Public Opinion in State Politics

Public Opinion in State Politics Public Opinion in State Politics Edited by Jeffrey E. Cohen Stanford University P...

Author: Jeffrey Cohen

76 downloads 1906 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Public Opinion in State Politics

Public Opinion in State Politics

Edited by

Jeffrey E. Cohen

Stanford University Press Stanford, California 2006

Stanford University Press Stanford, California ©2006 by the Board of Trustees of the Leland Stanford Junior University. All rights reserved. No part of this book may be reproduced or transmitted in any form or by any means, electronic or mechanical, including photocopying and recording, or in any information storage or retrieval system without the prior written permission of Stanford University Press. Printed in the United States of America on acid-free, archival-quality paper Library of Congress Cataloging-in-Publication Data Public opinion in state politics / edited by Jeffrey E. Cohen. p. cm. Includes bibliographical references and index. isbn-13: 978-0-8047-5300-5 (cloth : alk. paper) 1. State governments — United States — Public opinion. 2. Public opinion — United States. I. Cohen, Jeffrey E. jk2408.p83 2006 320.6093 — dc22 2006009397

To David and Ruth Nice, for their friendship

Contents

List of Tables and Figures Preface Acknowledgments Contributors 1. Introduction: Studying Public Opinion in the American States Jeffrey E. Cohen

ix xiii xv xvii

3

2. Does Familiarity Breed Contempt? Examining the Correlates of State-Level Conﬁdence in the Federal Government Paul Brace and Martin Johnson

19

3. State Residency, State Laws, and Public Opinion Barbara Norrander and Clyde Wilcox

38

4. Mexican-American and Cuban-American Public Opinion: Diﬀerences at the State Level? David L. Leal

53

5. Public Opinion in the States: Determinants of Legislative Job Performance John A. Hamman

79

6. The State Economy, the National Economy, and Gubernatorial Popularity Jeffrey E. Cohen and James D. King

102

viii

Contents

7. Ideological Cleavage, Political Competition, and Policy Making in the American States Charles J. Barrilleaux

121

8. The Civil State:Trust, Polarization, and the Quality of State Government Eric M. Uslaner

142

9. Public Opinion and Policy Making in the Culture Wars: Is There a Connection Between Opinion and State Policy on Gay and Lesbian Issues? Donald p. Haider-Markel and Matthew S. Kaufman

163

10. Citizen Inﬂuences on State Policy Priorities: The Interplay of Public Opinion and Interest Groups Saundra K. Schneider and William G. Jacoby

183

11. State-Level Opinions from National Surveys: Poststratiﬁcation Using Multilevel Logistic Regression David K. Park, Andrew Gelman, and Joseph Bafumi

209

12. Public Opinion in the States: A Quarter Century of Change and Stability Robert S. Erikson, Gerald C. Wright, and John P. McIver

229

13. Conclusions:Where We Have Been,Where Should We Go Jeffrey E. Cohen

254

Index

271

List of Tables and Figures

Tables 2.1. State-Level Conﬁdence in the Federal Government 2.2. Modeling State Conﬁdence in the Federal Government 3.1. The Inﬂuence of Demographic Traits and State Residency on Core Attitudes of Party Identiﬁcation and Ideology 3.2. The Inﬂuence of Demographic Traits, Core Attitudes, and State Residency on Attitudes on Abortion, Abortion Funding, and Parental Consent 3.3. The Inﬂuence of Demographic Traits, Core Attitudes, and State Residency on Attitudes on Capital Punishment and Aid to Blacks 3.4. The Eﬀect of State Laws, State Residency, Demographic Traits, and Core Attitudes on Public Opinion on Capital Punishment 4.1. Mexican-American Opinion Variation by State 4.2. Cuban-American Opinion Variation by State 5.1. Legislative Professionalism and Job Approval 1990–1991 (with Robust Standard Error Estimation) 5.2. The Probability of Approving State Legislative Performance for Diﬀerent Model Speciﬁcations and Independent Variable Values 6.1. Impact of Relative Unemployment on Gubernatorial Popularity 7.1. State Ideological, Population, Racial, and Partisan Diversity Scores and Ranks, and Group Strength Scores

26 31 42

44

45

48 66 68 92

93 112 129

ix

x

L i s t o f Ta b l e s a n d F i g u r e s

7.2. Inﬂuences of Ascriptive, Attitudinal, and Behavioral Competition on State Policy Liberalism 7.3. Baseline, Overidentiﬁed, and Reduced Models of State Policy Liberalism 7.4. Model of State Policy Liberalism with the Conditional Eﬀect of Ideological Diversity Appendix Table Chapter 7. Descriptive Statistics 8.1. Trust Estimates by State for the 1990s 8.2. Seemingly Unrelated Regressions for Government Quality Measures 8.3. Model of Corruption Perceptions in the American States 9.1. Selected Gay- and Lesbian-Related Policies in the American States 9.2. State Sodomy Law Repeal 9.3. Adoption of Hate Crime Law Including Sexual Orientation 9.4. Adoption of Same-Sex Marriage Bans 9.5. Determinants of State Scores on a Gay-Friendly Policy Index 10.1. The Impact of Public Opinion, Interest Groups, and Region on State Policy Priorities 10.2. The Impact of Public Opinion and Region on the Proportion of Interest Groups Within a State Working on Collective Goods and Particularized Beneﬁts 11.1. Years, Surveys, and Sample Size to Produce State-Level Opinion 11.2. Summary Statistics for CBS/New York Times, No Pooling, Complete Pooling, and Multilevel 11.3. Summary Statistics for CBS/New York Times, No Pooling, Complete Pooling, and Multilevel for States with Sample Sizes Less than 100 11.4. Summary Statistics for Weighted Survey Estimate and Multilevel Model 11.5. Mean Partisanship Estimates for Erikson,Wright, and McIver and Multilevel Model and Their Correlations 11.6. Mean Ideology Estimates for Erikson,Wright, and McIver and Multilevel Model and Their Correlations

132 134 135 137 150 153 157 168 174 175 176 177 201

203 211

218

219 220 222 223

List of Tables and Figures

12.1. Mean Ideological and Partisan Identiﬁcation over Seven Presidencies, CBS News/New York Times, Individual-Level Data 12.2. Reliability-Corrected Over-Time Correlations of State Ideological Identiﬁcation and State Party Identiﬁcation, 1977–2003, by Presidential Administration 12.3. Mean Reliability-Corrected Over-Time Correlations by Time Gap in Number of Presidential Administrations 12.4. Reliability-Corrected Cross-Sectional Correlations Between State Ideological Identiﬁcation and State Party Identiﬁcation 12.5. Reliability-Corrected Regressions of State Ideology and State Partisanship on Lagged State Ideology and Lagged State Partisanship 12.6. Reliability-Corrected Regressions of State Partisanship 2001–2003 (GW Bush) on State Partisanship and Ideology 1977–1980 (Carter) 12.7. (Non-reliability corrected) Correlations Between Ideological and Party Identiﬁcation by Region, for Six Presidencies 12.8. (Reliability-Adjusted) Variance of State Ideological Identiﬁcation and Party Identiﬁcation, by Presidencies 12.9. (Reliability-Adjusted) Regression of Presidential Vote on State Ideology and Partisanship, 1976–2004 Appendix Table Chapter 12. Current Estimates for State Partisanship and Ideology (from CBS/New York Times polls, 1996–2003)

xi

233

237 237 238

239

241 241 247 247

249

Figures 5.1. Trends in Institutional Job Performance and State Relative Unemployment 1980–1998 10.1. Mean Policy Score, 1998–2000 10.2. Mean State Score, 1998–2000 10.3. State Policy Priorities and Elazar Political Culture Categories 10.4. State Policy Priorities and Sharkansky Political Culture Scores 10.5. State Policy Priorities and Region

85 189 189 192 192 193

xii

L i s t o f Ta b l e s a n d F i g u r e s

10.6. State Policy Priorities and Electorate Partisanship 10.7. State Policy Priorities and Electorate Ideology 10.8. State Policy Priorities and Citizen Ideology 10.9. State Policy Priorities and Collective Goods Interest Groups 10.10. State Policy Priorities and Particularized Beneﬁt Interest Groups 11.1. Election Results by State versus (a) CBS/New York Times Weighted Means and Posterior Median Estimate 11.2. 1992 Election Results by State versus (a) Survey Weighted Means and (b) Posterior Median Estimate 11.3. Partisanship Self-Identiﬁcation in the United States: Erikson, Wright, and McIver versus Multilevel Model 11.4. Ideology Self-Identiﬁcation in the United States: Erikson, Wright, and McIver versus Multilevel Model 12.1. National Partisanship by Presidential Administration 12.2. Stability of State Ideology 12.3. Stability of State Partisanship 12.4. State Partisanship by Ideology 12.5. The Changing South: Partisanship from Carter to Bush 12.6. The Un-Changing South: Ideology from Carter to Bush 12.7. Partisan Change over Time: Reclaiming Democratic Voters 12.8. Partisan Histories of 2004 Red and Blue States

194 194 196 198 198 217 221 222 224 233 235 235 239 242 242 244 245

Preface

Over 20 years ago, Malcolm E. Jewell (1982) lamented that “we have given too little thought and devoted too little of our research resources attention and resources to the ﬁeld of state government and politics.” Public Opinion in State Politics represents the eﬀorts of 20 scholars to study an aspect of state politics that has generally eluded systematic and sustained inquiry: public opinion in the states. Until somewhat recently, the dearth of polls on the opinion of state publics often presented an overwhelming barrier to studying state-level public opinion and integrating such opinion into the study of state politics, policy making, and government.The explosion of public opinion polling over the past two decades or so by commercial ﬁrms, newspapers, and academic and nonproﬁt institutions has opened up the possibility of measuring public opinion in the American states. The studies included in Public Opinion in State Politics investigate statelevel public opinion from a number of angles. Some look at the factors that shape public opinion in the states, others look at the impact of state public opinion on politics and policy making in the states. A variety of measures of state public opinion are used including partisanship, global ideology, attitudes on particular issues, gubernatorial and state legislative approval, and trust in general and toward government in particular. New methodologies to measure state public opinion are presented (Chapter 1 by Park, Gelman, and Bafumi) and for the ﬁrst time in print, the dynamics of state public opinion are tracked (Chapter 12 by Erikson,Wright, and McIver). If Jewell’s exhortation a quarter of a century ago stimulated greater research eﬀort and attention to state politics, which seems to be the case— note the existence of a section of the American Political Science Association, the establishment of a dedicated journal, State Politics & Policy Quarterly, and an extensive recent literature review (Brace and Jewett, 1995)— then one hope of Public Opinion in State Politics is to stimulate continuing xiii

xiv

Preface

eﬀorts to study state-level public opinion, to ﬁnd new and better ways of measuring opinion at the state level, and increase the visibility of the study of state public opinion among political scientists, those who study state politics, and those interested in democratic politics and processes.

Sources for Preface Brace, Paul, and Aubrey Jewett. 1995.“The State of State Politics Research.” Political Research Quarterly 48 (September): 643–681. Jewell, Malcolm E. 1982. “The Neglected World of State Politics.” The Journal of Politics 44 (August): 638–657.

Acknowledgments

Any book, especially an edited one like this, is the product of many people. I want to thank the contributors to this volume.They agreed to come aboard before I had a contract in hand or even a publisher interested. In an age when the bottom line decides many publishing decisions, even for university presses, I greatly want to thank Stanford University Press for its willingness to publish this volume and for seeing the academic value of compiling the best in the way of research on state-level public opinion. Stanford’s editor, Amanda Moran, and the assistant editor, Jared Smith, ably steered us through the publication process, as did the rest of the production staﬀ. Special mention must be made of Amanda’s support and wise council on this project, which was also invaluable. I also want to thank Fordham University, especially my department chair, Bruce Berg; the dean of the faculty, Dominic Balestra; and the dean of the graduate school, Nancy Busch, for their personal support, a good working environment, ﬁnancial support, and release time to ﬁnish this manuscript. The Oﬃce of Research also provided some funding to help see this project through. Finally, I want to thank my wife for putting up with me with more good humor than I deserve.

xv

Contributors

Editor Jeﬀrey E. Cohen, Professor of Political Science, Fordham University

Contributors Joseph Bafumi, Assistant Professor of Government, Dartmouth University Charles J. Barrilleaux, Professor of Political Science, Florida State University Paul Brace, Clarence L. Carter Professor, Political Science, Rice University Robert S. Erikson, Professor of Political Science, Columbia University Andrew Gelman, Professor of Statistics and Political Science, Columbia University Donald Haider-Markel, Associate Professor of Political Science, University of Kansas John A. Hamman, Associate Professor of Political Science, Southern Illinois University William G. Jacoby, Professor of Political Science, Michigan State University Martin Johnson, Assistant Professor Department of Political Science University of California, Riverside Matthew S. Kaufman, planner at the Denver Regional Council of Governments; MA in Urban Planning, University of Kansas. James D. King, Professor of Political Science, University of Wyoming David L. Leal, Associate Professor of Political Science, University of Texas at Austin John McIver, Associate Professor of Political Science, University of Colorado Barbara Norrander, Professor of Political Science, University of Arizona xvii

xviii

Contributors

David K. Park, Assistant Professor of Political Science,Washington University, St. Louis Saundra K. Schneider, Professor of Political Science, Michigan State University Eric M. Uslaner, Professor of Political Science, University of Maryland Clyde Wilcox, Professor of Government, Georgetown University Gerald C.Wright, Jr., Professor of Political Science, Indiana University

