sobel 2001; DiMaggio, Evans, and Bryson 1996)

In their analysis of polarization the singular exception noted to their ‘absence of polarization’ finding is that of abortion. “If attitude polarization entails increased variance,

increased bimodality, and increased opinion constraint, then only attitudes toward abortion

have become more polarized in the past 20 years, both in the public at large and within most

subgroups” (DiMaggio, Evans, and Bryson 1996). Yet even that finding is suspect according to

some scholars (Mouw and Sobel 2001; Shaw 2003). Mouw and Sobel argue even on abortion

the American public has failed to polarize. Though Dimaggio and his colleagues maintain that

polarization on abortion is a real phenomenon (Evans, Bryson, and DiMaggio 2001).

WAGGING THE DOG: ELITES HOLD THE MEDIAN VOTER’S LEASH AND DO ALL THE BARKING

In his most recent take on the subject, Fiorina argues that sorting has markedly increased at the elite level, while very little sorting has occurred among the masses (see Table

2.1 for illustration of sorting). Abramowitz counters arguing that the masses, in some respects, have outpaced elites in partisan sorting. He finds that polarization in Congress has increased only half as much as the partisan polarization evident among voters (Abramowitz and Jacobson

2006). Fiorina explains the elite sorting disconnection from mass sorting as a function of the primary system (rather than gerrymandering), where candidates preemptively move to the relative extremes in anticipation of challenges from the ideological poles. “Because sorting produces a more homogenous and a more extreme primary electorate, the pressure increases for candidates to take consistently liberal or conservative positions on

Table 2.1: Partisan “Sorting” with No Ideological “Polarization” IDEOLOGY TIME PERIOD 1 TIME PERIOD 2

Liberals 65 Democrats, 35 Republicans 95 Democrats, 5 Republicans

Moderates 100 Independents 100 Independents

Conservatives 65 Republicans, 35 Democrats 5 Democrats, 95 Republicans

Source: Adapted from Fiorina & Levendusky, 2006

most issues, even when moderation would be more helpful in the general election. Thus sorted

partisans move candidates toward non-centrist positions“ (Fiorina and Levendusky 2006). All

well and good for the primaries, but that doesn’t really explain elite polarization outside of the

primary context. Running to the center in a general election is an ancient political axiom, and

primary challenges to sitting partisans are as rare as a blue moon in congressional races. So

why are elites diverging if the mass public is increasingly moderate and centrist? The classic

Downsian model illustrates the centripetal logic of party competition in an electoral setting with rational voters: the median voter theorem (MVT). This simple yet powerful insight provided a logic for the classic observation of candidates moving towards their parties extremes in primaries (or moving towards the median primary voter) only then to run to the center in general elections (move back towards the median voter in the electorate. DiMaggio et al. in their study of polarization argue explicitly that political polarization is implicated in median voter models. “Once the median is no longer the mode, majorities may form around either mode. If bimodality is great and the issue salient, rational candidates may embrace extreme positions in order to prevent a sit-out by purists or a third-party challenge” (DiMaggio, Evans, and Bryson

1996).

Yet in order for distributional changes to be relevant in median voter models, a step away from the stylized model developed by Downs is necessary. Downs argued that the location of equilibria in an election would be dependent on the shape of the distribution of citizen preferences (Downs 1957; Hotelling 1929). Downs was wrong (see Figure 2.1). The pure

MVT with complete turnout and sincere voting predicts that the median voter is, in fact, a

Condorcet winner: no position defeats the ideal point of the median voter in a pair-wise vote irrespective of distributional qualities (Black 1958). Thus, parties (or candidates) must move towards the center to maximize votes. However, there are countless elections where parties

and candidates have failed to converge to a single policy point or even a vector of policy points

in a continuum of policies. While 1950’s political competition tended to conform to moderate

and centrist expectations, current rhetoric and serious analysis alike decry the stark choice (as

opposed to an echo) that the parties present today and the ratcheted conflict that accompanies

it. Many analysts believe that diverging citizen preferences must be related to diverging

candidate and party positions, as the realignment literature implicitly assumes. Either this is the

consequence of a more polarized population, a more polarized electorate, or a nomination

process that generates polarized candidates, i.e. the primaries effect (Fiorina 1999).

Fiorina provides theoretical justification for his argument on polarization through the

use of a formal model of voter response to elite polarization (Fiorina, Abrams et al. 2005). It

employs a Downsian proximity model of the vote decision with which he attempts to explain

how the cultural dimension could become increasingly important while the voter positions

remain stable (Platt, Poole, and Rosenthal 1992). As can be seen in Figure 2.1, Fiorina argues

that elite polarization yields divergent candidates and parties. This thus elevates the

importance of a cultural or social dimension despite the fact that voter positions on this dimension remain the same. This is a key point in Fiorina’s arguments on the behavior of the elites relative to the masses. He argues that political and partisan elites, unanchored by the pragmatic and materially-oriented party machines of the past, have become beholden to extremist advocacy groups. As such they have polarized irrespective of the growing ideological centrism of the American electorate. The bi-modal distribution of polarized voters (and thus responsive political parties) in Figure 2.1 is a myth according to this view. Indeed, he titles his most recent work on the Culture War’s question “Disconnected: The Political Class versus the

People.” Fiorina asserts “there remains a critical missing piece in the prevailing portrait of a polarized American political order—the American people” (Fiorina and Levendusky 2006). His

Figure 2.1: Polarized Elites? Are Elites Located Distant from Median Voter and the Mass of Voters?

LEGEND Republican Party Elites Democratic Party Elites If the distribution of Fiorina’s Normal Distribution voters is normal and of voter ideal points centrist, then the party Bi-Modal Distribution of voter elites have moved away ideal points Median Voter Position from voters

If the distribution of voters is bimodal, then the party elites are located with their respective constituents

XMED

POLICY

argument is that, despite the research illustrating the centripetal tendency of two party politics

in America, “American politics today finds a polarized political class competing for the support of

a much less polarized electorate” (Fiorina and Levendusky 2006). Why? Because the

polarization of the Republican Party and the Democratic Party has occurred simultaneously,

there is no way for voters to punish one party over the other for its movement away from the

center of voters (as they did to the Republicans in 1964 and the Democrats in 1972). The

apparent emergence of a moral dimension in politics is not a consequence of voter polarization

on the social dimension, but rather a consequence of a change in the candidates at the elite

level. He suggests voters have moved in the opposite direction: becoming more moderate on a

range of social issues that include gay rights and abortion. Hence, as can be seen in Figure 2.1,

the party elites have become “disconnected” from the voters located at the center of the

issue/ideological dimension.

WHAT’S THE MATTER WITH THOMAS FRANK?

Recall Thomas Frank’s arguments that the lower class has been duped into voting for

Republicans on the cultural issues. But is Frank right? Has class-based voting declined? Has it been replaced by a now dominating social/cultural cleavage? Morton says no, arguing the stylized fact of an electorate, and particularly a lower class electorate, more concerned with social and cultural issues than economic issues is wrong. Citing recent election data from the

2004 election, Morton points out that lower-class values voters are much less likely to support

Republicans (30%) than upper-class values voters (60%), contrary to Frank and the Culture Wars thesis (Morton 2006). Stonecash’s Class and Party in American Politics presents a substantial time series examining determinants of partisanship since the 1950’s and concludes that

Republican gains have come among middle and upper-class voters, thus widening the traditional gap between the c lasses in party support (Stonecash 2000). This trend is also evident in

Nadeau and Stanley’s research on partisan trends among Southern whites (Nadeau and Stanley

1993). McCarty, Poole and Rosenthal present a similar finding in an analysis of class, partisanship, and presidential voting, showing class as an increasingly predictive favor—subject to fundamental economic and social changes—in electoral choice (McCarty, Poole, and

Rosenthal 2006). Stonecash and his coauthors conclude Democratic competitiveness in the post-New Deal realignment was rooted in their increasing appeal to working class voters outside of the South (Stonecash et al. 2000). In a later work focusing on congressional districts and partisan polarization, Stonecash shows that Democratic candidates increasingly win in districts with lower median incomes, higher urbanization, and a relatively high percentage of non-whites

(Stonecash, Brewer, and Mariani 2003; Stonecash 2005).

Larry Bartels, in his vivisection-like review of Frank’s work, argues that, despite bad analysis of the 2004 exit polls and Frank’s “stew of memoir, journalism and essay,” the theory that Republican success is a consequence of a growing cohort of working class conservatives is a myth (Bartels 2006). According to Bartels, the bottom third of the income distribution among working whites have become more reliably Democratic in presidential elections. He finds surprisingly little class differences in Democratic support from 1952 to 1974. However, from

1974 to 2004 he finds an exacerbation of the differences between the upper and lower thirds of the income distribution. The gap between lower-class whites and upper-class whites increases from 4% to 14% between the two periods (Bartels 2006). While he does find a decline in white working class party identification this is offset by an even greater decline among upper income voters. His analysis of issue trends in the National Election Studies suggests that working class whites have not become conservative over the last 30 years and that, indeed, the working class trends towards moderation in comparison to the more extreme social views of the upper class

(Bartels 2005).

In an update of the review, Bartels addresses Frank’s rejoinder to his review. Frank

suggests that the original thesis of What’s the Matter with Kansas was that the uneducated, not the lower income voters that Bartels examined, make up the working class. Bartels’ rebuttal identifies a particularly puzzling trend that tends to puncture a well-established conclusion regarding the partisan divisions of the 1930’s and the 1940’s. The traditional story of the New

Deal realignment was that it was strongly grounded in class conflict, with the poor solidly in the

Democratic camp aligned against greedy and rich Republicans. Yet Bartels finds that educational differences were more strongly associated with the partisan divisions in the 1950’s and the 1960’s. Contrary to Frank’s revision of the “What’s the Matter” thesis that suggested the uneducated have trended Republican, Bartels’ analysis suggests that education has vanished as a relevant partisan cleavage since the 1970’s. Income disparities in party affiliation and partisan political choice have escalated considerably at the expense of educational differences

(Bartels 2006). He again examines issue trends, but assesses here the extent to which education differences explain these trends. Once again, Bartels argues Frank’s take on the macro-trends in partisanship cannot stand up to scrutiny. The trend in educational effects on partisanship has actually declined over the last few decades.

One exception to Bartels’ generally damning case against Frank is on the abortion issue.

Bartels finds a strong and significant positive coefficient for the electoral significance of abortion among white voters without college degrees. Pro-life voters have been pulled towards the

Republican Party over the last twenty years. Before Frank dances a jig, however, Bartels also finds that this effect has been more significant among whites with college degrees, suggesting that the educational attenuation of vote choice, to the extent one exists, is not operative in the direction Franks argues. Furthermore, since 1996, among whites without college degrees abortion has declined in importance while it has increased in importance for whites with college

degrees (Bartels 2006). What’s more, the newer analysis conforms with Bartels’ earlier review in that economic issues continue to be relevant in structuring the electorate’s political choices and their impact is slightly larger among the white working class voters (Bartels 2006).

However, all this must be set against the backdrop of a significant trend upward in the number of college educated and bachelorette-earning adults in the United States over the past 50 years.

High-school graduation rates have increased from around 30% to nearly 90% of the 25-and- older population. The college degree-earning population has gone from well under 10% to just- under 30% of the population. Hence, it may be the case that the apparent trends Bartels identifies among the college and non-college educated sub-populations may be a consequence of the significant changes in composition those groups have undergone in the last 50 years. The effect of education on voting itself may not have changed while the composition of the group of

‘educated’ voters has, possibly explaining the puzzle.

Bartels’ lower income = support-for-Democrats finding presents a vexing puzzle regarding the observed empirical regularities in politics. On the one hand, it appears that lower class voters remain solidly in the Democratic camp and perhaps have increased their support for the party of Jefferson in the last 20 or 30 years. However, at the state level of analysis, states with lower per capita income and gross state product have trended on to the Republican side of the ledger. The reliably Blue states, on the other hand, are also the wealthiest states in the

Union. As can be seen in Table 2.2, the top states in per capita gross state product (GSP) and median household income (MHI) are reliable Blue states while the bottom states are just as reliably Red. The bottom ten states in MHI contain only one blue state, New Mexico. Likewise, only one red state (Virginia) cracks the top ten in MHI. Blue states overall have an average MHI of $55,000 while Red states trail well behind with an MHI of just above $47,000. The per capita

GSP differential isn’t as large, but the blue states still have a sizable advantage. For the red

states the average GSP is $34,000, but the blue states have an average per capita GSP just under

$40,000. So, evidence at the individual level suggests that lower-income voters continue to throw in their lot with the Democratic Party, that they have become increasingly Democratic over the last few decades, and that economic issues are a significant factor in their party identification and electoral choice. On the other hand, at the state-level, rich states overwhelmingly go Democratic in presidential elections while poor states are solidly Republican.

One answer is that this is a simple manifestation of the ecological fallacy—generalizing from the aggregate to the individual. Hence political observers such as David Brooks can project “the liberal stereotype onto the map” and “admit on behalf of everyone who lives in a blue zone that they are all snobs, toffs, wusses…and utterly out of touch with the authentic life of the people”

(Frank 2004). As Brooks puts colorfully, “We’re [Blue Staters] more sophisticated and cosmopolitan…But don’t ask us, please, what life in Red America is like. We don’t know…We don’t know what James Dobson says on his radio program…We don’t know about Reba and

Travis…Very few of us know what goes on in Branson, Missouri, even though it has seven million visitors a year, or could name even five NASCAR drivers…We don’t know how to shoot or clean a rifle” (Brooks 2001). Polarization is thus a mirage: a feverish, imagined divergence that is the product of mapping our own political biases on to the rest of America.

Morton suggests that the ecological fallacy is the answer, pointing out that Fisher’s

Paradox can disguise radically different individual support for Republicans among individual low and high income voters that wouldn’t show up at the aggregate level (Morton 2006). However, noting that there can be differences between individual characteristics and aggregate characteristics doesn’t rule out the possibility that the aggregate results are a real political phenomena rather than a statistical artifact. Gelman and his coauthors attempt to tackle this apparently intractable problem through a multi-level modeling of the aggregate and individual

Table 2.2: Income Polarization at the State Level by Presidential Voting 1996-2004*

MHI (median income) GSP (per million) GSP (per capita)

RED 47,227.30 225,901.40 34,072.78

BLUE 55,070.19 342,908.90 39,756.05

SWING 44,171.00 215,005.50 32,567.17

*See Appendix A for by state data effects on partisanship in yet another “What’s the Matter” article, this time with Connecticut as the foil (Gelman et al. 2005). Gelman et al. examine four separate levels of aggregation: the individual voter, the county, the state, and the national level. As Gelman and his colleagues note, given institutional aspects of the system such as electoral rules, parties, and geographic- based representation, elections are not “simple cumulations” of voter decisions and thus aggregate analysis can reveal real phenomena missed by a study of just individuals (Gelman et al. 2005). They find that the income and voting differences between Red and Blue states are real, despite the fact that richer voters within states tend to support Republicans. Their resolution of the apparent paradox is that within each state, income is positively correlated with

Republican vote choice, but average income varies by each state. Thus individual income

(positive) and state average income (negative) are independent predictors of presidential vote support.

The puzzle of a rich state such as Connecticut consistently supporting Democratic candidates is revealed to be not much of a puzzle at all. Rich voters in Connecticut are not all limousine liberals. Richer voters in Blue states tend to support Republicans just as they do in the

Red states. This is attenuated by the effect of the higher median income of the state such that

the net effect of income is more or less a wash. Gelman argues this debunks Frank and Brook’s

characterization of “latte” Democrats and “NASCAR” Republicans while, contra Bartels and

Fiorina, suggests the Blue/Red dichotomy is a real political and cultural phenomenon (Gelman et

al. 2005). Gelman provides little by way of explanation, theoretical or otherwise, for this pattern

but contends the pattern is real.

All this said, that income is still a relevant partisan cleavage and Republicans remain the

party that appeals to upper-income voters with Democrats favored by the working class does

not mean that social issues have failed to emerge as an important cleavage in partisan politics or

that it is irrelevant to the secular realignment of the political system. Employing a soft

conceptualization of realignment can allow for the fact that income still matters—still structures

the political cleavage of the parties—while at the same time changes along other issue

dimensions have resulted in significant shifts in the political coalitions of the parties that have

had a lasting impact on the electoral fortunes of both parties. As noted earlier, the new and the

old can persist through a secular realignment.

BLOWBACK: CRITICS TAKE ON FIORINA’S CRITICISM OF POLARIZATION

The argument over polarization wasn’t ended by Fiorina’s publication of The Culture

Wars Myth? Each of his main points remain a source of division among scholars. Abramowitz and Saunders, in a direct response to Fiorina’s myth thesis, find deep divisions in America between Democrats and Republicans, red state voters and blue state voters, as well as religious voters and secular voters within the mass electorate. Polarization is not confined to a small cadre of governmental officials and activists but rather is a phenomena of the general public and one reflective of an increasing trend in American politics (Abramowitz and Saunders 2005).

Abramowitz and Saunders take issue with Fiorina’s argument that there has been no increase in opinion polarization in the American public and that it is exclusively an elite phenomenon. Their analysis examines issue opinions from the 2004 election (ANES) in arguing that while most of the

American public is moderate, “there are sharp divisions between supporters of the two major parties that extend far beyond a narrow sliver of elected officials and activists” (Abramowitz and

Saunders 2005). They further suggest that Red and Blue state voters differ “fairly dramatically” in their social characteristics and political beliefs. This is consistent with Gelman et al.’s finding that there are apparently distinct cultures that inform the effect that aggregate and individual income (or class) levels have on partisan behavior and political beliefs (Gelman et al. 2005;

Abramowitz and Saunders 2005).

Abramowitz addresses the Fiorina argument suggesting weak sorting of the mass electorate by pointing out that, in fact, the American electorate is more educated and the educational process has resulted in an electorate more capable of understanding and using ideological concepts. This combined with elite polarization, providing a clearer set of ideological cues for the public to react to, has culminated in a more engaged, partisan, and polarized public.

This segment of the public is much larger than Fiorina and his colleagues define it. “In 2004, active citizens made up 46 percent of all Democratic identifiers and 49 percent of all Republican identifiers” (Abramowitz and Jacobson 2006). Only nonvoters, according to Abramowitz, are consistent with the characteristics Fiorina attributes to the mass electorate as a whole. Fiorina and Levendusky criticize Abramowitz’s polarization finding in suggesting it is an artifact of how he has coded the NES ideological scale (Abramowitz uses a recoded trichotomous ideology variable). Greater political activism is a consequence of greater party mobilization, ideological thinking is not on the rise in the American public, and, of course, “sorting” is not political polarization (Nivola and Brady 2006). Abramowitz suggests that changes in trends overtime

cannot be a consequence of recoding, since the same code is used in all years (Nivola and Brady

2006). These points of contention remain unresolved between these researchers, suggesting

further analysis is necessary to bridge their scholarly divide. Campbell and Cannon argue that

not only are the masses polarized today, they were polarized in the 1950’s when heterogenous

parties and centrist policies were the norm (Campbell and Cannon 2006). But it is not on

culture, as suggested by the culture warriors. Campbell and Cannon argue that “…it should be

remembered that for much of the twentieth century some of the most intense polarization in

nations around the world centered on the economic and political philosophy of Marxism”

(Campbell and Cannon 2006). Campbell argues there is direct evidence of mass polarization as

evidenced in the increase in polarization on the ideological measures in the NES. He finds a

decline in the “nonideologicals” in the mass electorate (Campbell and Cannon 2006).

SKEWERING THE CONVENTIONAL WISDOM

The Culture War skeptics present a strong case against the emergent polarization conventional wisdom. Fiorina points out that close elections do not connote a polarized electorate. Distinguishing political polarization from partisan sorting, he argues that the culture war is a myth. The American electorate has become increasingly centrist and moderate on social issues. DiMaggio et. al. fail to identify a polarization trend across a variety of cultural indicators, with the sole exception of abortion. And even there, Mouw & Sobel suggest that the

American electorate has not polarized on abortion. Bartels and Morton punch a hole in the hull of Frank’s culture war argument. Frank’s contention that economic and class divisions have receded in favor of social concerns takes on water in Bartels analysis, suggesting class remains and is in fact an increasingly predominant factor in partisan affiliation. Whether these arguments have sunk the culture wars thesis remains, however, an open and hotly contested question.

CHAPTER 3: RIGHTS & WRONGS & CULTURE WARS: TOWARDS AN UNCONVENTIONAL WISDOM ON POLARIZATION

“America is a divided nation. This is an indisputable fact. We are split on the war, split on abortion, split on the unions, split on what the Constitution really means and split on the role of religion in this nation.” - Cherry 2007

Having examined the evidence and arguments of the conventional wisdom and its critics, the task here is to assess where the polarization literature has gone wrong. There are problems with both the conventional consensus on the culture wars as well as the criticisms of that thesis. The research to date on polarization has poorly developed the concept of polarization, employed invalid and inaccurate measures of polarization, and failed to draw appropriate conclusions regarding polarization from the available data. A big part of this problem is the lack of consensus on what polarization is. Thus, the second task is to develop a rigorous conceptualization of political polarization: to determine what polarization is and what it is not. Third, I will develop empirical measures of polarization that will serve as the basis for the analysis in the subsequent chapters.

SECTION I: A POX ON ALL THEIR HOUSES: WHAT IS WRONG WITH THE POLARIZATION LITERATURE

Purple Politics: Getting Beyond Red vs. Blue

We have seen that there is evidence suggesting geographic distinctions between states and regions are indicative of territorial-based political cultures that have a significant part to play in the tale of political polarization. However, looking at political polarization solely from a

“Red vs. Blue” perspective misses a great deal of the empirical story. Indeed, the Red vs. Blue dichotomy maps poorly on to the true political changes that have wrought a more polarized political environment, requiring scholars to move past this false choice and examine the true

cultural, social, and political fissures that determine partisan politics as well as mass and elite political behavior.

The culture war proponents give us the argument that America has become more geographically polarized along cultural lines. On the other hand, there are those that suggest

American culture has moved away from distinctiveness towards a “standardization” that encourages similar perceptions of cultural issues throughout the country (Bowles and Pagano

2003). Homogeneity, in this instance, operates globally. We’ve all become more alike, period.

This ‘Americanization” of America is typified by chains such as Wal-Mart and Olive Garden where “the discreet hint of Tuscan decor and the passable wine list disguise the fact that there are 476 other Olive Gardens across North America, all with precisely the same menu”(Engel

2002). But is veal what we have in mind when we talk about cultural issues? The extent to which “Chain-America” and the influence of national popular culture produces cultural homogeneity ,or if it does, remains a controversial topic (Jia 2007). Less esoterically, there is the argument that most states are “purple” rather than red or blue. Furthermore, there is a great deal of partisan diversity even in the deeply red and blue states in the Union. Oklahoma,

Kansas, and North Carolina have Democratic governors, while California, and until very recently

New York and Massachusetts , has a Republican governor (Galston and Nivola 2006).

Then again, New England just defeated its last Republican hold out, Chris Shays, in the

November, 2008 election. Georgia (as is noted later once a bastion of Democratic power) has become one of the more consistently Republican states in America. The potentially confounding geographic impact on cultural and political views presents a problem to the study of political polarization. Important political polarization of the electorate may be masked in purely national-oriented studies, such as Fiorina’s examination of trends in views on abortion

Figure 3.1: Red, Blue & Purple – State Coded Composition, by Party, of the 110th United States Senate

http://en.wikipedia.org/wiki/File:110th_US_Congress_Senate3.PNG

and gay rights in the American electorate (Fiorina, Abrams, and Pope 2004). The contrasting trends of the homogenization of national culture versus and increasing heterogeneity between geographic regions and the cultures that manifest in those regions necessitates accounting for increasing or decreasing regional differences on social and other political issues. For example, it is possible for both trends (a trend towards homogeneity on the national level coupled with increasingly disparate views when the electorate is broken down in regional subcategories or reference groups) to be true. Any analysis of political polarization must account for trends in the body politic as well as changes in the grouping of that electorate that could yield polarizing factors independent of an overall centralizing trend. The American electorate as a whole could become more centrist on gay rights and abortion, while at the same time the reference groups and geographical ties that influence political opinion produce polarization in views between these groups and extreme views within them. So, while it is true that “Red vs. Blue” is an overly simplified and inaccurate measure of geographic polarization, this fact does not mean that geographic polarization hasn’t happened. The key is to sync the measures of geographic boundaries with the political cultures they serve as a proxy for.

The 50:50 Nation & Polarizing “Close” Elections: Being Right for the Wrong Reasons.

Journalistic treatments of the Culture Wars thesis have mostly expressed agreement with it, contributing to a perception of polarized politics irrespective of its empirical reality.

Their concurrence is based primarily on the ‘evidence’ of closely contested elections and in examining the culture war from an ‘up-close-and-personal’ standpoint. These stories are characterized by personal interviews of ideological characters and biopics on cities that illustrate the stark differences between conservatives and liberals and Republicans and Democrats— culturally, geographically, and on policy. “The polarization in US politics has both spawned and is being fed by a huge cultural rift that extends from the airwaves to the altar and from the

bookshelves to the bedroom. Alongside concrete issues like the economy, the war and healthcare, a bitter battle has emerged over individual lifestyle and moral values, oscillating between the personal, racial, social, sexual and religious, which is shaping the battleground for the forthcoming election” (Young 2004). As David Broder— that paragon of Beltway journalistic wisdom—warned, “This nation has rarely appeared more divided than it does right now”

(Broder 2000).

However as some scholars have noted, close elections, in and of themselves, are not evidence of polarization. A close election can be the product of a moderate, centrist and unimodal electorate just as much as it can be the result of a highly polarized and divided electorate. This does, of course, represent a divergence from traditional party positioning expectations given the central location of the median voter. Subjectively chosen communities that reflect stark partisan and cultural differences may make for good copy, but may not be representative of political, geographic, and social trends in the larger society (Wildermuth

2004). Jonsson notes Georgia’s remarkable adoption of the Republican party in a traditional

Democratic stronghold where John F. Kennedy secured his second largest margin of victory, in any state, over Nixon (Jonsson 2004). But is Georgia’s shift in to the Republican side of the ledger emblematic of a shift in the nature of national politics and the issues relevant to party identification? A trend of Republican victories is suggestive, but not dispositive. Partisan trends may be a consequence of polarization trends, however they are not necessary for nor sufficient evidence of polarization.

So while these accounts suggest to the reading public that there is a culture war going on and polarization is real, the primary evidence they marshal to demonstrate this fact may be misleading. Polarization may have indeed increased over the past several decades, but the data points commonly employed in popular accounts of polarization fail to make the case.

Size [Density] Matters: Dispersion across a Distribution vs. Average Location

Fiorina’s primary empirical evidence against polarization is based on time-series data of the average positions on a variety of cultural issues. He notes that abortion attitudes over the

‘polarization’ time period of the last thirty years has been relatively stable (Fiorina, Abrams, and

Pope 2004). Yet, as can be seen in Table 3.1, average policy positions in the electorate can be stable while the distributional characteristics shift drastically. One of Fiorina’s most important pieces of evidence against the Culture Wars thesis, just like the ‘close elections’ meme of the

Culture War proponents, doesn’t actually speak to political polarization. If we imagine Table

3.1 represents the distribution of opinion on a cultural issue such as abortion over time, then it is apparent that the distribution of opinion on abortion can radically change while the location of the mean and median of the distribution is stable over the three time periods. Clearly a

TABLE 3.1: ILLUSTRATION OF POLARIZATION NOT REFLECTED IN AVERAGE POSITIONS IN A HYPOTHETICAL POPULATION TIME PERIOD 1 TIME PERIOD 2 TIME PERIOD 3 Distribution of Pop. Distribution of Pop. Distribution of Pop. Population (n=8) Opinion on X Issue Opinion on X Issue Opinion on X issue Citizen 1 0 0 5

Citizen 2 0 1 5

Citizen 3 0 2 5

Citizen 4 0 5 5

Citizen 5 10 5 5

Citizen 6 10 8 5

Citizen 7 10 9 5

Citizen 8 10 10 5

MEAN 5.000 5.000 5.000 STAND. DEV. 5.345 3.780 0.000 Source: Compiled by the Author.

measure of polarization needs to account for the nature of the distribution of opinion on an issue and not merely the central tendency in an opinion dimension. Fiorina’s analysis that tracks the movement of the average issue positions of the electorate over time completely fails to account for the distribution of that opinion over that time period. Further we cannot assess whether the direction of a change in the average issue position is evidence of depolarization without knowing what change in the distribution of opinion on an issue that move reflects. A move to the center may be evidence of depolarization…it may also be evidence of polarization.

It isn’t about where you are. It is about where you’ve come from.

Sorting v. Polarization: Squares v. Rectangles

Fiorina argues that elite polarization is explained by the fact that activists polarize due to their attentiveness to politics, their high levels of information, and their sophisticated ideological outlook in comparison to an uninformed, uninterested, and unconstrained mass public, consistent with Converse’s evidence from a half-century ago. Rather than shifts towards the extremes in the mass electorate’s policy preferences, Fiorina and his colleagues argue that their choices have become increasingly polarized due to the polarization at the elite and activist level. Thus ideological conservatives have increasingly moved into the Republican camp and ideological liberals have moved into the Democratic camp as the parties at the elite levels have presented increasingly stark ideological choices in elections. This results in what Fiorina calls

“sorting” which, he argues, is distinct from polarization. “Some analysts prefer to refer to

[sorting] as ‘partisan polarization.’ We prefer the term ‘party sorting.’ Reserving the term polarization for bimodal distributions of opinion: voters are polarized on an issue if more of them cluster at the extremes than locate themselves in the center, or if they are moving from centrist positions toward the extremes” (Fiorina and Levendusky 2006). The sorting process that occurs absent preference polarization is illustrated in Table 3.1. Note that the number of

moderates in the electorate never changes nor does their choice to align as an “independent” rather than with one of the parties. What sorting entails is conservatives dealigning from the

Democratic Party and realigning with the Republican Party combined with the opposite shift in party allegiance among liberals. Thus the parties become more ideologically coherent while the distribution of opinion in the electorate remains unchanged.

Is sorting distinct from polarization? Fiorina’s distinction between sorting and polarization seems to hinge on what is being polarized rather than the distributional characteristics themselves. But polarization is generally applicable to any distribution, opinion or otherwise. Income, the attribute which economists have been primarily concerned with in defining and conceptualizing polarization, is not an opinion or issue dimension at all (Esteban and Ray 1994; Duclos, Esteban, and Ray 2004). Rather than defining sorting as a different species from partisan polarization, opinion polarization and partisan polarization (i.e. sorting) should be conceived as two types or classes of polarization. Just like a square and a parallelogram are both rectangles, so too is sorting and opinion polarization two kind of political polarization. The unifying symmetry between the variety of polarizations (issue, ideological, partisan, etc.), and thus the justification for this classification scheme, is that we can assess changes in their distributions, from polarized to depolarized, similarly. The particular attribute, be it partisanship, abortion opinion, or ideology, is irrelevant.

It follows that while the “Culture Wars” thesis is one form of political polarization, it is not synonymous with political polarization. As should be evident from the realignment discussion in Chapter 1 and the conceptualization of polarization outlined in Chapter 2, political conflict can occur across a plethora of dimensions of which the social issue dimension is just one. The New Deal involved economic polarization that produced the New Deal coalition and

the partisan divide that would characterize American politics for three generations. The Iraq war may have proved more polarizing in the ‘Oughts’ than abortion and gay rights. Perhaps a new issue of potential political conflict and distributional polarization will arise as a result of the financial crisis and recession. Certainly the November, 2008 election results do nothing to dissuade us from crediting such a possibility.

Elites & Masses: The Paradox of Elite Polarization

As noted in Chapter 2, Fiorina argues that the elites have polarized (or sorted) independent of the mass electorate, which has largely moved to the center on social issues. But

Fiorina’s argument regarding elite behavior presents a significant theoretical hurdle that he ultimately fails to clear. Political elites in a democratic republic must endure the judgment of the mass electorate periodically, and the United States is no exception. Hence Fiorina’s depiction of political elites as polarizing while the masses have centralized necessarily means that political elites have increasingly become unresponsive to the signals the mass electorate is sending. Fiorina speculates that primaries and ‘polarization entrepreneurs’ in Washington with big dollars behind them may explain this elite polarization independent of the masses.

However, both of those explanations are unsatisfying. Primary fights might explain some polarization during the primaries, but what about the general election? Given the unlikelihood of a sitting legislator losing a primary fight, why would these elites continue to respond to the subset of primary voters rather than the larger class of general election voters? Party is one answer. Given increasingly uncompetitive districts, the intra-district median would shift to the poles of the ideological distribution. Centrists from a national perspective would find competing for that extreme median voter difficult. Once in the general, the ‘extreme’ candidates rely upon the partisanship in the electorate to yield victory. As for the money in politics angle, how

does it make sense to raise a lot of money for re-election, only to lose that re-election because the candidate has wandered too far from his general election constituency?

One possible alternative explanation would be a polarized electorate. Where multiple modes exist, candidate divergence may be rational. The possibility of entry by a third party may cause parties to diverge from the median in order to discourage a third party challenge on their extremes in a polarized electorate(Fiorina 1999; Palfrey 1984). Indeed, entry at the candidate level might explain polarization at the congressional level where moderates may vote with hardliners in a move anticipatory of a third party challenge from hard-core ideologues in their districts such as challenge endured by Joe Lieberman for his pro-war stance (Fiorina 1999).

Hinich and Munger develop a theory of ideology which permits party divergence (Hinich and

Munger 1994). Incorporating previous studies that have pointed to incomplete information and uncertainty in voter policy locations and candidate locations, they argue that the creation and maintenance of an ideology by parties is a necessary component of political competition, given the traditional MVT model’s problem of representing choice in elections as having the same fundamental structure of economic decisions (Alverez and Franklin 1994; Bartels 1986; Calvert

1985; Enelow and Hinich 1981; Ferejohn and Noll 1978; Shepsle 1972; Wright and Goldberg

1985; Alverez 1997; Calvert 1986; Dalton, Beck, and Huckfeldt 1998; Feddersen and Pesendorfer

1999, 1997; Huckfeldt and Sprague 1988; Huckfeldt and Sprague 1987; Hinich and Munger

1994). In a political environment where Republicans have become much more consistently and strongly conservative (likewise for Democrats and liberals) vote-seeking parties rationally diverge in order to create a credible ideology which they can ‘sell’ to their constituents.

Establishing an ideological flag at one of the ‘poles’ in a bimodal distribution would account for the divergence Fiorina fails to adequately account for.

A ‘compassionate conservative’ with a centrist agenda may have to tack right in the political environment of 2004, whereas he could comfortably govern from the center in the

1970’s. This raises an interesting theoretical question: how polarized must the electorate be in order for candidates to diverge? The two possibilities that Fiorina mentions do not exhaust the possible distribution of voters. Other distributions exist which may have a significant numbers of voters located in the center of the distribution and yet still permit divergence, as Downs himself noted. But clearly Fiorina has a serious theoretical problem. He has adopted a

Downsian proximity model to explain the increase in the ‘weight’ associated with the cultural dimension; however his candidates behave in a decidedly un-Downsian fashion.

What mechanism produces candidate divergence (clearly contrary to classic Downsian logic) is an ambiguity in Fiorina’s initial argument, though he has attempted to account for it in later discussions (it remains “puzzling” to him despite his speculative explanations). He does not adopt Rabinowitz’s directional theory which calls for candidate divergence, yet his interest- group responsive model of party divergence on the issues irrespective of centripetal voter preferences may require it. Fiorina argues that electoral pressures on candidates have weakened while candidates have become more beholden to advocacy groups (with views on issues decidedly distant from the mass electorate’s mean). “The result is political activists and candidates whose ideological commitments run deeper than a generation ago, whose fear of losing elections is less than a generation ago, or both” (Fiorina, Abrams, and Pope 2004). It is a somewhat controversial position to take, and it has some apparent problems from a theoretical perspective and in terms of the empirical evidence (McKelvey and Ordeshook 1985, 1986;

Morton 1993; Palfrey and Poole 1987; Hinich and Munger 1994).

It is difficult to assess the relative strength of the electoral connection over the past 50

years, however even if the ‘risk’ of loosing for incumbents has been lessened through a variety

of factors, even Fiorina does not suggest it is not a or even the ‘prime mover’ of candidates for

office. Fiorina accounts for the ‘irrational candidates’ implication of his argument by suggesting

that as long as parties diverge a relatively equal distance from the mean, then voters cannot

punish the candidates of those parties for adopting stances away from the center of the

electorate distribution. The problem with this argument should be apparent: a party need

merely move some upsilon towards the mean in order to secure electoral success. It is difficult

to see how even if parties have become more beholden to advocacy groups that this would be

powerful enough to overcome the obvious strategy: move towards the center and win.

However, with an increasingly polarized electorate, parties, candidates, and political elites could

diverge in response without having to postulate irrational elites beholden to the special

interests.

SECTION II: POLITICAL POLARIZATION AS A CONCEPTUAL PROBLEM: FORMAL & EMPIRICAL FOUNDATIONS

“As the struggle proceeds, ‘the whole society breaks up more and more into two hostile camps, two great, directly antagonistic classes: bourgeoisie and proletariat.’ The classes, polarize, so that they become internally more homogenous and more and more sharply distinguished from one another in wealth and power” (Deutsch 1971)

Justice Potter Stewart, tasked with imposing a definition of obscenity in order to rule on the constitutionality of a fine imposed on a filmmaker for showing the French film The Lovers, said of obscenity that he could not intelligibly define it but that “I know it when I see it”

(Jacobellis v. Ohio 1964). It is tempting to do the same with polarization. Polarization is relatively easy to visualize but much more difficult to define. Much of the empirical discussion of political polarization has glossed over what polarization is. There is a great deal of references to polarization, but few efforts to rigorously define it. Some conceive it as increasing extremism

in the electorate or among social groups on issues, ideology, partisan and electoral choices.

Others look to election results.

Polarization Nuts & Bolts: Necessary and Sufficient Components

Polarization is a change in the distribution of opinion in a population or between

groups over an attribute such that the distribution contains more mass at the ‘poles’ relative

to unanimity on the attribute or a distribution with less mass at the poles of the distribution

(i.e. bimodality). Polarization is at its essence a relative concept. Much like one cannot define

“larger” without reference to something smaller with which to compare, polarization must be conceived relative to something. In order to properly conceptualize polarization, let’s first talk about the necessary ingredients for polarization. The minimum necessary components of polarization are 1) an attribute, 2) a population, and 3) a distribution. We start with an attribute. By attribute I merely mean some characteristic (belief, position, identity, etc.) which an individual, institution, organization, or even a population can have. A political issue is thus an attribute. It could be an attribute of policy, of a group, of an individual, of the aggregate electorate. The possibilities are infinite. The Culture War literature has mostly dealt with social issues, however any attribute could potentially be a source of polarization. Foreign policy, government spending, taxes, welfare policies, etc. are all potential issues which can be attributes.

The second condition is a population. At minimum, we must have at least two individuals in order to talk about polarization. While internal conflict is real, it is difficult to be polarized from one’s self. As a consequence, we must have a population in order to talk about polarization. Whether it is two individuals on a desert island or the citizenry of the United

States, polarization requires there be at least theoretically more than one position on the attribute.

Third, In order to have polarization, we must have a distribution. There must be a spread of points in relation to the attribute. Polarization suggests poles, and poles by definition must have separation. This observation has an important implication: the absence of polarization is the absence of a distribution. If we imagine a population of, say, 1000 individuals and every one of these individuals prefer exactly X amount of government spending, then there is no polarization of views on the amount of government spending in that population. Hence the ultimate reference point for polarization is its opposite: unanimity. While unanimity in a polity is rare (if it even exists), it provides a theoretical maximum by which we can compare other distributions (contemporaneous or over time) and assess polarization. We can thus compare multiple distributions in terms of their proximity to this theoretical maximum. Given three distributions, the distribution closest to ‘unanimity’ is the least polarized distribution.

Political Polarization: From Consensus to Conflict

Political polarization, as conceived here, is a phenomenon dependent on the distributional properties of aggregate political opinion in the American electorate, among groups relevant to political competition, and elite political actors which include government officials and opinion-makers. In other words, polarization as a political concept is the relative distribution of opinion in a politically relevant population or between politically relevant groups along either single or multiple issue or partisan dimensions. When we talk about a “polarized” opinion distribution in a static sense, we are contemplating the distribution of opinion relative to a “theoretical maximum” (DiMaggio, Evans, and Bryson 1996). Polarization as a process that occurs over a time period refers to the change in the distribution of opinion relative to this maximum or some other distribution (say, a previous time period) over some specified period of time (DiMaggio, Evans, and Bryson 1996).

One way to visualize polarization is to imagine three static distributions of opinion on an issue (the issue itself is not important). As can be seen in Figure 3.2, at one static stage and at the opposite end of “polarized,” indeed, the absence of polarization, is unanimous agreement.

The closest analog to the unanimity hypothetical in the real world of politics might be the issue of a “republican” form of government in the United States. Opinion on this issue is almost uniformly in favor. There is no ‘opposition’ to speak of. In the second state or stage of polarization we have an approximately ‘normal’ distribution of opinion on an issue, with most opinion located near the center with diminishing frequencies as you move in to the tails of possible ‘opinions’ on the issue (Figure 3.3). A relatively normal distribution of opinion exist might exist in regards to environmental regulation, with relatively few individuals wanting no environmental regulation on the one side of the distribution and relatively few individuals wanting total regulation of the environment on the other side. The third ‘stage’ of polarization is where relatively equal distributions of opinion are located at the “poles” of opinion on the issue and very little of the distribution is located in the center (Figure 3.4). An example of

“polarized” opinion in this sense might be the distribution of opinion among Palestinians and

Israelis in regards to the state of Israel and its role in the Middle East. The three static ‘states’ of polarization mentioned above are but three among an infinite continuum of possible distributions (see Figure 3.7 for two examples) with polarization as a process is conceived as movement over time across that continuum from “unanimous” to “polarized” opinion (Figure

3.5).

Whether discussing polarization from a dynamic or static disposition, defining a polarized distribution in such an instance requires that we reference another, less polarized, distribution. Figures 3.2, 3.3, and 3.4 illustrate three different kinds of archetypical distributions. Figure 3.2 illustrates what we might call a ‘consensus’ or essentially non-polarized

Figure 3.2: Hypothetical Policy Consensus

Policy A distribution of voter ideal points Policy A’ distribution of voter ideal points

POLICY A POLICY A’

Figure 3.3: Hypothetical Normal Policy Distribution

Distribution of voter ideal points

POLICY

Figure 3.4: Hypothetical Bi-Modal Policy Polarization

Distribution of voter ideal points

POLICY

Figure 3.5: Hypothetical Policy Multiple & Uniform Modal Non-Consensus

Policy A distribution of voter ideal points Policy A’ distribution of voter ideal points

POLICY A POLICY

distribution where most of the population ‘agrees’ on that particular policy (if we are assessing the distribution of opinion on a particular policy) and thus exhibits little to no spread along the opinion dimension. Figure 3.3 illustrates a ‘normal’ distribution of opinion on a policy where a predominant ‘modal’ preference on the policy is apparent in the population but substantial disagreement over where policy should be located exists in society and, indeed, the majority of the population have preferences located somewhere along the policy dimension other than at the modal or median position. Figure 3.4, finally, illustrates a bimodal distribution of policy where the center of the opinion dimension has been vacated and there are two relatively well structured groups of the population that exist some distance from one another on the policy dimension. The fact that the distribution in Figure 3.2 is less polarized than the distribution in

Figure 3.3 and the distribution in Figure 3.4 is the most polarized of the distributions is unambiguous. Indeed, the distribution of attitudes can take on an infinite number of different shapes (see Figure 3.5 for several examples).

In politics, attributes which have a distribution of opinion at unanimity or near unanimity don’t lend themselves well to the political process. This is especially so in the United

States, where the bar for successful partisan competition set by our first-past-the-post electoral system is set much higher than in proportional systems. There is little reason for a candidate or party to adopt a position in opposition to a unanimous or near unanimous position, as doing so could carry with it a penalty of lost elections and sapped strength in American political institutions. Not coincidentally, individuals and groups with beliefs and positions that run counter to a unanimous or near-unanimous position have difficulty getting access to the policy process. Parties ignore them. And attempts to organize politically independent of established parties and organizations run smack into Duverger and his law.

However, sometimes consensus positions—as a function of exogenous shocks,

demographic and population shifts, or merely the vagaries of time—become non-consensus

positions in the American electorate. This process of moving from consensus, where most

people agree on an issue, to conflict, where a substantial portion of the public disagree on an

issue, is at the heart of political polarization. When the political dynamic on an issue changes

such that there is a substantial portion of the American public in opposition to the rest of the

citizenry, this polarization is ripe for political conflict (see Figure 3.4). Polarization doesn’t

necessarily imply political conflict, but it is a necessary condition of it.

Political Polarization Requires Partisan Conflict

Political polarization requires partisan conflict. By partisan I do not mean it in the strict

sense of political parties in conflict with one another on some political issue (though that

certainly counts), but rather in the more general sense of group conflict. I have defined

polarization as a shift of mass in a distribution of opinion in a population or between groups

towards the poles (or away from the center) and have defined this in the political context in

relation to the absolute maximum of ‘unanimity’ on some political issue. Polarization occurs

when opinion on an issue moves from consensus to non-consensus. Absolute polarization is

where we get two masses of the population (or two groups) at the opposite ends of the extreme

poles of the continuum of possible positions on an issue. But polarization on some issue is

insufficient to produce conflict. It is a necessary but insufficient condition of political conflict.

In order for polarization to matter, politically speaking, then this polarization must be

galvanized as a partisan issue over which groups and/or parties compete and conflict within the confines of the political environment. There are a host of issues over which the American electorate is polarized, but which do not influence partisan choices, are not a subject of the policy process, and do not inform opinions on candidates, parties, and the political system.

Figure 3.6: Consensus to Conflict – Intra-Policy or Inter-Policy Polarization Over Time

POLICY A or An

T1 T2 T3

There are strong polarizing divisions in the American public over the Yankees, over the movie

Titanic, over the choice of Kris Allen as the next American Idol, but these polarizing topics are not a subject of partisan conflict. Relevant political polarization, or political polarization that we care about, is that which inspires and galvanizes groups to act politically and the American public to choose candidates and affiliate with parties, in part or in whole, as a consequence of where those candidates and parties stand on that issue. Political polarization on social issues thus suggests 1) the American public has shifted from a relative consensus on some social issues to a situation of non-consensus and 2) that the parties and political groups have adopted positions and engaged the policy process on that issue.

Relevant political polarization, the kind of polarization that can influence the probability of moderate group formation and the potential for compromise, is not only along the issue dimensions but also the variable salience of issues over time. Mere polarization of an issue does not in and of itself consist of a problematic condition for political compromise or moderation. A polarized issue that is not a subject of political competition lacks political relevance. While single issue and aggregation of issues can be assessed empirically, the existence of polarization is insufficient for a politically relevant cleavage. As noted in the realignment literature, the issue must not merely be cross-cutting and conflictual but also *salient* to the electorate in order to have significant social impact. A further implication of political polarization is that individual issue distributions need not become more polarized for partisan conflict and partisan polarization to occur. Rather than the underlying distribution of opinion on an issue shifting to the extremes, an issue that has a polarized distribution, which heretofore had not been a relevant political issue, may become politically salient.

SECTION III: MEASURES OF POLITICAL POLARIZATION

Vizzini: “HE DIDN'T FALL? INCONCEIVABLE!” Inigo Montoya: “You keep using that word. I do not think it means what you think it means.” – The Princess Bride

Having thus defined polarization as a concept, we must next move to a formal, empirical definition of polarization that can be tested. These empirical measures need to capture the fundamentals of polarization: distribution, polar location, and group conflict. The dispersion principle is simply that the more dispersed political opinion is in the aggregate, the more difficult it will be for the system to produce centrist / moderate policies. The bimodality principle, or polar location, suggests that to the extent that political opinions coalesce around two distinct poles, the greater the difficulty in producing centrist / moderate policies. The third principle, consolidation, suggests that the degree to which different opinions become more closely associated within groups, then the more intractable political competition is (Blau 1977;

DiMaggio, Evans, and Bryson 1996; Converse 1964; Blau 1977).

Dispersion. In order to measure dispersion, we need a measure that both reflects the relative distance that individual respondents differ from one another as well as taking into account the proportion of opinion located in the extremes relative to the center of the distribution. The traditional measure of dispersion (or inequality in the economics literature) is variance (or its cousin, standard deviation). As opinion dimensions become more polarized, variance should increase. Another related measure of dispersion is the coefficient of variation, which is a calculation of the standard deviation relative to the mean.

Bimodality. This principle reflects the underlying conceptualization of polarization. The

absolute polarized distribution is an extreme bimodal distribution: where exactly 50 percent of

the population is located at one extreme and the other 50 percent is located at the other

extreme. Again, a bimodal distribution is ripe for political conflict, given the implicit reduction in

the probability of centrist policies securing the support of compromising majorities. Here I use a measure of kurtosis and conflict to capture the ‘bimodality’ of opinions on political issues. I discuss these measures in detail later in this chapter as well as Chapter 4. Kurtosis, while an imprecise measure of the shape of the distribution, is sensitive to changes in the shape of the distribution and, consequently, correlated with changes in bimodality. The consensus measure provides a way of examining the dichotomous distribution of opinion on an issue relative to the

50/50 maximum polarization standard.

Consolidation. The consolidation measure refers to the relative agreement or consensus within groups and their relative disparity across groups. This “identity group” polarization is measured using a difference of means along issue dimensions between the groups to assess between group differences while we use the variance and kurtosis measures to assess within group consolidation on issue dimensions. This difference of means is captured in the group polarization measure.

Deutsch’s description of Marxian theory on social conflict in society provides an insight into how to conceptualize polarization. As Esteban and Ray argue, the two primary aspects of polarization are identification and alienation. Polarization is characterized by increasing identification with those similar to oneself along some relevant attribute coupled with increasing alienation from those dissimilar to oneself along that same attribute. Stated explicitly, there are three features of polarization identified by Esteban and Ray: 1) there must be a high degree of homogeneity within each group. 2) There must be a high degree of heterogeneity across groups. 3) There must be a small number of significantly sized groups

(isolated small groups or individuals are irrelevant) across the attribute dimension (Esteban and

Ray 1994). An empirical measure of group polarization must incorporate a measure of group

size and a measure of the ‘antagonisms’ or distances from all the other groups for a given policy,

issue, or ideological dimension. The measure developed here (GP) incorporates both aspects of

polarization.

Equation 3.2: Empirical Measure of Group Polarization

= ௠ { ( )[( ) + ( ) …+( ) ]} ଶ ଶ ଶ ܩܲ ෍ ݊ீଵ ͳെ݊ீଵ ܫீଵҧ െܫீଶҧ ܫீଵҧ െܫீଷҧ ܫீଵҧ െܫீ௠ҧ + {ீୀଵ ( )[( ) + ( ) …+( ) ]} … ଶ ଶ ଶ + { (1 ݊ீଶ ͳെ݊)[ ீଶ ܫீଶҧ )െܫ+ீଵҧ ( ܫீଶҧ െܫ) +ீଷҧ ( ܫீଶҧ )െܫ… +ீ௠ҧ ( ) ]} ଶ ଶ ଶ ଶ ீ௠ ீ௠ ீ௠ҧ ீଵҧ ீ௠ҧ ீଶҧ ீ௠ҧ ீଷҧ ீ௠ҧ ீ௟ҧ Where:݊ െ݊ ሺܫ െܫ ܫ െܫ ܫ െܫ ܫ െܫ

= the average position on for Groups 1 through M. … = the proportion for Groups 1 through M. ܫீଵҧ ǥܫீ௠ҧ ܫ ீଵ ீ௠ This݊ measure݊ of group polarization calculates the total distances between the defined groups in a particular policy, issue, or the ideological dimension and weights the summation of those distances by the sizes of the defined groups. This definition is consistent with the theoretical discussion of polarization mentioned above, as the maximal polarization in this measure would involve society dividing itself into two groups with the two groups locating themselves at the extremes of the relevant dimension. As the number of groups increases, necessarily the size of the groups decreases, and thus polarization declines. Also, if the groups move towards each other in the relevant dimension, polarization declines. One asset of this measure is that, even if the number of groups is over-defined (say, we have created foure categories of group but, in terms of their relative location, there really is only three groups) it will not affect the group polarization measure. Let’s say that Group A and Group B have almost identical positions on the attribute. If that is the case, the relative distance between them

approaches zero and correspondingly counts little towards the overall polarization coefficient.

Let’s consider a hypothetical example:

Table 3.2: Hypothetical Example of Partisan Polarization on Abortion AVERAGE PARTISAN GROUP PARTY ID PROPORTION POSITION ON ABORTION Republican .33 4.33 Independent .12 3.15 Democrat .44 1.89

Table 3.2 shows canned data from a hypothetical sample of the American electorate.

The proportions reported are the hypothetical percentages of the party identifiers in the

sample. The average position on abortion is on a hypothetical 7 point scale on support for

abortion rights. Recall Equation 3.2. Group polarization is equal to the sum of the squared

differences between the groups on the relevant dimension weighted by the size of the groups.

Here, the average partisan group positions on abortion are weighted by their proportion in the

sample. Recall the equation for the group weights: (1 ). The group weights for our

ீ௠ ீ௠ partisan groups in the hypothetical example are as follows:݊ െ݊

Table 3.3: Group Size Weights for Hypothetical Partisan Groups

Weight Equation Weight Calculation Weights ( ) .33 (1-.33) 0.221

࢔࢘ࢋ࢖(૚ െ ࢔࢘ࢋ࢖) .12 (1-.12) 0.106 ࢔࢏࢔ࢊ (૚ െ ࢔࢏࢔ࢊ ) .44 (1-.44) 0.246 ࢔ࢊࢋ࢓ ૚ െ ࢔ࢊࢋ࢓

The calculations for the partisan groups result in weights that reflect the size of the

groups within the sample population. Thus the distances between Independents and the other

partisan groups will not figure as prominently in the polarization calculation as the distances

between Republicans or Democrats and the partisan groups, as Independents are the smallest

Table 3.4: Hypothetical Partisan Individual Group Polarization Scores

PIG Polar PIG Polarization Equation PIG Polarization Calculation Score 0.221[(4.33 – 1.89)2 + (4.33 – 3.15)2] 1.620 RP = WR + ૛ ૛ ത࢘ࢋ࢖ തࢊࢋ࢓ ത࢘ࢋ࢖ ത࢏࢔ࢊ 0.106[(3.15 – 4.33)2 + (3.15 – 1.89)2] 0.316 IP = WI ቂ൫ࡵ െ ࡵ ൯+ ( ൫ࡵ െ ࡵ ) ൯ ቃ ૛ ૛ ത࢏࢔ࢊ ത࢘ࢋ࢖ ത࢏࢔ࢊ തࢊࢋ࢓ 0.246[(1.89 – 4.33)2 + (4.33-3.15)2] 1.807 DP = ቂW൫ࡵD െ ࡵ ൯ ࡵ+ (െ ࡵ ቃ ) ૛ ૛ ቂ൫ࡵതࢊࢋ࢓ െ ࡵത࢘ࢋ࢖൯ ࡵതࢊࢋ࢓ െ ࡵത࢏࢔ࢊ ቃ

group in the sample. Likewise, the Democrat distances will have the largest influence on

polarization as they are the largest group in the sample, assuming they are sufficiently distant

from the other groups. The analytic improvement GP represents is that it accounts for group

size and distance at the same time in a composite score for political polarization.

Table 3.4 shows the distance calculations summed for each partisan group, weighted by

the group size. The GP for partisan groups on abortion in our hypothetical example is, summing

across the partisan individual group polarization scores, 3.745. The score calculation (1.620 +

0.316 + 1.807 = 3.745) sums the weighted group polarization scores for a single measure of

polarization for that dimension. Again, note that the distances for the Independent group count

very little towards the partisan polarization score. Small groups contribute little to polarization

(if they contribute at all) as they are of an insufficient size in order to generate societal conflict

that is likely to have an impact on a national scale. The result is polarization scores for each

group and a total polarization score for all groups in the policy, issue, or ideological dimension of

interest.

Dynamic and the Static: Polarization vs. Polarized

As is apparent from the above discussion, there are two aspects of polarization, one

static and the other dynamic. Polarization as a process involves changes in a distribution of an

attribute or a meta-distribution of multiple attributes over time. Polarization as a state is, as

noted in the discussion of Esteban and Ray and Duclos and Esteban’s polarization measures, involves the distribution of an attribute or a meta-distribution of multiple attributes relative to an idealized non-polarized state or, to put it another way, a theoretical maximum. Here we might contemplate such a state as ‘total consensus’ or ‘unification’ where every member of the population is at an identical point. Hence I would argue that Figure 3.2 is a non-polarized distribution relative to the distributions in Figures 3.3, 3.4, and 3.5 because it is closer in characteristics of identification and alienation to the total consensus distribution of the population where everyone is located at a single point. Conversely, the distribution in Figure 3.4 is ‘polarized’ given it reflects the furthest diversion from that consensus of these distributions and is most illustrative of the bimodality principle. With GP we have a measure that can be used to assess static polarization between groups on a policy, issue, or ideological dimension (it would work in affect / valence dimensions as well). It can also be tracked over time to assess the dynamic process of polarization or its converse: depolarization.

Measures of Political Polarization: An Empirical Assessment

Having identified the measures of polarization based on a conceptualization of polarization derived from that developed by Esteban and Ray and later extended by Duclos and

Esteban, I explore the ability of these measures to describe the class of distributions that are relevant to political conflict. Specifically, I examine the measures of polarization we’ve identified to date and illustrate their utility in measuring polarization. I also demonstrate the distinctiveness of these measures as well as address how well “kurtosis” works as a bi-modality proxy (or, to put it another way, whether kurtosis means what we think it means when it comes to ‘peakedness’). Using canned data compiled by the author (see Figures 3.7A and 3.7B), seven different distributions are identified with measures of central tendency, dispersion (measure of alienation), and peakedness (a proxy for identification). The nine distributions provided here

are of empirical and theoretical interest. Distribution 1 is a flat distribution, similar to the hypothetical distribution considered by Esteban and Ray and as illustrated in Figure 3.4.

Distribution 2 is an example of a bi-modal distribution where the modes are located at the absolute extremes of the attribute in question. This distribution represents the ‘best case’ for polarization and is the limit of polarization identified by Esteban and Ray (Esteban and Ray

1994). Distribution 3 presents a bi-modal distribution, but the modes are located very close to the center of the attribute. This point illustrates the importance of alienation (in addition to identification) in conceptualizing polarization. While this distribution consists of two distinct poles, they are likely too close together on the attribute to provide much of an opportunity for political conflict. However, we cannot rule out the possibility. While the ‘identification’ of

Distributions 2 and 3 are exactly the same, Distribution 2 is a much more fruitful venue for political conflict and represents a more ‘polarized’ distribution on the attribute. Distribution 4 is an example of a ‘consensus’ issue polarization and thus illustrates the opposite of

‘polarization.’ Distribution 5 illustrates the case demonstrated by Esteban and Ray where a very alienated group lacks significant identification due to its size. They argue that the existence of such groups should not lend itself to significant social conflict given their low level of identification. Distribution 6 is the classic normal distribution where the bulk of the distribution is centrally located and it has a unimodal shape. Distribution 7 is a tri-modal distribution with three equally sized groups of the population density at equidistant points. Distribution 8 is a constant. Distribution 9 is a pure close bi-modal distribution, similar to Distribution 3 except that there are only values at 4 and 6 in this distribution, with the population equally distributed between them. Since the distributions employ values from the same range (1-10), direct comparisons of kurtosis and skewness are possible.

First, as noted in Table 3.1, it should be apparent that distributional characteristics are independent of measures of central tendency. In all but the small outlying group distribution, the means of each of these unique and radically different distributions is relatively the same

(around 5). Secondly, as Dimaggio et al. note, kurtosis is distinct from skewness. Kurtosis is positive when the distribution is concentrated and unimodal, indicating consensus. The highest value of kurtosis among the classes of distributions presented here is the ‘consensus’ found in

Distribution 4 (K = 7.814). Kurtosis is not sensitive, as skewness is, to the location of the peak of the distribution. Kurtosis becomes negative as distributions become more flat and then even further negative as they move towards bimodality. This can be seen in comparing kurtosis for

Distribution 1, a flat distribution (K = -1.225), with the kurtosis for Distribution 9, a pure bimodal distribution (K = -2.041). Both distributions have negative kurtosis, but the bi-modal distribution has a more negative kurtosis. A comparison of Distribution 3 and Distribution 7 shows that the kurtosis values conform to expectations when there are multiple groups or modes within the population. The tri-modal distribution exhibits less dispersion (V = 12.737) than that of the extreme bimodal distribution seen in Distribution 2 (V = 18.432) and a lower comparative kurtosis value (K = -1.434).

However, the kurtosis measure appears to have some rather serious problems as a measure of bimodality. For example, Distribution 2 and Distribution 3 are the same in terms of bimodality yet they have distinctly different kurtosis scores. Indeed, according to kurtosis, the normal distribution is more ‘bimodal’ than is Distribution 3. Kurtosis scores are nearly identical between the normal distribution found in Distribution 6 and the highly non-normal and skewed grouping in Distribution 5. The flat distribution in Distribution 1 has a lower kurtosis than the normal distribution in Distribution 6. Clearly Mouw and Sobel had good reason to dismiss kurtosis as a measure of political polarization and the DiMaggio study results on the bimodality

Figure 3.7A: Various Distributions (1-6) of Attribute X, N = 100

Distribution 1: Flat Distrubtion 4: Consensus 100 FLAT 100 CONSENSUS 90 90 80 = 5.5 80 = 5.05 70 70 60 = 2.887 60 = 1.438 50 ܺത = 8.333 50 ܺത = 2.068 40 40 Frequency Frequency ߪ 1.225 ߪ = 7.814 30 30 20 ܸ = 0 20 ܸ = 0.954 10 ܭ ൌ െ 10 ܭ 0 ܵ݇ 0 ܵ݇ 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 Attribute X Attribute X

Distribution 2: Extreme Bi-Modal Distribution 5: Small Outlying Groups EXT BI- SMALL 100 100 90 MODAL 90 OUTLIER 80 80 70 70 60 = 5.45 60 = 8.1 50 = 4.293 50 = 3.377 40 ܺത 40 ܺത Frequency Frequency = 18.432 = 11.404 30 ߪ 30 20 1.926 20 ߪ = 0.212 10 ܸ = 0.035 10 ܸ 1.912 0 ܭ ൌ െ 0 ܭ െ 1 2 3 4 5 6 7 8 9 10 ܵ݇ 1 2 3 4 5 6 7 8 9 10 ܵ݇ ൌ െ Attribute X Attribute X

Distrubtion 3: Close Bi-Modal Distribution 6: Normal Distribution CLOSE BI- NORMAL 100 50 90 MODAL 80 40 = 5.5 70 = 5.05 = 2.035 60 30 ത 50 = 1.696 ܺ = 4.131 40 ܺത 20 ߪ

Frequency 0.161 Frequency = 2.876 30 ߪ ܸ 20 = 1.752 10 = 0 10 ܸ = 0.554 ܭ ൌ െ 0 ܭ 0 ܵ݇ 1 2 3 4 5 6 7 8 9 10 ܵ݇ 1 2 3 4 5 6 7 8 9 10 Attribute X Attribute X

Figure 3.7B: Various Distributions (7-9) of Attribute X, N = 100

Distribution 7: Tri-Modal Distribution 8: Constant 100 TRI-MODAL 120 90 CONSTANT 80 = 5.3 100

70 = 3.569 80 60 = 5 ܺത = 12.737 50 60 = 0 40 ߪ 1.434 Frequency Frequency 40 ܺത = 0 30 ܸ = 0.157 20 ߪ ܭ ൌ െ 20 = . 10 ܸ = . 0 ܵ݇ 0 ܭ 1 2 3 4 5 6 7 8 9 10 1 2 3 4 5 6 7 8 9 10 ܵ݇ Attribute X Attribute X

Distribution 9: Pure Close Bi-Modal 100 90 80 PURE CLOSE 70 BI-MODAL 60 = 5 50 40 = 1.00 Frequency 30 ܺത = 1.01 20 ߪ 10 = 2.041 0 ܸ = 0 1 2 3 4 5 6 7 8 9 10 ܭ െ

Attribute X ܵ݇

principle are hence suspect. A comparison of Distribution 3 and Distribution 9 further illustrates the potential problem of using kurtosis as a proxy for bi-modality. One problem with using kurtosis as a measure of bi-modality is that it has difficulty assessing bimodality when the poles or modes are close to one another, such as the two-tailed gamma distribution(Moors 1988;

Finucan 1964; Balanda and MacGillivray 1988; Mouw and Sobel 2001; Groenveld and Meeden

1984; Darlington 1970; Hildebrand 1971; Kaplansky 1945; Moors 1986). Note that only 10% of the population ‘changes’ from Distribution 9 to Distribution 3.

The difference in the shape of the distributions is that I have ‘pinned’ the tails of the distribution to the extremes in Distribution 3. However, both of these distributions are essentially bi-modal and the modes are located in exactly the same place. Yet, note the radical difference in kurtosis. In Distribution 9, the kurtosis measure correctly identifies this as a bimodal distribution (K = -2.041). However, in Distribution 3, the kurtosis measure treats the two close bi-modal density points in the distribution as one, resulting in a kurtosis reflective of more of a ‘consensus’ or unimodal distribution. Contrast the kurtosis from Distribution 3 (K = 2.652) and that of the normal distribution illustrated in Distribution 6 (K = -0.165) and Distribution 4 (K

=7.814). This measure places Distribution 3 somewhere between the consensus distribution illustrated in Distribution 4 and the normal distribution in Distribution 6. Clearly it would be better to treat Distribution 3 as similar to that of Distribution 9, but the kurtosis measure does not. Kurtosis is suspect as a proxy for bimodality, though it may serve for relative comparisons.

Finally, as should be apparent from these distributions, kurtosis is distinct from variance.

As argued by Esteban and Ray as well as Dimaggio, dispersion (or alienation) is analytically distinct from bimodality (identification). This is best shown in a comparison of Distribution 3, where the distribution is distinctly bimodal but where most of the distribution is proximately

located (V = 2.876), and Distribution 2 where the bimodal distribution is identical but there is much greater dispersion of the population across the attribute (V = 18.432). Note that the kurtosis for both of these distributions is relatively the same.

These measures of identification and alienation will permit an empirical analysis of relevant social, foreign policy, and economic dimensions operative in the political space that assesses trends in polarization over time. While problematic in distinguishing certain kinds and gradations of distributions, kurtosis serves as a rough measure of the bimodality of a distribution. Over time, a shift towards a more negative kurtosis on an opinion distribution is indicative of increasing polarization over that time period. Increasing variance over time suggests that opinion has become less consensual (decreasing identification, increasing alienation), and hence also serves as a marker for polarization. The group polarization measure will permit the assessment of the degree to which politically and socially relevant groups have consolidated or separated on a particular issue, partisan, etc. dimension. Each measure provides a window on the nature of and changes in the distributions of political opinion in the mass electorate. Those changes (or lack thereof) will tell the tale of a cultural war in the

American public, or show it to be the myth that some skeptics believe it to be.

CHAPTER 4: EMPIRICAL MEASURES OF POLARIZATION

The first set of principles, measures, and models—specifically mean trends, bimodality, and dispersion—are employed to test trends in polarization across a plethora of issue and political dimensions in Chapter 5, 6 and 7 and are set out here in Section I. The second set of measures, the group polarization measures, assess the degree to which groups have become polarized on issue, political, and partisan dimensions. These measures are used in Chapters 8 &

9, and I report them in Section II. Chapter 10 includes a set of distance measures used to assess the relationship between mass polarization and elite polarization from the perspective of the citizens themselves. Chapter 11 assesses the mass and elite polarization relationship using measures of elite ideology derived from actual behavior rather than perceived placements.

Since the perceived distance measures created from the perceived placements of the parties and candidates on issues by respondents to the ANES relative to their own self-placements on the issue scales are set to work only in Chapter 10, I include the discussion of those methods to that chapter. Similarly, I leave the discussion of the models and measures used exclusively in

Chapter 11 to that chapter. Specific refinements of the models are detailed in the chapter or chapters in which that measure or model is used for data analysis. In the first set of measures I develop a method for assessing consensus and conflict, for measuring bimodality and dispersion in opinion distributions on issue dimensions, and for examining polarization both in terms of the salience of issue dimensions for partisan affect measures as well as the assessment of the most important national problems facing the United States per the ANES respondents from 1970 to

2008. In the second set, I used weighted and unweighted measures of average group distance on ideology, partisanship, and the abortion issue to answer important questions related to the culture wars and political polarization.

SECTION 1: MEASURES OF CONSENSUS, CONFLICT, BIMODALITY, AND DISPERSION. Method: Principles, Measures, & Expectations

In Chapter 5, the linear analyses of the data from the GRD and the GSS involve regressing the average yearly scores for the relevant variables against time. The models assess the time trends in political polarization, indicating whether or not attitudes on gay rights have become more polarized and conflictual, or less polarized and consensual.

Equation 4.1 - Polarization Trend Model ( ) +

݁ ݎܻܽ݁ ଵܤ଴ ൅ܤ ൌ ܸܦݏݕܴ݄݅݃ݐܽܩ There are two basic expectations that I test with the models. The first, the consensus expectation, posits that attitudes on gay rights have become more consensual since the 1970’s.

That the American public has become more unified on the subject of rights for homosexuals and attitudes towards them and their presence in society. The second, the conflict expectation, is essentially the converse of the consensus expectation. The conflict expectation posits that the

American public has become less consensual on the subject of gay rights. It asserts that the diversity of opinion on gay rights has increased, and that Americans have moved into opposing camps in their attitudes towards homosexuals and homosexuality.

Consensus Expectations Depolarizing on Gay Rights Issues over Time Eo: No trend in the consensus measure over time or a decline in consensus over time.

Ea: Consensus increases over time.

Conflict Expectations Polarizing on Gay Rights Issues over Time Eo: No trend in the consensus measure over time or an increase in consensus over time.

Ea: Consensus decreases over time.

Consensus Measure (CM) In order to conduct analysis of the consensus and conflict trends in public attitudes on gay rights in Chapter 5, the categorical responses must be converted to a single measure. The

GRD consists of frequency percentages across the polling question categories, not the individual responses that make up those percentages. The observational unit for the GRD is the poll and not the respondent. While it is possible to work backwards from the N and frequencies to produce means and standard deviations (something I do for a select number of polls), it is ultimately unnecessary to assess polarization of political issues (i.e. gay rights). We can examine the trends in gay rights attitudes in terms of consensus and conflict with a few manipulations of the available data. Conceptually, I define absolute conflict as a 50 / 50 distribution across the

Anti-Gay and Pro-Gay categories. Implicitly, absolute consensus is where 100 percent of the respondents fall into either the Anti-Gay or Pro-Gay categories. Figure 4.2 illustrates the three- step process to producing a measure of consensus (or the absence of it).

We start with the binary categories for the Anti-Gay and Pro-Gay attitudes in the

American public. Figure 4.2 has four hypothetical attitude distributions, ranging from near- absolutely consensus (95 to 5) to near-absolute conflict (55/45). The first step is to convert the category percentages to deviations from 50 percent (absolute conflict). Note that this conceptualization of absolute conflict is consistent with the theoretical arguments on polarization developed in Chapter 3 and is consistent with the group polarization measure I developed there (and discuss in empirical terms in Section II). Absolute conflict is defined as society divided into two oppositional groups of equal size. While the measure itself only permits two groups and the distance between the groups is fixed, this measure captures the size of the

‘pro’ side relative to the size of the ‘con’ side. If an issue is in a state of consensus, then we’d expect a high percentage of society to fall on one side of the issue or the other. If the issue is representative of a salient political conflict, however, we would expect opinion to be divided

FIGURE 4.2: CREATING A MEASURE OF CONSENSUS

FIGURE 4.2A: FOUR BINARY VARIABLES FIGURE 4.2C: ABSOLUTE VALUE OF CATEGORY DEVIATION FROM 50

100 50 90 45 80 40 70 35 60 30 50 ANTI-GAY 25 ANTI-GAY 40 PRO-GAY 20 PRO-GAY 30 15 20 10 10 5 0 0 C1 C2 C3 C4 C1 C2 C3 C4

FIGURE 4.2B: VARIABLE CATEGORIES RELATIVE TO 50 CENTER-POINT FIGURE 4.2D: COMBINED CONSENSUS MEASURE

50 100 40 90 30 80 70 20 60 10 ANTI-GAY 50 0 CONSENSUS MEASURE PRO-GAY 40 C1 C2 C3 C4 -10 30 -20 20 -30 10 -40 0 C1 C2 C3 C4 -50

relatively equally (and absolutely equally in the limit). In Figure 4.2A, we have a hypothetical rendering of four sets of responses on four different aspects of the gay rights issue. These could be, for example, gays in the military, gay marriage, gay adoption, and gay inheritance. For illustrative purposes, the distributions are quite distinct: The first gay rights issue is near consensus, while the last is at near absolute conflict. In Figure 4.2B, the Anti-Gay and Pro-Gay categories for each of the distinct questions on gay rights are represented in terms of their distance above and their distance below the 50 percent baseline. Poll Question C2 illustrates why one cannot simply choose one of the categories and use it in the polarization models. The

Anti-Gay and Pro-Gay categories for C2 do not sum to 100%. This is actually a frequent occurrence in the GRD. There are categories like “not sure” or “don’t know” that cannot be collapsed into one of our two substantive categories. Furthermore, there are middle categories

(neither approve nor disapprove) that should not be merged into one of the two categories.

Thus, this measure is necessary due to the fact that, even with the collapsed categories, the Anti-Gay and Pro-Gay category frequencies can and do add up to percentages below 100 percent. Only 27 of the 689 polls have Anti-Gay / Pro-Gay categories that sum to 100 percent

(3.92%). Since the deviation from 50 percent in the Anti-Gay category does not mirror the deviation from 50 percent in the Pro-Gay category, the consensus measure must combine the two for an aggregate measure across the two categories.

Once you have calculated deviations from 50 percent for both categories, the seconds step (Figure 4.2C) is simply taking the absolute value of those deviations. This leads to the third and final step of combining the two deviations into a single, composite consensus measure. As can be seen in Figure 4.2C, this consensus measure is a valid and reliable measure of the combined deviations from absolute conflict for the Anti-Gay and Pro-Gay categories. A

consensus score of 100 represents absolute consensus, and a consensus score of zero represents absolute conflict. The measure, henceforth referred to as CM, is thus ready for comparison across years and inclusion in the polarization trend regressions.

Mean, Dispersion, and Bimodality Measures of Polarization For the GSS data (and selected polls) in Chapter 5 and the closed-ended items from the

ANES in Chapter 6, I examine not only trends in consensus and conflict but also the extent to which public opinion on gay rights have become more dispersed on average over the time series. The dispersion measures are also useful in assessing changes on gay rights attitudes within and between the parties. I employ a measure of bimodality (kurtosis), a measure of average position (mean), and two measures of dispersion (standard deviation and coefficient of variation) to assess the degree that American public opinion on gay rights has fractured over the past 35 years.

Mean Trends. While not a measure of dispersion itself, the mean is sensitive to unequal shifts in the distribution of a variable and the trend in changes in the mean can move as a consequence of polarization. For example, in Chapter 5 I examine public opinion on a specific social issue: gay rights. The trend in the mean of gay rights attitudes can show whether the American public has continued to hold the same attitudes on homosexuals or whether that attitude has shifted towards the extreme values of the measures of gay rights attitudes. For example, if a variable ranges from 1 to 4 and the mean at period X1 is 1.6 while the mean at period X2 is 1.2, we could conclude that this reflects an increase in the consensus on the attribute for the population. If, however, the mean at X2 is 2.4, then we can conclude that we have seen a shift away from the previous consensus on the attribute.

Since polarization can occur independent of changes in the mean, shifts in the mean on an issue dimension to the extremes can reflect consensus while shifts to the center can reflect

an increase in conflict over an issue. However, for the most part, the variables included from the

ANES in the Chapter 6 analysis are already a subject of conflict over the course of the time series. There is no consensus on ideology, defense spending, or jobs in the United States. We may see an emerging consensus, however, on attitudes towards minorities and women.

Generally speaking with political variables, an increase in the mean which moves the mean to one extreme or another is an indication of consensus and not conflict.

This is the especially the case with affect measures such as thermometer scores. An average affect score of 50 is likely not the result of almost every respondent giving the group, party, or candidate a score in the middle of the scale. Rather, it is likely the consequence of a sizeable number of respondents giving scores in the sixties and seventies with an equal portion of respondents giving scores in the forties and thirties. Likewise, an average affect score of 70 suggests that most respondents have very positive feelings towards the group, party, or candidate and hence there isn’t a great deal of room for conflict in the related issue dimension

(if there is one). However, likelihood is not empirical certainty, and hence any interpretation of changes in the mean (if there is one) must be assessed in terms of the distribution of the variable prior to the change and the distribution afterwards. A failure to account for the distribution of opinion on the group makes the means ultimately uninterpretable in terms of polarization or depolarization. Where there is a reasonable spread of opinion on a group or issue, movements in the mean over time (if they occur) can be understood as moves towards consensus or conflict depending on the starting point (year) and whether opinion is moving closer or further away from the extreme of the scale the previous average was closest to. Hence a move from 65 to 75 is a move towards (rather than away from) the pole the mean is most proximate to and hence indicates increasing consensus. A move from 65 to 55 would be a move away from that poll and hence indicates conflict. The trend model assesses changes in the

average position on the issue or group over time. Whether a negative or positive coefficient is

indicative of polarization is dependent upon whether the pole that the mean opinion is

proximate to is located higher or lower on the scale.

Equation 4.2: Mean Model

( ) +

݁ ݎଵ ݕ݁ܽܤ଴൅ܤ௑ത ൌܫ

Mean Polarization on Issues Expectations Eo: No trend in the mean position.

E1: A significant trend towards the pole the average issue position is proximate to. (depolarization)

E2: A trend in the mean issue position away from the pole the average issue position is proximate to. (polarization)

Difference of Means. This measure is useful in comparing differences between groups and in

assessing how those differences change over time. Specifically here, the difference of means

(and standard deviations) will allow for the assessment of intra-party differences on gay rights

attitudes, and how those vary over time.

Bimodality - Polarization

The principle of bimodality is intended to capture the level of conflict within society by assessing the degree to which opinion on an issue or attribute is divided into two camps (see

Chapter 3). A measure which validly and reliably assesses the bimodality principle must capture the degree to which a distribution is located at one, two, or multiple modes and be sensitive to changes in that distribution. If the population is divided into two camps on an issue, then that would tend to increase identification (between those who share the same views) and increase perceptions of alienation (between those who do not share the same views). While the sizeable number in opposition may increase the necessity for making compromises (given majoritarian

American political institutions), the separation into two camps makes such compromise more

difficult and enhances conflict between the two groups as neither side can be dismissed as

irrelevant or incapable of securing political outcomes (as, say, small fringe groupings can be).

Bimodality Measures

As noted previously (see Chapter 3), a rough approximation of bimodality for a distribution is

captured in the fourth moment around the mean (standardized) otherwise known as kurtosis.

This measure provides an approximation of the ‘peakedness’ of the distribution. I’ve noted

some of the problems with using kurtosis to measure bi-modality. As a measure of peakedness,

kurtosis is a rough approximation rather than a direct measure of bimodality. While bimodal

distributions are less ‘peaked’ than normal distributions, differently shaped distributions present

special problems for kurtosis and thus the fourth moment about the mean is at best a rough

approximation of bimodality. Fine differences in bimodality between certain non-normal

distributions may produce inconsistent changes in kurtosis. That said, kurtosis is strongly

correlated with bimodality, as demonstrated in Chapter 3. The kurtosis measure is thus useful in

distinguishing between bimodal distributions and less-bimodal distributions on the whole.

Kurtosis is used here to measure the change in the shape of the distributions of issue

dimensions overtime.

Equation 4.3: Kurtosis = [ ÷ ] – 3 ସ ସ ݇ ቄ෍ ܺെ݉ ܰΤܵ ቅ As kurtosis = 0 is centered on t he normal distribution, kurtosis scores which fall below zero are indicative of a more bimodal distribution than the normal distribution (higher values connote a unimodal or single-peaked distribution). With respect to a trend over time, a negative trend in kurtosis indicates a shift towards bimodality in the distribution (conflict), while a positive trend indicates a shift towards unimodality (consensus). This trend is modeled with the average

kurtosis in a given year on an issue serving as the dependent variable with survey year as the

independent variable.

Equation 4.4: Bimodality Model

( ) +

݁ ݎଵ ݕ݁ܽ ܤ଴൅ܤ௞ ൌܫBimodality of Issue Dimensions Expectations Eo: No trend in the kurtosis of the distribution.

E1: A significant positive trend in the kurtosis of the distribution (depolarization).

E2: A significant negative trend in the kurtosis of the distribution (polarization).

Dispersion - Polarization

The interpretation of results for the dispersion measures is relatively straightforward. In order to measure dispersion, we need a measure that both reflects the relative distance that individual respondents differ from one another as well as taking into account the proportion of opinion located in the extremes relative to the center of the distribution. Increased dispersion indicates polarization as it reflects an increase in the distance between individuals and/or groups within society on that issue or attribute. If opinion is highly dispersed, then institutions and actors which seek to compromise on policy related to that opinion may find it difficult or impossible to get the requisite support at the popular level. Conversely, a constrained distribution of opinion (low dispersion) suggests a policy space ripe for compromise and perhaps even consensual politics. The standard deviation is a measure of the dispersion in the data. As such it is a good measure of the degree to which consensus on gay rights attitudes, abortion attitudes, economic issues attitudes, defense policy attitudes, etc. exists at the mass level and how that consensus has changed over time.

Dispersion Measures

The traditional measure of dispersion is variance, or its standardized version: the

standard deviation. As opinion dimensions become more polarized, variance (and thus the

standard deviation) should increase.

Equation 4.5: Standard Deviation

= ) 1 ଶ An increase in the standard deviationߪ onට an෍ issueሺݔെݔ indicatesҧ ൗܰെ that mass public opinion on the issue

has become more polarized on that issue. The more dispersed the population, the more likely

the issue will produce intractable political conflict rather than centrist / moderate compromise

polices. A more dispersed opinion distribution means that, in order to produce a compromise

policy, citizens must agree to a change in the status quo that is, on average, further distant from

their own ideal point. This increases conformity costs in the creation of public policy on the

issues where we observe greater dispersion. This trend is modeled with the average standard

deviation in a given year on an issue serving as the dependent variable with survey year as the

independent variable.

Equation 4.6: Bimodality Model

( ) +

ఙ ଴ ଵ ݁ ݎݕ݁ܽ ܤ൅ ܤൌ ܫ Dispersion Expectations on Issues Ho: No trend in the standard deviation on issues

Ha: A significant decrease in the standard deviation on issues (depolarization).

Ha: A significant increase in the standard deviation on issues (polarization).

Coefficient of Variation. The coefficient of variation is a normalized measure of dispersion. It is

the ratio of the standard deviation to the mean. The C.V. is recommended for use with ratio

measures, and is helpful because it states the standard deviation in terms of the mean. As such, it is useful for comparing different distributions of polling data with differing means. However, since it states the standard deviation in terms of the mean, it will be of less use in assessing a time trend where variance in the mean and standard deviation are correlated. As such, I employ this measure exclusively in Chapter 5.

Issue Dimension Measures for Party Likes and Dislikes & Mass Perceived National Problems

For the Republican and Democratic likes and dislikes, I create conservative and liberal response sets for the government philosophy, social, economic, and defense issue dimensions.

The categorization of the open-ended response sets is reported in Table 5.1. Note that the expected applicable response set is dependent upon whether it is a “like” or “dislike” variable. In this analysis I combine the conservative and liberal responses on the party likes and dislikes to get measures of the total number of social issue mentions (as well as mentions on the other kinds of issues).

Frequency Polarization

In assessing the open-ended responses here, I will be looking for a trend in the number of issue mentions across the time-series. For example and relevant to the culture wars thesis, if the number of social issue mentions as an “important national problem” or as a reason to like or dislike the two political parties have increased over the past three decades, then that is evidence suggesting an increase in the salience of the issue dimension (and thus evidence it is a dimension on which politically relevant polarization can occur). Furthermore, if social issue- mention trends deviate along party lines, then it would be evidence of partisan polarization and the increasing importance of social issues in political competition. For example, if citizens increasingly mention social issues as the reasons they like or dislike the parties then this would be evidence supporting the culture wars thesis and evidence suggesting partisan polarization on

the social issues. This trend is modeled by regressing year on the total number of issue mentions

for a survey year in the party likes and dislikes item as well as the national problems item.

Equation 4.7: Frequency Polarization Model

௡ ( ) +

݁ ݎଵ ݕ݁ܽܤ଴൅ܤ௜ ൌܯܫ ෍ Issue Mention Expectations ௜ୀ଴

Ho: No trend in the frequency of issue mentions on this dimension

Ha: A significant negative trend in the frequency of issue mentions (depolarization).

Ha: A significant positive trend in the frequency of issue mentions (polarization).

The methods applied in this analysis include an analysis of trend lines in the means and standard deviations and kurtosis for the closed-ended issue-related items from the ANES. It also includes OLS regression of the survey year on the frequency measures for the issue dimensions obtained from the open-ended party likes and dislikes items and the “national problem” open- ended item. In addition, I regress the mean and standard deviation measures on year to assess the direction and significance of these trends. I report these results for the thermometer and issue placement measures, the closed-ended self-placements on issue scales, and the count (or frequency) measures for the open-ended responses on the most important national problem and the Republican and Democratic likes and dislikes.

TABLE 4.1: ISSUE DIMENSION VARIABLE DEFINITIONS CREATED FROM ANES OPEN-ENDED RESPONSE SETS Government Government Social Social Economics Economics Defense Defense Conservative Liberal Conservative Liberal Conservative Liberal Conservative Liberal Against gov’t For gov’t activity Against social For Social Change Too Much Pro Planned Hard-Line, Anti- Soft-Liner on activity change Interference in Economy Communism Communism Private Economy Value property Value human Anti Separation of Pro Separation of Anti Government Pro Gov’t Isolationist Internationalist over human rights rights over Church & State Church & State Economic Aid Economic Aid property Anti-socialism Pro-socialism Anti Aid to Pro Aid to Anti Soc. Sec. Pro Soc. Sec. Strong Military Weak Military Education Education Expansion Expansion Conservative Liberal Pro Aid to Parochial Anti Aid to Anti Expansion Pro Expansion Oppose Détente Support Détente Education Parochial Education Unemployment Unemployment w/ Communist w/ Communist Benefits Benefits Countries Countries Pro-Far Right Anti-Far Right Anti Civil Rights Pro Civil Rights Anti Medicare and Pro Medicare and Pro Military Aid to Anti Military Aid to Medicaid Medicate Allies Allies Anti-Far Left Pro-Far Left Anti Civil Liberties Pro Civil Liberties Anti Public Pro Public Housing Anti Foreign Aid Pro Foreign Aid Housing Pro States Rights Anti States Rights Anti Pro Lower Taxes Higher Taxes Pro Israel / Anti- Anti Israel / Pro Environmentalism Environmentalism Arabs Arabs Allow Inequality Equality Law & Order – Law & Order – Soft- Keep Tax End Loopholes Anti Détente w/ Pro Détente w/ Hardliner liner Loopholes Red China Red China Pro Status Quo Anti Status Quo Traditional Public Permissive on Anti Price Supports Pro- Price Support Anti Détente w/ Pro Détente w/ Morality Public Morality for Farmers for Farmers Russia Russia Pro Work Ethic Anti Work Ethic Drugs – Hardliner Drug Legalization – Pro Right-to-Work Anti Right-to-Work Pro Defense of Anti Defense of Liberalization Laws Laws Iron Curtain States Iron Curtain States ------Anti-Abortion Pro-Abortion Fewer Labor More Labor Strikes Anti Castro’s Cuba Pro Castro’s Cuba Strikes ------Anti-Gun Control Pro-Gun Control Pro Nuclear Power Anti Nuclear Anti Leftists in Pro Leftists in Power Africa Africa ------Anti-Bussing Pro-Bussing Anti National Pro National Raise American Lower American Health Insurance Health Insurance Prestige Prestige ------Anti Gov’t Aid to Pro Gov’t Aid to Pro New Energy Less Fuel Pro Victory in Anti Victory in Cities Cities Vietnam Vietnam ------Anti Women’s Pro Women’s Anti Jobs Program Pro Jobs Programs Pro Free Trade / Anti Free Trade / Rights Rights Anti Tariffs Pro Tariffs ------Pro School Prayer Anti School Prayer Anti Gov’t Pro Gov’t Anti Trade w/ Pro Trade w/

Government Government Social Social Economics Economics Defense Defense Conservative Liberal Conservative Liberal Conservative Liberal Conservative Liberal Healthcare Healthcare Communists Communists ------Anti Gay Marriage Pro Gay Marriage Pro Drilling in Anti Drilling in Anti Amnesty for Pro Amnesty for Arctic Refuge Arctic Refuge Draft Dodgers Draft Dodgers ------Pro Death Penalty Anti Death Penalty Anti Transport – Pro Transport – Pro MIA/POW’s Anti MIA / POW’s Communication Communication Regulation Regulation ------Anti Affirmative Pro Affirmative Anti Labor Unions Pro Labor Unions Pro Kissinger Anti Kissinger Action Action Foreign Policy Foreign Policy ------Pro Standards for Anti Standards for Pro Rich Anti Rich Pro Military Anti Military School School Spending Spending ------Pro Clinton Pro Stem Cell Pro Small Business Anti Small Business Anti Nuclear Pro Nuclear Freeze Impeachment Research Freeze ------Pro School Pro Cloning Anti Welfare Pro Welfare Strong Homeland Weak Homeland Vouchers Mothers Mothers Security Security ------Anti Stem Cell Pro Gays & Lesbians Anti Poor Pro Poor Pro Desert Storm Iran Contra Research ------Anti Cloning Anti Christian Right Anti Blue Collar Pro Blue Collar Anti Chinese NATO war in Spying during Serbia Clinton Admin ------Anti Blacks Pro Blacks ------Anti Feminists Pro Feminists ------Pro Veterans Anti-Veterans ------Groups ------Anti-Minority Pro Minority ------Groups Groups ------Anti Gays & Pro Gays & Lesbians ------Lesbians ------Pro Christian Right Anti Christian Right ------Anti Hispanics Pro Hispanics ------

SECTION 2: MEASURES OF GROUP POLARIZATION

Expectations

Partisan Ideological Polarization

Eo: No observable trend in the ideological proximity of partisan identifiers.

Ea: A decrease in the distance between partisan identifiers on ideology. (depolarization)

Ea: An increase in the distance between partisan identifiers on ideology. (polarization)

Partisan Polarization on Abortion

Eo: No observable trend in the proximity of partisan identifiers on the abortion issue.

Ea: A decrease in the distance between partisan identifiers on abortion. (depolarization)

Ea: An increase in the distance between partisan identifiers on abortion. (polarization)

Income Group Partisan Polarization

Eo: No observable trend in the party identification of income groups.

Ea: A decrease in the distance between income groups in their party identification. (depolarization)

Ea: An increase in the distance between income groups in their party identification. (polarization)

Income Group Ideological Polarization

Eo: No observable trend in the ideological proximity of income groups

Ea: A decrease in the distance between income groups on ideology. (depolarization)

Ea: An increase in the distance between income groups on ideology. (polarization)

Religiosity Partisan Polarization

Eo: No observable trend in the party identification of religious and secular citizens.

Ea: A decrease in the distance between religious and secular citizens in their party identification. (depolarization)

Ea: An increase in the distance between religious and secular citizens in their party identification. (polarization)

Religiosity Ideological Polarization

Eo: No observable trend in the ideology of religious and secular citizens.

Ea: A decrease in the distance between religious and secular citizens on ideology. (depolarization)

Ea: An increase in the distance between religious and secular citizens on ideology. (polarization)

Models

As with most of the analyses, I examine the trends in polarization using simple and multiple regression models with the polarization measure as the dependent variable and year

(or a transformation of year into time periods) as the independent variable. Significant, positive coefficients indicate a trend towards polarization along the issue dimension in the model while negative coefficients show the opposite.

Equation 4.8: Simple Trend Model of Group Polarization

( ) +

the absence݁ of a significant relationship ݎଵ fromݕ݁ܽܤbe inferred଴൅ ܤൌܲܩNo apparent trend in polarization may between the group polarization measure and the year variable. Where the party of the presidential administration (ppa) is a theoretically-grounded explanation of group polarization, I include it as an independent predictor of group polarization. This multivariate model is described in Equation 6.2:

Equation 4.9: Multivariate Trend Model of Group Polarization

( ) ( ) +

݁ ଶ ݌݌ܽܤ൅ ݎଵ ݕ݁ܽܤ଴൅ܤൌܲܩ Analysis & Measures

As I noted in Chapter 3, polarization is characterized by increasing identification with those similar to oneself along some relevant attribute coupled with increasing alienation from those dissimilar to oneself along that same attribute. There are thus three features of polarization: 1) there must be a high degree of homogeneity within each group. 2) There must

be a high degree of heterogeneity across groups. 3) There must be a small number of significantly sized groups (isolated small groups or individuals are irrelevant) across the attribute dimension (Esteban and Ray 1994). An empirical measure of group polarization thus incorporates a measure of group size and a measure of the ‘antagonisms’ or distances from all the other groups for a given policy, issue, or ideological dimension. The measure developed here

(GP), incorporates both of these key aspects of polarization for the three sets of groups analyzed here: party identification groups, income groups, and religiosity groups.

While the groupings reflected in the party ID, income, and religiosity groups are defendable as valid and reliable constructs of the divisions along partisan, income, and religiosity dimensions, the cut-points established for these groups are somewhat arbitrary.

Naturally, if one collapsed categories of these groups it would increase GP overall and, likewise, if one parsed the groups further it would lead to a decrease in GP. However, this concern is mitigated by the fact this analysis examines changes in GP over time. Hence the number of groups remains constant for each survey year. Furthermore, any poor mapping of the groups used onto the actual distribution of mass in the population on these dimensions would work against a finding of polarization, as disparate identifiers on the issue dimension would be included in the wrong groups and thus artificially draw the means of those groupings together.

In other words, an inaccurate classification of groups might suggest the absence of polarization when it in fact exists, but would not lead to a finding of a false polarization trend (See Chapter 3 for GP explanation).

Equation 4.10: Empirical Measure of Weighted Group Polarization on Political Dimensions

= ௠ { ( )[( ) + ( ) …+( ) ]} ଶ ଶ ଶ ܩܲூ ෍ ݊ீଵ ͳെ݊ீଵ ܫீଵҧ െܫீଶҧ ܫீଵҧ െܫீଷҧ ܫீଵҧ െܫீ௠ҧ + {ீୀଵ( )[( ) + ( ) …+( ) ]} … ଶ ଶ ଶ ீଶ ீଶ ீଶ ீଵ ீଶ ீଷ ீଶ ீ௠ + { (1 ݊ ͳെ݊)[ ܫҧ )െܫ+ҧ ( ܫҧ െܫ) +ҧ ( ܫҧ )െܫ… +ҧ ( ) ]} ଶ ଶ ଶ ଶ ீ௠ ீ௠ ீ௠ ீଵ ீ௠ ீଶ ீ௠ ீଷ ீ௠ ீ௟ ݊ Where:െ݊ ሺܫҧ െܫҧ ܫҧ െܫҧ ܫҧ െܫҧ ܫҧ െܫҧ

= a political dimension (issues, partisanship, etc.) = a group of individuals associated on a political or social dimension ܫ = the average position on for Groups 1 through M. ீ …݊ = the sample proportion for Groups 1 through M. ܫீଵҧ ǥܫீ௠ҧ ܫ

݊ீଵ ݊ீ௠ This measure of group polarization calculates the total distances between the defined groups on the two dimensions of political conflict that I examine here: ideology and partisanship. It weights the summation of those distances by the sizes of the defined groups, using the ANES sample proportion for the group categories in each survey year. This definition is consistent with the theoretical discussion of polarization (Chapter3), as the maximal polarization in this measure would involve society dividing itself into two groups on the G dimension, with the two groups locating themselves at the extremes of the I dimension. If the groups separate in the I dimension, then that is evidence of a polarization trend. If the groups move towards each other in the I dimension, polarization declines. One asset of this measure is that, even if the number of groups is over-defined (say, we have created foure categories of group but, in terms of their relative location, there really is only three groups) it will not affect the group polarization trends identified in the analysis. Let’s say that Group A and Group B have almost identical positions on ideology. If that is the case, the relative distance between them approaches zero and correspondingly counts little towards the overall polarization coefficient on ideology and would not contribute to an increase or a decrease in the GP measure.

The unweighted group polarization measure simply calculates the squared distances between each of the group on a political dimension irrespective of the size of the group:

Equation 4.11: Empirical Measure of Unweighted Group Polarization on Political Dimensions

= ௠ {[( ) + ( ) …+( ) ]} ଶ ଶ ଶ ܷܩܲூ ෍ ܫீଵҧ െܫீଶҧ ܫீଵҧ െܫீଷҧ ܫீଵҧ െܫீ௠ҧ + {[(ீୀଵ ) + ( ) …+( ) ]} … ଶ ଶ ଶ ீଶ ீଵ ீଶ ீଷ ீଶ ீ௠ + {[ ܫҧ െܫҧ ) + (ܫҧ െܫҧ ) + ( ܫҧ െܫҧ ) … + ( ) ]} ଶ ଶ ଶ ଶ ሺܫீ௠ҧ െܫீଵҧ ܫீ௠ҧ െܫீଶҧ ܫீ௠ҧ െܫீଷҧ ܫீ௠ҧ െܫீ௟ҧ Where:

= a political dimension (issues, partisanship, etc.) = a group of individuals associated on a political or social dimension ܫ = the average position on for Groups 1 through M. ݊ீ ܫீଵҧ ǥܫீ௠ҧ ܫ

CHAPTER 5: FROM CONSENSUS TO CONFLICT – GAY RIGHTS AND THE CULTURE WARS

“Whence spring these inclinations, rank and strong? And harming no one, wherefore call them wrong?” -- Anonymous poem written in defense of Captain Nicholas Nicholls sentenced to death for sodomy, London 1833

“Adults may choose to enter upon this relationship in the confines of their homes and their own private lives and still retain their dignity as free persons. When sexuality finds overt expression in intimate conduct with another person, the conduct can be but one element in a personal bond that is more enduring. The liberty protected by the constitution allows homosexual persons the right to make this choice....” -- Justice Anthony Kennedy for the Majority, Lawrence v. Texas

“Today's opinion is the product of a Court, which is the product of a law-profession culture, that has largely signed on to the so-called homosexual agenda, by which I mean the agenda promoted by some

THE GROWING CONFLICT OVER GAYS AND GAY RIGHTS

The divisive and charged argument between Justices Kennedy and Scalia illustrate the fact that gay rights and attitudes towards homosexuals are at the cutting edge of contemporary social conflict and is one of the key fronts in the culture wars. However, it wasn’t always so. In the not too distant past, there was a relative consensus in the American public on homosexuals and gay rights. There was a general attitude of distaste for homosexuals and their lifestyles and a bare tolerance for even fundamental civil rights such as the right to hold a job. Homosexuals were not to be trusted in important positions like political office; they were not to be tolerated in jobs that required intimate contact with the public, nor any position which might expose children to gays or their lifestyle. This has changed over the last 35 years. The American public is more accepting of, tolerant of, and positive in its disposition towards homosexuals. This shift in public opinion, however laudable it may be, has also produced a backlash. The loss of consensus on homosexuals and gay rights has produced a cacophony of voices on the one hand demanding a halt to the mainstreaming of homosexuals (if not a complete reversal) with an equally fervent

(though smaller in relative size) segment of the population insisting on significant political changes to further normalize homosexuals, homosexuality, and homosexual relations.

Gay rights are a hot button issue in American social conflict. It is squarely on the public agenda in both legislative and legal environments. The parties have mobilized to provide alternative policy choices on gays and gay rights. Activists and political interest groups campaign for and against gay rights and hold officeholders to account on how they do or do not vote on gay rights measures. I will track the course the birth of the culture wars has paved on gay rights— from consensus to conflict. I will delineate the specific gay rights issues that have become more conflictual (or more consensual) over the past several decades. And I will illustrate how this particular front in the culture war has affected the political parties at the mass level and thus spurred the political and policy battles on gay rights we see today.

This relatively narrow focus on just one social issue serves several important purposes in furthering the general analysis of polarization. First, as mentioned earlier, it is a hot-button social issue that is topically relevant to the culture wars thesis and, yet, has received little attention in the culture wars literature. Some scholars have ignored gay rights in favor of social issues such as abortion, while other scholars have suggested that gay rights—while an important social issue and a potential factor in the culture wars discussion—is not currently subject to analysis due to volatility in public opinion and no apparent trends. I remedy that oversight in conducting a thorough examination of public opinion on gay rights and how it has trended over nearly four decades. Second, it allows for a direct examination of the validity and reliability of the empirical measures of polarization developed in Chapter 4. This proof of concept will provide the foundation upon which the later analysis of a number of issues and issue dimensions rests. Third, essentially serving as a case study of polarization, examining just gay

rights permits a depth of issue assessment that is not feasible with the broader analysis. The

ability to examine gay rights from a number of angles and to consider the variety of aspects to

this single social issue adds significant value to the analysis. Thus not only deepening our

contextual understanding of this specific social issue, but also sharpening our understanding of

polarization in public attitudes on an issue dimension.

DOES EMERGING CENTRISM ON GAY RIGHTS PROVE THE CULTURE WAR A MYTH?

In June, 2003, the Supreme Court issued an opinion in Lawrence v. Texas, striking down a Texas sodomy statute and overturning the precedent set in Bowers v. Hardwick, which had declined to find a constitutional right to private, consensual sexual relations a little over a decade prior. While the gay rights movement hailed the decision and culture warriors decried it, one thing is certain irrespective of the constitutional questions: Lawrence was not out of step with the trend in public attitudes on gay rights in America. Gays have steadily gained traction in the public consciousness and attitudes towards gays have shifted significantly in favor of tolerance and normalcy for homosexuals and homosexual relations. One of the prime pieces of evidence against the culture wars thesis is this growing centrism on public attitudes towards homosexuals and gay rights. There is no question that this is happening…and it is happening across demographics and even across party lines. Some culture war skeptics point to this as prima facie evidence against the culture wars thesis. That the aggregate decline in hostility towards gays and the slippage in opposition to gay rights is proof against an emergent divisive social conflict. They are wrong. The growing centrism on gay rights is powerful confirmatory evidence that the culture war has “gone hot.”

The mistake skeptics make in interpreting the data on attitudes towards gays and gay rights is to view centrism as commiserate with political moderation. As I discussed in Chapter 3, while a normal distribution of attitudes may be much more moderate than, say, a bi-polar

distribution with the poles shifted to the absolute extremes of an issue dimension, it is not the

case when compared to a consensual distribution of opinion with everyone in relative

agreement on the issue at hand. In the latter case, a shift towards centrism is, in fact,

polarization. It represents an increase in the probability of social conflict and an opening for

political activism and mobilization in the policy process. Consensus positions rarely find their

way onto the policy agenda, and if they do it is usually as a symbolic or affirmational gesture

rather than real, substantive policy. Consensus positions present no incentive for political

parties to diverge over related policy alternatives nor do they generate much activism against

the consensus position oriented towards persuading electorate-responsive officials (given the

small probability of success). When a consensus breaks down, with masses of the public attitude

on the issue shifting away from the status quo, the issue becomes ripe for politics. Now money

and votes can be procured by appealing to one side or the other of the emergent divide. Policy

alternatives can be generated that have some chance of successfully navigating the legislative

process. Both sides seek political solutions: either through progressive change or successfully

thwarting that effort at change. And as a consequence this issue becomes increasingly relevant

to partisanship and electoral politics.

CALM SEAS - THE COMMANDING CONSENSUS, 1970 – 1988

Homosexuality is a taboo in many societies, but it is particularly so in countries with

Judeo-Christian traditions. The United States is no exception, with nearly every state having had, at one time, sodomy laws outlawing homosexual relations. While some states had repealed their prohibitions on homosexual sex and many such provisions still on the books went largely unenforced, there was little evidence that the taboo on homosexuality and the societal rejection of homosexuals would ever change, baring extraordinary legal action. Certainly the policy process, subject to popular opinion, seemed an unlikely avenue for change. This is evidenced

explicitly in the Virginia Slims American Women’s Poll in 1974 asked 3880 respondents if they

would find it acceptable if their grown daughter had a homosexual relationship. 75% of those

respondents found it unacceptable with an additional 19% who would accept it but be unhappy

about it and would have a strained relationship with that daughter as a result. Only one percent

of those respondents reported that they would find the relationship acceptable.3 This is

powerful evidence of a strong consensus that perceives homosexuals as deviants, encourages

and endorses intolerance towards gays, and opposition to policy innovations on gay rights.

FIGURE 5.1: 1977 HARRIS SURVEY*: WHAT JOBS SHOULD HOMOSEXUALS BE ALLOWED TO HOLD?

*Harris Survey. Conducted by Roper Organization. June, 1977. Sample N: 1947 respondents.

The halcyon days of consensus on most issues related to gay rights was a mere three decades ago. The 1977 Harris Survey on public attitudes reveals the cut-point from consensus to conflict that existed over what jobs homosexuals could openly hold in society. While the

3 Virginia Slims American Women’s Poll conducted by Roper, April 1974. 3880 respondents. “For [a homosexual relationship], tell me for a daughter of yours who had just finished her schooling whether you would find it acceptable, or accept it but be unhappy about it, or not accept it and have the relationship very much strained as a result?”

100

consensus period certainly reflects a shunning of and hostility towards gays, it was not, as some

homosexual rights activists would paint it, equivalent to the imposition of Sharia law. Far from

public stonings, there was a consensus that homosexuals could hold jobs, make speeches, and

participate in civil society. A majority of Americans reported support for equal job opportunities

for homosexuals.4 While the consensus was accepting of homosexuals in blue collar jobs and positions which require little contact with the going public, there was considerable opposition to homosexuals in positions of authority (congressman, policeman), positions of public trust (social worker), positions which require intimate contact with others (doctor, psychiatrist), religious positions (priest), and positions which involve contact with children (principal, teacher). Though the Harris survey demonstrates that the United States in the 1970’s was far from Afghanistan under the Taliban in its attitude toward homosexuals, it also reveals that homosexuals were perceived as at the fringe of society. Well over half of the public endorsed outright discrimination against homosexuals in all but menial jobs.

The fact Americans were conflicted over the extension of basic civil rights to homosexuals illustrates that a powerful consensus of distaste towards homosexuals was operative. In 1983, Sixty-four percent of respondents said they would not vote for a candidate for president if he were homosexual despite the fact he was well-qualified for the office.5 A majority disapproved of homosexual relations between consenting adults for anyone, and less than ten percent found it personally approvable.6 Sixty-three percent reported that they were

4 Gallup Poll, June 1977. 2000 respondents. “In general, do you think homosexuals should or should not have equal rights in terms of job opportunities?” 5 Gallup Report, April 1983. 1517 respondents. “Between now and the political conventions in 1984 there will be discussion about the qualifications of presidential candidates…If your party nominated a generally well-qualified man for president and he happened to be homosexual, would you vote for him?” 6 Los Angeles Times Poll, September 1983. 1653 respondents. “What is your attitude toward homosexuality? Do you personally approve of homosexual relations between consenting adults…or do you oppose it for everyone?”

101

unsympathetic with the homosexual community,7 and ninety percent of respondents said they

would be upset if their child told them they were gay.8 Clearly the American public was more

than a little uncomfortable with homosexuals. Against the backdrop of the social consensus

against homosexuals, the AIDS epidemic could only have exacerbated fears of homosexuals.

While most Americans reported they did not avoid homosexuals because of the AIDS epidemic

and were not worried about catching it from them,9 majorities reported that they believed

discrimination against gays was on the rise as a consequence of AIDS10 and nearly eighty percent

that AIDS, a disease that in the 1980’s was almost exclusively a problem within the homosexual

community, posed a significant public health threat.11

SOWING THE WIND: AN EMERGENT SOCIAL CONFLICT ON GAY RIGHTS, 1988 – 1991

Whether a consequence of increasingly favorable representations in the movies and television, the organization of the gay rights movement and the founding of groups like ACT-UP, a reaction to the Supreme Court ruling in Bowers v. Hardwick that upheld the constitutionality of sodomy laws (only a bare majority reported approval), or some combination of exogenous and endogenous shocks to the social system, the tide began to turn in favor of homosexuals and

7 Los Angeles Times Poll, September 1983. 1653 respondents. “Would you say you are very sympathetic, somewhat sympathetic, somewhat unsympathetic, or very unsympathetic to the homosexual community?” 8 Los Angeles Times Poll, September 1983. 1652 respondents. “If you had a child who told you he or she was a homosexual, what do you think your reaction would be? Would you be very upset, not very upset or not upset at all?” 9 ABC News / Washington Post Poll, March 1987. 1511 respondents. “Would you say that you are worried that [a homosexual] might give you AIDS (Acquired Immune Deficiency Syndrome)?” 10 CBS News Poll, October 1986. 823 respondents. “Do you think there has been more discrimination against homosexual men since AIDS became a serious problem, or don’t you think the amount of discrimination against them is any different now than before?” 11 ABC News / Washington Post Poll, September 1985. “So far three-quarters of AIDS victims have been homosexual males. The rest of the victims have mainly been drug addicts or recipients of blood transfusions. Do you think that AIDS is spreading so that it is now a threat to the general public in the United States, or not?”

102

gay rights in the late 1980’s.12 The consensus against homosexuals was forever shattered. One

significant road sign of the radical change that was under way was in the number of respondents

who reported they had friends or acquaintances who were gay. In 1985, ninety-one percent of

respondents reported they had neither friends nor associates who were gay.13 In 1986, seventy-

eight percent of respondents reported that they did not have a friend who was gay.14 By 1992,

nearly half of the American public reported knowing someone who was gay or lesbian.15 A

fourth reported having a close, personal friend who was gay.16 New gay rights issues, such as gay marriage17 and gays in the military18 began to emerge during this time period that would become hot topics on the public agenda and embroil a president in controversy early in his tenure. A battle over gay rights in the political arena was brewing.

REAPING THE WHIRLWIND: THE CULTURE WAR ON GAY RIGHTS, 1991 - PRESENT

The cultural conflict over gay rights would come to a head in the 1990’s with the issue of gays in the military. President Clinton, just elected to office and thus ending twelve years of

Republican rule in the White House, stumbled into a political briar patch by signaling he would end the ban on gays in the military. President Clinton (and Dick Morris) may have suspected this

12 Gallup Report, July 1986. 1538 respondents. “The Supreme Court recently ruled that the Constitution does not give consenting adults the right to have private homosexual relations. Do you approve or disapprove of this ruling?” 13 ABC News Washington Post Poll, September, 1985. 1512 respondents. “Do you have a friend or someone you associate with on a regular basis who is a male homosexual?” 14 NBC News / Wall Street Journal Poll, January 1986. 1598 respondents. “Do you have any friends who are homosexual?” 15 CBS News / New York Times Poll, August 1992. 656 respondents. “Do you happen to personally know someone who is gay or lesbian?” 16 Harris Poll, October 1992. 1583 respondents. “Do you have any close personal friends who are gay or lesbian, or not?” 17 The first polling question in the IPOLL database on gay marriage was asked in the General Social Survey in February, 1988. “(Do you agree or disagree?)... Homosexual couples should have the right to marry one another.” 18 The first polling question in the IPOLL database on gay marriage was asked in a Los Angeles Times Poll in October, 1992. “Do you approve or disapprove of allowing openly homosexual men and women to serve in the armed forces of the United States? (If approve or disapprove, ask:) Do you (approve/disapprove) strongly or (approve/disapprove) somewhat?”

103

would be a relatively easy political win, given the consensus that had emerged supporting equal job opportunities for gays. They were wrong. Clinton’s effort quickly became a political firestorm and, consequently, he adopted a Third Way compromise position of “Don’t Ask, Don’t Tell.”

While this was intended to end the controversy, the American public was as closely divided on

DA,DT as they were gays in the military to begin with (54/47 against DA,DT).19 The battle over gays in the military was merely a prelude to the war that would be waged, and is still being waged, over gay marriage. In 1996, the United States Congress passed the Defense of Marriage

Act, defining marriage as only between couples of the opposite sex at the federal level. The

Supreme Court having paved the way with its ruling banning state sodomy laws in Lawrence v.

Texas, a number of legislative, judicial, and state referendum efforts have emerged in support of and opposed to gay marriage. Several state supreme courts have ruled that gay marriage is required under their state constitutions, and many state legislatures have moved to enact DOM- like provisions defining marriage as only between a man and a woman. With few exceptions, public opinion on contemporary gay rights issues is passionate and closely divided. While substantial majorities to this day are aligned against gay marriage, those in favor of it are no longer an inconsequential minority. Bare majorities exist in favor of civil unions.20 Nearly half of respondents reported that they have a friend, family member, or colleague who is gay.21 And strong majorities (around seventy percent) support extending benefits, social security, and inheritance rights to homosexual couples.22 But the consensus that has developed on benefits

19 ABC News Washington Post Poll, January 1993. 549 respondents. “Do you think people who join the military should be asked if they are homosexual, or not?” 20 Fox News / Dynamics Poll, May 2004. 900 respondents. “Do you believe gays and lesbians should be allowed to get legally married, allowed a legal partnership similar to but not called marriage, or should there be no legal recognition given to gay and lesbian relationships?” 21 Princeton Survey Research Associates / Pew Research Center for the People & the Press Political Typology Callback Poll, March 2005. 1090 respondents. “(And one last short list.)...Do you have a friend, colleague, or family member who is gay?” 22 Princeton Survey Research Associates / Henry J. Kaiser Family Foundation Views on Issues and Policies Related to Sexual Orientation Survey, February 2000. 2283 respondents. “(Next I'd like your opinion on

104

and job opportunities are the exception. The rule is close-quarters, evenly divided conflict. The

American public is closely divided on whether there should be a constitutional amendment

defining marriage as between a man and a woman.23 Americans are evenly divided on the

morality of homosexuality (49/49)24 and gay adoption (40/52 against).25

DATA SOURCES & VARIABLES - PUBLIC ATTITUDES TOWARDS GAY RIGHTS

While a general discussion of the picture of public attitudes on gay rights painted by various and sundry opinion polls over the last 40 years provides a sense of the movement from consensus to conflict—from relative peace to cultural war on gay rights—it does not establish the trend in a rigorous and systematic manner. Are gay rights a burgeoning front in the culture war? Did we have a consensus on homosexuals that has since eroded into two camps—roughly balanced, passionate, and hostile? Has the entry of gay rights in to the continuum of potential vote-getting political issues affected the disposition of the parties on gay rights as well as the policy process? Given these are empirical questions, empirical data is necessary to strike at an answer. I employ two separate data sets to assess the breadth and depth of public opinion on homosexuals and homosexual rights in the United States.

Data

The first set of data is a database of 689 polling questions on attitudes towards gays and gay rights compiled from 1971 to 2007. The poll questions and responses were culled from the

some gay rights issues.)...Do you think there should or should not be...health insurance and other employee benefits, Social Security, and inheritance rights for gay and lesbian domestic partners?” 23 Gallup / CNN / USA Today Poll, July 2003. 1003 respondents. “Would you favor or oppose a constitutional amendment that would define marriage as being between a man and a woman, thus barring marriages between gay or lesbian couples?” 24 CBS News / New York Times Poll, December 2003. 1057 respondents. “Do you think homosexual relations between adults are morally wrong, or are they okay, or don't you care much either way?” 25 Los Angeles Times Poll, March 2004. 1616 respondents. “Do you favor or oppose gay couples legally adopting children? (If Favor/Oppose, ask:) Do you strongly favor/oppose gay couples adopting children or only somewhat favor/oppose gay couples adopting children?”

105

IPOLL database developed by the Roper Center for Public Opinion Research.26 The observations

for the Gay Rights Database (GRD) were selected from the IPOLL database using a search

function that looked for polling questions with the word “gay” or “homosexual” in them. This

search found 1523 polling questions that met the selection criteria. From there I chose polling

questions that reflected public attitudes towards gays, homosexuals, gay rights, association with

gays, gay marriage, gays in the military, etc.

Excluded from the data set were fact-based questions, such as “On a scale of one to

seven, how would you rate your own personal risk of contracting 'AIDS'?”27 Also excluded where questions that asked the respondent about the future, such as “In the future, do you think 'AIDS' (Acquired Immune Deficiency Syndrome) will affect more people who are homosexual, or gay, or do you think it will affect more people who are heterosexual, or straight?”28 Some questions asked the respondent to rate one of the political parties,

government, or politicians have acted on gay rights. Since it was not clear from the question

whether the respondent thought the performance on gay rights issues was good or bad, those

questions were excluded as well. In short, if a question was determined to be unrelated to

public attitudes on gay rights, or what that attitude is was not clear, the question was excluded.

In order to keep a relatively even distribution across questions and to avoid repetitive entries,

some attitude-related questions were excluded when a sufficient number of questions on that

attitude had been obtained for that year.

26 IPOLL is a database of nearly half a million polling questions from 150 polling organizations catalogued since 1935. It includes data survey results from academic, commercial and media survey organizations such as Gallup Organization, Harris Interactive, Pew Research Associates, and many more. The data come from all the surveys in the Roper Center archive that have US national adult samples or samples of registered voters, women, African Americans, or any subpopulation that constitutes a large segment of the national adult population. 27 Los Angeles Times Poll, December 1985. 2308 respondents. 28 Los Angeles Times Poll, December 1985. 2308 respondents.

106

The response categories for each question were collapsed into a binary set: Pro-Gay

and Anti-Gay categories. This was done to facilitate cross-poll comparisons. The minimum

number of questions in the data base for each year is one polling question, the maximum

number is 76 (2004) with an average of 21.5 questions per year.29 There are polling questions

from 103 different polls in the database.30 There are a number of polls from which there is just one question included in the GRD. The most polling questions (74) in the database are from the

PSRA / Newsweek Poll. The average number of respondents per poll in the GRD is 1270.69 with a minimum of 549 respondents (ABC News / Washington Post Poll, January 1993) and a maximum of 8769 respondents (CBS News Exit Poll, November 1978).

The second data set consists of counts and statistics generated on the homosexual relations question asked consistently since 1973 in the General Social Survey.31 The exact question wording and response frequencies for the GSS attitude on homosexual relations item is reported in Appendix F.

Both the GRD and the GSS time-series data have advantages and disadvantages in examining trends in public attitudes toward homosexuals and gay rights. The GRD covers a larger variety of gay rights issues, and the time series of GRD polling questions permits an analysis of the specific gay rights issues contemporaneous to when they entered the public agenda and to track them over the period that they were a hot public issue. The GRD allows for a more nuanced assessment of attitudes towards gay rights and thus a more valid assessment of the continuum of gay rights public opinion. It also provides for a variety of different question wordings on the same issue, thus decreasing the chance that question wording overstates or

29 See Appendix C for a frequency table of polling questions per year for the time series. 30 See Appendix B for a list of polls included in the GRD. 31 Statistics generated from the General Social Survey Cumulative File, 1972-2006 courtesy of the Interuniversity Consortium for Political and Social Research.

107

understates gay rights attitudes for the entire sample. But this is a double-edged sword, as the

different formulations on a particular issue can influence respondent answers and hence muddy

the waters. Differences in responses may be a consequence of different question wording and

not an actual change in public opinion on gay rights. While the GSS data cannot differentiate

between opinion on gay marriage and civil unions, or gays in the military versus social security

benefits for gay couples, it does provide a consistent question wording across the time series.

This makes it potentially a more reliable measure of public opinion on gay rights, but it is

restricted to just the single aspect of gay rights covered by the GSS question on homosexual

relations.

Variables

The two primary source variables for the analysis are the Pro-Gay / Anti-Gay collapsed

(when necessary) binary variable from the GRD, and the attitudes towards homosexual relations

variable from the GSS. As is evident from the breakdown of the GSS variable in Appendix F, the

survey years 1975, 1978, 1983, and 1986 did not ask the “attitudes towards homosexual

relations” question. I interpolate those data points for the GSS analyses using the two most

proximate data points (survey year prior, survey year after) and the simple linear interpolation

method.32 This is done for easy of graphing, as exclusion of the interpolated data points does not affect the substantive results of the analyses.33

The primary classification variable for the GRD is the issue type variable, used to subset the data set by specific gay rights issues. Appendix D has a description of the kinds of questions included in each of the categories of the issue type variable. I will briefly review the categories

32 Generally, linear interpolation takes two data points, defined here as (xa,ya) and (xb,yb). The ) interpolant is given by: + ( ) at the point (x,y). ) ሺ௬್ ି௬ೌ 33 Models for the GSS analyses were run with and without the interpolated data points. There were no ݕൌݕ௔ ݔെݔ௔ ሺ௫್ ି௫ೌ significant changes in the model coefficients, standard errors, or the goodness-of-fit-measures.

108

used in this analysis. In the time-series conflict/consensus trend analysis, the operative rule was that a category must have a minimum of 14 poll questions to be included. The self-explanatory categories in the issue type variable are: gay marriage, gay adoption, morality of homosexual relations, the gay rights movement, civil unions, constitutional amendment, comfort, teach, inheritance, and benefits. All of those categories include questions on those clear gay rights topics (see Appendix D for further clarification). A bit less clear are the following: The “vote” category includes questions that ask the respondent whether they would vote for a candidate they knew to be homosexual. The “candidate” category includes questions where respondents indicate that they think whether or not the candidate for office is important, whether it would influence their vote, and if the public has a right to know that information. The “friends” category includes questions where the public reports on whether they have friends, associates, colleagues, or relatives who are gay. The “jobs” category references questions on homosexual job discrimination and equal opportunity rights in jobs for homosexuals. The “military” category refers to questions both about permitting openly gay soldiers to serve and the “Don’t Ask, Don’t

Tell” policy.

The year variable is self-explanatory. It was re-coded for the regression analyses

(ranging from 1 to 32 for the GDRB and 1-25 for the GSS) for ease in reporting the intercept parameter estimates in the tables. The party identification variable used for the GSS analyses is the standard ‘PARTYID’ variable used in their semi-annual surveys. The GSS uses a slightly different set of questions to produce the standard 7-point party identification scale. The GSS classifies partisan identifiers as either “strong” or “not very strong” Democrats or Republicans.

The independent / leaner category asks respondents which of the two parties they are “closer” to. Despite the different question wording, the scaling of the GSS party identification variable is identical to that of the party ID variable in the NES (0-6).

109

TABLE 5.1: GAY RIGHTS AND HOMOSEXUAL POLITICAL ISSUES BY QUESTION TYPE 1971-2007

MEAN ANTI-GAY 1ST YEAR ISSUE TYPE FREQ PER ANTI-GAY PER S.D. APPEARS GAY RIGHTS 53 7.69 49.48 13.61 1971 OFFSPRING 17 2.47 68.40 25.20 1974 CHRISTIAN 5 0.73 40.01 4.02 1977 JOBS 46 6.68 28.58 19.98 1977 LEGALITY 46 6.68 47.38 10.60 1977 CANDIDATE 41 5.95 55.36 15.93 1978 COMFORT 20 2.9 44.63 13.04 1978 MORALITY 34 4.93 59.44 12.04 1978 TEACH 18 2.61 52.11 16.61 1978 BORN / IMPORT 3 0.44 52.59 9.39 1983 MILITARY 34 4.93 45.69 12.32 1983 AIDS 17 2.47 51.97 26.60 1985 FRIENDS 15 2.18 53.20 20.90 1985 VOTE 14 2.03 51.88 17.89 1985 MOVEMENT 17 2.47 58.20 18.45 1987 PRIVACY 7 1.02 32.41 16.67 1987 GAY MARRIAGE 105 15.24 65.34 7.23 1988 ADOPTION 25 3.63 58.42 11.67 1989 CHURCH 7 1.02 56.76 7.72 1989 T.V. 12 1.74 57.11 12.66 1989 SPEECH 4 0.58 46.66 11.98 1990 BENEFITS 21 3.05 43.56 10.41 1991 HOMOSEXUALS 13 1.89 51.48 12.68 1992 HOUSING 4 0.58 18.88 7.038 1992 INHERITANCE 11 1.6 35.71 13.50 1992 PARTIES 19 2.76 45.03 7.07 1992 CIVIL UNIONS 36 5.22 50.34 6.85 1993 DOCTOR 4 0.58 47.65 9.22 1993 READ 1 0.15 64.84 . 1993 DENTIST 1 0.15 57.29 . 1994 SHOP 4 0.58 16.87 5.80 1994 PARENTS 3 0.44 45.37 13.14 1996 LIFESTYLE 1 0.15 34.02 . 1998 JUROR 1 0.15 12.77 . 1999 BOY SCOUTS 5 0.73 47.13 12.33 2000 AMENDMENT 1 0.15 56.99 . 2003 CONSTITUTION 24 3.48 50.12 6.81 2003

110

ANALYSIS – AGGREGATE POLARIZATION TRENDS IN PUBLIC OPINION ON GAY RIGHTS

Simple statistics for the gay rights analysis are reported in Table 5.1. The most frequent category in the GRD data set is gay marriage, largely a function of the number of polls included from 2004 and the fact that the number of polls and amount of polling has steadily increased over the last few decades. Also reported in Table 5.1 are the mean Anti-Gay attitudes in the GRD for the entire data set. The most consensual anti-gay rights attitude for the entire data set is the attitude towards the prospect of one’s child deciding he or she is gay. The most consensual pro- gay rights attitude is on gays having equal job opportunities and opposed to discrimination against gays in the workplace. The most contentious gay rights issues are civil unions and whether there should be a constitutional amendment to define marriage as only between a man and a woman. It is not a coincidence that these are also two of the more recent gay rights issues to enter the public agenda, both falling within the “culture wars” time horizon (1993 and 2003).

The only significant exceptions to the ‘recent = conflict’ rule are gay marriage and inheritance rights. Gay marriage has been consistently opposed by a strong majority of citizens since it was introduced to the policy environment. Inheritance rights have likewise enjoyed majority support since pollsters in the GRD began asking respondents to rate their opinion on gay rights to inheritance. If you take 1990 as a cut point and look at the issue categories with 10 or more polling questions, only inheritance falls outside the 43 – 52 range (Table 5.1). While a consensus may have emerged on granting homosexuals inheritance rights, the same cannot be said about gay marriage. When coupled with civil unions, gay marriage is one of the more contentious contemporary gay rights issues. Fourteen states have adopted gay marriage or civil union statutes (Figure 5.2). Furthermore, in response to the gay marriage ‘threat,’ a number of states have amended their constitutions to forbid gay marriage.

111

FIGURE 5.2: GAY MARRIAGE IN THE 50 STATES AND UNITED STATES TERRITORIES*

KEY SAME-SEX PARTNERSHIPS LAWS IN THE U.S.

Same-sex marriage

Unions granting rights similar to marriage

Legislation granting limited/enumerated rights

Same-sex marriages performed elsewhere recognized

No specific prohibition or recognition of same-sex marriages or unions

Statute bans same-sex marriage

4 Constitution bans same-sex marriage

* http://en.wikipedia.org/wiki/File:Samesex_marriage_in_USA.svg Constitution bans same-sex marriage and some or all other kinds of same-sex unions

112

Though gay marriage is a loser on the national stage (not unsurprisingly, both 2008

presidential candidates were against it), the strong minority on gay marriage has significant

political influence in the Northeast and the Northwest. In contrast, gay marriage is strongly

opposed in the Southern and Western states (otherwise known as fly-over country). As we saw

with Lawrence v. Texas as well as the Massachusetts and California State Supreme Court decisions ruling gay marriage to be a constitutional right, legal elites have pushed gay marriage on to the public agenda and thus sparked a national debate, interest group activity, referendums, and counter-legislation.

CONFLICT AND CONSENSUS MODELS ON SPECIFIC GAY RIGHTS ISSUES, 1971-2007

Tables 5.2 and 5.3 report the regressions of the consensus measure, by issue type, on the time series in the GRD. It is important to note that each regression is of the time trend for the years in which that type of polling question was asked. So a null-result doesn’t necessarily mean that this issue hasn’t become more conflictual or more consensual over the last three decades. It merely means that there is no statistically significant trend over the years that it was asked in that time. This points to one of the problems with assessing trends in social (or any kind of issue) issues over any significant time period using polling and survey data. Many issues are only asked about in public opinion polls once they have slipped from consensus to conflict (and, thus, become a subject of political competition). This is almost certainly why contentious issues like gay marriage, civil unions, and the constitutional amendment have null findings in the trend models. For the years where we have data, those issues haven’t changed much. Civil Unions was a 54/42 split against in 2000. The constitutional amendment breakdown was 45/50 for in 2003, when it was first asked. So there is no apparent trend towards conflict in the regression models.

If, however, we had data on those questions for the entire time series, we would likely see a trend from consensus to conflict. There is a selection effect working against finding

113

TABLE 5.2: CONFLICT TREND MODELS FOR GAY RIGHTS ISSUES 1971-2007 Trend: MODEL: Conflict or Intercept Parameter Estimate 2 CM = B0 + B1(YEAR) + E Consensus (S.E.) (S.E.) R N ISSUE CATEGORIES

Gay Rights CONF 79.310 -0.980 *** .032 685 Full Sample (5.686) (0.206)

Adoption CONF 159.834 -4.0315 *** .414 25 (28.396) (0.999)

Candidate CONF 87.626 -1.572 * .081 41 (20.305) (0.849)

Friends CONF 180.763 -4.897 *** .421 15 (41.376) (1.593)

Morality CONF 128.538 -2.946 *** .316 34 (22.119) (0.766)

Vote CONF 276.547 -7.519 *** .875 14 (25.068) (0.822)

Homosexuals ----- 46.021 -0.078 .001 14 (0.579) (0.977)

Movement ----- 92.657 -0.826 .011 17 (52.067) (2.077)

Parties ----- 48.346 -0.638 .034 17 (27.238) (0.847)

Civil Unions ----- 64.121 -1.260 .035 36 (37.791) (1.132)

* significant at .10 level ** significant at .05 level ***significant at .01 level

114

significant trend results for the more recent gay rights issues. If we don’t have data on the gay

rights issue for the 1971-1985 period when the consensus on gay rights existed—nor the period

of 1986 – 1992, when the big jump from conflict to consensus on gay rights occurs—then we are

unlikely to find a trend either way. Not coincidentally, pollsters failed to ask about civil unions

and constitutional amendments against gay marriage in the 1970’s. That these issues are the

subject of contentions debate in the policy arena today is powerful evidence that attitudes

towards gay rights have changed greatly over the last three decades.

Table 5.2 reports on the regression models of the consensus measure (CM) on the time-

series that have negative coefficients, indicating that those issues have become more conflictual

since the first year the pollsters began polling on them. The negative coefficient indicates a

decline in consensus in the mass public on the issue. The “Gay Rights” model is for the entire

GRD over the entire time-series. While the model fit is rather low (R2 =.032), there is statistically significant evidence that, on the whole, attitudes towards gay rights have moved from consensus to conflict. The regression coefficient indicates that gay rights attitudes have lost on average about a point on the consensus scale for each year of the time series. This result is substantively important, despite the low model-fitness statistic, when we consider how much of the deck is stacked against finding a statistically significant finding. As we know from Table 5.3, several of the gay rights issues move from conflict to consensus. Furthermore, several of the more conflictual gay rights issues were not collected over the full time-series and hence their almost certain movement from consensus to conflict is in the error term. So the fact that we find statistically significant findings for the full time-series is somewhat remarkable to begin with.

115

The largest movement from consensus to conflict is in the vote category, with respondents reporting more of a willingness to vote for a gay candidate by 7.52 points a year on average. The time-series independent variable accounts for nearly ninety percent of the variance in the model (.875). The consensus on gay adoption and respondents reporting they do not have gay friends has crumbled to the tune of about a four point loss in consensus each year.

While it is possible the number of gays in society has increased exponentially, the more likely explanation is that gays are no more comfortable being ‘out’ in their social settings and heterosexuals are more willing to admit associations with homosexuals. Both of those models account for about 40 percent of the model variance. The consensus on the immorality of homosexual relations (-2.95) and on whether a candidate’s homosexuality is a public issue (-

1.58) have also significantly declined, though the R2 for the candidate model is rather low (.081) in comparison to the other significant conflict models. The signs for the homosexual, gay rights movement, party, and civil unions are all in the conflict direction, but do not meet conventional standards of statistical significance. The conflict models in Table 5.2 indicate marked increase in conflict on gay rights as a whole and on a variety of fronts.

The story of gay rights attitudes in America is not exclusively a tale of conflict to consensus, as illustrated in Table 5.3. Table 5.3 reports on the gay rights issues which have moved from conflict to consensus over the course of the time-series, using the consensus measure (CM). There are three gay rights issues that have become more consensual since 1971.

The largest shift towards consensus is in public attitudes towards gays in the military. There was always a rather obvious disconnect between opposition to gays in the military and the overwhelming percentage of Americans who believe that gays should have equal job opportunity and workplace rights. It is clear from the four point per year increase in consensus on gays in the military that the argument to treat the military differently than other kinds of jobs

116

has not been persuasive. In May, 2007, seventy-eight percent of respondents asserted that

openly gay individuals should be able to serve in the military. Despite this, no official change in

the Don’t Ask, Don’t Tell policy has been adopted (though there have been a few rumblings

along those lines). This may be more a case of policy priorities than a real reluctance to change

the policy. On the other hand, a slight majority of Americans (52%) report support for DA,DT

when presented with the alternative of openly gay soldiers or forbidding gay soldiers from

serving entirely.34 As already noted previously, the American public has been supportive of

equal opportunity in jobs for homosexuals from the beginning of the time-series in the early

1970’s. This has only increased over the last thirty-plus years, by about three points of

consensus a year. The progress of time explains about forty percent of the variance in consensus

for the model years. Lastly, a consensus has formed on comfort with homosexuals. This is likely

related to the increasing number of homosexuals who are portrayed positively in the media and

on television. While Ellen as an openly gay sitcom character was highly controversial in the

1990’s, in a post Will & Grace world, a prominent gay leading character hardly merits comment in contemporary society.

Among the positive coefficient models that lack statistical significance, the model that is least likely to belong in this table is the constitutional amendment model. A public agenda issue at least as early as 2003, the American public has been evenly divided on the question for every year since. Opinion on the legality of homosexual relations, surprisingly, has not changed much over the course of the time-series despite data having been collected on it since 1977. The decision in Lawrence v. Texas is out-of-step with the majority of Americans. This

34 Gallup / USA Today Poll, July 2007. N = 1014 Respondents. “As you may know, under the current military policy, no one in the military is asked whether or not they are gay. But if they reveal that they are gay or they engage in homosexual activity, they will be discharged from the military. Do you personally think--gays should be allowed to serve openly in the military, gays should be allowed to serve under the current policy, or gays should not be allowed to serve in the military under any circumstances?”

117

TABLE 5.3: CONSENSUS TREND MODELS FOR GAY RIGHTS ISSUES 1971-2007 Trend: MODEL: Conflict or Intercept Parameter Estimate 2 CM = B0 + B1(YEAR) + E Consensus (S.E.) (S.E.) R N ISSUE CATEGORIES

Comfort CONS 4.008 1.594 * .144 20 (24.764) (0.917)

Military CONS -63.660 4.002 *** .323 34 (26.622) (1.023)

Jobs CONS 51.523 3.008 *** .422 46 (11.095) (0.531)

Legality ----- 23.624 0.451 .019 45 (12.486) (0.486)

Teach ----- 14.763 1.820 .145 18 (26.867) (1.103)

Constitution ------104.871 4.724 .033 25 (142.709) (4.435)

Benefits ----- 29.099 0.394 .003 20 (46.733) (1.647)

Gay Marriage ----- 50.944 0.355 .003 105 (18.570) 0.579

* significant at .10 level ** significant at .05 level ***significant at .01 level

118

is also the case with gay marriage, suggesting perhaps that public opinion has stabilized, if at a

much more conflictual baseline than in the past.

EXPLORING THE PATH FROM CONSENSUS TO CONFLICT ON GAY RIGHTS

One of the methodological criticisms of the collapsed-category method of examining trends in polarization is that it—through the combination of strong and less-strong respondents—overstates the oppositional nature of public attitudes. First, this criticism is mostly inapplicable to the polling data set, as most of the questions have binary responses sets to begin with. However, one could argue those response sets themselves are too restrictive on opinions and attitudes in the American public. While flat distributions may be described as polarized using this method, it cannot affect polarization over time.

The fact that the use of a binary measure of public opinion (either through the original response set or collapsed categories) is consistent over the time period covered by the data set vitiates against a false positive on polarization due to the number of categories. I am looking at trends over time on these measures, and hence trends over these consistently coded variables cannot be a consequence of using a binary response measure (the number of categories is constant throughout the data set). Second, collapsing the categories and thus treating all the polls as if they had binary response sets permits one to compare polling questions with categorical response options that vary. So one must either sacrifice validity through collapsed measures or sacrifice coverage in restricting the analysis to just questions with the same number of categories in their response set. Furthermore, the binary measure accurately and consistently captures the distribution of opinion on gay rights as to conflict versus consensus. It permits the identification of conflict or consensus, the questions themselves permit variation in the centrality vs. extremism of the gay rights issue (ranging from forbidding speech and job opportunities to permitting gay marriage and providing social security benefits), and the

119

characterization of debate on these issues in terms of two opposing viewpoints is a good proxy for the actual distribution of opinion in the mass public. Lastly, as illustrated in Table 5.4, the binary response measures may actually understate the degree of polarization in the public attitudes towards homosexual rights (note that the binary measures average about half of the variation in dispersion when compared to the four category measures).

In other words, it cuts both ways. While using a binary variable may overstate the polarization of a normally distributed variable or a relatively flat distribution, it would understate the polarization of an extreme bipolar distribution. That said, it is true that the use of a dichotomous measure may suggest polarization and conflict when the opinion distribution is relatively flat across a continuous measure. Thus the decision to use a dichotomous measure so as to include the maximum number and breadth of questions on gay rights represents a sacrifice in the validity of the measure. However, there remains an open question as to whether gay rights exhibits the flat distribution in opinion that would make the measure of consensus used here invalid. If gay rights opinion is not such a distribution, then this greatly mitigates the measurement problem. As such, I examine gay rights issues without collapsing categories by looking at a representative set of important gay rights issues from 2004 in Figures 5.3 reflecting the polarization on gay rights issues and a measure of the consensus that existed in the 1970’s in

Figure 5.5. In Figure 5.4 I generate a hypothetical ‘normal’ distribution on gay rights issues to provide a baseline for comparison.

120

FIGURE 5.3: POLARIZED ATTITUDES ON GAY RIGHTS, 2004

FIGURE 5.3A: GAY MARRIAGE CONSTITUTIONAL AMENDMENT FIGURE 5.3C: LEGALITY OF GAY MARRIAGE

50 55 45 50 40 45 35 40 35 30 30 25 25 20 Percent percent 20 15 15 10 10 5 5 0 0 strongly somewhat somewhat strongly legal strongly legal somewhat illegal illegal strongly approve approve disapprove disapprove somewhat

FIGURE 5.3B: GAY ADOPTION FIGURE 5.3D: GAY RIGHTS

50 50 45 45 40 40 35 35 30 30 25 25 20 percent 20 percent 15 15 10 10 5 5 0 0 favor strongly favor oppose oppose strongly approve approve disaprove disapprove somewhat somewhat strongly somewhat somewhat strongly

121

FIGURE 5.4: HYPOTHETICAL “CENTRAL” DISTRIBUTION ON GAY RIGHTS

50 45 40 35 30 25 20 percent 15 10 5 0 strongly somewhat somewhat strongly approve approve disapprove disapprove

FIGURE 5.5: CONSENSUS DISTRIBUTION ON THE UNACCEPTABILITY OF HOMOSEXUAL RELATIONSHIPS, 1974

100 90 80 70 60 50 percent 40 30 20 10 0 acceptable not acceptable

122

A COMPARATIVE ANALYSIS OF GAY RIGHTS ATTITUDE POLARIZATION ON SELECT MEASURES

Figure 5.3 reports four sets of responses to questions about contemporary gay rights issues from 2004: gay adoption, the legality of gay marriage, a constitutional amendment to define marriage as between a man and a woman (thus constitutionally barring gay marriage), and general attitudes towards gay rights (see Appendix C for exact question wording and poll information). Substantively speaking, these polls, ranging the breadth of contemporary gay rights issues, demonstrate that the American public has polarized on gay rights. All four of the distributions are bimodal or near-bimodal with more mass located in the extreme categories relative to the central categories (the extreme categories are also the two largest frequency categories for each of the issues). Figure 5.4 displays a hypothetical four-category centralized distribution with eighty-percent of the mass of public attitudes located in the two center categories. Figure 5.5 shows the consensus that existed on gay rights attitudes towards offspring being gay in the 1970’s. Over ninety-percent of respondents to the Virginia Slims poll reported that they would find their kid being gay unacceptable.35 The difference between the consensus against homosexual relations in contrast to the near-evenly divided and highly polarized distributions (conflict) in 2004 is plain.

Table 5.4 reports the univariate statistics on the four 2004 gay rights questions from

Figures 5.4, 5.5, and 5.6. These statistics capture two important aspects of polarization: dispersion (standard deviation, coefficient of variation) and bimodality (kurtosis). Table 5.4 also includes the Virginia Slims question on the acceptability of homosexual relations for ones offspring from 1974, reflecting the consensus position against homosexuality at that time. For a baseline comparison, I include the hypothetical near-normal four category variable with most of

35 Virginia Slims American Women's Poll, April 1974.

123

TABLE 5.4: DISPERSION & BIMODALITY STATISTICS FOR SELECTED GAY RIGHTS POLLS & HYPOTHETICAL DISTRIBUTION Standard Coefficient Poll Year Mean Deviation Kurtosis of Variation Gay Adoption 2004 2.783 1.232 -1.542 44.290 Constitutional 2004 2.467 1.323 -1.756 53.617 Amendment Gay Rights 2004 2.856 1.269 -1.508 44.423 Gay Marriage 2004 2.600 1.312 -1.743 50.492 Gay Adoption (binary) 2004 0.565 0.496 -1.934 87.793 Constitutional 2004 0.299 0.458 -1.223 153.261 Amendment (binary) Gay Rights (binary) 2004 0.463 0.499 -1.980 107.752 Gay Marriage (binary) 2004 0.482 0.500 -1.997 103.783 Homosexual 1974 0.979 0.143 43.186 14.591 Relationship Hypothetical Centralized Distribution ---- 2.5 0.807 -0.482 32.262

the mass of opinion located in the central categories (80% in center two categories, 20% in extreme categories) from Figure 5.4.

Table 5.4 points to the answers to several of the puzzles regarding polarization. It both illustrates the polarized nature of opinion on gay rights today vis-à-vis that in the 1970’s and provides a validity check for the polarization measures used in the aggregate analysis.36

Comparing the standard deviations of the Virginia Slims distribution, that of the hypothetical centralized distribution, and then the four gay rights issues from 2004 (4 cat), the linear increase in dispersion on the attribute is apparent. The standard deviations for the binary, collapsed opinions on gay rights from 2004 (0.500, 0.499, 0.496, and 0.458) are substantially larger than that of the opinion from 1974 (0.143). It is also the case with the coefficient of variation (14.591 for the 1974 poll; 32.262 for the hypothetical CD, and a 48.21 average COV for the 4 polls from

2004). The mean for the opinion on homosexual relationships from 1974 is near 1, indicating the

36 Note, converting the polarized four-category distributions to binary distributions doesn’t have much effect on the relative comparisons of these distributions on conflict/consensus, and if there is one it is to understate the polarization of the attribute distribution when the underling distribution is bimodal.

124

near-consensus of opinion at that time while the means for the opinions from 2004 are near .5

(much more likely to indicate conflict).37

The kurtosis results are consistent with what we know about the included distributions and strong evidence of the increase in bimodality from the 1970’s to the 20o0’s. The four contemporaneous measures of public attitudes towards gay rights (four-cat or binary) all are bimodal, highly conflictual, and polarized distributions both from an absolute sense and in comparison to the opinion on gay rights prior to the onset of the culture war. Both the full and collapsed category measures of gay rights opinion in 2004 are platykurtic and hence indicate bimodality. The Virginia Slims poll on attitudes on homosexual relationships from 1974 reflects the consensus against homosexuals that existed at that time. The standard deviation for the

1974 poll is quite small (0.143) and small relative to the mean, as illustrated in the coefficient of variation (14.591). The kurtosis for this poll is off the charts leptokurtic (43.186), reflective of the single-peaked distribution and the near-total mass of responses located at it (no bimodality).

Though the kurtosis measure inaccurately indicates slight bimodality in the centralized distribution, an ordinal comparison of the kurtosis results is consistent with the consensus-to- conflict expectation. It is thus helpful in looking at the relative peakedness differences between the distributions. Furthermore, note that the kurtosis of the distributions for both the two- category and the four-category versions of the polarized polling questions are relatively similar.

This tends to weigh against Fiorina’s argument that the Abramowitz transformation (and, by

37 The difference in the means for the four polarized distributions and the hypothetical centralized distribution is negligible (2.67 vs. 2.5). As I argued in Chapter 3, you can have highly dispersed distributions with similar means to that of centralized distributions. While means can change as a consequence of polarization they do not necessarily do so. The direction of the change if it does change, and thus whether it indicates polarization or depolarization, is dependent on the change relative to the previous distribution. A significant mean shift to the center is polarization given a previous consensus position, while it is depolarization given a previous bimodal opinion distribution. Changes in means are not interpretable in terms of polarization/depolarization taken independent of the prior distribution.

125

implication, the transformation here) artificially inflates polarization, at least with respect to

bimodality.

If we compare kurtosis relative to the other distributions, it illustrates the difference

between a consensus position and a centralized distribution. The normal (or near-normal)

distribution is not, as some skeptics treat it, the ultimate expression of consensus or

moderation. This is tangible evidence reinforcing the theoretical argument I made in Chapter 3:

whether or not a shift to a normal distribution on an attribute is indicative of depolarization or

polarization is dependent on the shape and dispersion of the prior state of the distribution. In

the case of gay rights, the aggregate shift to an on average ‘moderate’ position from the

consensus position evident in 1974 is polarization. Aggregate moderation, in this instance, is

evidence of increasing conflict (i.e. polarization). As we can see from the distributions in 2004

relative to that in 1974, opinion on gay rights has shifted from one extreme (consensus -

unimodality) to the other (conflict - bimodality).

POLARIZATION TRENDS IN GAY RIGHTS ATTITUDES, GSS 1973-2006

The consistent question-wording of the GSS homosexual relations item permit the direct comparison of polarization models on gay rights. The basic univariate statistics on gay rights from the GSS time-series are reported in Table 5.5. Note the sixth-tenths of a point increase in the mean as well as the three-tenths of a point increase in the standard deviation. Given the variable is scaled from zero to four; these are significant changes in the aggregate public opinion on homosexual relations towards greater polarization on gay rights over the time series.

While it is true the mean is moving towards the center on gay rights issues, as I noted previously, this reflects a move towards greater disagreement in the American public on gay rights. It is evidence of conflict, not, as some have concluded, moderation. Changes in means, as

126

they relate to polarization, must be assessed in context: in relation to the shape of the distribution prior to the change and what the direction of the shift in the mean indicates about that change. Take the example of a move of the mean towards the center of the scale. Some culture war skeptics have misinterpreted the change in mean attitudes on social issues

(attitudes that have become more centrally located on the NES scales) as evidence against polarization. We see this change in means with gay rights (i.e. trend of the mean moving towards the center of the attitudinal scale). But, in fact, this change is actually evidence of polarization. The prior distribution was consensual and unimodal. The shift of the mean towards

TABLE 5.5: GSS – PUBLIC OPINION ON HOMOSEXUAL RELATIONS 1973-2006 YEAR MEAN STAND DEV KURTOSIS ECD SAMPLE N 1973 1.559 1.037 0.809 61.7 1,448 1974 1.620 1.099 0.246 57.5 1,412 1975 1.658 1.125 0.016 55.9 - 1976 1.696 1.151 -0.214 54.2 1,426 1977 1.654 1.127 0.040 57.0 1,453 1978 1.637 1.119 0.171 57.9 - 1980 1.620 1.111 0.302 58.7 1,397 1982 1.587 1.086 0.532 61.4 1,771 1983 1.607 1.100 0.364 60.2 - 1984 1.627 1.114 0.195 59.0 1,412 1985 1.590 1.097 0.449 61.6 1,484 1986 1.552 1.068 0.821 63.9 - 1987 1.514 1.039 1.193 66.3 1,750 1988 1.544 1.063 0.915 64.0 937 1989 1.633 1.139 0.144 58.5 980 1990 1.555 1.068 0.810 63.5 872 1991 1.609 1.134 0.332 59.5 926 1993 1.851 1.264 -0.965 44.3 1,012 1994 1.863 1.281 -1.015 43.2 1,884 1996 2.021 1.340 -1.434 32.2 1,784 1998 2.076 1.349 -1.534 28.6 1,753 2000 2.068 1.348 -1.533 30.0 1,697 2002 2.182 1.382 -1.710 22.0 884 2004 2.108 1.366 -1.606 26.8 868 2006 2.174 1.379 -1.696 22.3 1,908

127

the center, in this case, indicates a loss of that consensus, as the consensus position from the

1970’s and the 1980’s was located near the extreme of one side of the gay rights attitude scale.

The change in the standard deviation permits a more straightforward interpretation: increasing standard deviations connote increasing dispersion which connotes increasing political polarization on gay rights.

The ECD measure (extreme category difference: top category minus bottom category on the scale) is the difference between the percentage of respondents that chose the “always wrong” category and those that chose the “nothing wrong at all” with homosexual relations category. This is an important measure with respect to polarization because it provides a direct measure of the changes in the mass at the tails of the opinion distribution on gay rights. Recall, it isn’t enough to have extreme opinions on an issue represented in society in order to generate social conflict. Those opinions have to be held by a sufficient mass of the citizenry in order to provide a venue for political and partisan activism and conflict. If the two extremes have equal mass, then the ECD measure will be zero. The ECD measure shows that there was a large difference between the two theoretical extremes of the gay rights opinion in the 1970’s and into the 1980’s, reflecting the consensus against gay rights that existed at that time. The declining difference between the two extremes tracks the loss of mass at the “anti-gay rights” extreme and the increasing mass at the “pro-gay rights” extreme. This decline in the difference between the two extreme categories is evidence of an increase in dispersion and bimodality. This difference was at its maximum in 1988, right around the point where public opinion on gay rights began to shift. The low water mark (illustrating the equalization trend in the two categories, i.e. conflict) was the 22 percent difference in 2002. There is a quite marked decline in the consensus on gay rights, with a large number of citizens declining to rate homosexual relations as “always wrong” and a corresponding increase in the number of Americans who

128

believe there is nothing wrong with homosexual relations. Also reported in Table 5.5 is the

sample N for each year the GSS collected data on the homosexual relations questions. Recall

that the statistics for the 1975, 1978, 1983, and 1986 years were interpolated (hence there is no

sample N to report).

Table 5.6 reports the polarization regressions of the trends in the mean, standard

deviation, coefficient of variation, kurtosis, and extreme category difference models. All indicate

a trend towards greater dispersion, and hence increasing conflict on gay rights—increasing

polarization in the attitudes of the American public on homosexuals.38 The best model, in terms

of goodness-of-fit (.669), is the trend model on the standard deviation in gay rights attitudes.

This demonstrates a significant increase in dispersion in opinion on gay rights for the mass

public, a key aspect of polarization. The positive coefficient (0.01) indicates an average increase

in the standard deviation (i.e. dispersion) of opinion on homosexual relations. The positive

coefficient in the mean model (0.019) shows that the average

TABLE 5.6 TRENDS IN POLARIZATION OF ATTITUDES ON HOMOSEXUAL RELATIONS 1973 - 2006 MODEL: EXTREME CAT GR = B0 + B1(YEAR) + MEAN MODEL S.D. MODEL KURTOSIS DIF MODEL E MODEL

P.E 0.019*** 0.010*** -0.072*** -1.202*** (S.D.) (0.003) (0.001) (0.014) (0.193)

Intercept -35.017 -18.900 142.393 2439.294 (S.D.) (5.662) (2.941) (27.384) (383.205)

R2 .647 .669 .542 .628 N 25 25 25 25 Pr < |t| <.0001 <.0001 <.0001 <.0001 * significant at .10 level ** significant at .05 level ***significant at .01 level

38 The COV (coefficient of variation) model is excluded as the measure cancels out the concomitant increases in variation in the mean and standard deviation over the time series.

129

response on gay rights in the GSS has moved away from the “always wrong” consensus of the

1970’s and early 1980’s to the contemporary divided opinion on gays today. Again, while this trend is towards the center of the distribution, it is evidence of increasing conflict on gay rights.

The ‘anti’ side of the gay rights debate has lost supporters as a percentage of the total population (mass) and the ‘pro’ side of the gay rights debate has gained: drawing the mean on gay rights attitude towards the middle. Both models explain over sixty percent of the variation in the standard deviation and the mean respectively.

The negative coefficient (-0.072) in the Kurtosis model indicates that the distribution of attitudes on gay rights has become more bi-modal over the past three decades. As can be seen in Table 5.5, the distribution goes negative in 1993 and is negative in every survey year after to the present. At the beginning of the series, the kurtosis for the GSS gay rights attitude measure is at a 0.809. By the end of the series, it stands at -1.696. Recall that kurtosis is centered at zero for the normal distribution (mesokurtic). So the distribution on gay rights attitudes flips from leptokurtic to platykurtic over the course of the past three decades. The model shows there has been a significant increase in bimodality in the distribution of opinion on gay rights for the mass public, explaining over half of the variance in kurtosis over the time series relative to the mean kurtosis value (.542). The ECD model shows the declining consensus against gay rights over the time period, with the AW category losing, on average, 1.202 percent of its mass every year of the time-series relative to the NW category. Over half of the variance in the Kurtosis model and sixty-three percent of the variance in the ECD model is explained by the progress of time. Both suggest increasing bimodality on gay rights. Overall, Table 5.6 shows powerful evidence of increasing polarization on gay rights in the American public and is completely consistent with the culture wars thesis. There is increasing dispersion on gay rights (standard deviation model),

130

the extreme categories have become more balanced as opinion has shifted away from extreme

opposition to gay rights while extreme support has increased (ECD model), there is increasing

bimodality as indicated by the negative coefficient from the kurtosis model, and the mean has

shifted away from the anti-gay side of the opinion scale, as we would expect in a shift from

consensus located near one pole of the scale towards conflict on the issue (mean model).

ANALYSIS – PARTISAN POLARIZATION TRENDS IN PUBLIC OPINION ON GAY RIGHTS

While we have established the loss of consensus on most gay rights issues and on gay rights overall that has occurred over the last three decades, as I noted in Chapter 3, it isn’t mere issue polarization that makes for a culture war. In order for a polarized issue to become politically relevant, it has to be relevant to political competition. The political parties are a prime mover in American political competition. Parties define and shape political conflict, educate and motivate their supporters on issues, and seek to gain supporters by taking advantage of issues that divide the American public. An issue that is polarized and where the parties align along the cleavage of opinion on that issue is the essence of what makes for a cultural war. In the

American political system, the threshold for an issue to gain salience among partisans and make its way into the debate over public policy is relatively large given the bias in the electoral system against minority factions (see Durverger’s Law). The GSS item on homosexual rights provides an opportunity to assess the trend in partisan opinion on gay rights and compare the trends in the differences both between the parties and within the parties.

Issues that divide the public and gain political salience present an opportunity for parties to gain supporters by appealing to one side or the other on the issue. Issues on which there is consensus, or which do not lend themselves to policy alternatives, or cross-cut well established partisan cleavages do not lend themselves to partisan and political competition (absent a realignment on the cross-cutting issue). Gay rights falls within the former classification. As a

131

consensus issue in the 1970’s, neither of the major parties had it as a major plank in their party platforms, it was not an issue upon which their candidates campaigned, and consequently the mass public did not identify the parties with a position on gay rights nor did they align themselves with the parties, in whole or in part, based on their attitudes towards gay rights and their perceptions of the parties vis-à-vis gay rights.