Public Opinion in State Politics

Chapter

Introduction: Studying Public Opinion in the American States

1

Jeffrey E. Cohen

Scholarly interest in state-level public opinion has grown in recent years. New data collections, methodologies, and theoretical approaches have all stimulated interest. One indication of this growing interest was the short course, “The Uses of Public Opinion Data in State Politics and Policy Research,” held at the 2003 American Political Science Association. Many of the participants at that short course are represented in these pages (e.g., Paul Brace, Robert Erikson, John McIver, Barbara Norrander, and Gerald Wright). Some may even like to think of state public opinion as an emergent subﬁeld in political science. Yet since the publication of Erikson, Wright, and McIver’s seminal Statehouse Democracy (1993) over a decade ago, no major book-length study of public opinion in the American states has been published. Public Opinion in State Politics provides the ﬁrst book on the topic in a decade in hopes of stimulating research on state public opinion by collecting in one convenient place some of the best recent research on the topic. Another aim of Public Opinion in State Politics is to increase the accessibility of work in the subﬁeld to scholars beyond those doing research on state public opinion. The issues and questions that state public opinion scholars deal with are relevant to those with interests in public opinion, public policy making, democratic theory and representation, and political development, among other subﬁelds. The authors of the chapters in Public Opinion in State Politics present a 3

4

Cohen

mix of senior scholars who have already made major contributions to the topic, as well as younger scholars who bring new insights and issues for study.The contributors oﬀer a variety of data, analytic methodologies, and substantive concerns. Despite the diversity across the chapters collected for this volume, several themes unify them. The ﬁrst theme that underlies the studies in this book is a concern for the role of public opinion in democratic politics, a fundamental question for democratic theory. Almost all theories of democracy require some minimal level of governmental responsiveness to public opinion in order to say that a polity is democratic. Do policy makers incorporate public opinion when making policy decisions? Can policy makers and other political leaders shape public opinion? These are some of the questions that the chapters in this volume address, but they do so by looking at the American states as the units of analysis. The focus on states as units of analysis represents a departure from the mode of much research on public opinion. Most public opinion research focuses on the individual, asking questions such as what opinions do people hold, how do they arrive at those opinions, or how stable and sophisticated are people’s opinions about politics? To study these questions, public opinion research relies heavily on surveys and more recently on experiments. However important the survey research brand of public opinion research is to our understanding of mass political opinion, it does not directly address many of the important questions of democratic theory. Democratic responsiveness and representation are fundamentally aggregate level processes. Policy makers respond and anticipate the preferences and reactions of groups of people, not individuals. Although an understanding of individual level properties of public opinion may provide a foundation for understanding public opinion in the aggregate (Page and Shapiro, 1992; Erikson, MacKuen, Stimson, 2002), understanding the quality of democracy requires an aggregate perspective, one that links the public opinion of groups of individuals, such as district constituents or blocks of voters, to the processes of government and policy making. States as analytic units serve this purpose well. All of the chapters in this volume treat public opinion as an aggregate phenomenon, although several chapters must look at individual-level opinion because of the particular question posed or because of data limitations (cf., Chapters 4 and 5). A second theme that ties the studies in this book concerns the increased importance of states as policy-making units. During the middle third of the twentieth century, the national government increasingly assumed responsibility for public policy, a function of the expansion of

Introduction 5

government into new policy realms, as well as national assumption of policy that was once the responsibility of the states and localities. During this era of national government policy expansion, the federal government often directed and mandated the shape of many state and locality policies. But for the past quarter century, responsibility for some policies has devolved from the national to the state governments, such as welfare, and states have increased their policy-making eﬀorts for other policy areas, such as economic development. The increased autonomy and scope of states as policy makers suggests the importance of studying the states. Third, the chapters in this volume exploit the states to study the linkages between public opinion, policy making, and democratic politics. Comparative state politics has long recognized the utility of making comparisons across the 50 states for building theory and testing hypotheses.All of the states possess a Madisonian political structure composed of checks and balances and separation of powers and they share in the same basic political culture.1 With these two characteristics of political systems held constant, one may be better able to isolate the eﬀects of other factors on policy making and democratic processes, such as public opinion.

State Politics and State Public Opinion Research The comparative study of state politics and policy making dates to V. O. Key’s magisterial study, Southern Politics, published in 1949, if not earlier. In the late 1960s and early 1970s the comparative study of state politics hit its high-water mark, with a stream of inﬂuential studies by scholars such as Richard E. Dawson and James A. Robinson (1963), Thomas Dye (1965, 1969a, 1969b), Richard Hoﬀerbert (1966), and Ira Sharkansky (1967, and with Hoﬀerbert, 1969), among others. But by the late 1970s, the excitement in pursuing comparative state studies began to fade, although many important studies were still to be published. Comparative state politics and policy was no longer the cutting edge topic it was a decade earlier (for a review of the state politics and policy literature see Brace and Jewett, 1995). One factor that depressed enthusiasm for comparative state studies was the lack of public opinion data across the states. Scholars in the 1960s and early 1970s developed many useful indicators of many aspects of state politics and policy processes, including aspects of policy outputs, political arrangements and structures, and state demographic and economic proﬁles. But without comparable data on public opinion, the comparative study of state politics and policies stalled. An important element of politics and policy-making processes, public opinion, was missing, and devel-

6

Cohen

oping public opinion measures for the states seemed unobtainable. Conducting surveys of public opinion at the time seemed prohibitively expensive, especially to academically based polling organizations, and organizations with the resources to conduct such studies seemed uninterested in tapping the opinions and attitudes of citizens in the states. Facing such a barrier to research progress, scholars interested in democratic theory and politics shifted to other research arenas where they felt they could make more progress.

Studying Public Opinion in the States The major complaint of the earlier generation of scholars about the absence of good comparative data on state public opinion can no longer be voiced so loudly, although as the studies in this volume demonstrate, gaping holes in our data on state public opinion still exist. Good data on some aspects of state-level public opinion, often for all 50 states, now exist; the major techniques and methodologies for deriving public opinion estimates are discussed below. They include pooling national surveys, conducting national surveys with states as important subunits of analysis, combining independent state-based surveys, simulating state opinion, and using individual-level national surveys to learn about individual opinions on state-level politics, policies, and issues. Each has its strengths as a way to gauge state-level opinion, as well as limitations.

Pooling National Surveys Erikson, McIver, and Wright (1987, 1989, 1993) and Wright, Erikson, and McIver (1985, 1987) pioneered the methodology of pooling national surveys to construct state-level opinion measures. Using CBS News/New York Times polls, they estimate the partisan and ideological make up of the states, two important global political orientations.The simplicity and elegance of this method led others to employ it to elicit other state-level opinion measures, most notably Brace and colleagues (2002, 2004), who employ the General Social Survey (GSS) to generate measures of state opinion on a range of speciﬁc issues, as well as trust toward federal government, as Brace and Johnson do in this volume (Chapter 2). Uslaner (Chapter 8) follows this path to construct a state-level measure of citizen trust. Several complaints have been leveled against this approach. In using national surveys, the number of cases per state reﬂects the relative population of each state. It can be diﬃcult to generate enough cases for reliable estimates from smaller states. To allow for reliable estimates for smaller

Introduction 7

states thus often requires gathering polls across relatively long periods of time, unless many national surveys exist over a short time period. When collapsing polls across relatively long time periods, one must assume stability in attitude under study at least for the length of the time frame used to generate the state estimates. In Erikson, Wright, and McIver’s initial 1985 report, this was a 7-year period (1976–1982), in their latter report (1993), the time frame is 13 years, and for Brace and colleagues the period lengthens at times to over 20 years (1974–1998). Second, national surveys do not use states as units from which to draw their national sample ( Jones and Norrander, 1996).The resulting variability of state estimates may not resemble the true variability in the states.

National Surveys with States Subunits of Analysis In response to the second critique noted above, Norrander and colleagues (2000, 2001; with Jones, 1996 and with Wilcox, 1999, and Chapter 3 in this volume) turned to the 1988–1990–1992 National Election Studies (NES), which used states as the primary sampling units to allow analysis of U.S. Senate elections. Two fundamental problems plague this approach. One, as NES has not replicated this design of using the states as primary sampling units, nor has any other national survey organization, the estimates of state opinion are becoming quite dated.Two, respondents were generally asked about national issues, not policies or issues speciﬁc to state policy making or politics, although national and state-level concerns often overlap.Those issues that fail to register on the national agenda or do not rise high on the national agenda, but appear on many state agendas, are absent but may be of vital importance in understanding the politics and policy making of state governments.

Combining Independent State-based Surveys A recent project by Thad Beyle, Richard Niemi, and Lee Sigelman (2002a, 2002b) has collected state-level popularity for the president, governor, and U.S. senators and provides us with another perspective on state public opinion—how state citizens view major political ﬁgures.2 These measures of state opinion come from state-level surveys, but across a variety of polling ﬁrms. Measures of state approval date to 1945 for presidents, 1958 for governors, and 1978 for senators, but most of the public approval readings are more recent—since the 1990s—when an explosion in opinion polls occurred.At this writing these data have been updated through early 2005. As a compilation from many survey organizations, including academic,

8

Cohen

commercial, and newspaper polls, one must be careful when making comparisons across the surveys.The sources of noncomparability that may exist include diﬀerent question wordings, diﬀerent response categories, diﬀerent placement within the survey protocol, house eﬀects, diﬀerent sampling frames, diﬀerent survey designs, diﬀerent interviewing techniques, and so on. Furthermore, not all of the polls went into the ﬁeld at the same time. Thus, we have opinion readings for some states at some time points but not for others. Cohen and King in this volume use these data to eﬀect in their study of the factors that aﬀect gubernatorial approval and oﬀer an extended discussion and example for comparing these popularity measures across the states (also see Beyle, Niemi, and Sigelman, 2002a, 2002b). These data have proved quite popular in recent years. The Fall 2002 issue of State Politics & Policy Quarterly was devoted entirely to studies using these data (Anderson and Newmark, 2002; Barth and Ferguson, 2002; Beyle, Niemi, Sigelman, 2002a, 2002b; Crew, Branham,Weiher, and Bernick, 2002; and Dometrius, 2002). Other recent studies utilizing these data include Bardwell (2003, 2005), Cohen (2003), Cohen and Powell (2005), Ferguson (2003), and Kang, Niemi, and Powell (2003). The presence and use of these data reveal another gap in our knowledge of state public opinion—we lack good data on state opinion about other political leaders and institutions, such as the state legislature and state courts. Hamman in this volume grapples with studying public opinion toward state legislatures in the face of highly limited data.

Simulating State Public Opinion Simulating opinion is the most venerable tradition in the study of state public opinion. The inability of collecting public opinion data across enough states to allow comparative state studies led ﬁrst to the use of demographic and socioeconomic data as surrogates for public opinion. Using demographics and socioeconomics as surrogate measures of opinion assumed that opinion holding within a group was reasonably uniform about the issue in question and that these social characteristics largely determined political preferences. Some scholars felt that the surrogate demographic approach to tapping state opinion was too blunt. In a series of reﬁnements, Pool, Abelson, and Popkin (1965) and Weber and associates (Weber, Hopkins, Mezey, and Munger 1972–1973;Weber and Shafer, 1972), still relied on demographic and socioeconomic variables, but tied them to opinion questions.With the combination of demographic categories and opinions of members of those categories, one could simulate state opinion if one knew the size of

Introduction 9

the group within a state. Still, such measures were at best indirect indicators of state-level public opinion. The assumptions that underlie simulation from demographics are open to challenge. First, these simulations assumed that the distribution of preferences of a group would not vary much across states, that, for instance, blacks or Latinos in New York would think similarly to blacks or Latinos in Mississippi. Leal, in Chapter 4 in this volume, tests for cross-state variation in the political attitudes of Latinos, ﬁnding signiﬁcant diﬀerences in the attitudes of Latinos from Cuba, from those in Texas, from those in California.The ﬁnding of Norrander and Wilcox in this volume that state residence may aﬀect public attitudes severely tests that assumption of similarity in attitude across states of people of the same demographic category. Seidman’s 1973 critique of simulations of state opinion led to the death of that approach shortly thereafter. First, the simulation technique then in use assumed independence across the diﬀerent demographic categories used to simulate opinion, but interaction eﬀects might at times aﬀect opinion. For instance, the interaction of race and region may lead southern and northern blacks (and whites) to hold diﬀerent political opinions (see Valentino and Sears, 2005, on diﬀerences between northern and southern whites). Second, the simulation depended on assuming that demographics not employed in the simulations are not relevant to the state estimates. However, typically only a handful of demographics are available, such as region, occupation, race, and education. Similarly, the model assumed that any two states are alike except for the diﬀerences in their demographic proﬁles. Other factors that may aﬀect opinion independent of demography (e.g., political leadership, media reporting, and variation in the economic cycle, for instance) are assumed to lack impact, a dubious assumption at best. A revival (or resurrection) of simulating state opinion has occurred in recent years. In this volume, Park, Gelman, and Bafumi (Chapter 11) present a new simulation approach using Bayesian statistics and hierarchical linear modeling. By controlling for place (state) they are able in their simulations, which also rely on demographics, to reduce the error variance and thus produce somewhat more precise estimates. Berkman and Plutzer (2004, forthcoming) adapt the Park-Gelman-Bafumi method to estimate the opinion in school districts units. Ardoin and Garand (2003) use a method similar to Berkman and Plutzer, what they call “top-down” simulation, to generate estimates of congressional district ideology.To the mix of demographics used to create these estimates, Ardoin and Garand add presidential election results at the congressional district level, which have been shown to be highly related in some elections to ideology. Still, these

10

Cohen

new simulation approaches are open to some of the charges made against the older variety, especially that the opinions of demographic groups do not vary across the states.