That began to change midway through the 1980’s. As the consensus on gay rights died and the emerging social conflict on gay rights emerged, the prospect for partisan competition, partisan alignment on either side of the gay rights debate, and the institution of policy alternatives on gay rights became a possibility. If this story on gay rights is correct, then we would expect that the mass public would increasingly associate the parties with a position on gay rights, and that partisan identifiers would, overtime, increasingly identify with the party’s position on gay rights (either through conversion of partisan identifiers to the pro or anti positions on gay rights or the adoption of a party id by supporters and opponents of gay rights).

Thus we would expect to see the party identifiers diverge on their attitudes on gay rights concomitant with the emerging cultural battle over homosexuals and their place in society. This is exactly what we see over this time period, as illustrated in Figures 5.7 - Figures 5.12. The broad classification of party identifiers reflects the emergent partisan polarization on gay rights and this polarization is evident for each of the party identification categories of respondents.

132

FIGURE 5.6: GSS – HOMOSEXUAL RELATIONS ARE WRONG 1973-2006, BY PARTY IDENTIFICATION 100.00%

95.00%

90.00%

85.00%

80.00%

75.00% REPUBLICAN INDEPENDENT 70.00% DEMOCRAT

65.00%

60.00%

55.00%

50.00%

FIGURE 5.7: GSS – HOMOSEXUAL RELATIONS ARE WRONG 1973-2006, REPUBLICAN /DEMOCRAT DIFFERENCE

25.00%

20.00%

15.00%

R-D WRONG 10.00%

5.00%

0.00%

133

Figure 5.6 illustrates the change in average opinion on homosexual relations among

Republican, Democrat, and Independent identifiers in the time series.39 Note that in the beginning of the series in 1973, Republicans, Democrats and Independents are unified in their opinion on whether homosexual relations are wrong. Ninety percent of the respondents, irrespective of party identification, believed that there was something wrong with homosexual relations. The implications for a ‘culture war’ here are clear. In the 1970’s there wasn’t a

“culture” war…certainly not on the hot button social issue of gay rights. There was an obvious consensus against homosexuals. No partisan advantage to be had by taking an opposing position and supporting gay rights (indeed, doing so at the time could have rated as political suicide).

While some separation occurs over the course of the next decade, there was still relatively little difference (about 5%) between Republicans, Democrats, and Independents on the ‘wrongness’ of homosexual relations. In the early part of the 1990’s, just about the time that Hunter identified the existence of an emerging culture war, Republicans and Democrats begin to increasingly separate on gay rights. As can be seen in Figure 5.7, the difference between

Republicans and Democrats tops ten percent and steadily increased through the end of the time series (maxing out at a 20% difference).

This trend in increasing partisan differences is apparent in each of the categories of partisanship on the 7-point scale. As one might expect, the largest disparity exists between strong Republican identifiers and strong Democratic identifiers. In 1973, there was no difference between strong Republicans and strong Democrats on homosexual relations (Figure 5.8). By

2006, forty percent of strong Democrats saw nothing wrong with homosexual relations compared with just fifteen percent of strong Republicans. While the differences are not as

39 Party ID defined in to 3 categories with weak identifiers and independent leaners included with their respective parties. The middle category consists of those respondents who identified as independent with no leanings.

134

FIGURE 5.8: GSS – HOMOSEXUAL RELATIONS ARE NOT WRONG 1973-2006, STRONG REPUBLICANS AND DEMOCRATS

50.00%

45.00%

40.00%

35.00%

30.00%

25.00% Strong Republican

20.00% Strong Democrat

15.00%

10.00%

5.00%

0.00%

1973 1974 1975 1976 1977 1978 1980 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1993 1994 1996 1998 2000 2002 2004 2006

FIGURE 5.9: GSS – HOMOSEXUAL RELATIONS ARE NOT WRONG 1973-2006, WEAK REPUBLICANS AND DEMOCRATS

45.00%

40.00%

35.00%

30.00%

25.00%

W REP 20.00% W DEM 15.00%

10.00%

5.00%

0.00% YEAR 1973 1974 1975 1976 1977 1978 1980 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1993 1994 1996 1998 2000 2002 2004

135

FIGURE 5.10: GSS – HOMOSEXUAL RELATIONS ARE NOT WRONG 1973-2006, INDEPENDENT LEANERS

60.00%

50.00%

40.00%

30.00% IND L REP IND L DEM 20.00%

10.00%

0.00% 1973 1974 1975 1976 1977 1978 1980 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1993 1994 1996 1998 2000 2002 2004 2006

FIGURE 5.11: GSS – HOMOSEXUAL RELATIONS ARE NOT WRONG 1973-2006, INDEPENDENTS

40.00%

35.00%

30.00%

25.00%

20.00% NOT WRONG 15.00%

10.00%

5.00%

0.00%

136

marked among weak identifiers (Figure 5.9) or independent leaners (Figure 5.10), there is still an apparent and distinct difference between Republicans and Democrats on gay rights completely consistent with the parties aligning in opposition on attitudes towards homosexuals. The alignment isn’t perfect, however. All three of the parties evidence the overall trend towards increasing tolerance of homosexuals. However, note that Republicans who think there is something wrong with homosexual relations has only declined about 15 percent over the last three decades. The increase in support for gays among Democrats is twice that (Figure 5.6). The behavior of independents is interesting in the later part of the time-series. In the late 1990’s independents tracked closer to the Democrats in their tolerance of homosexuals, but by 2006 they had ticked back to a position just below that of the Republicans. While one can only speculate as to the cause, the prominence of gay marriage as the prime gay rights issue in this decade and the legal interventionism on gay marriage may have produced a backlash against gay rights among Independents.

Finally, I examine the evidence on public attitudes on gay rights as it relates to the partisan differences on the polarization measures. Table 5.7 reports the means and standard deviations for the GSS homosexual rights item for the time series. Again, the trend away from the consensus against homosexual relations is apparent for both parties, but more marked among Democrats who register a half-point closer to the ”nothing wrong at all” category than

Republicans. I estimate six different partisan and party difference models on the trend in attitudes towards gay rights (Table 5.8). Intercepts, parameter estimates, and goodness-of-fit statistics are reported for each model. All of the models are statistically significant at the .10 level and all but one of the models is significant at the .01 level. I estimate trend models for the mean response on and the standard deviation of the GSS homosexual relations item for both

Republicans and Democrats. I further estimate a linear model on the trend in the difference

137

TABLE 5.7: PARTY MEANS & STANDARD DEVIATION FOR PUBLIC ATTITUDES ON HOMOSEXUAL RELATIONS 1971-2007 YEAR DEM MEAN DEM S.D. REP MEAN REP S.D. 1973 1.584 1.053 1.512 1.004 1974 1.669 1.142 1.510 0.987 1975 1.688 1.153 1.579 1.049 1976 1.706 1.164 1.648 1.111 1977 1.658 1.142 1.608 1.064 1978 1.654 1.142 1.542 1.018 1980 1.650 1.141 1.475 0.972 1982 1.579 1.079 1.548 1.051 1983 1.629 1.115 1.538 1.038 1984 1.678 1.151 1.528 1.025 1985 1.593 1.100 1.553 1.067 1986 1.572 1.086 1.497 1.013 1987 1.550 1.072 1.440 0.958 1988 1.625 1.129 1.473 0.991 1989 1.774 1.216 1.473 1.014 1990 1.685 1.155 1.440 0.973 1991 1.737 1.231 1.495 1.043 1993 2.054 1.345 1.642 1.136 1994 2.009 1.341 1.698 1.184 1996 2.152 1.383 1.818 1.262 1998 2.257 1.298 1.792 1.223 2000 2.189 1.371 1.807 1.231 2002 2.385 1.410 1.880 1.294 2004 2.401 1.422 1.843 1.249 2006 2.419 1.415 1.919 1.293

between the average Republican and Democrat mean response and standard deviation for the time series. The models for the means and standard deviations for both Republicans and

Democrats illustrate the loss of consensus and move towards conflict on gay rights issues. The positive coefficients on the mean trend indicate that both Republicans and Democrats have moved away from the previous consensus against homosexuals. The standard deviation models indicate that for both Republicans and Democrats, there is greater diversity on gay rights today than there was three decades ago.

138

TABLE 5.8: PARTISAN TRENDS IN PUBLIC ATTITUDES ON HOMOSEXUAL RELATIONS 1971-2007 Trend: MODEL: Conflict or Intercept Parameter Estimate 2 GR = B0 + B1(YEAR) + E Consensus (S.E.) (S.E.) R N

Mean Democrats CONF -50.167 0.026 *** .741 25 on Gay Relations (6.420) (0.003)

Standard Deviation CONF -20.101 0.011 *** .723 25 Dems on G. R. (2.759) (0.001)

Mean Republicans CONF -20.332 0.011 *** .519 25 on Gay Relations (4.403) (0.002)

Standard Deviation CONF -16.338 0.008 *** .601 25 Reps on G. R. (2.962) (0.001)

REP – DEM Mean CONF -29.835 0.015 *** .786 25 Difference on G.R. (3.271) (0.002)

REP – DEM S.D. CONF -3.763 0.002 * 0.132 25 Difference on G.R. (2.075) (0.001)

* significant at .10 level ** significant at .05 level ***significant at .01 level

However, note the distinct difference in the goodness-of-fit measure. The R2 for

Democrats on the mean trend model indicates that twenty percent more of the variance in the mean on gay rights is explained by the passage of time in comparison to the Republican mean model. The model explains seventy-five percent of the variance for the Democrats but only fifty percent for Republicans. There is a similar distinction between the Republican and Democrat

S.D. models, with the Democratic model explaining more of the variance than the Republican model. The answer, I believe, is found in the difference models. The positive coefficient in the R-

D mean difference model (0.015) shows that there is a significant increasing difference between

139

Republicans and Democrats on gay rights (R2 = .786). While the difference between S.D. is

significant, with Democrats experiencing greater intra-party dispersion than the Republicans, it

is a relatively small difference (.002) and a concomitant poorly fit model (R2 = .132).

In the 1970’s and the early years of the 1980’s a clear consensus against homosexuals and strong opposition to gay rights existed. As a consequence, the parties largely ignored the issue. The mass public did not differentiate the parties on gay rights, partisan elites had no incentive to stake out positions on gay rights or propose policy alternatives related to the issue, and gay rights was well off the public agenda for all but the small enclaves of gay rights activists.

This consensus died sometime in the middle of the 1980’s. Support for gay rights emerged and the same time opposition to gay rights declined. What emerged was an issue ripe for political and partisan conflict. The growing mass of citizens in favor of gay rights encouraged the

Democratic Party to woo those citizens (either to convert them to the party or as a consequence of the increasing number of Democratic party identifiers who supported gay rights) by offering policy alternatives such as the protection of homosexuals in the workplace and permitting gays to openly serve in the military. The parties became to be identified with specific positions on gay rights (Republicans opposed; Democrats in support) and gay rights supporters and opponents became an identifiable cleavage between the parties (and thus distinct groups that would need to be satisfied by policies in order to generate money and electoral support). This trend is apparent and significant in all of the partisan measures on gay rights opinion over the past three decades. The parties have clearly diverged on gay rights. In three short decades the Republicans and Democrats in the mass electorate went from having nearly identical viewpoints on gay rights to distinct and divergent views on this social issue. It is thus no surprise that gay rights are a significant player in the partisan battles in the culture war from the 1990’s to the present. The

140

FIGURE 5.12: WHICH PARTY IS CLOSEST TO RESPONDENT ON GAY MARRIAGE, 200540 100 90 80 70 60 50 SUPPORT 40 30 20 10 0 Democrats Republicans

American public is divided on gay rights, and hence are divided in their support for either party on gay rights issues (Figure 5.12).

CONCLUSION

I believe the evidence presented in this chapter paints a very clear picture on political polarization and the culture wars. In the early 1970’s and through the middle part of the 1980’s there existed a strong consensus in the American public that was barely tolerable of and expressly hostile towards homosexuals. In the later part of the 1980’s public attitudes began to undergo a radical transformation in attitudes on gay rights. An increasing number of Americans began to see “nothing wrong” with homosexual relations and were increasingly tolerant of and comfortable with gays. The result of this formation of a second and oppositional camp on gay rights was part and parcel of the emerging culture war in the early part of the 1990’s. Today, whether it be the spectacle of George Takei (Mr. Sulu) getting married to his long-time partner in San Francisco or the heated legal battles over Defense of Marriage acts versus state court-

40 CBS/New York Times Poll. February, 2005. N = 1111. Question: “Regardless of how you usually vote, which party comes closer to sharing your view on the legal recognition of gay couples, the Democratic Party or the Republican Party?”

141

imposed gay marriage, the battle lines are clearly drawn. There is social conflict on gay rights.

Here I have shown that a host of gay rights issues have moved from consensus to conflict (with but a few countervailing examples of gay rights attitudes moving from conflict to consensus).

Using two different datasets of public attitudes on gay rights, I have shown that there is an overall trend from consensus to conflict on attitudes towards homosexuals, homosexuality, and homosexual rights. Furthermore, I have shown that the parties have reacted to and aligned along this cultural divide. As public opinion on gay rights diversified, it created an opportunity for the parties to use gay rights to compete for votes. As a consequence the Democrats have aligned much more strongly with the pro-gay rights faction with Republicans sticking with the anti-gay rights faction. The parties have recognized, reacted to, and mobilized on the cultural battle over gay rights. And there is little evidence that this war will end any time in the near future.

This chapter has presented a clear case of increasing political polarization and conflict on one of the hot-button social issues of the day and one of the major fronts in the culture wars.

It further serves to demonstrate the application of specific, empirical measures of polarization to test whether opinion on an issue has become more dispersed, more bimodal, and more extreme (e.g. more polarized). However, this chapter on gay rights, while assessing polarization in an important social issue and a key aspect of the culture wars, is narrow in scope. There are other social issues relevant to partisan and political competition and conflict. Furthermore, there are non-social issue dimensions such as economic policy, government spending and regulation, and foreign policy in which polarization (or depolarization) can occur. In the next chapter I expand upon the analysis presented here by employing the measures of polarization to look not just at the range of other social issues that have sparked the culture war, but also at the other major dimensions of political conflict in the United States.

142

CHAPTER 6: MULTIDIMENSIONAL POLARIZED POLITICS: SOCIAL ISSUES, THE ECONOMY, & FOREIGN POLICY

“And it’s not surprising then [that] they get bitter, they cling to guns or religion or antipathy to people who aren’t like them or anti-immigrant sentiment or anti-trade sentiment as a way to explain their frustrations.” – Barak Obama, April 2008

ISSUE POLARIZATION - ‘YOU CAN HAVE MY GUN WHEN YOU PRY IT FROM MY COLD, DEAD HANDS’

Barak Obama made the quoted observation above to a group of California contributors in trying to explain why small towns in Pennsylvania care about social issues. Like Thomas Frank, he seems to think that there is a significant segment of the American electorate motivated by the culture war issues that Hunter originally identified as the turf over which secularists and traditionalists would battle. Putting aside Obama’s socio-psychological explanation for these issue positions among the Pennsylvanian rural electorate, clearly these social issues continue to resonate on the political scene. But is Hunter’s thesis—that these issues will increasingly dominate our politics, that voters will increasingly look to issue positions in the social and religious issue dimensions, and that this conflict will be increasingly strident and bitter— reflective in what we know about these issue dimensions and what the electorate thinks about them? Layman and Carsey find polarization at the level of the mass electorate across three major domestic policy agendas: social welfare, racial, and cultural issues which they term

“conflict extension” (Layman and Carsey 2002). Miller and Hoffman, similarly, find increasing salience on many social and moral issues during the 1970’s and 1980’s, resulting in a redefinition of what it means to be conservative and liberal. Thus orthodox religious denominations have increasingly categorized themselves as ‘conservative’ while progressive religious denominations have increasingly identified as ‘liberal’ (Miller and Wattemberg 1984; Miller and Hoffmann

1999). Lindaman and Haider-Markel find mixed evidence that mass partisans have polarized

143

along with elite partisans. They examine four issue areas, but only on the environment and gun control did they find significant increases in polarization at the mass and elite level (Lindaman and Haider-Markel 2002). Fiorina, as noted earlier, finds liberalizing trends on gay rights and that the electorate is increasingly moderate and centrist on social issues (Fiorina, Abrams, and

Pope 2004).

In Chapter 4 I established that significant political polarization has occurred at the level of the mass electorate on a key culture war issue: homosexual rights. But that is just one point on the continuum of political issues. There are other flashpoints of political competition and conflict in the culture wars that span the gamut of social issues: abortion, school prayer, religious monuments and expressions in the Public Square, crime and social order, etc.

Furthermore, beyond the culture war thesis, political polarization in other issue classes—such as the scope of government, economic & financial , defense, foreign policy, and regulatory issues— is just as important, if not more so, than that which occurs in the cultural arena. The attitudes of the American public on issues and policy is reflected in the aggregate opinions on the issues themselves, their attitudes towards groups relevant to that particular political dimension, and the parties which stake out positions and mobilize in response to opinions on these political dimensions.

Here I examine a wide variety of political issues and attitudes of the American public in the social, economic, and defense issue dimensions. The goal is to assess the extent to which the views of Americans have moved to the extremes versus moved to the center on multiple political issues in distinctly different dimensions, to assess the degree to which the views on these issues or groups have become dispersed or consolidated, and how those opinions interact with the partisan identification of the mass public. In short, I will look at the breadth of political issue dimensions in American politics and assess whether polarization, depolarization, or the

144

status quo on opinion distribution has remained relatively stable over the past three decades.

Using the measures of polarization and the methods assessing polarization over time, I examine

a variety of measures of mass public opinion across the assortment of political issues to

empirically test for trends in polarization and depolarization.

Data

The data for this analysis is culled from the American National Election Study (ANES)

cumulative file.41 I use the ANES studies from 1970-2004.42 The creation of the data set for analysis of polarization trends for the mass public is a two-step process. In the first step, univariate statistics are generated on the substantive variables from the ANES cumulative file.

Specifically, the means and standard deviations for the variables were output. The second step involves creating a time-series data set with the means and standard deviations for the relevant

ANES variables treated as individual variables themselves in the new data set. For example, let’s consider the thermometer score for the military from the ANES. The first step involves generating the means and standard deviations for the military thermometer score for each of the study years. In the second step, a data set is created where one variable is the mean military thermometer score for the study years. The other variable is the standard deviation for the military thermometer score for the study years. A time series data set was created that contains mean and standard deviation variables for all of the variables relevant to the polarization analysis.

41 The Cumulative Data File consists of variables derived from the 1948-2004 series of biennial ("time- series") SRC/CPS National Election Studies. The American National Election Studies / Time Series Studies are collected before and after presidential (pre and post surveys) elections. The off-year elections typically only have a post-election study. The ANES Cumulative Data File is a merged data set of all the time series studies from 1948-2004. The pooled data includes variables which appear in three or more studies and consists of 44,715 cases. 42 The data is sub-setted by year to include only studies from 1970-2004 as the previous data sets had few to none of the relevant substantive variables which are necessary for the polarization analysis. Furthermore, 1970-2004 covers the relevant time period to examine the culture wars thesis.

145

Variables

The substantive variables included in the analysis and the years for which data was collected on those variables are listed in Appendix G. Rather than detailing every variable included in the analysis, I will discuss them in terms of their types and classifications.

Ordinal Scale Variables

The ANES time-series includes a number of issue-oriented ordinals scales with which respondents can place themselves, the parties, and candidates in an issue space. While a number of interesting issues have come and gone in the ANES time-series, there are a number of issues that span the breadth of the time series that the ANES has collected data on consistently. Every issue variable rated on an ordinal scale in this analysis has at minimum a respondent self-placement on the issue.

The issues include respondent attitudes on government aid to Blacks, women’s role in society, government spending, defense spending, jobs, and abortion. The issue placement that is collected in the most study years for the respondents is the ideological placement variable

(asking respondents to place themselves and the parties and candidates on a 7-point ideological scale ranging from strongly conservative to strongly liberal). The variable for which the least number of years is included is the defense spending item, which the ANES did not begin collecting until 1980. The questions in the survey used to build the issue scales are consistent over the time period used in the analysis with the exception of the abortion item. 43

43 The abortion question wording was changed in 1980. Previously respondents were asked when abortion should be allowed with the following response set: 1. Abortion should never be permitted. 2. Abortion should be permitted only if the life and health of the woman is in danger. 3. Abortion should be permitted if, due to personal reasons, the woman would have difficulty in caring for the child. 4. Abortion should never be forbidden, since one should not require a woman to have a child she doesn't want. The question was changed in 1980 to better represent the issue as a question of law. The new question asked when abortions should be allowed by law. The response set for the new question (used from 1980 forward) was: 1. By law, abortion should never be permitted.

146

Thermometer Variables

The feeling thermometer variables ask respondents to rate groups in society on a scale

of 0-100, with 0 – 49 indicating the respondent does not feel favorably (cold) towards that group

and 51-100 indicating that they do feel favorable (warm) towards that group. Respondents are

instructed to give a 50 if they have a neutral disposition towards the group (neither warm nor

cold). The feeling thermometers relevant to political polarization assess the attitudes of

respondents towards groups associated with the relevant issue dimensions. Attitudes towards

groups such as evangelicals, Catholics, Protestants, Christian fundamentalists, gays and lesbians,

and blacks may provide insights into American social conflict. Trends in the views on the poor,

big business, welfare and labor groups can reveal polarization or depolarization on economic

issues. Opinion on the military can shed light on the defense issue dimension. And opinions on

the political parties and the federal government can provide perspective on partisan

polarization and shifts in the attitude towards the size and scope of government.

Open-Ended Variables

The final category of variables included in the polarization analysis is the open-ended

questions on party likes and dislikes as well as the “most important national problem”

questions.44 For both types of questions, the cumulative file contains collapsed categories that reflect the general types of open-ended answers to the open-ended questions. While these are useful in assessing polarization, the categories were not created with looking at the social,

2. The law should permit abortion only in case of rape, incest, or when the woman's life is in danger. 3. The law should permit abortion for reasons other than rape, incest, or danger to the woman's life, but only after the need for the abortion has been clearly established. 4. By law, a woman should always be able to obtain an abortion as a matter of personal choice. 44 See Appendix H for open-ended collapsed category codes.

147

economic, defense issue dimensions specifically. For that purpose, the fully coded Republican

and Democrat likes and dislikes are included in the analysis.45

SOCIAL CONFLICT AND MULTIVARIATE POLITICAL POLARIZATION

In order to examine social conflict in the context of other kinds of political polarization

(i.e. other dimensions of politics), I employ a variety of measures developed from the ANES to cover the broad classifications of American political issues: social issues, economic issues (i.e.

Jobs), foreign policy and defense issues, and the scope of government (i.e. spending). While the primary focus remains on the social issue dimension, mass attitudes on economics, foreign policy, and on government influence the same actors and institutions as those affected by the culture wars. The culture wars do not exist in a vacuum.

Furthermore, political polarization is not restricted to the culture wars. Polarization along other issue dimensions can be just as destructive as cultural/social polarization. While it is not a specific subject of this analysis, the polarization and partisan conflict in the wake of 9-11 over the wars in Afghanistan and Iraq have had profound impacts on policy, the democratically responsive political actors and institutions, and on the competition between parties. Thus it is important to avoid myopic focus on just one aspect of the changing landscape of the past three decades in terms of the public agenda, public attitudes, and policy. While polarization on one dimension does not necessarily imply the absence of polarization on others, the degree to which one set of issues becomes increasingly the focus of party and political conflict can change the nature of that competition as well as crowd out the other issues.

45 Republican and Democrat likes and dislikes are coded consistently in the ANES time-series from 1972- 2004. The codes for the open-ended responses are recorded in the master code appendix: “PARTY- CANDIDATE 1972 AND LATER.”

148

POLITICAL POLARIZATION AND ATTITUDES TOWARDS GROUPS

To get a gauge of the changing public attitudes on political issues, I examine the yearly trends in public attitudes towards groups strongly associated with the issue dimensions. There is a clear conceptual connection between certain issues and specific political groups. Religious groups, gays and lesbians, and feminists are associated with the social issues that primarily concern them. The Catholic Church, for example, has taken strong public stands against the legalization of abortion and same-sex marriage. Labor unions and businesses are associated with the economic issues and policies they compete over. Finally, attitudes towards the military provide a rough proxy for the disposition of the public on foreign policy and military spending.

Table 6.1 reports regressions of changes in the average mean response on feeling thermometers by ANES respondents towards the relevant political groups. It should be noted from the outset that examining attitudes on issues using groups is at least one step removed from the subject of interest. There is nothing necessary about one’s attitude towards the military that speaks to a particular position on foreign policy or whether that dimension is relevant in an individual’s partisan affiliations, policy preferences, or electoral decisions. Still, if you oppose gay marriage, gay adoption, gays in the military, etc. there is a likely correlation between those issue position and your attitudes towards gays and lesbians as a group. Another caveat: some of the items were collected over a small number of survey years. While I report regressions where the normality assumption is suspect, the regressions provide a data reductive method for examining the trends. Where the n is small, caution is necessary in taking significance and model-fit results at face value. However, regression is a robust method of analysis and clues can be found even where data is scarce.

149

TABLE 6.1: TREND REGRESSIONS OF PUBLIC ATTITUDES ON ISSUE DIMENSION-RELATED GROUPS (MEANS) Trend: MODEL: Polarization Intercept Parameter Estimate = + ( ) + Y/N (S.E.) (S.E.) R2 N ഥ PARTYࡲࢀࢄ ࡮૙ ࡮૚ ࢟ࢋࢇ࢘ ࢋ Major Party Candidates Y 419.290 -0.181 *** .574 13 (93.973) (0.047)

House Candidates Y 286.637 -0.116 *** .467 14 (71.149) (0.036)

ECONOMICS Big Business N -190.449 0.122 * .210 12 (149.138) (0.055)

Poor Y 352.763 -0.141 ** .292 15 (121.370) (0.061)

Environmentalists Y 1120.538 -0.528 ** .512 9 (388.390) (0.195)

DEFENSE Military ------211.855 0.142 .226 12 (165.195) (0.083)

SOCIAL Blacks N -314.818 0.191 *** .478 17 (102.548) (0.052)

Christian ------165.931 0.110 .076 7 Fundamentalists (341.349) (0.171)

Feminists N -244.030 0.149 .547 5 (156.719) (0.078)

Gay & Lesbians Y - 1.094 *** .907 9 2142.687 (0.132) (263.643)

Whites Y 481.206 -0.205 *** .630 15 (86.674) (0.043)

* significant at .10 level ** significant at .05 level ***significant at .01 level

150

TABLE 6.2: TREND REGRESSIONS OF PUBLIC ATTITUDES ON ISSUE DIMENSION-RELATED GROUPS (STANDARD DEVIATION) Trend: MODEL: Polarization Intercept Parameter Estimate = + ( ) + Y/N (S.E.) (S.E.) R2 N ࡲࢀPARTY࢙ࢊ ࡮૙ ࡮૚ ࢟ࢋࢇ࢘ ࢋ Democratic Party Y -215.811 0.119 *** .560 13 (63.725) (0.032)

Major Party Candidates Y 109.881 0.048 * .395 13 (44.732) (0.022)

ECONOMICS Big Business N 149.648 -0.065 * .292 12 (63.529) (0.055)

Poor Y -68.661 0.043 ** .324 15 (34.759) (0.017)

Environmentalists ----- 111.226 -0.044 .034 9 (192.057) (0.096)

DEFENSE Military N 231.926 -0.106 . 723 12 (41.253) (0.020)

SOCIAL Blacks N 136.268 -0.058 *** .222 17 (56.024) (0.028)

Feminists N 130.462 -0.054 *** . 797 5 (31.706) (0.016)

Gay & Lesbians ----- 22.792 0.002 .001 9 (128.028) (0.064)

Catholics Y -85.045 0.052 * .372 8 (55.675) (0.027)

Protestants Y 449.412 0.050 *** .727 4 (257.565) (0.022)

Whites ------17.141 0.017 .273 15 (40.069) (0.020)

* significant at .10 level ** significant at .05 level ***significant at .01 level

151

Partisan Groups

Tables 6.1 and 6.2 report the mean and standard deviation trend models for the issue-

related group thermometer scores. The evidence in Table 6.1 indicates that there is increasing

partisan polarization at the mass public level. The significant trends in the means of the feeling

thermometer responses on House candidates as well as the major party candidates is consistent

with increasing dispersion in attitudes towards the two parties. This could reflect increasingly

negative feelings by partisans of one party towards the other. This supposition is supported by

the results from the partisan models in Table 6.2. Both attitudes towards Democratic Party and

major party candidates have evidenced increasing dispersion since the 1970’s. Attitudes about

the Democratic Party have become more disperse at a rate of .119 on the thermometer scale for

each year of the time series, and this model explains over half of the variation in thermometer

scores for the Democratic Party (.560). The increase in the standard deviation for the major

party feeling thermometer apparent in Table 6.2 underscores the evidence from Table 6.1:

increasing partisan polarization.

Economic Groups

One merging theme of this analysis is the lack of a trend in polarization on economic

issues (attitudes towards big business have slightly improved). This isn’t to say that economic

issues are not a major source of political conflict, but rather that that conflict has not become

more polarized over the past three decades. Attitudes towards big business were already near

the 50/50 cut point in the 1970’s and not much has changed in that regard in the 21st century.

One exception to that general finding, however, is the significant negative trend in attitudes towards the poor in the mass public. However, while support for the poor has declined over the time-series, attitudes towards the poor are still relatively positive (65.51 average feeling thermometer, 2002).

152

Defense / Foreign Policy Group – the Military

There is no apparent trend in attitudes toward the military. The average feeling thermometer rating of the military is near 70 for the entire time-series with only small oscillations. If there is growing polarization on foreign policy, it is not reflected in attitudes towards the United States armed forces. Most Americans have had, on average, very positive feelings towards the military irrespective of whether there was an ongoing conflict or the relative popularity of that conflict in the American public. While the Iraq war became extremely unpopular in the later half of the 2000’s, this seems to have had little to no impact on attitudes towards the military.

Social & Religious Groups

The results from Table 6.1 do show that attitudes towards religious groups have changed significantly. While there is no apparent trend in attitudes towards Christian fundamentalists, the first time data on this feeling thermometer was collected was 1988, just short of the burgeoning culture war. The average FT attitude for Christian fundamentalists at that time was 51.34, near the absolute point of group conflict. Attitudes since have changed little, oscillating between 51 and 53 for the survey years in which the item is collected. This is not so for evangelicals and protestants. While neither coefficient is significant (likely due to the extremely small N’s), the trends for both are negative, indicating decreasing support in the public. Protestants rate a good 10 points higher than evangelicals in the time series, most likely due to the greater number of citizens who identify as protestant relative to evangelicals.

However the consistent trend suggests two things: religiosity may be on the decline (and thus igniting the culture war) and the increasing association of the religious with the political. The greater dispersion in attitudes towards Catholics apparent in Table 6.2 could be a reflection of decreasing religiosity, a reaction to the pedophile priest scandal, or a reaction to the Church’s

153

strong position and activism on abortion. While this is just speculation, the pattern of declining

positive feelings towards religious groups suggests conflict.

Though the trend on attitudes towards gays & lesbians is positive, as you will recall from

Chapter 5, this is not necessarily evidence of the absence of polarization. The first time the item

on attitudes towards gays and lesbians was asked in the ANES was 1984, with an average feeling

thermometer score of 29.9. While attitudes towards gays and lesbians have become more

tolerant (gaining over a point on the thermometer score each survey year – largest trend

coefficient), the 48.37 mean attitude in 2004 illustrates the ‘great divide’ that now exists on gay

rights in the wake of the culture war.

The declining dispersion of attitudes towards Blacks and feminists suggests that the

racial and gender conflicts that reached their peak in the 1960’s and the 1970’s are less an

object of political conflict today. The attitudes towards feminists, however, remain well within

the range of political conflict (56.14 FT, 2004). Attitudes towards blacks were well on the

positive side of the distribution and have only moved closer towards consensus over the last

thirty years, gaining 10 points on the feeling thermometer scale for the time series.

TRENDS IN MASS PUBLIC OPINION ON ISSUE DIMENSIONS

The simple univariate statistics for the time-series data on the closed-ended issue self- placement items are reported in Table 6.3. The three primary measures of polarization employed in this analysis are the means, standard deviations, and kurtosis scores for each of the self-placements on the ideology scale (liberal-conservative), the jobs and guaranteed standard of living scale (rating the degree to which respondents wish government to provide job assistance versus leaving individuals to make their way in the job market on their own), the defense spending scale (whether the federal

154

TABLE 6.3 SIMPLE STATISTICS ON MEANS, S.D.’S & KURTOSIS FOR ISSUE SELF-PLACEMENTS FOR TIME SERIES VARIABLE N MEAN STAND DEV. MIN MAX MEANS Ideology 19 4.250 0.089 4.130 4.452 Jobs 19 4.321 0.188 3.962 4.690 Defense 14 4.074 0.382 3.497 4.633 Spending 14 4.245 0.297 3.714 4.659 Aid to Blacks 20 4.467 0.204 4.081 4.822 Women 19 2.546 0.517 1.833 3.511 Abortion 19 2.768 0.136 2.497 2.971 STANDARD DEVIATIONS Ideology 19 1.387 0.074 1.265 1.542 Jobs 19 1.862 0.079 1.737 2.014 Defense 15 1.525 0.089 1.411 1.720 Spending 14 1.613 0.049 1.514 1.700 Aid to Blacks 20 1.804 0.136 1.594 2.027 Women 19 1.793 0.243 1.467 2.273 Abortion 19 1.073 0.050 0.981 1.170 KURTOSIS Ideology 19 -0.475 0.182 -0.911 -0.240 Jobs 19 -0.872 0.159 -1.106 -0.544 Defense 15 -0.285 0.223 -0.645 0.107 Spending 14 -0.516 0.094 -0.673 -0.340 Aid to Blacks 20 -0.730 0.233 -1.096 -0.370 Women 19 0.309 1.316 -1.322 2.796 Abortion 19 -1.316 0.138 -1.659 -1.077 government should increase or decrease it), the aid to blacks scale (rating how much the government should help minority groups), the government spending scale (rating whether government should provide more or less services and spending), the abortion scale (reporting the respondent’s position on abortion), and the women’s role in society scale (rating whether respondent believes women should have an equal role with men or whether women’s place is in the home). All of the issue scales are 7-point scales except for the Abortion scale (4-point scale).

Table 6.3 is organized into three tiers reporting the average, standard deviation, and min/max values for the mean, standard deviation, and kurtosis for the responses on the issue

155

scales for the period of 1970 to 2008. So, for example, the mean reported in Table 6.4 for

Ideology (4.3) is the average mean ideology for respondents to the ANES from 1972 to 2008.

Not all the scales where collected for all of the survey years during this period. Estimates

of the mean, standard deviation, and kurtosis for an issue were interpolated where there was an

immediately proximate survey year before and after the year where the issue is to be

interpolated in which the issue scale was included. If this condition was not met, the value for

that issue scale was set to missing. I interpolate the data points for the ANES closed-ended

analyses using the two most proximate data points (survey year prior, survey year after) and the

simple linear interpolation method.46 There were relatively few points that needed to be interpolated for most of the survey years.47 The exceptions were 2002 and 2006. In 2002, only

the ideology scale was an included item for the ANES. No issue scale data was collected. In 2006,

there was no ANES Time Series Study. As such, all issue scale data for those years was

interpolated. Regression models were estimated with and without the interpolated data. No

substantial changes in statistical significance, model fitness, or the size of the coefficients were

found, and there was no change in the direction of the signs for the coefficients.

The statistics in Table 6.3 illustrate a state of conflict for all but one of the issue scales

included (Women’s role in society). For the 7-point issue scales the standard deviation is at or

above 1.5 points on the scale, indicating a relatively dispersed distribution of opinion on these

issues. Furthermore, the distributions on the issue scales are nearly all playtkurtic, suggesting a

relatively bimodal distribution of opinion on those issues (and thus ripe for political conflict).

The lone exception here is the women’s role scale, which has a leptokurtic distribution,

indicating single-peakedness (consensus). Excluding the women’s role scale, the least

46 Generally, linear interpolation takes two data points, defined here as (xa,ya) and (xb,yb). The ) interpolant is given by: + ( ) at the point (x,y). ) ሺ௬್ ି௬ೌ 47 Interpolated points (other than 2002, 2006): Abortion (1974), Women (1986), Jobs (1998), and Defense ݕൌݕ௔ ݔെݔ௔ ሺ௫್ ି௫ೌ (1998).

156

‘conflictual’ issue included in the analysis is defense spending. It has a kurtosis score--while still

on the bimodal side of the ‘normal distribution’ divide—that is closest to zero of the playtkurtic

or unimodal distributions. Furthermore, of the 7-point issue scales it has the smallest average

standard deviation over the time series. However, as I will demonstrate, this average result is

somewhat misleading and the story of polarization on defense spending and foreign policy is a

bit more complex (and hence murkier) than it would seem at first blush.

The most significant outlier of the issue scales is clearly that of women’s equal role in

society. It has the largest range between its minimum (1.833) and maximum (3.511) average

self-placements as well as the minimum and maximum standard deviation and kurtosis for the

time series. Indeed, women’s role in society and defense spending are the only issues which

cross the “normal distribution” zero boundary48 over the time series, indicating that a shift between a more bimodal to a more unimodal distribution occurred between 1970 and 2008. For at least some time in the covered period, the mass opinion on these two issues exhibited traits of consensus.

Jobs and Defense Spending Trend Models

In order to examine the trend in polarization on mass opinion in these issue dimensions and assess the direction and statistical significance of these trends, I regress the data on the mean, standard deviation, and kurtosis for the issues on survey year. The regressions for the economic, ideological, and defense / foreign policy dimensions are reported in Table 6.4. No trend in polarization is apparent for the defense spending and jobs issue dimensions. There are no significant trends in bimodality or dispersion in the opinion of the mass public for either of these issues. There has been a significant negative trend over the time series for the jobs scale, indicating the public has become less supportive of government involvement in providing jobs.

48 A normal distribution, or mesokurtic distribution, has a kurtosis of zero.

157

TABLE 6.4: POLARIZATION TREND MODELS FOR MULTIPLE ISSUE DIMENSIONS Trend: MODEL: Polarization Intercept Parameter Estimate = + ( ) + Y/N (S.E.) (S.E.) R2 N IDEOLOGY ࡵ ࡮૙ ࡮૚ ࢟ࢋࢇ࢘ ࢋ IDEOLOGY MEAN ---- 2.844 0.001 .008 19 (3.788) (0.001)

IDEOLOGY S.D. Y -8.643 0.005 *** .581 19 (2.066) (0.001)

IDEOLOGY KURTOSIS Y 21.362 -0.011 *** .460 19 (5.737) (0.002)

GOVERNMENT PHILOSOPHY SPENDING MEAN Y -40.680 0.023 *** .403 14 (15.778) (0.047)

SPENDING S.D. ---- 1.712 -0.001 .001 14 (3.391) (0.010)

SPENDING KURTOSIS Y -8.928 0.004 * .140 14 (5.017) (0.002)

ECONOMICS JOBS MEAN N 16.846 -0.006 * .142 19 (7.468) (0.003)

JOBS S.D. ---- 6.535 -0.002 .111 19 (3.203) (0.001)

JOBS KURTOSIS ---- -1.051 0.001 .000 19 (6.829) (0.030)

DEFENSE DEFENSE SPENDING N -56.964 0.031 *** .449 15 MEAN (19.510) (0.010)

DEFENSE SPENDING ---- 2.170 -0.001 .001 15 S.D. (5.516) (0.020)

DEFENSE SPENDING ---- -16.490 0.008 .106 15 KURTOSIS (13.063) (0.007)

* significant at .10 level ** significant at .05 level ***significant at .01 level

158

On defense spending there is a significant positive trend on defense spending, with the public becoming more supportive of increased spending on defense over the course of the time series.

Overall, there is no trend in the dispersion of opinion on defense spending. Bimodality decreased overall over the past three decades. This average trend, however, masks what is essentially a non-linear trend in bimodality for defense spending. I will take a closer look at polarization on defense spending and foreign policy later in chapter 7.

Ideological and Government Spending Trend Models

The issue dimensions in which polarization is apparent are the ideological and government spending dimensions. On both ideology and government spending, the distribution of public opinion has become increasingly bimodal since 1972 and 1982 respectively. The strong platykurtic trend in ideology indicates that the mass public has shifted away from the center, with either more conservatives and liberals (as opposed to centrists) or more conservatives and liberals identifying with the ideological extremes. The time-trend model of kurtosis explains just short of half of the variation in the bimodality of ideology (.460). Furthermore, there is an increase in the overall dispersion of the distribution, suggesting that the public is increasingly identifying with the extremes of the ideological distribution. This strong, positive trend shows that the average distances among the ideological self-identifiers have increased (0.005) over the time period. It suggests more citizens are identifying themselves as “very conservative” and/or

“very liberal” in comparison to the other categories or identifying themselves as “somewhat conservative” and/or “somewhat liberal” in comparison to the moderate category. While there is no statistically significant trend in dispersion for government spending, the kurtosis model suggests a statistically significant trend in bimodality on government spending. The government spending bimodality model, however, is not as good a fit as the ideological model (.140).

159

Social Issue Trends

The models for the three social issue items from the ANES with sufficient coverage for

the time period are reported in Table 6.10. Although the degree to which these items rate as

“issues” and issues relevant to the Culture Wars varies. Clearly abortion is both a social issue

and one of particular importance to the culture wars. Indeed, in many ways, abortion is the sine

qua non of the culture war

TABLE 6.5: POLARIZATION TREND MODELS FOR THE SOCIAL ISSUE DIMENSION Trend: Parameter MODEL: Polarizatio Intercept Estimate = + ( ) + n Y/N (S.E.) (S.E.) R2 N SOCIAL ࡵ ࡮૙ ࡮૚ ࢟ࢋࢇ࢘ ࢋ AID TO BLACKS MEAN N -17.122 0.011 *** .397 20 (6.267) (0.003)

AID TO BLACKS S.D. ---- 8.486 -0.003 .086 20 (5.133) (0.002)

AID TO BLACKS N -15.091 0.007 * .135 20 KURTOSIS (8.584) (0.004)

WOMEN MEAN N 92.791 -0.045 *** .974 19 (3.612) (0.001)

WOMEN S.D. N 43.808 -0.021 *** .958 19 (2.143) (0.001)

WOMEN KURTOSIS N -223.948 0.113 *** .929 19 (15.013) (0.008)

ABORTION MEAN Y -14.854 0.009 *** .537 19 (3.970) (0.002)

ABORTION S.D. Y -6.984 0.004 *** .833 19 (0.876) (0.001)

ABORTION KURTOSIS Y 20.092 -0.011 *** .769 19 (2.845) (0.001)

* significant at .10 level ** significant at .05 level ***significant at .01 level

160

debate. Aid to Blacks is a social issue as well, given the heated debate over the role of

government in helping Blacks overcome the dual legacies of slavery and segregation. The third

issue—women’s equal role in society—is related to social concerns, but the item itself does not

necessarily imply government intervention. The item rates public opinion on what should be the

condition of women in society, not on whether government should have a role in determining

that condition. As such, the “equal role” item is broader than the scope of the other two

included social issues. That said, it certainly captures public attitudes towards the role of women

in society and thus women in the workplace. Given the religious-based opposition to non-

traditional roles for women, its relevance to the culture wars is apparent.

The results in Table 6.5 demonstrate that not all social issues during the culture war

period have been characterized by greater conflict and polarization. On the Aid to Blacks issue

dimension, there is a statistically significant leptokurtic trend in bimodality, suggesting a move

towards consensus on aid to Blacks. The positive coefficient (0.011) in the mean model on aid to

blacks indicates that Americans have moved away from government solutions to racial

problems in society since the 1970’s.

The most remarkable trend among the three social issues is a depolarization trend in

opinion on the equality of women in society. Nearly all of the variance in the average opinion on

women’s role in society is explained by the linear progression of time (R2 = .974). The negative coefficient (-0.045) indicates that citizens were becoming increasingly supportive of equality between women and men in society. This trend is consistent with a growing consensus in favor of gender neutrality. We see a significant decline in the dispersion of opinion on women’s equality in the time series (-0.021) and a positive trend in kurtosis, indicating that the distribution of opinion on women’s equality is increasingly unimodal. The ANES social issue measures have not uniformly depolarized, however. The mean, standard deviation, and kurtosis

161

models for abortion all show significant trends over the past 30+ years and are indicative of increasing polarization. The positive coefficient (0.009) suggests that, on average, Americans have become more Pro-Choice over the course of the time-series. At the same time, the abortion issue has become increasingly divisive and polarized. There is a statistically significant increase in dispersion (0.004) and a negative trend in kurtosis (bimodality) for abortion (-0.011).

These models explain over seventy-five percent of the variation in opinion on abortion for the time series (SD R2 = .833; K R2 = .769). In brief, there is strong evidence that opinion on women’s equality has depolarized. There is some evidence of aid to Black’s depolarization, and there is strong evidence that public opinion on abortion is increasingly polarized.

CONCLUSION

I report a summary assessment of polarization and polarization trends in Table 6.6. The analysis of political polarization across multiple issue dimensions in the aggregate reveals a great deal of ‘static’ polarization in the sense that on most of these issue dimensions public opinion was ‘polarized’ to begin with and the changes over the course of the time series have not changed the fundamental conflict that exists within these issue dimensions. In addition, there has been a great deal of ‘dynamic’ polarization—or polarization trends—as can be seen in the case of aggregate ideology, aggregate opinion on abortion, and aggregate opinion on spending.

From a static perspective, to assess average opinion distributions and dispersion in issue dimensions requires a point of reference from which to assess whether the observed opinion and the observed dispersion is ‘polarized’ or not. For average opinion, I have suggested that polarization is much more probable when that opinion is near the center rather than when it is on the extremes. Average opinion close to the center is likely a product of opposing opinions rather than consensus squarely in the middle of the distribution. The disposition of dispersion

162

TABLE 6.6: DIMENSIONAL POLARIZATION AND POLARIZATION TRENDS DIMENSION Current Polarization Polarization Trend AVERAGE POSITION Ideology Y о

Jobs Y ј

Defense Y о

Spending Y љ

Aid to Blacks Y љ

Women N љ

Abortion Y ј

DISPERSION Ideology Y ј

Jobs Y о

Defense Y о

Spending Y о

Aid to Blacks N о

Women N љ

Abortion Y ј

BIMODALITY Ideology Y ј

Jobs Y о

Defense Y о

Spending Y ј

Aid to Blacks Y љ

Women N љ

Abortion Y ј

163

and bimodality within the opinion distributions provide confirmatory evidence for this supposition. For example, two of the polarized opinion distributions that exhibit increasing polarization, ideology and spending, are also two opinion distributions with the global average mean opinion nearest the center of the seven point issue scale (both at 4.25). Opinion on the equality of women in society, which exhibits significant depolarization trends, has an average mean position on the scale considerably distant from the center (2.55). However, this is not to say that all opinion distributions with centrally-located means exhibit conflict, that all trend towards the center are increasing conflict, and that all trends away from the center are evidence of increasing consensus. These are suggestive, not determinative. Indeed, you can have increasing polarization or depolarization without the mean position moving at all. In order to assess the polarization of the mean position for all citizens, one must take into account the distribution of opinion in terms of dispersion and bimodality.

In the section on dispersion inTable6.6, it is apparent that the distributions on these issues are fairly dispersed. All of the issues have average standard deviations above 1 with most of the variables closer to a standard deviation of 2 rather than to 1. The two most dispersed issue opinion distributions are aid to blacks (1.804) and jobs (1.862). Only two of the issues exhibit significant increasing dispersion since the 1970’s: ideology and abortion. Defense spending, government spending, and aid to Blacks exhibit no apparent trend in dispersion over the time series, at least in the aggregate, linear models estimated in Table 6.4 and Table 6.5.

Public opinion on the equality of women in society exhibits marked and significant declining dispersion.

The measure of polarization which exhibits the most polarizing trends over the past thirty-plus years is the kurtosis measure (bimodality). Two issue dimensions (abortion, government spending) and the ideological dimension have become more bimodal over this

164

period. Exhibiting declining bimodality are the aid to Blacks item and, once again, the women’s equality item. The one issue on which there is depolarization across all three measures is the women’s equality item, while the one issue that shows characteristics of polarization across all three measures is the abortion issue.

Overall, along multiple dimensions of political issues, there is strong evidence of polarization. On several issues, public opinion has moved towards the center of conflict, has an increasingly dispersed range of opinion, and a more bimodal shape. However, polarization has not occurred across the board. In several polarized distributions there was no apparent trend in polarization over this time. And on women’s ideology a previously polarized distribution has become distinctly depolarized.

Having examined the broad, linear trends in polarization on these political dimensions, in the next chapter I take a closer look at the trends in polarization. Have the linear trend models masked important, nonlinear relationships? Secondly, I look at potential causes of polarization. Are these random fluctuations in opinion, is it a function of elite manipulation, or are the trends in polarization on social, economic, government philosophy, and foreign policy dimensions comprehensible in terms of an attentive public reacting to exogenous political events and/or public policies advocated and enacted by the parties in the government?

165

CHAPTER 7: DETERMINANTS OF POLARIZATION

“The president spoke words of comfort with tear drops in his eyes / Then he led us as a nation into a war based on lies." – Sheryl Crow, “God Bless This Mess”

"The problem with socialism is that eventually you run out of other people's money to spend." – Margret Thatcher, 1976

““Affirmative action is the attempt to deal with malignant racism by instituting benign racism.” - Elliot Larson

“If men could get pregnant, abortion would be a sacrament.” - Florynce R. Kennedy, 1973

In Chapter 5 I took a narrow focus and examined the breadth and extent of polarization in public opinion on gay rights. In Chapter 6 I broadened that focus to include the entire social issue dimension as well as other types of issues such as economic issues and issues on foreign policy and examined aggregate linear polarization trends for these issues. In both of these prior analyses I focused on the questions of whether and to what extent polarization had occurred using several measures of polarization to answer those questions. In this chapter, I continue to ask whether polarization has or has not occurred over the “culture wars” time period from the

1970’s to the present. However, rather than look at the overall trend in polarization, I take a closer look at the trends in the public opinion on political issues by examining the trends in detail rather than summary form. The linear models employed in the previous chapters may have masked non-linear trends that actually reflect polarization on issue dimensions.

Furthermore, there may be early or late-breaking trends that is lost in the summary findings from the full time series. Secondly, I take an initial step at assessing the causes of polarization.

What are the factors that cause public opinion to polarize or depolarize? Are these random, chaotic events with no rhyme or reason? Or, is political polarization rational in the aggregate: a

166

response by society and groups within society to real world political events, policy promises made by candidates and parties, and the actual implemented policies of the government?

A causal model of public opinion is, in many ways, the Great White Whale, of political behavior analysis. The degree to which perturbations in public opinion are responsive to actual events, the manipulations of elite actors, perceptions of politics screened by a changing media landscape, implemented policies, the economy, personal circumstances, cultural and social group factors, etc. has been the subject of a great deal of theorizing and analysis in the discipline since Walter Lippmann pontificated on public opinion in the 1920’s and the empirical study of public opinion that began in earnest in the 1950’s. Studies of the relationship between public opinion and public policy have mostly shown that policy-makers are responsive to public opinion and that policy elites drive public opinion towards their views on policy (Shapiro et al. 1990;

Page 1979; Stimson 2004; Zaller 1992). As Harwood Childs noted in an early study linking public policy and public opinion, “the relationship between public opinion and public policy varies greatly from issue to issue. The influence of public opinion varies from virtually no influence to enormous influence. Influence may be exerted quickly or slowly; it may change over time or remain constant, and its impact may be direct or indirect” (Childs 1965). As Childs argued, the extent of the influence of public opinion on policy depends on factors including the degree of agreement within the public, the intensity of that opinion, and the extent of organized support for and against the public position. To that I can add the polarization of opinion on an issue.

Identifying causes of public opinion trends is a complex and difficult undertaking, plagued by poor and inconsistent measures of attitudes, interrelated and highly correlated causal variables, and significant omitted-variable bias. As Childs argued, the relationship between public opinion and public policy is cyclic and dynamic. In other words, public opinion not only influences public policy but, as I will show here, public policy also influences public

167

opinion. However, contrary to Childs’ supposition that the public becomes more accepting of a policy change once it is made, I will show that public policy opinion can move in the opposite direction: signaling disapproval of the adopted policy rather than acceptance. Indeed, this formulation is consistent with Sharp’s “thermostatic sequence” where public policy responds to public opinion and then public opinion provides feed back where public opinion adjusts to the policy change. Changes in policy responding to public opinion can move the policy too far, leading to the public moving in the opposite direction on the issue (Sharp 1999). Thus the importance of answering not merely the ‘what’ of polarization but also the ‘why’ of polarization is clear. While definitive answers may be allusive, analyses which identify important determinants of public attitudes on political issues, however contingent these models may be, are of substantive significance. Theoretically, we can define the potential relationships as follows:

Partisan Model. Public opinion polarization on issues is conditioned by the reactions of

the electorate to the partisanship of those in control of government policy. In

particular, partisans of the opposite party are more likely to polarize in response to the

other party in control of the policy agenda.

Policy Model. Public opinion polarization is responsive to the actual implemented policy

and changes in the levels or status of that policy. If, for example, defense spending

increases, then the public polarizes or depolarizes relative to the change in the status

quo and their preferred level of defense spending.

Events Model. Public opinion polarization is responsive to exogenous shocks to the

system, relatively independent of the current policy makers or their policy prescriptions.

These events independently affect the average issue positions of the public. For

168

example, the September 11th attacks could produce a significantly higher preferred level

of defense spending.

Note, these models are not mutually exclusive. In other words, the acceptance of one does not necessitate the rejection of the other models of public opinion influence. Rather, each could be a factor influencing public opinion. I will examine three potential explanations of political polarization: partisanship, exogenous events or ‘shocks’ to the political system, and implemented public policy on the issues themselves.

CASUAL FACTORS OF PUBLIC OPINION POLARIZATION: PARTISANSHIP, POLITICAL EVENTS, AND PUBLIC POLICY

Partisanship as an exogenous influence on public opinion can take on many forms. We can assess the partisan identification of the citizenry itself, public opinion on the relationship between the parties and the issues, the parties’ activities as an organization, and the parties as represented by elected and appointed officials in the government. It is this later conceptualization of ‘partisanship’ that I use as an independent causal influence on public attitudes on the issue dimensions. Specifically, I use as a proxy for the partisanship in the government the party affiliation of the presidential administration. While the federal government has often been split between the Republican and Democratic Parties (divided government), the president is the most visible individual within the government, and the public expectations that presidents will be decisive during political crises and in implementing public policy is well documented. There is thus a natural implication that presidents can influence public opinion through their policy advocacy and implementation of policy through executive administration and leadership on a legislative agenda. Using the party affiliation of the president as an independent factor causing shifts in public opinion trends serves two ends, one methodological the other substantive. As a measure, it permits a statistical analysis of presidential impact on public opinion while avoiding the classic n=1 problem associated with

169

empirical studies of and including the president. In addition to assessing presidential administrations in terms of partisanship, using survey year rather than the individual president as the unit of analysis is useful in resolving the small sample problem that presidents present.

Substantively this decision is justifiable given that presidents tend to have policy positions that mostly comport with that of their parties, and even when they diverge, from the perspective of the mass public, the president and the president’s party are strongly associated. Given the well- established positions of parties on major political issues, it is reasonable to expect that citizens generally view that policy does or will reflect the policy preferences of the party that controls the White House. This, in turn, could affect the public’s preferences for policy or policy changes on that issue dimension.

The second cause I consider in this analysis is exogenous political events. I define these as “exogenous shocks” to the political environment in the sense that they are events that are not directly the product of government policy. The stock market crash in 1989, the fall of the

Berlin Wall, the September 11th attacks, these are events that can be distinguished from overt and direct policy decisions such as the 1986 tax reform bill, the Reagan Cold War defense buildup, and President Bush’s decision to invade Iraq in 2003. While policy decisions may have created a situation which precipitated an exogenous event (e.g. the limited policy response of the Clinton administration to Al Qaeda attacks in the 1990’s), or policy may be made in response to an exogenous event (e.g. the U.S. invasion of Afghanistan in the wake of the 9-11 terrorist attacks), the events are clearly separate from and thus independent of the decision making and thus direct control of the policymakers themselves. These happenings can cause dramatic responses in the average opinion and opinion distributions of the American citizenry on one or more issue dimensions. Indeed, sometimes these events precipitate an extended and relatively

170

stable time-period where public opinion finds a new equilibrium quite distinct from the previous

period.

The third independent variable in this analysis of polarization determinants is public

policy itself. I test, where possible, if public opinion polarization is responsive to the actual

policies on the issues. Now, for some policies this isn’t practical, as the policy itself hasn’t

changed over the period, data on the policy is not available, or what is available fails to capture

the essence of the policy itself. For example, in examining the abortion issue, we could look at

the number of abortions each year, but is that a measure of abortion policy? Perhaps, in some

indirect sense, it is. However, it could just as easily be argued that public responsiveness to

abortion policy involves Supreme Court decisions, abortion legislation, and the continued

aftershocks from the landmark decision in Roe v. Wade. It may be possible to measure

‘abortion policy’ and include it in an analysis, but I do not do so here, except to assess abortion- related events in the context of polarization tends. However, on other issues, there is a clearly relevant and central policy that is measurable over the time period and thus I can assess public opinion and whether it varies with actual changes in the policy. For example, consider the opinion on levels of defense spending. The actual level of defense spending is clearly the central policy output relevant to public opinion on this issue. As such, I include a measure of this actual policy output in my analysis of polarization causes.

By taking into account multiple potential causes of polarization in addition to assessing the polarization trends in a detailed rather than summary fashion, a more nuanced and substantively rich depiction of polarization in American public opinion on political issue dimensions emerges. This analysis suggests polarization is comprehensible as a rational, aggregate reaction to prime movers in public policy such as the presidential administrations, the exogenous events directly related to the relevant political issues, and to actual policy outputs

171

produced by the federal government since the 1970’s. From the broad, ideological dimension

to the relatively narrow issue of abortion, I examine polarization in the context of the policy

actors taking positions, advocating for policies, and implementing policy changes. I examine

polarization in relation to real-world events that have structured and shifted American public

opinion—sometimes in dramatic fashion. And I look at the programs, policy-proscriptions, and

declared positions offered by administrations and perceived by the public both in terms of the

policies themselves and as screened through the lense of partisanship. I present a

sophisticated, multivariate analysis of the determinants of political polarization.

IDEOLOGY DISPERSION & BIMODALITY: CONSISTENT IDEOLOGICAL POLARIZATION OF THE AMERICAN PUBLIC

As one polarization scholar argued, a key test of political polarization is the degree to which the American public ideology has become more dispersed and more coherent in the ‘two opposing camps’ sense. The increasing trend in the dispersion of the distribution of mass ideology is depicted in Figure 7.1 divided by presidential administrations over the course of the data period. There was a slight increase in dispersion through the Nixon and into the Carter

FIGURE 7.1: DISPERSION TREND IN MASS IDEOLOGY, 1972-2008 2 1.9 Nixon / Carter Reagan / Bush I Clinton Bush II 1.8 Ford 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1

IDEOLOGY S.D.

172

years, however this declined or held steady through all of the 1980’s into George H.W. Bush’s term. With the election of Clinton the dispersion in ideology reached a new and higher plateau, but it remained steady throughout his presidency until the last two years (the run-up to the

2000 presidential election). Whether a consequence of the divisive and controversial 2000 presidential election of George W. Bush, the attacks on September 11th, or the 2002 War in Iraq, the dispersion in ideology for the mass public increased substantially in the first decade of the

21st century. By 2008 the dispersion in mass ideology was as high as it had ever been during the entire time series. On average the distribution of ideology at the mass level has become more spread out (more polarized), suggesting a worsening political environment for compromise and a target-rich environment for the purveyors of political and partisan conflict .The trend of increasing dispersion in the ideological distribution of the American public is consistent with the trend of increasing bimodality as shown in Figure 7.2. While the 1990’s and the Clinton administration coincided with an increasingly polarized ideological distribution relative to the

FIGURE 7.2: BIMODALITY TREND IN MASS IDEOLOGY, 1972-2008 0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 War in -0.8 Iraq -0.9 -1

IDEOLOGY KURTOSIS

173

previous two decades, the election of George W. Bush in 2000 was proximate to the most significant playtkurtic spike in the time series. Between 1998 and 2000 the bimodality (kurtosis) of ideology goes from the closest it ever was to the normal distribution at any point in time in the series (-0.24) to a new low, indicating that the mass public had increasingly shifted into two ideological camps. Indeed, the bimodality of ideology for the entire Bush presidency is lower than any of the points for any of the other years in the time series. Hence on two major indicators of polarization—dispersion and bimodality—the American public has become significantly and substantively more polarized on ideology. Relative to the 1970’s, the average distance between ideological identifiers is greater and the distributional shape of ideology indicates conservatives grouping with fellow conservatives and likewise for liberals. Either through a reduction in the number of moderate and slightly ideological identifiers, or a spike in the tails of the ideological distribution, there is considerably more ideological conflict in

America. The war in Iraq further polarized the country, as 2002 set the high watermark of bimodality in the ideological distribution. The change in bimodality between 1998 and 2002 is quite remarkable. In the span of four years, bimodality on ideology set its global maximum and global minimum for the time series.

The most apparent intervening events between those two survey years are the election of President Bush in 2000 and the war in Iraq in 2002. The 2000 election was a highly controversial and thus potentially polarizing election. The dispute centered mostly over who, between George W. Bush and Al Gore, won the state of Florida. Initially reported on election eve as going to Gore, the final count showed George W. Bush with a narrow lead, and hence the decisive electoral votes necessary to win the presidency. Subsequent recount and challenges narrowed Bush’s lead, but it was certified by the Florida Secretary of State at 537 votes. Gore appealed to the Florida Supreme Court, which ordered a statewide recount of 70,000 rejected

174

FIGURE 7.3: BIMODALITY TREND IN PUBLIC OPINION ON GOVERNMET SPENDING & SERVICES, 1972-2008

0 -0.1 -0.2 -0.3 -0.4 -0.5 -0.6 -0.7 -0.8 -0.9 -1 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

SPENDING KURTOSIS NT ballots. This recount process was ruled unconstitutional by the United States Supreme Court in violation of the equal protection clause in the constitution in a 5-4 vote. While Gore conceded following the court decision, many Democrats in and out of elected office believed the result to be illegitimate. Adding to the potentially polarizing nature of the event was the fact that Gore won the popular vote (+543,895). The bitterness over such a hotly contested presidential election and the subsequent controversy and public spectacle of an election contest had a significant impact on the polarization of the American public on ideology.

While 2004 and subsequent survey years have shown less bimodality, each year of the

Bush presidency reflects a more bimodal distribution of ideology than in any previous year since

1972. To the extent that ideology serves as a proxy for the political agenda for issues and the conflict of social groups, the evidence on dispersion and bimodality in ideology suggests a more conflictual political space and a policy arena less friendly to compromise. The evidence of increasing political polarization at the mass level is strong and robust.

175

GOVERNMENT SPENDING BIMODALITY & DISPERSION: MIXED EVIDENCE OF POLITICAL POLARIZATION

The evidence for polarization on government spending and services is not as clear as that for the ideological dimension. There is no apparent trend in the dispersion of opinion on the government role in helping citizens acquire and keep jobs. However, there is some evidence that this issue has become the subject of more political conflict within the mass electorate.

Figure 7.3 depicts the trend in the bimodality of opinion on government spending since 1982.

The trend is relatively stable, though there is significant volatility in the measure in the 1990’s.

While 1998 was a low point in the ebb and flow if ideological kurtosis, it is the most bimodal distribution of opinion on government spending in the time series. While it recovers somewhat for the 2000 survey year, there is a fairly steady decline in kurtosis for the subsequent years of the Bush administration. However, even at its most leptokurtic point, opinion on government spending is non-normal with bimodal characteristics. The absence of a trend here should not be surprising. Whether or not and to what extent the government should provide services or the market is one of the fundamental questions of democratic politics and has likely been a subject of division between social groups, parties, and ideological camps for both the mass public and elites since the ancients invented the word “government.”

As I noted at the beginning of the chapter, some issues lend themselves to a direct assessment of the relationship between opinion polarization and the actual public policy on the issue. The nature of government spending does provide an opportunity to assess the degree to which the bimodality of mass public opinion on government services and spending is related to actual (as opposed to perceived) levels of spending and services as well as assess the trends in opinion on government spending relative to real spending levels. It further can be suggestive as to what is causing variation in opinion on government spending as well as what spurs changes in

176

FIGURE 7.4: U.S. GOVERNMENT SPENDING AS PERCENTAGE OF GDP, FISCAL YEARS 1970-201049 50

20 1970 1971 1972 1973 1974 1975 1976 1977 1978 1979 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008 2009 2010

GOVERNMENT SPENDING (%GDP)

49 Fiscal Years 2009 & 2010 are estimated.

177

FIGURE 7.5: BIMODALITY TREND (-Z-SCORE) ACTUAL SPENDING % GDP (Z-SCORE), PUBLIC OPINION ON SPENDING & SERVICES (Z – SCORE), 1980-2008 3

-1

-2

-3

GOV'T SPENDING % GDP (Z) GOV'T SPENDING PO MEAN (Z) GOV'T SPENDING PO BIMODALITY (-Z)

178

the polarization of opinion on government spending. Figure 7.4 shows government spending as

a percentage of GDP (Gross Domestic Product) from 1970 to 2008, with the fiscal years of 2009

and 2010 estimated according to budget projections.50 The U.S. Government spending has

oscillated around 35% of GDP since the 1980’s. However, note the significant uptick in 2008

followed by a substantial jump in the projected figures for the first year of the Obama

administration.

Figure 7.5 converts government spending as a percentage of GDP, the average opinion

on the level of government spending and services from the ANES, and the bimodality measure

(kurtosis) for government spending to Z-scores based on the means and standard deviations for

the data series between 1982 and 2008. This permits an analysis of changes in the polarization

trends for government spending opinion and actual government spending on the same scale.

Thus changes in the opinion variables are directly comparable to changes in the actual levels of

government spending. The Z-Score calculation can be found in Equation 7.1:

Equation 7.1: Z-Score for Government Spending

ZGS = ீௌ೔ିீௌതതതത೔ ఙ೔ Where: = the ith observed value of government spending for year. = the ith mean value of government spending for year. ௜ .the ith standard deviation of government spending forݔ year =ܵܩ ௜ തതതത ݔܵܩ ߪ௜ ݔ This allows for a direct comparison of the variables despite the fact that they are scaled using

disparate units. As such the values for government spending (%GDP), average opinion on

50 Government spending data compiled from the U.S. Treasury’s site on Historical Debt: www.treasurydirect.gov/govt/reports/pd/histdebt/histdebt.htm. GDP data compiled from the data published at www.usgovernmentspending.com.

179

government spending, and the bimodality of government spending are translated into

differences from their respective means and standardized relative to the Z-distribution (Z-

scores) centered on a mean of zero and a standard deviation equal to 1. Hence the 1.14 Z-score

on government spending bimodality in 1990 indicates that the kurtosis for government spending

in that year was 1.14 standard deviations above the mean of government spending for the full

time series. In order to match the year values of government spending with the bi-yearly ANES

data on government spending opinion, I interpolate intervening values for opinion on

government spending and services for the intervening years (see footnote 43 for the model of

linear interpolation used in this analysis).

The trend lines in Figure 7.5 appear to be fairly independent of one another on the

whole. Through the 1980’s, public opinion on government spending, both in terms of the

average opinion and the bimodality of the opinion distribution, track closely with actual levels of

government spending as a percentage of GDP. Interestingly, conflict in the late 1980’s over

government spending was more unusual than the actual levels of spending or the opinion on

levels of spending would indicate. While it is just speculation, this rise in conflict is coincident

with Black Monday, the 1987 world-wide stock market crash that saw the Dow Jones Industrial

TABLE 7.1: DEVIATION MODELS REGRESSING GOV’T SPENDING LEVELS (%GDP) ON GOV’T SPENDING PUBLIC OPINION Parameter MODEL: Intercept Estimate 2 = + ( % ) + (S.E.) (S.E.) R N

SPENDING࢖࢕ (MEAN)૙ ૚ ࡳࡰࡼ -0.001 -0.049 .002 27 ࡿ ࡮ ࡮ ࡿ ࢋ (0.196) (0.200)

SPENDING (KURTOSIS) -0.001 0.088 .008 27 (0.195) (0.199)

* significant at .10 level ** significant at .05 level ***significant at .01 level

180

Average plunge over 500 points and lose over twenty percent of its value (22.61%). The decade of the 1990’s has the most volatile shifts in opinion on government spending in addition to the trend lines moving more independently than at any other time. In terms of the average opinion on government spending, it peaks in the early part of the 1990’s but takes a precipitous drop proximate to the 1994 “Republican Revolution” and the historic shift of control of both Houses of Congress to the Republican Party. However, this trend reverses itself and by 2000, the average opinion on government spending is right at the mean (4.25) for the time series.

Opinion shifts substantially in 2008 towards increased government spending, undoubtedly in part a reaction to the financial collapse and stock market crash in the fall of 2008 in the lead-up to the November elections. The bimodality in government spending for the 1990’s reflects extreme oscillations relative to the mean, perhaps in reaction to, at first, the Clinton administration’s unpopular health care proposal in 1993 and then to the equally unpopular government shut down in 1995 that Clinton successfully (from the standpoint of public opinion) depicted as the Republicans’ responsibility. This volatility gives way to a steady bimodal trend in the Bush administration. While public opinion on government spending is apparently sensitive to partisan manipulation and exogenous events, actual government spending increased from

1994 to the present day independent of partisan control of the White House, Congress, or external events (that is, external events do not seem to have resulted in any spikes or troughs in government spending outside of the most recent economic collapse).

Table 7.1 employs the real government budget outlays (% of GDP) as an independent variable in regression models of public opinion on government spending and services. The absence of any apparent linear relationship in the three variables on government spending nor any covariance between actual spending and opinion on spending evident in Figure 7.5 is consistent with the findings (or lack thereof) in Table 7.1. The models using government spending to predict the

181

average public opinion on spending levels and the bimodality in the distribution of opinion on

government spending and services fail to explain the variance in either of those measures given

standard levels of significance (.10, .05, & .01). Indeed, these linear models contribute almost no

improvement over the mean model of government spending (.002 and .008 R2). While opinion on government spending does appear responsive to partisan control of government and economic issues, there is no evidence it is responsive to the actual levels of government spending itself as a percentage of GDP.

D EFENSE SPENDING / FOREIGN POLICY BIMODALITY – A RATIONAL PUBLIC REACTS TO THE REAL WORLD

As I pointed out in assessing the linear models of trends in the issue self placements reported in Table 7.4, the OLS regression model of defense spending overtime is a poor fit for the data. The linear model of defense doesn’t work because the defense spending

FIGURE 7.6: BIMODALITY TREND IN MASS OPINION ON DEFENSE SPENDING, 1980-2008

1 Clinton 0.8 Administration 0.6 G.W. Bush’s 0.4 ‘Axis of Evil’ Berlin Wall Speech 0.2 Comes Down Reagan’s 0 ‘Star Wars’ -0.2 Speech

-0.4 Bush’s ‘New World Order’ War in -0.6 Reagan’s Speech Iraq -0.8 ‘Evil Empire’ Speech -1 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

DEFENSE KURTOSIS

182

trend is distinctly non-linear, as is apparent in Figure 7.6. The conflict over defense policy, as

captured in the playtkurtic distribution on defense policy, increased from the outset of Ronald

Reagan’s presidency and only began to improve as George H. W. Bush succeeded Reagan in

1988. Shortly thereafter the Berlin wall came down and the slow-motion collapse of the Soviet

Union is coincident with a marked decrease in the bimodality for the distribution of opinion on

defense spending. This was followed by a period of relative peace (dubbed by George H.W. Bush

as the “New World Order”) and no commitments of significant forces of U.S. ground troops in

active military conflict during the Clinton administration (the Serbian-Kosovo war was mostly

conducted through the air). This would all change with the dual terrorist attacks on New York

and Washington D.C. on September 11th, 2001. Those attacks spurred a new buildup in defense spending with two separate land wars in Iraq and Afghanistan. The measure of bimodality in

Figure 7.6 tracks with these broad foreign policy and defense policy-related events: trending increasingly bimodal during the Cold War, then strongly unimodal during the New World Order, and finally trending once again towards bimodality during the Bush administration in the first decade of the 21st century.

While there appears to be a strong relationship between the exogenous foreign policy events that defined distinct defense policy periods over the past thirty-plus years, this is not the final word on opinion polarization on defense spending. As I noted earlier, there are multiple potential causal factors operating independently and in concert to influence and structure public opinion on defense spending and, by implication, foreign policy. In order to take a more sophisticated look at defense spending opinion, we need to incorporate the other factors which theory and experience suggest are potentially significant factors. One such is the defense spending levels themselves. While the raw levels of defense spending would be one and perhaps appropriate measure of defense spending, I look at defense spending as a percentage

183

of GDP. This controls for confounding factors such as inflation and budget-creep, and it permits a more valid measure of defense spending—assessing how much of the economic pie defense- related spending consumes in a given year. I report the actual level of defense spending from

1980 to 2003 as a percentage of GDP in Figure 7.7. Note that spending on defense peaked in the mid-1980’s, at the height of the Cold War, and steadily declined in the post-Soviet Union period of George H.W. Bush’s “New World Order” and on through the span of the Clinton administration. It begins to tick back up post-911 as the prosecution of two wars—Afghanistan and Iraq—required significant increases in defense appropriations.

FIGURE 7.7: DEFENSE SPENDING AS A PERCENTAGE OF GROSS DOMESTIC PRODUCT, 1980-200351 7 6.5 6 Clinton 5.5 Administration 5 4.5 4 3.5 Berlin Wall 3 Comes Down 2.5 2 September 11th 1.5 Attacks 1 0.5 0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003

Defense Spending % GDP

51 “Table 3.1: Outlays by Superfunction and Function: 1940--2003," in Office of Management and Budget, Historical Tables, Budget of the United States Government, Fiscal Year 2005 (2004), Washington, pp. 45-- 52

184

In March 1983, Ronald Reagan gave two famous speeches that set the agenda for his presidency on foreign policy: a more aggressive, judgmental and confrontational stance towards the Soviet Union matched with a substantial defense spending build-up. On March 8,

1983, President Reagan deliveredan address to a meeting of the National Association of

Evangelicals in Orlando, Florida. In it, he referred to communism as "the focus of evil in the modern world," and it quickly became known as his "Evil Empire Speech." Reagan had similar language in a speech to the British Parliament in 1982, but aides struck the language before

Reagan delivered it. The Evil Empire

speech was delivered at a time when “Let me share with you a vision Congress was debating a resolution in of the future which offers hope. It is that we embark on a support of a "nuclear freeze," a doctrine program to counter the awesome Soviet missile threat supported by the Soviet Union that with measures that are defensive. Let us turn to the very would have prevented the deployment strengths in technology that of U.S. cruise and Pershing II Missiles in spawned our great industrial base and that have given us the Europe. On March 7, President Reagan quality of life we enjoy today…a shield that could protect us from had met with a conservative leaders and nuclear missiles just as a roof Hawkish officials in regards to the protects a family from the rain"

– Ronald Reagan, 1983 nuclear freeze legislation moving

through Congress. The President reiterated his opposition to the nuclear freeze, and meeting participants urged him to go public on the topic and make an appeal directly to the people. According to a contemporaneous report by the President's National Security Advisor Judge William Clark, the president added the famous paragraphs to the speech he delivered the next day to the National Association of

Evangelicals. As the National Center for Public Policy Research put it, “Those additional

185

paragraphs turned it from a routine, if worthy, speech to one that electrified dissidents behind the Iron Curtain and appalled Reagan's domestic opposition, including much of the press”

(Reagan 1983). While the speech itself is not a cause of defense spending, this speech (and the subsequent presidential speeches mentioned here) are a marker for major changes in foreign policy as a consequence of presidential vision and/or historical circumstance.

In the same month, Regan delivered a speech to the nation on national security and the defense budget. On March 23rd, 1983, Reagan delivered his speech, which focused on the debate in Congress over defense spending cuts and the ‘nuclear freeze,’ the threat of the continued buildup of nuclear arms by the Soviet Union, the necessity of addressing the defense spending crisis due to neglect of the defense budget during the Carter administration, and his administration’s commitment to guaranteeing “peace through strength.” Towards the end of the speech Reagan advocated the Strategic Defense Initiative (though SDI wasn’t mentioned specifically), a defense spending program aimed at developing a technological solution to the threat of intercontinental ballistic missile attacks through, in part, the deployment of satellites into space (Reagan 1983). Dr. Carol Rosin is credited with first referring to the program as “Star

Wars,” a derisive term intended to suggest the whole concept of missile defense was fantastical.

Hailed by supporters as a major factor in the collapse of the Soviet Union, SDI was a very controversial program. Mikhail Gorbachev spoke out strongly against SDI. And Reagan’s domestic opponents attacked SDI, the deployment of missiles in Europe, and the “cowboy” diplomacy of the Reagan administration. All of this played into an atmosphere of intense partisan and political conflict over defense policy, diplomacy, Reagan’s foreign policy, and the

Cold War. This is evident in the playtkurtic trend in the kurtosis of defense spending opinion

(Figure 7.6). Whatever the ultimate cause of the collapse of the Soviet Union during George

186

H.W. Bush’s administration in the early part of the 1990’s52, the consequence was a significant shift in actual defense spending and the perception of defense spending in the American electorate. As is apparent from Figure 7.7, the reduction in defense spending, at least as a

percentage of GDP, began in the first year of the GHW

FG Bush administration and declined steadily through

Now, we can see a new 2000 with the sole exception of 1992. In the wake of world coming into view. A the successful and low-casualty war in the Gulf in world in which there is the very real prospect of a 1990-91, George H. Bush, as many erstwhile

new world order… A prognosticators before him, asserted an end to war as world where the United Nations, freed from cold a function of traditional balance of power international war stalemate, is poised politics or the adventurism of opportunistic nation. The to fulfill the historic vision “New World Order” that George H. Bush declared in of its founders. A world in which freedom and his March 6th, 1991 speech involved the use of a respect for human rights coalition, backed by international organizations such as find a home among all nations. NATO and the United Nations, to enforce an

- George H. W. Bush, 1991 international ‘law’ against aggressor states and an ED idealistic vision of a world in which war (rather than coalition-backed police actions) is an anachronism.