An Old Standby: National Surveys on State-Level Topics Although we now possess state-level opinion data for a much larger range of attitudes than was the case two decades ago, several gaps still exist. As noted above, we know little about issues that states grapple with but do not percolate to the national agenda. Also, we know little about public attitudes toward state leaders—even the impressive JAR ( Job Approval Ratings) data compendium is more about state-level attitudes to those who hold federal oﬃce (president, U.S. senator) than state oﬃce (governor).We know precious little about attitudes, for instance, about state legislators, state judiciaries, and other state executives. Occasionally, however, national surveys ask respondents about these state leaders. For instance, in 1990 and 1991 the ABC/Washington Post polls asked respondents whether they approved or disapproved of the job that their state legislature was doing. Presumably, other national surveys have asked about other aspects of state politics. Hamman, in Chapter 5, analyzes these state legislative data, which require extraordinary care.With such surveys we get a portrait of how the people in the nation feel about state leaders, but we can say little about state publics. Moreover, characteristics of state legislatures vary by state size—large state legislatures tend to be more professional and often have larger district sizes than small state legislatures.The level of state legislative professionalism and district size may aﬀect public attitudes toward state legislatures, but unless we control for this covariance between state size, state legislative characteristics, and state opinion, the attitudes of those in large states may swamp that of those living in smaller states. Admittedly, then, the data on state public opinion that now exist have their limitations. Many of the chapters in this volume speak to the limitations of existing state public opinion data.Yet compared to the situation that scholars of the 1960s through 1980s faced, even the limited data currently available represent a quantum improvement and allow scholars to address questions only dreamed about a generation ago. The chapters in this volume represent among the best and most exciting work on the topic of the role of public opinion in state politics and policy making, and by implication, in democratic processes. And as all good research, the chapters in this volume raise many new questions.

Introduction 11

A Preview of the Chapters I have organized the chapters in this book into three sections, although many of the chapters can easily fall into more than one of these categories.The ﬁrst deals with factors that aﬀect state-level opinion, while the second section looks at the impact of state opinion on state politics and policy making. The ﬁnal section oﬀers directions for future research, although one could say that about the other chapters as well. In section three, Park, Gelman, and Bafumi present their new simulation methodology and Erikson,Wright, and McIver update their state ideology and partisanship data to account for change in state-level opinion, a research direction that one previously could only dream of. In the ﬁnal chapter of this book, I assess where the studies in this volume have taken us and where we need to go. In the ﬁrst part of this book, several authors deal with factors that shape state-level opinion. Paul Brace and Martin Johnson, in Chapter 2, “Does Familiarity Breed Contempt? Examining the Correlates of State-Level Conﬁdence in the Federal Government,” ask whether the federal context aﬀects attitudes to national government across state publics. Using the methodology developed in Brace, Sims-Butler, Arceneaux, and Johnson (2002), Brace and Johnson pool General Social Surveys from 1975 to 1998 to develop measures of conﬁdence in the federal government. They ﬁnd that the presence of the federal government in a state aﬀects the way a state’s citizens view the national government: when the federal government owns more land within the state, citizens view the federal government with less conﬁdence, but as federal employment increases, so does conﬁdence in the national government. Their results echo Cohen and King, ﬁnding that the national context seeps into and helps shape state public opinion. In Chapter 3, “State Residency, State Laws, and Public Opinion,” Barbara Norrander and Clyde Wilcox ask whether state residence aﬀects public opinion, a question that contrasts but also complements the question of federal impacts that Brace and Johnson pose. From a political culture argument, states can be seen as relatively autonomous political units, with distinctive sets of public policies and political traditions. Does the environment of the state aﬀect a person’s attitude toward public policies? Using the pooled Senate National Election Surveys (SNES), which provide Norrander and Wilcox with a representative sample for all 50 states, they ask whether variation in abortion laws across the states aﬀect attitudes toward abortion. Based on their estimates, Norrander and Wilcox

12

Cohen

calculate that state residence accounts for from 10 to 25 percent of the variance in abortion opinion. People who live in more conservative abortion policy states will tend to hold more conservative opinion than those who live in more liberal abortion policy states. David L. Leal also addresses the question of whether state residence aﬀects the attitudes of its citizens in Chapter 4, “Mexican-American and Cuban-American Public Opinion: Diﬀerences at the State Level?” Leal focuses on the fastest growing and now largest minority group in the United States, Hispanics. Much research treats Hispanics as a homogenous group when it comes to politics, but Leal notes that in California, Hispanic political activity and opinion tends to be more liberal than that of Hispanics who reside in Texas. Besides the obvious diﬀerences in national origin of Hispanics in the United States, does state of residence help account for any diﬀerences in Hispanic political opinion? Leal ﬁnds that MexicanAmericans in Texas and California, in a multivariate analysis, actually appear more similar than diﬀerent, which calls into question the impact of state residence on citizen opinion. Taking Leal and Norrander and Wilcox together suggests that we need to better understand the conditions on which state residence aﬀects citizen behavior and opinion. Leal also shows the utility of using a small number of states for comparative analysis. John A. Hamman turns to an often forgotten state policy-making institution, the state legislature, in his contribution, Chapter 5, “Public Opinion in the States: Determinants of Legislative Job Performance.” Hamman’s study is hampered by the dearth of cross-state data on public evaluations of the state legislature. Scouring the National Network of State Polls, Hamman found 124 surveys across 13 states that asked respondents about their attitudes toward the state legislature. The limited number of states and the spottiness of the data preclude much analysis. Then Hamman turns to two national polls in 1990 and 1991 that ask about state legislative performance.Thus, unlike most of the other studies in this volume, this part of Hamman’s analysis looks at individual-level attitudes. Hamman ﬁnds, consistent with other studies, that the more professional the state legislature, the lower its job performance ratings. This is indeed an ironic result, which calls for more research and the need for more cross-state data on public evaluations of the state legislature. Touching on federalism themes (also see Brace and Johnson above), Jeﬀrey E. Cohen and James D. King, in Chapter 6, “The State Economy, the National Economy, and Gubernatorial Popularity,” compare the impact of the national and state economies on public attitudes toward the governor. Governors often tout their eﬀorts to bring jobs into their states. Thus, it makes sense to ask whether a state’s citizens hold their governor

Introduction 13

accountable for the state’s job climate. But national economic factors also impinge on state economies. Given the large impact of the national economy, will the public still hold the governor accountable for the performance of the state economy? Cohen and King use the recently released Oﬃcial Job Approval Rating database (Beyle, Niemi, and Sigelman, 2002a and 2002b), which provides them with approximately 2,000 monthly gubernatorial popularity observations across a 20-year period. The JAR data set allows them to employ a modiﬁed pooled cross-sectional time series design: They ﬁnd that both national and state economic factors aﬀect state public opinion toward the governor. Chapters in the second section of this book look at the consequences of state public opinion on state government and policy. Most studies of state-level opinion look at opinion in the aggregate. In so doing, they tend to rely on measures of central tendency, like means or the proportions of the public that hold a particular opinion. Charles J. Barrilleaux takes a diﬀerent tack in Chapter 7, “Ideological Cleavage, Political Competition, and Policy Making in the American States,” by focusing on the spread or diversity of opinion. Barrilleaux makes the case that states with homogeneous opinion present diﬀerent information and constraints to policy makers compared to states with heterogeneous or dispersed opinion, that is, when opinion is spread widely around its central tendency, even if the compared states have the same mean opinion. Using the Erikson,Wright, and McIver data on state ideology, he ﬁnds that more dispersed or diverse opinion weakens the impact of ideology on state policy outputs. Barrilleaux’s chapter reminds us that single measures cannot fully describe the complexity of aggregate state opinion. Eric M. Uslaner, in Chapter 8, continues the theme of the linkage between a state’s citizens and the outer environment in his contribution, “The Civil State: Trust, Polarization, and the Quality of State Government.” He asks whether trust toward others in the mass public aﬀects the quality of governmental performance. Besides creating a state-level measure of public trust toward others, Uslaner’s other major innovations include treating trust as an independent variable, something that may aﬀect governmental performance, as opposed to treating trust and social capital as something to be explained. Uslaner ﬁnds that the quality of government performance improves for most of his measures when mass trust is higher, suggesting an important linkage between the way that the public thinks about others and the ability of government to deliver services and policies. Like Norrander and Wilcox, Donald Haider-Markel and Matthew S. Kaufman deal with social policies in Chapter 9, “Public Opinion and

14

Cohen

Policy Making in the Culture Wars: Is There a Connection Between Opinion and State Policy on Gay and Lesbian Issues?” Haider-Markel and Kaufman collect a number of measures of state attitudes toward gay and lesbian issues and ask whether such opinions aﬀect state policies toward gays and lesbians. They ﬁnd that in some instances public opinion aﬀects state policies, but in other instances, it does not. Haider-Markel and Kaufman hypothesize that when opinion is divided and contentious, policy makers will be more sensitive to public opinion in building policy. But when opinion is consensual or lopsided, to use their term, other factors, such as policy entrepreneurs, will have a larger impact on policy. The Haider-Markel and Kaufman chapter is a big step forward in identifying the conditions when public opinion will aﬀect policy. Similar to Haider-Markel and Kaufman, Saundra K. Schneider and William G. Jacoby also look at the impact of state public opinion on policy making in Chapter 10, “Citizen Inﬂuences on State Policy Priorities: The Interplay of Public Opinion and Interest Groups.” But Schneider and Jacoby ask about the relative importance of public opinion and interest groups, an enduring question of democratic responsiveness. Beyond comparing the impact of interest groups and public opinion on public policy, Schneider and Jacoby oﬀer a new measure of state public policy.They create a measure of state policy priorities, which makes comparisons of policy spending levels across all 10 major budget categories, using spatial proximity modeling, a form of unfolding. Based on their new policy indicator, Schneider and Jacoby suggest that policies may be either collective goods oriented or more particularistic. The impact of public opinion varies across the two types of policies, with stronger direct eﬀects found for collective goods. But in an important reﬁnement in our understanding of the comparative impact of interest groups and public opinion, Schneider and Jacoby argue that public opinion indirectly aﬀects both collective goods and particularistic policy through the impact of public opinion on interest group formation.Thus, the relationship between public opinion, interest groups, and policy is quite complex and conditional; the conventional view that public opinion and interest groups compete in the policy-making process is too simplistic. In Chapter 11, “State-Level Opinions from National Surveys: Poststratiﬁcation Using Multilevel Logistic Regression,” David K. Park, Andrew Gelman, and Joseph Bafumi critique the Erikson-Wright-McIver methodology and return to the older simulation methodology, but with signiﬁcant reﬁnements. To Park, Gelman, and Bafumi, a major limitation of the Erikson-Wright-McIver methodology is its insensitivity to shortterm change in opinion. Park, Gelman, and Bafumi return to simulation,

Introduction 15

but increase the number of categories compared to earlier research (Weber et al., 1972–1973;Weber and Schaﬀer, 1972) and employ multilevel logistic regression to estimate state opinion. As a result, Park, Gelman, and Bafumi (2003) are able to estimate state-level opinion in more reﬁned time units than Erikson,Wright, and McIver (1993) or Brace et al. (2002). Their method produces quite precise estimates of diﬀerent examples of public opinion, opening up a new avenue for capturing dynamic properties in estimating state-level public opinion. Robert Erikson, Gerald Wright, and John McIver, in Chapter 12,“Public Opinion in the States: A Quarter Century of Change and Stability,” update their seminal work on state mass partisanship and ideology, asking whether the patterns observed through the 1980s, the end point of their original study, still hold as we enter the twenty-ﬁrst century. They ﬁnd strong stability over time in mass ideology, and while mass partisanship is also highly stable, patterns of change are evident in the partisanship data, notably the conversion of the once rock-solid Democratic south into a Republican-advantaged region. Erikson,Wright, and McIver also ﬁnd that ideology and partisanship are strongly correlated in the 1990s, unlike their lack of association in 1970s and 1980s data. According to their analysis, over-time partisanship realigned to converge with the ideological predispositions of states’ mass publics.

Notes 1. However, some scholars have noted variation in political culture across the states, notably Daniel Elazar (1966), who distinguished among individualistic, moralistic, and traditionalistic political cultures. Elazar’s venerable typology still ﬁnds it way into current research; for instance, see the chapter by Schneider and Jacoby in this volume. 2. These data can be accessed from the website, http://www.unc.edu/~beyle/jars .html.

References Anderson, Jennifer L., and Adam J. Newmark. 2002. “A Dynamic Model of U.S. Senator Approval, 1981–2002.” State Politics & Policy Quarterly 2 (Fall): 298– 318. Ardoin, Phillip J., and James C. Garand. 2003.“Measuring Constituency Ideology in U.S. House Districts: A Top-Down Simulation Approach.” Journal of Politics 65 (November): 1165–1191. Bardwell, Kedron. 2003. “Not All Money Is Equal: The Diﬀerential Eﬀect of Spending by Incumbents and Challengers in Gubernatorial Primaries.” State Politics & Policy Quarterly 3 (Fall): 294–308.

16

Cohen

———. 2005. “Reevaluating Spending in Gubernatorial Races: Job Approval as a Baseline for Spending Eﬀects.” Political Research Quarterly 58 (No. 1): 97–106. Barth, Jay, and Margaret R. Ferguson. 2002.“American Governors and Their Constituents:The Relationship between Gubernatorial Personality and Public Approval.” State Politics & Policy Quarterly 2 (Fall): 268–284. Berkman, Michael, and Eric Plutzer. 2004. “School Budgets and Local Teacher’s Unions:The Mediating Role of Political.” Paper presented at the annual meeting of the American Political Science Association Hilton Chicago and the Palmer House Hilton, Chicago, IL. ———. Forthcoming. Ten Thousand Democracies: Politics and Public Opinion in America’s School Districts.Washington, DC: Georgetown University Press. Beyle, Thad, Richard G. Niemi, and Lee Sigelman. 2002a. “Approval Ratings of Public Oﬃcials in the American States: Causes and Eﬀects.” State Politics & Policy Quarterly 2 (Fall): 213–216. ———. 2002b.“Gubernatorial, Senatorial, and State-level Presidential Job Approval: The U.S. Oﬃcials Job Approval Ratings ( JAR) Collection.” State Politics & Policy Quarterly 2 (Fall): 215–231. Brace, Paul, Kevin Arceneaux, Martin Johnson, and Stacy G. Ulbig. 2004. “Does State Ideology Change over Time?” Political Research Quarterly 57 (December): 529–540. Brace, Paul, and Aubrey Jewett. 1995.“The State of State Politics Research.” Political Research Quarterly 48 (September): 643–681. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using New Data.” American Journal of Political Science 46 ( January): 173–189. Cohen, Jeﬀrey E. 2003. “The Polls: State-level Presidential Approval: Results from the Job Approval Project.” Presidential Studies Quarterly 33 (March): 211–220. Cohen, Jeﬀrey E., and Richard J. Powell. 2005.“Building Public Support from the Grassroots Up: The Impact of Presidential Travel on State-Level Approval.” Presidential Studies Quarterly 35 (March): 11–27. Crew, Robert E., Jr., David Branham, Gregory R. Weiher, and Ethan Bernick. 2002. “Political Events in a Model of Gubernatorial Approval.” State Politics & Policy Quarterly 2 (Fall): 283–299. Dawson, Richard E., and James A. Robinson. 1963. “Inter-Party Competition, Economic Variables, and Welfare Policies in the American States.” Journal of Politics 25 (May): 265–289. Dometrius, Nelson C. 2002. “Gubernatorial Approval and Administrative Inﬂuence.” State Politics & Policy Quarterly 2(Fall): 251–268. Dye, Thomas R. 1965. “Malapportionment and Public Policy in the States.” Journal of Politics 27 (August): 586–601. ———. 1969a. “Income Inequality and American State Politics.”American Political Science Review 63 (March): 157–162. ———. 1969b. “Executive Power and Public Policy in the States.” Western Political Quarterly 22 (December): 926–939. Elazar, Daniel. 1966. American Federalism:A View from the States. New York:Thomas Y. Crowell.

Introduction 17 Erikson, Robert S., Michael B. MacKuen, and James A. Stimson. 2002. The Macro Polity. New York: Cambridge University Press. Erikson, Robert S., John P. McIver, and Gerald C.Wright, Jr. 1987.“State Political Culture and Public Opinion.” American Political Science Review 81 (September): 797–813. Erikson, Robert S., Gerald C.Wright, Jr., and John P. McIver. 1989.“Political Parties, Public Opinion, and State Policy in the United States.” American Political Science Review 83(3): 729–750. ———. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Ferguson, Margaret Robertson. 2003. “Chief Executive Success in the Legislative Arena.” State Politics & Policy Quarterly 3 (No. 2): 158–182. Hoﬀerbert, Richard I. 1966. “The Relation between Public Policy and Some Structural and Environmental Variables in the American States.” American Political Science Review 60 (March): 73–82. Jones, Bradford S., and Barbara Norrander. 1996. “The Reliability of Aggregated Public Opinion Measures.” American Journal of Political Science 40 (February): 295–309. Kang, Insun, Richard G. Niemi, and Lynda W. Powell. 2003. “Strategic Candidate Decisionmaking and Competition in Gubernatorial Nonincumbent-Party Primaries.” State Politics & Policy Quarterly 3 (Winter): 353–366. Key,V. O. 1949. Southern Politics in State and Nation. New York: Knopf. Norrander, Barbara. 2000.“The Multi-Layered Impact of Public Opinion on Capital Punishment Implementation in the American States.” Political Research Quarterly 53 (December): 771–793. Norrander, Barbara. 2001.“Measuring State Public Opinion with the Senate National Election Study.” State Politics & Policy Quarterly 1 (March): 113–127. Norrander, Barbara, and Clyde Wilcox. 1999. “Public Opinion and Policymaking in the States: The Case of Post-Roe Abortion Policy.” Policy Studies Journal 27 (December): 707–722. Page, Benjamin I., and Robert Y. Shapiro. 1992. The Rational Public: Fifty Years of Trends in Americans’ Policy Preferences. Chicago: University of Chicago Press. Park, David K., Joseph Bafumi, and Andrew Gelman. 2003. “Bayesian Multilevel Estimation with Poststratiﬁcation: State-Level Estimates from National Polls.” Political Analysis 12 (Autumn): 375–385. Pool, Ithiel de Sola, Robert P.Abelson, and Samuel L. Popkin. 1965. Candidates, Issues, and Strategies. Cambridge, MA: MIT Press. Seidman, David. 1973. “Simulation of Public Opinion: A Caveat.” Public Opinion Quarterly 39 (Autumn): 331–342. Sharkansky, Ira. 1967. “Government Expenditures and Public Services in the American States.” American Political Science Review 61 (December):1066 – 1077. Sharkansky, Ira, and Richard I. Hoﬀerbert. 1969. “Dimensions of State Politics, Economics, and Public Policy.” American Political Science Review 63 (September): 867–879. Valentino, Nicholas A., and David O. Sears. 2005.“Old Times There Are Not For-

18

Cohen

gotten: Race and Partisan Realignment in the Contemporary South.” American Journal of Political Science 49 ( July): 672–688. Weber, Ronald E.,Anne H. Hopkins, Michael L. Mezey, and Frank Munger. 1972– 1973.“Computer Simulation of State Electorates.” Public Opinion Quarterly 36: 49–65. Weber, Ronald E., and William R. Shaﬀer. 1972. “Public Opinion and American State Policy-Making.” Midwest Journal of Political Science 16(4): 683–699. Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship and Ideology with Survey Data.” Journal of Politics 47 ( June): 469–489. ———. 1987.“Public Opinion and Policy Liberalism in the American States.” American Journal of Political Science 31(3): 980–1007.

Chapter

Does Familiarity Breed Contempt? Examining the Correlates of State-Level Conﬁdence in the Federal Government

2

Paul Brace and Martin Johnson

Federalism is a central aspect of American government. The original justiﬁcation for the federal form of government was to preserve personal liberty by dividing the power of the state. Madison argued in Federalist 51 that two levels of government could control each other while controlling themselves. Early federalism was built on the principle of dual sovereignty but since the end of the Civil War sovereignty has tended to concentrate in the national government. Nonetheless states remain integral to American government, electing oﬃcials, taxing, and spending for a variety of functions. Modern federalism has grown complex, with overlapping authorities and signiﬁcant funding ﬂowing from the national government to subnational units. As power further accumulated at the federal level during the 1960s and early 1970s, the American public appears to have developed an increasingly negative view of the national political system. While Miller (1974) and Citrin (1974) debated the meaning of this precipitous drop in public trust in government, they agreed a sizeable shift in public sentiment occurred. The decline to a lower level of support has been stable, furthermore, and many observers of the contemporary political scene describe a “crisis in conﬁdence” (e.g., Dionne, 1991; Hibbing and Theiss-Morse, 1995). Gallup periodically asks “How much of the time do you think you can trust the government in Washington to do what is right: just about always, most of 19

20

Brace and Johnson

the time, or only some of the time?”Typically somewhere between 50 and 60 percent of those surveyed express a lack of conﬁdence, saying that they trust the government in Washington only some of the time or never. While an impressive body of research considers alternative explanations for the relatively low level of government trust among individuals (Hibbing and Theiss-Morse, 1995) and the decline in trust observed three decades ago at the national level (Citrin, 1974; Miller, 1974), we have not considered how trust or conﬁdence in the federal government is distributed across the United States. This disjunction between concerns about trust and interests in federalism is ironic given that federalism’s Latin root fœdus, a covenant, shares its etymological lineage with the verb ﬁdere, meaning “to trust” and underlying English words like conﬁde. When we speak about federalism we are quite literally invoking trust or conﬁdence. Unfortunately, contemporary research on U.S. federalism has ignored this dimension just as students of trust or conﬁdence in government have largely ignored the institution of federalism (but see Bowler and Donovan, 2002; Hetherington and Nugent, 2001; Ulbig,Wade, and Martorano, 2004). In this chapter, we ask how conﬁdence in the federal government varies among state populations. Like other contributors to this volume, we use national opinion data to estimate state-level opinions. Our focus is on whether and how opinions regarding trust in the federal government vary among the states. These opinions can illustrate the degree of national integration in the United States and reveal conditions where support is more or less forthcoming.We begin by considering how questions of public trust, conﬁdence, and legitimacy have been approached in past studies.We then present a measure of public trust in the national government at the state level, using well-established methods. We are also interested in identifying correlates of conﬁdence in the federal government, focusing on the socio-demographic characteristics of states, the presence of federal government operations in states, and several political factors including mass ideology and political engagement.We conclude by reﬂecting on the implications of the sources of state-level support for the national government.

Public Conﬁdence in Systems of Federal Governance In a federal system, the distribution of conﬁdence in the national government could have important consequences for governing. Pronounced rivalries among states could result in federal policies that promote fragmentation. Prior to the Civil War, southerners believed tariﬀs favored by northern states had a negative eﬀect on their economies.These economic

Correlates of State-Level Confidence in the Federal Government 21

hostilities ultimately contributed to secession and armed conﬂict. Antebellum politics was dominated by intense sectional rivalries over policies with distinct regional advantages and disadvantages. In contemporary politics, such sectional rivalries are less evident and levels of trust may be more uniformly distributed across the country.1 Fragmentation and even disintegration of federal systems is not uncommon. In Weimar Germany, the Soviet Union, Canada, and even the United States, sectional pressures inhibited national integration and promoted fragmentation and even national disintegration. Despite similar patterns of social and economic development, not all modern nations display uniformly high levels of national integration. Canada, with literacy and communication levels parallel to those in the United States, exhibits resilient levels of sectionalism (Kornberg, 1990; Kornberg, Clarke, and Leduc, 1978; Kornberg, Clarke, and Stewart, 1979; but see Atkinson, Coleman, and Lewis, 1980). Residents of British Columbia and Quebec commonly express signiﬁcantly less support for the Canadian national government than residents of other provinces, and these diﬀerences play a substantial role in Canadian politics. The topic of national integration has been a mainstay of cross-national comparative research. Robert Jackman observes that “the process of national integration, assisted by such factors as the growth of levels of mass literacy and of means of communication, is usually described as one where political allegiances are focused increasingly at the national level over time, while there is a corresponding decrease in the salience of local or more parochial allegiances” ( Jackman, 1972, 512). While developed nations commonly exhibit greater national integration, general levels of satisfaction with government vary substantially across Western European democracies (Anderson and Guillory, 1998). Likewise, levels of support for speciﬁc institutions also vary dramatically in Western Europe (Gibson, Caldeira, and Baird, 1998). The conditions that promote or inhibit positive public sentiments about national governments or their institutions are of considerable interest to students of national and cross-national politics. As noted, in the study of politics within the United States, there has not been a parallel interest in this topic at the subnational level. In their inﬂuential study of satisfaction with government, Hibbing and Theiss-Morse (1995) alert us to diﬀerential support across institutions of U.S. government and draw attention to important diﬀerences in support for the executive, legislative, and judicial branches.These authors also acknowledge but do not address the implications of federalism for conﬁdence and trust in government. This dimension of public support needs further consideration.

22

Brace and Johnson

In the midst of the emergent and burgeoning interest in state-level public opinion, the lack of attention to state-level support for the national government seems remarkable for at least three reasons. First, in one way or another, conﬁdence in government is a core concern of the study of political attitudes. Second, whatever variation is evident in these attitudes across the states variation can illustrate the degree of national integration or sectionalism within the contemporary United States.Third, examining state-level variation in conﬁdence in the national government can help identify the conditions that encourage or inhibit government legitimacy.

The Study of Political Trust and Conﬁdence The study of conﬁdence in government has been a mainstay of attitudinal research. Political trust is deﬁned as a basic evaluative orientation toward the government (Stokes, 1962) based on how people perceive the government is operating according to their normative expectations (Miller, 1974).The dramatic drop in Americans’ political trust occurring in the latter 1960s prompted intense academic and journalistic debate. Opinions diﬀer as to the source of the decline but there is apparent consensus that trust is essential to the proper functioning of democracy. When support for institutions declines, legitimacy is called into question and governments become hard-pressed to be eﬀective without the ﬂexibility aﬀorded by public trust (Hetherington, 1998). Distrust could reduce approval for incumbent governments and make it hard for leaders to solve problems (Brace and Hinckley, 1992; Neustadt, 1991). This incapacity could breed further distrust. Studies of variations in political trust at the national level have revealed strong connections to the presidential vote (Hetherington, 1999), feelings about the incumbent president (Hetherington, 1998), and white support for racial policies (Hetherington and Globetti, 2002). Distrustful voters were more likely to support George Wallace in 1968 (Abramson, 1983) and Perot in 1992 (Abramson,Aldrich, and Rohde, 1994).The distribution of trust over time and across the population has thus had meaningful consequences in elections and for policy.

Dimensions of Trust in Government As conceptualized by the National Election Study, political trust was a general concept. The trust questions were intended to evaluate how respondents felt about the honesty and other ethical qualities of public oﬃcials, as well as the eﬃciency and correctness of government decisions

Correlates of State-Level Confidence in the Federal Government 23

(Stokes, 1962). The introduction and four speciﬁc items regarding political trust focus on people’s orientations toward “the government,”“people in government,” or “the government in Washington” (National Election Studies, 1995–2000).While the introductory paragraph and the ﬁrst question in this series speaks of “the people in Washington,” three of the four questions only mention “the government” or “people in government” and could in principle extract feelings about oﬃcials in any of the branches of government and at any level of government. We now know that trust is not monolithic as a variety of polls reveal signiﬁcant variation in the public’s support for the institutions of the national government (Hibbing and Theiss-Morse, 1995). Speciﬁcally, Americans have more conﬁdence in the Supreme Court than they do in the presidency, which they support more than the Congress. Moreover, polls conducted by the Gallup Organization reveal dramatic diﬀerences in public support for government depending on the level of government addressed. Many respondents express as high amounts of trust in state or local government as they do in the national government. For example, when Gallup polled opinions about each level of government in June 1997, only 3 percent expressed high support for the national government, while 18 and 21 percent expressed high support for state and local government respectively.The pronounced decline in government trust is not uniform across institutions of the federal government nor across levels within the federal system. But does conﬁdence in government meaningfully vary across states in the United States? For students of state government, it might be helpful if we had comprehensive and comparable data on trust in both national and subnational levels of government across the states. Unfortunately, such data do not exist.2 It is possible, however, to construct a political trust measure that better isolates attitudes about the institutions of the federal government. We examine diﬀerences in levels of support for the federal government among the states using survey items about government trust included on repeated administrations of the General Social Survey (GSS), disaggregated to the state level and validated using suitable diagnostic tools (Brace, Sims-Butler, Arceneaux, Johnson, 2002). This measure allows us to judge the extent to which attitudes directed expressly at the major institutions of the federal government (the Supreme Court, the president, and Congress) correspond to each other, indicating high or low degrees of consistency in how respondents feel about these institutions. This statelevel measure of trust in the federal government can be subjected to analyses to consider the forces promoting variable trust across the states.