War would—if not disappear—certainly change in a fundamental way. And efforts by ambitious nations to act outside of the framework of international law would be met by a united world and a coalition of nations ready to punish the transgressor.

52 Some credit Regan’s aggressive foreign policy stance, the strong commitment to SDI and the defense spending build-up (walking out on arms talks over it in Iceland) and the inability of the Soviets to match it as the crucial, deciding factor in the collapse of the USSR. Others point to the structural failures of the communist system and/or the Gorbachev reforms of “glasnost” and “perestroika” in 1985 and 1987 respectively as precipitating the end of the Soviet Union. (see Figure 5.9).

187

While the United States would commit to support of security organizations aimed at preventing regional warfare, Bush explicitly ruled out committing U.S. land forces and placed regional security at the foot of regional powers (Bush 1991). This scaled back interventionism was matched with Bush’s argument that decreased defense spending would bring about economic benefits (the Peace Dividend). This "I know…that the United States cannot--indeed, we should not-- reluctance to commit forces on the ground, but be the world's policeman, and I know this is a time, with the Cold rather to rely heavily on air power, and War over, that so many emphasis on defense spending cuts would Americans are reluctant to commit military resources and continue to reflect itself in American foreign our personnel beyond our shores." policy in the Clinton administration. -William J. Clinton, 1994

Clinton promised dramatic defense cutbacks and, while initially subscribing to ‘holding budgets’ rather than initiating a rollback of

Cold War defense spending, his administration witnessed a marked decline in defense spending.

As Figure 7.7 illustrates, defense spending declines from 5% of GDP in 1992 to 3% of GDP in

2000. This represents a 40% reduction in defense spending over the eight years of the Clinton administration. The decline in defense spending was marked by a philosophical move away from intervention. Under Clinton, the only significant commitment of ground troops was in Haiti, a quintessential police action in the mold of “New World Order” foreign policy. The Serbian conflict was almost exclusively prosecuted by air power, and Clinton withdrew forces from

Bosnia after a high-profile botched operation resulted in several U.S. marine casualties. The defense spending cutbacks, base closures, and an apparent reluctance on the part of Clinton

(either due to non-interventionist foreign policy philosophy or a concern that Clinton’s own war- protesting and draft avoidance would make the serious commitment of ground troops by him a

188

political problem beyond the efficacy of the military decision) to commit troops on the ground resulted in a minimalist interventionist strategy during the Clinton years. Indeed, even in such minor altercations as the Haiti police action, where the U.S. met no significant military opposition, Clinton couched the actions in terms of a reluctance to intervene in international affairs (Blumenthal 1994).

On September 11th 2001 at 8:46 am, hijacked American Airlines Flight 11 was flown into

the North Tower of the World Trade Center in New York City

FG by Mohammed Atta, an Al Qaeda terrorist. This attack was

States like these, and followed by another, when United Airlines Flight 175 crashed their terrorist allies, into the WTC South Tower. A third attack was made on the constitute an axis of evil, arming to Pentagon (American Airlines Flight 77) and a fourth—perhaps threaten the peace of intended for the White House or the Capitol Building—was the world. By thwarted by passengers and crew on United Flight 93 after seeking weapons of mass destruction, learning of the fates of the other three planes in flight over

these regimes pose a passenger cell phones. Catastrophic failure of the integrity of grave and growing danger. the two WTC towers, as a consequence of the structural

-George W. Bush, damage done from the fire and concussive force of the two 2002 fully fueled jetliners, caused each of the towers to collapse. ED Khalid Sheik Muhammad, under the direction of Al Qaeda

leader Osama Bin Laden, planned and coordinated the attacks. Al Qaeda had taken shelter in Afghanistan under the control of the Taliban, a fundamentalist Muslim government with a harsh interpretation of Sharia Law. In October 2001,

U.S. forces invaded Afghanistan and overthrew the Taliban in Operation Enduring Freedom. The

September 11th attacks would precipitate a major shift in the nascent Bush administration’s

189

foreign policy and spur a substantial surge in defense spending. While candidate Bush had expressed a fairly modest foreign policy aims that bordered on isolationism, President Bush initiated an aggressive policy centered on two pillars: the preemption doctrine and the terrorism sponsor doctrine, both of which would come to be known generally as the “Bush doctrine.” The new disposition towards states that sponsor terrorism provided the justification for prosecuting wars in both Afghanistan and Iraq. Under the doctrine, as articulated by

President Bush in his 2002 State of the Union speech (known as the “Axis of Evil” speech), the

United States would not distinguish between terrorists threatening attack on America and the states that gave them safe harbor. Furthermore, under the preemption doctrine, the United

States would not wait until an attack was initiated, but rather asserted that threats could and would be addressed before an attack could be made. This doctrine is articulated in the 2002 government document: National Security Strategy for the United States. “It is an enduring

American principle that this duty obligates the government to anticipate and counter threats, using all elements of national power, before the threats can do grave damage.”

On March 20th, 2003, Operation Iraqi freedom was initiated and the Iraq war was underway.

Though initially an unvarnished military success, the toppling of Saddam Hussein would give way to an insurgency and an intense sectarian civil war between Shiites and Sunni’s that would seriously undercut support for the mission in Iraq in the public. While the Bush doctrine was the philosophical basis for the war, in the public campaign to win support for the war the administration emphasized the threat posed by Iraq given its believed stockpiles of weapons of mass destruction. The failure to find significant stockpiles by the Iraq Survey Group would prove a significantly polarizing event, as it lead the opposition to the war and the Bush administration to charge that Bush had “lied” us into war. The September 11th attacks were the impetus for the largest “Rally around the Flag” effect since we began measuring public opinion (Figure 7.8).

190

George W. Bush was the recipient of the highest job approval rating for a president in the history of public opinion polling, topping out at 89% approval in October 2001, the month after the attacks.

These numbers were unsustainable given a return to some sense of normalcy in politics.

Clearly the prosecution of the War on Terror, particularly in Iraq, served to seriously erode public support for the Bush administration (Figure 7.8). Though Bush narrowly won reelection over John Kerry in 2004, that year would mark the last in which his approval rating would exceed his disapproval rating. With the security difficulties in Iraq, the difficulties in establishing a stable, democratic Iraqi government with the participation of Iraqi Sunni and Shiite leaders, and the growing number of casualties brought on by the Iraqi Insurgency, Bush lost the confidence of the public on the war. By 2007, the NIE (National Intelligence Estimate) had declared a civil war underway in Iraq, and the anti-war movement was putting on large demonstrations across the country. Though the appointment of General Petraeus and the 2006 “surge” of forces reestablished security and reduced casualty counts, the die was cast on the Bush administration and the Iraq war in the eyes of the public.

While pop stars writing colorful song lyrics accusing Bush of ‘lying’ the country into war and large anti-war demonstrations on the Mall in Washington, D.C. suggest polarization and political conflict over foreign policy, they are not sufficient markers in and of themselves. A

March 2003 poll showed that only 5% of the public reported having participated in an anti-war protest or having made a public demonstration against the war (Bowman 2008). Furthermore, there is nothing to say that either pop stars or large protest demonstrations are representative of the public at large, or even of a substantial portion of the public. Large protests against the war were organized in 2003, when approval for the war was at 65% and approval of President

191

FIGURE 7.8: GEORGE W. BUSH JOB APPROVAL RATING (GALLUP / USA TODAY POLL), FEBRUARY, 2001 – DECEMBER, 200853

100

80 Invasion of NIE: Civil War in Iraq 70 Iraq Iraqi Insurgency 60

September 11th 50 Attacks 40

30 Interim Report

20 Iraq Survey Group: No WMD Found 10

0 01 01 01 02 02 02 03 03 03 04 04 04 05 05 05 06 06 06 07 07 07 08 08 08 01 01 02 02 03 03 04 04 05 05 06 06 07 07 08 08 01 02 03 04 05 06 07 08 ------Jun Jun Jun Jun Jun Jun Jun Jun Oct Oct Oct Oct Oct Oct Oct Oct Feb Apr Feb Apr Feb Apr Feb Apr Feb Apr Feb Apr Feb Apr Feb Apr Aug Dec Aug Dec Aug Dec Aug Dec Aug Dec Aug Dec Aug Dec Aug Dec

BUSH APPROVAL BUSH DISAPPROVAL

53 The data for Figure 5.8 was obtained from the “Polling Report” website which collects and report polling data from the gamut of polling organizations in the United States. www.pollingreport.com/BushJob1.htm. It is the approval and disapproval numbers for the national adult sample taken by the Gallup Organization / USA Today poll in answer to the question: “Do you approve or disapprove of the way George W. Bush is handling his job as president?" The job approval data in Figure 5.8 is averaged by month for the entire Bush administration. The maximum number of polls in a month included in the series is seven, the minimum number is one.

192

Bush was in the sixties. Intensity of opposition does not connote the frequency of opposition.

The bimodality trend in public opinion on defense spending in Figure 7.6, however, provides a

measure of conflict independent of these unreliable polarization markers. It shows that conflict

within society was significant in the 1980’s, declined in the 1990’s, and has spiked in the first

decade of the 21st century. But what is causing these shifts of American public towards and

away from “two camps” status? The linear model of defense spending bimodality (Table 7.4)

doesn’t fit well because, as is apparent in Figure 7.6, the trend in bimodality on the defense

spending issue scale is decidedly nonlinear. While there is an overall positive trend for the time-

series due (at least in part) to the substantial move towards consensus coincident with the

collapse of the Soviet Union up until the events of September 11th, 2001 and the subsequent

war in Iraq, this trend masks the significant bimodality during the height of the Cold War in the

1980’s as well as precipitous decline in consensus on defense and foreign policy in the post-9/11

world. But why have the trends in the division of the American public over defense spending

been so volatile?

In order to answer this question, as with government spending, I can assess trends in

actual defense spending with the trends derivative of public attitudes on defense spending from

the ANES using Z-Scores:

Equation 7.2: Z-Score for Defense Spending

ZDS = ೔ ೔ Where: ஽ௌ ି஽ௌതതതത th = the i observed value of defense spending forఙ ೔ year. = the ith mean value of defense spending for year. ௜ .the ith standard deviation of defense spending forݔ year =ܵܦ ௜ തതതത ݔܵܦ ߪ௜ ݔ As I noted with government spending, this allows for a direct comparison of the variables

despite the fact that they are scaled using different scales and units. As such, the values for

193

FIGURE 7.9: BIMODALITY TREND & DEFENSE SPENDING TREND EXPRESSED AS Z SCORES, 1980-2008 3

-1

-2

-3 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

DEFENSE SPENDING - %GDP (Z) DEFENSE BIMODALITY KURT (Z)

FIGURE 7.10: BIMODALITY TREND (-Z-SCORE) & DEFENSE SPENDING (Z-SCORE) IN WAR & PEACE, 1980-2008 3 COLD WAR NEW WORLD IRAQ WAR ORDER 2 Berlin Wall Comes Down 1

0 Glasnost Perestroika

-1

Persian Gulf th -2 War September 11 Attacks

-3 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997 1998 1999 2000 2001 2002 2003 2004 2005 2006 2007 2008

DEFENSE SPENDING - %GDP (Z) DEFENSE BIMODALITY KURT (-Z)

194

defense spending (%GDP), average opinion on defense spending, and the bimodality of defense spending are translated into differences from their respective means and standardized relative to the Z-distribution (Z-scores) centered on a mean of zero and a standard deviation equal to 1.

The z-scores for defense spending bimodality and for actual defense spending are reported in

Figure 7.9. In order to match increases onthe defense spending scale with increases in bimodality (conflict), I flip the z-scores for defense spending kurtosis and report the negative z- score for defense spending (Figure 7.10).

As Figure 7.10 shows, the variance in the bimodality of defense spending and actual defense spending as a percentage of GDP track very closely together. Indeed, it is easier to point out incongruities than it is congruities. In the early part of the 1990’s, defense spending bimodality is well below the actual defense spending levels in terms of deviation from the mean.

In 1992, we see more polarization, above that of the declining defense spending levels. Finally, in the later part of the 2000’s polarization is again on the rise, outstripping even the significant relationship between the mean public opinion on increasing or decreasing spending relative to actual defense spending levels. As actual defense spending levels become more extreme relative to the mean defense spending levels for the period, public opinion on what the level of defense spending should be moves in the opposite direction.

There are three readily apparent explanations for the trend in bimodality we observe from 1980 to 2008. Briefly, they are public responsiveness to the status quo on defense spending, partisanship, and defense-related exogenous shocks and foreign policy. The first and perhaps simplest explanation for the shifts in bimodality of defense spending is the mass public reacting to changes in the status quo on defense spending. Indeed Figures 7.10 and 7.11 seem to show this is pretty much the entire story. But to conclude that is premature. There are apparent

195

FIGURE 7.11: BIMODALITY TREND (-Z-SCORE) ACTUAL DEFENSE SPENDING % GDP (Z-SCORE), PUBLIC OPINION ON DEFENSE SPENDING (Z-SCORE), 1980-2008 3

Status Quo on Status Quo on 0 Defense Spending Defense Spending

-1

-2

-3 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 8 8 8 8 8 8 8 8 8 8 9 9 9 9 9 9 9 9 9 9 0 0 0 0 0 0 0 0 0 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 19 20 20 20 20 20 20 20 20 20

DEFENSE SPENDING - %GDP (Z) DEFENSE MEAN (Z) DEFENSE BIMODALITY KURT (-Z)

196

relationships, in both means and kurtosis, between the presidential administration and exogenous foreign policy events as well.

While we only have two switches in the party of the presidential administration (GHW

Bush Æ Clinton; Clinton Æ GW Bush), bimodality in defense spending as well as the mean preferred defense spending level are distinctly different between Republican and Democratic administrations. There was more conflict/polarization, as measured by kurtosis, in Republican administrations as opposed to the Democratic administration in the time series. Why might this be? It is possible that the Left and its anti- war faction are more active in relation to defense spending when Republicans control the White House (and hence foreign policy and the military), given their presumably Hawkish disposition, as opposed to Democrats. Or, at least in the case of

Bill Clinton, who was not only a Democrat but who had protested the Vietnam War, had avoided the draft, and instituted the “Don’t Ask, Don’t Tell” policy on gays in the military.

That said, there are some indications that declining FG conflict over defense spending and foreign policy during

For many liberal Democratic administrations may be more general— activists, opposing the war was really extending beyond the personal characteristics of Bill Clinton. about opposing Since Barak Obama ascended to the presidency, the anti-war George W. Bush. When Bush fervor on the Left seems to have dissipated. Byron York disappeared, so did reports on straw poll taken by Democratic pollster Stanley their anti-war Greenberg at the 2009 “YearlyKos,” a convention organized passion. by the founder of the Daily Kos blog, a popular Leftwing site - Byron York, 2009 (York 2009). As York notes, “No group was more angrily

ED opposed to the war in Iraq than the netroots activists

197

clustered around the left-wing Web site DailyKos. It's an influential site, one of the biggest on the Web, and in the Bush years many of its devotees took an active role in raising money and campaigning for anti-war candidates.” As part of the straw poll, Greenberg presented attendants with a list of policy priorities such as health care or the environment. He asked people to list the two priorities they believed "progressive activists should be focusing their attention and efforts on the most." Health care came first followed by the environment.

Military involvement in Iraq and Afghanistan was well down the list in eighth place. On the question of what two issues "you, personally, spend the most time advancing currently," attendants again listed health care reform first. Coming in dead last, named by just one percent of attendants, was “working to end U.S. involvement in Iraq and Afghanistan” (York 2009). This is despite the fact that, as of the conference, there were 130,000 troops in Iraq and a plan for

68,000 troops in Afghanistan by the end of the year. Perhaps, just as only Nixon could go to

China, Democrats are the only presidents who can go to war without generating intense anti- war opposition.

The third possible explanation for the variance in polarization on defense spending and foreign policy are the exogenous foreign policy-related events that are, in many respects, independent of the policy-making process, yet at the same time spur (even dictate) foreign policy decisions by government officials and defense policy actors. The Cold War dates back to the 1950’s and thus any president taking office had to grapple with the threat of the Soviet

Union, the bipolar world of the Superpowers, and policies such as détente, which enjoyed a status quo position in the 1980’s. The end of the Soviet Union is a good example. It is an exogenous event (certainly from the perspective of the Bush administration, whether you credit

Reagan Era foreign-policy with the collapse or not), and thus any president in office when it happens would have to respond to it. However, that response is not preordained: a range of

198

policy choices is available to the administration. To put it another way, defense spending was

bound to decrease following the end of the Cold War no matter who was the president, but by

how much, and where American intervention in the international system would happen next, is

very much within the purview of presidential choice. Likewise, defense spending was bound to

increase in the wake of September 11th. And a war in Afghanistan against the Taliban probably

would have happened had there been a President Gore in the White House. But the war in Iraq,

and the foreign policy philosophy which demanded it, may not have been. This suggests an

interactive effect between the party of the president (a proxy for foreign policy philosophy given

the differing positions of the parties on military intervention and how the parties are perceived

on defense by the mass public) and foreign policy events.

TABLE 7.2: DEVIATION MODELS REGRESSING DEFENSE SPENDING LEVELS (%GDP) ON DEF. SPENDING PUBLIC OPINION. Parameter MODEL: Intercept Estimate 2 = + ( % ) + (S.E.) (S.E.) R N

DEFENSE࢖࢕ SPENDING૙ ૚ (MEAN)ࡳࡰࡼ 0.000 -0.478 *** .229 29 ࡰࡿ ࡮ ࡮ ࡰࡿ ࢋ (0.166) (0.169)

DEFENSE SPENDING (KURTOSIS) 0.000 0.859 *** .737 29 (0.859) (0.099)

* significant at .10 level ** significant at .05 level ***significant at .01 level Table 7.2 reports a simple regression with the z-scores for actual defense spending (%

GDP) as the predictor of the z-score for the mean position on defense spending in the general

public as well as the z-score for bimodality in defense spending opinion levels.

Equation 7.3: Simple Regression of Actual Defense Spending on D.S. Public Opinion

(% ) ) +

ܦܵ௣௢ ൌܤ଴൅ܤଵሺܦܵ ீ஽௉ ݁

199

For the full time series, the public is almost perfectly middle-of-the-road. The mean public opinion on defense spending levels for the full time series (mean of the mean) is 4.074, reflecting a slight preference for more defense spending over the time span. But as can be seen in Figure 7.11, this average masks substantively important variance in the average public position on defense spending. During the 1980’s during the Reagan administration and at the height of the Cold War, the preference on defense spending is below the mean, indicating a general preference for less spending (relative to the defense spending that was going on).

During the 1990’s, the public preferred more defense spending, on average, than was the case with the Clinton administration’s deep defense spending cuts from 1994 through the end of his presidency. And while the preference for more defense spending spiked after September 11th, as the war in Iraq drug on that average preference for defense spending regressed towards the mean defense spending preference level (Figure 7.11).

The coefficient for actual defense spending in Table 7.2 suggests this inverse relationship is significant and substantively important. For a single unit increase in actual defense spending, preference for defense spending in the public declines -0.478 Z-score units.

However, this model has a relatively weak fit to the data, with only about 30% of the variance in average defense spending public opinion explained by actual defense spending levels (% GDP).

There is a strong central tendency in American public opinion on defense spending. Republican administrations appear to have endorsed defense spending levels in excess of that preferred by the public, while Democratic administrations have implemented deeper defense spending cuts above and beyond what is preferred by the public. The relationship between actual defense spending and bimodality in public opinion on defense spending is positive and much stronger than that of defense spending and public opinion on defense spending. For every single unit increase in actual defense spending, there is a 0.859 unit increase in defense spending opinion

200

polarization (kurtosis), and this model explains over seventy percent of the variance in defense

spending public opinion bimodality.

To test all three of the theories on defense spending public opinion, I create dummy variables

for the partisanship of the presidential administration and the foreign policy period. Table 7.3

reports the structure of the dummy variables. The presidential administration partisanship

dummy is a simple binary categorical variable with the value of one for a Republican president

and zero for a Democratic president. The three level variable for foreign policy period gives a -1

for the New World Order period of foreign policy extending from the fall of the Berlin Wall in

1989 to the September 11th attacks in 2001. The Cold War period (value = 0) extends for all years prior to the fall of the Berlin Wall and the War on Terror period (value = 1) extends from the September 11th attacks through to the present. Table 7.4 adds the measure for presidential administration partisanship to the actual defense spending model, examining the variance in defense spending public opinion in a multivariate analysis.

Equation 7.4: Multiple Regression - Defense Spending Public Opinion Model

( ) +

ݐݕ௣௔൯ ൅݁ݎଶ൫݌ܽܤ Ψீ஽௉ ൯ ܵܦଵ൫ܤ଴൅ܤ௣௢ ൌܵܦ TABLE 7.3: CODING SCHEME FOR FOREIGN POLICY PERIOD & PRESIDENTIAL PARTY VARIABLE CATEGORIES VALUE PARTY OF PRESIDENTIAL ADMINISTRATION

REPUBLICAN 1

DEMOCRAT 0

FOREIGN POLICY PERIOD

COLD WAR 0

NEW WORLD ORDER -1

WAR ON TERROR 1

201

The model on average public opinion on defense spending levels is greatly improved by the

inclusion of presidential administration party in the regression. There is a fifty percent

improvement in the overall fit of the model by including party as a predictor variable, as the

multivariate model explains close to half of the variance in average defense spending opinion (R2

= .467). While actual defense spending has an independent, significant affect on average defense spending public opinion, the multivariate analysis makes it clear that public preference on defense spending levels is highly dependent upon what party controls the White House. This could be a function of opposition party portrayal of the president as more extreme on defense spending than he actually is (I address partisan perceptions of the opposition in Chapter 7), or it could be due to institutional and partisan constraints on presidential policymaking. Partisans have incentives to portray an opposition president as more extreme than their actual policy positions may warrant. And there are significant institutional roadblocks in the American system of separated powers that can limit the extent to which a president can move the status quo. As reported by Vin Weber, President Clinton had promised “dramatic” defense spending cuts as he took office in 1992 (Weber 1993). However, the deep cuts in Reagan/Bush spending didn’t

TABLE 7.4: D. M. REGRESSING DEFENSE SPENDING & PARTY OF PRESIDENTIAL ADMIN ON D.S. PUBLIC OPINION.

MODEL: Intercep DEFENS = + ( ) + ( ) t E SPEND PARTY R2 N + ࡰࡿ࢖࢕ ࡮૙ ࡮૚ ࡰࡿΨࡳࡰࡼ ࡮૛ ࢖ࢇ࢚࢘࢟࢖ࢇ DEFENSE SPENDINGࢋ (MEAN) -1.029 -0.810 *** 1.356 *** .46 29 (0.333) (0.173) (0.398 7 ) DEFENSE SPENDING (KURTOSIS) 0.435 0.999 *** -0.573 *** .78 29 (0.213) (0.111) (0.256 0 ) * significant at .10 level ** significant at .05 level ***significant at .01 level

202

materialize in his first budget. Defense Secretary Lee Aspin declared, “"What we're doing is kind of treading water." Why did Clinton adopt what was described as a ‘holding budget’ rather than institute the steep cuts he preferred: a $122 billion dollar cut over four years—twice what he had proposed on the campaign? Institutional resistance, as Colin Powell referred to the budget as “fundamentally flawed” as well as resistance from centrists within his own party, such as Senator Sam Nunn (Weber 1993). Public perception of presidential defense spending may be influenced by promises and proposals that, through the course of the policy process, become watered-down significantly in comparison to the original intentions.

The bimodality model does not show as dramatic an improvement vis-à-vis the mean model with the inclusion of party as an independent variable. However, the party of the presidential administration is a significant predictor of defense spending. The sign of the coefficient is negative, indicating that there is less conflict for Republican administrations.

However, this is due largely to the first Bush administration, which had low levels of conflict despite relatively high levels of defense spending. Still, it suggests that when you control for actual levels of spending the influence of partisanship is not in the expected direction. Yet, as

I’ve noted, presidents influence actual defense spending levels. There may be an interactive relationship between these two variables that needs to be accounted for.

Table 7.5 adds the categorical variable for foreign policy period to both models and reports coefficients, standard errors and model-fit statistics. Thus we now have a model with variables representing all three theories explaining defense spending public opinion competing in the same multivariate model:

Equation 7.5: Multiple Regression - Defense Spending Public Opinion Model

( ) + + ( )

௣௢ ଴ ଵ Ψீ஽௉ ଶ ௣௔ ଷ ൅݁ ܲܲܨ ܤ ݐݕ ൯ݎ൫݌ܽ ܤ ൯ ܵܦ൫ ܤ൅ ܤൌ ܵܦ

203

TABLE 7.5: D. M. REGRESSING DEFENSE SPENDING, PARTY & FOREIGN POLICY PERIOD ON D.S. PUBLIC OPINION. MODEL: Parameter Standard = + ( % ) + + ( ) Estimate Error + ࡰࡿ࢖࢕ ࡮૙ ࡮૚ ࡰࡿ ࡳࡰࡼ ࡮૛൫࢖ࢇ࢚࢘࢟࢖ࢇ൯ ࡮૜ ࡲࡼࡼ DEFENSE SPENDING (MEAN) ࢋ Intercept -0.357 0.320

Actual Defense Spending -0.826 *** 0.140

Party of Presidential Administration 0.626 * 0.372

Foreign Policy Period 0.687 *** 0.178

PR > |F| <.0001 R2 .666 N 29 DEFENSE SPENDING (KURTOSIS) Intercept 0.736 *** 0.235

Actual Defense Spending 0.992 *** 0.103

Party of Presidential Administration -0.899 *** 0.274

Foreign Policy Period 0.307 ** 0.131

PR > |F| <.0001 R2 .820 N 29 * significant at .10 level ** significant at .05 level ***significant at .01 level

For both the mean and kurtosis models, the foreign policy period is a significant

predictor of defense spending opinion. In the mean model, the positive coefficient indicates that

the American public prefers higher levels of defense spending in times of war, such as the Cold

War period and the War on Terror. Again we see a significant improvement in the predictive

capacity of the multivariate model, with just over sixty-six percent of the variance in average

defense spending opinion explained. In the bimodality model, inclusion of foreign policy period

improves the overall fit of the model to .820, meaning that over eighty percent of the variance

in defense spending bimodality is explained by actual defense spending, the party of the

204

president, and the exogenous event-determined foreign policy period. Given that the war

periods have higher values than the New World Order period, it isn’t surprising to see that, as

foreign policy period increases, so does the polarization in defense spending opinion. What is

interesting is that all three variables have independent significant influence on defense spending

opinion, controlling for the other variables in the model.

Table 7.6 reports the full models for both average defense spending opinion and the

bimodality of defense spending with interactions.54

Equation 7.6: Multiple Regression - Defense Spending Public Opinion Model w/ Interactions

( ) + + ( ) ( )

ிி௉כ௉௉஺ ସ ஺஽ௌכ௣௢ ଴ ଵ Ψீ஽௉ ଶ ௣௔ ଷ ஺஽ௌ ݐ݊݅ ܤݐ ൅݊݅ ܤ ݐݕ ൯ݎ൫݌ܽ ܤ ൯ + ( ܵܦ൫ ܤ൅) ܤൌ+ ܵܦ

௉௉஺כହ ி௉௉ ݐ defense spending݁ level continues to have a significant, negative݊݅ ܤIn the mean model, actual influence on defense spending public opinion (-1.283). Even controlling for the other theoretical explanations and interactions, the public’s opinion on what the level of defense spending should be is responsive to defense spending levels. The negative coefficient indicates an inverse relationship between actual defense spending and defense spending opinion. As defense spending increases (becoming more extreme relative to the mean) for the time period, the public wants less spending on defense. As defense spending decreases (becoming more extreme relative to the mean), the public wants more spending. Partisanship (of the presidential administration) has a substantial and significant effect on average defense spending opinion independent of the other included variables (2.197). Beyond the exogenous shocks to the system and the actual defense spending levels, the party of the president is a significant factor.

54 The primitive term for Foreign Policy Period is excluded as including it creates perfect colinearity between the independent variables.

205

TABLE 7.6: REGRESSING DEFENSE SPENDING, PARTY & FPP ON D.S. PUBLIC OPINION WITH INTERACTIONS

MODEL: Parameter Standard = + ( ) + ( ) Estimate Error + ( ) + ࡲࡲࡼכࡼࡼ࡭ ࡮૝ ࢏࢔࢚࡭ࡰࡿכ࡮૜ ࢏࢔࢚࡭ࡰࡿ ڮ ࡰࡿ࢖࢕ ࡮૙ DEFENSE SPENDING (MEAN) ࡼࡼ࡭ ࢋכ࡮૞ ࢏࢔࢚ࡲࡼࡼ Intercept -2.014 *** 0.489

Actual Defense Spending -1.283 *** 0.506

Party of Presidential Administration 2.197 *** 0.504

Foreign Policy Period ------

ADS*PPA 0.606 0.508

ADS*FFP 0.533 ** 0.217

FPP*PPA 0.850 *** 0.167

PR > |F| < .0001 R2 .775 N 29 DEFENSE SPENDING (KURTOSIS) Intercept 0.270 0.397

Actual Defense Spending 1.218 *** 0.410

Party of Presidential Administration -0.484 0.409

Foreign Policy Period ------

ADS*PPA -0.158 0.412

ADS*FFP 0.387 ** 0.176

FPP*PPA 0.427 *** 0.135

PR > |F| <.0001 R2 .852 N 29 * significant at .10 level ** significant at .05 level ***significant at .01 level

206

Specifically, citizens prefer higher defense spending during Republican administration years

relative to Democratic administration years.

Confirming earlier speculation, there are significant interaction effects between defense

spending levels and the foreign policy period. Naturally events like the Cuban missile crisis, the

Soviet invasion of Afghanistan, and the terrorist attacks on September 11th, can precipitate increases in defense spending in and of themselves. Likewise, liberalization in the USSR

(Perestroika), the fall of the Berlin Wall, and the collapse of the Soviet Block result in demilitarization. The combined effects of changes in actual defense spending and foreign policy events—that serves either to make their world look safer or more dangerous—influence public opinion on defense spending levels. While the overall effect of defense spending remains negative (even with the interaction), foreign policy events tends to mitigate this inverse relationship. There is also a significant interaction between the foreign policy periods and the party of the president. The positive coefficient indicates that times of war combined with

Republican administrations tend to produce higher preferred defense spending levels than otherwise.

Turning to the model of bimodality, levels of defense spending as a percentage of GDP tend to produce more polarization, independent of the other variables and interactions. The strong, positive coefficient (1.218) indicates that there is more bimodality in defense spending opinion coincident with higher levels of defense spending. Defense spending levels and the party of the president has an independent effect on defense spending opinion. The positive coefficient (0.387) suggests that it matters what party the president is when defense spending increases. Republicans tend to provoke more polarization when they make changes to defense spending than Democrats. Interestingly, when accounting for the interactive effect of the

207

foreign policy period and the party of the president, partisanship is not a significant predictor of

defense spending opinion. This suggests that the public is not reacting to party alone, but rather

to the party of the president in the context of exogenous events. Bush’s partisanship may have

engendered more conflict in defense spending than a Gore presidency would have engendered,

but September 11th, and the foreign policy decisions that resulted because of it, were necessary

factors. In a world in which September 11th doesn’t happen and Bush continued the minimalist foreign intervention policy he had signaled in the campaign, it is likely that the divisions in society over defense spending wouldn’t have become as polarized as they did in response to the wars in Iraq and Afghanistan.

There is significant and substantively important polarization of defense policy opinion, both in terms of the average public-preferred level of defense spending and the conflict over defense spending (kurtosis). The poorly performing linear trend model on bimodality for

FIGURE 7.12: BIMODALITY TRENDS IN DISTRIBUTIONS OF MASS OPINION ON SOCIAL ISSUES, 1972-2008 3.6 3.4 3.2 3 2.8 2.6 2.4 2.2 2 1.8 1.6 1.4 1.2 1 Year of the 0.8 0.6 0.4 Woman 0.2 0 -0.2 -0.4 -0.6 -0.8 -1 -1.2 -1.4 -1.6 -1.8 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

ABORTION KURTOSIS WOMEN KURTOSIS AID TO BLACKS KURTOSIS

208

defense spending opinion masked an important relationship between defense spending

opinion,actual levels of defense spending, partisanship, and the defense-related events of the

past three decades. Clearly the partisanship of the presidential administration, the primary actor

in American foreign policy, affects the preferred defense spending levels of the American public.

Partisanship influences the preferred levels of defense spending, and presidents are the most

significant players in determining the defense budget. But this effect is mediated by foreign

policy events as well as institutional and political constraints. There is an inverserelationship

between defense spending and defense spending opinion. As defense spending becomes

extreme relative to the average defense spending level, the preferences of the American public

on defense spending moves in the opposite direction. Contrary to some cynics who believe the

public is unresponsive to the actions of their government and budgetary decisions, the conflict

over defense spending and the average opinion on defense spending levels indicates that the

public is paying attention and adjusting their views dependent upon changes in philosophy,

policy and spending.

SOCIAL ISSUE TRENDS: POLARIZED ON MINORITIES, POLARIZING ON ABORTION, DEPOLARIZED ON WOMEN

Figure 7.12 illustrates the bimodality trends in the three ANES social issues. The largest substantive change in bimodality for any of the issues considered in this or other chapters is readily apparent in the kurtosis change on women’s equality. It reflects a profound shift in opinion of the American public on the right of women to fully participate in all the rights and responsibilities that men have traditionally enjoyed, and the consequent implication that government should, at minimum, be gender neutral in its treatment of women and men if not actively working to provide opportunities for women. In the 1970’s the polarization of society on women’s rights was relatively indistinguishable from that on aid to Blacks, with an average kurtosis for female equality sitting below -1 for the decade, just as was the case for aid to Blacks.

209

However, the two trends break off in the 1980’s with aid to Blacks remaining playtkurtic while

opinion on women’s equal role in society takes an extreme leptokurtic turn. In 1992, labeled by

many in the media as “The Year of the Woman” given the number of successful Senate seat bids

made by women that year including Patty Murray of Washington state and Diane Feinstein in

California, the kurtosis of opinion on women in society crosses the ‘normal’ distribution

threshold. From that point on, there is a steady and steep incline in the trend towards

unimodality (consensus) in opinion on female equality public opinion. By 2008, opinion on this

social issue is by far the most consensual issue of all the issues analyzed here.

Abortion opinion has displayed a significant polarizing trend in both the dispersion and

bimodality of public opinion on the social issue. While abortion was already a bimodal

distribution in 1972 (-1.077), abortion opinion has become more bimodal over the past thirty

years. Note that the Bush administration evidences a consistently increasing conflict over

abortion as measured by kurtosis. Perhaps Bush’s overt Christianity (having famously cited Jesus

as the philosopher he admired the most

during the 2000 campaign for the "[If] the right of privacy means anything, it is the right of the Republican presidential nomination as individual, married or single, to well as making his conversion story a part be free from unwarranted governmental intrusion into of his public persona) contributed to the matters so fundamentally affecting a person as the further playtkurtic distribution of opinion decision whether to bear or beget a child." on abortion during his presidency. Of

-Eisenstadt v. Baird, 1972 course there has been a steady, near-

monotonic loss of consensus on abortion

since Roe v. Wade was decided by the USSC. Two blips in abortion opinion polarization are coincident with major decisions by the Supreme Court. There was a brief positive uptick in

210

kurtosis following the Supreme Court decision in Webster v. Reproductive Health Services, where

the Court upheld a Missouri law imposing restrictions on state funds for abortions and

permitting regulation of abortion by states in the first trimester (Webster v. Reproductive Health

Services 1989). This case distinguished the Roe precedent, and was a step away from the pro- abortion stance articulated in Roe. This change in public opinion on abortion was followed by a bimodal blip coincident with the case in Planned Parenthood of Southeastern Pennsylvania v.

Casey. Though many Court observers had speculated that Roe v. Wade might be overturned in

1992, in fact the Court reaffirmed Roe on stare decisis grounds, though it dispensed with the artificial trimester requirements in favor of a “viability” standard (Planned Parenthood of

Southeastern Pennsylvania v. Casey 1992). With those two exceptions, opinion on abortion has evidenced a very consistent trend towards greater conflict and bimodality—i.e. greater polarization.

Women in Society: Depolarization, “The Year of the Woman,” & the Separability of Abortion

While the shift in opinion on women’s equal rights is remarkable, a more subtle but just as interesting aspect of this trend is its separability from the abortion issue. Proponents of

Second-Wave feminism and the women’s rights groups that have organized to lobby on issues related to it, such as the National Organization for Women (NOW), have linked the equal rights of women and abortion (reproductive rights) for some time. This conceptualization of the abortion decision as integral to equal rights for women was adopted by the Supreme Court in the land mark case of Roe v. Wade, where the court applied the right to privacy discovered in

Griswold v. Connecticut to the abortion decision. As Justice Blackmun wrote in his majority opinion, “We, therefore, conclude that the right of personal privacy includes the abortion decision…” (Roe v. Wade 1973). However, there is little evidence that the general public has adopted this viewpoint. Indeed, it appears that public opinion on equal rights for women is

211

distinct separable from that of abortion both in terms of the global status of the mean opinion

on the two social issues, the dispersion of opinion on the two issues, and the bimodality of

opinion on them. Furthermore, they display distinct and opposite trends across the two primary polarization measures. While opinion on equal rights for women has become very leptokurtic, abortion has evidenced a significant playtkurtic trend (Figure 7.13). While the dispersion of abortion has remained relatively steady over the period (declining somewhat in the 1980’s but recovering to previous levels in the 2000’s), there has been a steady decline over the past three decades in the equality for women ANES item (Figure 7.14). Only in the mean is there a trend that suggests a connection between abortion and equal rights, with the average opinion on abortion becoming more “Pro-Choice” at the same time that there is a significant increase in support for an equal role for women in society relative to men.

FIGURE 7.13: BIMODALITY TREND IN MASS OPINION ON ABORTION, 1972-2008 0 -0.2

-0.4 Bush Admin Roe v. Wade -0.6 Planned Parenthood v. Casey -0.8 Webster v. Reproductive Health Services -1 -1.2 -1.4 -1.6 -1.8 -2

ABORTION KURTOSIS

212

FIGURE 7.14: DISPERSION TREND IN MASS OPINION ON EQUALITY FOR WOMEN, 1972-2008 3

ERA 2.5 Ratification Fails 2

1.5

Year of the 1 Woman

0.5

0 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

WOMEN S.D.

However, the average position on abortion remains much closer to the center of the distribution than the equality for women average placement. The average opinion on equality for women has moved significantly towards the extremes of the distribution, with an average of 1.833 on a

7 point scale in favor of an equal role in 2008. The Abortion trend does not display a shift in the same ballpark as that of the equality for women item.

An illustrative example of the separability of abortion and equal rights for women is the campaign to ratify the Equal Rights Amendment to the constitution. When Americans were convinced the ERA was about providing equal rights for women, it enjoyed strong support. For example, in 1975, respondents to a Gallup poll were asked, whether they favored the ERA which gave women “equal rights and equal responsibilities.” Respondents supported the measure 58

213

to 2455. However, when language was included suggesting that it could remove protections in the law specifically for women, such as the Harris survey conducted the same year, the issue was significantly more conflictual (45 against, 40 support).56 Opponents of the ERA, such as

Phyllis Schlafly, were able to successfully identify the act with the feminist movement (i.e.

abortion and advocacy of non-traditional roles for women). A survey in 1979 by the National

Federation of Business & Professional Women asked respondents whether ERA advocates were

“women’s libbers” seeking to change traditional roles of women.57 There was an even split of

respondents on the question, with 44% agreeing and 51% in disagreement.

The ERA was first proposed in the 1940’s. It passed the Congress and was ratified by

most states—well into the 1970’s there was significant momentum in favor of ratification. The

rate of ratification was particularly rapid in 1972 and 1973. Indeed, all but three states had

ratified the ERA as of the end of the 1970’s. But the tide had turned as of the late 1970’s. In

1980 the Republican Party removed it from their platform. The ERA, in the wake of Roe v. Wade, was increasingly identified with the abortion-supporting feminist groups such as NOW and was attacked for costing women on the legal front. By 1982, the official end-date of the extended ratification period (extended by Congress in 1978), five states who had passed ratification measures had rescinded their ratification votes: Idaho, Kentucky, Nebraska, Tennessee, and

South Dakota. The ratification movement was stopped short and, despite efforts to circumvent the rescissions and ratification deadline, the ERA was never ratified. While the ERA was strictly

55 Gallup poll conducted by the Gallup Organization. March, 1976. N = 1,582. Question: “Do you favor or oppose this (Equal Rights) Amendment (which would give women equal rights and equal responsibilities?” 56 Harris Survey conducted by Louis Harris & Associates. April, 1975. N = 1,568. Question: “Do you tend to agree or disagree that the Equal Rights Amendment should be opposed because it would wipe out many of the laws that have given women special protection for many years? 57 Survey by National Federation of Business & Professional Women conducted by Louis Harris & Associates. November, 1979. N = 1,500. Question: “I'm going to read you some statements about various groups that oppose and support the ERA (Equal Rights Amendment). For each, please tell me whether you tend to agree or disagree.)... Advocates of the Equal Rights Amendment are mainly women's libbers who would totally change the traditional role of women.”

214

about equal rights, it enjoyed consensus support. When it was associated with issues such as

abortion, opinion on the ERA became polarized and killed the opportunity for attaining the

super-majority threshold required for a constitutional amendment.

THE RISE OF CONFLICT OVER MINORITY RIGHTS: THE POLARIZING EFFECT OF AFFIRMATIVE ACTION

Public opinion on aid to Blacks hasn’t changed a great deal since the 1970’s. While there has been a significant shift away from support of government involvement in helping Blacks overcome past discrimination, the polarization of t he distribution of opinion on aid to Blacks has varied little over the time period. Furthermore, the global trend for the time series masks some important developments in polarization of opinion on aid to Blacks. While the overall trend in bimodality is leptokurtic for aid to blacks (Figure 7.12), note that since 2000 there has been a steady negative trend in the kurtosis of public opinion on this issue, with the most recent

FIGURE 7.15: DISPERSION TREND IN MASS OPINION ON AID TO BLACKS, 1970-2008 3 CA Prop. 209 2.5 Passes 2

1.5 U. Michigan 1 Cases

0.5

0 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004 2006 2008

BLACKS S.D.

215

distribution of opinion on aid to blacks just as bimodal as it was in 1972 (Figure 7.12). Though

bimodality declined in the 1980’s the overall opinion on aid to blacks (rather than the trend) is

distinctly bimodal for the breadth of the time series. The least conflictual kurtosis score for aid

to Blacks is the -0.40 in 1980, well below the score for a normal distribution (zero) and far from

unimodal consensus. As can be seen in Figure 7.13, a similar trend is apparent in the dispersion

of opinion on government aid to Blacks. Dispersion in opinion on aid to Blacks falls in the 1980’s,

and then oscillates between 2 and 1.5 standard deviations until 2000, when it begins a steady

trend upwards. Just as with the bimodality measure, the dispersion of aid to Blacks in 2000 is

the same as it was in 1972.

One possible explanation for the resurgence of polarization in public opinion on aid to

Blacks is the continued political relevance and salience of a specific aid to Blacks policy:

affirmative action. In 1978 there was strong support among Blacks and Whites for affirmative

action programs. In 1978, a survey commissioned by the National Conference of Christians &

Jews asked respondents whether t hey supported affirmative action policies (as long as there

were no rigid quotas. Figure 7.16 reports the breakdown of the response for blacks and

whites.58 As can be seen, overwhelming (near consensual) support existed for non-quote affirmative action programs. Thirty years later, however, affirmative action is a much more controversial public policy. In Figure 7.17 the results of a 2005 Gallup poll reflects a much more divided public on the necessity of affirmative action.59 Support for affirmative action has

declined among Blacks and Whites since the 1970’s. The level of support for affirmative action

among Blacks in 2005 is equivalent to the level of support for affirmative action among Whites

58 Racial & Religious Minorities & Women Poll, conducted by Louis Harris & Associates. October, 1978. N = 2456. Sample Breakdown: 1673 Whites, 783 Blacks. Question: “All in all, do you favor or oppose affirmative action programs in industry for ...blacks... provided there are no rigid quotas?” 59 Gallup Poll, conducted by the Gallup Organization. June 6-25, 2005. N = 2,264. Sample Breakdown: 807 Whites, 802 Blacks. Question: “Do you favor or oppose affirmative action programs for racial minorities?”

216

in 1978. While AA programs still have supermajority support among Black respondents, a majority of Whites now oppose affirmative action. Indeed, this has become a considerably controversial (polarized) public policy issue between 1978 and the present for White voters. For the full sample, 50% of Americans support affirmative action while 42% oppose it. Over the course of the time series and into the 2000’s, affirmative action public opinion shifts to state very close to the cut-point for pure conflict.

We see this emerging conflict on t he policy stage in the form of the state-by-state campaign to enact “Civil Rights Initiatives” designed to enshrine the outlawing of affirmative action plans in state constitutions. This movement was originally founded by political activist and former Regent of the University of California, Ward Connerly. Connerly formed the

California Civil Rights Initiative Campaign, which worked to get Proposition 209 onto the 1996 state ballot for constitutional amendments in the California state elections that year. The language of Prop. 209 included: “The state shall not discriminate against, or grant preferential treatment to, any individual or group on the basis of race, sex, color, ethnicity, or national origin in the operation of public employment, public education, or public contracting.” While couched in the language of anti-discrimination laws, the law would have the effect of preventing governments from developing and initiating affirmative action programs. While supporters viewed this as a next step in establishing equality for all before the law, critics argued it rolled back the civil rights victories of the 1960’s and the 1970’s. Connerly and his supporters have taken the campaign beyond California, successfully getting a similar ballot resolution passed in

Michigan in 2006 (Hadley 2005).

However, the more recent developments in affirmative action politics haven’t been uniform victories for opponents of AA. In the companion cases of Gratz v. Bollinger and Grutter

217

FIGURE 7.16: PUBLIC OPINION ON AFFIRMATIVE ACTION BY RACE, 1978 100 90 80 70 60 50 SUPPORT 40 OPPOSE 30 20 10 0 WHITE BLACK

FIGURE 7.17: PUBLIC OPINION ON AFFIRMATIVE ACTION BY RACE, 2005 100 90 80 70 60 50 SUPPORT 40 OPPOSE 30 20 10 0 WHITE BLACK

218

v. Bollinger (known collectively as the University of Michigan Cases), while ruling specific provisions of the university’s affirmative action admissions program unconstitutional, the

Supreme Court refused to rule affirmative action itself unconstitutional. Justice O’Conner, writing for the majority in Grutter, held that the U.S. Constitution “does not prohibit the law school's narrowly tailored use of race in admissions decisions to further a compelling interest in obtaining the educational benefits that flow from a diverse student body.” These rulings left affirmative action ensconced firmly within the salient policy debate and it remains an issue upon which opinion in the United States is polarized. The fact that debate over affirmative action is current, lively, bimodal, and conflictual suggests that it is at least part of what has driven the dispersion and bimodality of opinion on Aid to Blacks to 1970’s levels. While the previous levels were almost certainly the product of former segregationists and racial animus, the current conflict over race appears to be of a different character. Rather the current question on which the American public is polarized is whether the pendulum has swung too far the other direction in regards to government efforts to ameliorate past discrimination and provide a ‘leg up’ to

Blacks in jobs, admissions to schools, and other areas of society where Blacks were previously excluded.

CONCLUSION

In this chapter I examined a variety of measures of the primary political issue dimensions in which political conflict exists in the United States. I found increasing polarization on ideology, social issues, defense policy, and government spending. The polarization of public opinion along the issue dimension was apparent in the detailed analysis of public opinion variance on the ANES issue scales. For all but one of the issue dimensions considered, there is evidence of consistent polarization over the time series. Even among issues in which there are

219

global depolarization trends (aid to Blacks), the average kurtosis for these distributions is playtkurtic, indicating public opinion is relatively split into two opposing camps.

The trends in political polarization over the past thirty years reveal comprehensible attitude adjustments to political events, the policies implemented by the parties in control of the branches of the federal government, and to exogenous shocks to the system from stock market crashes to terrorist attacks. On some issues, such as the equality of women in society, there was substantial evidence of depolarization. On all measures of political polarization: mean trend, dispersion, and bimodality, opinion on gender equality has become increasingly consensual. This consensus has emerged in support of equality between men and women. While average public opinion on government jobs had no apparent trend in polarization, opinion on economic issues are fairly dispersed and the distribution is bimodal.

On other issue dimensions, there was strong evidence of polarization. On ideology, a proxy for opinion across the range of political issues, there was evidence of increasing bimodality in its distribution. The American public has shifted towards the ideological poles. On government spending, there was evidence that opinion had become more bimodal over the time series. However, the trends in opinion on government spending did not track well with actual levels of government spending, indicating that what drives opinion on government spending is at least somewhat independent of the actual level of government spending. Defense spending public opinion, on the other hand, while evidencing little polarization in the global model, tracked strongly with actual levels of defense spending. The investigation into defense spending polarization revealed that, along with actual levels of defense spending, partisanship

(the party of the presidential administration) and exogenous foreign policy events interacted with the trends in actual levels of defense spending to condition and influence public opinion on

220

defense spending. One interesting relationship reveled in the analysis of defense spending trends is the inverse relationship between actual defense spending and public opinion on the level of defense spending. When defense spending becomes extreme—either due to extreme increases in defense spending or extreme cuts in defense spending (relative to the mean level of defense spending)—the American public trends in the opposite direction.

On the social issue dimension, there is the previously mentioned significant depolarization of opinion on general equality. However, there was substantial evidence of polarization in the social issue dimension as well. On abortion, the American public has steadily become more bimodal and more dispersed. Indeed, the fact that abortion opinion has remained playtkurtic (indeed, evidencing a significant trend towards bimodality), despite the remarkable unimodal trend on women’s equality reveals an important story about American politics, namely the failure of the feminist movement to successfully associate abortion with women’s rights in the eyes of the public. And finally, while there was a significant global trend of depolarization in the dispersion and kurtosis on government aid to blacks, it is apparent that this trend has reversed itself in more recent years. As the American public has become less supportive of ameliorative policies for minorities such as affirmative action, the opinion on aid to Blacks has become more conflictual, i.e. more polarized. This chapter presents strong and consistent evidence of political polarization, not just on the culture wars, but on the gamut of political issue dimensions in American politics. While several issue dimensions failed to exhibit polarization trends (i.e. economic issues), most of the issues—whether they were social, economic, or defense-related—exhibited dispersed and bimodal opinion distributions even at the beginning of the time series. There was significant social issue polarization on the abortion issue as well as increasing polarization late in the time-series on aid to blacks. The only issue that exhibited consistent (and strong) tendencies towards consensus was in public attitudes towards the equal

221

role of women. Americans may increasingly agree that women should have a stature in society on par with women but they agree on little else. And, furthermore, they are increasingly conflictual on social issues and foreign policy.

222

CHAPTER 8: POLARIZATION IN ISSUE DIMENSIONS OF PARTISAN AFFECT AND PERCEIVED NATIONAL PROBLEMS

Public sentiment is everything. With public sentiment nothing can fail; without it nothing can succeed. He who molds public sentiment goes deeper than he who enacts statutes or decisions possible or impossible to execute. – Abraham Lincoln

ISSUE SALIENCE AND THE CULTURE WARS

As Larry Bartels once observed, different people care about different issues (Niemi and

Bartels 1985). Issue salience has long been recognized as an important factor in determining the voting behavior of the electorate and evaluations of parties and candidates. As Converse pointed out in his classic study of belief systems, people generally have just a few issues that are of particular importance to them and thus only a few issues they pay much attention to even when they are attentive to politics (Converse 1964). To the degree that issues matter, those issues matter relative to how important the public perceives them to be. Shapiro found that salience was relevant in the spatial calculus of voting (Shapiro 1969). RePass found that the salience of issues condition the degree to which partisan evaluations condition electoral choice

(RePass 1971). And Rabinowitz et al. argue that “any issue singled out as personally most important plays a substantially greater role for those who so view it than it does for others”

(Rabinowitz, Prothro, and Jacoby 1982). Brody found a relationship between press coverage of issues and relates those changes to patterns in presidential approval (Brody 1991). Edwards et al. use issue salience weights to assess issue impact on presidential approval, finding that issues vary over time in their salience to the mass public and as to their impact on presidential approval. Salience is relevant in evaluating presidential performance (Edwards, Mitchell, and

Welch 1995). Some scholars have found that accounting for salience reduced the explanatory

223

power of issue-based voting models, but by and large issue salience is recognized as an important factor in political behavior.

Issue salience clearly relates to political polarization. The more relevant an issue is to current political debates and partisan competition, the more likely citizens are to polarize on the issue. The more important an issue is to the candidate and party evaluations as well as the electoral choice of the mass public, the more likely political polarization is to occur and the more likely that polarization is relevant to the policy process. As I noted in the previous chapter, polarization of issues are tied to objective conditions such as unemployment, international tensions, and racial conflict. Hence the relative importance of these issue dimensions in terms of evaluating parties and identifying national problems is an important part of the story on political polarization and public policy.

The open-ended responses to the “important national problem” and party “likes and dislikes” provide a unique opportunity to assess how the public views the parties in terms of their issue positions and how this has changed over time as well as trends in what the mass public identifies as important national problems. Unlike closed-ended items, the open-ended questions on partisan affect and important national problems do not impose a structure upon the respondent’s answers. Questions that specifically identify social issues and ask respondents their opinion may be valid and reliable measures of public attitudes on social issues, but it does not and cannot tell us how important respondents believe these issues to be nor does it tell us how those issues relate to the perceptions of the parties or perceptions of the problems themselves relative to other problems. Though rank ordering items would serve this purpose, we lack long and consistent time series that allow respondents to rank order several issues, let alone the issue dimensions themselves. The open-ended responses, however, allow respondents to spontaneously identify the one or two aspects of the parties that could be

224

classified along issue dimensions as well as the one or two issues that they view as the most important problems facing the country. Respondents are not forced onto preconceived formulations of these issues, open-ended items are free from the pitfalls associated with issue descriptions and particular question wordings, and they are not primed with a mention of that issue, which might muddy the waters in assessing salience. The most significant drawback of these items is the low response rates on open-ended questions, however this problem is somewhat mitigated by looking at the issues in broad classifications rather than the specific issue mention itself. Furthermore, the aggregate analysis here examines the full sample, and thus permits the maximum number of responses for assessing the salience and thus relevance of the issue dimensions on partisan affect and in the national problems item. While a breakdown of these responses may reveal interesting relationships, increasingly parsing the data raises the specter of false findings.

In the aggregate, this can provide us an accurate measure of the importance of particular issues relative to other issues or issue dimensions. A rank ordering of the issues that emerges from an analysis of the open-ended items is independent of what researchers believe to be relevant and important political issues. Such an analysis can also paint a picture of what issues citizens are increasingly or decreasingly associating with what they like or dislike about the Republican and/or Democratic Parties. A significant increase in the number of unsolicited issue mentions indicates greater salience for that issue dimension and is a reflection of the perceived public agenda at the level of the mass electorate. Using the measure classifying the open-ended responses on the government philosophy, social, economic, and foreign policy dimensions, we can compare shifts in the perceptions of those dimensions over the course of the time series.

225

ANALYSIS: RISING ATTENTION TO SOCIAL ISSUES IN THE MASS PUBLIC

Methods & Models

There are two important methodological issues to address in analyzing the count data from the ANES open-response set items. First is the limited number of survey years in which these questions were asked. With only 12 survey years with open-ended responses, the analysis is a small-N study. While ordinary least squares regression is robust, among the possible small- sample violations of the OLS assumptions are non-normality in the distribution of the data

(biasing standard errors and significance tests) and heteroscedasticity (non-constant error variance). In order to remediate this problem, I use robust standard errors for the open-ended models, estimating the asymptotic covariance matrix of the estimates under the hypothesis of heteroscedasticity. The standard error obtained from the asymptotic covariance matrix is less sensitive to violations of OLS assumptions such as normality, homoscedasticity, and the absence of outliers and overly influential data points (a particularly vexing problem with small samples).

With robust standard errors, we have more confidence in the reported significance tests and estimates of the standard errors in the small-N analysis.

The open-ended response items are count variables. As such, using OLS to estimate the polarization models may yield inefficient, inconsistent, and biased estimates (Long 1997). The data does not contain an undue number of zero-value observations and, given the relatively large number of counts, the distributions of the count variables should relatively normal. While the count data is relatively well distributed—suggesting OLS procedures may be appropriate—I estimate Poisson regressions for all of the count models to illustrate the robustness of the results. The Poisson regression model estimates the probability of a count determined by a

Poisson distribution. Given t he inherent problems of using a simple Poisson regression where conditional variance exceeds the conditional mean (overdispersion), I model t he counts using

226

the Negative Binomial Regression Model (NBRM). Where overdispersion is absent, the NBRM

reduces to the PRM (Long 1997). The NBRM is estimated using a maximum likelihood function:

Equation 8.1: Maximum Likelihood Function for NBRM

= ࡺ ( | )

ࡸሺࢼȁ࢟ǡࢄ ෑ ࡼ࢘ ࢟࢏ ࢞࢏ Discussion – Increasing Social Issue Salience ࢏ୀ૚

I report the univariate statistics on the open-ended issue items in Table 8.1 for the time period between 1976 and 1998. For all of the variables except the “Racial Problems” variable, the mean is several times the size of the standard deviation, suggestive of the relative normality

TABLE 8.1: UNIVARIATE STATISTICS FOR ANES OPEN RESPONSES – PARTISAN LIKES & DISLIKES & NATIONAL PROBLEMS STANDARD VARIABLE N MEAN DEVIATION MIN MAX

GOVERNMENT 12 48.917 23.693 23 103 PHILOSOPHY

SOCIAL ISSUES 12 150.167 62.279 77 253

ECONOMIC 12 129.833 32.296 87 178 ISSUES

FOREIGN 12 147.667 36.338 83 229 POLICY

PUBLIC ORDER 12 219.833 167.933 31 650

RACIAL 12 13.917 11.325 1 42 PROBLEMS

SOCIAL 12 422.417 159.027 204 707 WELFARE

227

of these count distributions. Interestingly enough, the “Racial Problems” variable has the lowest average count of all the open-ended variables. While the closed-ended analysis indicated racial issues remain a subject of significant political polarization, it doesn’t rate highly as one of the most important national problems at this time. That said, of the three national problem categories, the “Racial Problems” category is the most restrictive (the public order and social welfare categories include more kinds of issues).

The social issues category has the highest average count for the time period for all open- ended responses on partisan likes and dislikes (150.167). The largest number of mentions for all categories is on social welfare (422.417). Interestingly, the frequency of foreign policy mentions is significantly higher (147.667) than that of economic policy mentions (129.833), despite the apparent decline in the salience of foreign policy issues in the 1990’s. The lowest frequency of mentions on the partisan likes and dislikes item is on government philosophy issues. Despite the growing ideological coherence of the Democratic and Republican parties, respondents are still much more likely to mention specific issues such as the economy, abortion, or the war in Iraq than to cite ideological concerns. This is consistent with Converse’s classic finding that few citizens think in ideological terms. This result does not mean that ideology cannot serve as a proxy for an amalgam of political issues.

Figure 8.1 reports a histogram of the total open-ended responses on the partisan “likes and dislikes” item on the social issue dimension for the time series. It is immediately apparent that the number of social issue mentions for partisan affect has substantially increased since

1976. Hovering near 100 mentions through 1984, the frequency of mentions increases by 150% over the course of a decade on the social issue dimension. Note that this increase is consistent with the culture wars thesis. The low point for social issue mentions was 1980, coincident with

Ronald Reagan’s ascendency to the White House. Given the economic angst over stagflation, the

228

FIGURE 8.1: ANES OPEN-ENDED PARTISAN LIKES & DISLIKES RESPONSE TREND ON SOCIAL ISSUES, 1976-1998 300

250

200

150

100

0 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998

SOCIAL ISSUE MENTIONS

Iranian hostage crisis, and the advancement of communism by an internationally aggressive

Soviet Union, this is not a surprising result. Between 1976 and 1986, the frequency of mentionsfor social issues in the partisan affect measures tops 100 mentions only once, in 1984

(102). The steady increase of social issue mentions throughout the 1980’s is coincident with the rising political schism in American society identified by Hunter in the original culture wars thesis. While 1996 and 1998 social issue mentions fall off from the high in 1994 (253), both of those years have significantly higher frequencies of social issue mentions in comparison to any of the survey years prior to 1990. Clearly there has been a marked increase in the salience of social issues for the American public and they increasingly identify the parties— both in terms of what they like and what they do not like about the parties—based on their perceived positions on social issues.

Figure 8.2 illustrates the trend in issue mentions for the partisan “likes and dislikes” measure for all four of the issue dimensions. Government philosophy mentions have been relatively low and fairly constant over the time-period, with a minor blip upwards in the early

229

part of the 1990’s. It peaks at 100 mentions in 1994, coincident with the Republican Revolution that resulted in a turnover of both the Senate and the House of Representatives to the

Republican Party, the latter of which had not been in Republican hands since 1954. Newt

Gingrich’s congressional campaign, taking advantage of public angst over Clinton’s failed health care plan and the perception that Democrats had tilted Left and were out of touch with the

American people, was perhaps the most ideological off-year election in history. It was centered on specific, poll-tested legislative proposals labeled “The Contract with America.” It is thus not particularly surprising to see ideological and government philosophy issues gain in salience over this time period.

There is an apparent increase in economic issue mentions, but it isn’t a substantially large increase. Economic issue mentions go from about 90 mentions in 1972 to a peak of 180 mentions in 1988 (coincident with the 1987 “Black Monday” stock market crash) and settles in around 140 mentions for the rest of the time series. Foreign policy mentions peak in 1984 (230

FIGURE 8.2: RESPONSE TRENDS ON GOVERNMENT, SOCIAL, ECONOMIC ISSUES, AND FOREIGN POLICY, 1976-1998

280 260 240 220 200 180 160 140 120 100 80 60 40 20 0 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998

GOV'T PHILOSOPHY MENTIONS SOCIAL ISSUE MENTIONS ECONOMIC ISSUE MENTIONS FOREIGN POLICY MENTIONS

230

mentions), undoubtedly a consequence of a presidential election centered on the United States posture towards the Soviet Union and an explicit dove (Walter Mondale) on the Democratic ticket. As established in Chapter 6, defense and foreign policy public opinion is sensitive to foreign policy events and the policy positions of presidential administrations. The significant increase in defense spending advocated and implemented by Ronald Reagan was likely a significant factor in raising the salience of foreign policy in public opinion along with the Cold

War. Foreign policy mentions steadily decline throughout the 1990’s. In 1988, economic issues displace foreign policy mentions as the most frequently mentioned issue by respondents. With the Cold War having ended and a Democratic president reluctant to commit troops abroad, this is not a surprising result, as was apparent in the assessment of defense policy opinion in the

1990’s in Chapter 7. While there is no data on the likes and dislike item for the first decade of the 21st Century, there is little doubt that this measure would have spiked in the wake of 9-11 and the resulting wars in Iraq and Afghanistan. Foreign policy is an increasingly inviting area for partisan conflict and political polarization.

Standing out from the other dimensions is the significant polarization trend on the social issue dimension. In the 1970’s social issues were third of the four issue dimensions and lagged well behind foreign policy concerns. By the 1990’s, social issue mentions were the most frequent open-ended response and, despite the decline from the peak year of 1994, is the most frequent of the issue dimensions through the end of the time series. Social issues were more frequently mentioned even in 1992, the presidential campaign in which James Carville, a campaign advisor for Bill Clinton, famously quipped that “It’s the economy, stupid.” Apparently it was also social issues.

231

TABLE 8.2: OLS WITH RSS - TIME TRENDS IN ANES OPEN-ENDED RESPONSES – PARTISAN LIKES & DISLIKES MODEL: Trend: Parameter Polarizat Intercept Estimate ࢔ = + ( ) + ion Y/N (R.S.E.) (R.S.E.) R2 N ෍ ࡵࡹ࢏ ࡮૙ ࡮૚ ࢟ࢋࢇ࢘ ࢋ ࢏ୀ૙ GOVERNMENT y -2872.529 1.470 * .200 12 PHILOSOPHY (1587.578 (0.801) )

SOCIAL ISSUES Y - 7.154 *** .686 12 14065.000 (1.515) (3002.236 )

ECONOMIC ISSUES y -4142.311 2.150 ** .231 12 (1751.358 (0.882) )

FOREIGN POLICY N 5330.541 -2.608 ** .268 12 (2339.134 (1.176) )

TABLE 8.3: POISSON REGRESSION OF TIME TRENDS IN ANES PARTISAN LIKES & DISLIKES MODEL: Trend: Parameter Polarization Intercept Estimate PE ( | , = ࡺ ( | ) Y/N (R.S.E.) (R.S.E.) Z Pr > |Z| N ࡸ ࢼ ࡵࡹ ࢅ ෑ ࡼ࢘ ࢏࢓࢏ ࢟࢏ ࢏ ୀ૚ GOVERNMENT PHILOSOPHY Y -56.384 0.030 .038 12 (28.939) (0.015)

SOCIAL ISSUES Y -91.855 0.049 < .0001 12 (19.377) (0.010)

ECONOMIC ISSUES Y -28.137 0.017 .018 12 (14.010) (0.007)

FOREIGN POLICY N 40.192 -0.018 .027 12 (15.894) (0.008)

232

Table 8.2 reports OLS regression models with bootstrapped (robust) standard errors of

the trend in issue dimension mentions on the “likes and dislikes” partisan affect item from the

ANES.

Equation 8.2: OLS with RSS – Time Trends in Issue Mentions

࢔ ( ) +

෍ ࡵࡹ࢏ ൌ࡮૙൅࡮૚ ࢟ࢋࢇ࢘ ࢋ ࢏ୀ૙ Note there is a significant trend for all four of the political issue dimensions. Both frequency of government philosophy mentions (1.470) and economic issue mentions have been on the rise

(2.150). The spike in mentions of government philosophy in the 1990’s clearly contributes to the positive coefficient for the model. However, the model is a relatively poor fit, likely a consequence of the return to the ‘equilibrium’ government philosophy mentions in 1996.

Economic issue mentions have become more frequent as well, though the economic mentions fall off in the later part of the 1990’s as the economy was in recovery and we were on the cusp of the “dotcom” bubble. There is a significant trend for foreign policy mentions, but it is a negative trend: foreign policy mentions decline about 2.5 mentions per survey year. Again, the end of the Cold War is the best explanation for this result. Social issue mentions evidence the largest increase—7.154 mentions per survey year—tripling the coefficients of the other issue dimensions. The model-fit statistic indicates that the time series explains 68% of the variation in social issue mentions, and this is by and far the best fit model of the four issue dimensions. This is significant evidence of political polarization on the social issue dimension.

Table 8.3 reports the NBRM regression model with bootstrapped (robust) standard errors for the four issue dimensions in the partisan likes and dislikes item. The maximum likelihood model for issue mentions is reported in Equation 8.3. The ML estimate is the parameter that maximizes the log of the likelihood for the linear model ( ) for the frequency

233 ߙǡ ߚ

counts for issue mentions, where the ML estimate for survey year is the parameter that makes

the observed data most likely.

Equation 8.3: OLS with RSS – Time Trends in Issue Mentions

= ࡺ ( | )

ࡸሺࢼȁࡵࡹǡ ࢅ ෑ ࡼ࢘ ࢏࢓࢏ ࢟࢏ ࢏ୀ૚ where: imi = frequency of issue mentions in a survey year yi = survey year

The models in Table 8.3 confirm the robustness of the OLS-estimated models for partisan likes and dislikes reported in Table 8.2. The parameter estimate that maximizes the likelihood of observing the increase in issue mentions is largest for social issues, consistent with the OLS models. Furthermore, while all of the models are significant at the .05 level, only the model for social issues is significant at the .01 level (Z Pr > |Z| < .0001). The ML model coefficients for all of the issue dimensions are consistent with the betas reported for the OLS models. The ML models for social issues, economic issues, and government philosophy have positive coefficients, while the ML model for foreign policy has a negative coefficient.

Table 8.4 reports on trends in the other open-ended item asked consistently over the

ANES time series: the “most important problem” question using OLS estimates with bootstrapped (robust) standard errors60. Survey respondents were asked to respond with what they believed to be the most important problems facing our nation at the time. The Public Order category for the open-ended “most important national problem” item includes mentions that near-uniformly fall within the social issue dimension. The social welfare category, while encapsulating a variety of economic issues, likely has some overlap with issues related to government philosophy and the social dimension. The ‘Public Order’ category includes most of

60 See Equation 5.14 for the OLS regression equation for these open-ended response models.

234

TABLE 8.4: OLS WITH RSS - TIME TRENDS IN ANES OPEN-ENDED RESPONSES - IMPORTANT NATIONAL PROBLEMS MODEL: Trend: Parameter Polarization Intercept Estimate ࢔ = + ( ) + Y/N (R.S.E.) (R.S.E.) R2 N ෍ ࡵࡹ࢏ ࡮૙ ࡮૚ ࢟ࢋࢇ࢘ ࢋ ࢏ୀ૙ PUBLIC ORDER y -23214.000 11.794 ** .257 12 (10214.000) (5.152)

RACIAL PROBLEMS Y -1844.554 0.943 * .355 12 (953.620) (0.481)

SOCIAL WELFARE N -7730.536 4.103 .035 12 (1198.000) (6.021)

TABLE 8.5: POISSON REGRESSION OF TIME TRENDS IN ANES IMPORTANT NATIONAL PROBLEMS MODEL: Trend: Parameter Polarization Intercept Estimate PE ( | , = ࡺ ( | ) Y/N (R.S.E.) (R.S.E.) Z Pr > |Z| N ࡸ ࢼ ࡵࡹ ࢅ ෑ ࡼ࢘ ࢏࢓࢏ ࢟࢏ ࢏ୀ૚ PUBLIC ORDER y -104.381 0.055 .016 12 (45.390) (0.023)

RACIAL PROBLEMS Y -137.292 0.070 .032 12 (65.192) (0.033)

SOCIAL WELFARE N -13.275 0.010 .507 12 (29.126) (0.015)

235

the hot-button social issues such as abortion and gay marriage.61 The parameter estimate for the public order model indicates a significant,positive trend in concern over social issues

(11.794). Public order mentions increase over the time series just short of twelve mentions for each survey year, and this is by far the largest coefficient among the three national problems models.

While the “racial problems” model is also significant, it increases at just short of one mention per survey year (R2 = .355). It is not surprising—given the relationship between the mean and standard deviation in the raw data—that this was the only model for which the use of robust standard deviations resulted in an increase in the estimated standard error for the model

(and thus a decrease in statistical significance). There is no apparent statistically significant trend in social welfare mentions, though the parameter estimate for the model is positive. The ML estimates for the three national problems models are reported in Table 8.5.62 As with the partisan affect models, the results of the maximum likelihood estimations of the national problems counts is consistent—both in terms of statistical significance and the direction / size of the coefficients—with the OLS models reported in Table 8.4. Both the “Public Order” and “Racial

Problems” models are significant and in the expected direction. The Social Welfare ML model, as with the OLS model, is statistically insignificant. The substantial increase in social issue mentions on the national problems item is significant and substantively important evidence of political polarization on the social issue dimension and in support of the culture wars thesis.

61 See Appendix H for the total response set categories and descriptions for the National Problems item from the ANES. 62 The maximum likelihood equation used for the regression models in Table 5.16 is reported in Equation 5.15.

236

CONCLUSION

I have examined the open-ended responses to two items from the ANES: partisan

affect and national problem items. While I found significant increases for three out of the four

issue dimensions (government philosophy, economic, and social), my analysis found that foreign

policy had declined in salience for the American public. Again, as mentioned earlier, foreign

policy opinion for this time period is distinctly non-linear, and the open-ended response survey

years end short of 2001 and the September 11th terrorist attacks. On both partisan affect and the identification of important national problems, there has been a significant increase in the absolute salience and relative position vis-à-vis the other issue dimensions for social issues.

Indeed, the social issue dimension evidenced the largest increase in counts and was the most frequently mentioned issue dimension from 1988 forward. This chapter, as was the case in

Chapter 6, presents strong and consistent evidence of political polarization, not just on the culture wars, but on the gamut of political issue dimensions in American politics.

237

CHAPTER 9: GROUP POLARIZATION OF PUBLIC ATTITUDES I: PARTISAN POLARIZATION

Polytics ain’t beanbag” – Martin J. Dooley via columnist Finley Peter Dunne

“We simply must look beyond partisan goals and find common ground as Americans. It is imperative that the Members of Congress recognize that partisanship will not serve the American people.” – Michael Crapo

“Some have dug into unyielding ideological camps that offer no hope of compromise.” - Barak Obama

Having examined political polarization in public opinion and attitudes on gay rights, abortion, and other hot-button social issues in Chapter 5, 6, 7 and 8, I turn my attention now to the structure of polarization by examining polarization at the group level. I developed the theoretical foundation for a group-based analysis in Chapter 3. The proximity of individuals to one another along a relevant political dimension encourages identification with those who see the political world the same way that you do. Secondly, a substantial distance between you and those you identify with relative to another group of citizens results in alienation from that other group and can lead to social conflict. The greater the distance between groups, the more intense is alienation and the higher the probability of greater conflict between the groups in politics. Parties are a natural vehicle for social conflict between groups, and they are a natural source of identification and alienation in politics themselves (i.e. serving as political groups). If the parties become more ideological coherent (increasing identification) and separate on the ideological dimension (increasing alienation), then the obvious implication is greater political conflict and a decreasing available space for political compromise. Similarly, if the parties become more consistent on social issues and move away from each other in terms of average positions on the cultural issues, then the consequence would be a partisan culture ‘war.’

238

I will examine political polarization on both counts, analyzing partisan polarization on

ideology as well as on one of the hot button (if not the hot button) social issues: abortion. This

analysis will show that significant partisan polarization has occurred in the ideological

dimension as well as on abortion. Republicans have become significantly more conservative,

while Democrats have become substantially more liberal. On abortion, Republicans have moved

in a Pro-Life direction on the abortion scale, but even more so, Democrats have become

increasingly Pro-Choice.

In some respects, the debate over the culture wars reduces to the debate over the

importance of abortion in American politics. The 1973 USSC decision in Roe v. Wade is widely

seen as a catalyst for Christian political activism, and particularly within the Republican Party

(Layman 2001, 1996, 1999). Every Republican presidential nominee since 1973 has been either

staunchly Pro-Life or adopted the Pro-Life position in order to secure the nomination. And it is

likewise for Democrats and a strong Pro-Choice position. Whatever their substantive affect on

partisan politics, the party platforms have had widely disparate positions on abortion since Roe.

While these points of fact suggest abortion is an important political issue for the political parties and that it could be a potential point of cleavage and polarization, they do not constitute evidence in and of themselves. For that, we need a systematic assessment of the positions of the parties on abortion and the relative distance between the parties on that issue dimension.

The group polarization measure of the partisan groups on abortion accomplishes exactly that.

While the culture war story is very much centered on specific social issues, the larger ideological dimension is a clearly relevant political space where polarization is a substantively important phenomenon. First, ideology serves as a proxy for the n-issue space in which political competition occurs. As such, it provides a short-cut to answering questions about global

239

polarization in the political space that a focus on specific issues or issue dimensions cannot

provide individually. Second, while specific issues exhibit varying degrees of salience and

passion in the American public, ideology is ubiquitous across the American political landscape.

It is a salient political dimension for the full time series considered in this analysis and it serves

as a perceptual screen for the mass public whether citizens are strong ideological identifiers or

not. That is not to say that examining trends in ideological group polarization renders an

examination of specific issue polarization irrelevant or redundant. Ideology can serve as a proxy

for the issues on the public agenda, but it is a rough proxy. Furthermore, to the extent it serves

as a proxy for issues, it is an aggregate proxy. As we’ve seen in Chapter 5, 6, 7, and 8, there is a

great deal of nuance and interesting political phenomena both within issue dimensions and

between issue dimensions. This is why I include one of the more important social issues,

abortion, in the analysis of partisan polarization.

The evidence presented here on group polarization speaks to two fundamental sets of

questions on political polarization and the culture wars: 1) Is there partisan polarization? Are

the parties divergent and to what extent are the divergent on ideology? What is the trend in

partisan polarization—are the partisan groups increasingly ideologically coherent at the mass

level)? 2) Have the parties polarized on social issues? Is there a polarization trend and if so to

what extent have partisan groups contributed to polarization on the social dimension?

PARTISAN POLARIZATION ON IDEOLOGY: HETEROGENEOUS OR HOMOGENOUS PARTIES?

Partisan polarization is one of the major points of contention and controversy in the polarization literature. Whatever its label and ultimate source (ideological citizens identifying more consistently with parties or party identifiers shifting to the extremes of ideology), the degree to which the mass public has become more coherent and divergent is a significant

240

polarization indicator. Some scholars believe that very little partisan polarization has occurred:

“the amount of [polarization] at the level of ordinary voters hardly seems commensurate with the sorting of Democrats and Republicans at the elite level.” I will address the direct relationship between mass and elite ideology in Chapters 11 and 12. Here I will address the extent and degree to which partisan polarization has occurred at the mass level. As Fiorina puts it, “if [partisan polarization] has been taking place in response to elite polarization, we would expect to see an even tighter relationship between partisanship and ideology among voters…”

(Fiorina and Levendusky 2006). The analysis here empirically tests exactly that expectation.

If there is increasing partisan polarization at the level of the mass public, and this polarization has translated into increasingly polarized parties in the government, then the clear implication of such a trend is a significant impact on the public policy process. Explicitly, we would expect policy alternatives to the status quo by the respective parties to be more ideologically extreme. In other words, we would expect more ambitious policy agendas seeking greater change relative to the status quo. Furthermore, the prospects of compromise on public policy should decline. Not only would the proposed policy alternatives of the majority party be more extreme relative to the status quo, but the ideologically extreme minority party would view these policy proposals as further distant from their own set of acceptable policies, as those policies have moved towards the pole as well.