24

Brace and Johnson

Measuring State Confidence in the Federal Government Since the early 1970s, the GSS has included a series of questions to gauge the conﬁdence Americans have in the people responsible for a variety of formal and informal U.S. institutions. The underlying question includes a variety of prompts for speciﬁc institutions and industries, but generally asks, “I am going to name some institutions in this country. As far as the people running these institutions are concerned, would you say you have a great deal of conﬁdence, only some conﬁdence, or hardly any conﬁdence at all in them?” The list of institutions includes the three branches of the U.S. government: Executive branch of the federal government, Congress, and U.S. Supreme Court.3 Between 1975–1998, the GSS cumulative ﬁle reports 26,107 responses to all three questions.4 We create an individual-level scale of conﬁdence in the federal government from these three items to serve as the basis for a state-level conﬁdence in the national government.The individual responses to questions about the executive, legislative, and judicial branches were recoded with “a great deal of conﬁdence” = 100; “only some conﬁdence” and “don’t know” = 0; and “hardly any conﬁdence at all” = –100. Refusals are left out of the scale, treated as missing data. The three items produce a usable scale measuring conﬁdence in people with power across branches of the federal government (Cronbach’s D = 0.65). We average across the three items to measure individual conﬁdence in the federal government, producing an indicator with a potential range from –100 to 100. We are interested in the correlates of the conﬁdence in the federal government expressed across the American states. Before reaching substantive questions about what contextual, economic, and political factors may be related to conﬁdence in the federal government, we must ask whether conﬁdence in the federal government actually varies across states. It could be the case that a uniformity to conﬁdence in government varies individually or temporally, inﬂuenced by the partisan competition of government, the competitiveness of politics, or simply the mood of voters, either their policy mood, or growing cynicism. However, the premise of our further investigation is that the residents of states have an aggregate conﬁdence in the federal government distinguishable from the conﬁdence expressed by residents of other states. In order to assess this basic measurement question—whether it is appropriate to pool individual respondents by their states and compute aggregate measures of state conﬁdence in government—we turn to several well-established tools used by students of state public opinion and the political orientation of state electorates.

Correlates of State-Level Confidence in the Federal Government 25

Erikson,Wright, and McIver (1993) recommend using the split-halves approach developed by scholars of psychological and education testing devices (see Carmines and Zeller, 1979). For these tests, observations in the GSS cumulative ﬁle are split into subsets two diﬀerent ways. Dividing respondents in even- and odd-year administrations of the GSS, we compute state-level means for these subsets. The correlation between these even- and odd-year state means underlies a Spearman-Brown prophecy statistic5 used to assess the reliability of our aggregate measure of state trust in government. A similar Spearman-Brown statistic was computed splitting the GSS cumulative ﬁle into earlier (1975–1986) and later (1987– 1998) groups in order to assess the stability of state-level conﬁdence in the federal government. Spearman-Brown prophecy formula scores greater than 0.7 are considered an indication that a measure is reliable (or stable, as the case may be); scores between 0.6 and 0.7 indicate moderate reliability; and measures with scores below 0.6 are thought unreliable. The state-level measure of conﬁdence and government has a reliability (even-odd) Spearman-Brown coeﬃcient of 0.74, in the reliable range. However, its stability (early-late) score is 0.55.That said, when states with small sample sizes in the cumulative ﬁle (fewer than 100 observations) are removed,6 the reliability score improves slightly (0.76) and the stability score improves somewhat more (0.6), suggesting that sampling error associated with small sample sizes is at least partially responsible for the apparent instability of collective state judgments about the federal government during this 24-year period. Jones and Norrander (1996) elaborate on a diﬀerent technique to assess the appropriateness of pooling data for analysis in state-level aggregates. They discuss a generalizability coeﬃcient,7 developed by O’Brien (1990), designed to assess the relative within- and between-unit variance of data provided by respondents (R) nested within aggregates (A). Jones and Norrander (1996) report similar thresholds for evaluating the appropriateness of pooling data at the state level with the O’Brien statistic, Er 2, with 0.7 suggesting generalizable data. The conﬁdence in government measure has a generalizability score well within the range of acceptability, Er 2 = 0.79. This score does not change when the data is restricted to respondents in states with cumulative ﬁle sample sizes greater than 100. Given our conﬁdence in the stability of this measure of conﬁdence in the federal government, particularly among states with cumulative GSS sample sizes greater than 100, we will focus this analysis here on the 39 states meeting this data requirement.We list state scores on this aggregate measure of state-level conﬁdence in the federal government in Table 2.1. Negative numbers indicate the balance of state opinion leans toward hav-

Table 2.1 State-Level Confidence in the Federal Government State

New Hampshire Rhode Island North Dakota Maryland Utah Massachusetts West Virginia Minnesota New Jersey Wisconsin Connecticut Georgia Louisiana New York Michigan Virginia Iowa Texas Kansas Illinois Alabama Arizona Tennessee Florida Pennsylvania Missouri South Carolina California Montana Kentucky Ohio Washington Colorado Oklahoma Arkansas Oregon North Carolina Indiana Mississippi

N

(122) (102) (181) (399) (132) (508) (391) (536) (898) (745) (438) (681) (323) (1,793) (1,369) (953) (212) (1,377) (323) (1,022) (671) (386) (780) (1,156) (1313) (703) (387) (2,593) (142) (182) (1,338) (386) (652) (297) (221) (355) (860) (587) (181)

State Confidence in Federal Government

11.75 8.66 7.00 6.60 6.06 3.15 2.30 1.68 0.74 0.63 0.15 –0.05 –0.31 –0.37 –0.58 –0.87 –0.94 –1.45 –1.55 –2.84 –3.63 –3.80 –3.89 –3.98 –4.11 –4.55 –5.17 –5.45 –6.57 –6.59 –6.93 –7.17 –7.72 –8.19 –9.35 –9.39 –10.85 –10.90 –19.15

Source: General Social Survey Cumulative File.

Rank

1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24 25 26 27 28 29 30 31 32 33 34 35 36 37 38 39

Census Region

Northeast Northeast Midwest South West Northeast South Midwest Northeast Midwest Northeast South South Northeast Midwest South Midwest South Midwest Midwest South West South South Northeast Midwest South West West South Midwest West West South South West South Midwest South

Correlates of State-Level Confidence in the Federal Government 27

ing “hardly any conﬁdence at all” across the branches of the federal government, on average. Positive evaluations suggest on average respondents lean toward state residents having a great deal of conﬁdence in the branches of the federal government. Not surprisingly, more states—almost three-fourths of the states with available and reliable data—have generally low levels of conﬁdence in the federal government than have positive evaluations of the central government. The U.S. Census Bureau divides the United States into four large regions: Northeast, South, Midwest, and West.8 The average levels of state conﬁdence in the federal government for the four regions are (n of states in parentheses): Northeast, 2.85 (7); Midwest, –1.90 (10); South, –4.31 (15); and West, –4.86 (7). While Southern and Western states tend to have low levels of conﬁdence in the federal government,Table 2.1 shows that several Northeastern and Midwestern states also have relatively low levels of conﬁdence in the federal government. Further, several Southern and Western state populations have a level of conﬁdence in Washington greater than the mean. On average, however, the Northeastern states have a higher level of conﬁdence in the federal government than the other three regions, which have similar levels of conﬁdence in the federal government.9

Understanding State Confidence in the Federal Government What accounts for variation in a state’s aggregate conﬁdence in the federal government? Here, we examine three broad groups of variables we anticipate might relate to state-level variance in conﬁdence in the federal government.We may divide these inﬂuences into (1) federal presence, (2) population attributes, and (3) political processes.

Federal Presence: Relations with the National Government States diﬀer considerably in their relations with the federal government. Notably, the federal government is a major employer in some states and plays a critical role in shaping their economic development (Brace, 1993).We would expect state-level conﬁdence in the federal government to vary positively with levels of employment by the federal government. If a state’s population is receiving ﬁnancial beneﬁts directly from the federal government as an employer, for example, receiving a paycheck from a federal government agency or bureau, we anticipate that working for the federal government will aﬀect a higher level of trust in the government. Combined then with the multiplier eﬀects of these investments on a state’s economy, federal employment should be associated with relatively better feelings about the government.

28

Brace and Johnson

Not all relations with the federal government are positive.The federal government controls roughly 20 percent of all land in the United States. Tensions over this control came to a peak in the 1970s. The Sagebrush Rebellion represented something of a regional realignment in Western states (Cawley, 1997), when states frustrated by this federal presence clamored for local control of these lands. Ronald Reagan supported the Sagebrush Rebellion and appointed James Watt as his Secretary of the Interior. Watt moved to promote private ownership of public lands but this eﬀort ultimately failed. With Reagan in the White House, the Sagebrush Rebellion became dormant, but it was not eliminated. Many people continue to view government presence as an encroachment and the Sagebrush Rebellion has not gone away (Lewis, 2004). From cabinet and judicial appointments, to acts of civil disobedience and even violence, the control of public lands by the federal government shapes politics in meaningful ways and could be expected to shape public attitudes about the federal government. We expect that conﬁdence in the federal government will be inversely related to the percentage of a state’s land that is controlled by the federal government.

Population Attributes: Income and Complexity State populations vary in many ways. Most notably states vary in their levels of wealth and complexity. Some states are wealthy and urbanized. Such states may have very diﬀerent needs of the federal government than states that are less aﬄuent and wealthy. We examine per capita personal income and urbanization, mainstays of state-level analysis (Hero, 1998). Individuals living in states with stronger economies and higher levels of personal well being presumably are beneﬁting from the federal arrangement and are likely to be more supportive. Urbanization may aﬀect how a state’s residents regard the federal government, because of the historical association between the federal government and cities. Caraley (1992) argues that in the years following the Great Depression, the federal government invested resources in cities in order to address the gap between urban infrastructure needs and the ability of cities to raise ﬁnancial resources, covering more than one-ﬁfth of city expenditures by 1980. Consequently, states with larger urban populations may have a stronger evaluation of the federal government.

Political Processes: Awareness, Political Orientations, and Participation Among citizens, knowledge and awareness of political institutions inﬂuences feelings of conﬁdence and legitimacy in those institutions

Correlates of State-Level Confidence in the Federal Government 29

(Gibson, Caldeira, and Baird, 1998).We attempt to gauge these feelings by examining the relationship between variations in state levels of education and conﬁdence in the federal government. We anticipate a positive relationship between how well a state’s residents are educated and conﬁdence in the federal government. We also consider several political variables, including ideology, partisanship, and presidential turnout. Owing to the traditional Republican and conservative aversion to government generally, we expect more liberal and more Democratic states to have a higher conﬁdence in the federal government. In addition, we expect political activity to be related to support for the federal government and anticipate that states with higher turnout will exhibit more conﬁdence in the federal government. At the same time, however, very low levels of turnout might be indicative of satisfaction and complacency as much as disengagement. Consequently we consider the possibility that turnout and conﬁdence in the federal government could also manifest a curvilinear relationship.

Data and Measures Given that our dependent variable represents a multiyear average of state-level public opinion, albeit reasonably stable, we measure our independent variables with a similar approach, averaging across the values of each variable included in the equation at multiple time points. For most of these indicators, we mean reported information for each state in 1970, 1980, and 1990. For indicators of federal presence in a state, we include both the percentage of state land that is owned by the federal government and the percentage of the state workforce employed by the federal government.The percentage of state land that is federally owned is reported both in the annual Statistical Abstract of the United States (U.S. Department of Commerce, 1995) and regularly in an inventory of federally owned property (U.S. General Services Administration, 1971, 1983). Although the amount of federally owned property in each state changes little from decade to decade, we average across the percentage of state land owned by the federal government at 10-year intervals to better reﬂect the tendency of the federal government to own state property over the 22 years reﬂected in the dependent variable.The percentage of the state workforce employed by the federal government is easily computed using data from the Bureau of Economic Analysis detailed state-level income and employment tables (U.S. Department of Commerce, 2004). We use similar sources of data and techniques to measure educational attainment, income, and urbanization in the states.The Statistical Abstract

30

Brace and Johnson

regularly reports the percentage of each state’s residents with a high school diploma, and the percentage of state residents living in metropolitan areas (U.S. Department of Commerce, 1973, 1984, 1995). We obtain data on state per capita income in constant dollars from the Bureau of Economic Analysis summary personal income estimates (U.S. Department of Commerce, 2004). Again, for each of these measures, we average across observations in 1970, 1980, and 1990 in order to reﬂect the tendencies with states during the 1975–1998 time frame informing our dependent variable. The political variables are measured somewhat diﬀerently. Our measures of state political ideology and partisanship, drawn from Erikson, Wright, and McIver (1993), also reﬂect multiple time points, relying on survey data from the late 1970s to mid-1980s. However, the slight diﬀerences in the time frame of our dependent variable and these data are not particularly worrisome given the general stability of these state political orientations during the years considered here (Brace,Arceneaux, Johnson, and Ulbig, 2004). We measure political participation by averaging statelevel turnout in each presidential election 1976 to 1996, provided by the Federal Election Commission (2004). We include in the models reported below a quadratic speciﬁcation for presidential turnout, in order to account for the possibility that very low levels of turnout are associated with more conﬁdence in government than moderate participation, reﬂecting satisfaction with the federal government.