Using the group polarization measure I developed in Chapter 3 and the average D-W

Nominate scores for the Republican and Democratic legislators in the House and the Senate from 1972 to 2004, I will attempt to answer these questions on political polarization by assessing whether these groups have ‘polarized’ on partisan and ideological dimensions over the course of the past three decades and the relationship between that polarization and partisan

241

polarization in Congress. First I address the question: Is there increasing ideological coherence

among partisan identifiers? What role do the Republican and Democratic Parties play in that

polarization over time? Secondly, I ask: Is there a relationship between elite and mass partisan

polarization? What does this relationship tell us about the public policy agenda and the

prospects for bipartisan compromise? And finally, I look at the parties in the mass public on a

specific culture wars issue: abortion. Have the parties increasingly diverged on the abortion

issue? A systematic assessment of the aggregation of attitudes in the partisan groupings and

along these important political and issue dimensions represents a substantively important step

towards a greater understanding of partisan polarization in the United States. Linking partisan

polarization in the public with partisan polarization in the Congress can help us understand a

policy process increasingly hostile to compromise.

Data

The data for this analysis are culled from the American National Election Study (ANES)

cumulative file.63 I use the ANES studies from 1970-2004.64 The creation of the data set for analysis of polarization trends for the mass public is a two-step process. In the first step, univariate statistics are generated on the substantive variables from the ANES cumulative file.

63 The Cumulative Data File consists of variables derived from the 1948-2004 series of biennial ("time- series") SRC/CPS National Election Studies. The American National Election Studies / Time Series Studies are collected before and after presidential (pre and post surveys) elections. The off-year elections typically only have a post-election study. The ANES Cumulative Data File is a merged data set of all the time series studies from 1948-2004. The pooled data includes variables which appear in three or more studies and consists of 44,715 cases. 64 The data is sub-setted by year to include only studies from 1970-2004 as the previous data sets had few to none of the relevant substantive variables which are necessary for the polarization analysis. Furthermore, 1970-2004 covers the relevant time period to examine the culture wars thesis.

242

individual variables themselves in the new data set. For example, let’s consider the party identification variable as a group identification variable and ideology as our ‘issue’ dimension variable. The party ID variable has three categories: Republicans, Independents, and Democrats.

The first step involves generating the means and frequencies for ideology for each of the categories in each of the study years. In the second step, a data set is created where there are three ideology variables that reflect the mean position on ideology for respondents within each of the party ID categories. A time series data set was created that contains mean and frequency variables for all of the variables relevant to the polarization analysis.

Variables

As mentioned above, there are two ‘types’ of variables used in this analysis. The first type is a group-identification variable: an ordinal classification of the population along some relevant dimension. The group identification variable for this analysis is party identification. The party ID variable used for classification is a three-category variable that collapses strong partisans, weak partisan identifiers, and independent leaners all into an aggregate ‘party’ category. So, for example, “Republicans” in this variable are respondents who either identified themselves as strong Republicans, weak Republicans, or Independents who lean towards the

Republican Party.

The two political dimension variables are ideology and opinion on abortion. The ideology variable is a seven-point scale where respondents self-identified on the dimension ranging from “Extremely Liberal” to “Extremely Conservative.”65 On abortion, respondents to the ANES were asked when abortion should be allowed and given the following options from the beginning of the time series up through 1980 of 1) asserting abortion should never be

65 Seven Point Ideology Scale: 1) Extremely Liberal 2) Liberal 3) Slightly Liberal 4) Moderate; Middle of the Road 5) Slightly Conservative 6) Conservative 7) Extremely Conservative.

243

permitted, 2) asserting abortion should be forbidden except where the life or health of the

woman is in danger, 3) asserting abortion should be permitted for personal reasons (such as

difficulty carrying the child) and 4) asserting abortion should never be forbidden as no woman

should have to carry a child to term she does not want. In 1980 there was a substantial change

in the language of the abortion question with more emphasis put on what the respondent

thought the law should be on abortion. All four of the options begin with some variation of “by

law” and the options were substantially changed in the two middle categories. In option

number 2, now the health option was omitted and rather the respondent was asked to indicate

whether they would make legal exceptions to an absolute ban on abortion for rape, incest, and

the life of the mother. In option number 3, respondents were asked if they would further

expand the number of allowable justifications for an abortion under the law beyond the

exceptions delineated in option number 2 but where “a need for the abortion has been clearly

established.” The first option (never any abortions) was not substantially altered and only a

minor change to the 4th option was made, using the language of “personal choice” rather than a child the woman “does not want.” While there is good reason to believe these substantive changes in the abortion question had a substantive impact on the distribution of abortion (see

Mouw & Sobel, 2001), it impacts only three of the survey years in the analysis (and one of those is omitted in the contribution analysis). Where a year from this period was a significant regression outlier (3 or more standard deviations), it was omitted from the analysis.

Data Presentation Organization

To depict the trend in polarization, I report the calculated polarization score for

each survey year and provide illustrations of the variance in the group polarization measure

over time. Three types of tables are used to demonstrate group polarization in this analysis.

The first type of table is a decomposition table, which shows the calculated group polarization

244

measure broken down by each grouping or classification’s contribution to the overall group

polarization measure. For example, consider partisan polarization on ideology. The

decomposition table will show the individual contributions of Republicans, Independents, and

Democrats to the total group polarization score for each year in the time series. If you look at

the decomposition for, say, 1986, the Republican score will be the Republican identifiers’

squared distances on ideology from each of the other party id classifications (Democrats and

Independents). The last column is the group polarization score for the classification variable

(party identification in this case), which sums each of the polarization scores for all of the other

classifications (see Equation 9.2).

The second type of table is a contribution table. The contribution table reports the percent contribution for each of the individual categories in the classification variable to the group polarization score. So if, say, the Republican squared distances on ideology compose 44% of the summed squared distances for all the categories in 1986, then the table will report a 44% for Republicans in that year. Each percent contribution column is paired with a mean deviation column for each of the categories for the classification variable used for the group polarization score. Let’s say that the mean percent contribution for Republicans on partisan polarization on ideology for the full time series is 40%. Thus the mean deviation for Republicans in 1986 would be 4.000. If the percent contribution falls below the mean, then the mean deviation will be negative (if the percent contribution for Republicans had been 36 rather than 44, then the mean deviation for Republicans in 1986 would have been -4.000).

For each analysis, I include the weighted and the unweighted group polarization score.

The weighted score is the group polarization measure defined in Equation 9.2. This weights the distances of group I to groups N through groups M based on the size of group I. Again, as I argued in Chapter 3, group size is an essential component of the total conflict in society with a

245

political dimension. Small groups may be extreme relative to other groups, but their capacity to contribute to societal conflict is limited both as a consequence of the ‘threat’ posed by such a group to others in society and also given majoritarian, democratic institutions and electoral systems that discriminate against small groups in making policy and translating votes to seats in the government. That said, the trends in the movement of groups on a political dimension relative to the other groups irrespective of size are an important aspect of polarization. How extreme the Republicans have become relative to Independents and Democrats, regardless of how many Republican identifiers there are, is a significant factor in political conflict. Trends in group distances are of intrinsic interest. Furthermore, while group size conditions political conflict, it must also be noted that the perceived ‘threat’ of opposing groups may not be sensitive to small changes in the size of a group. The group polarization measure I developed in

Chapter 3 assumes that the effect of group size on conflict is continuous and linear. However, the latent relationship between group size and conflict may be ordinal and/or nonlinear. Given this, I include the unweighted group polarization as a separate contribution table and I include trend regressions for both weighted and unweighted group polarization. The unweighted contribution table includes the percent contribution of a category (Republicans) to the total unweighted group polarization score, which is simply the sum of the squared distances on that particular political dimension (partisan or ideological). Thus the unweighted percent contribution for Republicans is a pure distance measure for that group relative to all other partisan groups on the political dimension.

The third table type is a regression table, reporting the model statistics for the trend models on group polarization and the trend models on the mean deviation of the percent contribution of each category in the classification variable to the group polarization score. The first regression assesses whether total group polarization has occurred, and whether each group

246

has become polarized relative to the other groups. The second tests whether a trend has

occurred in the percent contribution of the group to the total group polarization score for both

weighted and unweighted group polarization. So each analysis of groups on either ideology or

partisanship includes a decomposition table of group polarization on the ideological or partisan

dimension, a table of regressions for each category as well as group polarization for the full

classification variable (showing the trend in polarization for, say, party ID and the polarization of

each party identification group: the Republican regression would assess whether Republicans

have increasingly polarized on the ideological dimension relative to the other groups), two

contribution tables (weighted and unweighted) including the percent contribution to

polarization for each category and the mean deviation for each category in each year of the

analysis, and a table of the regressions for the weighted and unweighted group polarization

measures showing the trend in the mean deviation in the contribution of each group to the

group polarization score over the time series.

MASS PARTISAN POLARIZATION ON IDEOLOGY – DECOMPOSITION AND CONTRIBUTION TO GROUP POLARIZATION

As noted earlier, some polarization scholars have suggested that little partisan polarization has occurred at the level of the mass electorate and certainly not commiserate with the polarization that has occurred the elite level. This view is mistaken on both counts. There has been significant partisan polarization (or sorting, depending on your preference for terminology) over the last three decades. Furthermore, partisan polarization is just as relevant to social conflict and the culture wars as non-partisan polarization on issues in the mass electorate. Parties are a fundamental political institution in American society and the primary mechanism through which political and social conflict is translated. It is parties that put forward policy platforms, parties that populate the electoral institutions of government, and parties that shape and influence (and are shaped and influenced by) the mass electorate. If parties diverge

247

TABLE 9.1: PARTY ID – DECOMPOSITION OF PARTISAN GROUP POLARIZATION ON IDEOLOGY, 1972-2004

YEAR Republican GP Independent GP Democrat GP GPPID 1972 0.38904 0.05760 0.50049 0.94713 1974 0.44606 0.07431 0.50261 1.02298 1976 0.63629 0.10281 0.72772 1.46683 1978 0.49748 0.09902 0.51693 1.11343 1980 0.36929 0.05968 0.61532 1.04430 1982 0.66645 0.09806 1.26506 2.02957 1984 0.60661 0.06535 0.54337 1.21532 1986 0.42881 0.05442 0.58473 1.06796 1988 0.81130 0.08100 0.71103 1.60333 1990 0.48434 0.06718 0.86878 1.42031 1992 0.79990 0.09609 0.77702 1.67302 1994 1.10393 0.11880 1.45902 2.68175 1996 1.26947 0.10782 1.36976 2.74705 1998 0.87420 0.09498 1.23556 2.20473 2000 1.12631 0.14229 1.57994 2.84854 2002 1.94442 0.11344 1.79780 3.85566 2004 1.62303 0.16227 2.07954 3.86484

ideologically, then that has important implications for the range of alternatives made available on the public agenda, the nature of political conflict at the elite level, and the potential for political compromise in the elective institutions of government.

Table 9.1 reports the group polarization scores on ideology for partisanship and the decomposition of the group polarization measure, giving the individual contributions to the total measure for the three-category partisan identifiers. Note that the measure for each of the groups reflects polarization for Republicans, Democrats, and Independents. Independents contribute relatively little to the trend in ideological polarization for partisan identifiers, both the Republican Party identifiers and the Democratic Party identifiers have increasingly contributed to ideological polarization (Figure 9.2). The group polarization score for party ID on ideology in 1972 was nearly 1.0 (.947). By 2004, that score would come up just short of four

(3.86), a three-fold increase in partisan polarization on ideology. The big jumps occur in 1984 (a

248

half point increase to 1986) and from 1992 to 1994 and 2000 to 2002 (a full point increase). All

during the “culture wars” period and evidencing significant mass polarization. Figure 9.1

illustrates the significant increase in partisan polarization on the ideological dimension over the

course of the time series.

Through the 1970’s there was a fairly flat trend in partisan polarization. It was largest in

1976, in the first post-Watergate presidential election (1.46), but for the most part hovered near

a group polarization score of 1.0. In the 1980’s we saw partisan polarization spike in 1982, but

for the most part the partisan polarization was only slightly higher in the 1980’s relative to the

1970’s. This all changes in 1988, as partisan polarization on ideology established a new and

higher plateau (around 1.5). This would continue until 1992. Between 1992 and 1994,

proximate to Clinton’s failed attempt to reform health care and leading into the Republican

Revolution, partisan polarization on ideology begins to significantly and substantially increase

more than it had at any other point of time since the 1970’s. By 1994, partisan polarization is

approaching 3, doubling the partisan polarization on ideology that we witness in the 1970’s.

The first decade of the 21st century and the Bush administration is no exception to this trend.

Indeed, while there was a small decline in 1998, partisan polarization spiked even more in the

2000’s: peaking near a partisan group polarization score of 4 in 2004. This score is also the maximum polarization score for the time series.

Figure 9.2 shows the relative contributions of Republicans, Democrats, and

Independents to the group polarization score. Recall, these are weighted scores, so independents have a small and declining contribution to partisan polarization given their size relative to the parties and the fact that independent identifiers have declined since the 1970’s.

That said, there is a slight increase in the contribution of independents to polarization

249

FIGURE 9.1: PARTY ID GROUP POLARIZATION ON IDEOLOGY OVER TIME, 1972-2004 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0

Party ID GP

FIGURE 9.2: DECOMPOSITION OF PARTISAN GROUP POLARIZATION ON IDEOLOGY OVER TIME, 1972-2004 2.2 2.1 2 1.9 1.8 1.7 1.6 1.5 1.4 1.3 1.2 1.1 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

REPUBLICANS INDEPENDENTS DEMOCRATS

250

despite its decreasing weight, likely due to the increasing distance of Republicans and

Democrats from Independents rather than movement on ideology for the Independent identifiers themselves. Interestingly, Democrats have higher group polarization scores than

Republicans for most of the series. The three data points where the Democrats spike above the

Republicans demonstrate the face validity of the measure of partisan polarization. In 1982, in response to Reagan’s election, Democrats spike from a polarization score of 0.6 to a score just short of 1.3. Democrats again surge ahead of the Republicans on polarization in 1994, likely in response to the Republican Revolution. And finally, Democrats substantially outpace

Republicans in polarization in 2004, coincident to the prosecution of the Iraq war.

TABLE 9.2: TREND REGRESSIONS OF PARTY ID GROUP POLARIZATION ON IDEOLOGY, 1972-2004 Trend: MODEL: Polarization Intercept Parameter Estimate 2 GP(I) = B0 + B1(YEAR) + E Y/N (S.E.) (S.E.) R N

Republican Y -73.574 0.037 *** .695 17 Identifiers (12.727) (0.006)

Independent Y -3.943 0.002 *** .465 17 Identifiers (1.118) (0.000)

Democrat Y -83.667 0.043 *** .723 17 Identifiers (13.527) (0.007)

Party ID Group Y -161.185 0.082 *** .745 17 Polarization (24.663) (0.012)

251

The regression models of the group polarization measures for each partisan identifier category and for party ID as a whole demonstrate the significant across-the-board partisan polarization that has occurred since the 1970’s (Table 9.2).66 Note that while Independent identifiers have significantly contributed to partisan polarization, substantively speaking their impact is negligible, again, given their size relative to the other parties and the fact that the two parties are the greatest contributors to partisan polarization. The polarization of Independents relative to the other parties has only increased by .002 for every year in the time series.

Compare that to the Republican increased contribution to group polarization (.037) and that of the Democrats (.043). Both Republican and Democrat identifiers have a much more substantial

(and near equal) contribution to the group polarization score on ideology for party ID.

The models for the Republican and Democrats proportionately reduce the error over the mean by over 70 percent in the case of the Democrats and nearly 70 percent for the

Republicans. Clearly this is strong evidence that the partisan identifiers in the mass public have polarized. Partisan polarization on ideology has increased by 0.082 per for every survey year since 1972, Survey year (I.e. the progression of time) explains 74.5% of the variation in partisan polarization. The trend model for Independent identifiers performs poorly in comparison, explaining less than half of the variation in partisan polarization over the time series. And again, most of that is likely attributable to the Republicans moving to the Right and Democrats moving to the Left.

66 For all regression models reported in Chapter 6, the following significance standards are observed: * significant at .10 level ** significant at .05 level ***significant at .01 level

252

TABLE 9.3: PERCENT CONTRIBUTION & MEAN DEVIATION WEIGHTED PARTISAN G.P. ON IDEOLOGY, 1972-2004 YEAR REP R-DEV IND I-DEV DEM D-DEV 1972 41.08% -1.42% 6.08% 0.75% 52.84% 0.67% 1974 43.60% 1.10% 7.26% 1.94% 49.13% -3.04% 1976 43.38% 0.88% 7.01% 1.68% 49.61% -2.56% 1978 44.68% 2.18% 8.89% 3.56% 46.43% -5.74% 1980 35.36% -7.14% 5.71% 0.39% 58.92% 6.75% 1982 32.84% -9.66% 4.83% -0.50% 62.33% 10.16% 1984 49.91% 7.41% 5.38% 0.05% 44.71% -7.46% 1986 40.15% -2.35% 5.10% -0.23% 54.75% 2.58% 1988 50.60% 8.10% 5.05% -0.28% 44.35% -7.82% 1990 34.10% -8.40% 4.73% -0.60% 61.17% 9.00% 1992 47.81% 5.31% 5.74% 0.41% 46.44% -5.73% 1994 41.16% -1.34% 4.43% -0.90% 54.41% 2.24% 1996 46.21% 3.71% 3.92% -1.40% 49.86% -2.31% 1998 39.65% -2.85% 4.31% -1.02% 56.04% 3.87% 2000 39.54% -2.96% 5.00% -0.33% 55.46% 3.29% 2002 50.43% 7.93% 2.94% -2.39% 46.63% -5.54% 2004 41.99% -0.51% 4.20% -1.13% 53.81% 1.64%

TABLE 9.4: PERCENT CONTRIBUTION & MEAN DEVIATION UNWEIGHTED PARTISAN G.P. ON IDEOLOGY, 1972-2004 YEAR REP R-DEV IND I-DEV DEM D-DEV 1972 44.86% 1.38% 17.14% -0.49% 38.00% -0.89% 1974 47.61% 4.13% 18.75% 1.11% 33.64% -5.24% 1976 46.76% 3.28% 18.07% 0.43% 35.17% -3.71% 1978 49.48% 6.01% 21.68% 4.05% 28.83% -10.05% 1980 40.77% -2.70% 16.70% -0.94% 42.53% 3.64% 1982 39.69% -3.78% 16.81% -0.83% 43.49% 4.61% 1984 47.01% 3.53% 18.24% 0.61% 34.75% -4.14% 1986 42.35% -1.12% 16.69% -0.95% 40.96% 2.08% 1988 46.56% 3.08% 17.94% 0.30% 35.50% -3.38% 1990 36.39% -7.08% 17.61% -0.02% 45.99% 7.10% 1992 47.10% 3.62% 18.31% 0.68% 34.59% -4.30% 1994 37.92% -5.56% 17.16% -0.47% 44.92% 6.03% 1996 46.07% 2.59% 17.65% 0.02% 36.28% -2.61% 1998 41.54% -1.93% 16.67% -0.97% 41.79% 2.90% 2000 40.13% -3.35% 16.76% -0.88% 43.11% 4.23% 2002 44.20% 0.72% 16.96% -0.68% 38.85% -0.04% 2004 40.63% -2.84% 16.71% -0.93% 42.66% 3.77%

253

In Tables 9.3 and 9.4 I report the percent contribution to and mean deviation for the weighted and unweighted group polarization for each of the party categories. Recall that the distinction between the two types of group polarization is that the weighted group polarization weights the group ideology differences by the size of the group (percentage of party identifiers for that category in the sample) while the unweighted table reports percentage contribution to a group polarization measure that treats each group equally, irrespective of size. Note that the

Republican and Democratic Parties account for almost all of the percent contribution to group polarization. The maximum contribution for the Independent respondents is 8.89% in 1978, and the percentage contribution declines consistently as we move forward to the end of the time series. The minimum contribution for Independents is 2.94%, in 2002. The impact of the size of the group of Independents for the weighted measure of partisan polarization on ideology is apparent in comparing the percent contributions for Independents in Table 9.3 and the percent contributions for Independents in Table 9.4. The percent contribution to the unweighted measure for Independents varies between 16% and 22%. Furthermore, there is no apparent linear trend in the contribution of Independents in the unweighted measure, suggesting the trend observed in Table 9.3. is mostly due to the declining number of Independent identifiers since 1972.

While the Republican and Democratic percent contributions are clearly affected by their relative group sizes in the weighted measure, it is not nearly as dramatic as the effect on

Independent contributions. For the most part there is ordinal consistency between the weighted and unweighted contributions for the identifiers with the two political parties, but there are differences, suggesting the size of the group has an influence on contribution. For example, the largest contribution of the Republican identifiers to weighted partisan polarization on ideology is in 1988 (50.68%), but it is just the second largest contributor to unweighted group

254

FIGURE 9.3: UNWEIGHTED MEAN DEVIATION TRENDS IN AVERAGE IDEOLOGY FOR PARTISAN IDENTIFIERS 12.00% 11.00% 10.00% 9.00% 8.00% 7.00% 6.00% 5.00% 4.00% 3.00% 2.00% 1.00% 0.00% -1.00% -2.00% -3.00% -4.00% -5.00% -6.00% -7.00% -8.00% -9.00% -10.00% -11.00% -12.00% 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

DEMOCRATS REPUBLICANS

polarization (46.56%). The implication is that not only were Republican identifiers more distant from the other party identifiers in 1988, but there were more Republican identifiers in 1988 resulting in a greater contribution to the weighted group polarization measure.

If we look at the weighted group polarization in Table 9.3, we see that there are rather significant oscillations in percent contribution between the Republican and Democratic Parties.

For example, in 1982 Republicans account for only 32.84% of the group polarization score, which is 9.66% below the mean contribution for Republicans in the time series. But in 1984, in the next survey year, that contribution climbs to 49.91%, which is 7.41% above the mean contribution to partisan polarization. There is a similar trend with the Democrats. In 1982, the

Democrats account for 62.33% of partisan polarization on ideology, coming in at 10.16% above the mean contribution. In 1984, that drops to 44.71%, below the mean contribution. While relative size plays some role in this oscillation, the oscillations are still apparent in the unweighted group polarization contribution table. If you look at Figure 9.3, you can see the up and down nature of the percent contributions for Republican and Democratic identifiers in the

255

unweighted group polarization score. However, there does appear to be some patterns in the contributions. Democrats were well below the mean through most of the 1970’s, and they were well above the mean for most of the 2000’s. Conversely, Republicans were big contributors to group polarization in the 1970’s relative to the other party identifiers while less so in the other, more recent, decades. Republican relative contribution has been on the decline.

TABLE 9.5: WEIGHTED & UNWEIGHTED MEAN DEVIATION TREND REGRESSIONS OF PID GP ON IDEOLOGY, 1972-2004 YEAR PPA MODEL: Intercept P.E. P.E. 2 GPMD(PID) = B0 + B1(YEAR) + B2(PPA) (R.S.E.) (R.S.E.) (R.S.E.) R N + E WEIGHTED GROUP POLARIZATION

Republican -143.100 0.073 -1.921 .052 17 Identifiers (188.135) (0.095) (2.169)

Independent 221.312 -0.111 *** -0.772 * .674 17 Identifiers (42.509) (0.023) (0.438)

Democrat -78.177 0.039 2.694 .058 17 Identifiers (187.781) (0.095) (2.293)

UNWEIGHTED GROUP POLARIZATION

Republican -327.973 -0.164 * -2.318 .257 17 Identifiers (174.475) (0.057) (1.613)

Independent 113.164 -0.057 ** -1.046 * .355 17 Identifiers (51.175) (0.026) (0.602)

Democrat -441.137 0.221 *** 3.364 * .316 17 Identifiers (208.061) (0.076) (2.013)

256

In Table 9.5, I regress the weighted and unweighted group polarization score percent contributions represented as a deviation from the mean contribution for the time series for each of the identifier groups. The model uses robust standard errors. Also included in the models is the party of the presidential administration. The reason to include presidential partisanship should be apparent. In one respect, it is indirectly influenced by the number of party identifiers. Secondly, we could see a greater contribution for out-of-power parties to partisan polarization on ideology. It is certainly plausible that Democrats may polarize in response to a Republican presidency, and likewise for Republican identifiers during a

Democratic administration. In the weighted models, there are no significant linear trends in the percent contribution of Republicans or Democrats. Nor is there a significant relationship between the percent contribution of the partisan identifiers and presidential partisanship.

There is a strong statistical relationship between survey year and the percent contribution of

Independents, reflective of the negative trend in that categories contribution to the group polarization score. As I noted previously, this is largely the consequence of the declining number of Independent identifiers since the 1970’s. For every survey year, the contribution of

Independent identifiers declines by -0.111, and the model accounts for 67.4% of the variation in contribution. There is a significant, negative relationship between the presidential administration partisanship and the contribution of Independent identifiers to the group polarization score (-0.772). This is likely a consequence of the fact that a great deal of the decline in Independent identifiers occurred in the 1990’s, when President Clinton was in office.

When we treat the groups equally, ignore the relative size of the groups, there are an interesting set of patterns not apparent for the weighted group polarization measures. Just looking at the relative distances between the categories, there is a significant positive trend for the Democratic party and its contribution to polarization over the time series. For every survey

257

year there is a 0.221 increase in the percent contribution of Democrats above the mean

contribution. Democrats have become increasingly distant from the other identifiers on

ideology, contributing more to group polarization. Interestingly, there is a significant negative

trend for Republican identifiers (-0.164) along with the still-negative trend for Independents (-

0.057). Republicans are contributing less to polarization over the full time series. The

presidential partisanship effect on group polarization is in the expected direction and significant

in all models. For the Republican percent contribution to group polarization, presidential

partisanship is negative, declining -2.318 as the percent contribution for Republicans increases.

In other words, Republicans are contributing more to group polarization during Democratic

administrations. Similarly, the coefficient for presidential partisanship is positive for Democratic

identifiers. Democrats are contributing more to group polarization during Republican

administrations. It seems apparent that out-of-power parties react to the other party

controlling the White House by shifting towards the ideological extremes. These models

account for between 25% and 31% of the variance in percent contribution as measured as a

deviation from the mean contribution to unweighted group polarization.

PARTISAN POLARIZATION: THE DEATH OF COMPROMISE & THE PARADOX OF FAILED “MANDATES” FOR CHANGE

In 2004, President Bush won a narrow but decisive victory over the Democratic candidate, John Kerry. Republicans not only secured the White House, but also picked up seats in the House of Representatives (3) and in the United States Senate (4). President Bush felt emboldened to pursue a significant political agenda in the wake of this election. In a press conference shortly after the election he announced: “Let me put it to you this way: I earned capital in the campaign, political capital, and now I intend to spend it….I earned some capital.

I’ve earned capital in this election—and I’m going to spend it for what I told the people I’d spend it on, which is—you’ve heard the agenda: Social Security, and tax reform, moving this economy

258

forward, education, fighting and winning the war on terror.” Six months later, the president’s plan to partially privatize social security through personal accounts for younger workers lay in shambles(Ortiz 2007). Despite a unified Republican government, Bush’s primary public policy agenda item had failed, barely into his second term. Why? And was this unique to just Bush, or did this failure have broader implications?

This wasn’t the first time an elected president, with a unified Congress, failed to get even incremental action on the primary policy goal for his administration. Bill Clinton, elected in

1992 to the presidency and with a 56 to 44 seat majority in the Senate and while losing a few seats (9), maintained a cavernous 91 seat margin over the Republicans (258-176). Again, the newly elected candidate had campaigned hard on an issue (in this case, health care), and sought to enact significant reform on the issue in the wake of his electoral victory. The plan did not disturb the employer-funded health insurance the current system is built on, however it created an overarching governmental health care superstructure (health maintenance organizations) which would mandate universal health care and regulate health care coverage. Despite unified government, by September 1994, Senate Majority Leader George Mitchell declared health care reform (having been dubbed by opponents as “HillaryCare”) dead (Bok 2003).

What happened to these two presidents? Supposedly at the height of their power, with few institutional obstacles in their path and control of the entire federal government, these presidents failed to even obtain incremental change. It is not as if the health care or social security status quos are immune to change. Social Security reforms were enacted in 1977 and

1981. Medicare and Medicaid were not only instituted, but have been modified several times since the programs were initially setup. What is the answer? Some attribute it to powerful interest groups. The health care industry spent millions on opposition ads, including the famous

259

“Harry and Louise” ads which depicted ‘regular’ citizens with worries over Clinton’s plan. The

AARP was firmly aligned against President Bush’s social security plan. While these factors were

certainly important, the overarching factor, I believe, was the increase in partisan polarization.

The parties in the electorate had polarized, producing more ‘landslide’ districts, resulting in

increasingly ideologically divergent parties in Congress.

In the polarized era, advocates of significant change to the status quo face an already

interested and passionate opposition that constitutes a substantial proportion of the public.

Newt Gingrich, writing in the wake of President Bush’s social security plan bogging down in the

legislative process, argued, “I don’t think [President Bush] can get complex reform through…It’s

too hard with the AARP opposing you and all of the Democrats lined up against it.” Likewise,

Hillary Clinton reported being a bit wiser in the wake of health care defeat, “I learned some

valuable lessons about the legislative process, the importance of bipartisan cooperation and the

wisdom of taking small steps to get a big job done.” But was there “bipartisan cooperation” to

be had? Unified government was a sufficient basis for significant policy change in the past.

What about unified government in the 1990’s and the first decade of the 21st century had changed from the halcyon days of unified government under Franklin Roosevelt and Lyndon

Johnson—both successful in producing major policy changes and instituting significant new programs. Indeed, presidents have accomplished more legislative success under divided government, such as the major reductions in the marginal tax rates achieved by President

Reagan in the 1980’s.

These anecdotes suggest two aspects of partisan polarization. Intractable conflict has become more likely in competition over the policy agenda because the distance between the starting points on the two sides has become greater. Hillary Clinton may have learned the value

260

of bipartisanship, but that doesn’t mean that elected officials are more likely to propose changes to the policy status quo likely to induce it. A polarized electorate elects polarized candidates and officials. Even if they want to compromise, the shadow of future elections where a more ideologically consistent and polarized constituency is more likely to punish them for their heresy. Witness the fate of Pennsylvania Senator Arlen Specter, who was forced to switch parties as a consequence of his vote for President Obama’s stimulus bill. Facing a strong primary challenge for his seat in Pat Toomey, Specter faced sure defeat in the 2010 Republican primary. Partisan polarization reduces the compromise space in two ways: 1) It forces the policy agenda of the party in power in the direction of their more extreme members rather than the center. 2) The opposition’s rejection zone is further distant from the center.

In Figure 9.4, I depict a hypothetical scenario where bipartisan compromise on a policy change from the status quo is possible. I represent the continuum of possible policies as a line, on which the Republican and Democratic Parties have a subset of that continuum which are acceptable policies from the perspective of the parties in Congress (would command a majority of the party identifiers). I characterize the “rejection space” as the policy choices that exist outside the range of acceptable policies for either party. In the Figure 9.4 hypothetical, the status quo policy is within the range of acceptable policies for the Republicans, but outside of the range of acceptable policies for the Democrats. In this scenario, Democrats have a majority in the House, and thus can outvote the Republicans in a party-line vote. However, given not

FIGURE 9.4: BIPARTISAN POLICY COMPROMISE BETWEEN IDEOLOGICALLY MODERATE PARTIES

Compromise Status Quo Space Policy Rejection Space Rejection Space

Republican Democratic Policy Range Policy Range 261

only the relative proximity of the range of potential policies but also the overlap in the two acceptable policy ranges, the possibility for a bi-partisan compromise is apparent. Especially on major and highly visible policies, the preference for bipartisan action over party-line policies should be apparent. Majority parties have an incentive to avoid stark policy choices between the two parties, as it could pave the way for electoral defeat in the next election cycle. Bringing the other party in on a policy change (in this case a policy alternative proposed within the compromise space) guards against the use of that policy by the minority party as a cudgel. In this scenario, the Democrats can propose a policy alternative that moves the status quo into the range of their acceptable policies, but also garners support from the Republicans. A bipartisan compromise and a successful change in the status quo is the result.

In Figure 9.5, I construct a hypothetical that is represents a stark break from that of

Figure 9.5. In this scenario, the two parties are polarized on the relevant issue dimension (or in the general ideological dimension). Note that the status quo lies outside both parties range of acceptable policies.

However, unlike the Figure 9.4 hypothetical where a compromise was possible in the center of the dimension, the center in this case is entirely subsumed within the rejection space.

Neither party finds moderate polices an acceptable alternative. There is no overlapping compromise space. Republican policy proposals will exclusively consist of polices that move the status quo substantially to the Right, while Democratic proposals will only include policies well to the Left of the center of the policy dimension. The likely result: intractable conflict (i.e. no change in the status quo) or a party-line policy imposed by the majority party. That policy will be a much more significant change in the status quo than those likely in the Figure 9.4 hypothetical. As such, it may result in no change in the status quo at all: members of the

262

FIGURE 9.5: INTRACTABLE POLICY CONFLICT BETWEEN IDEOLOGICALLY POLARIZED PARTIES

Status Quo Policy RS Rejection Space RS

Democratic Republican Policy Range Policy Range majority party may be unwilling to enact such a substantial change in the status quo given that they could be held responsible for that vote in their less-safe districts in the following election.

At the same time, there is little incentive to propose moderate policies because a) the ideological extreme members of the majority party are unlikely to support it and b) it will not garner support from any of the Republicans, thus gaining bipartisan support.

While it is only anecdotal evidence, there is some indication this is the kind of scenario operative in the defeat of Bill Clinton’s health care plan in 1993, the defeat of President Bush’s proposal to reform social security in 2005, and there are strong similarities between the hypothetical and the ongoing debate over President Obama’s proposal to reform health care.

That all three involved unified government suggests perhaps that the presidents believed that this would allow them to enact a stronger change, given their control of the government. Note that all three proposals represented significant changes to the status quo. Bush sought a privatization of social security, while Obama and Clinton sought to bring health care providers, accountable for 1/7th of the economy, under the purview of government oversight.

Compromise measures failed to garner the support of ideologically extreme members of the majority party. And ultimately (in the case of Bush and Clinton), there was no change to the status quo policy.

263

There is a great deal of the debate over Obama’s health care proposal reminiscent of the Figure 9.5 hypothetical. Obama’s health proposals, including a public option for health insurance designed to ‘compete’ against private insurers, represents a significant change in the status quo. And to this point, compromise policies have a) failed to garner Republican votes and b) been resoundingly rejected by the ideological extremists in the Democratic Party. Witness the House Democrats insistence on the inclusion of the public option in their proposals despite the fact that it is considered a non-starter in the Senate. The Speaker of the House, Nancy Pelosi insists: “There is no division. We all [House Democrats] support a public option” (Miller 2009).

Whether or not the result of this will be no change in the status quo and yet another failed

“presidential mandate” for policy change remains to be seen. But the signs suggest that may be the path that health care reform is on.

Of course, these examples constitute just anecdotal evidence suggesting that partisan polarization has reduced or eliminated the compromise space and generated intractable political conflict. A more systematic examination of policy proposals to change the status quo is beyond the scope of this analysis, but it constitutes the next step in assessing the relationship.

However, there are essential predicates to the scenarios I’ve described: namely that partisan polarization has occurred at the electoral level, that this has produced partisan polarization in the elites (a reciprocal relationship would also be consistent with the scenarios), and thus it drives presidents to propose more radical changes to the status quo in response to their polarized constituents, it produces more polarized elected officials voting on legislation, and thus leading to more conflict and less compromise. I have already shown that there is partisan polarization in the electorate. Next, I will show the partisan polarization at the elite level and assess the relationship between partisan polarization at the mass level and elite partisan polarization.

264

FIGURE 9.6: PARTISAN POLARIZATION IN CONGRESS, 1954-2004

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2 MeanDifference: D-W Nominate Scores 0.1

HOUSE SENATE

Figure 9.6 reports the difference between the average D-W Nominate score for

Republicans and Democrats from 1954 to 2004. The D-W Nominate score (first dimension) is generally interpretable as the Right-Left ideological dimension. Note that the difference between the two parties has substantially increased since the 1950’s in both the House of

Representatives and the Senate. From an average difference 0.40 for the Senate and 0.49 for the House in 1954, the difference fifty years later has doubled, with an ideological difference between the parties in the House above 0.90 and an ideological difference in the Senate of 0.80.

The Senate partisan polarization is particularly striking, given the conventional wisdom that partisan polarization in the House is exclusively due to gerrymandering and the creation of

‘landslide’ districts in recent decades. The only change along the lines of district change in the

Senate was the inclusion of the states of Alaska and Hawaii to the Union, adding four new senators to the chamber. Otherwise, the territories from which senators are elected has remained fixed since 1954. Despite this, the polarization in the Senate has been nearly as substantial as that for the House.

265

FIGURE 9.7: TRENDS IN Z-STANDARDIZED PARTISAN POLARIZATION AT THE MASS AND ELITE LEVELS, 1972-2004 3

-1

-2

-3 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

PARTISAN GROUP POLARIZATION Z-SCORE HOUSE NOMINATE PARTY DIFFERENCE Z-SCORE (R-D) SENATE NOMINATE PARTY DIFFERENCE Z-SCORE (R-D)

In order to compare the partisan polarization in Congress at the elite level with the observed partisan polarization at the mass level as captured by the group polarization measure,

I convert the House and Senate average Nominate differences to Z-scores based on their deviation from the time-series global mean for each of the variables for the period extending from 1972 to 2004. Thus the scores for each year on partisan polarization in Congress are the z- distribution standardized scores. I convert the group polarization score to a Z-Score using the same method. With both elite and partisan polarization scored on the same scale, a direct comparison of their trends is possible (Figure 9.7). The z-score calculation is reported in equation 9.1.

Equation 9.1: Z-Score for Partisan Polarization

ZPP = ೔ೣ ೔ೣ Where: ௉௉ ି௉௉തതതത th = the i observed value of partisan polarizationఙ ೔ೣfor year. = the ith mean value of partisan polarization for year. ௜௫ .the ith standard deviation of partisan polarization forݔ year =ܲܲ തതതത௜௫ ݔܲܲ ߪ௜௫ 266 ݔ

As is apparent in Figure 9.7, there has been significant and substantial partisan polarization at both elite and mass levels since 1972. Furthermore, the magnitude of the polarization at the mass level tracks closely with polarization at the elite level. The negative Z-scores for the congressional D-W Nominate (ideological) differences and the negative Z-scores for the mass partisan polarization are in the 1970’s and the 1980’s. In 1992, the partisan polarization for both elites and masses crosses over to positive standardized partisan polarization, with the partisan polarization at the elite and mass levels in 2004 at or near two standard deviations above the mean. Recall, the negative z-scores do not indicate depolarization or moderation, but rather these partisan polarization scores fall below the global mean.

Table 9.6 reports regression models using mass partisan polarization to predict elite partisan polarization. Note the strong relationship between the trend in partisan polarization for the public and the difference in ideology for the Republican and Democratic Parties in both the House and the Senate. In the House model, there is an average increase in partisan polarization of 0.130 for each survey year, and the model explains over eighty percent of the variance in elite partisan polarization. The story is similar for the Senate model. The per-unit increase in partisan polarization is 0.097, slightly less than that in the House model. And the

TABLE 9.6: CONGRESSIONAL MODELS OF ELITE PARTY DIFFERENCES & MASS PARTISAN POLARIZATION, 1972-2004 Trend: MODEL: Polarization Intercept Parameter Estimate 2 CNSR-D = B0 + B1(GPPID) + E Y/N (R.S.E.) (R.S.E.) R N

House Nominate Y 0.425 0.130 *** .821 17 Party Difference (0.021) (0.009)

Senate Nominate Y 0.457 0.097 *** .779 17 Party Difference (0.028) (0.013)

267

model performance is equally strong: 77.9% of the variance in elite partisan polarization in the

Senate is explained by partisan polarization at the mass level.

What this shows is that there has been a substantial increase in partisan polarization at the mass level. Republican identifiers have become much more conservative. Democratic identifiers have become much more liberal. There has been a concomitant increase in partisan polarization at the elite level. Republican legislators have become much more conservative.

Democratic legislators have become much more liberal since the 1970’s. As Figure 9.7 and the models in Table 9.6 indicate, there is a strong relationship between mass and elite partisan polarization both in direction (upward) and in magnitude. While the direct evidence of increasing intractable policy conflict and a related decline in the probability of compromise on major policy changes to the status quo presented here is anecdotal, this strong relationship between mass and elite polarization is very suggestive that the state of affairs in American politics today is closer to that depicted in Figure 9.5 rather than that shown in Figure 9.4.

Republican and Democratic Presidents are more likely to play to their base constituencies, and propose more extreme policy alternatives. The legislators tasked with crafting these proposals into actual laws are less inclined to compromise, less inclined to acquiesce to moderate alternatives, and more likely to take a ‘this change or no change’ stance. The prospects of party- line policies increase the chance of a ‘no policy change’ outcome, as it exposes vulnerable party members in the next election cycle. The result: intractable conflict, party-line voting, and increasingly intense opposition from the party out of power. Whether this is the case or not, it is clear that the last thirty years plus has witnessed a remarkable and significant increase in the ideological consistency of and ideological distance between the Republican and Democratic parties in the mass public and at the elite level.

268

PARTISAN POLARIZATION ON ABORTION

Having established partisan polarization in the ideological dimension, I turn here to the issue dimension, and specifically one of the most important issues for the culture wars: abortion. The abortion issue is the sine qua non of the culture wars. Increasing partisan polarization on the abortion issue would be further evidence that the parties are increasingly motivated on the social issue dimension and that public debate on social issues and the public policy process is increasingly structured by partisan competition. I report the group polarization measure for partisanship on the abortion issue along with the decomposition of the measure into party ID categories in Table 9.7.

Table 9.7 shows significant partisan polarization on the abortion issue, mostly due to the polarization of Republicans and Democrats on the issue. In 1972, there was little to no abortion polarization (0.015). By 1990, each of the individual category contributions to polarization on abortion by the Republicans and Democrats would exceed the total polarization on abortion in

TABLE 9.7: PARTY ID – DECOMPOSITION OF PARTISAN GROUP POLARIZATION ON ABORTION, 1972-2004

YEAR Republican GP Independent GP Democrat GP GPPID 1972 0.00760 0.00135 0.00638 0.01533 1974 0.00491 0.00165 0.00759 0.01415 1976 0.00221 0.00195 0.00880 0.01296 1978 0.00556 0.00101 0.00991 0.01648 1980 0.00378 0.00361 0.01543 0.02282 1982 0.00338 0.00077 0.00923 0.01337 1984 0.00743 0.00340 0.02230 0.03313 1986 0.00750 0.00374 0.00671 0.01795 1988 0.00664 0.00224 0.00382 0.01270 1990 0.02136 0.01513 0.04978 0.08627 1992 0.05765 0.00804 0.04273 0.10842 1994 0.03484 0.00343 0.04133 0.07961 1996 0.04417 0.00481 0.07744 0.12642 1998 0.07268 0.00754 0.08076 0.16098 2000 0.05157 0.01032 0.10698 0.16887 2002 0.07762 0.01016 0.11564 0.20342 2004 0.10367 0.01000 0.12430 0.23796

269

1972. Furthermore, the maximum polarization contribution for Republicans (0.104) and

Democrats (0.124) is the last year in the time series, 2004. This suggests that not only has

polarization on abortion significantly increased over the course of the time series, but that

partisan polarization on abortion is an ongoing contemporary phenomenon.

Figure 9.8 plots the total partisan polarization from 1972 to 2004. Partisan polarization

throughout the 1970’s and well in to the 1980’s was relatively flat. From 1988 to 1990,

however, there is a substantial spike in partisan polarization on abortion. Abortion polarization

goes from 0.012 to 0.086—over a 700% increase in partisan polarization—in these two years.

Note that this time period synchs up well with the culture wars thesis. Except for 1994, which

witnessed a slight dip in partisan polarization on abortion, from 1988 forward partisan

polarization has monotonically increased biannually until the end of the time series in 2004.

Hence, the global maximum for partisan polarization on abortion is in 2004 (0.238).

Interestingly, while we might have expected partisan polarization to stall-out given the return of

foreign policy to prominence as one of the major political fault line in American politics, the

increase in partisan polarization during the first decade of the 21st century is comparable with the partisan polarization from the 1990’s. Since the 1970’s there has been a dramatic increase in partisan polarization on abortion.

The decomposition of partisan polarization by party identification category is reported in Figure 9.9. The dramatic increase in partisan polarization on abortion is apparent, with the two major parties as the biggest contributors to partisan polarization. A particularly interesting finding in the decomposition figure is the fact that it is not the case that Republicans and

Democrats were the two biggest contributors to partisan polarization since the 1970’s. Up until 1988, during the low-polarization period, it was sometimes the case that the Independents outstripped either the Republicans or Democrats in contribution to partisan polarization.

270

FIGURE 9.8: PARTY ID GROUP POLARIZATION ON ABORTION OVER TIME, 1972-2004 0.3

0.25

0.2

0.15

0.1

0.05

0 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

ABORTION GROUP POLARIZATION

FIGURE 9.9: DECOMPOSITION OF PARTISAN GROUP POLARIZATION ON ABORTION OVER TIME, 1972-2004 0.3

0.25

0.2

0.15

0.1

0.05

0 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

REPUBLICANS INDEPENDENTS DEMOCRATS

271

Indeed, in 1980, 1984, 1986, and 1988, the Independents were larger contributors to partisan

polarization than the Republicans. In 1988, though it is a low polarization survey year, the

Independents were the largest contributors to partisan polarization on abortion. This leads us

to 1990, the most unusual data point in the entire series, perhaps in all of the polarization

analyses. 1990 not only witnessed a dramatic increase in partisan polarization across the board,

but it was the Independents that were responsible for the largest contribution to this

polarization. While interesting, there is no obvious reason this would be the case. Though, as

noted in Chapter 7, this was in the wake of the 1989 Webster decision. After 1990, partisan polarization exhibits trends in the expected direction. By 1992, the ‘normal’ order, with

Republicans and Democrats the two primary contributors to partisan polarization, had reasserted itself.

Table 9.8 reports the decomposition and total partisan polarization trend models for abortion from 1972 to 2004. There is strong evidence of partisan polarization overall (R2 = .807)

with an average increase of 0.007 in polarization on abortion for every survey year. The

Republican and Democrat identifiers have identical parameter estimates and standard errors,

though the Democratic model explains slightly more of the variation in partisan polarization (R2

= .784) over the Republican model (R2 = .745). Independents contribute to the trend in

increasing partisan polarization, but not as much as the two major parties (0.001). The overall

fit is not nearly as good, explaining less than 50% of the variance in the dependent variable (R2 =

.455).

As I noted earlier, the trend in partisan polarization for the individual partisan categories

reflects a separation of the two major political parties that began in 1990 and has steadily

272

TABLE 9.8: TREND REGRESSIONS OF PARTY ID GROUP POLARIZATION ON ABORTION, 1972-2004 Trend: MODEL: Polarization Intercept Parameter Estimate 2 GP(I) = B0 + B1(YEAR) + E Y/N (S.E.) (S.E.) R N

Republican Y -5.648 0.003 *** .745 15 Identifiers (0.922) (0.001)

Independent Y -0.590 0.001 *** .455 15 Identifiers (0.181) (0.000)

Democrat Y -7.383 0.003 *** .784 15 Identifiers (1.080) (0.001)

Party ID Group Y -13.621 0.007 *** .807 15 Polarization (1.857) (0.001)

advanced through to 2004. The percent contributions of each partisan category for the weighted and unweighted group polarization measure can be found in Tables 9.8 and 9.9. The maximum contribution for Republicans was 52% in 1988 for the weighted polarization measure.

The unweighted maximum for Republicans was in 1992, where Republicans accounted for just below 50% of the percent contribution to partisan polarization on abortion. The low point came on the weighted contribution was, unsurprisingly, in 1976. Not only was this prior to the

Christian Right movement to influence the Republican Party, but it was also the first post-

Watergate presidential election year (Carter), depressing the number of Republican identifiers in the sample. That said, the weighted measure downgrades the Independent category contribution much more significantly in 1976 than for the Republicans, cutting their contribution by two thirds relative to the unweighted Independent contribution. What little partisan polarization there was in the 1970’s is mostly attributable to the Democrats, again, given their size within the population. The maximum and near maximum contributions for the Democratic identifiers come in 1976, 1980, and 1982 respectively. The Democrats averaged about 68% of

273

TABLE 9.9: PERCENT CONTRIBUTION & MEAN DEVIATION WEIGHTED PARTISAN G.P. ON ABORTION, 1974-2004 YEAR REP R-DEV IND I-DEV DEM D-DEV 1974 34.68% -1.05% 11.65% 1.95% 53.66% -1.05% 1976 17.05% -18.68% 15.05% 5.34% 67.90% -18.68% 1978 33.74% -1.99% 6.13% -3.57% 60.13% -1.99% 1980 16.56% -19.17% 15.84% 6.13% 67.62% -19.17% 1982 25.28% -10.45% 5.74% -3.97% 69.04% -10.45% 1984 22.43% -13.30% 10.27% 0.56% 67.31% -13.30% 1986 41.78% 6.05% 20.82% 11.12% 37.38% 6.05% 1988 52.28% 16.55% 17.62% 7.92% 30.08% 16.55% 1990 24.76% -10.97% 17.54% 7.84% 57.70% -10.97% 1992 53.17% 17.44% 7.42% -2.29% 39.41% 17.44% 1994 43.76% 8.03% 4.31% -5.39% 51.92% 8.03% 1996 34.94% -0.79% 3.80% -5.90% 61.26% -0.79% 1998 45.15% 9.42% 4.68% -5.02% 50.17% 9.42% 2000 30.54% -5.19% 6.11% -3.59% 63.35% -5.19% 2002 38.16% 2.43% 5.00% -4.71% 56.85% 2.43% 2004 43.57% 7.84% 4.20% -5.50% 52.24% 7.84%

TABLE 9.10: PERCENT CONTRIBUTION & MEAN DEVIATION UNWEIGHTED PARTISAN G.P. ON ABORTION, 1974-2004 YEAR REP R-DEV IND I-DEV DEM D-DEV 1974 35.05% 0.92% 29.17% 1.43% 35.78% -2.35% 1976 17.47% -16.65% 36.83% 9.09% 45.70% 7.56% 1978 41.67% 7.54% 16.67% -11.07% 41.67% 3.53% 1980 16.72% -17.41% 40.53% 12.80% 42.75% 4.61% 1982 30.95% -3.17% 20.24% -7.50% 48.81% 10.67% 1984 19.51% -14.61% 32.17% 4.44% 48.31% 10.17% 1986 31.44% -2.68% 48.62% 20.88% 19.94% -18.20% 1988 35.71% 1.59% 46.43% 18.69% 17.86% -20.28% 1990 19.55% -14.57% 48.34% 20.61% 32.10% -6.04% 1992 49.71% 15.58% 22.44% -5.29% 27.85% -10.29% 1994 40.35% 6.23% 16.73% -11.00% 42.91% 4.77% 1996 36.10% 1.97% 17.72% -10.02% 46.19% 8.05% 1998 46.00% 11.88% 17.62% -10.11% 36.38% -1.76% 2000 30.76% -3.36% 20.36% -7.38% 48.88% 10.74% 2002 37.32% 3.19% 18.21% -9.52% 44.47% 6.33% 2004 42.03% 7.91% 16.67% -11.06% 41.29% 3.15%

274

the contribution to partisan polarization for those years. As was the case with partisan polarization on the ideological dimension, weighting the partisan polarization measure on abortion substantially affects the contribution of Independents to polarization. In Table 9.9, the contribution of Independents to polarization, while sometimes eclipsing one of the major parties, is never the top categorical contributor to partisan polarization. This is not true of the unweighted contribution of Independents (Table 9.10). From 1986 to 1990, the Independent category accounted for nearly half of the partisan polarization on abortion.

I estimate linear trend models for the mean deviation in the percent contribution of each partisan category to the polarization on abortion (Table 9.11). In assessing the linear polarization trends, I omit the 1972 data point for both empirical and theoretical reasons.

Empirically, the 1972 polarization data point was an unusual outlier. While one should be cautious in eliminating data points, there is good reason to exclude the 1972 data on abortion.

Slope estimates of small-N regressions are particularly susceptible to the influence of outliers.

The ANES used a split-sample in 1972, and as such the abortion question was asked of only half of the 1972 sample. Further sub-setting the data by partisanship may have resulted in unrepresentative estimates of the partisan positions on abortion. Theoretically, 1972 preceded the 1973 Roe v. Wade USSC decision that sparked the cultural battle over abortion. By excluding 1972, the trend depicts the percent contribution for the categories (mean deviation) to abortion polarization since that landmark decision.

For the weighted polarization measures, the linear models reveal interesting trends in the contribution of Republican identifiers to the abortion issue. There is a statistically significant positive coefficient for the Republican model, indicating that Republican contribution to abortion polarization has increased since the 1970’s. For every survey year, there is a 0.596

275

increase in the percent contribution for Republicans to abortion polarization. This model

accounts for about 25% of the variation in percent contribution for the time series. While

Republican identification has increased since the 1970’s, this trend is not exclusively a

consequence of that fact. Republican contribution to the abortion polarization measure

increased in the unweighted model as well (0.504). Interestingly Democrats, when you account

for their declining adherents since the 1970’s up until 2004, evidence no significant trend in

their contribution to abortion polarization. This is also the case when it comes to the

unweighted model (R2 = .003). Substantively speaking and ignoring the lack of statistical significance for the Democratic coefficient, there is clearly a much smaller increase in magnitude when compared to that of Republicans. It seems clear that Republicans have contributed more to abortion polarization than Democrats have since the 1970’s. Irrespective of whether we weight the groups by size, Republicans evidence a significant increasing contribution to abortion polarization while the Democrats do not. Those who suggest that Republicans were ‘captured’ by the Christian Right may have exaggerated their case, but clearly the influence of conservative religious citizens increasingly identifying with and influencing the Republican Party has had an effect on their collective position on abortion. For Independents, they have contributed less to abortion polarization since the 1970’s in both the weighted (-0.314) and unweighted regression

(-0.564) models. Recall that Independents were significant contributors to abortion polarization in the late 1980’s and early 1990’s but after 1992 there is a sharp decline in the contribution of

Independents, and this trend remains constant through 2004. The regression coefficients for

Independents in Table 9.11 capture this change. The weighed model explains just over 25% of the variation in percent contribution, while the unweighted model accounts for 22% of the change in the Independent contribution to abortion polarization.

276

TABLE 9.11: WEIGHTED & UNWEIGHTED MEAN DEVIATION TREND REGRESSIONS OF PID GP ON ABORTION, 1974-2004 YEAR MODEL: Intercept P.E. 2 GPMD(PID) = B0 + B1(YEAR) + B2(PPA) + E (R.S.E.) (R.S.E.) R N WEIGHTED GROUP POLARIZATION

Republican -1186.546 0.596 *** .244 16 Identifiers (414.440) (0.208)

Independent 623.886 -0.314 ** .264 16 Identifiers (278.130) (0.140)

Democrat 562.967 -0.283 .005 16 Identifiers (381.646) (0.192)

UNWEIGHTED GROUP POLARIZATION

Republican -1003.165 0.504 ** .219 16 Identifiers (439.957) (0.221)

Independent 1121.281 -0.564 *** .190 16 Identifiers (400.694) (0.201)

Democrat -118.116 0.060 .003 16 Identifiers (547.511) (0.157

AN EXAMPLE OF PARTISAN POLARIZATION ON ABORTION AND PUBLIC POLICY: THE PBA BAN

Unlike with ideology, there are no relatively objective measures of the policy outputs for

Congress on abortion or the abortion policy positions for congresspersons for the full time series. While interest group ratings could serve as a proxy for abortion positions for congress, I do not include an analysis with that data here (though it is an avenue of future research). What can be done, however, is to look at a suggestive example of the partisan polarization on abortion and the effect this has had on abortion public policy. One such suggestive example is partial birth abortion, a highly controversial abortion procedure that exists at the extremes of abortion debate in the United States.

277

FIGURE 9.10: 2003 ABC NEWS OPINION POLL - PUBLIC OPINION ON ABORTION BY SITUATION67 100 90 80 70 60 50 40 30 20 SHOULD BE LEGAL 10 SHOULD BE ILLEGAL 0

In 2003, an ABC News opinion poll attempted to assess the nature of abortion opinion across a range of possible situations where a woman might seek an abortion (Figure 9.10). It included the normal abortion ‘exceptions’ of life, health, rape, and incest, but it also included newer abortion issues such as abortions of physically impaired babies and partial birth abortion.

Partial birth abortion is an abortion procedure involving a partial delivery of the fetus in order to complete the abortion, hence the term. It has been highly controversial since it became a target of the abortion debate, with the Partial Birth Abortion Ban Act of 2005 having termed the procedure “gruesome and inhumane.” Whatever the merits of the argument, as the 2003 ABC

News poll illustrates, the American public is aligned in near-consensus against it. Nearly 70% of respondents to the poll (69%) said they believed the procedure should be illegal. Only 21% reported that it should be legal. While the public has long been opposed to abortions in

67 ABC News Poll conducted from January 16th to January 20th, 2003. N = 1,133 adult sample with a +/- 3 point margin of error. Respondents were presented with several situations where an abortion decision could be made, and asked whether they thought abortion “should be illegal” or “should be legal” in those situations.

278

“unwanted child” situations, that position is relatively conflictual, with over 55% of the public aligned in opposition but 41% of the public in favor of permitting abortions in that situation.

That is an over 10% difference on the ‘opposed’ side and an over 20% difference between respondents who believe partial birth abortion should be legal and those who support the legality of ending unwanted pregnancies.

In a depolarized, non-culture war political environment, we would not anticipate a great deal of controversy associated with the PBA ban. With the public solidly behind the ban, a major controversy over its enactment should be unlikely in a non-partisan, depolarized abortion issue dimension. As Table 9.12 illustrates, however, this is not the case. The table gives the partisan breakdown and caucus percentage for the 2005 vote on the Partial Birth Abortion Ban

Act. It was voted on in both the House and the Senate, with the Harkin Amendment the only substantive difference between the two bills. The Harkin Amendment expressed support for the

Roe v. Wade decision (which also enjoys majority support in polling). That amendment was

TABLE 9.12: 2005 CONGRESSIONAL VOTES ON PARTIAL BIRTH ABORTION BAN ACT68 Congressional Against Parties For PBA Ban Caucus % PBA Ban Caucus % HOUSE Republicans 218 98.21% 4 1.80%

Democrats 63 31.34% 138 68.66% SENATE Republicans 47 94.00% 3 6.38%

Democrats 17 35.42% 31 64.58%

68 Data culled from reports on House Roll Call No. 530 and Senate Roll Call No. 402. Twelve House members and two senators did not participate in the vote, and are excluded from the calculation of caucus percentage.

279

eliminated in conference. That in itself is suggestive of the Figure 9.5 scenario, as Republicans rejected a Democratic amendment which might have induced more bipartisan support for the bill.

The vote illustrates clearly that even on a bill which reflects a strong, consensus position for the American people, the behavior of Republicans and Democrats in Congress reflects the evidence of strong polarization on the issue between their constituencies that this chapter has demonstrated. In the House 98% of Republicans supported the ban and 94% in the Senate did likewise. While this outstrips the level of public opposition, at least as was evident in the ABC poll from two years prior, it is fair to say that 100% of rational congresspersons should follow where 70% of the American public leads. However, that is not the case with the Democrats.

Strong majorities of the Democratic caucuses in both the House and the Senate opposed the ban. While the Democratic caucus was certainly more split than the Republicans (with over 30% defections in both houses), the fact over 65% of all the Democrats in Congress voted against the ban is a testament to extent of partisan polarization on Abortion. While some might attribute this to the influence of pro-choice interest groups such as NARAL, the evidence in this chapter presents a more electorally-plausible, constituent-centered explanation. Democrats opposed the ban because their constituents tilt strongly towards the Pro-Choice position. Republicans likewise on pro-life issues, though in this case they also had a majority of the public to rely upon in justifying their votes. While one shouldn’t go too far in over-interpreting a single vote from

Congress or a single poll of the public on partial birth abortion, the vote and the poll are consistent with partisan polarization on abortion and consistent with representatives to

Congress responding to that polarization in the policy process. One way to think about it is to do so in terms of a counterfactual: how does this abortion vote comport with what we could reasonably expect if the Republicans and Democrats in Congress were responding to the

280

‘average’ voter and the strong consensus in the public against partial birth abortion or is it inconsistent with it? I would argue we would expect a near consensus on a partial birth abortion ban in the absence of partisan polarization on abortion. Since the parties in the mass public have diverged on abortion, we see this reflected in the partisan polarization on the PBA vote.

CONCLUSION

In this chapter I present compelling evidence of significant partisan polarization in public attitudes on the ideological dimension. I tie that ideological polarization at the mass electorate level directly to the voting behavior of members of Congress, a part of the American political elite. The relationship between partisan polarization on ideology at the mass level and partisan polarization level is strong, and the correspondence of polarization on ideology between the mass and elite levels is quite striking. I explore the nature of this relationship in terms of which way the causal arrow points in Chapter 12.

In terms of the contributions of identifiers of either party on ideological polarization, I show that the Democrats have had an increasing role in the polarization of ideology, irrespective of the number of Democratic identifiers in the electorate. At the same time, Republicans had a corresponding decline in their contribution to ideological polarization. However, both of these trends are mediated by group size, as weighting the measures by group size revealed no significant trend for either Republicans or Democrats. The significant trend in the weighted group polarization measure of party identifiers on ideology was for the Independents, evidencing a declining contribution of polarization. As noted earlier, Independents have declined as a portion of the population since the 1970’s. Interestingly, Republican and

Democratic contribution to partisan polarization on ideology is directly related to the party of

281

the presidential administration. When a Democrat is president, Republicans contribute more to partisan polarization on the ideological dimension. When a Republican is president, it is the

Democrats in the mass public who shift to the extremes on ideology. This reaction to the presidential administration suggests that, for those citizens who identify with one of the two major parties, dissatisfaction with their party being out of power tends to induce an aggregate move towards the ideological extremes.

On abortion, as with ideology, I find strong evidence of partisan polarization in the

American public. While the 1970’s and most of the 1980’s exhibited little partisan polarization on abortion, with the contribution of both parties to polarization sometimes eclipsed by that of

Independents, this all changes in the latter half of the 1980’s. From 1988 forward, I find a substantial increase in partisan polarization on abortion that increases nearly monotonically until the end of the time series in 2004. Both Republican and Democratic identifiers were strong contributors to this polarization trend, though Republicans outstripped Democrats in percent contribution to abortion polarization in both the weighted and unweighted group polarization trend models. The case of the Partial Birth Abortion Ban Act of 2005 suggests this polarization at the level of the mass electorate has translated into polarized partisan behavior in Congress.

Despite a consensus in the American public against partial birth abortion, over 65% of

Democrats in both houses voted against the ban.

282

CHAPTER 10: GROUP POLARIZATION OF PUBLIC ATTITUDES II: CLASS & RELIGIOSITY POLARIZATION

“NES America is not a place that I recognize. It might as well be the moon…Working-class conservatism exists. Yes, the phenomenon is complex, but nevertheless it is an important, if not the preeminent reason for the continuing electoral weakness of the Democratic Party.” – Thomas Franks

“Americans increasingly vote as they pray, or don't pray [at all].” – Michael Barone

In Chapters 5, 6, 7, and 8, I established that significant political polarization has occurred at the level of the mass public over the last three decades in the dimensions of political conflict in American society. In Chapter 9 I examined partisan polarization, an important phenomenon in American politics. Using the group polarization measure I have developed, I examined the grouped party identifiers and found that the parties have become increasingly ideologically coherent and polarized on social issues, like abortion. While the previous analyses have examined polarization globally, at the level of the mass electorate, it is important to examine polarization from the perspective of the social and political groups that compose the mass public and structure the partisan and political competition. Though polarization is likely operative at the primary (kin) group level, political polarization is most relevant to the competition and conflict between secondary groups (peer groups, interest groups, demographic groups, etc.), as it is these groups that both serve to inform and motivate individual citizens to become politically active and to which formal political organizations (i.e. political parties) structure their appeals of allegiance in exchange for policy concessions.

SOCIAL GROUPS: ORGANIZING AND DRIVING POLITICAL POLARIZATION

Social groups are foundational human associations upon which societies are built and changes in the political activity and partisan affiliation of groups is a fundamental force for

283

change in our partisan and electoral systems, the issue environment of the public policy agenda, and the conflict within society irrespective of whether that influence serves pluralist ideals. A full group-based analysis would involve examining all or most all of the relevant social groups that compose the party coalitions and the constituents of elected officials. The complete group-based assessment is beyond the scope of this study; however I will look at two groups specifically relevant to the debate over polarization in American politics: class and religiosity. I will examine political polarization in two distinct but related dimensions (partisanship and ideology) with respect to two these major social groupings using the following classifications: levels of income, and levels of religiosity.

Partisanship is the most significant vehicle for political conflict in American society and thus polarization of groups along a partisan dimension reflects 1) the degree to which these groups have become ‘politicized’ and thus have become active participants in partisan politics 2) the degree to which party elites must be responsive to the issue and policy concerns of the groups that have increasingly identified with their side of the partisan divide.

The larger ideological dimension is, as I noted in Chapter 9, a useful proxy for the political issue space as well as a dimension on which we have reliable data for the breadth of the time series. Dispersion of the classes on ideology would suggest a shift in the politics of citizens with differing social and economic status. Particularly if the poor have become more conservative since the 1970’s it may indicate a successful appeal to that demographic in non- economic dimensions (i.e. the social issue dimension). Furthermore, if the religious and secular citizens have become ideologically divergent, it suggests an increasing role for religiosity in politics and a new source of social conflict, just as Hunter posited in his original take on the

284

culture wars. In order to study political polarization and the culture wars the relationship between social groupings and ideology must be at the forefront of the empirical analysis.

The evidence presented here on group polarization speaks to two fundamental questions on political polarization: 1) Has the poor in America been influenced by the emerging culture war and consequently associated themselves increasingly with the Right and the

Republican Party, as Thomas Frank argues? Have the poor been duped into taking the Culture

War opiate? 2) Is religion playing a larger role in American politics? Is your propensity to attend church a better predictor of your political beliefs today than a few decades ago?

Class is a relevant social characteristic both as a source of social identification and association in the American public, and thus of intrinsic importance in political competition, and as a potential source of political and partisan polarization. Class structures the political and partisan inclinations of the public. Issues related to income levels and social class define and distinguish the political parties and are a major source of political conflict in most if not all economically advanced societies. Furthermore, class is of particular relevance to the culture war. For those who believe that class is the be-all and end-all of electoral choice and partisan affiliation, social issues represent a confounding and even immoral basis for structuring partisan competition. Beyond the normative affront some take to the possibility of social issues trumping class in partisan and political choices by the mass public, there is the question of whether it does. There is an empirical question: have social issues increasingly trumped class for voters and partisan identifiers? If they have, then given the widely accepted division on social issues among partisan and political elites, we would expect the classes to respond accordingly. If poor and blue collar, working class voters have prioritized social issues above and

285

beyond their economic concerns, then this should reflect in the relative average partisan identifications for these classifications.

Finally, I consider the fundamental empirical question that goes directly t o the existence of the culture war itself as Hunter originally envisioned it. Hunter’s formulation of the culture wars suggested a growing schism between religious citizens and secular citizens in

American society. Given the uncompromising nature of divisions on moral questions like abortion and gay rights, this growing division, reflected in our elected institutions, would lead to increasing conflict: a culture war. Just as with the class-related polarization issues, Hunter’s thesis presents a specific empirical question that requires further examination and explication: is there an increasing partisan and/or ideological divide between the religious members of the mass public and the secular identifiers? Has growth in the size and political coherence of either side of the divide contributed to political conflict? I test these questions directly by grouping the mass public in terms of religiosity in order to assess the trends in the polarization (or depolarization) of the religious versus the secular on the ideological and partisan dimensions.

Using the group polarization measure I developed in Chapter 3 and defined for empirical analysis in Chapter 4 (both weighted and unweighted group polarization), I will attempt to answer these questions on political polarization by assessing whether these groups have

‘polarized’ on partisan and ideological dimensions over the course of the past three decades.

Have the poor shifted towards the Republican Party given the increasing social conservatism of

Republican political elites and activists despite the apparent incongruence in economic policy between the poor and the Republicans? In other words, have the poor been ‘duped’ into becoming Republicans because culture trumps economics for these voters? Have the religious and non-religious groups increasingly defined themselves in terms of the political dimension and

286

thus increasingly identified with one party over the other? Have the faithful diverged from

seculars on partisanship and ideology? A systematic assessment of the aggregation of attitudes

in these groupings and along these important political dimensions represents a substantively

important step towards a greater understanding of political polarization in the United States.

Data

The data for this analysis are culled from the American National Election Study (ANES)

cumulative file.69 I use the ANES studies from 1970-2004.70 The creation of the data set for analysis of polarization trends for the mass public is a two-step process. In the first step, univariate statistics are generated on the substantive variables from the ANES cumulative file.

Specifically, the means and frequencies for the variables were output. The second step involves creating a time-series data set with the means and frequencies for the relevant ANES variables for each group or category in the identification variable from the ANES and these are treated as individual variables themselves in the new data set. For example, let’s consider the party identification variable as a group identification variable and ideology as our ‘issue’ dimension variable. The party ID variable has three categories: Republicans, Independents, and Democrats.

69 The Cumulative Data File consists of variables derived from the 1948-2004 series of biennial ("time- series") SRC/CPS National Election Studies. The American National Election Studies / Time Series Studies are collected before and after presidential (pre and post surveys) elections. The off-year elections typically only have a post-election study. The ANES Cumulative Data File is a merged data set of all the time series studies from 1948-2004. The pooled data includes variables which appear in three or more studies and consists of 44,715 cases. 70 The data is sub-setted by year to include only studies from 1970-2004 as the previous data sets had few to none of the relevant substantive variables which are necessary for the polarization analysis. Furthermore, 1970-2004 covers the relevant time period to examine the culture wars thesis.

287

of the party ID categories. A time series data set was created that contains mean and frequency

variables for all of the variables relevant to the polarization analysis.

Variables

As mentioned above, there are two ‘types’ of variables used in this analysis. The first

type is a group-identification variable: an ordinal classification of the population along some

relevant dimension. The three group identification variables in this analysis are party

identification, income levels, and religiosity. The party ID variable used for classification is a

three-category variable that collapses strong partisans, weak partisan identifiers, and

independent leaners all into an aggregate ‘party’ category. So, for example, “Republicans” in this

variable are respondents who either identified themselves as strong Republicans, weak

Republicans, or Independents who lean towards the Republican Party. The income the

classification variable uses the five income categories created for the ANES time series.71 The religiosity group identification variable sorts respondents by their church attendance. This is a rough measure of the respondent’s dedication to his or her religion. The religiosity measure used here collapses the top two and bottom two attendance categories to create a four category measure of religiosity. On the top end, respondents who said they attended church every week or nearly every week were included in the same category. At the bottom of the church attendance scale, respondents who reported they never attend church (but do not claim to be atheist or agnostic) were combined with respondents who said they were either atheists or agnostics.

The two political dimension variables are party identification and ideology. The party identification variable in this chapter is used as a proxy for partisanship as a political dimension,

71 Income categories in the ANES cumulative file are defined thusly: category 1 – 0 to 16th percentile, category 2 – 17 to 33rd percentile, category 3 – 34 to 67th percentile, category 4 – 68 to 95th percentile, and category 5 is the 96 to 100th percentile.

288

and it is not the same one used for the partisan group classifications in Chapter 9 (which defines

partisanship using three categories: Republican, Independent, and Democrat). This party

identification variable is a five category variable with the weak partisan identifiers and

independent-leaners collapsed into the same category. Strong partisans and independents are

classified in separate categories.72

Data Presentation Organization

The organization of the data presentation in this chapter is the same as that in Chapter

9. I first present a decomposition table which reports the total group polarization score for the

grouping variable on the ideological or partisan dimension in addition to each categorical

component of that group polarization score. For example, the decomposition table of class on

the partisan dimension reports the total class polarization on partisanship as well as the

component scores for each class category. For each decomposition table there is a regression

table reporting the trends in polarization for total group polarization and the component

category group polarization. The second type of table is a contribution table. I report two contribution tables for each dimensional analysis. The first table reports the percent contribution and the mean deviation of that contribution on the weighted group polarization score—the group squared distances weighted by the size of the group. The second table reports the percent contribution and the mean deviation for the unweighted group polarization score.

For each set of contribution tables there is a regression table that reports trend models in the mean deviation for the percent contribution of each of the categories to the total group polarization score (both weighted and unweighted). I use data pictures to illustrate interesting and significant trends and relationships in group polarization.

72 PID2 – 1 = Strong Democrat, 2 = Weak Democrat / Independent Leans Democrat, 3 = Independent, 4 = Weak Republican / Independent Leans Republican, 5 = Strong Republican.

289

ARE SOCIAL ISSUES THE OPIATE OF THE WORKING CLASS?

In What’s the Matter With Kansas?, Thomas Franks argues that the working class of

America have been duped into supporting t he Republican party due to its conservative social issue positions (which, in fact, are just a convenient electoral ploy in order to pursue a conservative economic agenda at odds with working class interests). “The hallmark of backlash conservatism is that it approaches politics not as a defender of the existing order or as a genteel aristocrat but as an average working person offended by the arrogant impositions of the (liberal) upper class.” Franks believes that the Democrats have failed to earn working class support both as a function of the clever ‘framing’ of elections by Republicans as referendums on social issues and family values and as a consequence of the Democratic Party’s failure to emphasize economic liberalism (and thus appeal directly to the poor and working class interests). As the

Franks quote at the beginning of the chapter indicates, Franks argues that growing working class conservatism has lead to working class defections on social issues and cost the Democratic Party elections. Franks tells a gripping tale backed with anecdotes and narrative…but has Franks built it on sand?

Class Polarization on Partisanship

Table 10.1 reports the group polarization for the income classes and the decomposition of GP by each income class category from 1970 to 2004. It should be apparent from just a cursory examination of this table that little polarization has occurred. The polarization of the

Poor was 0.094 in 1970. It was 0.103 in 2004. The only consistent trend apparent is a relatively slight increase in the polarization of the richest of income classes on party identification. The poor, on the other hand, are just as polarized in 1970 (0.09) as they were in 2000 (0.09). There is very little difference between the polarization for the ‘working’ class category in 1970 (0.12) and that in 2000 (0.15), which is also the peak in group polarization for the “poor” category. For the

290

lower middle class category, we see the same pattern of group polarization that we did with the working class. There is a slight increase in group polarization in the later years (peaking in 1996 with a 0.43) but there is very little evidence of substantial polarization over the times series.

There’s even less polarization for the upper middle class (0.32 peak in 1994). Figure 10.1 shows the flat trend line for income group polarization on partisanship. But for the slight increase in polarization in the mid-nineties where it briefly tops a polarization score of 1, the average distance between income groups in terms of their partisan affiliation is essentially constant. The poor and working class party identification in the 1970’s is virtually indistinguishable from that of any other decade in the analysis. Though some of the larger polarization scores occur in the later years for the lower and upper middle class, the polarization trend for these categories is slight. Looking to the polarization on income as a class variable, the polarization peak occurs in

1996 (1.125) while the low was in 1972 (0.154). So there is some increase in polarization among the income classes. But this increase is negligible substantively speaking.

Table 10.2 gives the regression models for the polarization of each of the groups on party identification. The “poor” income group polarization coefficient lacks statistical significance and is virtually indistinguishable from zero. This is a substantial blow to the Franks thesis of a lower class duped into voting Republican due to social issues, his so-called “working class conservative.” While there are significant polarization trends for the other groups and for income overall, substantively there has been no substantial uptick in class polarization on partisanship, as is evident in Figure 10.1. The Class GP on partisanship indicates just a slight increase in polarization on party ID. Overall, there is only a 0.016 increase in polarization for every survey year, mostly a product of the moderate spike in class polarization in 1994 and

1996. Furthermore, the best fitted model for increasing polarization is not for the poor or

291

TABLE 10.1: INCOME CLASSES – DECOMPOSITION OF GROUP POLARIZATION ON PARTY ID, 1970-2004 Working Lower Mid Upper Mid YEAR Poor GP GP GP GP Rich GP GPINC 1970 0.09438 0.12519 0.11352 0.09167 0.01312 0.43789 1972 0.03755 0.02333 0.04043 0.02912 0.02401 0.15444 1974 0.09477 0.04546 0.33516 0.06899 0.02104 0.56542 1976 0.05475 0.10337 0.12182 0.06425 0.02964 0.37383 1978 0.04396 0.06263 0.10711 0.04328 0.02387 0.28085 1980 0.05055 0.12183 0.12028 0.09192 0.04982 0.43442 1982 0.12442 0.21887 0.13669 0.10871 0.06763 0.65632 1984 0.08535 0.23352 0.21842 0.11744 0.05739 0.71212 1986 0.06131 0.15555 0.12743 0.09538 0.05626 0.49592 1988 0.06883 0.10487 0.12285 0.09902 0.03117 0.42674 1990 0.04942 0.09408 0.15363 0.10716 0.02193 0.42622 1992 0.05706 0.15659 0.18771 0.10973 0.06332 0.57441 1994 0.15331 0.28548 0.40814 0.32424 0.11313 1.28431 1996 0.1223 0.26865 0.43336 0.17554 0.12521 1.12506 1998 0.04472 0.05284 0.30413 0.07385 0.04576 0.52131 2000 0.09404 0.45408 0.13425 0.14453 0.04303 0.86993 2002 0.09836 0.30295 0.17211 0.11063 0.05391 0.73795 2004 0.10268 0.15182 0.20996 0.07673 0.06478 0.60597

FIGURE 10.1: INCOME GROUP POLARIZATION ON PARTISANSHIP OVER TIME, 1970-2004 5 4.5 4 3.5 3 2.5 2 1.5 1 0.5 0

Income GP on Party ID

292

TABLE 10.2: TREND REGRESSIONS OF INCOME GROUP POLARIZATION ON PARTY ID, 1970-2004 Trend: MODEL: Polarization Intercept Parameter GP(PD) = B0 + B1(YEAR) + Y/N (S.E.) Estimate R2 N E (S.E.)