Findings Table 2.2 reports the results of several analyses of state-level trust in the federal government. We begin by considering our separate explanatory categories. In the ﬁrst column, we consider only the federal-presence indicators. As expected, federal land ownership is associated with less conﬁdence in the government while a higher level of federal employment in a state is associated with greater conﬁdence in the government. These results indicate opinions about the federal government are shaped in meaningful ways by public experiences with the federal government. To those living in a state with high levels of federal employees, the federal government is a source of jobs and income. To those living in a state where Washington controls huge tracts of land, the federal government may be viewed as an encroaching presence that inhibits economic growth and development. Quite clearly, levels of feelings of conﬁdence about the federal government vary based on the nature of a state’s relationship to the federal government. The model in the second column includes only the socioeconomic

Table 2.2 Modeling State Confidence in the Federal Government a Model 1

Model 2

Model 3

Model 4

Model 5

Federal Presence

Socioeconomics

Political Variables

Full Model

Reduced Model

ß (robust SE)

ß (robust SE)

ß (robust SE)

ß (robust SE)

ß (robust SE)

–0.109** (0.053) 1.490*** (0.480) 0.180 (1.222) 0.069 (0.103)

–0.085** (0.033) 1.150** (0.534)

Percent of state land –0.082*** federally owned (0.025) Per worker federal 0.746† employment (0.527) Per capita personal income (in thousands) Percentage of population living in metropolitan areas Percentage of residents with high school diploma Political ideology (Erikson,Wright, and McIver) State partisanship (Erikson,Wright, and McIver) Average presidential vote turnout 1976–1996 (%VAP) Average presidential vote turnout squared Constant

1.379** (0.615) –0.064 (0.069)

–4.023** (1.710)

–13.735*** (4.961)

N = 39 F2,36 = 6.4***

N = 39 F2,36 = 2.96*

R2 = 0.11

R2 = 0.10

–0.108 (0.184)

–0.071 (0.244)

0.214* (0.114)

0.165 (0.191)

0.006 (0.113)

–0.002 (0.098)

–2.549** (1.194)

–3.470** (1.372)

–1.467 (1.126)

0.025** (0.011)

0.034** (0.013)

0.016† (0.010)

71.145* (37.134)

80.837** (39.984)

28.420 (31.028)

N = 39 N = 39 N = 39 F5,33 = 2.99** F9,29= 5.09*** F4,34 = 6.34*** R2 = 0.15

R2 = 0.39

R2 = 0.22

*** p < 0.01, ** p < 0.05,* p < 0.1 (two-tailed tests); † p < 0.1 (one-tailed) a This analysis is restricted to states with 1975–1998 cumulative GSS sample sizes greater than 100, thus excluding Alaska, Delaware, South Dakota,Vermont, and Wyoming. Hawaii, Idaho, Maine, Nebraska, Nevada, and New Mexico are not included in the GSS cumulative file. Source: General Social Survey Cumulative File.

32

Brace and Johnson

factors we discuss above. Here we ﬁnd only per capita personal income appears to have a signiﬁcant relationship with conﬁdence in the federal government. It is signed in the positive direction, suggesting that more aﬄuent states are more conﬁdent in the political and economic system in place. Urbanization does not exhibit a signiﬁcant relationship to support for the federal government. Model 3 includes the political variables. As expected, more liberal states have more conﬁdence in the federal government than conservative states. However, there appears to be no signiﬁcant relationship between partisanship and conﬁdence in the federal government. Political participation manifests the curvilinear relationship with conﬁdence in the federal government we anticipate. This indicates that conﬁdence in the federal government decreases as a function of turnout, with low turnout states exhibiting higher conﬁdence in the federal government.The signiﬁcance of squared turnout indicates that this pattern reverses itself for the states with the highest turnout. Overall, it appears that states with politically complacent and politically active populaces are more conﬁdent in the federal government than states exhibiting middling levels of turnout. We combine the alternative explanations in the fourth column. Here we ﬁnd that, controlling for other potential correlates of federal government conﬁdence, both federal presence indicators are strengthened and remain signed in the anticipated direction. However, several factors signiﬁcantly related to conﬁdence in the federal government are not robust to the inclusion of other variables. Income and ideology both fail to reach conventional levels of statistical signiﬁcance, although they remain signed in the anticipated direction.This is not the case with participation, which also has a stronger relationship with conﬁdence in government in the fully speciﬁed model.

Discussion Employing a well-established set of methodological tools and a familiar source of public opinion data, we are able to estimate measures of trust in the federal government. Importantly, this conﬁdence varies systematically across the states, related to the federal government’s varying status as landlord and employer, and political participation. This measure has the potential to further inform our understanding of state politics and contemporary intergovernmental relations, as well as public policy outcomes (see Hetherington, 1998, 1999; Hetherington and Globetti, 2002). Over the course of U.S. history, the nation has experienced various phases of federalism ranging from the “dual federalism” of the nineteenth

Correlates of State-Level Confidence in the Federal Government 33

century, the “cooperative federalism” of much of the twentieth century, to the “new federalisms” of Nixon and then Reagan. Throughout the nineteenth century the national government provided promotional and support services to the state governments, which in turn undertook most of the substantive tasks of governing. Each era marked a diﬀerent level of penetration by the federal government into governance, and presumably each era elicited diﬀering feelings from the citizenry about the legitimacy of the federal government. We cannot go back in time to examine variations in conﬁdence in the federal government. We can, however, comment on the levels of conﬁdence exhibited by the public across the states in the contemporary era. Perhaps the best way to appreciate these levels of conﬁdence may be to consider what they are not, rather than what they are.They are not indicative of pervasive sectionalism. Although states in the Northeast tend to be more conﬁdent in the federal government than states in other regions, across these other regions there appears to be a general lack of conﬁdence in the federal government.Thus slight regional diﬀerences notwithstanding, we ﬁnd that state diﬀerences in trust in the federal government vary systematically with characteristics of states other than their region. We did not ﬁnd clear patterns indicating federal support by economic winners and opposition by economic losers, or notable partisan divisions in support for the federal government. In the United States, rich and poor states tend to exhibit similar levels of conﬁdence in the federal government, as do Democratic and Republican and urban and rural states. This result, in and of itself, should not be underestimated because we do not ﬁnd the kind of “center” versus “periphery,” “have” versus “have-not” or cultural divisions that commonly threaten other federal regimes. There are, nonetheless, noteworthy diﬀerences among the states. States are diﬀerentially beneﬁted and burdened by the federal government and public conﬁdence in the federal government reﬂects these relationships in an entirely plausible manner. Given this, we would expect changes in government expenditure and employment to play a critical role in shaping state-level conﬁdence in the federal government. At the same time, the outcome of the politics of public land use, which pervades many debates in Washington, may serve to enhance or erode conﬁdence in the federal government in those states where federal holdings are high. In the end, the seemingly high level of national unity we ﬁnd would seem to be indicative of an evolved, nationally integrated democracy where questions regarding the role of the federal government are not alarmingly divisive. At this point in our nation’s evolution, however, forces promoting privatization may conﬂict with forces promoting protecting

34

Brace and Johnson

public lands. Moreover, devolution and federal retreat in many areas may fundamentally alter federal employment patterns among the states. The diﬀerences that we ﬁnd may not be alarming, but they are also not random.The politics of federal contraction and those of public lands will play a decisive role in the distribution of conﬁdence in the federal government in the future.

Notes 1. In his investigation of sectional diﬀerences in contemporary American politics, so dominated by a perceived distinction between “red” and “blue” states, Fiorina ﬁnds “similar proportions regard the government as almost always wasteful and ineﬃcient—relative to the red states, the blue states are clearly not wellsprings of support for big government” (2005, 11).There are diﬀerences between states in assessments of the federal government but these diﬀerences are only partially a function of aggregate partisan leanings perhaps. 2. Ulbig, Wade, and Martorano (2004) use state-level survey data on trust in state and national governments but, due to the limited availability of relevant data, are only able to examine attitudes about Congress in 13 states and state institutions in eight states. 3. An alternative form of the question was used in 1984 for about one-third of GSS respondents, 485 cases. These observations are not used in the present analysis. 4. As noted, the questions appeared on two of three forms of the GSS in 1984. NORC (National Opinion Research Center) did not appear on the survey in 1985, and were asked of two-thirds of respondents in 1988–1998. 5. The Pearson’s correlation coeﬃcient r12 is adjusted using the SpearmanBrown prophesy formula: 2r12 1+r12 6. Alaska, Delaware, South Dakota, Vermont, and Wyoming each contribute fewer than 100 respondents to the GSS cumulative ﬁle, 1975–1998. Hawaii, Idaho, Maine, Nebraska, Nevada, and New Mexico provide no observations in the GSS cumulative ﬁle during these years. 7. O’Brien’s (1990) generalizability coeﬃcient for the R:A Design contemplates the mean square, an estimate of the population variance between aggregate units, MS(a), and the mean square for individual-level scores within the aggregated units, MS(r:a), using the formula: 2

Er =

[MS(a) — MS(r:a)] MS(a)

MS(a) and MS(r:a) were estimated using the One-Way ANOVA procedure in SPSS.

Correlates of State-Level Confidence in the Federal Government 35 8. As deﬁned by the U.S. Census Bureau, the Northeast includes Connecticut, Massachusetts, Maine, New Hampshire, New Jersey, New York, Pennsylvania, and Vermont.The South includes Alabama,Arkansas, Delaware, Florida, Georgia, Kentucky, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina,Tennessee,Texas,Virginia, and West Virginia.The Midwest includes Illinois, Indiana, Iowa, Kansas, Michigan, Minnesota, Missouri, Nebraska, North Dakota, Ohio, South Dakota, and Wisconsin. The West includes Alaska, Arizona, California, Colorado, Hawaii, Idaho, Montana, Nevada, New Mexico, Oregon, Utah, Washington, and Wyoming. The Census Bureau divides these regions into divisions (e.g., states of the Northeast are separated into the Middle Atlantic made up of New Jersey, New York, and Pennsylvania, and a New England division with the other six states in the region). However, we focus on the regional divisions in part because cutting the 50 states into four regions rather than nine smaller divisions uses ﬁve fewer degrees of freedom and leads to comparing groups of 7, 10, 15, and 7 states instead of groups of states with an average size of 5. 9. The mean level of conﬁdence in the federal government across the Northeastern states we have data for was signiﬁcantly higher than conﬁdence in government observed across the other regions. In the aggregate, Northeastern states were more conﬁdent in the federal government than Western states, (t = 2.71, p < 0.05), Southern states (t = 2.62, p < 0.05), and Midwestern states (t = 1.87, p < 0.1).The diﬀerences in means among the other census regions did not reach conventional levels of statistical signiﬁcance.

References Abramson, Paul R. 1983. Political Attitudes in America. New York: W. H. Freeman. Abramson, Paul R., John H. Aldrich, and David W. Rohde. 1994. Change and Continuity in the 1992 Election.Washington, DC: Congressional Quarterly Press. Anderson, Christopher J., and Christine A. Guillory. 1997. “Political Institutions and Satisfaction with Democracy:A Cross-National Analysis of Consensus and Majoritarian Systems.” American Political Science Review 91 (March): 66–81. Atkinson, Michael M.,William D. Coleman, and Thomas J. Lewis. 1980. “Regime Support in Canada: A Comment.” British Journal of Political Science 10 ( June): 402–410. Bowler, Shaun, and Todd Donovan. 2002. “Why Do People Like their State and Local Government more than the Federal Government?” Paper presented at the annual meeting of the Midwest Political Science Association, Chicago, IL. Brace, Paul. 1993. State Government and Economic Performance. Baltimore, MD: Johns Hopkins University Press. Brace, Paul, Kevin Arceneaux, Martin Johnson, and Stacy G. Ulbig. 2004. “Does State Ideology Change over Time?” Political Research Quarterly 57 (December): 529–540. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using New Data.” American Journal of Political Science 46 ( January): 173–189.

36

Brace and Johnson

Brace, Paul, and Barbara Hinckley. 1992. Follow the Leader. New York: Basic Books. Caraley, Demetrios. 1992. “Washington Abandons the Cities.” Political Science Quarterly 107 (Spring): 1–30. Carmines, Edward G., and Richard A. Zeller. 1979. Reliability and Validity Assessment. Beverly Hills, CA: Sage. Cawley, L. MacGregor. 1997. Federal Land, Western Anger: The Sagebrush Rebellion and Environmental Politics. Lawrence: University of Kansas Press. Citrin, Jack. 1974. “Comment:The Political Relevance of Trust in Government.” American Political Science Review 68 (September): 973–988. Dionne, E. J. 1991. Why Americans Hate Politics. New York: Simon and Schuster. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Federal Election Commission. 2004.“About Elections and Voting:Voter Registration and Turnout Statistics.” Washington, DC: Federal Election Commission . Fiorina, Morris P., with Samuel J. Abrams and Jeremy C. Pope. 2005. Culture War? The Myth of a Polarized America. New York: Pearson Longman. Gibson, James L., Gregory A. Caldeira, and Vanessa A. Baird. 1998.“On the Legitimacy of National High Courts.” American Political Science Review 92 ( June): 343–358. Hero, Rodney E. 1998. Faces of Inequality: Social Diversity in American Politics. New York: Oxford University Press. Hetherington, Marc J. 1998. “The Political Relevance of Political Trust.” American Political Science Review 92 (December): 791–808. ———. 1999. “The Eﬀect of Political Trust on the Presidential Vote, 1968–96.” American Political Science Review 93 ( June): 311–326. Hetherington, Marc J., and Suzanne Globetti. 2002. “Political Trust and Racial Policy Preferences.” American Journal of Political Science 46 (April): 253–275. Hetherington, Marc J., and John D. Nugent. 2001.“Explaining Public Support for Devolution: The Role of Political Trust.” In John R. Hibbing, and Elizabeth Theiss-Morse, eds., What Is it About Government That Americans Dislike? New York: Cambridge University Press, pp. 134–151. Hibbing, John R., and Elizabeth Theiss-Morse. 1995. Congress as Public Enemy: Public Attitudes Toward American Political Institutions. New York: Cambridge University Press. Jackman, Robert W. 1972. “Political Parties,Voting, and National Integration:The Canadian Case.” Comparative Politics 4 ( July): 511–536. Jones, Bradford S., and Barbara Norrander. 1996. “The Reliability of Aggregated Public Opinion Measures.” American Journal of Political Science 40 (February): 295–309. Kornberg, Alan. 1990. “Political Support in Democratic Societies: The Case of Canada.” Journal of Politics 52 (August): 709–716. Kornberg, Allan, Harold D. Clarke, and Lawrence Leduc. 1978. “Some Correlates of Regime Support in Canada.” British Journal of Political Science 8 (April): 199– 216.