Poor N -1.822 0.001 .086 17 (1.598) (0.001)

Working Y -10.855 0.006 ** .277 17 (4.596) (0.002)

Lower Middle Y -9.533 0.005 * .202 17 (4.999) (0.003)

Upper Middle Y -5.500 0.003 * .195 17 (2.942) (0.001)

Rich Y -3.379 0.002 *** .324 17 (1.278) 0.000

Income Group Y -31.089 0.016 *** .329 17 Polarization (0.016) (0.006)

working class but rather for the “rich” (R2 =.325). Not only is the “rich” category the best fit, but the decline in goodness of fit for the polarization model declines almost monotonically as we move down the income scale. The worst fit is for the upper middle class category (unsurprising given the weak polarization scores apparent in Table 10.1). So while there is a polarization trend that meets the standards of statistical significance for four of the five categories, there is very little apparent income group polarization.

The weighted and unweighted contributions of each of the class categories to the group polarization measure on partisanship can be found in Tables 10.3 and 10.4 respectively. If one examines the weighted group polarization percent contributions for the class categories, it is

293

apparent that there has been relatively little change in the contribution of the Poor to polarization. And, in fact, that contribution peaked early in the time series (1972). Indeed, the first two years, 1970 and 1972, are the two largest contributions of the “Poor” category to group polarization. The lows are relatively recent (1992 and 1998) and the contribution of the Poor doesn’t exceed 17% at any point after the 1972 survey year. The working class numbers are fairly stable as well, except for the unusual contribution figure in 2000, where the working class contributed 52 percent of the group polarization measure for that year. The unusual years for the lower middle class were 1974 (59.28) and 1998 (58.345), all likely strongly influenced by variation in group size. For example, the lower middle class unweighted figure for 1974 was

33.69%, more than twenty percent less than the weighted number. The “Rich” are downgraded in their percent contribution due to their small size relative to the population, as is evident in

Table 10.3. No apparent trend is evident from 1970 to 2004 for the contribution of the rich to the group polarization measure on partisanship. It ranges from a low of three percent to a high of 11.47% in 1980.

The unweighted group polarization in Table 10.4 tells us how much each category contributes to polarization independent of their size (treating each group as if they are of equal size). To the extent there was any trend in the contribution to group polarization on partisanship by the poor it is a negative one, as again the largest contributions come in the

1970’s. For the most part, there is a flat contribution trend line for the poor to unweighted group polarization. It decline sharply from 1970 to 1980, and then stabilizes until a slight spike upward during the George W. Bush administration (Figure 10.2). But note this uptick doesn’t

294

TABLE 10.3: CLASS – WEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON PARTY ID, 1970-2004 WC-DEV LMC- UMC- YEAR POOR P-DEV WC LMC DEV UMC DEV RICH R-DEV 1970 21.55% 7.22% 28.59% 2.16% 25.92% -6.66% 20.93% 2.89% 3.00% -5.61% 1972 24.31% 9.98% 15.11% -11.32% 26.18% -6.41% 18.86% 0.81% 15.55% 6.94% 1974 16.76% 2.43% 8.04% -18.39% 59.28% 26.69% 12.20% -5.84% 3.72% -4.89% 1976 14.65% 0.31% 27.65% 1.22% 32.59% 0.00% 17.19% -0.86% 7.93% -0.68% 1978 15.65% 1.32% 22.30% -4.13% 38.14% 5.55% 15.41% -2.63% 8.50% -0.11% 1980 11.64% -2.70% 28.04% 1.62% 27.69% -4.90% 21.16% 3.12% 11.47% 2.86% 1982 18.96% 4.62% 33.35% 6.92% 20.83% -11.76% 16.56% -1.48% 10.30% 1.69% 1984 11.99% -2.35% 32.79% 6.36% 30.67% -1.91% 16.49% -1.55% 8.06% -0.55% 1986 12.36% -1.97% 31.37% 4.94% 25.70% -6.89% 19.23% 1.19% 11.34% 2.73% 1988 16.13% 1.80% 24.57% -1.85% 28.79% -3.80% 23.20% 5.16% 7.30% -1.31% 1990 11.59% -2.74% 22.07% -4.35% 36.04% 3.46% 25.14% 7.10% 5.15% -3.47% 1992 9.93% -4.40% 27.26% 0.83% 32.68% 0.09% 19.10% 1.06% 11.02% 2.41% 1994 11.94% -2.40% 22.23% -4.20% 31.78% -0.81% 25.25% 7.20% 8.81% 0.20% 1996 10.87% -3.46% 23.88% -2.55% 38.52% 5.93% 15.60% -2.44% 11.13% 2.52% 1998 8.58% -5.75% 10.14% -16.29% 58.34% 25.75% 14.17% -3.88% 8.78% 0.17% 2000 10.81% -3.52% 52.20% 25.77% 15.43% -17.15% 16.61% -1.43% 4.95% -3.66% 2002 13.33% -1.00% 41.05% 14.63% 23.32% -9.26% 14.99% -3.05% 7.30% -1.31% 2004 16.94% 2.61% 25.05% -1.37% 34.65% 2.06% 12.66% -5.38% 10.69% 2.08%

TABLE 10.4: CLASS – UNWEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON PARTY ID, 1970-2004 WC-DEV LMC- UMC- YEAR POOR P-DEV WC LMC DEV UMC DEV RICH R-DEV 1970 25.83% 10.20% 34.26% 7.41% 14.54% -3.48% 10.94% -0.75% 14.45% -13.37% 1972 18.03% 2.40% 18.22% -8.63% 10.02% -8.00% 10.00% -1.69% 43.78% 15.96% 1974 23.78% 8.15% 11.76% -15.10% 33.69% 15.67% 10.52% -1.17% 20.25% -7.56% 1976 19.24% 3.61% 26.34% -0.51% 19.03% 1.02% 10.00% -1.69% 25.36% -2.46% 1978 18.04% 2.41% 24.06% -2.79% 24.00% 5.99% 11.20% -0.49% 22.68% -5.14% 1980 10.80% -4.83% 29.08% 2.23% 12.35% -5.66% 13.49% 1.80% 34.28% 6.46% 1982 19.96% 4.33% 33.97% 7.12% 12.22% -5.79% 10.21% -1.48% 23.63% -4.19% 1984 14.12% -1.50% 33.27% 6.41% 15.51% -2.50% 11.39% -0.30% 25.69% -2.12% 1986 10.62% -5.00% 29.85% 2.99% 10.55% -7.47% 10.41% -1.28% 38.56% 10.75% 1988 11.72% -3.90% 25.85% -1.00% 11.28% -6.73% 10.86% -0.83% 40.27% 12.46% 1990 12.62% -3.01% 26.25% -0.60% 22.67% 4.66% 15.87% 4.18% 22.60% -5.21% 1992 11.07% -4.55% 25.07% -1.79% 18.10% 0.08% 10.43% -1.26% 35.32% 7.51% 1994 11.54% -4.08% 23.48% -3.38% 15.20% -2.82% 14.82% 3.13% 34.96% 7.15% 1996 10.95% -4.67% 24.14% -2.71% 17.40% -0.62% 10.07% -1.62% 37.43% 9.62% 1998 11.83% -3.79% 12.38% -14.48% 38.06% 20.04% 12.50% 0.81% 25.24% -2.57% 2000 14.51% -1.11% 43.41% 16.55% 10.93% -7.09% 13.32% 1.63% 17.83% -9.99% 2002 16.73% 1.10% 36.14% 9.28% 15.87% -2.14% 12.66% 0.97% 18.61% -9.20% 2004 19.86% 4.24% 25.86% -0.99% 22.85% 4.84% 11.72% 0.03% 19.72% -8.10%

295

FIGURE 10.2: UNWEIGHTED MEAN DEVIATION TRENDS IN AVERAGE PARTY IDENTIFICATION FOR RICH AND POOR CITIZENS 20.00%

15.00%

10.00%

5.00%

0.00%

-5.00%

-10.00%

-15.00%

-20.00% 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

RICH POOR

even reach the 1970’s level of percent contribution to group polarization for the poor. The two biggest categorical contributors in the unweighted scores are the working class on the one hand and the rich on the other. The peak for the rich was in 1988 (40.27%) while the peak for the poor was in 1970 (25.83%). It does appear that the rich contributed more in the 1990’s (local max = 40.27%) than they have in the 2000’s (local max = 19.72%). However, there is little by way of an overall trend in contribution to group polarization on partisanship since 1970 by the rich.

In Table 10.5 I report a multiple regression of the weighted and unweighted group polarization measure using survey year (time trend) and the party of the presidential administration as independent variables. As in previous analyses, the logic of including a presidential variable is to test whether the public polarizes in reaction to the party of the president, given their generally distinct policy positions and the consequence of having a president who shares or does not share your own partisanship in office.

296

TABLE 10.5: WEIGHTED & UNWEIGHTED MEAN DEV. TREND REGRESSIONS OF CLASS GP ON PARTY ID, 1970-2004 YEAR PPA MODEL: Intercept P.E. P.E. 2 GPMD(PD) = B0 + B1(YEAR) + E (R.S.E.) (R.S.E.) (R.S.E.) R N WEIGHTED GROUP POLARIZATION

Poor 444.693 -0.225 *** 2.921 ** .495 18 (144.655) (0.075) (1.092)

Working -725.653 0.363 7.365 .227 18 (481.588) (0.242) (4.858)

Lower Middle 194.606 -0.095 -9.416 ** .163 18 (498.873) (0.251) (4.570)

Upper Middle 101.041 -0.051 0.232 .021 18 (149.955) (0.076) (1.856)

Rich -14.732 0.008 -1.102 .031 18 (167.929) (0.084) (1.032)

UNWEIGHTED GROUP POLARIZATION

Poor 412.372 -0.208 ** 2.145 .296 18 (199.017) (0.100) (1.553)

Working -373.008 0.185 6.971 ** .216 18 (340.290) (0.189) (3.060)

Lower Middle -83.511 0.044 -5.792 * .138 18 (325.096) (0.164) (3.397)

Upper Middle -140.782 0.071 *** 0.488 .191 18 (43.607) (0.022) (0.762)

Rich 221.312 -0.092 -3.639 .049 18 (42.509) (0.210) (3.640)

297

The multivariate group polarization model for the weighted data reveals few significant

trends in the percent contribution of class categories to partisanship. The one significant trend

for the class categories is a decreasing trend in percent contribution for the poor (-0.225),

accounting for about a 50% reduction in error over the mean model (R2 = .495). This trend

holds up even when we hold group size constant in the unweighted models (-0.208), though the

model fit is not nearly as good (R2 = .216). Presidential administration partisanship is also a significant predictor of the poor’s percent contribution (mean deviation), with that contribution increasing in Republican administrations (2.91). The contributions of the poor were much larger during the Reagan and George H.W. Bush administrations than they were during the Clinton administration. However, as we can see in the unweighted model, this result is spurious in the sense that it is an artifact of the group size for the poor in this administrations rather than, as it might appear, a polarizing reaction to the presidential partisanship. In the ‘poor’ model, presidential partisanship falls out as a significant predictor when we treat the groups as equal in size and assess merely categorical contribution to the group deviations on partisanship.

One consistent effect of presidential partisanship in both the weighted and unweighted models is its affect on the contribution to group polarization on partisanship for the lower middle class. In the weighted model the percent contribution of the lower middle class decreased by 9.416 percent for every survey year during Republican administrations relative to the Democratic administrations. And even ignoring the relative sizes of the class groups, there was still a significant negative coefficient for presidential partisanship (-5.792). The upper middle class has increasingly polarized on partisanship in the unweighted model (0.071), though the fit isn’t particularly good (R2 = .191). Still, this could be a consequence of the southern white middle class shifting towards the Republican Party.

298

TABLE 10.6: INCOME CLASSES – DECOMPOSITION OF GROUP POLARIZATION ON IDEOLOGY, 1972-2004 Working Lower Mid Upper Mid YEAR Poor GP GP GP GP Rich GP GPINC 1972 0.05351 0.02027 0.03992 0.03103 0.017609 0.16234 1974 0.10327 0.18419 0.19969 0.13021 0.087964 0.70533 1976 0.10625 0.08668 0.10427 0.10427 0.082215 0.48367 1978 0.03133 0.04056 0.03745 0.03449 0.038062 0.18189 1980 0.20054 0.06373 0.14366 0.09998 0.079960 0.58787 1982 0.0468 0.03458 0.12065 0.06114 0.055696 0.31887 1984 0.06622 0.02790 0.04517 0.04658 0.021039 0.20690 1986 0.02395 0.01494 0.02102 0.01640 0.009678 0.08599 1988 0.12468 0.03594 0.08370 0.07116 0.023086 0.33857 1990 0.01129 0.02043 0.01931 0.02955 0.012304 0.09288 1992 0.01775 0.00510 0.00761 0.01219 0.002078 0.04472 1994 0.03852 0.09069 0.07372 0.13873 0.007440 0.34910 1996 0.03909 0.01562 0.04595 0.09779 0.004620 0.20306 1998 0.02057 0.01217 0.01389 0.01683 0.002655 0.06611 2000 0.03102 0.06990 0.05013 0.06880 0.044526 0.26437 2002 0.09672 0.05470 0.06227 0.05995 0.042407 0.31604 2004 0.16242 0.03949 0.07441 0.05110 0.040288 0.36771

Also showing a significant coefficient for presidential partisanship is the working class model (6.971), even though the trend variable does not reach statistical significance.

Furthermore, the positive coefficient indicates that the working class has contributed more to group polarization on partisanship during Republican rather than Democratic administrations. If we treat the class categories equally (as Franks certainly does), we find exactly the opposite relationship that Franks argues for in What’s the Matter with Kansas? The working class has become more Democratic: polarizing in response to Republican administrations (such as the administration of George W. Bush). The upper classes react in the opposite direction. In other words, there is no evidence that the culture wars and social issue polarization has altered the fundamental political commitments of the classes.

299

Class Polarization on the Ideological Dimension

The story doesn’t get in any better for Franks when we turn our attention to the ideological dimension. Table 10.6 reports the group polarization for income categories on ideology. The group polarization for class on ideology exhibits little to no apparent linear trend in the raw, total score. The peak polarization for class on ideology is in 1980 (0.588). The lower polarization scores come mostly in the middle of the time series, with a low in 1992 (0.044).

While the polarization in the 2000’s isn’t as large as that in the 1970’s, the classes were more polarized on ideology then than in the 1990’s. The polarization scores for the poor and working class groups are relatively stable with slight oscillations over the time period. However, there is an apparent uptick in the polarization for the ‘poor’ category during the George W. Bush administration. The peak contribution for the Poor was in 1980 (0.201), but the second largest contribution comes in 2004 (0.162).

Examining the regression models for ideology in Table 10.7, the trend among income groups indicates a slight decrease in ideological polarization over the past three decades. The only significant coefficient of polarization is for the “rich” category (-0.001), but for all of the categories the polarization coefficient is negative. These results give little support for the argument that the poor and working class citizens have become more conservative since the

1970’s. Rather than a lower income class of voters buying into the Republican message on the culture wars, the evidence on the ideological and partisan dimensions indicates that the political conflict between economic classes has been remarkably stable over the last thirty-some years.

Note that unlike with partisan polarization, many of the lower scores for ideology occur at points of time proximate to the identification of the culture wars and some of the more significant years for social issue polarization, as I have shown in previous analyses. For the

300

TABLE 10.7: TREND REGRESSIONS OF INCOME GROUP POLARIZATION ON IDEOLOGY, 1972-2004 Trend: MODEL: Polarization Intercept Parameter Estimate GP(I) = B0 + B1(YEAR) + E Y/N (S.E.) (S.E.) R2 N

Poor N 1.926 -0.001 .026 17 (3.024) (0.001)

Working N 3.143 -0.002 .114 17 (2.304) (0.001)

Lower Middle N 4.814 -0.002 .195 17 (2.578) (0.001)

Upper Middle N 1.296 -0.001 .022 17 (2.196) (0.001)

Rich Y 3.068 -0.001 ** .252 17 (1.396) (0.000)

Income Group N 14.247 -0.007 .130 17 Polarization (9.679) (0.005)

“poor” category, once of its lowest polarization scores was in 1998 (0.02), a start-off point for much of the polarization we have seen on gay rights, abortion, and partisanship. The “working” class category peaked in terms of polarization in 1974 (0.184). The negative trend is particularly apparent in the “rich” category, with this category contributing almost nothing to the overall polarization score from 1984 on. These negative trends are apparent in the parameter estimates reported in Table 10.7. The goodness-of-fit statistic reflects the absence of a polarization trend for all of the category models and the income group polarization score.

The best fit is for the “rich” category model, and that is only a modest .252. It is barely a twenty-five percent improvement over simply using the mean of the category. The model fit for

301

income as a whole classification is a mere .130. One of the poorest fits is the “poor” category

(pun intended) with an R-Square of .026 (essentially no improvement over the mean by using

the trend model). It isn’t much better with the “working class” model (R2 = .114). Note these are

the categories of citizens that Franks argues have trended strongly towards Republicans and

conservatives because of family values appeals and social issue positions of the Republican

Party.

There hasn’t been much of a trend in group polarization of class in the ideological

dimension. However, I examine the contributions of the class categories as a percentage of

total contribution and as a mean deviation from the global mean for the category in Tables 10.8

and 10.9. The maximum contribution to the weighted group polarization measure by the poor

does come in 2004, the election year that Franks focuses on in What’s the Matter with Kansas?

Unfortunately for Franks, the poor in 2004 where more ideologically polarized to the Left— not in the conservative direction. The lower contributions to polarization (which might be consistent with a conservative trend) are scattered and unpatterned throughout the time series.

The ‘Poor’ contribution was near 11% in 1990, 1994, and 2000. The trend in contributions and the relative size of that contribution in the weighted models for the poor is very similar to that in the unweighted models, indicating the size of the category as a proportion of the population is not a significant factor in the percent contribution of the poor to ideological polarization. The working class contributions also seem to be fairly randomly distributed throughout the past three decades, with a low in 1996 (7.69%) and a high in 2000 (26.44%), just four years later. The trends for both the weighted and unweighted measures in the working class contribution to group polarization of class on ideology are fairly consistent—consistently inconsistent.

302

TABLE 10.8: CLASS – WEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON IDEOLOGY, 1972-2004 WC-DEV LMC- UMC- YEAR POOR P-DEV WC LMC DEV UMC DEV RICH R-DEV 1972 32.96% 7.55% 12.49% -4.10% 24.59% 1.66% 19.11% -4.90% 10.85% -0.22% 1974 14.64% -10.77% 26.11% 9.53% 28.31% 5.38% 18.46% -5.55% 12.47% 1.41% 1976 21.97% -3.45% 17.92% 1.34% 21.56% -1.37% 21.56% -2.45% 17.00% 5.93% 1978 17.22% -8.19% 22.30% 5.71% 20.59% -2.34% 18.96% -5.05% 20.93% 9.86% 1980 34.11% 8.70% 10.84% -5.74% 24.44% 1.51% 17.01% -7.00% 13.60% 2.54% 1982 14.68% -10.74% 10.84% -5.74% 37.84% 14.91% 19.17% -4.84% 17.47% 6.40% 1984 32.01% 6.59% 13.48% -3.10% 21.83% -1.10% 22.51% -1.50% 10.17% -0.90% 1986 27.85% 2.44% 17.37% 0.79% 24.44% 1.52% 19.07% -4.94% 11.25% 0.19% 1988 36.83% 11.41% 10.62% -5.97% 24.72% 1.79% 21.02% -2.99% 6.82% -4.25% 1990 12.16% -13.26% 22.00% 5.41% 20.79% -2.14% 31.82% 7.80% 13.25% 2.18% 1992 39.69% 14.28% 11.40% -5.18% 17.02% -5.91% 27.26% 3.25% 4.65% -6.42% 1994 11.03% -14.38% 25.98% 9.39% 21.12% -1.81% 39.74% 15.73% 2.13% -8.93% 1996 19.25% -6.16% 7.69% -8.89% 22.63% -0.30% 48.16% 24.15% 2.28% -8.79% 1998 31.11% 5.70% 18.41% 1.82% 21.01% -1.92% 25.46% 1.45% 4.02% -7.05% 2000 11.73% -13.68% 26.44% 9.86% 18.96% -3.97% 26.02% 2.01% 16.84% 5.78% 2002 30.60% 5.19% 17.31% 0.72% 19.70% -3.23% 18.97% -5.04% 13.42% 2.35% 2004 44.17% 18.76% 10.74% -5.85% 20.24% -2.69% 13.90% -10.11% 10.96% -0.11%

TABLE 10.9: CLASS – UNWEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON IDEOLOGY, 1972-2004 WC- LMC- UMC- YEAR POOR P-DEV WC DEV LMC DEV UMC DEV RICH R-DEV 1972 27.28% 1.68% 16.81% 0.85% 10.50% -1.00% 11.32% -4.05% 34.09% 2.52% 1974 13.08% -12.52% 24.04% 8.08% 10.13% -1.37% 10.02% -5.35% 42.73% 11.16% 1976 23.00% -2.59% 13.61% -2.35% 10.04% -1.46% 10.01% -5.37% 43.34% 11.77% 1978 15.70% -9.90% 19.02% 3.06% 10.24% -1.25% 10.90% -4.48% 44.14% 12.57% 1980 30.07% 4.47% 10.68% -5.28% 10.35% -1.14% 10.30% -5.07% 38.60% 7.03% 1982 15.37% -10.23% 10.99% -4.98% 22.08% 10.58% 11.75% -3.62% 39.82% 8.24% 1984 34.16% 8.56% 12.39% -3.57% 10.00% -1.50% 14.09% -1.28% 29.36% -2.21% 1986 24.15% -1.44% 16.69% 0.73% 10.12% -1.37% 10.42% -4.95% 38.61% 7.04% 1988 28.16% 2.56% 11.75% -4.21% 10.19% -1.30% 10.35% -5.03% 39.56% 7.98% 1990 10.11% -15.48% 20.01% 4.05% 10.00% -1.49% 15.36% -0.01% 44.51% 12.94% 1992 47.11% 21.51% 11.16% -4.81% 10.03% -1.46% 15.85% 0.48% 15.85% -15.72% 1994 13.34% -12.26% 34.30% 18.34% 12.62% 1.13% 29.16% 13.79% 10.58% -21.00% 1996 25.47% -0.13% 10.21% -5.75% 13.43% 1.93% 40.84% 25.46% 10.05% -21.52% 1998 37.94% 12.34% 19.87% 3.90% 12.12% 0.62% 19.87% 4.49% 10.21% -21.36% 2000 11.87% -13.73% 16.56% 0.60% 10.12% -1.38% 15.72% 0.35% 45.74% 14.16% 2002 30.96% 5.36% 13.11% -2.85% 11.25% -0.25% 13.60% -1.78% 31.09% -0.49% 2004 47.38% 21.78% 10.14% -5.82% 12.22% 0.72% 11.77% -3.60% 18.49% -13.08%

303

The story for the poor is similar to that of the rich, though the relatively fewer number of ‘rich’ citizens means that there is a stronger group-size influence on its overall contribution to polarization than with the poor. There is greater variation in the percent contribution of the rich to class polarization on ideology. In the weighted measure, it was as high as 21% in 1978 and as low as 2% in 1994. The larger contributions of the rich do seem to have been located in the earlier survey years in the time series rather than the later years. Particularly in the 1990’s the rich were relatively small contributors to group polarization. The ideological position of the lower middle class is remarkably stable, ranging from 10 to just over 12 percent in every year since 1972. However, there is much more variation in their contribution when you take into account the size of the lower middle class, with 1982 standing out as a year where their size greatly enhanced their contribution to class polarization on the ideological dimension.

The paucity of significant intra-class polarization on ideology, if not apparent already, is definitive in the contribution trend models in Table 10.10. At the extremes of class, for both the poor and the rich, there has been no trend in contribution to class polarization on ideology for either the weighted or unweighted measures. In the weighted model, the only significant time trend is for the lower category of the middle class, and that is a negative trend (-0.208) indicating a declining contribution to polarization. There is an interesting juxtaposition of the upper middle class and the rich apparent in the unweighted trend models, where the upper middle class has contributed more to group polarization (0.345) while the rich have moved in the opposite direction, contributing less over time (-0.668).

The weighted percent contribution models and the unweighted percent contributions both have two categories that have significant trends in their contribution to polarization dependent upon the party of the presidential administration. While the upper middle class doesn’t have a significant trend in their percent contribution when you account for group size, it

304

TABLE 10.10: WEIGHTED & UNWEIGHTED MEAN DEV. TREND REGRESSIONS OF CLASS GP ON IDEOLOGY, 1972-2004 YEAR PPA MODEL: Intercept P.E. P.E. 2 GPMD(IDEO) = B0 + B1(YEAR) + E (R.S.E.) (R.S.E.) (R.S.E.) R N WEIGHTED GROUP POLARIZATION

Poor -365.275 0.183 3.424 .050 17 (496.563) (0.250) (4.962)

Working 34.869 -0.017 -0.111 .008 17 (295.027) (0.149) (0.023)

Lower Middle -78.177 -0.208 *** 3.195 ** .337 17 (187.781) (0.065) (1.572)

Upper Middle -486.527 0.248 -9.164 ** .369 17 (332.019) (0.167) (4.017)

Rich 206.961 -0.100 2.804 .082 17 (348.399) (0.175) (3.182)

UNWEIGHTED GROUP POLARIZATION

Poor -760.453 0.383 -1.721 .119 17 (513.068) (0.258) (5.517)

Working 121.082 -0.060 -3.287 .072 17 (227.478) (0.115) (3.486)

Lower Middle -38.469 0.019 0.158 .005 17 (86.300) (0.043) (1.010)

Upper Middle -641.478 0.325 ** -8.355 ** .450 17 (318.639) (0.135) (3.751)

Rich 1319.266 -0.668 *** 13.206 *** .538 17 (475.590) (0.252) (4.787)

305

is significant in the equal-size models and in both measures the upper middle class are greater contributors to polarization during Democratic administrations rather than Republican administrations (-9.164 weighted; -8.355 unweighted). Again, as mentioned earlier, this could be an aftershock of the southern white middle class becoming increasingly Republican since the

1960’s. The lower middle class are an increasing contributor to class polarization on ideology during Democratic administrations (3.195) and, perhaps not unsurprisingly so are the rich

(13.206). Given the differing policies on the top marginal tax rates between Democrats and

Republicans, there is a certain logic to the rich contributing more to polarization during

Democratic rather than Republican administrations on the ideological dimension.

CLASS CONCLUSIONS

While there are some interesting trends in the individual contributions of the classes to partisanship and ideology, there is little evidence of strong intra-class trends in ideological or partisan polarization, nor is there evidence of such trends for overall class as a group. The evidence on ideology and polarization all point to one conclusion: Frank’s story of unwitting poor and working class voters co-opted by the Republicans through, as Obama might put it, appeals on their guns and relation, is largely mythical. There’s nothing the matter with Kansas.

Or, at least, the working poor voting Republican based on the party’s stance on social issues is not the problem. If anything, it is the wealthier Americans who have driven what class polarization is apparent in partisanship and ideology.

306

RELIGIOSITY, SECULARIZATION AND THE CULTURE WARS

“This [rise of secularism], not the supposed right-wing religious revival that conservatives champion and liberals dread, is the newest new thing in American political life, and the trend that's likely to have the most impact on the culture wars over the next decade or so.” – Ross Douthat

In 1990, those reporting no affiliation with a religious tradition were 8% of the population. In 2009, a survey conducted by Trinity College indicates that number has risen to

15%, a near doubling of the atheist/agnostics in the country (Kosmin and Keysar 2009).

According to the survey most nones are neither atheists nor theists but rather agnonstics and deists (59%). The 1990’s witnessed 1.3 million Americans join the ranks of the “Nones” every year. This makes “Nones” the fastest growing segment of the national religious landscape.

Why have religiosity factors only recently exhibited partisan and political polarization, given the long religious tradition in North America? The answer: the growth of seculars. As we saw in the discussion on polarization theory, a key factor in polarization as an effective determinant of political conflict is identification. It isn’t enough to be divided, there must be a substantial proportion of the population located at those distant positions (DiMaggio, Evans, and Bryson

1996; Duclos, Esteban, and Ray 2004; Esteban and Ray 1994) As Hout and Fischer document, the proportion of Americans who reported no religious preference doubled from 7 percent to 14 percent in the 1990’s. Furthermore, the percentage of adults who had been raised with no religion increased from 2 percent to 6 percent (Hout and Fischer 2002). While they did not find a rise in religious skepticism (given that the no religion identifiers hold conventional religious beliefs), they are alienated from organized religion and there is a distinct political dimension, as the rise of the “no religion” identifiers occurred exclusively among liberals and moderates. As

Paul Waldman notes, research has shown that evangelicals who live near atheists become more politically conservative. “Whatever the answer is, the possibility does seem real for secularism to achieve a new awakening of its own as a political and social movement.” As he notes, non-

307

believers can now claim their first publicly open member of Congress (Pete Stark of California), and they have their own lobbying group (Waldman 2007).

This suggests that as secularism becomes more of an identity, it too could generate further The encroachment of secularism political activism. As atheists perceive themselves seems to push at least some as atheists, a self-identified group in an adversarial evangelicals to identify more closely with their own religious tribe, and to position relative to the larger religious out-group, vote accordingly. One secular person in your town is a lone lost the impetus to seek political redress of perceived soul; ten are a threat to your way of life. (Waldman 2007) inequities should become stronger. Indeed this reactionary polarization is exactly the sort that tends to lead to political conflict, as not only do the current majorities view the insurgent group as more of a threat, but the growing opposition group increasingly associate based on the group characteristic and feel a connection with those citizens who share it.

Again, as long as traditional social mores were consensus issues, it didn’t matter what its relative placement on the secular-religious scale was. With the growth of seculars (i.e. the increase in density), the religious skeptics and alienated believers were not merely distant from traditionalists on social issues, but they, according to alienation-identification polarization, increasingly identify with fellow seculars (Duclos, Esteban, and Ray 2004; Esteban and Ray 1994).

In other words, the greater the number of seculars in society the more likely they are to identify with one another as a non-religious group and the more likely that is to factor in to political conflict. Group identification promotes group-oriented behavior, and that includes the realm of partisan politics.

308

Religiosity Polarization on Partisanship

Table 10.11 demonstrates how increasing secularization has contributed to the culture wars.73 It reports the polarization scores for religiosity on party identification from 1970 to

2004. Note the near-absence of polarization on party identification in the 1970’s (0.022). Two factors contribute to this: 1) the public was much more religious at the time and thus the religiously observant was a politically diverse class of citizens, 2) the seculars were a minute portion of the public (accounting for less than 10 percent of respondents). This would all change over the course of the next three decades. By 2004,

TABLE 10.11: RELIGIOSITY – DECOMPOSITION OF GROUP POLARIZATION ON PARTY ID, 1970-2004

YEAR Every Week Twice a Twice a Year Never GPRelig Month 1970 0.01412 0.00189 0.00303 0.00322 0.02226 1972 0.02078 0.00710 0.03712 0.01992 0.08492 1974 0.06328 0.01577 0.01356 0.00961 0.10222 1976 0.03015 0.01163 0.01078 0.00717 0.05972 1978 0.00692 0.00505 0.00460 0.00738 0.02394 1980 0.01945 0.00816 0.01787 0.00913 0.05461 1982 0.01864 0.00385 0.00422 0.00269 0.0294 1984 0.04767 0.02241 0.02181 0.01434 0.10623 1986 0.00918 0.00688 0.01725 0.00471 0.03801 1988 0.00813 0.00894 0.00643 0.00401 0.02751 1990 1.23448 0.40747 0.62383 3.27935 5.54514 1992 0.04175 0.02460 0.01116 0.03913 0.11664 1994 0.13524 0.03334 0.02049 0.05550 0.24458 1996 0.11939 0.08447 0.06120 0.18933 0.45439 1998 1.74048 0.47939 0.56822 3.78839 6.57648 2000 2.02698 0.49852 0.51876 3.42353 6.46779 2002 2.79615 0.71137 0.47253 4.11316 8.09322 2004 1.62612 0.76093 0.50814 4.13745 7.03264

73 “Seculars” are defined differently than in the Trinity College survey. Here seculars are defined based on their church attendance (religiosity), whereas the seculars in the American Religious Identification survey conducted by Trinity College defined seculars explicitly in terms of their stated religious beliefs. The ‘no religious identification’ group of individuals is subsumed within the larger group of those who never attend church.

309

seculars were a significant portion of the population (coming in just under 30 percent) and, as is apparent in Table 10.11, major contributors to the polarization on party identification (4.137 in

2004). As argued by Wideman and as is apparent from the group polarization measure of religiosity on partisanship, the rise of the seculars has inspired political activity and thus party identification on the part of the religiously observant. The reaction of religious traditionalists to the rise of the seculars and their increasing identification with the Democratic Party has been a corresponding increasing identification with the Republican Party. Note that the other extreme category, those citizens who attend church every week, is the second most significant contributor to group polarization in 2004. If there is such a thing as political polarization, the polarization of church attendees (or non-attendees) on party identification in 2004 is a prime example.

The OLS models in Table 10.12 regress the trend in group polarization for each of the categories and for religiosity on time. The results are consistent with the religious/secular partisan polarization evident in the group polarization scores in Table 10.11. Each category of religiosity has significantly contributed to polarization on party identification. The regression coefficients for each of the categorical models are significant at the .01 level and on average all of the models explain about half of the variation in religiosity group polarization on party identification. Furthermore, the most or second-most significant contributor to the religiosity polarization on party ID—both in terms of model fit (.537) and the size of the coefficient

(0.118)—are the seculars (those who never attend church). While the “Twice a Month” model beats the seculars in overall explanatory value, the coefficient for the seculars is more substantively significant. Again, note that the two categories that are the most substantial contributors to group polarization are the two extreme categories. However the seculars

(0.118) out ‘polarize’ the religiously religious by about two fold (0.061). This very much

310

TABLE 10.12: TREND REGRESSIONS OF RELIGIOSITY GROUP POLARIZATION ON PARTY ID, 1970-2004 Trend: MODEL: Polarization Intercept Parameter GP(PD) = B0 + B1(YEAR) + Y/N (S.E.) Estimate R2 N E (S.E.)

Every Week Y -122.326 0.061 *** .525 18 (28.471) (0.014)

Twice a Month Y -37.806 0.019 *** .586 18 (7.983) (0.004)

Twice a Year Y -30.776 0.016 *** .472 18 (8.184) (0.004)

Never Y -234.337 0.118 *** .537 18 (54.609) (0.027)

Religiosity Group Y -424.245 0.215 *** .548 18 Polarization (96.795) (0.049)

conforms to what one would expect of a ‘culture war’ between the religious and secular members of society. Turning to the percent contribution of each of the religiosity groups to partisan polarization, the significant trend in secular contribution to partisan polarization is more than apparent in both the weighted and unweighted group polarization measures (Tables

10.13 & 10.14). The last four years of the time series (1998, 2000, 2002, & 2004) all exceed the mean secular contribution to partisan polarization by 20% or more. While the weighted measure downgrades the secular contributions from 1970 to 1986, after that year the weight has some but relatively little impact on the percent contribution of the secular group. Figure

10.3 shows the weighted and unweighted partisan polarization trends in percent contribution for the ‘never’ church attendees (i.e. seculars). From 1970 to 1986 the weighted measure downgrades the contribution of seculars to partisan polarization, while after 1986 through 2004

311

TABLE 10.13: RELIGIOSITY – WEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON PARTY ID, 1970-2004

YEAR EW EW-DEV TM TM-DEV TY TY-DEV NEV NEV- DEV 1970 63.43% 25.55% 8.49% -6.39% 13.61% -3.98% 14.47% -15.19% 1972 24.47% -13.41% 8.36% -6.52% 43.71% 26.12% 23.46% -6.19% 1974 61.91% 24.02% 15.43% 0.55% 13.27% -4.32% 9.40% -20.25% 1976 50.49% 12.60% 19.47% 4.59% 18.05% 0.46% 12.01% -17.65% 1978 28.91% -8.98% 21.09% 6.21% 19.21% 1.63% 30.83% 1.18% 1980 35.62% -2.27% 14.94% 0.06% 32.72% 15.13% 16.72% -12.93% 1982 63.40% 25.52% 13.10% -1.79% 14.35% -3.24% 9.15% -20.50% 1984 44.87% 6.99% 21.10% 6.22% 20.53% 2.94% 13.50% -16.15% 1986 24.15% -13.73% 18.10% 3.22% 45.38% 27.79% 12.39% -17.26% 1988 29.55% -8.33% 32.50% 17.62% 23.37% 5.78% 14.58% -15.08% 1990 22.26% -15.62% 7.35% -7.53% 11.25% -6.34% 59.14% 29.49% 1992 35.79% -2.09% 21.09% 6.21% 9.57% -8.02% 33.55% 3.90% 1994 55.29% 17.41% 13.63% -1.25% 8.38% -9.21% 22.69% -6.96% 1996 26.27% -11.61% 18.59% 3.71% 13.47% -4.12% 41.67% 12.02% 1998 26.47% -11.42% 7.29% -7.59% 8.64% -8.95% 57.61% 27.95% 2000 31.34% -6.54% 7.71% -7.17% 8.02% -9.57% 52.93% 23.28% 2002 34.55% -3.33% 8.79% -6.09% 5.84% -11.75% 50.82% 21.17% 2004 23.12% -14.76% 10.82% -4.06% 7.23% -10.36% 58.83% 29.18%

TABLE 10.14: RELIGIOSITY – UNWEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON PID, 1970-2004

YEAR EW EW-DEV TM TM- TY TY-DEV NEV NEV- DEV DEV 1970 - 47.64% 21.81% 15.57% 12.22% 13.23% -5.03% 23.47% -4.64% 1972 16.50% -9.33% 18.07% -9.73% 34.34% 16.08% 31.09% 2.98% 1974 40.24% 14.41% 33.51% 5.72% 12.64% -5.62% 13.61% -14.51% 1976 33.12% 7.29% 35.14% 7.35% 16.26% -2.00% 15.46% -12.66% 1978 17.73% -8.11% 38.91% 11.11% 14.05% -4.21% 29.30% 1.19% 1980 22.84% -2.99% 34.78% 6.98% 25.49% 7.24% 16.89% -11.22% 1982 44.59% 18.76% 27.93% 0.13% 14.99% -3.27% 12.55% -15.56% 1984 30.64% 4.81% 37.58% 9.78% 17.24% -1.02% 14.53% -13.58% 1986 14.84% -10.99% 33.62% 5.83% 37.83% 19.57% 13.68% -14.43% 1988 17.75% -8.08% 49.67% 21.87% 17.91% -0.34% 14.72% -13.39% 1990 16.24% -9.59% 14.79% -13.01% 19.45% 1.19% 49.52% 21.41% 1992 23.32% -2.51% 36.78% 8.99% 15.68% -2.57% 24.21% -3.90% 1994 38.44% 12.61% 28.34% 0.54% 14.44% -3.82% 18.78% -9.34% 1996 16.66% -9.17% 31.01% 3.21% 18.20% -0.05% 34.13% 6.01% 1998 20.14% -5.69% 14.77% -13.02% 15.74% -2.52% 49.35% 21.24% 2000 23.08% -2.75% 14.38% -13.42% 14.33% -3.93% 48.21% 20.10% 2002 23.16% -2.67% 15.35% -12.44% 13.50% -4.75% 47.98% 19.87% 2004 18.02% -7.81% 20.13% -7.66% 13.29% -4.96% 48.55% 20.44%

312

FIGURE 10.3: WEIGHTED & UNWEIGHTED MEAN DEVIATION OF AVERAGE PARTY ID FOR SECULAR CITIZENS 40.00% 35.00% 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00% -5.00% -10.00% -15.00% -20.00% -25.00% -30.00% -35.00% -40.00% 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

WEIGHTED SECULARS UNWEIGHTED SECULARS

FIGURE 10.4: WEIGHTED MEAN DEVIATION OF AVERAGE PARTY ID FOR RELIGIOUS & SECULAR CITIZENS 40.00% 35.00% 30.00% 25.00% 20.00% 15.00% 10.00% 5.00% 0.00% -5.00% -10.00% -15.00% -20.00% -25.00% -30.00% -35.00% -40.00% 1970 1972 1974 1976 1978 1980 1982 1984 1986 1988 1990 1992 1994 1996 1998 2000 2002 2004

VERY SECULAR VERY RELIGIOUS

313

it weights its contribution more heavily, a consequence of the growing number of seculars in

America. Note that, while the weighting increases the contribution to partisan polarization, the larger component of the trend is a significant shift in the average partisan identification of seculars towards the Democratic Party. In other words, seculars are responsible for a significant portion of the religiosity partisan polarization.

On the other side of the religiosity dimension, the regular church attendees have contributed less and less to partisan polarization since the 1970, though they remain the second largest contributor to partisan polarization of the religiosity groups (second to the seculars).

The peak contribution of the ‘every week’ church attendees was in the first year of the time series, 1970 (63.43%) in the weighted measure. The second lowest level of contribution to partisan polarization for this group is in the last year of the time series, 2004 (23.12%). It trails the global minimum in 1990 by only about half a percent (22.26%). From 1986 on, the percent contribution of the regular church attendees falls below the mean contribution for the full time- series. This is less a function of declining polarization among the very religious and more a function of the growing contribution of the seculars.

The secular-religious trends in weighted group polarization on partisanship are illustrated in Figure 10.4. While there is a great deal of year-to-year variation for both the religious and secular groups, the overall trends noted in Tables 10.13 and 10.14 are evident.

The “very religious” (i.e. every week church attendees) are big contributors to partisan polarization for the religiosity groups in the early part of the time series, while the seculars take over in the 1990’s and by 1996 are dominating the very religious group in partisan polarization.

There is an interesting action-reaction relationship between secular and religious polarization evident in 1988 (Figure 10.4). As mentioned earlier, the 1990’s witnessed a remarkable spike in the number of secular-identifying citizens in the United States. In 1990, the seculars reach their

314

TABLE 10.15: WEIGHTED & UNWEIGHTED MEAN DEV. TREND REGRESSIONS OF RELIGIOSITY GP ON PID, 1970-2004 YEAR MODEL: Intercept P.E. 2 GPMD(PD) = B0 + B1(YEAR) + E (R.S.E.) (R.S.E.) R N WEIGHTED GROUP POLARIZATION

Every Week 1314.372 -0.661 ** .229 18 (565.597) (0.284)

Twice a Month 212.029 -0.107 .028 18 (248.855) (0.125)

Twice a Year 1173.078 -0.590 *** .282 18 (420.272) (0.211)

Never -2699.192 1.358 *** .590 18 (387.778) (0.195)

Culture War -5398.672 2.717 *** .590 18 (EW – NEV) (775.517) (0.390)

UNWEIGHTED GROUP POLARIZATION

Every Week 837.020 -0.421 * .180 18 (425.461) (0.214)

Twice a Month 574.398 -0.289 .084 18 (449.831) (0.226)

Twice a Year 358.608 -0.178 .070 18 (291.469) (0.146)

Never -1767.007 0.889 *** .425 18 (422.639) (0.213)

Culture War -3532.033 1.778 *** .424 18 (EW-NEV) (845.840) (0.426)

315

peak in contribution to partisan polarization. This is followed by a ‘reaction’ peak contribution by the very religious in 1994. Which itself gave way to the secular domination of partisan polarization from 1998 forward.

Returning our attention to Tables 10.13 and 10.14, the middle religiosity categories show no increasing partisan polarization. Both of the middle categories show declining contributions in the weighted measure—registering below-10% contributions on multiple years after 1988. The unweighted contributions, which are fairly stable over the time series suggests this decline is a consequence of fewer citizens identifying as semi-regular or even occasional church attendees. The maximum percent contribution to the weighted partisan polarization measure for the month church attendees was in 1988 (49.67%). The largest contribution on weighted group polarization for the ‘once or twice a year’ church attendees group was in the prior survey year, 1986 (45.38%). Both minimums come in at 1990 (7.35%) or later (7.23% in

2004).

The percent contribution regression models for religiosity on partisanship are reported in Table 10.15 for both weighted and unweighted group polarization. For the weighted models, the two most substantively significant trends are the declining contribution to partisan polarization by the ‘every week’ church attendees (-0.661) and the increasing contribution of the seculars who ‘never’ attend church (1.358). The contribution of the very religious to partisan polarization has declined by over half a percent for every survey year, while the secular contribution has risen by 1.4% in every survey year of the time series. The linear trend model for seculars explains nearly 60% of the variation in mean deviation of percent contribution to partisan polarization (R2 = .590). The “Culture War” model tracks the absolute difference between the percent contribution of the every week church attendees and the secular citizens

316

(mean deviation). The difference in contribution between the very religious and secular respondents increased nearly 3% for every survey year (2.717), and this model also accounts for close to 60% of the model error (R2 = .590). The one other significant trend is the declining contribution of the “twice a year” church attendees (-0.590), with it declining over half a percent a survey year since 1970. The reduction in model error for the ‘infrequent’ church attendees was nearly 30% over the past 34 years (R2 = .282). This appears to be exclusively a function of the declining number of citizens who report going infrequently, as the unweighted model for this category is not statistically significant (R2 = .070). Likely these are citizens who were weakly attached to their church or religion and, over the course of the past thirty-plus years, have moved into the seculars and now never attend church.

In the unweighted group polarization models (weights each category of religiosity equally), the significant trends identified in the weighted models (excepting the “twice a year” model) are also significant, though the coefficients are uniformly lower than in the weighted models. Clearly the decline of religiosity accentuated the divergence of the religiosity groups on party identification. The very religious group (every week) decreasingly contributes to partisan polarization (-0.421) and the secular group increasingly contributes to the group polarization measure on partisanship (0.889). The “Culture War” model indicates that the divergence of religious and secular citizens has increased just short of two percent per survey year irrespective of the change in the size of the religiosity groups (1.778). This model explains about 40% of the variation in the group polarization measure on partisanship (R2 = .424).

Religiosity Polarization on Ideology

The significant group polarization of religiosity on partisanship apparent in the previous analysis is echoed in the significant group polarization of religiosity in the ideological dimension. Table

317

10.16 has the group polarization measure for the religiosity groups on ideology. Again it is apparent that significant polarization has occurred over the time series. All four categories have witnessed substantial increases in their contributions to group polarization since the 1970’s, with once again the most significant contributor being that of the seculars. Note that in 1972 the secular contribution to polarization on ideology is half that of the every-week church attendees.

If you will recall, it is not merely the relative distance on the dimension which determines a group’s contribution to polarization but also the size of the group relative to the population.

With seculars on the rise over the course of the time series, they increasingly contributed to the ideological polarization of the religiosity groups.

TABLE 10.16: RELIGIOSITY – DECOMPOSITION OF GROUP POLARIZATION ON IDEOLOGY, 1972-2004

YEAR Every Week Twice a Twice a Year Never GPRelig Month 1972 0.32455 0.03316 0.08888 0.14816 0.5947 1974 0.26127 0.03251 0.08706 0.15143 0.5323 1976 0.33951 0.04222 0.08750 0.16114 0.6304 1978 0.25486 0.02464 0.06610 0.11530 0.4609 1980 0.02515 0.00496 0.02071 0.02244 0.0733 1982 0.43544 0.04723 0.09801 0.20988 0.7906 1984 0.26373 0.04897 0.06668 0.09589 0.4753 1986 0.19571 0.02114 0.04873 0.09908 0.3647 1988 0.24625 0.04383 0.08142 0.16137 0.5329 1990 2.42579 0.65785 0.78970 5.19809 9.0714 1992 0.21625 0.02188 0.04348 0.07366 0.3553 1994 0.44598 0.04488 0.05285 0.18129 0.7250 1996 0.28759 0.05301 0.14369 0.11405 0.5983 1998 2.68358 0.86009 0.77779 5.84816 10.1696 2000 3.14466 0.86400 0.71009 5.11155 9.8303 2002 3.77038 1.03513 0.61226 5.50609 10.9239 2004 2.74913 0.85054 0.67359 5.50295 9.7762

318

Beginning in the 1990’s it is the seculars that are the most significant drivers of group polarization and by 2004 their contribution is nearly twice that of the every-day church attendees.

Table 10.17 reports the trend models for the religiosity groups on ideology. While the secular model does not provide the best proportionate reduction in error of the categorical models (that honor belongs to the every-day-a-weekers), substantively speaking the seculars have seen the largest increase in their contribution to ideological polarization (.177 for each survey year). The middle two categories contribute relatively little to the trends in group polarization on ideology. Both the “Twice a Month” category (0.029) and the “Twice a Year” category (0.021) have relatively small coefficients. Their impact combined is only half that of

TABLE 10.17: TREND REGRESSIONS OF RELIGIOSITY GROUP POLARIZATION ON IDEOLOGY, 1972-2004 Trend: MODEL: Polarization Intercept Parameter Estimate GP(I) = B0 + B1(YEAR) + E Y/N (S.E.) (S.E.) R2 N

Every Week Y -184.463 0.093 *** .535 17 (44.687) (0.022)

Twice a Month Y -56.813 0.029 *** .552 17 (13.285) (0.007)

Twice a Year Y -40.422 0.021 *** .464 17 (11.295) (0.006)

Never Y -349.191 0.177 *** .510 17 (88.834) (0.045)

Religiosity Group Y -630.889 0.319 *** .549 17 Polarization (156.064) (0.079)

319

the regular, weekly church attendees (0.093) and not even in the same ballpark as the seculars.

That said, while their substantive impact is negligible, they both have a positive effect on religiosity group polarization on ideology and the model fit statistic indicates that both models improve on the mean by an over 50 percent reduction in error. For the full religiosity classification, there is a 0.319 (four tenths of a point) increase in polarization for each survey year for the regression model. The R-Square for this model (.549) indicates a good fit and thus yet another example of increasing political polarization consistent with the culture wars thesis.

The percent contributions of religiosity to group polarization for the weighted and unweighted measures on the ideological dimension are reported in Tables 10.18 and 10.19. The decline in the contribution of the most frequent church attendees of the religiosity groups isn’t as steep in the ideological dimension as it was on partisanship, but it is significant and pronounced. There are higher percent contributions in the weighted measure than we saw with partisanship, as the “every week” category accounts for at least 60% of the group polarization on ideology in 1992 and 1994. The two middle categories trail behind the two extreme categories significantly in their contribution to weighted polarization on ideology. The maximum contribution for the “twice a month” category was 10.3% in 1984 and, for the most part, the contribution of this category lags well below 10% for the time series. While the “twice a year” group had a spike in its contribution in 1996 (24.02%), this was the only time since 1988 that their contribution to polarization exceeded 10%.

The unweighted contribution to group polarization for the “every week” group tracks with the weighted contribution, however it is between 5 and 15% less of a contribution in the unweighted measure. The size of the group clearly accentuated its contribution in the 1970’s and 1980’s and its subsequent decline downgrades its contribution in the 2000’s. The secular group exhibits the same steady trend towards greater contribution to the polarization of

320

TABLE 10.18: RELIGIOSITY – WEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON IDEOLOGY, 1972-2004

YEAR EW EW-DEV TM TM-DEV TY TY-DEV NEV NEV- DEV 1972 54.57% 8.94% 5.58% -1.76% 14.95% 1.86% 24.91% -9.03% 1974 49.08% 3.45% 6.11% -1.23% 16.36% 3.27% 28.45% -5.49% 1976 53.86% 8.22% 6.70% -0.64% 13.88% 0.79% 25.56% -8.38% 1978 55.30% 9.66% 5.35% -1.99% 14.34% 1.26% 25.02% -8.93% 1980 34.31% -11.32% 6.77% -0.57% 28.25% 15.17% 30.61% -3.33% 1982 55.08% 9.44% 5.97% -1.36% 12.40% -0.69% 26.55% -7.40% 1984 55.49% 9.85% 10.30% 2.97% 14.03% 0.94% 20.17% -13.77% 1986 53.66% 8.03% 5.80% -1.54% 13.36% 0.28% 27.17% -6.77% 1988 46.21% 0.58% 8.22% 0.89% 15.28% 2.19% 30.28% -3.66% 1990 26.74% -18.89% 7.25% -0.08% 8.71% -4.38% 57.30% 23.36% 1992 60.86% 15.23% 6.16% -1.18% 12.24% -0.85% 20.73% -13.21% 1994 61.51% 15.88% 6.19% -1.14% 7.29% -5.80% 25.01% -8.94% 1996 48.07% 2.44% 8.86% 1.53% 24.02% 10.93% 19.06% -14.88% 1998 26.39% -19.24% 8.46% 1.12% 7.65% -5.44% 57.51% 23.56% 2000 31.99% -13.64% 8.79% 1.46% 7.22% -5.86% 52.00% 18.06% 2002 34.51% -11.12% 9.48% 2.14% 5.60% -7.48% 50.40% 16.46% 2004 28.12% -17.51% 8.70% 1.37% 6.89% -6.20% 56.29% 22.35%

TABLE 10.19: RELIGIOSITY – UNWEIGHTED PERCENT CONTRIBUTION & MEAN DEVIATION ON GP ON IDEO., 1972-2004

YEAR EW EW-DEV TM TM- TY TY-DEV NEV NEV- DEV DEV 1972 39.30% 6.02% 12.87% -2.10% 12.54% -3.51% 35.29% -0.41% 1974 31.29% -1.99% 13.02% -1.95% 15.28% -0.77% 40.41% 4.71% 1976 38.05% 4.76% 13.02% -1.94% 13.47% -2.58% 35.46% -0.24% 1978 43.45% 10.16% 12.62% -2.34% 13.45% -2.61% 30.48% -5.22% 1980 24.26% -9.03% 17.37% 2.41% 24.26% 8.21% 34.13% -1.57% 1982 38.44% 5.15% 12.63% -2.34% 12.85% -3.20% 36.08% 0.38% 1984 42.22% 8.93% 20.45% 5.49% 13.12% -2.93% 24.21% -11.49% 1986 38.82% 19.58% 12.69% -13.65% 13.11% -5.92% 35.38% 25.37% 1988 33.62% 0.34% 15.22% 0.25% 14.19% -1.87% 36.97% 1.27% 1990 20.09% -13.20% 15.03% 0.06% 15.49% -0.56% 49.39% 13.69% 1992 46.42% 13.13% 12.57% -2.39% 23.49% 7.44% 17.52% -18.18% 1994 48.11% 14.83% 14.48% -0.48% 14.13% -1.92% 23.28% -12.42% 1996 32.66% -0.63% 15.83% 0.87% 34.78% 18.73% 16.73% -18.97% 1998 20.00% -13.29% 17.07% 2.11% 13.87% -2.18% 49.06% 13.36% 2000 23.51% -9.78% 16.36% 1.40% 12.88% -3.18% 47.26% 11.56% 2002 23.08% -10.20% 16.51% 1.55% 12.93% -3.12% 47.47% 11.77% 2004 22.54% -10.74% 16.65% 1.68% 13.04% -3.01% 47.77% 12.07%

321

TABLE 10.20: WEIGHTED & UNWEIGHTED MEAN DEV. TREND REGRESSIONS OF RELIGIOSITY GP ON IDEO, 1972-2004 YEAR MODEL: Intercept P.E. 2 GPMD(IDEO) = B0 + B1(YEAR) + E (R.S.E.) (R.S.E.) R N WEIGHTED GROUP POLARIZATION

Every Week 1293.784 -0.651 *** .280 17 (357.216) (0.180)

Twice a Month -192.466 0.097 *** .402 17 (32.217) (0.016)

Twice a Year 617.985 -0.311 *** .270 17 (212.553) (0.107)

Never -1720.521 0.865 *** .376 17 (411.000) (0.208)

Culture War -3439.824 1.730 *** .376 17 (EW – NEV) (1144.129) (0.415)

UNWEIGHTED GROUP POLARIZATION

Every Week 837.020 -0.421 * .180 17 (425.462) (0.214)

Twice a Month 574.398 -0.289 .084 17 (449.831) (0.226)

Twice a Year 353.608 -0.178 .070 17 (291.469) (0.146)

Never -1767.007 0.889 *** .425 17 (422.639) (0.213)

Culture War -3532.033 1.778 *** .424 17 (EW-NEV) (845.840) (0.426)

322

religiosity groups on the ideological dimension. It contributes nearly half of the polarization for each of the survey years in the first decade of the 21st century (47.26% - 47.77%). Again we see the impact of group size on the secular contribution to polarization. In the 1970’s its weighted percent contribution lags behind its unweighted percent contribution. In 1972 seculars contributed 35.29% of the unweighted polarization on ideology, while it only contributed

24.91% in unweighted ideology. This relationship flips in 1990. In 2004, the unweighted group polarization was 47.77%, but the weighted group polarization was higher: 56.29%.

The percent contributions for both weighted and unweighted group polarization of the religiosity groups on ideology are given in Table 10.20. While the middle categories exhibited no significant trends in their percent contributions in the partisanship model with one exception, both middle categories exhibit significant contribution trends in the weighted models. The

“twice a month” group has a statistically significant increase in its contribution (0.097), while the

“twice a year” group had a declining part in group polarization over the time series (-0.311). The trend for the group that reports attending church a few times a month accounts for more of the variation in the group polarization measure (R2 = .402) than does the twice a year model (R2 =

.270). Both significant trends are a consequence in the variation of group size, as when each category is not weighted by their percentage in the population (as approximated by sample size), the linear trends are not significant.

The two extreme religiosity groups exhibit the same trends on ideology that they did on partisanship. The ‘every week’ religiosity group has a strong negative trend in percent contribution for the weighted model (-0.651) and the unweighted model (-0.421). The two most significant factors in these trends is the rise of the seculars as a contributor to polarization and the slight but real decline in the size of the “every week” group. The seculars significantly

323

increase in their percent contribution to ideological polarization as we advance from the

beginning of the time series in 1972. In the unweighted model, the secular group contributes

1.778% more to ideological polarization for each survey year. In the weighted model it is less,

but still a significant trend and the largest change in percent contribution among the religiosity

groups (0.865). For both the unweighted and weighted models, the difference between the two

“Culture War” categories has significantly increased. The absolute distance between the

contributions of the most religious citizens and the least religious citizens is, on average, 1.7%

for every survey year in the weighted model and about 1.7% in the unweighted model. No

matter whether you take into account group size or not, there has been a substantial and

significant increase in the ideological polarization of the two extreme religiosity groups.

CONCLUSION: THE CULTURE WARS ARE A CRISIS OF FAITH IN CHRIST, NOT MARX

Using the group polarization measure, I have examined polarization on two different dimensions (partisanship and ideology) for two distinct groups (class and religiosity). Despite the colorful story-telling of Thomas Franks, I find little evidence of income group polarization.

And what polarization exists is dominated by the ‘rich’ rather than the poor or working class groups as suggested by Franks. There is significant polarization among the religiosity groups. A significant proportion of the partisan and ideological polarization among religiosity groups is attributable to the rise of the seculars in modern America. It is along the religious divide that we have become increasingly divided ideologically and along which the parties have increasingly divided themselves in electoral and policy competition. While there is slight polarization of the classes, it is neither the poor nor the working class that have been the prime movers for polarization in partisanship or ideology. Hunter’s original insight was correct. The growing culture wars have a distinct religious dimension with partisans and ideologues increasingly divided based on their belief in God and affiliation with religious institutions and organizations.

324

CHAPTER 11: PERCIEVED POLITICAL POLARIZATION: PARTIES, CANDIDATES, AND THE MASS PUBLIC

“Reality is merely an illusion, albeit a very persistent one.” – Albert Einstein

“Talking perceptions, people. Do we really see each other for what we really are, or do we just see what we want to see, the image distorted by our own personal lenses? I lost someone today and the funny thing is, I don't even know who she was.” - Jeff Melvoin

TRENDS IN ATTITUDES ON ISSUES: PERCEPTIONS OF DISTANCE FROM PARTIES AND CANDIDATES

The previous chapters analyzing polarization across multiple issue dimensions using measures of bimodality, dispersion, and mean trends have demonstrated that for a number of social issues and non-social issues, the American public has become more polarized since the

1970’s. I have also demonstrated a strong relationship between polarization trends at the level of the mass public and actual public policy, presidential partisanship, and exogenous shocks to the political environment relevant to the polarized public attitudes on the issue dimensions. I have shown that the salience of social issues has grown since the 1970’s, and that partisan polarization has occurred in ideology and abortion. I have punctured the myth of the working class conservative, and identified the growing political activation and polarization of the religious versus the secular citizenry. Throughout the course of these analyses, I have tied trends in the mass public on polarization with anecdotal and systematic evidence of elite polarization. But the exact relationship between the mass public and elites on polarization has yet to be delineated. Are party elites polarizing on the ideological, partisan, and issue dimensions in response to the polarization at the mass level? Or is mass polarization driven by elite polarization, as is the relative consensus in the mass-elite literature? Is it one of these two relationships, or is it a combination of the two through some reciprocal process?

325

I assess these questions in this chapter using perceptual data from the ANES on the positions of the parties and candidates relative to the reported positions of the mass public.

How does the American voter view the parties? Are elites increasingly polarized and extreme relative to the public’s own relatively moderate positions? If so, has this occurred independent of polarization at the mass public level or in spite of attitude stability or even depolarization on the issues in public opinion? How the respondents themselves perceive the parties and candidates can provide a window into the mass-elite relationship on polarization.

An important part of the political polarization story is the relationship between the mass public and elites on the issue dimensions. Fiorina argues that elites have polarized on social issues while the mass public attitudes on social issues have remained stable. Thus elites have moved away from the voters and citizens they are supposed to be responsive to. Thus an empirical question is born: have elites polarized relative to the mass public? In the early part of the Twentieth century, tips on horse racing were known to circulate among the ‘authorities’ such as owners, jockeys, and punters. The most trustworthy sources on the likely winner were those “closest” to the horse (stable boys, trainers, etc.). One step ‘closer’ to the horse than the inner circle of horse aficionados was, of course, the horse itself. Thus was born the colloquialism: if you want the truth, go straight to the horse’s mouth. We can do exactly that, in examining political polarization in the ideological and issue dimensions. Respondents in the

ANES time series were asked not only to place themselves on issue and ideological scales, but also to place the parties and candidates on these dimensions as well. We can thus assess directly the perception of the citizenry on the distance between themselves and the parties and candidates respectively on social, defense, and economic issue dimensions. If elites have moved away from their constituents, then citizens should increasingly locate both parties and the candidates of both parties further distant from their own positions. Of course elites can move

326

away from a part of the electorate while staying proximate or even moving closer to other parts

of the electorate. Indeed, if we accept the polarization thesis then we should expect

Republicans to view the Democratic Party and its candidates further from them, and likewise for

Democratic identifiers and elite Republicans. Thus the increasing dispersion between the

parties and candidates would be intelligible as responsiveness to the public rather than neglect

of it.

Data

The data for this analysis is culled from the American National Election Study (ANES) cumulative

file.74 I use the ANES studies from 1970-2004.75 Distance measures are calculated on the seven- point issue scales from the ANES where respondents were asked to place themselves and the parties and/or the candidates.

Variables

The substantive variables included in the analysis and the years for which data was collected on those variables are listed in Appendix G. The ANES time-series includes a number of issue-oriented ordinals scales with which respondents can place themselves, the parties, and candidates in an issue space. While a number of interesting issues have come and gone in the

ANES time-series, there are a number of issues that span the breadth of the time series that the

ANES has collected data on consistently. Every issue variable rated on an ordinal scale in this analysis has at minimum a respondent self-placement on the issue. However, in order to

74 The Cumulative Data File consists of variables derived from the 1948-2004 series of biennial ("time- series") SRC/CPS National Election Studies. The American National Election Studies / Time Series Studies are collected before and after presidential (pre and post surveys) elections. The off-year elections typically only have a post-election study. The ANES Cumulative Data File is a merged data set of all the time series studies from 1948-2004. The pooled data includes variables which appear in three or more studies and consists of 44,715 cases. 75 The data is sub-setted by year to include only studies from 1970-2004 as the previous data sets had few to none of the relevant substantive variables which are necessary for the polarization analysis. Furthermore, 1970-2004 covers the relevant time period to examine the culture wars thesis.

327

calculate distance measures, the study must also include the placement of the candidates or the parties in the issue dimension.76 The candidate placements always include the placement of the presidential candidates in a presidential election year. In some of the presidential and off-year elections, the House and Senate candidates for the respective parties are included. In a maximum of sixteen (liberal/conservative scale) of the study years the party placements are also collected, though the party placements were not collected in all study years in which the respondent self-placements on the issue were included.

The issues include respondent attitudes on government aid to Blacks, women’s role in society, government spending, defense spending, and jobs. The issue placement that is collected in the most study years for the respondent, candidates, and parties is the ideological placement variable (asking respondents to place themselves and the parties and candidates on a 7-point ideological scale ranging from strongly conservative to strongly liberal). Also included are the candidate and party placements on party. While the abortion self-placement variables are asked for all but one of the study years, the candidates and parties were placed by the respondent only in the study years 1980, 1992, 1996, 1998, 2000 and 2004. The ANES

Cumulative Data File does not include the respondent placements of parties and candidates on the abortion scale. Given the importance of the abortion issue to understanding the culture wars, I merged the abortion party and candidate placement variables from the individual studies into the cumulative file so I could include them in this analysis. Given that there is only one year in which they appear prior to 1992, there is insufficient data to assess statistically significant trends. This does not mean the abortion data is meaningless. It is an important indicator of

76 See Appendix G for the study years in which respondent placements, candidate placements, and party placements were collected on the relevant issues for the polarization analysis.

328

perceived distance on the social issue dimension. As such I examine differences in the abortion distances outside of the framework of a linear trend model.

Expectations

Below are precise statements of expectations given either polarization or depolarization trends. In the models that assess the perceptions of all respondents, we can assess whether the public perceives a particular party and/or its candidates has moved closer to their position or further away over time. If political polarization has occurred at the elite level independent of the masses, we would expect to see that the distances between the mass positions on the issues and the elite positions have increased. Some scholars term this ‘alienation’ in the literature: elites have become more alienated from the public on the issues.

However, as I noted above, merely identifying alienation between the average positions of the mass public and the perceived positions of the elites is insufficient to conclude that the elites have polarized while the public has moderated. Elites may be responding to an electorally significant portion of the public which has polarized as well (or they may have polarized together). I consider the most likely model for such a phenomenon: partisan polarization. If the electorate has experienced partisan polarization, then we would find alienation at the aggregate mass public level that is fundamentally misleading on the polarization question. If mass Republican identifiers have polarized from Democratic identifiers, and the Republican elites have moved in response to (or coincident to) this mass polarization of their base, then we would expect two specific conditions: 1) Republicans at the mass level would not perceive an increasing distance between themselves and elite Republicans. 2) Republicans at the mass level would perceive a substantial increase in the distance between themselves and the Democratic elites. And the same would go for Democratic identifiers and Republican elites. The precise expectations of partisan polarization are laid out in the Base and Opposition models.

329

Perceived Partisan Proximity on Issues Expectations (Distance)

Elites (Candidates) / Mass Divergence (Alienation) Eo: No trend in the mean absolute distance from both Republican and Democratic Candidates.

Ea: A decrease in the mean absolute distance from both Republican and Democratic Candidates (non-alienation).

Ea: An increase in the mean absolute distance from both Republican and Democratic Candidates (alienation).

Elites (Parties) / Mass Divergence (Partisan Polarization) Parties and Partisan Identifiers Base Model Eo: No trend in the mean absolute distance between Republican identifiers and the Republican Party or in the distance between Democratic identifiers and the Democratic Party (constituent status quo).

Ea: An increase in the mean absolute distance between Republican identifiers and the Republican Party or an increase in the distance between Democratic identifiers and the Democratic Party (constituent alienation).

Ea: A decrease in the mean absolute distance between Republican identifiers and the Republican Party and/or an increase in the distance between Democratic identifiers and the Democratic Party (constituent responsiveness).

Parties and Partisan Identifiers Opposition Model Eo: No trend in the mean absolute distance between Republican identifiers and the Democratic Party or in the distance between Democratic identifiers and the Republican Party (partisan polarization status quo).

Ea: A decrease in the mean absolute distance between Republican identifiers and the Democratic Party or a decrease in the distance between Democratic identifiers and the Republican Party (partisan depolarization).

Ea: An increase in the mean absolute distance between Republican identifiers and the Democratic Party or an increase in the distance between Democratic identifiers and the Republican Party (partisan polarization).

Measures

Distance Measures

One test of the elite / mass divergence hypothesis is to examine the perceived issue

distance of the candidates to the mass public. Using the ANES, we can place the candidates and

respondents in the issue space. We can then calculate the distance between the candidates and

330

the respondents using Euclidian geometry. Thus the issue distance measures are the relative,

perceived difference between a respondent and each of the candidates, parties, or elected

officials on an issue dimensions (e.g. aid to Blacks). . Let Sij denote the self-placement of voter j

on an issue dimension i, and let the Democratic and Republican Party or those parties

candidates’ perceived locations on issue i by respondent j be Dij and Rij.

Equations 11.1 & 11.2: Perceived Distance from Respondent to Parties or Candidates

௥௜௝ ௜௝ ூ௃ ݀ = หܵ െܴ ห

ௗ௜௝ ௜௝ ூ௃ We can assess the proximity of݀ the respondedหܵ െܦ to หone candidate or party relative to the other by calculating the absolute value of the distances between the candidates or parties and taking the difference of these absolute distances. Again, let Sij denote the self-placement of voter j on an issue dimension i, and let the Democratic and Republican Party or those parties candidates’ perceived locations on issue i by respondent j be Dij and Rij. Then, distance on issue i is the absolute difference between self-placement and the perceived candidate [party, official] location. The relative difference on issue i, for respondent j, is:

Equation 11.3: Perceived Partisan Proximity Measure

௜௝ ௜௝ ூ௃ ௜௝ ௜௝ where a positive value indicates ݀closer หissueܵ െܴ proximityห െ ห ܵto theെܦ Democraticห presidential candidate or Democratic Party, a negative value indicates the respondent is closer to the Republican candidate or party, and a value of zero indicates indifference between the two. A larger magnitude for issue distance indicates greater perceived difference between the presidential candidates or political parties (Bough et al. 2004).

331

Continuing with the logic set out above, we can also measure the total perceived

distance from the respondent position to each of the candidates, or alternatively, both parties.

Rather than assess which candidate the respondent perceives as closer to him or her on a

particular political issue, this measure assesses the total distance from the respondent’s

declared position on the issue and the position they believe the candidate has on that issue. The

total distance on issue i, for respondent j, is:

Equation 11.4: Total Perceived Distance Measure

= +

௜௝ ௜௝ ூ௃ ௜௝ ௜௝ For those respondents who݀ gaveหܵ non-responses,െܴ ห หܵ ifെܦ the respondentห failed to replace him or herself on an issue, the distance value was set to zero. If a respondent failed to place one of the candidates, the perceived candidate position for that respondent was set to the mean candidate position for that issue. If a respondent refused to answer, they were set to missing.

DISCUSSION: INCREASING PERCIEVED INTRA & INTER-POLARIZATION ON IDEOLOGY AMONG PARTISANS

As I have argued in previous chapters, the ideological dimension is a convenient proxy for the sum of the issue dimensions at play in the public policy debate. Tables 11.1 reports distance measures for ideology (a proxy for the aggregate issue space). Abortion distance measures are assessed in Table 11.2 and the distance measure trend models for public opinion on government’s role in providing jobs can be found in Table 11.3. Distance measures for defense spending and government spending are located in Appendix I.77 I do not include those models in the main text because there are no significant linear trends for defense spending and the only significant model for government spending is a positive trend in the perceived distance between respondents and the Democratic presidential candidates and the Republican House

77 Appendix I has tables on OLS regression trend models for the distance measures for government spending & defense spending.

332

TABLE 11.1: TREND MODELS FOR PERCEIVED DISTANCE MEASURES ON IDEOLOGY

Parameter Standar 2 MODEL: DV (DIST) = B0 + B1(YEAR) + E N Intercep Estimate d Error R t POSITIONS R Ideology Self-Placement 1 -2.470 0.003 * 0.002 .164 7 Democratic Party Position 1 4.785 -0.001 0.004 .002 6 Democratic Prez Cand Position 9 -0.046 0.002 0.013 .002 Republican Party Position 1 -14.991 0.010 *** 0.003 .456 6 Republican Prez Cand Position 9 -8.128 0.007 * 0.003 .393 RELATIVE DISTANCE FROM PARTIES & CANDIDATES R – DP (Dem Party) 1 -21.589 0.011 ** 0.005 .284 6 R – DPC (Dem Prez Cand) 9 -11.942 0.007 0.012 .043 R – DHC (Dem House Cand) 1 -27.660 0.014 ** 0.006 .365 0 R – DSC (Dem Sen Cand) - - - - - R – RP (Rep Party) 1 26.270 -0.014 *** 0.003 .628 6 R – RPC (Rep Prez Cand) 9 22.796 -0.012 *** 0.003 .668 R- RHC (Rep House Cand) 1 17.698 -0.009 0.006 .202 0 R – RSC (Rep Sen Cand) - - - - - RELATIVE PARTISAN PROXIMITY (REP DISTANCE – DEM DISTANCE) |R – RP| – |R – DP| 1 6.127 -0.003 0.004 .040 6 |R – RPC| – |R – DPC| 9 -10.710 0.005 0.007 .071 |R – RHC| – |R – DHC| 1 50.074 -0.026 *** 0.007 .612 0 |R – RSC| – |R – DSC| - - - - - TOTAL RELATIVE PARTISAN DISTANCE (REP DISTANCE + DEM DISTANCE) |R – RP| + |R – DP| 1 -47.857 0.025 *** 0.006 .567 6 |R – RPC| + |R – DPC| 9 -36.822 0.022 *** 0.003 .858 |R – RHC| + |R – DHC| 1 -45.357 0.023 ** 0.008 .511 0 |R – RSC| + |R – DSC| - - - - - * significant at .10 level ** significant at .05 level ***significant at .01 level

333

candidates. This is not to say that there is nothing of significance to be found in these models.

The fact that there are no significant increases in the total perceived distance between the

candidates or parties on defense spending suggests the public does not believe the elites are out

of touch with them on foreign policy and national defense issues. On government spending it is

essentially the same story, except that there was one significant trend in total distance: a

significant decline in the total distance between the public and the Republican and Democratic

House candidates. The issue distance tables contain four different classes of measures. The first

section includes respondent self- placements and the placements of parties and candidates on

the specific issue scale. The second section reports the distances between the respondent and

the parties or candidates. So a distance measure here would be the perceived distance between

the respondent and the Republican Party. The third section reports the relative absolute

differences between the perceived distance between the respondent and the Republican Party

and the respondent and the Democrat Party (proximity scores). The fourth section reports total

distance measures on the issue for the parties and candidates. So a total distance measure adds

the perceived distance between the Republican Party or candidate and the Democratic Party or

candidate for a measure of the sum of the absolute distances from the respondents and both

political parties and candidates.

Ideological Distance Measures

On ideology there is a significant trend in the average respondent self-placement in a

conservative direction. The American public has become more conservative over the past three

decades by, on average, .003 points every survey year (R2 = .164). It is not a strong trend, nor is it a substantively large change in the average ideology of the American public, but it is statistically significant. Furthermore, the American public sees a conservative trend in the

Republican Party and its presidential candidates. The perceived conservative tilt in the

334

Republican Party is particularly strong (0.010) relative to the coefficients of the other models,

with the linear trend model explaining 45% of the variance in the Republican Party’s ideological

position (R2 = .456). The trend for Republican presidential candidates is similar but not as strong

(0.007), explaining just under 40% of the variation (R2 = .393).

On the relative distances, the models show significant positive trends in the distance

between the respondents and the Democratic Party (0.011) and the Democratic House

Candidates, suggesting that Democrats are increasingly to the ideological Left of the average

respondent. The significant negative coefficients for the Republican Party and Republican

presidential candidates show that Republican elites have increasingly moved to the Right

relative to the average respondent. The Republican relative distance models illustrate strong

predictive power, with the linear trends accounting for over 60% of the variance over the mean

distance for the Republican Party (R2 = .628) and the Republican presidential candidates (R2 =

.668).

In the absolute distance measures, the relative partisan proximity model for House candidates demonstrates that, from the viewpoint of the ‘average’ member of the public, the

Republicans are further distant from the Democrats. Interestingly, this is not the case for the

Republican Party or its presidential candidates. While both the Republican and Democratic parties are perceived to have polarized relative to the average citizen, there isn’t strong evidence that either has polarized more than the other. There is strong evidence, however, that the Democratic and Republican elites have increasingly polarized on the ideological dimension.

The total relative partisan distance measures, combining the absolute perceived distance for both Democrats and Republicans, show strong positive linear trends. The average citizen perceives both parties (0.025), both parties’ presidential candidates (0.022), and both parties’

House candidates (0.023) as increasingly distant from their own position. The presidential

335

candidate model explains over 85% of the variance in the distances, and exhibits a strong, positive linear coefficient (R2 = .858). The other two models explain over 50% of the variance in distances. The total distance measures indicate that the average respondent finds the parties further distant from him or herself today than in the past.

At first blush, the total distance measures would appear to be evidence in favor of Fiorina’s conjecture that elites have polarized away from the masses. However, jumping to that conclusion is unwarranted. Recall that political polarization implies greater dispersion. The total distance measures are relative to the average (mean) respondent. Hence these measures indicate that the parties have polarized relative to the average member of the mass public. The culture wars thesis (political polarization) predicts that the Republican and Democratic Parties will move towards the poles of the distribution relative to the center point. So the question isn’t whether the parties have moved away from the center on any particular issues. The question is whether the Parties have diverged independent of the voters.

Abortion Distance Measures

As noted earlier, the abortion distance measures are assessed in Table 11.2.

However, the statistical test for the abortion distances is different than those used for the other issue dimensions. The abortion issue placements for the parties and candidates were only collected in five survey years.