Correlates of State-Level Confidence in the Federal Government 37 Kornberg, Alan, Harold D. Clarke, and Marianne C. Stewart. 1979. “Federalism and Fragmentation in Political Support in Canada.” Journal of Politics 41 (August): 889–906. Lewis, Neil A. 2004.“Another Clash on a Judicial Nominee, but the Issue Is New.” New York Times. February 6. National Election Studies, Center for Political Studies, University of Michigan. 1995–2000. The NES Guide to Public Opinion and Electoral Behavior.Ann Arbor, MI: University of Michigan, Center for Political Studies [producer and distributor]. . Neustadt, Richard E. 1991. Presidential Power and the Modern Presidents. New York: Free Press. O’Brien, Robert M. 1990.“Estimating the Reliability of Aggregate-Level Variables on Individual-Level Characteristics.” Sociological Methods and Research 18 (May): 473–504. Stokes, Donald E. 1962. “Popular Evaluations of Government: An Empirical Assessment.” In Harlan Cleveland and Harold D. Laswell, eds. Ethics and Bigness: Scientiﬁc,Academic, Religious, Political, and Military. New York: Harper and Brothers, pp. 61–72. Ulbig, Stacy G., Michelle Wade, and Nancy Martorano. 2004. “Punishing the Statehouse for the Sins of Congress: Displaced Anger and State Government Performance Ratings.” Paper presented at the annual meeting of the Southern Political Science Association, New Orleans, LA. U.S. Department of Commerce, Bureau of the Census. 1973. Statistical Abstract of the United States.Washington, DC: U.S. Government Printing Oﬃce. ———. 1984. Statistical Abstract of the United States. Washington, DC: U.S. Government Printing Oﬃce. ———. 1995. Statistical Abstract of the United States. Washington, DC: U.S. Government Printing Oﬃce. U.S. Department of Commerce, Bureau of Economic Analysis. 2004. Regional Economic Accounts: Annual Personal Income.”Washington, DC: U.S. Department of Commerce, Bureau of Economic Analysis . U.S. General Services Administration. 1971. Inventory Report on Real Property Owned by the United States Throughout the World as of June 30, 1970.Washington, DC: U.S. Government Printing Oﬃce. ———. 1983. Summary Report of Real Property Owned by the United States Throughout the World as of September 30, 1980. Washington, DC: U.S. Government Printing Oﬃce.

Chapter

State Residency, State Laws, and Public Opinion

3

Barbara Norrander and Clyde Wilcox

“Location, location, location” is the mantra of real estate agents in describing the worth of property. Echoing this theme, social scientists are increasingly readopting a geographic perspective in analyzing public opinion and electoral outcomes. New measures of state ideology and partisanship ignited a wealth of studies. Wright, Erikson, and McIver (1985) developed their inﬂuential measure of state ideology and partisanship by pooling national media surveys over a number of years. Berry, Ringquist, Fording, and Hanson (1998) created yearly indicators of citizen ideology based on congressional ratings and election returns. Brace and his colleagues (2002) pooled General Social Surveys to create their measures of state public opinion. Finally, Jones and Norrander (1996) and Norrander (2001) demonstrate that the pooled 1988–1992 Senate National Election Studies (SNES) can be used to produce measures of state ideology and partisanship as well. State ideology has been successfully linked to a host of state policies, including laws governing rape, alcohol, drugs, budgets, economic development, welfare, and the death penalty (Berger, Neuman, and Searles, 1991; Berman and Martin, 1992; Call, Nice, and Talarico, 1991; Johnson and Meier, 1990; Meier, 1992; Meier and Johnson, 1990; and Nice, 1991, 1992) State ideology also can be used to judge the behavior of state supreme court justices (Brace, Langer, and Hall, 2000). Furthermore, states are being studied for their contribution to national outcomes, such as presidential elections (Campbell, 1992; Erikson, McIver, and Wright, 1987; Holbrook, 1991; Jackson and Carsey, 1999a, 1999b; Shelley and Archer, 1994). Because of data availability, most attention has been paid to state ideology or state partisanship.Yet ideology and partisanship are not always suc38

State Residency, State Laws, and Public Opinion 39

cessful surrogates for opinions on a wider range of issues. Even with improvements in measuring issue positions (Sullivan, Piereson, and Marcus, 1978) and a tighter connection between partisanship and ideology in recent years (Abramowitz and Saunders, 1998), the American public still cannot be described as holding tightly constrained issue opinions that are shaped by an overarching ideology. Fortunately, measures of state-level opinion on speciﬁc issues have been developed from the Senate National Election Surveys ( Jones and Norrander, 1996; Norrander, 2001) and the General Social Surveys (Brace, Sims-Butler, Arceneaux, and Johnson, 2002). These aggregated issue attitudes have been successfully linked to state policies on abortion (Norrander and Wilcox, 1999) and capital punishment (Mooney and Lee, 2000; Norrander, 2000) and to the number of women in state legislatures (Arceneaux, 2001). A complete understanding of public opinion in the states requires the examination of a host of speciﬁc issues as well as ideology and partisanship.

State Variations in Public Opinion As state public opinion is increasingly linked to state policies and national politics, a better understanding of why public opinion varies by states is needed.The simplest explanation for state-level variations in public opinion would be that compositional eﬀects of state electorates account for much of the variation in aggregate public opinion. In this vein, Cook, Jelen, and Wilcox (1993a) found in a study of abortion opinion in six states that interstate diﬀerences could be accounted for by distributional diﬀerences in education, race, age, and religion. In addition, the premise that demographic composition structures public opinion in the states was the foundation for the early computer simulations of state public opinion (Pool,Abelson, and Popkin, 1965;Weber, Hopkins, Mezey, and Munger, 1972–1973). State residency, per se, could have independent inﬂuences on public opinion in a number of ways. State laws vary and laws have legitimizing eﬀects on public opinion. For example, Norrander (2000) notes approval of capital punishment is 7 percentage points higher in states with death penalty laws versus those that do not have such statutes. State laws on partisan or nonpartisan voter registration and rules for primary participation inﬂuence levels of partisan identiﬁcation (Burden and Greene, 2000; Finkel and Scarrow, 1985; Norrander, 1989). Groups organize and promote their views, such that states with an actively organized group could strengthen the opinions of their supporters and perhaps cause countermobilization among their opponents. In this

40

Norrander and Wilcox

vein, Cook, Jelen, and Wilcox (1993b) in a study of abortion opinion in 42 states found that the distribution of Catholic and Protestant residents mattered. Protestants in heavily Catholic states became more supportive of abortion rights than Protestants in other states. Others attribute the inﬂuence of states on public opinion to variations across the states in political culture. Elazar (1966) divided the states into three categories: traditionalistic, individualistic, and moralistic. Although these political cultures refer primarily to styles of governing, they also have been linked to public policy (Lowery and Sigelman, 1982) and state opinion (Norrander, 2000). Meanwhile, Erikson, McIver, and Wright (1987) deﬁne “state political culture as only that portion of state public opinion that cannot be accounted for by the group characteristic of the state electorate” (798). In this chapter, we test the ﬁrst of these alternative explanations for state public opinion: the eﬀect of state laws.We begin our analysis of individuallevel opinion on ideology, partisanship, and several issues by allowing individual demographic traits to explain their maximum variance. Next we add measures of state residency to document the unique contributions of states on individuals’ opinions. Finally, we test how much of these unique state residency eﬀects can be explained by prior state laws.

Data and Methods State public opinion can be studied using the pooled Senate National Election Study (Miller, Kinder, Rosenstone, and National Election Studies, 1999). Unlike other NES surveys, the Senate study employs a state-based sampling frame. Thus, the SNES provides representative samples of all 50 states. Jones and Norrander (1996) demonstrate the reliability of many of the issue questions in the SNES as indicators of aggregated opinion for the states. In this study, the concentration is in the other direction. How much inﬂuence does state residency have on public opinion? We adopt a common methodology for judging the inﬂuence of state residency on public opinion (Cook, Jelen, and Wilcox, 1993a; Erikson, McIver, and Wright, 1987).We control for a variety of demographic traits to eliminate, as much as possible, the inﬂuence of varying population characteristics as reasons for state-level diﬀerences in opinion. Second, we add dummy variables for each state (minus one) to judge the independent inﬂuence of state residency on public opinion. We also designate as core political attitudes partisanship and ideology. These core attitudes will be included as determinants of attitudes on speciﬁc issues (three abortion measures, capital punishment, and government aid to blacks).

State Residency, State Laws, and Public Opinion 41

Although the SNES provides an ideal sampling frame, it provides less than ideal measures of demographic characteristics. Basic measures of gender, race, ethnicity, job status, marital status, union membership, age, and income are available. The only indicator of religious view is denominational aﬃliation. Still, including these basic demographic traits should account for much of the inﬂuence of an individual’s background on public opinion.1 Methodologically, OLS regression analyses were ﬁrst run using only demographic variables to explain attitudes. State residency was then added by including 49 dummy variables. The excluded state in each analysis is the state whose citizens fall closest to the median opinion on the issue nationwide. Thus, Oregon is the excluded category for the partisanship model, Wisconsin for ideology, Texas for abortion attitudes, Pennsylvania for government funding of abortions for poor women, Florida for parental consent for abortions for women under 18, Kansas for the death penalty, and Iowa for government aid to black people. The impact of state residency is gauged by the change in R2 when the 49 dummy variables are added to the model based on demographic traits.This methodology allows for demographic composition to explain its maximum eﬀect, with the state dummy variables picking up any additional unique inﬂuence beyond demographic factors. For the issue-speciﬁc questions (abortion, capital punishment, aid to black people), a change in R2 also is computed for the addition of the two core attitudes (partisanship and ideology) to the demographic variable model. State residency is added after the core attitudes. The signiﬁcance of the change in R2 is indicated by the F-statistic and its level of signiﬁcance. The percent of explained variance due to demographic, core attitudes, and state residency also is indicated in the tables.

Results Table 3.1 examines the inﬂuence of state residency and demographic traits on the core attitudes of party identiﬁcation and ideology. Personal demographic traits explain 11.2 percent of the variation in party identiﬁcation. State residency adds an additional 1.4 percent to the explanation of partisan preferences. For ideological self-identiﬁcation, demographic traits account for 5.5 percent of the total variance and state residency provides an additional 1.5 percent of explained variance. The amount of variance in individual attitudes attributable to state residency at ﬁrst appears to be quite small. The explained variances for demographic traits and state residency reported here are similar in size to those

Table 3.1 The Influence of Demographic Traits and State Residency on Core Attitudes of Party Identification and Ideology Party Identification b (SE)

Demographics Education Homemaker Working Unemployed Retired Student Female Married White Black Native American Asian American Latino No religion Conservative Christian Jew Catholic Union Age Income States Total R2 Number of Cases

0.05** (0.01) 0.60** (0.17) 0.42** (0.15) 0.20 (0.19) 0.52** (0.16) 0.48* (0.21) –0.26** (0.05) 0.14** (0.05) 0.56** (0.22) –1.35** (0.24) –0.04 (0.30) 0.52 (0.32) –0.38** (0.14) –0.63** (0.11) –0.02 (0.06) –1.68** (0.26) –0.50** (0.07) –0.80** (0.07) –0.01** (0.00) 0.08** (0.1)

7,127

* p d0.05; ** p d 0.01 Source: Senate National Election Study.