OLS regressions with only five data points are unreliable and may be biased and inefficient due to the violation of important assumptions such as normality. As such, I designed an independent sample t-test to produce difference of means statistics for three pairs of survey years, using the first year in which the distance measures are obtainable (1980) as the baseline

336

TABLE 11.2: ABORTION DISTANCE MEASURES – DIFFERENCE OF MEANS TESTS B/W 1980 & 1992, 1996, & 2000 VARIABLE 80-92 Pr > |t| 80-96 Pr > |t| 80-04 Pr > |t| POSITIONS -0.161 -0.079 0.020 R Abortion Self-Placement <.0001 .0356 .6460 (0.035) (0.037) (0.043) -0.434 -0.497 Democratic Party Position ------<.0001 <.0001 (0.037) (0.043) Democratic -0.483 -0.475 -0.415 <.0001 <.0001 <.0001 Prez Cand Position (0.027) (0.029) (0.033) -0.459 -0.401 Republican Party Position ------<.0001 <.0001 (0.029) (0.034) Republican 0.162 -0.083 0.252 <.0001 .0206 <.0001 Prez Cand Position (0.034) (0.036) (0.041) RELATIVE DISTANCE FROM PARTIES & CANDIDATES 0.402 0.564 R – DP (Dem Party) ------<.0001 <.0001 (0.056) (0.062) 0.316 0.392 0.431 R – DPC (Dem Prez Cand) <.0001 <.0001 <.0001 (0.042) (0.045) (0.052) 0.378 0.414 R – RP (Rep Party) ------<.0001 <.0001 (0.048) (0.054) -0.251 0.073 -0.164 R – RPC (Rep Prez Cand) <.0001 .2139 .0114 (0.051) (0.059) (0.065) RELATIVE PARTISAN PROXIMITY (REP DISTANCE – DEM DISTANCE) -0.021 0.051 |R – RP| – |R – DP| ------.7258 .4498 (0.097) (0.067) -0.179 -0.036 -0.064 |R – RPC| – |R – DPC| <.0001 .5527 0.356 (0.612) (0.077) (0.069) TOTAL RELATIVE PARTISAN DISTANCE (REP DISTANCE + DEM DISTANCE) 0.050 0.022 |R – RP| + |R – DP| ------.3581 .7253 (0.054) (0.061) 0.055 0.152 0.012 |R – RPC| + |R – DPC| .2727 .0051 .8446 (0.050) (0.054) (0.059)

for comparison. The public was significantly more Pro-Choice in 1992 and 1996 relative to

1980. However, sometime between 1996 and 2004 this Pro-Choice shift reversed itself. While there isn’t a significant difference between the public’s position on Abortion in 2004 relative to

337

1980, the coefficient is positive indicating that the abortion position in 1980 was further towards the Pro-Choice side of the abortion issue dimension than 2004.

The party comparisons indicate that the placement of the Democratic and Republican parties has become more Pro-Choice since the 1980’s. This is a strange result for Republicans, considering the partisan polarization on abortion that was apparent in the Chapter 9 analysis.

George H.W. Bush was viewed as more pro-life than Ronald Reagan, either due to the shifting abortion views of the public or perhaps given Reagan’s inconsistent record on abortion from his days as governor of California. Bob Dole was seen as more Pro-Choice than Reagan, though the difference is small (-0.083). The difference between George W. Bush and Reagan’s placement on the abortion scale is largest, likely owing to Bush’s self-identification as a religious conservative, a Christian-convert, and the prominence he gave faith and faith-based politics in his administration.

There was only one statistically significant difference in the relative partisan differences.

The comparison of 1980 to 1992 suggests that respondents saw Clinton as further distant from

George H.W. Bush on abortion when compared to Reagan and Jimmy Carter. In the total absolute distances, there was little evidence that average voters perceived the parties or the candidates much differently relative to 1980. Only in the 1980 to 1996 comparison was there a decrease in the total distance between the average respondent and the presidential candidates

(0.152) with a t-test that met the standard for statistical significance (.0051). Overall, while it was apparent that views on abortion and the placements had changed, there was no compelling evidence that citizens were placing the parties or the candidates at positions on the abortion dimension polarized from their own positions on abortion.

338

TABLE 11.3: TREND MODELS FOR PERCEIVED DISTANCE MEASURES ON JOBS

Parameter Standar MODEL: DISTANCE = YEAR N Intercep Estimate d Error R2 t JOBS – Respondent Placements R Jobs Self Placement 12 -32.918 0.018 *** 0.005 .673 Democratic Party Position 12 -14.121 0.009 * 0.005 .248 Democratic Prez Cand Position 9 -40.870 0.022 ** 0.009 .460 Republican Party Position 12 -36.363 0.021 *** 0.004 .736 Republican Prez Cand Position 9 -40.759 0.023 *** 0.006 .676 JOBS - Respondent Relative Distance from Parties & Candidates R – DP (Dem Party) 12 16.656 -0.008 0.006 .165 R – DPC (Dem Prez Cand) 9 34.029 -0.017 0.009 .320 R – DHC (Dem House Cand) 6 15.686 -0.007 0.010 .133 R – DSC (Dem Sen Cand) - - - - - R – RP (Rep Party) 12 43.206 -0.022 *** 0.007 .488 R – RPC (Rep Prez Cand) 9 36.703 -0.019 * 0.009 .341 R- RHC (Rep House Cand) 6 44.081 -0.022 * 0.011 .497 R – RSC (Rep Sen Cand) - - - - - JOBS - Respondent Relative Partisan Proximity (Rep Distance – Dem Distance) |R – RP| – |R – DP| 12 3.583 -0.002 0.005 .013 |R – RPC| – |R – DPC| 9 -17.929 0.009 0.007 .172 |R – RHC| – |R – DHC| 6 30.870 -0.016 0.011 .378 |R – RSC| – |R – DSC| - - - - - JOBS – Total Relative Partisan Distance (Rep Distance + Dem Distance) |R – RP| + |R – DP| 12 41.590 -0.021 0.027 .159 |R – RPC| + |R – DPC| 9 31.635 -0.014 0.016 .103 |R – RHC| + |R – DHC| 6 7.525 -0.001 0.010 .004 |R – RSC| + |R – DSC| - - - - - * significant at .10 level ** significant at .05 level ***significant at .01 level

Jobs Distance Measures

The government role in jobs provision linear trend models are presented in Table 11.3.

Immediately apparent should be the statistically significant shift in the average position of the public, both parties, and both candidates. However, all these trends are in the same direction.

The public has shifted towards a position of less government responsibility in the provision of jobs, but their collective perception of the Democratic Party (0.009), the Democratic presidential candidates (0.022), the Republican Party (0.021), and the Republican presidential candidates

339

(0.023) is that they’ve all shifted in that direction as well. However, while of the positional

models evidence positive linear coefficients, the Democratic Party lags behind relative to the

other parties and candidates and the public itself, and this model has the lowest goodness-of-fit

for the linear trend (R2 = .248). The model trend that tracks closest to that of the American public is the Republican Party trend and, this is also the best performing of the positional linear trends. The Republican party model explains almost 75% of the variation in the average placement of the party on the jobs scale (R2 = .746). The models of the public positional trend

(R2 = .673) and the placement of the Republican presidential candidates (R2 = .676) perform equally well, accounting for approximately 67% of the model variation relative to the mean position on the scale. Finally, while the Democratic presidential candidate model has a beta coefficient with a near-equal magnitude of that of the Republicans, the model doesn’t perform nearly as well, accounting for less than 50% of model variation (.460).

The relative partisan difference models are perfectly comprehensible given the positional trends identified above. While the Democratic models are all insignificant, their coefficients slope in the same negative direction indicating declining distance between the respondents and the Democrats. However, the Republican models are also negatively sloped and they are statistically significant. The relative perceived distances between the average public position and their placement of the Republican Party (-0.022), the Republican presidential candidates (-0.019), and the Republican House candidates (-0.022) have all declined. The

Republican models explain between 40% and 50% of the variation in relative partisan distance.

Given that both parties and the party candidates are perceived to have become less supportive of government intervention in the jobs market along with the American public, the results from the relative and total absolute distance measures should be expected. The public does not appear to have grown closer to either party, given the absence of statistically

340

significant linear trends in the relative absolute partisan distance measures. And, naturally, the

American public does not believe the Republican and Democrats have polarized on jobs relative

to their own average position on the scale, as is apparent from the statistically insignificant total

absolute partisan difference models. Overall, while there was a sizable shift in public opinion on

jobs and their placement of both parties and candidates on the issue, since that movement was

all in the same direction there is no apparent perceived polarization on the government role in

jobs provision. The fault line on jobs has shifted, but all the partisan political actors are

perceived to have shifted right along with it.

MASS PARTISAN PERCEPTIONS OF THE DISTANCE FROM ELITE PARTIES AND CANDIDATES

At first blush, the total distance measures would appear to be evidence in favor of

Fiorina’s conjecture that elites have polarized away from the masses. However, jumping to that conclusion is unwarranted. Recall that political polarization implies greater dispersion. The total distance measures are relative to the average (mean) respondent. Hence these measures indicate that the parties have polarized relative to the average member of the mass public. The culture wars thesis (political polarization) predicts that the Republican and Democratic Parties will move towards the poles of the distribution relative to the center point. So the question isn’t whether the parties have moved away from the center on any particular issues. The question is whether the Parties have diverged independent of the voters.

For that, we need to take a closer look at the mass public. One possible explanation for the greater perceived distance between the average member of the mass public and the political parties is that they have diverged relative to the average voter but in response to a concomitant shift towards the poles of the distribution by partisan identifiers. In other words, rather than elites shifting to the poles despite the mass public, the party elites may be shifting in

341

TABLE 11.4: TOTAL PARTISAN DISTANCE MEASURES FOR PARTY IDENTIFIERS

Parameter Standar 2 MODEL: DV (DIST) = B0 + B1(YEAR) + N Intercep Estimate d Error R E t IDEOLOGY R |R – RP| + |R – DP| 16 -51.846 0.027 *** 0.006 .637 R |R – RPC| + |R – DPC| 9 -49.457 0.026 ** 0.011 .456 R |R – RHC| + |R – DHC| 10 -39.174 0.021 0.012 .265 R |R – RSC| + |R – DSC| - - - - - I |R – RP| + |R – DP| 16 -23.670 0.013 0.009 .120 I |R – RPC| + |R – DPC| 9 -3.248 0.002 0.013 .005 I |R – RHC| + |R – DHC| 10 -73.347 0.038 * 0.017 .385 I |R – RSC| + |R – DSC| - - - - - D |R – RP| + |R – DP| 16 -32.192 0.017 ** 0.006 .339 D |R – RPC| + |R – DPC| 9 -13.326 0.008 0.010 .078 D |R – RHC| + |R – DHC| 10 -54.203 0.028 *** 0.009 .539 D |R – RSC| + |R – DSC| - - - - - JOBS R |R – RP| + |R – DP| 12 39.905 -0.019 0.012 .201 R |R – RPC| + |R – DPC| 9 -16.756 0.011 0.008 .229 R |R – RHC| + |R – DHC| 6 26.306 -0.012 0.019 .094 R |R – RSC| + |R – DSC| - - - - - I |R – RP| + |R – DP| 12 28.827 -0.014 0.011 .137 I |R – RPC| + |R – DPC| 9 64.571 -0.032 *** 0.010 .584 I |R – RHC| + |R – DHC| 6 101.035 -0.050 0.030 .412 I |R – RSC| + |R – DSC| - - - - - D |R – RP| + |R – DP| 12 -23.602 0.013 * 0.007 .228 D |R – RPC| + |R – DPC| 9 2.245 -0.001 0.012 .001 D |R – RHC| + |R – DHC| 6 -12.445 0.007 0.007 .178 D |R – RSC| + |R – DSC| - - - - - * significant at .10 level ** significant at .05 level ***significant at .01 level response to the segment of the mass public that they most count on for votes and funds. If this is the case, then not only would this divergence be intelligible as an electoral strategy, but also it could directly contribute to the increase in distance reflected in the total distance measure.

Because as the Democratic and Republican identifiers shifted towards the poles, they would perceive a greater distance between themselves and the other party irrespective of whether the party had changed its position on the issue dimension at all. Of course, if they did shift towards

342

TABLE 11.5: TOTAL PARTISAN DISTANCE MEASURES FOR PARTY IDENTIFIERS

Parameter Standar 2 MODEL: DV (DIST) = B0 + B1(YEAR) + N Intercep Estimate d Error R E t SPENDING R |R – RP| + |R – DP| 11 20.352 -0.009 0.017 .035 R |R – RPC| + |R – DPC| 6 34.614 -0.016 0.016 .209 R |R – RHC| + |R – DHC| 9 37.707 -0.018 0.023 .092 R |R – RSC| + |R – DSC| - - - - - I |R – RP| + |R – DP| 11 38.007 -0.018 0.016 .131 I |R – RPC| + |R – DPC| 6 39.381 -0.019 0.011 .440 I |R – RHC| + |R – DHC| 9 30.404 -0.015 0.024 .050 I |R – RSC| + |R – DSC| - - - - - D |R – RP| + |R – DP| 11 -17.179 0.010 0.014 .047 D |R – RPC| + |R – DPC| 6 -5.335 0.004 0.019 .009 D |R – RHC| + |R – DHC| 9 -80.641 0.041 *** 0.013 .587 D |R – RSC| + |R – DSC| - - - - - DEFENSE SPENDING R |R – RP| + |R – DP| 10 22.952 -0.011 0.020 .034 R |R – RPC| + |R – DPC| 7 23.147 -0.011 0.032 .021 R |R – RHC| + |R – DHC| - - - - - R |R – RSC| + |R – DSC| - - - - - I |R – RP| + |R – DP| 10 74.716 -0.037 *** 0.012 .562 I |R – RPC| + |R – DPC| 7 71.596 -0.035 * 0.016 .502 I |R – RHC| + |R – DHC| - - - - - I |R – RSC| + |R – DSC| - - - - - D |R – RP| + |R – DP| 10 56.618 -0.028 0.029 .101 D |R – RPC| + |R – DPC| 7 41.375 -0.020 0.029 .088 D |R – RHC| + |R – DHC| - - - - - D |R – RSC| + |R – DSC| - - - - - * significant at .10 level ** significant at .05 level ***significant at .01 level their own constituents, that would also contribute to a greater perceived distance between party identifiers for one party and the perceived distance between them and the other party.

One indicator that this might be the case is found in Table 11.4. Note that it is the

Republican and Democrat identifiers that are driving the significant trend results on total perceived distance. For all the issue dimensions, independents perceive an increase in total distance on only one measure (Republican House candidates on the ideological dimension). And

343

even there, the model barely achieves significance and has a relatively moderate goodness-of-fit

(.385). The significant increases in total perceived partisan distance from elite actors are exclusively in the party identifier models: mass Republicans perceiving increasing distance between themselves and Democratic elite actors, and likewise for mass Democrats and

Republican elites. Democrats perceive increasing total distance on jobs, and both Republican and Democratic identifiers see increasing total distance between themselves and the parties and candidates on the ideological dimension.

In Table 11.5, we see the same trend. Independent identifiers perceive a significant decline in the distance between themselves and the parties (-0.037) and the candidates (-0.035) on defense spending. Whereas the linear trend of increasing distance is in an oppositional model: Democratic identifiers perceiving increasing distance between themselves and House candidates on defense spending (0.041). All three of the models account for between 50% and

60% of the variance in total absolute distance. While the total distance measures suggest partisan polarization rather than alienation given the perceptions of Independent identifiers relative to the partisan identifiers, it isn’t direct evidence. It is possible, though unlikely, that partisan identifiers attribute the increasing total distance on these political dimensions equally to both political parties or even exclusively to their own party. We could have disaffection rather than party polarization.

In order to directly test partisan polarization, I examine the absolute distances between party identifiers and the parties and candidates individually. In Table 11.6 I report on the absolute distances between party identifiers and the Republican and Democrat parties. Note, if the partisan constituent responsiveness hypothesis is correct, we would expect Republican identifiers to perceive little to no increase in the distance between themselves and the

344

TABLE 11.6: REPUBLICAN AND DEMOCRATIC PARTY DISTANCES FROM PARTY IDENTIFIERS

Parameter Standar 2 MODEL: DV (DIST) = B0 + B1(YEAR) + N Intercep Estimate d Error R E t IDEOLOGY R |R – RP| 16 11.988 -0.006 ** 0.002 .320 R |R – DP| 16 -39.858 0.021 *** 0.006 .432 D |R – RP| 16 -43.841 0.023 *** 0.004 .675 D |R – DP| 16 11.649 -0.006 0.004 .164 JOBS R |R – RP| 12 34.276 -0.017 *** 0.005 .532 R |R – DP| 12 5.629 -0.002 0.007 .006 D |R – RP| 12 -53.521 0.028 ** 0.010 .410 D |R – DP| 12 29.919 -0.015 ** 0.006 .413 SPENDING R |R – RP| 10 1.506 -0.001 0.004 .004 R |R – DP| 10 18.846 -0.009 0.018 .026 D |R – RP| 10 -42.086 0.022 0.015 .193 D |R – DP| 10 24.908 -0.012 * 0.007 .267 DEFENSE SPENDING R |R – RP| 10 40.349 -0.020 0.012 .249 R |R – DP| 10 -17.997 0.010 0.027 .017 D |R – RP| 10 23.259 -0.011 0.030 .016 D |R – DP| 10 33.359 -0.017 ** 0.006 .464 * significant at .10 level ** significant at .05 level ***significant at .01 level

Republican Party while, at the same time, we would expect them to perceive a substantively large increase in the distance between themselves and the Democratic Party (likewise for

Democratic party identifiers vis-à-vis the two parties). The results in Table 11.8 are consistent with this prediction across all of the issue dimensions and for most of the individual measures.

The results for the ideological dimension are particularly noteworthy. Republican and

Democratic identifiers perceive either no trend or a declining trend in the distance between themselves and their own parties. Whereas both Republican and Democratic identifiers reported significant and substantively large increases in the distance between themselves and the opposite party over the time series. This table in particular demonstrates that the

345

polarization on ideology, accurately perceived by the mass public and partisan identifiers, is a rational response to the issue positions of their respective constituents. The Republicans perceive a declining distance between themselves and Republican elites (-0.006), and mass

Democrats identify a similar negative trend in the distance between themselves and Democratic elites, though it is statistically insignificant (-0.006). Furthermore, by a magnitude between two and three times, the explanatory power of the oppositional trend models outperforms the constituent models. The linear trend in the distance between mass Democrats and the

Republican party explains almost 68% of the model variation (R2 = .675). While not as strong, the Republican oppositional model relative to the Democratic Party (R2 = .432) outperforms the

Republican constituent model (R2 = .320) by a significant margin. For spending and defense spending, the only significant models are the constituent models. Both the model on government spending (-0.012) and the model on defense spending (-0.017) show declining distance between the Democratic party identifiers and their elites, explaining between 25% and

50% of the model variation. The jobs model has significant decreasing distance between both the Republican constituents and the Republican Party (-0.017)and the Democratic constituents and the Democratic Party (-0.015). Furthermore, Democratic identifiers perceived increasing distance between themselves and the Republican Party (0.028). All three of these models explained between 40% and 50% of the model variation.

As noted in Chapter 2, culture war skeptics argue that elites have polarized independent of and despite of mass moderation over the past few decades. And more recently, Fiorina expresses doubt that even partisan sorting can explain why elites have ideologically diverged from the mass electorate. The puzzle for the skeptics is thus why elites have polarized away from the mass electorate they are supposed to be responsive to. There is an obvious and easy answer to this apparent puzzle: the elites haven’t grown more distant from the mass electorate,

346

at least not in terms of alienation from their constituents. Indeed, the opposite has occurred.

Partisans at the mass level believe their own party elites have either remained constant or even

grown closer to them since the 1970’s. As measured by the perceived Euclidian distance from

the candidates, neither the parties nor the candidates have diverged from the mass public on

several issue dimensions (e.g. jobs). And where they have diverged, the divergence of the

parties is explained almost entirely by the perceptions of the oppositional party identifiers.

Mass Republicans believe that the Democratic elites have polarized relative to their own

positions on the issues and ideology, and the same goes for mass Democrats vis-à-vis Republican

elites.

CONCLUSION

I examined the perceived distance on the political issues between the mass public and the parties and candidates. On jobs and ideology, however, there is a significant trend in the average respondent self-placement. Both trends are in a conservative direction, with the

American public becoming more conservative and less in favor of government responsibility for job creation over the past three decades. On the ideological dimension, respondents report greater distance between themselves and the Democratic Party and its candidates, while the regression coefficients on the absolute distances between respondents and Republicans are negative, suggesting that the mass public has moved closer to the Republican Party on the ideological dimension. This is coupled by a perception among the mass public that the

Republican Party has become more conservative over the last three decades, according to the significant coefficients reported in the first section of Table 11.4. Just as with abortion, the total distance measures indicate that the parties are more distant from the average respondent today than in the past. On jobs, government spending, and defense spending, there are no significant findings of increasing polarization between respondents and the parties or candidates. Indeed,

347

there are negative coefficients for most of the total distance models reported on these issues, indicating a declining perceived distance between the respondents and the partisan elites. This is particularly so on the ‘government responsibility for job creation’ dimension, with respondents reporting that both the Republicans and the Democrats have moved closer to them on this dimension.

There is strong evidence that, to the degree that the parties have diverged from one another and from the center of the mass of the electorate it is in response to divergence on the same issue dimensions by their own party identifiers. Party identifiers perceive little to no distance between themselves and their parties on the issue dimensions on average, while they perceive a substantial increase in the distance between themselves and the opposite party over the course of the time series. Thus what perceived polarization between the mass electorate and the party elites I find is accounted for, not by an alienation of the elites from the voters, but rather by partisan polarization. It is a function of the partisan polarization that has occurred since the 1970’s. Republican identifiers have polarized on the issues and ideological dimension relative to Democratic identifiers. Interestingly, given this fact, their perceptions of the parties and candidates could have remained constant, and there still would have been an increase in the distance between partisans of one party and the elites of the opposite party, as that constant position would be further distant from the new and polarized position of the partisan identifiers. What I find is that Republicans at the mass level view the Democrats as increasingly distant from their own positions, and likewise for mass Democrats and Republican elites.

Polarization, not alienation, has occurred. Both the mass and elites, the party in the electorate and the party in government, have polarized on the issue and ideological dimension. In the next chapter, I continue to assess the relationship between mass and elite polarization. I will use an objective measure of elite ideology in Chapter 12, rather than perceived ideological placement

348

of the elite partisans employed in this chapter. Using the D-W Nominate scores for Congress, I attempt to assess the causal direction between mass and elite polarization as well as continue to assess the degree to which both have polarized on the ideological dimension.

349

CHAPTER 12: WHO IS WAGGING WHOM? RESPONSIVE PUBLICS, RECEPTIVE ELITES

“Party conflict is pervasive in the House.” – Stonecash

““(We) used to be able to do more together on a bipartisan basis than seems possible these days. I’m not sure exactly why.” – Vice President Dick Cheney

The relationship between the mass electorate and the elites of American politics is a complicated and oft-debated question that has received a great deal of attention from scholars and remains a hotly contested issue in political science. It is particularly relevant on the question of political polarization in the electorate. First, elite polarization could be a causal variable explaining increasing polarization at the mess level. Second, the extent and nature of the link between masses and elites can explain voting behavior, party platform and issue advocacy decisions at the elite and activist level. Third, these factors may operate casually on significant aspects of electoral trends and realignments. These inter-related factors are aspects of the endogenous relationship between partisanship, ideology, issues, and voting behavior.

The puzzle of partisan polarization is one of the more interesting and mysterious developments of the late Twentieth and early Twenty-First centuries. How is it that in an increasingly independent, split-ticket voting, valence and symbolic politics-oriented environment where the days of the party machine are long past (as a consequence of a series of

Progressive era reforms), the behavior of legislators in Congress has become increasingly polarized and partisan? As Stonecash et al. note, “Members of each party are more likely to join together and vote against the other party. The parties increasingly adopt sharply different policy positions” (Stonecash, Brewer, and Mariani 2003). This argument is consistent with the trends in

DW-Nominate scores of Congress over the past 40 years (Poole and Howard 1997; Poole and

Rosenthal 2001). In the House, the 105th Congress was sharply divided in comparison to House

350

nominate scores two decades previous. Contrary to Darcy’s earlier finding that congressional parties were no more constrained than the public (Darcy 1980), the parties have become more ideologically coherent and ‘constrained,’ in Conversian terms, and further distant from one another (Ono 2005). This trend is apparent in Figure 12.1, tracking the change in the average 1st dimension D-W Nominate score for congressional Republicans and Democrats in each chamber over the last 50 years. Jacobson finds a growing disparity in congressional support for presidential initiatives, arguing that the degree to which the president and members of the opposing party in Congress share constituencies has declined (Jacobson 2002). The natural consequence of this is a decline in bi-partisan legislative compromise, the heightening of political conflict over legislative action, and increased combativeness with a president from the opposition party. Jones suggests that party polarization is a more significant factor in policy agenda stalemates (gridlock) than divided government (Jones 2001). Ono argues that the two party caucuses in Congress have become more ideologically unified, as evidenced in the increase in party unity scores, and more divergent from one another (Ono 2005).

One explanation for partisan polarization in Congress is that the political party apparatuses have been “hijacked” by amateur, ideologically extreme activists (Fiorina, Abrams, and Pope 2004). The ‘storming of the gates’ to the Republican party nominating conventions by religious conservatives is a significant factor in the partisan polarization identified by Layman

(Layman 2001). Others have attributed it to the increasing influence of party organizations on candidate selection (Fleisher and Bond 2004; Fleisher and Bond 2000). Stonecash and his colleagues argue that partisan polarization in the Congress is a consequence of a long-term secular realignment and social change. Constituencies in both parties have become more homogenous and more distinct from one another. “Realignment has brought the Democrats an electoral base that is less affluent, urban, and non-white. Republicans have acquired an electoral

351

Figure 12.1: Average Partisan House & Senate 1st Dimension DW-Nominate Scores 1954-2004

base that is more affluent, suburban and rural, and primarily white” (Stonecash, Brewer, and

Mariani 2003). Stonecash makes a district-based argument, positing that the districts from which congresspersons are elected and reelected have become more homogenous and thus contributing to the realignment, partisan polarization in the House, and the differing issue positions the parties have taken up in this newly aligned era (Stonecash 2005; Stonecash,

Brewer, and Mariani 2003; Stonecash et al. 2000; Stonecash 2000). However, as is apparent in

Figure 12.1, this trend has also occurred in the Senate (Poole and Rosenthal 1984), suggesting the effect of partisan polarization may not be dependent on changes in the district but may be a function of factors operating at a higher level of aggregation (Gelman et al. 2005). Ono’s argument that the growing incumbent advantage and uncompetitive districts has little to with redistricting but rather has an origin in factors that have operated on a national basis (party affiliation) is consistent with this view (Ono 2005).

352

On a contrarian note, while Brady and Han recognize that party elites have clear ideological differences and that parties have become more cohesive, they argue there is more bipartisan action in Congress now than in the past, perhaps due to the ambiguous ‘polarization’ of the masses. They argue “the difficulty of interpreting polarization in the electorate ultimately constrains elites” (Brady and Han 2006). Brady and Han define bipartisan unity as the

“percentage of partisans who vote with the majority of their party when it is not a party vote”

(Brady and Han 2006). So, when majorities in both parties pass legislation together, this is an act of bipartisanship. They conclude that “today’s level of bipartisan unity is close to as high as it has ever been.” Campbell and Cannon argue this inflates ‘bipartisanship’ as most of these votes are

“noncontroversial,” such as reprimanding Iran for threatening behavior and designating the birthplace of President Clinton a National Historic Site. Campbell, in response, makes an argument in every way the opposite of that of Fiorina’s and, to a lesser extent, Brady and Han.

While most scholars have reached a consensus that partisan polarization has occurred while opinion polarization, to the extent it exists, is much murkier, Campbell argues just the contrary.

He asserts that polarization “may not always be well represented by the parties” (Campbell and

Cannon 2006).

Mass partisan polarization is a consequence of the degree to which segments of the electorate, elites, or sub groups within the electorate have increasingly identified with one party rather than the other. This increasing identification may be recursively related to partisan politics. The groups themselves become more political and partisan in their outlook, and the politics of the parties and the elites as a consequence cater more to the interests of these increasingly loyal constituents. Layman and Carsey argue that the traditional theory of conflict displacement (as one dimension of partisan conflict emerges, the previous dimension must decline in salience and relevance) is inaccurate. Their “conflict extension” model suggests that

353

multiple dimensions of issue conflict can coexist (what could be termed ‘ideological realignment’ as I will argue later). As such one need not argue that economic class is no longer relevant to suggest that partisans have aligned on one or more other issue dimensions (Layman 1999;

Layman and Carsey 2002, 2003). Layman in particular argues that the social issues, or “Culture

War” dimension has become more salient and that partisan conflict at the mass and elite level has become increasingly structured along this dimension (Layman 1996, 1997, 1998, 1998, 1999,

2001; Layman and Carmines 1997). Bolce and De Maio concur, arguing antipathy towards

Christian fundamentalists has been on the rise in a society that increasingly divides itself along the lines that Hunter first theorized (Bolce and De Maio 1998, 1999, 1999). Resolving some methodological issues in previous studies, Brooks and Manza provide evidence of limited but significant changes in group-specific voting coupled with much larger changes in religion-based partisanship and party coalitions (Brooks and Manza 2004; Brooks 2002).

What direction does the arrow point in the mass / elite relationship? The consensus in the literature is that the elites wear the pants in the family (Dalton 1987; Herrera 1992;

Hetherington 2001; Layman 1996; Miller 1986; Aldrich 1995; Poole and Howard 1997; Collie and

Mason 2000; Alverez 1997; Alverez and Nagler 1995; McClosky, Hoffmann, and O'Hara 1960).

Elites signal to a largely uninformed and unconstrained public what the relevant and salient issues of the day are and structure the political conflict around those issues. In response, the attentive public shifts its positions to conform with the elites they identify with (Arnold 1990).

As noted earlier, Darcy argues that consensus among congressional elites and the mass public is similar, though the issues on which consensus exists differs. And while the congressional elites exhibited greater internal constraint, when external constraints were taken into account, this difference washed away given that the congressional elites were drawn from a narrower band of society (Darcy 1980). Though more recent trends in congressional polarization suggest

354

otherwise (Poole and Rosenthal 1984, 2001). Zaller’s seminal work on public opinion addresses the relationship between mass and elite behavior, examining the process by which individuals express political opinions and the relationship between those opinions at the mass and elite levels. Zaller explicitly rejects Converse’s model of a ‘constrained’ public with a structured belief system, arguing instead that the expressed opinions found in public opinion surveys reflect the receptiveness of those individuals to the elite ‘message’ on the issues. Attentive publics are exposed to a great deal of information, but filter that information through their prior commitments and beliefs (i.e. conservatives pay attention to Rush Limbaugh and Sean Hannity while ignoring and discounting Keith Olberman and Jon Stewart). Thus politically aware individuals exhibit greater knowledge but stable preferences, while the masses are the weather vane of politics: exposed to few political messages but more inclined to accept the ones they receive or are ‘primed’ to accept at that particular moment in time (Zaller 1992). Other research has looked to the nature of issues themselves, identifying ‘easy’ and ‘hard’ issues with the easier issues more likely to be employed by elites to move the mass public in their direction (Conover,

Gray, and Coombs 1982).

Fiorina expressly argues that elites have taken up increasingly polarized positions as a result of interest group and activist pressure while the mass electorate has increasingly moved toward a consensus on moderation and tolerance on these social and cultural issues. He goes so far as describing this elite polarization coupled with mass moderation as the “Hijacking of

American Democracy” (Fiorina, Abrams, and Pope 2004). This view, however, is incongruent with two major arguments on the logic of partisan behavior and the relationship between elites and the masses in terms of shaping political opinion. Zaller argues that the adoption of opposing issues by political elites can have a “polarization” effect where the masses follow the cues of their party’s elite and adopt the position of their party (truest among the most attentive

355

individuals in the electorate). However, if public sentiment strongly moves towards a consensus position, as Fiorina argues is the case with many of the culture issues in the culture wars, Zaller asserts that political elites of both parties will adopt that position and bring the public towards that position, thus producing a “mainstream effect” (Zaller 1992). Elite position-taking in this circumstance should reflect no partisan division and produce greater consensus within the electorate not, as Fiorina suggests, a greater consensus occurring as parties adopt increasingly divergent positions on the issue.

One of the models of mass politics assumes that elites drive and determine the nature of polarization. Polarization scholars have concurred, either siding with the belief that elites are pulling the electorate (or that portion of the electorate responsive to politics) along for the ride, or with the conclusion that elite politics occurs independent of the mass electorate. Is mass polarization the result or the cause of polarization among the elites? Is it that public opinion has become polarized, forcing the parties to move with the public in order to maintain electoral support? Or has the electorate become more coherently aligned along partisan dimensions because of a polarization of the parties themselves? Or both?

Some scholars have strongly argued that polarization in the electorate is the direct result of polarization among elites (Carmines and Layman 1997; Carmines and Gopoian 1981; Adams

1997; Abramowitz and Saunders 1998; Carmines and Woods 2002; Layman and Carsey 2002;

McCloskey 1960; Hetherington 2001). Whether it is a matter of some issues falling out of favor while new issues gain salience or an “expansion” of conflict across a host of issues, these scholars has argued that elites lead and masses follow. Certainly there is strong evidence of partisan polarization in Congress (see Figure 12.2). The ideological distance between the

Democrat and Republican parties has increased substantially over the last 50 years, as measured

356

by the first dimension D-W Nominate scores. Layman and Carmine argue that when Democrats and Republican elites are polarized on an issue, and party identifiers become cognizant of those differences, then a significant portion of those individuals respond by adjusting their party ties to conform to their issue positions (conversion) while others respond by adjusting their issue positions to conform with the party identification (adaptation). Thus the mass public responds to the polarization at the elite level (Layman and Carsey 2002). Hetherington finds that greater partisan polarization in Congress has clarified the parties’ ideological positions as perceived by the American public, and thus has increased party importance and salience on the mass level

(Hetherington 2001). For Fiorina, elite polarization occurs in spite of a largely centrist and moderate mass electorate and thus is a significant breakdown in the institution of electoral democracy in America. Elites have certainly polarized (see Figure 12.2). Furthermore, they have done so independent of gerrymandering. The polarization trend in Figure 12.2 is apparent for both the House (subject to gerrymandering) and the Senate (not subject to gerrymandering).

FIGURE 12.2: PARTISAN POLARIZATION IN CONGRESS, 1954-2004

0.9

0.8

0.7

0.6

0.5

0.4

0.3

0.2 MeanDifference: D-W Nominate Scores 0.1

HOUSE SENATE

357

For Fiorina, this elite polarization occurs independent of the masses, hence the mass electorate is largely unresponsive to shifts in the alignment of elites (Fiorina and Levendusky

2006). The roots of this orientation of the polarization literature can be found in the decades of work on mass public opinion. Converse first posited an uninformed and unstable electorate– essentially unequipped to participate in elections as traditionally conceived—and a variety of scholars have provided arguments to that effect since (Converse 1964; Blau 1977). This aspect of mass behavior provides the foundation for elite-driven mass politics and, consequently, polarization. This stands in contrast to the literature purporting to demonstrate that mass electorates are responsive and rational (Page and Shapiro 1992; Alverez 1997; Bartels 1986;

Conover and Feldman 1989; Gerber and Green 1998; Popkin 1994). Whether it is a matter of responsive electorates, rational publics, or low-information rational voters, polarization in spite of the disposition of the voting public is inherently problematic. In any electoral environment where voters–however imperfectly or incompletely–perceive exogenous shocks, economic and political indicators, and the behavior of elected officials, polarization without the sanction of constituents spells political suicide to elites so inclined.

Furthermore, the most significant study of elites can be found in the decades of literature on the U.S. Congress where hyper-responsiveness to constituent concerns in service, position-taking, political advertisements, etc. has been the norm rather than the exception

(Jacobson 2000, 2002; Hill and Hurley 1979; Collie and Mason 2000; Mayhew 1974). Without a responsive electorate, there is no electoral connection that elites are obligated to respect, and hence the distributive rationale for congressional action would be non-operative.

Fiorina’s elite polarization story (see Figure 12.3) essentially doesn’t make sense in a world where electorates are rational and responsive, as a unilateral shift to the poles of the

358

FIGURE 12.3: FIORINA’S ELITE POLARIZATION WITH CENTRIST ELECTORATE*

T: Functioning Electoral System with Parties Located Near Voters

DEM VOTERST REP

ELITET ELITET

T+1: Increasing Polarization of Partisan Elites Away from Voters

DEM VOTERST+1 REP

ELITET+1 ELITET+1

T+2: “Hijacked” Political System with Elites Polarized from Voters

DEM VOTERST+2 REP

ELITET+2 ELITET+2

*Points on line reflect the hypothetical mean position of the voters and partisan elites

359

electoral distribution would lead to Downsian party marginalization. Rationalizations offered to explain elite non-responsiveness, such as Fiorina’s theory that elites anticipate ideological challenges from the poles in primaries, fundamentally detach the elected official from his or her interest in re-election (as opposed to standing for election, or receiving interest group benefits, or rewarding partisans, etc.). The argument does not merely suffer from theoretical problems, but also an unresponsive electorate is incongruent with apparent aggregate rationality in terms of short-term effects on elections, increasing homogenization of districts tracking with increased elite polarization, and the effectiveness of political advertising campaigns. The evidence suggests that voters are paying attention to some signals from the political environment and that they hold officials accountable in elections. And while district polarization (as a function of gerrymandering) may account for polarization in the House, it does not explain polarization among candidates for offices with constituencies at higher levels of aggregation.

I argue that elite polarization is recursively related to mass polarization (see Figure 12.4). It is true that elites attempt to influence the mass public through the framing and emphasis of certain issues, but it is just as true that elites strive to detect those issues which have become salient to the electorate and shape their position-taking accordingly. Scholars examining the factors of polarization in American politics miss half of the story when ignoring the influence that shifts in the mass electorate have on elite behavior. Electorates change, they change in comprehensible ways, the issues and opinions of aggregate electorates shift sometimes independent of elite preferences, and these shifts produce shifts in electoral behavior. As a result, the optimum party platform in any given election changes as well. Why don’t elites merely determine the optimum political positions on the relevant set of issues and adopt those as their party platform? There are numerous factors that prevent this simple strategy. Elites operate in a world of uncertainty. They are uncertain as to what issues have become politically

360

FIGURE 12.4: RATIONAL POLARIZATION GIVEN SHIFTS IN CONSTITUENCIES

Type 1: Partisan Voting Constituencies Signal Divergence on Salient Issues*

DemVT+1 DemVT RepVT RepVT+1

Type 1: Elites Diverge in Response to Voter Polarization

D-EliteT+2 D-EliteT+1 R-EliteT+1 R-EliteT+2

Type 2: Elites Signal Divergence on Salient Issues

D-EliteT+1 D-EliteT R-EliteT R-EliteT+1

Type 2: Partisan Voting Constituencies Respond to Elite Polarization

DemVT+2 DemVT+1 RepVT+1 RepVT+2

*Points on line reflect the mean position of the voters and partisan elites

361

salient among the public. They are uncertain as to what portion of the electorate will turnout and in what proportions, hence making the nature of the responsive public unclear. They are uncertain as to what exogenous shocks will influence the public and what impact those will have on public opinion (Alverez and Franklin 1994; Bartels 1986; Calvert 1985; Enelow and Hinich

1981; Ferejohn and Noll 1978; Shepsle 1972; Wright and Goldberg 1985; Hinich and Munger

1994).

Given this, elites attempt to gain and influence information. They make efforts to gather signals from the public as to what issues they care about through polls and electoral results…hence reducing uncertainty as to the electoral calculus. However, even if they manage to determine what the optimum position is, they may not be in a position to adopt it. Position- taking does not occur in a vacuum. The positions that parties and officials can take are bounded by the positions they have taken in the past, and hence reputations are relatively sticky. As such, elites attempt to influence publics through party platforms, campaign messages, statements, press conferences, etc…in an effort to guide public opinion to regard as salient those issues and policies where the party has strength and on which the electorate could be decisive (Alverez

1997; Calvert 1986; Dalton, Beck, and Huckfeldt 1998; DeSart 1995; Feddersen and Pesendorfer

1997, 1999; Huckfeldt and Sprague 1987; Huckfeldt and Sprague 1988; McKelvey and

Ordeshook 1985, 1986; Morton 1993; Palfrey and Poole 1987). Emergent issue cleavages are, as

Petrocik and others have argued, thus a function of changes in the mass electorate that yield shifts in the party platforms and messages of the two parties and a function of the elites within parties shifting their message so as to expand their base or to take advantage of a perceived avenue of electoral opportunity. Where one party successfully produces a credible platform or policy output that appeals to a sufficiently large portion of the electorate (or conversely, where

362

one party alienates a sufficiently large portion of their political base) then the conditions are set

for realignment.

MODELS OF POLARIZATION Polarization is a phenomenon where some combination of the above factors serves to push the parties, groups, elites and/or the aggregate electorate apart. These centrifugal forces pull voters towards the poles of political opinion and polarize their beliefs, opinions, and actions relative to other groups in the political space. The predicates of polarization and the consequences of polarization are:

Aggregate Polarization

1) Mass Ideological Polarization: where the distribution of the mass electorate along the ideological dimension shifts away from the center and towards one of or both of the poles. This is reflected in greater dispersion of opinion across the ideological dimension. Defined by: a. Aggregate shifts from the center to one or both of the poles on issue, policy, or ideological dimensions. b. Within-group (or party) homogenization along issue, policy, or ideological dimensions. c. An increase in the distance between the in-group and out-group on issue, policy, or ideological dimensions.

2) Elite Ideological Polarization: where the distribution of the elites (politicians, activists, etc.) on the ideological dimension shifts away from the center and towards one of or both of the poles. This is reflected in greater dispersion of opinion across the ideological dimension. Defined by: a. Aggregate shifts from the center to one or both of the poles on issue, policy, or ideological dimensions. b. Within-group (or party) homogenization along issue, policy, or ideological dimensions. c. An increase in the distance between the in-group and out-group on issue, policy, or ideological dimensions.

363

Polarization Causes: Mass vs. Elite

3) Electorate-Driven Polarization: where partisan electorates become more homogenized along issue or policy dimensions, reducing the uncertainty parties have as to the policy preferences of their respective electorates and permitting them to construct clear responses in terms of positions and policies that the more homogenized electorate can receive. Aspects: a. Become more polarized along current policy or ideological dimensions b. Coalesce around new salient issues or policies on which they are more polarized

4) Elite-Driven Polarization: where partisan elites become more homogenized along issue or policy dimensions and hence reduce the signal-to-noise ratio for electorate decisions. Aspects: a. Have divergent beliefs as to the direction of a move along an issue, policy, or ideological dimension position that will yield an expansion of their partisan electoral base. b. Have divergent beliefs as to which position to adopt on a newly salient issue that has entered the policy space.

Polarization Causes: Constituents, Partisans, & Partisan Constituents

5) Party Image Polarization: polarization of or in response to the perceived positions of the political parties. a. As the electorate or identifiers perceive a shift from the center to one or both of the poles on the issue, policy or ideological dimensions of the party elites, they respond accordingly: shifting party allegiance consistent with their issue positions or changing their positions to conform to their party (i.e. polarize). b. As elites perceive a shift in how they are perceived by the public, they seek to capitalize on this change by shifting in the polar direction congruent with the public or the partisan electorate’s perception.

6) Constituent Polarization: polarization of or in response to the constituents of the elected officials of the political parties. a. As constituents perceive shifts from the center to one or both of the poles on the issue, policy or ideological dimensions of the party elites, they respond accordingly: shifting party allegiance consistent with their issue positions or changing their positions to conform to their party (i.e. polarize). b. As elites perceive a shift in their constituent’s position towards one of the poles on issue, policy, or ideological dimensions, they seek to capitalize on this change by shifting in the polar direction congruent with their constituent preferences, or are replaced by someone who will.

All of these phenomena–either through responsive publics, responsive elites, or some combination of the two–can yield political polarization. We can assess polarization by coupling

364

shifts in aggregate issue positions of the electorate with nominate scores in Congress over the same time period.

MEASURES OF POLARIZATION

Mean Deviation from T1. In order to measure the changing distribution of belief in the mass public and among partisan identifiers and elites, we need a measure of the average ideological position of these groups and sub-groups. Deviations from past average ideological positions will allow us to assess whether the mass public is moving towards one pole or the other as well as assess the difference in ideological positions between groups.

Dispersion. In order to measure dispersion, we need a measure that both reflects the relative distance that individual respondents differ from one another as well as taking into account the proportion of opinion located in the extremes relative to the center of the distribution. The traditional measure of dispersion (or inequality in the economics literature) is variance. As opinion dimensions become more polarized, variance (or alternatively, the standard deviation) should increase.

Consolidation. The consolidation measure refers to the relative agreement or consensus within groups and their relative disparity across groups. This “identity group” polarization is measured using a difference of means for ideology between the groups to assess between group differences while we use the standard deviation measures to assess within group consolidation on the ideological dimension. The difference being that identity group polarization is characterized by decreasing variance within the group.(DiMaggio, Evans, and Bryson 1996)

365

POLARIZATION HYPOTHESES

American Public Hypotheses Mass Public Ideological Polarization over Time Ho: No change or centripetal change in the distribution of ideology in the mass public over time.

Ho1: Constant or decreasing variance in the ideology of the mass public over time. Ho2: Constant or centripetal change in mean ideology of the mass public over time.

Ha: Centrifugal change in the distribution of ideology over time.

Ha1: Increasing variance in the ideology of the mass public over time. Ha2: Centrifugal change in mean ideology of the mass public over time.

Elite Ideological Polarization over Time Ho: No change or centripetal change in the distribution of elite ideology over time.

Ho1: Constant or decreasing variance in elite ideology over time. Ho2: Constant or centripetal change in mean elite ideology over time.

Ha: Centrifugal change in the distribution of ideology.

Ha1: Increasing variance in the ideology of the mass public over time. Ha2: Centrifugal change in mean ideology of the mass public over time.

Mass Elite Ideological Polarization Ho: No relationship between mass ideology and elite ideology.

Ha1: Simple: mass ideology determines elite ideology. Ha2: ME Aftershock: Lagged mass ideology determines elite ideology.

Elite Mass Ideological Polarization Ho: No relationship between elite ideology and mass ideology.

Ha1: Simple: elite ideology determines mass ideology. Ha2: EM Aftershock: Lagged elite ideology determines mass ideology.

Constituents Elite Ideological Polarization Ho: No relationship between constituent ideology and elite ideology.

Ha: Squared differences between constituent ideology and elite ideology decline over time

Political Party Hypotheses Inter-Partisan Identifier Ideological Polarization over Time Ho: No change or a decrease in the average ideological difference between Republican identifiers and Democratic identifiers over time.

366

Ha: Increase in the average ideological difference between Republican identifiers and Democratic Identifiers over time.

Intra Partisan Identifier Ideological Polarization over Time Ho: No change or centripetal change in the distribution of ideology among partisan identifiers over time.

Ho1: Constant or Increasing variance in the ideology of the partisan identifiers over time. Ho2: Constant or centripetal change in mean ideology of partisan identifiers over time

Ha: Centrifugal change in the distribution of ideology among partisan identifiers over time.

Ha1: Decreasing variance in the ideology of partisan identifiers over time. Ha2: Centrifugal change in mean ideology of partisan identifiers over time.

Party Identifier Party Elite Ideological Polarization Ho: No relationship between party identifier ideology and party elite ideology.

Ha1: Simple: Party identifier ideology determines party elite ideology. Ha2: ME Aftershock: Lagged party identifier ideology determines party elite ideology.

Party Elite Party Identifier Ideological Polarization Ho: No relationship between party elite ideology and party identifier ideology.

Ha1: Simple: Party elite ideology determines party identifier ideology. Ha2: EM Aftershock: Lagged party elite ideology determines party identifier ideology.

DATA SOURCES, VARIABLES, & METHODS

Data Sources

There are two primary sources of data for this analysis. One is the series of American

National Election Studies from the Cumulative Data File. The cumulative data file consists of variables derived from the 1948-2004 series of biennial ("time-series") SRC/CPS National

Election Studies78. My analysis uses data ranging from 1954-2004. The second source of data for this analysis is the series of congressional ideology (D-W Nominate) scores developed by Poole

78 RELEASE VERSION: 20051031 (Oct 31, 2005)

367

and Rosenthal. Use of the D-W Nominate scores was necessary for cross-Congress

comparisons.79 Analysis was conducted using SAS version 9.0. The dataset employed in this

analysis was constructed using a multi-step process. First, the Poole-Rosenthal D-W Nominate

scores (hereafter referred to as nominate scores) were merged with the cumulative file by year,

state, and district. This created a data set where each respondent in the NES cumulative file had

corresponding nominate scores for the congressional representatives from his or her district and

senators from his or her state. The second stage in constructing the dataset for analysis involved

calculating means and standard deviations for all relevant variables in the merged dataset. Then,

using those means and standard deviations, I created a set of variables which translated the

means and standard deviations for the respondent ideology and legislator nominate scores into

z-scores for the partisan groupings and the full sample.

= Equation 12.1: Z-Scores ௒೔ೣି௒തೣ ఙೣ Where:

= the observed value of Y for year. = the mean value of Y for year. ௜ .the standard deviation of Y ݔfor year = ܻ ത ݔܻ ߪ௫ ݔ The third stage involved estimating averages for each of the relevant computed and

merged variables for each year of the biennial NES time series. Those values were then entered

into a separate dataset for the purpose of statistical analysis. The ultimate product is a data set

of average respondent ideology and average legislator nominate scores by year (1954-2004) for

the means, standard deviations, and z-scores as calculated from the second data set. The total

sample size for the final dataset is 27 observations corresponding to each NES biennial survey.

79 See (Poole and Rosenthal 2001) for further description of D-W Nominate Scores

368

Variables

The means and standard deviations for selected variables included in the analysis can be found in Table 12.1. The Party ID variable used for this analysis collapses the traditional 7-point party identification scale into a 3-point party identification scale with Republican and

Democratic leaners collapsed into the Republican and Democrat categories with only non- leaning independents included in the Independent category. The ideology variable is the traditional 7-point scaling of ideology that the NES incorporated in 1972.

In addition, the thermometer for conservatives is included as a proxy for ideology. The Pearson’s correlation between ideology and the conservative thermometer from the cumulative NES cross-section is just .637 (P<.0001), however the conservative thermometer was asked as early as 1964, and thus it allows for several more degrees of freedom in any statistical model using it to assess political polarization. The conservative thermometer was not asked of respondents to the 1978 NES. The conservative thermometer for 1978 is set at the mean of the two closest years (1976 & 1980) to retain biannual continuity. However, given the distance between the conservative thermometer and ideology, the models assessing elite-mass and partisan polarization use the 7-point ideology measure. Except where the scales were the same, difference variables use average z-scores in their calculations. Method

Modeling the Relationship between Mass Ideology and Objective Measures of Elite Ideology

The polarization analyses consist of five distinct types of models utilizing OLS regression and GLS regression where appropriate. The first set of models asses simple regressions of year on the mass, elite, and partisan ideologies. Given the limited number of observations in the data set, more robust multivariate regression models simply are not possible, as they quickly expend the available degrees of freedom. For full sample ideological

369

Table 12.1: Selected Means & Standard Deviations

Variables N Means Standard Deviations Party ID (M) 27 3.601 0.166

Ideology (M) 17 4.261 0.086

Ideology (SD) 17 1.372 0.0624

Conservatives Therm (M) 21 52.222 1.189

Conservatives Therm (SD) 21 15.279 0.829

Republican Ideology (M) 17 4.933 0.168

Republican Ideology (Z) 17 0.487 0.075

Republican Con Therm (M) 21 59.879 1.905

Republican Con Therm (Z) 21 0.499 0.075

Democratic Ideology (M) 17 3.696 0.143

Democratic Ideology (Z) 17 -0.408 0.101

Democratic Con Therm (M) 21 46.513 1.860

Democratic Con Therm (Z) 21 -0.372 0.100

House Nominate Score 27 -0.014 0.061

Senate Nominate Score 27 -0.060 0.054

House Republican Nom Score 27 0.317 0.081

House Democratic Nom Score 27 -0.292 0.064

370

means, positively sloped parameter estimates indicate a shift towards the upper pole

(conservative) of the ideological distribution. For full sample ideological standard deviations, a

positive slope indicates an increase in political polarization, as the dispersion of ideology will

have increased over the time series. For the partisan variables, the direction of the slope and its

relationship to polarization is determined by which party it is. For example, for the Republican

ideological mean, a positive slope would indicate polarization, as the average ideology of

Republicans would be shifting in a more conservative direction. However, for the Democratic

ideological mean, a positive slope would indicate moderation, as the average ideology of

Democrats would be shifting the center. This distinction is best illustrated using the Z-scores.

The Republican mean Z-Score for ideology is 0.487. A positive slope would indicate that average

Republican ideology is moving away from the mean. The Democratic mean Z-score for ideology

is -0.408. Here a positive slope would indicate that average Democratic ideology is moving

towards the mean. These models are reported in Table 12.2.

( ) + Equation 12.2

݁ ݎଵ ݕ݁ܽܤ଴൅ܤൌܱܧܦܫ The second set of models (Table 12.4) assess the simple regression of year on the difference between Republican and Democratic ideology as well as the difference between

Republican and Democratic feeling thermometers on conservatives as a group. The table reports the unstandardized means and the standardized z-score for both ideology and the ideology proxy.

) ( ) + Equation 12.3

݁ ݎଵ ݕ݁ܽܤ଴൅ܤሺோ௘௣௨௕௟௜௖௔௡௜ௗ௘௢௟௢௚௬ି஽௘௠௢௖௥௔௧௜௖௜ௗ௘௢௟௢௚௬ ൌܨܫܦܱܧܦܫ The third and the fourth set of models include both simple OLS regressions and generalized least squared (GLS) regressions for the models assessing ideological polarization, partisan polarization, and elite versus mass causation of that polarization. As mentioned earlier,

371

whether or not elites lead and the masses follow is a significant source of contention among

scholars, though the emergent consensus suggests that elites condition changes in mass

attitudes. The models reported in Tables 12.4 and 12.5 assess polarization and elite-mass

causation empirically.

For each group under consideration, there are eight regression models to test this

relationship. Four models test a mass Æ elite causal relationship using the average ideology of respondents to predict the average ideology (nominate scores) of the elites (legislators).

Whereas the other four modes test an elite Æ mass inferential model where the average

nominate scores are used to predict average respondent ideology. For each group, a simple OLS

regression is tested for both the mass Æ elite and the elite Æ mass models. However, given that this analysis is a cross-sectional time-series, accounting for the potentially serious problem of serial correlation necessitates employing a GLS model using autoregression techniques. For each of the three sets of independent variables, a GLS model is estimated. The Durbin-Watson statistic testing for first order autocorrelation and the probability of positive autocorrelation are reported in Tables 12.4 and 12.5 for the appropriate models.

In order to assess the theoretical problem of mass vs. elite causation (i.e. which is the chicken and which is the egg) I use two lagged independent variables (2 year lag & 4 year lag) of mass and elite ideology. Thus three possible causal relationships are employed testing both mass Æ elite and elite Æ mass polarization. The first model sans lagged variables tests whether there is a simultaneous relationship between mass ideology and elite ideology. If mass and elite ideology are both predictive in the same year, then I conclude this is strong evidence supporting the recursive model of mass and elite polarization. Indeed, the recursive model is supported, even if we find lagged effects, when there is significant within-year causation between mass and

372

elite ideology. If lagged mass ideology (2 or 4 years) predicts elite ideology, then the mass Æ

elite model is supported. If, however, lagged elite ideology (2 or 4 years) predicts mass ideology,

then the elite Æ mass model finds support. In other words, if the mass or elite ideology from

1972 predicts the mass or elite ideology for 1974 or 1976, then we have a strong temporal basis

for pointing the causal arrow in one direction or the other. The last set of models in Table 12.5

assess whether the differences between Republican and Democratic partisan identifiers is

driving the observed differences between Republican and Democratic legislators, and vice versa.

Equations 12.4 – 12.9: Models for Mass Æ Elite & Elite Æ Mass Causation

( ) ( ) + Equation 12.4 ( ) ( 2 ) + Equation 12.5 ଴ ଵ ݋݈݋݃ݕ 4 ݁ ) + Equation 12.6݁݀݅ݏݏܽ݉) ܤ൅ ܤݐ݁݅݀݁݋݈݋݃ݕ) ൌ݈݅݁)ܱܧܦܫ ଴ ଵ Equation 12.7 ݁ ݈݀݁݃݃݃ܽݎ݋݈݋݃ݕ) +ݕ݁ܽ݁݀݅ݏݏܽ݉) ܤ൅ ܤݐ݁݅݀݁݋݈݋݃ݕ) ൌ݈݅݁ )ܱܧܦܫ ଴ ଵ Equation 12.8 ݁ + ( ݈݀݁݃݃ܽݎ݋݈݋݃ݕ 2 ݕ݁ܽ݁݀݅ݏݏܽ݉) ܤ൅ ܤݐ݁݅݀݁݋݈݋݃ݕ )ൌ݈݅݁)ܱܧܦܫ ଴ ଵ ݐ݁݅݀݁݋݈݋݃ݕ 4 ݁ ) + Equation 12.9݈݅݁) ܤ൅ ܤ݋݈݋݃ݕ) ൌ݁݀݅ݏݏܽ݉)ܱܧܦܫ ଴ ଵ ݁ ݈݀݁݃݃݃ܽݎݐ݁݅݀݁݋݈݋݃ݕ ݕ݈݁ܽ݅݁ ܤ൅ ܤ݋݈݋݃ݕ ൌ݁݀݅ݏݏܽ݉ ܱܧܦܫ ݁ ݈݀݁݃݃ܽݎଵ ݈݁݅ݐ݁݅݀݁݋݈݋݃ݕ ݕ݁ܽܤ଴൅ܤ݋݈݋݃ݕ ൌ݁݀݅ݏݏܽ݉ ܱܧܦܫ

The fifth set of models assess the squared differences between normalized legislator nominate scores from the two parties and normalized constituent ideologies scores by year for both the House and the Senate. The squared difference is calculated in order to retain the ideological differences between the constituent and his or her representative while permitting an assessment of the absolute difference between a legislator and his or her constituents. Using constituent ideologies means that a legislator’s ideological distance from respondents,

Republican and Democrat, from his district or, if he is a senator, from his state. Furthermore, there is an important difference between this set of models and those previous. For the squared difference variables, the differences were first calculated in the second data set. This was necessary as the constituent model assesses distances between a legislator and that legislator’s constituents. The aggregation makes this kind of analysis impossible in the third data set. The

373

problem with this analysis from the outset is that the ANES sample is not appropriate for state

and district level aggregations. While the constituent models reported in Table 12.6 are at the

party and house levels of aggregation, the differences between any one included legislator and

his constituent may rest on just one respondent…or it could be twenty respondents. While these

problems may average out over the long run, they do present a peculiar problem for this

analysis. The assessment of constituent effects may further necessitate examining only the

constituents from the party of the legislator (Fenno’s re-election constituency), however that

would only exacerbate the previously mentioned problem. While this set of models is presented

here tentatively, we hesitate to rely on them without further diagnostics.

( ) ( ) + Equation 12.4 ଶ ௖௢௡௦௧௜௧௨௘௡௧௜ௗ௘௢௟௢௚௬ି௟௘௚௜௦௟௔௧௢௥௜ௗ௘௢௟௢௚௬ ଴ ଵ ݁ ݎMOVEMENTݕ݁ܽ ܤ൅ ܤANALYSIS: ELITE RESPONSIVENESS TO MASS IDEOLOGICALൌ & ܨܫܦܱܧܦܫFINDINGS

Ideological Polarization over Time First, I will address the primary theoretical and empirical problem that students of political polarization grapple with: namely, well, polarization…is there any? The short answer is, yes. A great deal! Table 12.2 reports the linear trends in political ideology for the mass public, elite politicians, partisan identifiers, and party elites (legislators). The first model regressing year on party identification demonstrates a slight Republican trend over the course of the time series. It is thus not particularly surprising that the trend in mean ideology indicates a shift towards the conservative end of the ideological spectrum for the mass public in the aggregate.

Both represent evidence for mass ideological polarization, as both means are trending towards the poles of the distribution rather than the center. There was no apparent trend in average respondent feelings towards conservatives, though this finding does underscore the fact one must be cautious in using it as a stand in for ideology. Then again, perhaps the additional years in the time series accounts for the conservative FT’s poor showing. Of most interest in the

374

Table 12.2: Ideology, Conservative FT’s, and Nominate Regression Models by Year for Full Sample Standard 2 MODEL: DV = B0 + B1(year) + N Intercept Parameter Estimate Error R e Party ID (m) 27 -11.696 0.008 *** 0.001 0.544

Party ID (sd) 27 6.580 -0.002 ** 0.001 0.173

Ideology (m) 17 -2.425 0.003 * 0.002 0.155

Ideology (sd) 17 -6.728 0.004 *** 0.001 0.434

Conservative FT (m) 21 72.223 -0.010 0.022 0.011

Conservative FT (sd) 21 -27.263 0.021 0.014 0.103

Repub Ideology (m) 17 -19.158 0.012 *** 0.002 0.528

Repub Ideology (z) 17 -9.175 0.005 ** 0.001 0.428

Dem Ideology (m) 17 24.356 -0.010 *** 0.003 0.533

Dem Ideology (z) 17 16.772 -0.009 *** 0.001 0.737

House Rep Nom 27 -7.571 0.004 *** 0.001 0.609

Senate Rep Nom 27 -4.722 0.003 *** 0.001 0.540

House Dem Nom 27 7.508 -0.004 *** 0.001 0.955

Senate Dem Nom 27 8.044 -0.004 *** 0.001 0.854

* significant at .10 level ** significant at .05 level ***significant at .01 level

aggregate models is the significant positive trend in the standard deviation for respondent

average ideology. An increase in dispersion for ideology is confirmatory evidence in favor of

thecentrifugal change hypotheses for mass public ideological polarization. Ideologically

speaking, there is growing dispersion in the aggregate electorate as well as a shift in the average

ideological position towards the extremes of the ideological continuum and not to the center as

Fiorina and his fellow Culture Wars skeptics have suggested. Note in particular the standard

375

deviation model reported in Table 12.2. I find a statistically significant increase in the variance in ideology over the time series. This runs directly counter to Fiorina’s and the other Culture Wars skeptics who suggest that there has been a centripetal trend among the mass electorate over the last 40 years.

Partisan Polarization over Time

While one must sift through the aggregate findings for subtle bits of evidence on our theoretical questions, there is no such problem when it comes to the partisan models and partisan polarization. What is of interest is that both the partisan identifiers in the mass public model s and the partisan elite models evidence a strong trend in ideological polarization. The later has been well documented by Poole and Rosenthal (Poole and Rosenthal 1984; McCarty,

Poole, and Rosenthal 2006), but the former, as our previous discussion aptly demonstrated, is a major bone of contention. Fiorina argues that the party elites have become beholden to activists and interest groups while ignoring the mass electorate. They polarize to service the interested and passionate few at the expense of a largely centrist but unorganized public. But the results of this analysis indicate that the mass public is polarizing to some degree, and that partisan constituencies in the mass electorate are polarizing to a much larger degree. This is not a small point. An isolated party elite beholden to the organized (i.e. rich) and the passionate (I.e. the poles) could be a significant problem for American democracy. Certainly Fiorina thought so when he sounded the warning. But a party elite that is responsive to its adherents in the electorate—that may not be a problem at all. It sounds a lot like what parties are for. Indeed, it wasn’t so long ago that the APSA was explicitly calling for “responsible party government” where there were clear partisan differences, party discipline, and thus party responsibility. Is this a case of be careful what you wish for…or is the “hijacking of American democracy” really

376

just responsible party government with sinister music playing in the background? Table 12.2

finds strong evidence that the Democrat and Republican identifiers in the mass public have

Table 12.3: Models Regressing Differences b/w Republican & Democrat Identifiers by Year

Parameter Standard MODEL N Intercept Estimate Error R2

IDEOR-D(m) = B0 + B1 (year) + e 17 -43.515 0.023 *** 0.004 0.681

IDEOR-D(z) = B0 + B1 (year) + e 17 -25.946 0.014 *** 0.002 0.668

CONFTR-D(m) = B0 + B1 (year) + 21 -395.978 0.206 *** 0.025 0.787 e

CONFTR-D(z) = B0 + B1 (year) + e 21 -23.575 0.012 *** 0.001 0.816

* significant at .10 level ** significant at .05 level ***significant at .01 level

polarized along the ideological dimension. Both in terms of the direction of the mean (towards

the poles) and in the relative placement of the two parties on the ideological scale (the Z-score

model), there is strong statistical evidence that both parties have moved away from the other.

There is increased ideological disparity between Republican and Democratic identifiers since

1970.

Looking to Table 12.3, we can see that looking at the question of polarization in terms

of the difference between Republican and Democratic identifiers places the issue in stark relief.

There is strong evidence of an ever widening divide between Republican and Democratic

identifiers, even in our previously disappointing conservative feeling thermometer. Republican

377

and Democratic identifiers in the American electorate are moving away from each other and

taking up further distance positions on the ideological spectrum.

Mass Æ Elite vs. Elite Æ Mass Polarization

The part of this analysis most fraught with pitfalls is the assessment of temporal causation in mass and elite polarization. Is it a) mass-driven polarization or is it b) elite driven polarization? Or is it, as Fiorina argues, in fact, c) elite polarization irrespective of mass behavior

(polarized or no). Table 12.4 provides some powerful evidence on this point. It is evident from this table that there is no support for the recursive model of ideological polarization between the mass public and political elites. Average ideological scores for the mass public are not correlated with the nominate scores for legislators of the same year in the House, Senate, nor the two combined. Furthermore, there’s little support for the Elite Æ Mass hypothesis, as none of the models that use lagged nominate scores to predict contemporaneous average mass public ideology scores is significant. While this specific finding doesn’t speak to Fiorina’s argument, it does run counter to a significant sub-literature which suggests that elites drive mass political opinion (Hetherington 2001; Conover, Gray, and Coombs 1982; Converse 1964;

Sullivan, Piereson, and Marcus 1978; Mutz 2006; Hill and Hurley 1979). As can be seen in Table

12.4, I find no evidence that elite polarization is driving trends in the ideological disposition of the American public.

There is evidence in support of the Mass Æ Elite hypothesis. In all three sets of models, the lagged ideology of the mass public was a significant predictor of elite ideology. In the House and

Senate models, the ideological polarization of the public is a significant predictor of the polarization in our elite legislative bodies on a two year lag (Table 12.4). In the full congressional model, this relationship is significant at the .01 level, giving strong evidence that the apparent

378

relationship is not due to chance. For every single unit of change in the ideology of the masses there is a near full point (.749) change in the ideology of Congress. About half of the variation in the model is explained using the 2 year lagged ideology of the masses (0.549).

This suggests that elite polarization was, in fact, preceded by mass polarization. This result is consistent with my alternative formulation which argues that elites are sensitive to the opinions of the mass public and responsive to shifts in the aggregate views of the electorate.

The causal arrow points from the masses to the elites and not, as the consensus in the literature suggests, from the elites towards the masses. And this elite-responsive-to-mass citizenry relationship is apparent in both the House of Representatives and the Senate models, arguing against redistricting or district characteristics as the prime mover mass-elite ideology. This evidence alos runs directly counter to the Fiorina conjecture of a runaway political elite dashing to the poles while ignoring the largely centrist public. Rather than elites diverging completely independent of mass centrism, I find that the mass public has polarized on ideology, and that elite polarization has lagged behind the polarization in the mass public. It is the mass public that has departed for the poles first, while elite legislators have followed—polarizing in response to mass ideological polarization. This suggests that Fiorina’s expressed concern over the democratic process in light of political polarization is somewhat overblown. While political polarization in the legislature reduces the probability of compromise and incremental politics, it is not in and of itself undemocratic. When elites polarize in response to mass polarization, this is an indicator of representative democracy, not a departure from it. This responsiveness is evident at the level of the general mass public and the elites in the congressional institutions, but the most powerful evidence of elite/mass responsiveness is among partisan elites and mass partisans.

379

Table 12.4: Mass Public Respondents & Elite Legislators - Simple & Autoregressive Models of Ideology MODEL: DV = B0 + B1(IV) + e N Analysis Intercept Parameter Estimate Standard Error R2 D-W PR < DW House Nom = Ideology (m) 17 OLS 0.118 0.118 0.181 0.028 House Nom = Ideology (m) 17 GLM 0.148 -0.040 0.098 0.012 1.192 0.008 H Nom = Ideology (m) lag 2Y 17 GLM -0.905 0.207 ** 0.092 0.277 1.066 0.016 H Nom = Ideology (m) lag 4Y 17 GLM -0.181 0.036 0.137 0.006 0.823 0.003 Ideology (m) = House Nom 17 OLS 4.269 0.234 0.359 0.028 Ideology (m) = House Nom 17 GLM 4.270 0.244 0.363 0.031 1.860 0.293 Ideology (m) = H Nom lag 2Y 17 GLM 4.258 -0.073 0.433 0.002 1.8350 0.280 Ideology (m) = H Nom lag 4Y 17 GLM 4.252 -0.234 0.440 0.021 1.104 0.137

Senate Nom = Ideology (m) 17 OLS -0.388 0.084 0.130 0.028 Senate Nom = Ideology (m) 17 GLM 0.175 -0.047 0.092 0.018 1.132 0.021 SNom = Ideology (m) lag 2Y 17 GLM -0.821 0.187 * 0.094 0.232 1.201 0.035 S Nom = Ideology (m) lag 4Y 17 GLM -0.624 0.141 0.124 0.106 1.399 0.139 Ideology (m) = Senate Nom 17 OLS 4.270 0.324 0.497 0.028 Ideology (m) = Senate Nom 17 GLM 4.271 0.355 0.497 0.035 1.905 0.345 Ideology (m) = SNom lag 2Y 17 GLM 4.266 0.140 0.527 0.005 1.846 0.302 Ideology (m) = S Nom lag 4Y 17 GLM 4.277 0.349 0.465 0.042 1.667 0.181

Congress Nom = Ideology (m) 17 OLS -0.460 0.101 0.145 0.031 Congress Nom = Ideology (m) 17 GLM 0.170 -0.046 0.878 0.019 1.113 0.018 CNom = Ideology (m) lag 2Y 17 GLM -0.007 0.739 *** 0.140 0.549 1.895 0.331 C Nom = Ideology (m) lag 4Y 17 GLM -0.020 0.409 * 0.194 0.176 1.716 0.216 Ideology (m) = Congress Nom 17 OLS 4.270 0.307 0.442 0.031 Ideology (m) = Congress Nom 17 GLM 4.272 0.331 0.443 0.038 1.886 0.317 Ideology (m) = C Nom lag 2Y 17 GLM 4.262 0.014 0.508 0.001 1.795 0.284 Ideology (m) = C Nom lag 4Y 17 GLM 4.266 0.047 0.502 0.001 1.608 0.146 * significant at .10 level ** significant at .05 level ***significant at .01 level

380

While the models assessing the ideological trends of the mass public and political elites

failed to support the recursive hypotheses, there is ample evidence of a recursive relationship

between party identifiers and partisan elites in Table 12.5. All of the models that regress

partisan identifier ideology on party elite ideology and models that do the opposite are

significant with strong reduction in error. The best performing models for the House, as

measured by R-Square, are the elite Æ mass models, with the contemporaneous model performing the best, explaining just short of half the variation in mass ideology (.499). It outperforms the lagged models, suggesting that increasing polarization among party identifiers and party elites is a simultaneous event. And while Republican identifiers are strongly responsive to Republican elites, the Democrat models perform markedly better than the

Republican models. The R-Square for the elite Æ mass models for the Democrats is nearly 15%

higher than in the Republican models (.631). The coefficients for the Democratic models are

also, on average, larger than the counterpart Republican models. The model coefficient for the

contemporaneous elite Æ mass model is 2.499, while the corresponding coefficient for the

Republican elite Æ mass model is 1.290, a full point difference on the ideological scale.

The difference models examining the trend in ideological differences between partisan

identifiers and the ideological differences in Congress are particularly strong, indicating that

partisan identifiers have responded to the ideological polarization of Congress and that

congressional legislators have responded to their partisan constituents becoming less

ideologically diverse and more ideologically extreme. Note, here we are looking at the

differences between the mass identifier ideological scores and the elite legislative nominate

scores. Again, the coefficients for the elite Æ mass models are significantly larger than those for

the mass Æ elite models, though all meet typical standards of statistical significance (0.428 vs.

2.050 on the high end). This is highly significant partisan polarization along ideological lines both

381

Table 12.5: Mass Party Identifiers & Party Elite (Legislators) Simple & Autoregressive Models of Partisan Ideology 2 MODEL: DV = B0 + B1(IV) + e N Analysis Intercept Parameter Estimate Standard Error R D-W PR < DW House Rep Nom = R Ideology (m) 17 OLS -1.560 0.386 *** 0.100 0.499 House Rep Nom = R Ideology (m) 17 GLM -0.544 0.182 ** 0.085 0.246 0.544 0.001 HR Nom = R Ideology (m) lag 2Y 17 GLM -0.551 0.184 ** 0.082 0.281 0.657 0.001 HR Nom = R Ideology (m) lag 4Y 17 GLM -0.479 0.171 * 0.098 0.217 0.537 0.001 R Ideology (m) = House Rep Nom 17 OLS 4.488 1.290 *** 0.334 0.499 R Ideology (m) = House Rep Nom 17 GLM 4.482 1.305 *** 0.393 0.441 1.754 0.210 R Ideology (m) = HR Nom lag 2Y 17 GLM 4.506 1.287 *** 0.478 0.341 1.720 0.190 R Ideology (m) = HR Nom lag 4Y 17 GLM 4.481 1.437 *** 0.461 0.427 1.848 0.276

House Dem Nom = D Ideology (m) 17 OLS -1.277 0.256 *** 0.050 0.640 House Dem Nom = D Ideology (m) 17 GLM -0.994 0.180 *** 0.052 0.462 1.183 0.026 HD Nom = D Ideology (m) lag 2Y 17 GLM -1.233 0.242 *** 0.072 0.467 1.205 0.033 HD Nom = D Ideology (m) lag 4Y 17 GLM -1.046 0.190 ** 0.088 0.294 0.784 0.002 D Ideology (m) = House Dem Nom 17 OLS 4.523 2.499 *** 0.483 0.640 D Ideology (m) = House Dem Nom 17 GLM 4.525 2.506 *** 0.512 0.631 1.609 0.131 D Ideology (m) = HD Nom lag 2Y 17 GLM 4.567 2.70 *** 0.521 0.666 1.652 0.152 D Ideology (m) = HD Nom lag 4Y 17 GLM 4.586 2.832 *** 0.537 0.681 1.667 0.161

HR – DR Nom = RIDEO – DIDEO 17 OLS 0.146 0.428 *** 0.066 0.736 HR – DR Nom = RIDEO – DIDEO 17 GLM 0.286 0.316 *** 0.073 0.573 1.350 0.058 HR – DR Nom = RIDEO – DIDEO L2 17 GLM 0.277 0.340 *** 0.087 0.542 1.311 0.056 HR – DR Nom = RIDEO – DIDEO L4 17 GLM 0.330 0.317 *** 0.116 0.405 0.839 0.004 RIDEO – DIDEO = HR – DR Nom 17 OLS 0.076 1.719 *** 0.266 0.736 RIDEO – DIDEO = HR – DR Nom 17 GLM 0.057 1.748 *** 0.309 0.695 1.861 0.282 RIDEO – DIDEO = HR – DR Nom L2 17 GLM 0.035 1.843 *** 0.354 0.660 1.847 0.273 RIDEO – DIDEO = HR – DR Nom L4 17 GLM -0.053 2.050 *** 0.344 0.732 2.000 0.500 * significant at .10 level ** significant at .05 level ***significant at .01 level

382

in the mass public and among elite legislators. While Fiorina might be tempted to dismiss this as sorting, this result runs counter to Fiorina’s argument that little mass sorting has occurred and only moderate sorting has occurred at the elite level. I find strong partisan polarization at the elite and the mass level, and that the trends in mass and elite polarization over time are recursively related.

Constituent vs. Representative Ideological Differences over Time

Table 12.6 reports models regressing the squared difference between the z-standardized ideology of respondents in the states/districts of each Republican and Democrat House representative or senator for the full set of time series data. Thus the model looks at the squared difference between the standardized ideology of the constituent minus the standardized ideology of the elite legislator who represents those constituents. The model thus doesn’t assess whether constituents are further polarized relative to their representatives (or vice versa), but whether there has been a linear increase in the distance between the constituents of representatives and the representatives themselves. The constituent models are intended to assess the difference between a mass public -elite responsiveness based on party identifier positions and aggregate partisan elite ideology versus elite partisan legislators responding to changes in their geographical constituents. The overall trend in these models is of increasing distance between Republican legislators and their constituents while, at the same time, there is decreasing distance between Democratic legislators and their states and districts. While the

Republican Senate model is insignificant, the coefficient is in the positive direction, and the House

Republican model indicates significant constituent-representative polarization (0.013). Indeed, the

House model explains half of the variation in the squared differences between Republican elite representative ideology and Republican mass constituent ideology. On the Democratic side of things, there is a significant decline in the polarization of representatives and constituents in the House (-0.008) and the Senate (-0.001). The Democratic models range between 20% and 35% of the variance explained in the distance between constituents and representatives.

383

Table 12.6: Models Regressing Squared Difference b/w Normalized Constituent Ideology and Normalized Legislator Nominate Score by Year

2 MODEL: DV = B0 + B1(YEAR) + e N Intercept Parameter Estimate Standard Error R

DV: (HR ConIdeo Z– HR Nom Z)2 17 -24.74 0.013 *** 0.003 0.507

DV: (HD ConIdeo Z – HD Nom Z)2 17 14.581 -0.008 ** 0.004 0.227

DV: (SR ConIdeo Z – SR Nom Z)2 17 -8.526 0.005 0.003 0.135

DV: (SD ConIdeo Z – SD Nom Z)2 17 2.440 -0.001 ** 0.867 0.341

* significant at .10 level ** significant at .05 level ***significant at .01 level

384

Why do the House models outperform the Senate models for the Republicans? Indeed, the coefficients are larger for the House models over the Senate models for both political parties. The effect of redistricting (gerrymandering) is a likely suspect. Over the time series under consideration we have a significant change from more Democrats and fewer Republicans in the earlier periods to, as a consequence of the Republican revolution, more Republicans and fewer Democrats in the later periods for the U.S. Congress. Thus there were almost certainly more Democrats representing marginal districts and states in the early period of the time series, while more Republicans represented marginal districts and states in the later periods. More marginal districts and states would have the consequence of more Democratic-leaning constituents for Republicans in the later period and vice versa in the earlier period. This would explain the significant positive coefficients in the Republican models and the negative coefficients in Democratic models. A more nuanced constituent-based argument for the partisan and ideological polarization would require looking at only same-party identifiers among constituents. This more thorough cut of the data, along with an analysis that takes into account district characteristics, will be necessary before any definitive conclusion could be reached.

CONCLUSION

This analysis finds significant ideological polarization in the mass public, significant ideological polarization among party elites and party identifiers, a growing divide between party identifiers in the mass public, and a strong linear trend in partisan and ideological polarization over time. Furthermore, there is evidence that political elites are responsive to perturbations in mass ideological beliefs and there is strong evidence of a recursive relationship between the ideology of party identifiers and that of party elites.

385

The story Fiorina gives us where elites polarize irrespective of mass public attitudes simply doesn’t hold water. It doesn’t make sense in a world where electorates are rational and responsive, as a unilateral shift to the poles of the electoral distribution would lead to Downsian party marginalization. And indeed, we find evidence that legislators are cognizant of the fact they must face the voters and are knowledgeable of the fact that those voters have become more ideologically polarized. Rationalizations offered to explain elite non-responsiveness, such as Fiorina’s theory that elites anticipate ideological challenges from the poles in primaries, fundamentally detach the elected official from his or her interest in re-election (as opposed to standing for election, or receiving interest group benefits, or rewarding partisans, etc.). The argument does not merely suffer from theoretical problems, but also an unresponsive electorate is incongruent with apparent aggregate rationality in terms of short-term effects on elections, increasing homogenization of districts tracking with increased elite polarization, and the effectiveness of political advertising campaigns. The evidence suggests that voters are paying attention to some signals from the political environment and that they hold officials accountable in elections. And while district polarization may account for polarization in the

House, it does not explain polarization among candidates for offices with constituencies at higher levels of aggregation.

I find that partisan elite polarization is recursively related to partisan mass polarization.

While it is certainly true that elites attempt to influence the mass public through the framing and emphasis of certain issues, it is just as true that elites strive to detect those issues which have become salient to their partisan constituents and shape their position-taking accordingly.

Scholars examining the factors of polarization in American politics miss half of the story when ignoring the influence that shifts in the mass electorate have on elite behavior. Electorates change, they change in comprehensible ways, the issues and opinions of aggregate electorates

386

shift sometimes independent of elite preferences, and these shifts produce shifts in electoral behavior. As a result, the optimum ideological disposition for a Republican or Democratic party elite has changed over time. And that shift has been decidedly towards the poles and not the center.

387

APPENDIX A: States by Presidential Voting, Median Household Income, & Gross State Product

STATE BLUE/RED* MHI (median GSP (per GSP (per income) million) capita) Alabama RED $40,554 $165,796 $29,697 Alaska RED 64,333 44,517 43,748 Arizona RED 49,889 247,028 33,441 Arkansas RED 38,134 95,371 27,875 California BLUE 59,948 1,812,968 41,663 Colorado RED 55,212 236,324 41,798 Connecticut BLUE 65,967 216,266 50,332 Delaware BLUE 54,610 60,118 59,288 Florida RED 47,804 734,519 33,718 Georgia RED 49,136 396,504 35,362 Hawaii BLUE 63,746 61,532 38,083 Idaho RED 46,253 51,149 30,896 Illinois BLUE 54,124 609,570 39,514 Indiana RED 47,448 246,439 34,058 Iowa BLUE 47,448 129,026 35,662 Kansas RED 47,451 117,305 34,242 Kentucky SWING 40,267 154,184 29,842 Louisiana SWING 40,926 261,146 32,923 Maine BLUE 45,888 48,108 30,305 Maryland BLUE 68,080 268,685 39,161 Massachusetts BLUE 62,365 351,514 46,721 Michigan BLUE 47,950 381,963 33,468 Minnesota BLUE 55,082 254,970 41,295 Mississippi RED 36,338 88,546 24,062 Missouri SWING 45,114 229,470 33,468 Montana RED 43,531 34,253 27,942 Nebraska RED 47,085 80,093 36,441 Nevada SWING 55,062 127,213 39,813 New Hampshire BLUE 62,369 57,341 37,666 New Jersey BLUE 67,035 465,484 44,885 New Mexico BLUE 41,452 76,178 31,986 New York BLUE 53,514 1,103,024 46,617 North Carolina RED 44,670 399,446 36,489 North Dakota RED 43,753 26,385 34,446 Ohio SWING 46,597 466,309 34,609 Oklahoma RED 41,567 139,323 29,545 Oregon BLUE 48,730 158,223 37,633 Pennsylvania BLUE 48,576 531,110 34,828 Rhode Island BLUE 53,514 46,900 36,292 South Carolina RED 43,329 152,830 29,642 South Dakota RED 43,424 33,934 35,842 Tennessee RED 42,367 243,869 34,321 Texas RED 47,548 1,141,965 36,920 Utah RED 55,109 105,658 32,357 Vermont BLUE 49,907 24,543 34,472 Virginia RED 59,562 382,964 41,702 Washington BLUE 55,591 311,270 39,616 West Virginia SWING 37,060 57,711 24,748 Wisconsin BLUE 50,578 232,293 35,390 Wyoming RED 51,731 31,514 39,130 *Blue / Red = 3 out of 4 elections (1992,1996,2000,2004) for either party. Swing = 2 out of 4. *MHI reported in Table 1.1 is for 2007. * GSP reported in Table 1.1 is calculated over a three year period 2004-2006.

388

APPENDIX B: GRD - Frequency Table of Polls on Gays and Gay Rights 1971-2007

CUM CUM POLLS & SURVEYS FREQ PER FREQ PER 1996 Survey of American Political Culture 4 0.58 4 0.58 2005 National Hispanic Survey 1 0.15 5 0.73 ABC News / Facebook Poll 1 0.15 6 0.87 ABC News / Washington Post Poll 26 3.77 32 4.64 ABC News Poll 11 1.6 43 6.24 Active Center Holds Survey 1 0.15 44 6.39 Adoption Survey 1 0.15 45 6.53 America's Evangelicals 1 0.15 46 6.68 American Public Opinion About Privacy At Home and At Work 10 1.45 56 8.13 American Values in the 80's 1 0.15 57 8.27 Associated Press / IPSOS-Public Affairs Poll 1 0.15 58 8.42 Associated Press / Media General Poll 1 0.15 59 8.56 Associated Press Poll 8 1.16 67 9.72 Attitudes Toward Smoking and the Tobacco Industry Survey 1 0.15 68 9.87 Barna Report 1993-1994 Absolute Confusion 2 0.29 70 10.16 Barna Report 1994-1995 Virtual America 1 0.15 71 10.3 CBS / New York Times Poll 3 0.44 74 10.74 CBS News / NY Times Poll 39 5.66 113 16.4 CBS News Exit Poll 1 0.15 114 16.55 CBS News Poll 8 1.16 122 17.71 CNN / Opinion Research Corporation Poll 4 0.58 126 18.29 Consumers in the Information Age 1 0.15 127 18.43 Defense of Marriage Act Poll 3 0.44 130 18.87 Democracy Corps Survey 39 5.66 169 24.53 Evangelical Christianity in the U.S. 1 0.15 170 24.67 Fair Juror Survey 1 0.15 171 24.82 Family Circle Ethics Poll 1 0.15 172 24.96 Family Research Council Survey 1 0.15 173 25.11 For Goodness Sake Survey 1 0.15 174 25.25 Fox News / Opinion Dynamics Poll 12 1.74 186 27 Free Expression and the American Public 2 0.29 188 27.29 GSS 1 0.15 189 27.43 Gallup / CNN / USA Today Poll 38 5.52 227 32.95 Gallup / CNN Poll 1 0.15 228 33.09 Gallup / Newsweek Poll 12 1.74 240 34.83 Gallup / USA Today Poll 3 0.44 243 35.27 Gallup Poll 62 9 305 44.27 Gallup Report 8 1.16 313 45.43 Gallup/PDK Poll of Public Attitudes Toward the Public Schools 3 0.44 316 45.86

389

CUM CUM POLLS & SURVEYS FREQ PER FREQ PER Garth Analysis Survey 1 0.15 317 46.01 General Social Survey 2 0.29 319 46.3 Gordon Black / USA Today Poll 1 0.15 320 46.44 Great American TV Poll 5 0.73 325 47.17 Harris Poll 4 0.58 329 47.75 Harris Survey 14 2.03 343 49.78 If Women Ran America 2 0.29 345 50.07 Judical Confirmation Survey 2 0.29 347 50.36 Kaiser Family Foundation Survey on Americans and AIDS/HIV 1 0.15 348 50.51 Los Angeles Times Poll 53 7.69 401 58.2 NBC News / Wall Street Journal Poll 29 4.21 430 62.41 NPR Poll 3 0.44 433 62.84 National Family Values 2 0.29 435 63.13 New Democratic Electorate Survey 3 0.44 438 63.57 New Models National Brand Poll 3 0.44 441 64.01 PSRA / Newsweek Poll 74 10.74 515 74.75 Parents Magazine Poll 1 0.15 516 74.89 Peole, The Press & Politics Poll - New Political Landscape 1 0.15 517 75.04 People & The Press -- Mood of America Survey 1 0.15 518 75.18 People, The Press & Politics Poll 2 0.29 520 75.47 Pew Internet & American Life Project Poll 2 0.29 522 75.76 Pew News Interest Index / Believability Poll 1 0.15 523 75.91 Pew News Interest Index Poll 7 1.02 530 76.92 Pew News Interest Index Poll / Homosexuality Poll 6 0.87 536 77.79 Pew Research Center Political Survey 1 0.15 537 77.94 Pew Research Center for TP&TP Political Typology Poll 1 0.15 538 78.08 Pew Research Center for TP&TP State of the Union Poll 1 0.15 539 78.23 Pew Research Center for TP&TP Typology Poll 1 0.15 540 78.37 Pew Social Trends Poll 1 0.15 541 78.52 Quinnipac University Poll 3 0.44 544 78.96 Reader's Digest Poll 1 0.15 545 79.1 Religion and Public Life 1 0.15 546 79.25 Roper / Ladies' Home Journal Poll 1 0.15 547 79.39 Roper / U.S. News & World Report Poll 1 0.15 548 79.54 Roper Commerical Survey 1 0.15 549 79.68 Roper Report 77-7 2 0.29 551 79.97 Roper Report 85-7 1 0.15 552 80.12 Roper Report 86-10 1 0.15 553 80.26 Roper Report 87-2 2 0.29 555 80.55 Roper Report 87-7 1 0.15 556 80.7 Roper/U.S. News & World Report Poll 2 0.29 558 80.99

390

CUM CUM POLLS & SURVEYS FREQ PER FREQ PER TIPP / Investor's Business Daily / Christian Science Monitor 4 0.58 562 81.57 Poll TV Poll 1 0.15 563 81.71 Taking America's Pulse III - Intergroup Relations Survey 3 0.44 566 82.15 Tarrance Group Poll 2 0.29 568 82.44 The Civic and Political Health of the Nation Survey 1 0.15 569 82.58 Time / CNN / Harris Interactive Poll 8 1.16 577 83.74 Time / CNN / Yankelovich, Clancy & Shulman Poll 40 5.81 617 89.55 Time / SRBI Poll 5 0.73 622 90.28 Time / Yankelovich, Skelly & White Poll 4 0.58 626 90.86 Times Mirror News Interest Index 1 0.15 627 91 U.S. News & World Report / Bozell Worldwide Poll 1 0.15 628 91.15 U.S. News & World Report Poll 15 2.18 643 93.32 Views on Issues and Policies Related to Sexual Orientation 25 3.63 668 96.95 Survey Virginia Slims American Women's Poll 3 0.44 671 97.39 Voice of Mom Survey 2 0.29 673 97.68 Voter Attitudes on Political Campaigns Survey 1 0.15 674 97.82 Washington Post / Harvard / KFF American Values Survey 1 0.15 675 97.97 Washington Post / Harvard / KFF Race Relations Poll 2 0.29 677 98.26 Washington Post / KFF / Harvard Americans on Values 4 0.58 681 98.84 Washington Post / KFF / Harvard Political Independents 1 0.15 682 98.98 Survey Washington Post Poll 5 0.73 687 99.71 What Americans Expect from the Public Schools Survey 1 0.15 688 99.85 Women on Their Own in Unmarried America Survey 1 0.15 689 100

391

APPENDIX C: GRD – Questions by Year

YEAR FREQUENCY PERCENT CUM FREQ CUM PER 1971 1 0.15 1 0.15 1974 1 0.15 2 0.29 1977 21 3.05 23 3.34 1978 6 0.87 29 4.21 1979 1 0.15 30 4.35 1980 2 0.29 32 4.64 1981 1 0.15 33 4.79 1982 10 1.45 43 6.24 1983 8 1.16 51 7.4 1985 23 3.34 74 10.74 1986 10 1.45 84 12.19 1987 20 2.9 104 15.09 1988 6 0.87 110 15.97 1989 16 2.32 126 18.29 1990 5 0.73 131 19.01 1991 13 1.89 144 20.9 1992 29 4.21 173 25.11 1993 60 8.71 233 33.82 1994 33 4.79 266 38.61 1995 7 1.02 273 39.62 1996 38 5.52 311 45.14 1997 22 3.19 333 48.33 1998 44 6.39 377 54.72 1999 16 2.32 393 57.04 2000 54 7.84 447 64.88 2001 9 1.31 456 66.18 2002 15 2.18 471 68.36 2003 50 7.26 521 75.62 2004 76 11.03 597 86.65 2005 40 5.81 637 92.45 2006 30 4.35 667 96.81 2007 22 3.19 689 100

392

APPENDIX D: GRD Issue Category Descriptions

ISSUE TYPE ISSUE CATEGORY DESCRIPTIONS GAY RIGHTS Attitudes towards homosexuality, gay rights, homosexual relations, bars, etc. OFFSPRING Attitudes on homosexual contact / interaction with own children and gay progeny CHRISTIAN Whether homosexuals can be good Christians JOBS Whether homosexuals should have equal right to job opportunities, specific jobs LEGALITY Whether homosexual sexual activity should be legal CANDIDATE Is the homosexuality of a candidate important; is it information the public should know or a private matter COMFORT How comfortable the respondent feels with and around homosexuals MORALITY Does the respondent think that homosexuality is immoral, moral, or not a moral issue TEACH Any questions that involves teaching by homosexuals, schools with homosexuals, and school officials who are homosexuals. BORN / IMPORT If respondent thinks that homosexuals are born that way MILITARY Whether the respondent thinks gays should be allowed to serve openly in the military and the Don’t Ask, Don’t Tell policy AIDS Attitudes towards homosexuals in light of the AIDS epidemic FRIENDS Whether the respondent has friends, acquaintances, relatives or associates who are gay VOTE Would the respondent vote for a candidate they knew to be homosexual or a party that supported gay rights MOVEMENT Attitudes towards the gay rights movement, gay rights organizations PRIVACY Is homosexuality or homosexual relations a private matter GAY MARRIAGE Does the respondent support the legalization of gay marriage ADOPTION Whether the respondent supports or opposes gay adoptions CHURCH Attitudes towards gay religious leaders and whether the respondent would remain in a church with an openly gay minister T.V. Do gays receive too much attention on television, should Ellen’s character be openly gay, are sympathetic portrayals of gays on television a problem SPEECH Should gays be permitted to give a speech BENEFITS Should gays be allowed to receive job benefits, social security benefits, and other benefits that normally accrue to heterosexual couples HOMOSEXUALS Attitudes towards homosexuals as a group / community HOUSING Should discrimination against gays in housing be permitted INHERITANCE Should gays enjoy inheritance rights should their partner die

393

ISSUE TYPE ISSUE CATEGORY DESCRIPTIONS PARTIES Does respondent support / associated with one of the parties due to their gay rights position CIVIL UNIONS Whether gays should be able to form legal unions that convey marriage-like benefits but are not legally defined as marriage DOCTOR Should gays be permitted to serve as doctors READ Should kids read a book that portrays homosexuals sympathetically DENTIST Should gays be permitted to serve as doctors SHOP Would respondent be comfortable with shopping at an establishment that employed homosexuals PARENTS Can gays be just good parents, as good as heterosexuals LIFESTYLE How acceptable is the gay lifestyle to respondent JUROR Should gays be permitted to serve as jurors BOY SCOUTS Does respondent believe the Boy Scouts should be permitted to exclude gay boy scouts / gay scoutmasters, does respondent agree with USSC decision allowing the Boy Scouts to exclude gays AMENDMENT Should the respondent’s state amend their constitution to define marriage as between a man and a woman CONSTITUTION Should the U.S. Constitution be amended to define marriage as between a man and a woman

394

APPENDIX E: Chapter 4 Graphed Poll Questions

Virginia Slims American Women's Poll* Question on Acceptability of Daughters Homosexual Relationship, April 1974

Question: (Now, here is a list of things some young people find acceptable today. Would you read down that list, and for each one, tell me for a daughter of yours who had just finished her schooling whether you would find it acceptable, or accept it but be unhappy about it, or not accept it and have the relationship very much strained as a result?) Having a homosexual relationship.

CATEGORY PERCENT Find it Acceptable 1 Accept, but Unhappy 19 Not Acceptable 75 * Virginia Slims American Women's Poll. Conducted by Roper. April, 1974. Sample N: 3880 respondents.

Harris Survey* Question on Jobs & Homosexuals, June 1977

Question: If qualified in every other way, do you feel a person who admits to being a homosexual should be allowed to hold down a job as... An [insert profession]...or not?

PROFESSION SHOULD BE ALLOWED SHOULD NOT BE ALLOWED Artist 86 7 Factory worker 85 7 Clerk 80 12 TV News Commentator 72 19 Company President 67 22 Congressman 53 37 Doctor 48 40 Social Worker 43 48 School Principal 33 58 Counselor 27 63 School Teacher 34 55 Minister / Priest / Rabbi 40 50 Psychiatrist 40 48 Policeman 48 42 * Harris Survey. Conducted by Roper Organization. June, 1977. Sample N: 1947 respondents.

395

Newsweek / PSRA Poll* Question on Gay Marriage Constitutional Amendment, January 2004

Question: (Please tell me how much you approve or disapprove of each of the following Bush administration policies and proposals.) What about...passing a Constitutional amendment, if necessary, to ban gay marriage in all states? Do you strongly approve, somewhat approve, somewhat disapprove, or strongly disapprove of this Bush administration policy or proposal?

CATEGORIES PERCENT strongly approve 36 somewhat approve 10 somewhat disapprove 13 strongly disapprove 33 * PSRA / Newsweek Poll. Conducted by Princeton Survey Research Associates International. January, 2004. Sample N: 1233 respondents.

ABC News / Washington Post Poll * Question on Legality of Gay Marriage, March 2004

Question: Do you think it should be legal or illegal for homosexual couples to get married? (If Legal/Illegal, ask:) (Is that strongly or somewhat)?

CATEGORIES PERCENT legal strongly 24 legal somewhat 14 illegal somewhat 11 illegal strongly 48 * Poll conducted by ABC News / Washington Post. March, 2004. Sample N: 1202 respondents.

Los Angeles Times Poll* Question on Approval of Gay Adoption, March 2004

Question: Do you favor or oppose gay couples legally adopting children? (If Favor/Oppose, ask:) Do you strongly favor/oppose gay couples adopting children or only somewhat favor/oppose gay couples adopting children?

CATEGORIES PERCENT favor strongly 21 favor somewhat 19 oppose somewhat 11 oppose strongly 41 * Poll conducted by Los Angeles Times. March, 2004. Sample N: 1616 respondents.

396

Los Angeles Times Poll* Question on Approval of Gay Rights, March 2004

Question: Generally speaking, do you approve or disapprove of homosexual or gay rights -- or haven't you heard enough about that yet to say? Do you approve/disapprove strongly or approve/disapprove somewhat?

CATEGORIES PERCENT favor strongly 27 favor somewhat 16 oppose somewhat 6 oppose strongly 36 * Poll conducted by Los Angeles Times. March, 2004. Sample N: 1616 respondents.

397

APPENDIX F: GSS TABLE – Public Opinion on Homosexual Relations

QUESTION: What about sexual relations between two adults of the same sex̂̂do you think it is always wrong, almost always wrong, wrong only sometimes, or not wrong at all?

YEAR Nearly Sometimes Not Wrong N Always Wrong Always Wrong Wrong At All AW - NWAA 1973 72.7 6.6 7.6 11 61.7 1,448 1974 70.5 5 7.9 13 57.5 1,412 1975 - - - - - 0 1976 70.1 6.2 7.9 15.9 54.2 1,426 1977 71.9 5.8 7.5 14.9 57 1,453 1978 - - - - - 0 1980 73.3 6 6.1 14.6 58.7 1,397 1982 74.8 5.3 6.5 13.4 61.4 1,771 1983 - - - - - 0 1984 73.3 5 7.4 14.3 59 1,412 1985 75.3 4 7 13.7 61.6 1,484 1986 - - - - - 0 1987 78.2 4.1 5.8 11.9 66.3 1,750 1988 76.8 4.7 5.7 12.8 64 937 1989 74.2 4.1 6 15.7 58.5 980 1990 76.3 4.8 6.1 12.8 63.5 872 1991 75.5 4.1 4.4 16 59.5 926 1993 66.3 4.3 7.3 22 44.3 1,012 1994 66.5 4 6.2 23.3 43.2 1,884 1996 60.4 5.2 6.2 28.2 32.2 1,784 1998 58 5.7 6.9 29.4 28.6 1,753 2000 58.8 4.5 8 28.8 30 1,697 2002 55 4.9 7.1 33 22 884 2004 57.6 4.7 6.9 30.8 26.8 868 2006 55.1 5 7.1 32.7 22.3 1,908 Total 68.2 5 6.8 19.7 43.2 29,058

398

APPENDIX G: ANES CUMULATIVE FILE VARIABLES INCLUDED IN ANALYSIS 1964-2004

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0203 Thermometer: Protestants X X X X X X X VCF0204 Thermometer: Catholics X X X X X X X X X X X VCF0205 Thermometer: Jews X X X X X X X X X X VCF0234 Thermometer: Christian Fundamentalists X X X X X X X VCF9003 Thermometer: Evangelical Groups X X X VCF0206 Thermometer: Blacks X X X X X X X X X X X X X X X X X X X X VCF0207 Thermometer: Whites X X X X X X X X X X X X X X X X X X VCF0217 Thermometer: Chicanos/Hispanic X X X X X X X X X VCF0233 Thermometer: Illegal Aliens X X X X VCF0208 Thermometer: Southerners X X X X X X X VCF0209 Thermometer: Big Business X X X X X X X X X X X X X X X X VCF0223 Thermometer: Poor People X X X X X X X X X X X X X X X VCF0220 Thermometer: People on Welfare X X X X X X X X X X X X VCF0210 Thermometer: Labor Unions X X X X X X X X X X X X X X X X X X VCF0232 Thermometer: Gays and Lesbians X X X X X X X X X VCF0253 Thermometer: Feminists X X X X X VCF0229 Thermometer: Environmentalists X X X X X X X X X VCF0213 Thermometer: Military X X X X X X X X X X X X X X VCF0211 Thermometer: Liberals X X X X X X X X X X X X X X X X X X X X VCF02X2 Thermometer: Conservatives X X X X X X X X X X X X X X X X X X X X VCF0231 Thermometer: Federal Government X X X X X X X VCF0201 Thermometer: Democrats X X X X X X X X X VCF0202 Thermometer: Republicans X X X X X X X X X VCF02X8 Thermometer: Democratic Party X X X X X X X X X X X X X

399

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0224 Thermometer: Republican Party X X X X X X X X X X X X X VCF0290 Thermometer: Both Major Parties X X X X X X X X X X X X X VCF0291 Thermometer Index: Major Party X X X X X X X X X X X X X VCF0413 Thermometer: Index Major Party Presidential Candidate X X X X X X X X X X VCF0412 Thermometer: Both Major Party Presidential Candidates X X X X X X X X X X VCF0424 Thermometer: Democratic Presidential Candidate X X X X X X X X X X VCF0425 Thermometer: Democratic Vice-presidential Candidate X X X X X X X X X X VCF0426 Thermometer: Republican Presidential Candidate X X X X X X X X X X VCF0427 Thermometer: Republican Vice-presidential Candidate X X X X X X X X X X VCF0906 Thermometer: Democratic House Candidate X X X X X X X X X X X X X X VCF0907 Thermometer: Republican House Candidate X X X X X X X X X X X X X X VCF1017 Thermometer: Both Major Party House Candidates X X X X X X X X X X X X X X VCF1018 Thermometer Index: Major Party House Candidates X X X X X X X X X X X X X X VCF9056 Thermometer: Senate Democratic Candidate X X X X X X X X X X X X VCF9057 Thermometer: Senate Republican Candidate X X X X X X X X X X X X VCF0503 Democratic Party Placement: Liberal-Conservative Scale X X X X X X X X X X X X X X X X VCF0504 Republican Party Placement: Liberal-Conservative Scale X X X X X X X X X X X X X X X X VCF9101 Democratic House Candidate: Liberal-Conservative Scale X X X X X X X X X X VCF9106 Republican House Candidate: Liberal-Conservative Scale X X X X X X X X X X VCF9088 Democratic Presidential Cand: Liberal-Conservative Scale X X X X X X X X VCF9096 Republican Presidential Cand: Liberal-Conservative Scale X X X X X X X X X VCF0801 Thermometer Index: Liberal-Conservative X X X X X X X X X X X X X X X X X X X X VCF0803 R Placement: Liberal-Conservative Scale X X X X X X X X X X X X X X X X X VCF0804 R Placement: Collapsed Liberal-Conservative Scale X X X X X X X X X X X X X X X X X VCF0849 Summary: Liberal-Conservative Position of R X X X X X X X X X X X VCF0508 Democratic Party Placement: Govt Health Insurance Scale X X X X X X

400

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0509 Republican Party Placement: Govt Health Insurance Scale X X X X X X VCF9093 Republican Presidential Cand: Govt Health Insurance Scale X X X X VCF9085 Democratic Presidential Cand: Govt Health Insurance Scale X X X X VCF0806 R Placement: Government Health Insurance Scale X X X X X X X X X X X VCF0513 Democratic Party Placement: Guaranteed Jobs Scale X X X X X X X X X X X VCF0514 Republican Party Placement: Guaranteed Jobs Scale X X X X X X X X X X X VCF9100 Democratic House Cand.: Guaranteed Jobs and Living Scale X X X X X X VCF9105 Republican House Cand.: Guaranteed Jobs and Living Scale X X X X X X VCF9087 Democratic Presidential Cand: Guaranteed Jobs-Living Scale X X X X X X X X X VCF9095 Republican Presidential Cand: Guaranteed Jobs-Living Scale X X X X X X X X X VCF0808 R Opinion: Guaranteed Jobs and Income X X X VCF0809 R Placement: Guaranteed Jobs and Income Scale X X X X X X X X X X X X X X X X VCF0517 Democratic Party Placement: Aid to Blacks Scale X X X X X X X X X X X X X VCF0518 Republican Party Placement: Aid to Blacks Scale X X X X X X X X X X X X X VCF9098 Democratic House Candidate: Aid to Blacks Scale X X X VCF9103 Republican House Candidate: Aid to Blacks Scale X X X VCF9084 Democratic Presidential Cand: Aid to Blacks Scale X X X X X X X X VCF9092 Republican Presidential Cand: Aid to Blacks Scale X X X X X X X X VCF0830 R Placement: Aid to Blacks Scale X X X X X X X X X X X X X X X X X VCF0537 Democratic Party Placement: Women Equal Role Scale X X X X X X X VCF0538 Republican Party Placement: Women Equal Role Scale X X X X X X X VCF9102 Republican House Candidate: Women's Equal Role Scale X X X X VCF9097 Democratic House Candidate: Women's Equal Role Scale X X X X VCF9083 Democratic Presidential Cand: Women's Equal Role Scale X X X X X X X VCF9091 Republican Presidential Cand: Women's Equal Role Scale X X X X X X X VCF0834 R Placement: Women Equal Role Scale X X X X X X X X X X X X X X X

401

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0541 Democratic Party Placement: Govt Services/Spending Scale X X X X X X X X X X X VCF0542 Republican Party Placement: Govt Services/Spending Scale X X X X X X X X X X X VCF9099 Democratic House Candidate: Govt Services/Spending Scale X X X X X X X X X VCF9104 Republican House Candidate: Govt Services/Spending Scale X X X X X X X X X VCF9086 Democratic Presidential Cand: Govt Services/Spending Scale X X X X X X VCF9094 Republican Presidential Cand: Govt Services/Spending Scale X X X X X X VCF0839 R Placement: Government Services/Spending Scale X X X X X X X X X X X VCF0549 Democratic Party Placement: Defense Spending Scale X X X X X X X X X X VCF0550 Republican Party Placement: Defense Spending Scale X X X X X X X X X X VCF9081 Democratic Presidential Cand: Defense Spending Scale X X X X X X X VCF9089 Republican Presidential Cand: Defense Spending Scale X X X X X X X VCF0843 R Placement: Defense Spending Scale X X X X X X X X X X X VCF9082 Democratic Presidential Cand: Cooperation with USSR Scale X X X VCF9090 Republican Presidential Cand: Cooperation with USSR Scale X X X VCF0709 Ticket-splitting: President/House X X X X X X X X X X VCF0710 Ticket-splitting: President/Senate X X X X X X X X X X X VCF0711 Does Always Vote for the Same Party X X X X X X X X VCF0717 Did R Try to Influence the Vote Others During the Campaign X X X X X X X X X X X X X X X X X X X X X VCF0718 Did R Attend Political Meetings/Rallies During the Campaign X X X X X X X X X X X X X X X X X X X X VCF0719 Did R Work for Party or Candidate During the Campaign X X X X X X X X X X X X X X X X X X X X VCF0720 Did R Display Candidate Button/Sticker During the Campaign X X X X X X X X X X X X X X X X X X X X VCF0721 Did R Donate Money to Party or Cand. During the Campaign X X X X X X X X X X X X X X X X X X X X VCF0722 Has R Ever Written a Letter to a Public Official X X X X VCF0723 Campaign Participation Count [1 of 2] X X X X X X X X X X X X X X X X X X VCF0723a Campaign Participation Count [1 of 2] X X X X X X X X X X X X X X X X X X X X VCF0740 Did R Contribute to Political Party During the Campaigns X X X X X X X X X X X X X

402

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0741 Did R Contribute to PAC During the Campaigns X X X X X X X X X X X VCF0742 Did R Give Money to Individual Cand. During the Campaigns X X X X X X X X X X X X X VCF0743 Does R Belong to Political Organization or Club X X X X X X VCF0728 Media Exposure Count X X X X X X X X X X VCF0724 Did R Watch TV Programs about the Election Campaigns X X X X X X X X X X X X X X X X VCF0725 How Many Programs about Campaigns on the Radio [1 of 2] X X X X X X X X X X X X X X VCF0726 How Many Articles about Election Campaigns in Magazines X X X X X X X X X X X X VCF0727 How Many Articles about Election Campaigns in Newspapers X X X X X X X X X X X X X X X X VCF0735 Candidate Code of R Vote: House Candidate X X X X X X X X X X X X X X VCF0736 Party of R Vote: House Candidate X X X X X X X X X X X X X X X X X X X X X VCF0746 Did Any Religious/Moral Group Try to Influence R's Vote X X X VCF0747 Did R Obtain Info about Cands/Parties/Issues at Church X X X VCF0845 R Opinion: Authority of the Bible [1 of 2] X X X X X X X VCF0850 R Opinion: Authority of the Bible [2] X X X X X X X VCF0846 Is Religion Important to R X X X X X X X X X X X X VCF0847 How Much Guidance Does R Have from Religion X X X X X X X X X X X X VCF0851 R Opinion: Newer Lifestyles Contribute to Society Breakdown X X X X X X X X X VCF0852 R Opinion: Should Adjust View of Moral Behavior to Changes X X X X X X X X X VCF0853 R Opinion: Should be More Emphasis on Traditional Values X X X X X X X X X VCF0854 R Opinion: Tolerance of Different Moral Standards X X X X X X X X X VCF0837 R Opinion: When Should Abortion Be Allowed X X X X VCF0838 R Opinion: By Law, When Should Abortion Be Allowed X X X X X X X X X X X X VCF9043 R Opinion: When Should School Prayer Be Allowed [1 of 2] X X X X X X X VCF9051 R Opinion: When Should School Prayer Be Allowed [2 of 2] VCF9043a Strength of Opinion: When School Prayer Should be Allowed X X X X X X X VCF0876 R Opinion: Law to Protect Homosexuals Against Discrimination X X X X X

403

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0876a R Opinion Strength: Law Against Homosexual Discrimination X X X X X VCF0877 R Opinion: Favor/Oppose Gays in the Military X X X X VCF0877a R Opinion Strength: Favor/Oppose Gays in Military X X X X VCF0878 R Opinion: Should Gays/Lesbians Be Able to Adopt Children X X X VCF0867 R Opinion: Affirmative Action in Hiring/Promotion [1 of 2] X X X X X X X X VCF0867a R Opinion: Affirmative Action in Hiring/Promotion [2 of 2] X X X X X X X X VCF0825 R Opinion: How Likely for U.S. to be at War/in Bigger War X X X X VCF0826 R Opinion: Did U.S. Do Right Thing Getting Involved in War X X X X X VCF0827 R Opinion: How Should U.S. Proceed in Current War X X X X VCF0848 How Concerned is R about Conventional War X X X X X X X X X VCF0828 R Opinion: Should Government Cut Military Spending X X VCF0844 R Opinion: How Willing Should U.S. Be to Use Military Force X X X VCF0875 R Opinion: Mention X- Most Important National Problem X X X X X X X X X X X X X X X X X X X VCF0875a R Opinion: Mention 2- Most Important National Problem X X X X X VCF0875b R Opinion: Mention 3- Most Important National Problem X X X X X X X X X X X X X X VCF0314 Likes Number of Mentions: Democratic Party X X X X X X X X X X X X X X X X VCF0315 Dislikes Number of Mentions: Democratic Party X X X X X X X X X X X X X X X X VCF0316 Affect (Likes-Dislikes): Democratic Party X X X X X X X X X X X X X X X X VCF0317 Salience (Likes-Dislikes): Democratic Party X X X X X X X X X X X X X X X X VCF0318 Likes Number of Mentions: Republican Party X X X X X X X X X X X X X X X X VCF0319 Disikes Number of Mentions: Republican Party X X X X X X X X X X X X X X X X VCF0320 Affect (Likes-Dislikes): Republican Party X X X X X X X X X X X X X X X X VCF0321 Salience (Likes-Dislikes: Republican Party) X X X X X X X X X X X X X X X X VCF0322 Affect (Likes-Dislikes): Major Parties X X X X X X X X X X X X X X X X VCF0323 Salience (Likes-Dislikes): Major Parties X X X X X X X X X X X X X X X X VCF0324 Relative Salience (Likes-Dislikes): Major Parties X X X X X X X X X X X X X X X X

404

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0374 R likes anything about Democratic party X X X X X X X X X X X X X X VCF0375a #1 Full Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0375b #1 Collapsed Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0375c #1 Full Like - Democratic Party 1952-1968 X X VCF0375d #1 Collapsed Like - Democratic Party 1952-1968 X X VCF0376a #2 Full Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0376b #2 Collapsed Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0376c #2 Full Like - Democratic Party 1952-1968 X X VCF0376d #2 Collapsed Like - Democratic Party 1952-1968 X X VCF0377a #3 Full Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0377b #3 Collapsed Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0377c #3 Full Like - Democratic Party 1952-1968 X X VCF0377d #3 Collapsed Like - Democratic Party 1952-1968 X X VCF0378a #4 Full Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0378b #4 Collapsed Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0378c #4 Full Like - Democratic Party 1952-1968 X X VCF0378d #4 Collapsed Like - Democratic Party 1952-1968 X X VCF0379a #5 Full Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0379b #5 Collapsed Like - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0379c #5 Full Like - Democratic Party 1952-1968 X X VCF0379d #5 Collapsed Like - Democratic Party 1952-1968 X X VCF0380 R dislikes anything about Democratic party X X X X X X X X X X X X X X VCF0381a #1 Full Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0381b #1 Collapsed Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0381c #1 Full Dislike - Democratic Party 1952-1968 X X VCF0381d #1 Collapsed Dislike - Democratic Party 1952-1968 X X X X X X X X X X X X X X X X

405

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0382a #2 Full Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0382b #2 Collapsed Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0382c #2 Full Dislike - Democratic Party 1952-1968 X X VCF0382d #2 Collapsed Dislike - Democratic Party 1952-1968 X X VCF0383a #3 Full Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0383b #3 Collapsed Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0383c #3 Full Dislike - Democratic Party 1952-1968 X X VCF0383d #3 Collapsed Dislike - Democratic Party 1952-1968 X X VCF0384a #4 Full Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0384b #4 Collapsed Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0384c #4 Full Dislike - Democratic Party 1952-1968 X X VCF0384d #4 Collapsed Dislike - Democratic Party 1952-1968 X X VCF0385a #5 Full Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0385b #5 Collapsed Dislike - Democratic Party 1972 and later X X X X X X X X X X X X X X VCF0385c #5 Full Dislike - Democratic Party 1952-1968 X X VCF0385d #5 Collapsed Dislike - Democratic Party 1952-1968 X X VCF0386 R likes anything about Republican party X X X X X X X X X X X X X X VCF0387a #1 Full Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0387b #1 Collapsed Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0387c #1 Full Like - Republican Party 1952-1968 X X VCF0387d #1 Collapsed Like - Republican Party 1952-1968 X X VCF0388a #2 Full Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0388b #2 Collapsed Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0388c #2 Full Like - Republican Party 1952-1968 X X VCF0388d #2 Collapsed Like - Republican Party 1952-1968 X X VCF0389a #3 Full Like - Republican Party 1972 and later X X X X X X X X X X X X X X

406

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0389b #3 Collapsed Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0389c #3 Full Like - Republican Party 1952-1968 X X VCF0389d #3 Collapsed Like - Republican Party 1952-1968 X X VCF0390a #4 Full Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0390b #4 Collapsed Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0390c #4 Full Like - Republican Party 1952-1968 X X VCF0390d #4 Collapsed Like - Republican Party 1952-1968 X X VCF0391a #5 Full Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0391b #5 Collapsed Like - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0391c #5 Full Like - Republican Party 1952-1968 X X VCF0391d #5 Collapsed Like - Republican Party 1952-1968 X X VCF0392 R dislikes anything about Republican party X X X X X X X X X X X X X X VCF0393a #1 Full Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0393b #1 Collapsed Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0393c #1 Full Dislike - Republican Party 1952-1968 X X VCF0393d #1 Collapsed Dislike - Republican Party 1952-1968 X X VCF0394a #2 Full Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0394b #2 Collapsed Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0394c #2 Full Dislike - Republican Party 1952-1968 X X VCF0394d #2 Collapsed Dislike - Republican Party 1952-1968 X X VCF0395a #3 Full Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0395b #3 Collapsed Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0395c #3 Full Dislike - Republican Party 1952-1968 X X VCF0395d #3 Collapsed Dislike - Republican Party 1952-1968 X X VCF0396a #4 Full Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0396b #4 Collapsed Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X

407

6 6 6 7 7 7 7 7 8 8 8 8 8 9 9 9 9 9 0 0 0 VAR VARIABLE LABEL 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 6 8 0 2 4 VCF0396c #4 Full Dislike - Republican Party 1952-1968 X X VCF0396d #4 Collapsed Dislike - Republican Party 1952-1968 X X VCF0397a #5 Full Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0397b #5 Collapsed Dislike - Republican Party 1972 and later X X X X X X X X X X X X X X VCF0397c #5 Full Dislike - Republican Party 1952-1968 X X VCF0397d #5 Collapsed Dislike - Republican Party 1952-1968 X X VCF0704 Party of R Vote: President- Major Candidates X X X X X X X X X X X VCF0704a Party of R Vote: President- 2 Major Parties X X X X X X X X X X X VCF0707 Party of R Vote: Congressman X X X X X X X X X X X X X X X X X X X X X

408

Appendix H: Collapsed Categories for Open-Ended Responses

QUESTION: What do you think are the most important problems facing this country?

VALUES CATEGORIES 1 AGRICULTURAL ECONOMICS; BUSINESS; CONSUMER ISSUES (includes foreign investment, tariffs/protection of U.S. industries, international trade deficit/balance of payments, immigration, 2 interstate commerce/transportation; does not include unemployment [09], defense spending [03], foreign [03] or government spending on domestic social welfare [09])

FOREIGN AFFAIRS AND NATIONAL DEFENSE (includes: foreign aid, defense spending, the 3 space program; does not include: international trade deficit [02])

GOVERNMENT FUNCTIONING 4 (not "the economy" [02])

LABOR ISSUES 5 (not unemployment [09])

6 NATURAL RESOURCES PUBLIC ORDER (includes: crime, drugs, civil liberties and non 7 racial civil rights, women's rights, abortion rights, gun control, family / social / religious / moral 'decay,' church and state, etc.)

RACIAL PROBLEMS (note: this primarily includes civil rights issues and racial equality; monetary assistance to 8 minorities is primarily found in code 9, however there is a slight overlap: see Note 7 for specific codes; note especially 1988 code 300 and 1966- 1972 codes 61-63)

SOCIAL WELFARE (includes: population, child care, aid to 9 education, the elderly, health care, housing, poverty, unemployment, 'welfare' etc.)

97 Other problems (incl. specific campaign issues)

409

QUESTION: Is there anything in particular that you like about the Republican [or Democrat] party [or candidate]?

VALUES CATEGORIES 11 People within party (0001-0097)

12 Party characteristics (0101-0197)

21 Candidate experience, ability (0201-0297)

22 Candidate leadership qualities (0301-0397)

23 Candidate personal qualities (0401-0498)

24 Candidate party connections (0500-0597)

31 Government management (0601-0697)

32 Government activity/philosophy (0801-0897)

33 Domestic policies (0901-1097)

34 Foreign policies (1101-1199,1300-1310)

35 Group connections (1201-1299)

40 Miscellaneous (0701-0797,8877)

50 Events unique to one campaign (5001-5004)

410

Appendix I: Selected Chapter 5 Issue Distance Measures

TABLE I.1: SPENDING DISTANCE MEASURES

Parameter Standar MODEL: DISTANCE = YEAR N Intercep Estimate d Error R2 t SPENDING – Respondent Placements R Spending Self Placement 11 -25.253 .015 0.012 .147 Democratic Party Position 11 -1.484 0.003 0.005 .039 Democratic Prez Cand Position 6 3.451 0.001 0.005 .006 Republican Party Position 11 -6.009 0.005 0.008 .035 Republican Prez Cand Position 6 -33.148 0.018 0.015 .276 SPENDING - Respondent Relative Distance from Parties & Candidates R – DP (Dem Party) 11 -20.005 0.010 0.013 .059 R – DPC (Dem Prez Cand) 6 -33.08 0.016 * 0.008 .503 R – DHC (Dem House Cand) 9 -30.657 0.015 0.014 .142 R – DSC (Dem Sen Cand) - - - - - R – RP (Rep Party) 11 -26.362 0.014 0.011 .148 R – RPC (Rep Prez Cand) 6 -13.510 0.007 0.014 .061 R- RHC (Rep House Cand) 9 -37.660 0.019 *** 0.006 .594 R – RSC (Rep Sen Cand) - - - - - SPENDING - Respondent Relative Partisan Proximity (Rep Distance – Dem Distance) |R – RP| – |R – DP| 11 -5.673 0.003 0.011 .008 |R – RPC| – |R – DPC| 6 -12.678 0.006 0.009 .105 |R – RHC| – |R – DHC| 9 -1.639 0.001 0.010 .003 |R – RSC| – |R – DSC| - - - - - SPENDING – Total Relative Partisan Distance (Rep Distance + Dem Distance) |R – RP| + |R – DP| 9 -15.288 0.008 0.012 .068 |R – RPC| + |R – DPC| 6 52.959 -0.025 0.016 .383 |R – RHC| + |R – DHC| 9 36.712 -0.016 ** 0.007 .428 |R – RSC| + |R – DSC| - - - - - * significant at .10 level ** significant at .05 level ***significant at .01 level

411

TABLE I.2: DEFENSE SPENDING DISTANCE MEASURES

Parameter Standar MODEL: DISTANCE = YEAR N Intercep Estimate d Error R2 t DEFENSE SPENDING – Respondent Placements R Defense Spending Self 11 -0.090 0.002 0.024 .001 Placement Democratic Party Position 10 -4.061 0.004 0.008 .026 Democratic Prez Cand Position 7 -25.009 0.014 0.014 .174 Republican Party Position 10 18.881 -0.006 0.010 .054 Republican Prez Cand Position 7 27.999 -0.011 0.022 .053 DEFENSE SPENDING - Respondent Relative Distance from Parties & Candidates R – DP (Dem Party) 10 -9.697 0.005 0.020 .008 R – DPC (Dem Prez Cand) 7 34.820 -0.0117 0.024 .090 R – DHC (Dem House Cand) - - - - - R – DSC (Dem Sen Cand) - - - - - R – RP (Rep Party) 10 -22.809 0.011 0.025 .023 R – RPC (Rep Prez Cand) 7 -7.559 0.003 0.028 .003 R- RHC (Rep House Cand) - - - - - R – RSC (Rep Sen Cand) - - - - - DEFENSE SPENDING - Respondent Relative Partisan Proximity (Rep Distance – Dem Distance) |R – RP| – |R – DP| 10 15.186 -0.008 0.016 0.027 |R – RPC| – |R – DPC| 10 -15.469 0.008 0.021 .028 |R – RHC| – |R – DHC| - - - - - |R – RSC| – |R – DSC| - - - - - DEFENSE SPENDING – Total Relative Partisan Distance (Rep Distance + Dem Distance) |R – RP| + |R – DP| - - - - - |R – RPC| + |R – DPC| 7 76.156 -0.037 0.045 .118 |R – RHC| + |R – DHC| - - - - - |R – RSC| + |R – DSC| - - - - - * significant at .10 level ** significant at .05 level ***significant at .01 level

412

BIBLIOGRAPHY

Abramowitz, Alan I., and Gary C. Jacobson. 2006. Disconnected, or Joined at the Hip? In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Abramowitz, Alan I., and Kyle L. Saunders. 1998. Ideological Realignment in the U.S. Electorate. Journal of Politics 60 (3):634-652.

———. 2005. Why Can’t We All Just Get Along? The Reality of a Polarized America. The Forum (2), http://www.bepress.com/forum.

Adams, Greg D. 1997. Abortion: Evidence of Issue Evolution. American Journal of Political Science 41:718-37.

Aldrich, John H. 1995. Why Parties? The Origin and Transformation of Political Parties in America. Chicago: University of Chicago Press.

Alverez, R. Michael. 1997. Information and Elections. Ann Arbor: University of Michigan Press.

Alverez, R. Michael, and Charles H. Franklin. 1994. Uncertainty and Political Perceptions. Journal of Politics 56 (3):671-88.

Alverez, R. Michael, and Jonathan Nagler. 1995. Economics, Issues and the Perot Candidacy: Voter Choice in the 1992 Presidential Election. American Journal of Political Science 39:714-44.

Arnold, Douglas R. 1990. The Logic of Congressional Action. New Haven, CT: Yale University Press.

Balanda, Kevin P., and H. L. MacGillivray. 1988. Kurtosis: A Critical Review. American Statistician 42:111-119.

Bartels, Larry. 2005. What's the Matter with What's the Matter with Kansas? In American Political Science Association Annual National Conference. Washington, D.C.

———. 2006. What's the Matter with What's the Matter with Kansas? Quarterly Journal of Political Science 1:201-26.

Bartels, Larry M. 1986. Issue Voting under Uncertainty: An Empirical Test. American Journal of Political Science 30 (4):709-28.

Berelson, Bernard R., Paul R. Lazarsfeld, and William N. McPhee. 1954. Voting: A Study of Opinion Formation in a Presidential Campaign. Chicago: University of Chicago Press.

Bishop, Bill. 2004. THE GREAT DIVIDE: Divided electorate is a natural for a bitter, issueless campaign. The American Statesman.

413

Black, Duncan. 1958. The Theory of Committees and Elections. Cambridge: Cambridge University Press.

Blau, Peter M. 1977. Inequality and Heterogeneity: A Primitive Theory of Social Structure. New York: Free Press.

———. 1977. A Macrosocial Theory of Social Structure. American Journal of Sociology 83.

Blumenthal, Sidney. 1994. A Reluctant Warrior - With Haiti, Clinton Redefines the Post-Cold War Presidency. L.A. Times, September 18th, 1994, online.

Bok, Derek. 2003. Political Leadership in the Great Health Care Debate of 1993-1994. In Public Discourse in America: Conversation and Community in the Twenty-First Century, edited by S. P. Steinberg and J. Rodin. Philadelphia: University of Pennsylvania Press.

Bolce, Louis, and Gerald De Maio. 1998. The Impact of the Christian Fundamentalist Symbol on Party Coalitions: A Reference Group Theory Approach In American Political Science Association. Boston.

———. 1999. The Anti-Christian Fundamentalist Factor in Contemporary Politics. Public Opinion Quarterly 63 (4):508-42.

———. 1999. Religious Outlook, Culture War Politics, and Antipathy toward Christian Fundamentalists. Public Opinion Quarterly 63:29-61.

Borusk, Alan J. 2004. Red, blue and a bit bruised: A divided Wisconsin looks beyond the vote. Milwaukee Journal Sentinel, Nov. 1st, 2004.

Bough, Brian, James Endersby, Donald M. Gooch, and Monica Klimek. 2004. Vote Choice and Spatial Perception: Distance as a Determinant of the Presidential Vote 1984-2000. In Midwest Political Science Association National Conference. Chicago, ILL: MWPSA.

Bowles, Samuel, and Ugo Pagano. 2003. Economic Integration, Cultural Standardization, and the Politics of Social Insurance. In University of Siena Economics Working Papers. Siena, Italy: Social Science Research Network.

Bowman, Karyln H. 2008. Public Opinion on the War in Iraq. AEI Public Opinion Studies, http://www.aei.org/docLib/200701121_roody2.pdf.

Brady, David W., and Hahrie C. Han. 2006. Polarization Then and Now: A Historical Perspective. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Brewer, M.B. 1979. Ingroup bias in the Minimal Intergroup Situations: A Congitive Motivational Analysis. Psychological Bulletin 86:307-24.

Broder, David. 2000. Burying the Hatchet. Washington Post, November 30th, 2000.

Brody, Richard A. 1991. Assessing the President. Stanford: Stanford University Press.

414

Brooks, Clem. 2002. Religious Influence and the Politics of Family Decline Concern: Trends, Sources, and U.S. Political Behavior. American Sociological Review 67 (2):191-211.

Brooks, Clem, and Jeff Manza. 2004. A GREAT DIVIDE? Religion and Political Change in U.S. National Elections, 1972-2000. Sociological Quarterly 45 (3):421-450.

Brooks, David. 2001. One Nation, Slightly Divisible. In The Atlantic. Washington, D.C.

Buchanan, Patrick J. 1992. 1992 Republican National Convention Speech. Houston, TX.

Bush, George H.W. 1991. "New World Order" Speech to Congress. http://www.al- bab.com/arab/docs/pal/pal10.htm.

Calvert, Randall L. 1985. Robustness of the Multidimensional Voting Model: Candidate Motivations, Uncertainty, and Convergence. American Journal of Political Science 29 (1):69-95.

———. 1986. Models of Imperfect Information in Politics. Chur: Harwood.

Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. Chicago: University of Chicago Press.

Campbell, Angus, and Homer C. Cooper. 1956. Group Differences in Attitudes and Votes. Ann Arbor, MI: Survey Research Center, Institute for Social Research, University of Michigan.

Campbell, Angus, Gerald Gurin, and Warren Miller. 1954. The Voter Decides. Evanston, IL: Peterson.

Campbell, Angus, and Donald Stokes. 1959. Partisan Attitudes and the Presidential Vote. In American Voting Behavior, edited by E. Burdick and A. J. Brodbeck. Glencoe, IL: Free Press.

Campbell, James E., and Carl M. Cannon. 2006. Polarization Runs Deep, Even by Yesterday's Standards. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Carmines, Edward G., and J. David Gopoian. 1981. Issue Coalitions, Issueless Campaigns: the Paradox of Rationality in American Presidential Elections. Journal of Politics 43 (4):1170- 1189.

Carmines, Edward G., and Geoffrey C. Layman. 1997. Issue Evolution in Postwar American Politics: Old Certainties and Fresh Tensions. In Present Discontents: American Politics in the Very Late Twentieth Century, edited by B. E. Shafer. Chatham, NJ: Chatham House.

Carmines, Edward G., and James Woods. 2002. The Role of Party Activists in the Evolution of the Abortion Issue. Political Behavior 24 (4):361-77.

Cherry, Brian. 2007. The Great Divide. USA Partisan, June, 28, 2007.

415

Childs, Harwood. 1965. Public Opinion: Nature, Formation, and Role. Princeton, New Jersey: Van Nostrand.

Collie, Melissa P., and John Lyman Mason. 2000. The Electoral Connection Between Party and Constituency Reconsidered: Evidence from the U.S. House of Representatives, 1972-94. In Continuity and Change in House Elections, edited by D. W. Brady, J. F. Cogan and M. P. Fiorina. Stanford, CA: Stanford University Press.

Conover, Pamela Johnson, Virginia Gray, and Steven Coombs. 1982. Single-issue Voting: Elite- Mass Linkages. Political Behavior 4 (4):309-331.

Conover, Pamela Johnston, and Stanley Feldman. 1989. Candidate Perception in an Ambiguous World: Campaigns, Cues, and the Inference Processes. American Journal of Political Science 33 (4):912-39.

Converse, Philip. 1964. The Nature of Belief Systems in Mass Publics. In Ideology and Discontent, edited by D. Apter. Glencoe, IL: Free Press.

Dalton, Russell J. 1987. Generational Change in Elite Political Beliefs: The Growth of Ideological Polarization. Journal of Politics 49 (4):976-97.

Dalton, Russell J., Paul Allen Beck, and Robert Huckfeldt. 1998. Partisan Cues and the Media: Information Flows in the 1992 Presidential Election. American Political Science Review 92 (1):111-26.

Darcy, Robert Emmett. 1980. Consensus, Constraint, and Political Polarization in Recent Presidential Elections. Political Behavior 2 (2):147-61.

Darlington, Richard B. 1970. Is Kurtosis Really "Peakedness?" The American Statistician 24 (2):19-22.

DeSart, Jay A. 1995. Information Processing and Partisan Neutrality: A Reexamination of the Party Decline Thesis. Journal of Politics 57 (3):776-95.

Deutsch, M. 1971. Conflict and its Resolution. In Conflict Resolution: Contributions of the Behavioral Sciences, edited by C. G. Smith. Notre Dame: University of Notre Dame Press.

DiMaggio, Paul , John Evans, and Bethany Bryson. 1996. Have Americans' Social Attitudes Become More Polarized? American Journal of Sociology 102:690-755.

Dionne Jr., E. J. Class Warfare and Political Polarization. Washington Post Online 2006 [cited. Available from www.realclearpolitics.com/articles/2006/03/class_warfare_and_political_po.html.

Downs, Anthony. 1957. An Economic Theory of Democracy. New York: Harper and Row.

Duclos, Jean-Yves, Joan Esteban, and Debraj Ray. 2004. Polarization: Concepts, Measurment, Estimation. Econometrica 72 (6):1737-1772.

416

Edsall, Thomas B. 1986. Shifts in Political Alignments. Dissent 33:151-55.

Edwards, III George C., William Mitchell, and Reed Welch. 1995. Explaining Presidential Approval: The Significance of Issue Salience. American Journal of Political Science 39 (1):108-134.

Eldersveld, Samuel J. 1952. The Independent Vote: Measurement, Characteristics, and Implications for Party Strategy. American Political Science Review 46 (3):732-53.

Enelow, James M., and Melvin J. Hinich. 1981. A New Approach to Voter Uncertainty in the Downsian Spatial Model. American Journal of Political Science 25 (3):483-93.

Engel, Matthew. 2002. Special Report George Bush's America: A Galaxy Far, Far Away. The Guardian, Feburary 26, 2002.

Esteban, Joan-Maria, and Debraj Ray. 1994. On the Measurement of Polarization. Econometrica 62 (4):819-51.

Evans, John H. 1996. "Culture Wars" or Status Group Ideology as the Basis of US Moral Politics. The International Journal of Sociology and Social Policy 16 (1):15-34.

———. 1997. Worldviews or Social Groups as the Source of Moral Value Attitudes: Implications for the Culture Wars Thesis. Sociological Forum 12 (3):371-404.

———. 2003. Have Americans’ Attitudes Become More Polarized?—An Update. Social Science Quarterly 84 (1):72-90.

Evans, John H., Bethany Bryson, and Paul DiMaggio. 2001. Opinion Polarization: Important Contributions, Necessary Limitations. American Journal of Sociology 106 (4):944-959.

Feddersen, Timothy J., and Wolfgang Pesendorfer. 1997. Voting Behavior and Information Aggregation in Elections With Private Information. Econometrica 65 (5):1029-1058.

———. 1999. Elections, Information Aggregation, and Strategic Voting. Proceedings of the National Academy of Sciences of the United States of America 96 (19):10572-10574.

Ferejohn, John A., and Roger G. Noll. 1978. Uncertainty and the Formal Theory of Political Campaigns. American Political Science Review 72 (2):492-505.

Finucan, H. M. 1964. A Note on Kurtosis. Journal of the Royal Statistical Society. Series B (Methodological) 26 (1):111-112.

Fiorina, Morris P. 1978. Economic Retrospective Voting in American National Elections: A Micro- Analysis. American Journal of Political Science 22 (2):426-443.

———. 1999. Whatever Happened to the Median Voter? In Massachusetts Institute of Technology Conference on Parties and Congress. Cambridge, MA.

417

Fiorina, Morris P., Samuel J. Abrams, and Jeremy C. Pope. 2004. Culture War? The Myth of a Polarized America, Great Questions in Politics Series. New York: Longman.

Fiorina, Morris P., and Matthew S. Levendusky. 2006. Disconnected: The Political Class versus the People. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Fleisher, R., and J.R. Bond. 2000. Congress and the President in a Partisan Era. In Polarized Politics: Congress and the President in a Partisan Era, edited by R. Fleisher and J. R. Bond. Washington, D.C.: CQ Press.

Fleisher, Richard, and John R. Bond. 2004. The shrinking middle in the US Congress. British Journal of Political Science 34 (3):456-9.

Frank, Thomas. 2004. What's the Matter with Kansas? How Conservatives Won the Heart of America. New York: Henry and Holt Company.

Galston, William A., and Elaine C. Kamarak. 2005. The Politics of Polarization. Washington, D.C.: ThirdWay.

Galston, William A., and Pietro S. Nivola. 2006. Delineating the Problem. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institution Press.

Geer, John G. 1992. New Deal Issues and the American Electorate, 1952-1988. Political Behavior 14 (1):45-65.

Gelman, Andrew, Boris Shor, Joseph Bafumi, and David Park. 2005. Rich state, poor state, red state, blue state: What’s the matter with Connecticut? . In Midwest Political Science Association meeting. Chicago.

Gerber, Alan S., and Donald P. Green. 1998. Rational Learning and Partisan Attitudes. American Journal of Political Science 42 (3):794-818.

Glatter, Lesli Linka. 2007. You Don't Want to Know. In House, edited by B. Singer. United States: Fox.

Groenveld, Richard A., and Glen Meeden. 1984. Measuring Skewness and Kurtosis. The Statistician 33 (4):391-99.

Hadley, Eryn. 2005. Did the Sky Really Fall? Ten Years After California's Proposition 2009. BYU Journal of Public Law 20 (103).

Herrera, Richard. 1992. The Understanding of Ideological Labels by Political Elites: A Research Note. Western Political Quarterly 45:1021-35.

Hetherington, Marc J. 2001. Resurgent Mass Partisanship: The Role of Elite Polarization. American Political Science Review 95 (3):619-631.

418

Hildebrand, David K. 1971. Kurtosis Measures Bimodality? The American Statistician 25 (1):42- 43.

Hill, Kim Quaile, and Patricia A. Hurley. 1979. Mass Participation, Electoral Competitiveness, and Issue-Attitude Agreement between Congressmen and Their Constituents. British Journal of Political Science 9 (4):507-511.

Hinich, Melvin J., and Michael C. Munger. 1994. Ideology and the Theory of Political Choice. Edited by J. E. Jackson and C. H. Achen, Michigan Studies in Political Analysis. Ann Arbor: University of Michigan Press.

Hotelling, Harold. 1929. Stability in Competition. Economic Journal 39:41-57.

Hout, Michael, and Claude S. Fischer. 2002. Why More Americans Have No Religious Preference: Politics and Generations. American Sociological Review 67 (2):165-190.

Huckfeldt, Robert, and Carol W. Kohfeld. 1989. Race and the Decline of Class in American Politics. Champaign, IL: University of Illinois Press.

Huckfeldt, Robert, and John Sprague. 1988. Choice, Social Structure, and Political Information: The Information Coercion of Minorities. American Journal of Political Science 32 (2):467- 482.

Huckfeldt, Robert, and John Sprague. 1987. Networks in Context: The Social Flow of Political Information. American Political Science Review 81:1197-216.

Hunter, James Davison. 1991. Culture Wars: The Struggle to Define America. New York: Basic Books

Jacobellis v. Ohio. 1964. In United States Reports: United States Supreme Court.

Jacobson, Gary C. 2000. The Electoral Basis of Partisan Polarization in Congress. In Annual Meeting of the American Political Science Association.

———. 2002. Partisan Polarization in Presidential Support: The Electoral Connection. In Annual Meeting of the American Political Science Association.

Jacobson, Gary C., and Thomas B. Edsall. 2006. Why Other Sources of Polarization Matter More. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Jia, Panle. 2007. What happens when Wal-Mart Comes to Town: An Empirical Analysis of the Discount Retailing Industry. Cambridge, MA: Massachusetts Institution of Technology.

Jones, David R. 2001. Party Polarization and Legislative Gridlock. Political Research Quarterly 54 (1):125-141.

Jonsson, Patrik. 2004. Republican America: How Georgia Went 'Red'. The Christian Science Monitor, July 15, 2003, 1-6.

419

Kaplansky, Irving. 1945. A Common Error Concerning Kurtosis. Journal of the American Statistical Association 40 (230):259.

Key, V. O. 1955. A Theory of Critical Elections. Journal of Politics 17:3-18.

Kiewiet, D. Roderick. 1983. Macroeconomics and Micropolitics: The Electoral Effects of Economic Issues. Chicago: University of Chicago Press.

Kinder, Donald R., and D. Roderick Kiewiet. 1984. Sociotropic Politics: The American Case. In Controversies in Voting Behavior, edited by R. G. Niemi and H. F. Weisberg. Washington, D.C.: Congressional Quarterly Press.

Kosmin, Barry A., and Ariela Keysar. 2009. American Nones: The Profile of the No Religion Population. In American Religious Identification Survey. Hartford, Connecticut: Trinity College.

Kristof, Nicholas D. 2004. Living Poor, Voting Rich. New York Times, November 6th, 2004, A19.

Layman, Geoffrey C. 1996. The Culture Wars in the States: Religious Polarization among State Party Elites and State Electorates. Paper read at Midwest Political Science Association, April, at Chicago.

———. 1997. Religion and Political Behavior in the United States: The Impact of Beliefs, Affiliation, and Commitment from 1980-1994. Public Opinion Quarterly 61:288-316.

———. 1999. Culture Wars in the American Party System: Religious and Cultural Change Among Partisan Activists since 1972. American Politics Quarterly 27:89-121.

———. 2001. The Great Divide: Religious and Cultural Conflict in American Party Politics. New York: Columbia University Press.

Layman, Geoffrey C., and John C. Green. 1998. The Changing Religious Voter: The Impact of Belonging, Believing, and Behaving in the 1960s and 1990s. Paper read at Midwest Political Science Association, April, at Chicago.

Layman, Geoffrey C., and Thomas M. Carsey. 1998. Why Do Party Activists Convert? An Analysis of Individual-Level Change on the Abortion Issue. Political Research Quarterly 51:723-50.

———. 1999. Ideological Realignment in Contemporary American Politics: General Ideological Polarization Rather Than Conflict Displacement. Paper read at American Political Science Association, September, at Atlanta.

Layman, Geoffrey C., and Edward G. Carmines. 1997. Cultural Conflict in American Politics: Religious Traditionalism, Postmaterialism, and U.S. Political Behavior. Journal of Politics 59:751-77.

Layman, Geoffrey C., and Thomas M. Carsey. 2002. Party Polarization and "Conflict Extension" in the American Electorate. American Journal of Political Science 46 (4):786-802.

420

———. 2002. Party Polarization and Party Structuring of Policy Attitudes: A Comparison of Three NES Panel Studies. Political Behavior 24 (3):199-236.

———. 2003. Erratum: Party Polarization and 'Conflict Extension' in the American Electorate. American Journal of Political Science 47 (2):388.

Lazarsfeld, Paul, Bernard Berelson, and Hazel Gaudet. 1948. The People's Choice. New York: Columbia University Press.

Lindaman, Kara, and Donald P. Haider-Markel. 2002. Issue Evolution, Political Parties, and the Culture Wars. Political Research Quarterly 55 (1):91-110.

Lipset, Seymour M., and Stein Rokkan. 1967. Cleavage Structures, Party Systems, and Voter Alignments. In Party Systems and Voter Alignments, edited by S. M. L. a. S. Rokkan. New York: Free Press.

Lipset, Seymour Martin. 1960. Political Man. Baltimore: Johns Hopkins University Press.

Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables, Advanced Quantitative Techniques in the Social Sciences Series. Thousand Oaks: SAGE Publications.

Lord, C., L. Ross, and M. Lepper. 1979. Biased Assimilation and Attitude Polarization: The Effects of Prio Theories on Subsequently Considered Evidence. Journal of Personality and Social Psychology 37:2908-2109.

Maass, A., P. Corvino, and L. Arcuri. 1994. Linguistic Intergroup Bias and the Mass Media. Revue de Psychologie Sociale 1:31-43.

Manza, Jeff, Michael Hout, and Clem Brooks. 1995. Class Voting in Capitalist Democracies Since WWII. Annual Review of Sociology 21:137-63.

Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven, CT: Yale University Press.

McCarty, Nolan, Keith T. Poole, and Howard Rosenthal. 2006. Polarized America: The Dance of Ideology and Unequal Riches. Cambridge, MA: MIT Press.

McCloskey, Herbert. 1960. Issue Conflict and Consensus Among Party Leaders and Followers. American Political Science Review 54 (2):406-427.

McClosky, Herbert, Paul J. Hoffmann, and Rosemary O'Hara. 1960. Issues Conflict and Consensus Among Party Leaders and Followers. American Political Science Review 54 (2):406-427.

McKelvey, Richard D., and Peter C. Ordeshook. 1985. Sequential Elections with Limited Information. American Journal of Political Science 29 (3):480-512.

———. 1986. Information, Electoral Equilibria, and the Democratic Ideal. Journal of Politics 48 (4):909-937.

421

Miller, Alan S., and John P. Hoffmann. 1999. The Growing Divisiveness: Culture Wars or a War of Words? Social Forces 78 (2):721-745.

Miller, Arthur H., and Martin P. Wattemberg. 1984. Politics from the Pulpit: Religiosity and the 1980 Elections. Public Opinion Quarterly 48:301-17.

Miller, S.A. 2009. Public Option Splits House Democrats. The Washington Times, September 9th, 2009.

Miller, Warren E., and M. Kent Jennings. 1986. Parties in Transition: A Longitudinal Study of Party Elites and Party Supporters. New York: Sage.

Moors, J. J. A. 1986. The Meaning of Kurtosis: Darlington Reexamined. The American Statistician 40 (4):283-84.

———. 1988. A Quantile Alternative for Kurtosis. The Statistician 37 (1):25-32.

Morton, Rebecca B. 1993. Incomplete Information and Ideological Explanations of Platform Divergence. American Political Science Review 87:382-92.

———. 2006. Polarized over Policy or Voting on Valence? A War Between the States? In Analyzing Elections, edited by R. B. Morton. New York: W.W. Norton & Company.

Mouw, T., and M. E. Sobel. 2001. Culture Wars and Opinion Polarization: The Case of Abortion. American Journal of Sociology 106 (4):913-44.

Mutz, Diana C. 2006. How the Mass Media Divide Us. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Nadeau, Richard, and Harold W. Stanley. 1993. Class Polarization in Partisanship among Native Southern Whites, 1952-90. American Journal of Political Science 37 (3):900-919.

Niemi, Richard A., and Larry Bartels. 1985. New Measures of Issue Salience: An Evaluation. Journal of Politics 47 (4):1212-1220.

Nivola, Pietro S., and David W. Brady, eds. 2006. Red and Blue Nation? Characteristics and Causes of America's Polarized Politics. 2 vols. Vol. 1. Washington, D.C.: Brookings Institution Press.

Nixon, Richard Milhouse. 1962. Six Crises. New York: Doubleday.

O'Keefe, Mark. 2004. A Divide Forms When Politics Battles Religion. Houston Chronicle, Sat 02/14/2004, 1.

Ono, Keiko. 2005. Electoral Origins of Partisan Polarization in Congress: Debunking the Myth. Extensions: Journal of the Carl Albert Congressional Research and Studies Center 1-8.

422

Ortiz, Hector. 2007. The Battle for Social Security: Four Historical and Institutional Explanations of Bush's Social Security Reform. In Annual Meeting of the Midwest Political Science Association. Palmer House Hotel, Chicago, IL.

Page, Benjamin I., and Calvin C. Jones. 1979. Reciprocal Effects of Policy Preferences, Party Loyalties, and the Vote. American Political Science Review 73:1071-90.

Page, Benjamin I., and Robert Y. Shapiro. 1992. The Rational Public: Fifty Years of Trends in Americans' Policy Preferences, American Politics and Political Economy Series. Chicago, IL: The University of Chicago Press.

Palfrey, Thomas. 1984. Spatial Equilibrium with Entry. Review of Economic Studies 51:139-56.

Palfrey, Thomas R., and Keith T. Poole. 1987. The Relationship between Information, Ideology, and Voting Behavior. American Journal of Political Science 31 (3):511-530.

Petrocik, John R. 1981. Party Coalitions. Chicago: University of Chicago Press.

Planned Parenthood of Southeastern Pennsylvania v. Casey. 1992. In United States Reports: United States Supreme Court.

Platt, Glenn, Keith T. Poole, and Howard Rosenthal. 1992. Directional and Euclidean Theories of Voting Behavior: A Legislative Comparison. Legislative Studies Quarterly 17 (4):561-572.

Poole, Keith T., and R. Howard. 1997. Congress: A Political-Economic History of Roll Call Voting. New York: Oxford University Press.

Poole, Keith T., and Howard Rosenthal. 1984. The Polarization of American Politics. Journal of Politics 46 (4):1061-1079.

———. 2001. D-Nominate after 10 Years: A Comparative Update to Congress: A Political- Economic History of Roll-Call Voting. Legislative Studies Quarterly 26 (1):5-29.

Popkin, Samuel L. 1994. The Reasoning Voter: Communication and Persuasion in Presidential Campaigns. Chicago: University of Chicago Press.

Rabinowitz, George, James W. Prothro, and William Jacoby. 1982. Salience as a Factor in the Impact of Issues on Candidate Evaluations. Journal of Politics 44:41-63.

Reagan, Ronald. 1983. Address to the Nation on National Security edited by W. House. Washington, D.C.

———. 1983. The Evil Empire Speech. http://www.nationalcenter.org/ReaganEvilEmpire1983.html.

RePass, David E. 1971. Issue Salience and Party Choice. American Political Science Review 65:389-400.

Roe v. Wade. 1973. In United States Reports: United States Supreme Court.

423

Samuelson, Robert J. 2004. How Polarization Sells. Washington Post, June 30th, 2004, A21.

Schlesinger, Arthur M., Jr. 1958. The Coming of the New Deal. New York: Houghton Mifflin.

Shapiro, Catherine R., David W. Brady, Richard A. Brody, and John A. Ferejohn. 1990. Linking Constituency Opinion and Senate Voting Scores: A Hybrid Explanation. Legislative Studies Quarterly 15 (4):599-621.

Shapiro, Michael J. 1969. Rational Political Man: A Synthesis of Economic and Social Psychological Perspectives. American Political Science Review 63:1106-1119.

Sharp, Elaine B. 1999. The Sometime Connection: Public Opinon and Social Policy. Albany, N.Y.: State University of New York Press.

Shaw, Greg M. 2003. The Polls-Trends Abortion. Public Opinion Quarterly 67:407-429.

Shepsle, Kenneth. 1972. The Strategy of Ambiguity, Uncertainty, and Competition. American Political Science Review 66:551-68.

Stimson, James A. 2004. Tides of Consent: How Public Opinion Shapes American Politics. New York: Cambridge University Press.

Stonecash, Jeffrey M. 2000. Class and Party in American Politics. Boulder, CO: Westview Press.

———. 2005. Scaring the Democrats: what's the matter with Thomas Frank's argument? The Forum 3 (3):article 4.

Stonecash, Jeffrey M., Mark D. Brewer, and Mack D. Mariani. 2003. Diverging Parties: Social Change, Realignment, and Party Polarization. Edited by L. C. Dodd, Transforming American Politics. New York: Westview Press.

Stonecash, Jeffrey M., Mark D. Brewer, Eric Petersen, Mary P. McGuire, and Lori Beth Way. 2000. Class and Party: Secular Realignment and the Survival of Democrats outside the South. Political Research Quarterly 53 (4):731-752.

Sullivan, John, James Piereson, and George Marcus. 1978. Ideological Constraint in the Mass Public: A Methodological Critique and Some New Findings. American Journal of Political Science 22:233-249.

Sunstein, Cass R., David Schkade, and Lisa M. Ellman. 2006. Are Judges Political?: An Empirical Analysis of the Federal Judiciary. Washington, D.C.: Brookings Institution Press.

Tajfel, Henri, and John Turner. 1979. An Integrative Theory of Intergroup Conflict. In The Social Psychology if Intergroup Relations, edited by W. G. Austin and S. Worchel. Monterey, CA: Brooks-Cole.

Tajfel, Horace. 1970. Experiments in intergroup discrimination. Scientific American 223:96-102.

424

———. 1982. Social Identity and Intergroup Relations. Cambridge, England: Cambridge University Press.

Taylor, Donald, and Fathali Moghaddam. 1994. Social Identity Theory. In Theories of Intergroup Relations: International Social Psychological Perspectives, edited by D. Taylor. Westport, CT: Praeger Publishers.

Tufte, Edward. 1975. Determinants of the Outcomes of Midterm Congressional Elections. American Political Science Review 69:812-26.

Waldman, Paul. 2007. Religion and the Threat Effect. The American Prospect.

Weber, Vin. 1993. Tactical retreat - Bill Clinton's defense budget. National Review, May 10th, 1993.

Webster v. Reproductive Health Services. 1989. In United States Reports: United States Supreme Court.

Wildermuth, John. 2004. Red State, Blue State: California's political map reflects the nation -- Dems capture metro area while vast interior goes Republican. San Francisco Chronicle, Sunday, November 7, 2004.

Wilson, James Q. 1962. The Amateur Democrat: Club Politics in Three Cities. Chicago: University of Chicago Press.

Wolfe, Alan, and Andrew Kohut. 2006. Myths and Realities of Religion in Politics. In Red and Blue Nation? Characteristics and Causes of America's Polarized Politics, edited by P. S. Nivola and D. W. Brady. Washington, D.C.: Brookings Institute Press.

Wright, John R., and Arthur S. Goldberg. 1985. Risk and Uncertainty as Factors in the Durability of Political Coalitions. American Political Science Review 79 (3):704-18.

York, Byron. 2009. For the Left, war without Bush is not war at all. Washington Examiner, August 18th, 2009.

Young, Gary. 2004. Culture War Casualties : Polarization in US politics Reflects the Huge Rift that has Emerged over Individual Lifestyle and Moral Values. The Guardian.

Zaller, John. 1992. The Nature and Origins of Mass Opinion. New York: Cambridge University Press.

Zeller, Tom. 2004. Ideas & Trends; One State, Two State, Red State, Blue State. The New York Times, August 31, 2008, 2.

425

VITA

Donald M. Gooch was born in Joliet, Illinois in May of 1975. Mr. Gooch is an Eagle

Scout, and a past member of Alpha Phi Omega. He is currently an Assistant Professor of Political

Science in the Department of History & Political Science at Arkansas Tech University. He has been at ATU since July, 2007. Mr. Gooch is an alumnus of the University of Central Arkansas and has a Masters degree in Political Science from the University of Arkansas at Fayetteville, earned in 2001. He entered the doctoral program in political science at the University of Missouri,

Columbia in 2001. Professor Gooch is an expert in American politics and public policy and teaches a broad spectrum of classes in these fields such as courses in political behavior, constitutional law, research methods, and American political institutions (i.e. the U.S. Congress).

He is the faculty sponsor of the Pre-Law Society and the Young Conservatives at ATU and the

Pre-Law advisor for the SS&P department. Donald Gooch is a dedicated fan of the St. Louis

Cardinals and the Chicago Bears.

His current research agenda involves studying trends in political polarization in the

American electorate and the distributional relationship on issue and partisan opinions between elites and the mass public. He also is continuing to pursue a variety of projects that include a comparative analysis of campaign finance laws, an assessment of the effect mass electorate issue positions relative to perceived candidate issue positions have on voting behavior, a paper on the consequences of majority-minority districts on voting behavior in Congress, and an analysis of rational bureaucratic decision-making in response to federal performance benchmarks.

426