R2

Ideology %

0.112**

88%

0.014** 0.127**

11%

b (SE)

–0.04** (0.01) 0.23 (0.14) 0.04 (0.13) 0.09 (0.16) –0.07 (0.13) –0.01 (0.17) –0.20** (0.04) 0.22** (0.04) –0.17 (0.17) –0.56** (0.19) –0.47 (0.25) 0.15 (0.26) –0.10 (0.12) –0.67** (0.08) 0.09 (0.05) –1.16** (0.22) –0.16** (0.06) –0.20** (0.05) 0.012** (0.002) 0.004 (0.01)

7,035

R2

%

0.055**

79%

0.015** 0.070**

21%

State Residency, State Laws, and Public Opinion 43

described by Erikson, McIver, and Wright (1987) for their analysis of the pooled CBS News/New York Times polls. In addition, the impact of state residency on party identiﬁcation is more than such individual demographic traits as income (change in R2 of 0.006**) and only slightly less than union membership (change in R2 of 0.018**). State residency accounts for 11 (0.014/0.127) to 21 (0.015/0.070) percent of the total variance explained. Regression coeﬃcients for individual states (not shown) indicate that living in Oklahoma, West Virginia, or Hawaii increases attachment for the Democratic party at a statistically signiﬁcant rate (p = 0.05) while residing in Alaska increases Republican preferences. For ideology, living in the South increases conservatism as the following states have statistically signiﬁcant coeﬃcients: Alabama, Arkansas, Florida, Georgia, Louisiana, Oklahoma, South Carolina, and Virginia. Living in Nebraska also contributes a conservative inﬂuence on ideological identiﬁcation. Public attitudes on three abortion questions are explored in Table 3.2. In this analysis, the core attitudes of partisanship and ideology are included as predictive variables. Demographic variables alone explain from 4.9 percent (parental consent) to 7 percent (abortion position) of the overall variance in individual attitudes. Core attitudes contribute an additional 3.3 to 5.4 percent of explained variance. State residency explains an additional 1.9 to 2.7 percent of the variation in abortion attitudes. Living in Kentucky, Mississippi, South Dakota, or West Virginia increases the chances that an individual will express a pro-life opinion while living in Arizona, Connecticut, Maine, New Hampshire, New Jersey, Nevada, and New York increases the likelihood that one will hold a pro-choice opinion. More conservative attitudes on government funding of abortions for poor women are found in Louisiana and South Dakota while more liberal attitudes are found in Nevada and Rhode Island. No state had a regression coeﬃcient signiﬁcant at the 0.05 level in the parental consent model, but at the 0.10 level, residents in Indiana, Nevada, and Tennessee have more conservative opinions and residents of Maine, Oregon, and Vermont have more liberal opinions than their demographic traits or core attitudes would predict. Demographic traits and core attitudes are more important than state residency in explaining abortion attitudes, but state residency accounts for 18 to 20 percent of the total variation explained. Table 3.3 examines the same three sets of variables in explaining public opinion on capital punishment and government aid to minorities. For attitudes on capital punishment, state residency explains as much of the variation in individual-level opinions as do the two core attitudes of party identiﬁcation and ideology. (In fact, state residency explains a smidgeon

Table 3.2 The Influence of Demographic Traits, Core Attitudes, and State Residency on Attitudes on Abortion,Abortion Funding, and Parental Consent Abortion Funding

Abortion b (SE)

Demographics Education Homemaker Working Unemployed Retired Student Female Married White Black Native American Asian American Latino No Religion Conservative Christian Jew Catholic Union Age Income Core Attitudes Party identification Ideology States Total R2 Number of cases

R2

%

b (SE)

R2

Parental Consent %

b (SE)

R2

%

–0.04** (0.01) 0.13 (0.11) –0.08 (0.10) –0.02 (0.12) –.15 (.11) –0.25 (0.13) –0.07* (0.03) 0.10** (0.03) –0.14 (0.14) 0.12 (0.15) –0.06 (0.20) 0.21 (0.20) 0.07 (0.09) –0.32** (0.07) 0.33** (0.04) –0.33* (0.17) 0.31** (0.04) –0.01 (0.04) 0.005** (0.001) –0.04** 0.070** 51% (0.01)

–0.06** (0.01) 0.31 (0.17) 0.16 (0.15) 0.15) (0.19 0.19 (0.16) 0.16 (0.20) 0.27** (0.05) 0.13* (0.05) 0.49* (0.20) 0.35 (0.22) 0.36 (0.30) 0.87** (0.30) 0.07 (0.14) –0.22* (0.09) 0.70** (0.09) –0.49* (0.22) 0.38** (0.06) 0.12 (0.07) –0.00 (0.00) –0.02 0.064** 45% (0.01)

–0.07** (0.01) 0.36* (0.16) 0.21 (0.14) –0.14 (0.18) 0.22 (0.15) 0.22 (0.19) –0.17* (.05) 0.15** (0.05) –0.08 (0.19) –0.18 (0.22) 0.07 (0.29) –0.09 (0.29) –0.08 (0.13) –0.41** (0.09) 0.15 (0.09) –0.51* (0.22) 0.25** (0.06) –0.03 (0.06) –0.002 (0.002) –0.03* 0.049** 49% (0.01)

0.03** (0.01) 0.14** (0.01)

0.10** (0.01) 0.17** (0.02)

0.05** (0.01) 0.15** (0.02)

0.039** 29% 0.027** 20% 0.136**

6,517

* p d 0.05; ** p d 0.01 Source: Senate National Election Study.

0.054** 38% 0.025** 18% 0.142**

4,378

0.033** 33% 0.019** 19% 0.100**

4,380

Table 3.3 The Influence of Demographic Traits, Core Attitudes, and State Residency on Attitudes on Capital Punishment and Aid to Blacks Capital Punishment b (SE)

Demographics Education Homemaker Working Unemployed Retired Student Female Married White Black Native American Asian American Latino No religion Conservative Christian Jew Catholic Union Age Income Core Attitudes Party identification Ideology States Total R2 Number of cases

R2

–0.03** (0.01) –0.09 (0.15) 0.02 (0.13) 0.02 (0.17) –0.01 (0.14) –0.06 (0.17) –0.33** (0.04) 0.19** (0.05) 0.21 (0.17) –0.55** (0.20) –0.03 (0.26) 0.31 (0.26) 0.01 (0.12) –0.21** (0.08) 0.01 (0.08) –0.50** (0.19) –0.16** (0.05) 0.06 (0.06) –0.004* (0.002) –0.003 0.059** (0.011) 0.07** (0.01) 0.09** (0.01)

4,338

* p d 0.05; ** p d 0.01 Source: Senate National Election Study.

Aid to Blacks %

53%

0.0259**

23%

0.0263** 0.111**

24%

b (SE)

–0.02* (0.01) –0.24 (0.16) –0.13 (0.14) 0.04 (0.17) –0.07 (0.15) –0.65** (0.18) –0.10* (0.04) 0.07 (0.05) 0.50** (0.19) –1.09** (0.21) 0.38 (0.27) 0.48 (0.28) –0.18 (0.13) 0.07 (0.09) –0.20** (0.05) 0.23 (0.24) 0.19** (0.06) 0.09 (0.06) 0.001 (0.002) 0.005 (0.011) 0.06** (0.01) 0.16** (0.01)

6,538

R2

%

0.060**

51%

0.035**

30%

0.024** 0.118**

20%

46

Norrander and Wilcox

more.) State residency and these core attitudes each contribute one-fourth to the total variation explained. State residency accounts for 2.4 percent of the total variation in opinions on government aid to blacks but accounts for 20 percent of the explained variance. Residents of Florida and Oklahoma are more likely to express support of the death penalty than their demographic and core attitude background would predict, and residents of Vermont are more liberal.At the 0.10 level of signiﬁcance, residency in Illinois, Missouri, and Utah increases support for the death penalty and living in Maine, North Dakota, and Rhode Island decreases support. Residency in many southern states (Alabama,Arkansas, Florida, Georgia, Louisiana, Maryland, Mississippi, North Carolina, Oklahoma, South Carolina, and Tennessee) and a variety of other states (Idaho, Illinois, New Jersey, Nevada, and Wyoming) leads to more conservative attitudes on government aid to minorities while only Vermont residency appears to increase liberalism on this attitude.

Understanding the Inﬂuence of State Residency Results from the analyses in Tables 3.1 through 3.3 indicate state residency has a signiﬁcant impact on individual-level public opinion. But what is the source of this inﬂuence? One possibility is that state residency is simply picking up missed measures of demographic variation. However, it is unlikely that untapped demographic variation would consistently match state residency. State residency may impact public opinion because states have a speciﬁc political milieu, whether it is political culture or political history.The concept of political culture, however, is often used as a catchall for other unmeasured eﬀects.Another possibility is that state residency reﬂects concrete political conditions in a state such as variations in state laws. Laws may provide a legitimation eﬀect to one side of a public opinion issue. It is this last point that can be most clearly tested. State laws on and usage of capital punishment have been remarkably stable over time, despite the disruption caused by the 1972 Supreme Court ruling in Furman v. Georgia. Most states that employed capital punishment prior to the ruling simply rewrote their laws to conform to the newer standards. After the 1970s, few states changed their position on capital punishment and execution rates from the pre- and post-Furman eras are highly correlated, with Pearson’s r = 0.61, signiﬁcance =0.01 (Norrander, 2000). Given the longevity of state policies on capital punishment, such laws are more likely to have a legitimation eﬀect. To analyze the contribution of state laws to the explanation given by state residency to public opinion, a dummy variable was developed for the presence or absence of a death penalty law in each state in 1989. (The cap-

State Residency, State Laws, and Public Opinion 47

ital punishment question was ﬁrst asked in the 1990 wave of the SNES.) The state law variable was added to the model containing demographic traits, core attitudes, and state residency variables.2 To avoid collinearity between the dummy variables for state laws and state residency, the median state (in terms of public opinion) in each category of the state law variable was excluded: Oregon for states having death penalty laws and Wisconsin for states without such laws.The model was tested by sequentially adding demographic traits, core attitudes, state laws, and ﬁnally state residency. Changes in explained variances (R2) are used to demonstrate the importance of each component to the model. Results in Table 3.4 show that the presence of a state law contributed a small but statistically signiﬁcant portion of the model’s ability to explain individual-level opinions on capital punishment (change in R2 due to the law = 0.007**). State laws contributed about 6 percent (0.007/0.111) to the overall explained variance. More important, state laws constitute about one-fourth of the explanation that previously came from state residency. Twenty-seven percent (0.007/0.026) of a state’s inﬂuence on public opinion is due to state law, while 73 percent (0.019/0.026) is attributable to other state characteristics. Abortion laws in the states are of more recent vintage. After Roe v.Wade overturned state bans on abortion, states enacted a number of laws qualifying the right to an abortion.These laws date from 1975 and cover topics such as prohibiting use of government funds for abortions for poor women, parental consent for minors, spousal consent, and various measures intended to constitute informed consent, such as waiting periods or distribution of printed materials. Abortion laws may have less inﬂuence on public opinion because of their shorter history. Speciﬁc abortion laws also may be less well known to the public than capital punishment laws. Pending executions and death penalties imposed at the conclusion of murder trials are stories covered on the nightly news. Abortion laws are publicized on enactment but are unlikely to be discussed at other times.With less discussion of these laws, they may have a lesser legitimation eﬀect. However, the public may have a vague sense of how much abortion is regulated in their states and this may inﬂuence public opinion. Four abortion laws were coded for their presence in 1987, the year prior to the ﬁrst asking of the abortion questions in the SNES. Dummy variables were created for government funding bans and parental consent laws.These match the two speciﬁc abortion policy questions asked in SNES. For analyzing general abortion attitudes, an abortion law index was formed by summing the number of the four abortion laws in place in a state in 1987.To prevent collinearity, the median state for the “one” and “zero” categories of the

48

Norrander and Wilcox Table 3.4 The Effect of State Laws, State Residency, Demographic Traits, and Core Attitudes on Public Opinion on Capital Punishment Without State Laws

Capital Punishment Demographic traits Core attitudes State laws State residency Total R2 Government Funding of Abortion Demographic traits Core attitudes State laws State residency Total R2 Parental Consent for Abortions Demographic traits Core attitudes State laws State residency Total R2 Abortion Law Index Demographic traits Core attitudes State laws State residency Total R2

0.059** 0.026** 0.026** 0.111** 0.064** 0.054** 0.025** 0.142** 0.049** 0.033** 0.019** 0.100** 0.070** 0.039** 0.027** 0.136**

With State Laws

0.059** 0.026** 0.007** 0.019** 0.111** 0.064** 0.054** 0.004** 0.021** 0.142** 0.049** 0.033** 0.000 0.019** 0.100** 0.070** 0.039** 0.007** 0.021** 0.136**

Entries are R2 and change in R2. * p d 0.05; ** p d 0.01 Source: Senate National Election Study.

dummy variables were excluded (funding = Alaska and Arkansas; parental consent = Georgia and New Mexico) and for the abortion index, the median states for the “zero” versus all other categories were excluded (Virginia and Connecticut). The bottom section of Table 3.4 presents results for analysis of abortion laws.The index of state laws can account for 26 percent (0.007/0.027) of the inﬂuence of state residency on general abortion attitudes. This is an eﬀect similar in magnitude to that found for capital punishment laws. Laws on government funding of abortion have a smaller inﬂuence on public opinion, accounting for 16 percent (0.004/0.025) of the inﬂuence of state residency. Finally, laws on parental consent for abortions have no inﬂuence on state opinion.The public may be unaware of the speciﬁc abortion laws in a state, which accounts for the lack or lesser inﬂuence for parental consent and gov-

State Residency, State Laws, and Public Opinion 49

ernment funding laws. However, a general sense of abortion laws may be guiding general abortion attitudes.

Discussion State residency was demonstrated to be a small but important determinant of public opinion on a host of issues, from core attitudes such as partisanship and ideology, to opinions on the speciﬁc issues of abortion, capital punishment, and government aid to minorities. State residency adds 10 to 25 percent to the explained variation in individual-level opinion, while demographic factors contribute approximately half of the explained variation. The core attitudes of partisanship and ideology contribute from one-fourth to one-third of the explanation of attitudes on the speciﬁc issues. A variety of reasons may account for the relationship between state residency and public opinion. In this analysis, the inﬂuence of state laws was tested. With only a single measure of public opinion and state laws, it is diﬃcult to untangle the direction of inﬂuence. A longer time perspective allows for the historical chain of interactions between public opinion and public policy to be untangled (Norrander, 2000). Nevertheless, analyses presented in this research provide a maximum inﬂuence of state laws on public opinion. If the causal direction is solely from state laws to public opinion, in a legitimation eﬀect, state laws account for up to 25 percent of the inﬂuence of state residency on public opinion. The causal arrow may go in the other direction, despite the fact that state laws were measured prior to public opinion. If public opinion in a state is relatively stable, then our measure of public opinion in the 1990s may reﬂect public attitudes in the states in the 1970s when the capital punishment and abortion laws were being rewritten in response to Supreme Court cases. These earlier attitudes may have helped to shape laws in the state and these earlier attitudes continue to shape public opinion today. Nevertheless, even in this last scenario where prior public opinion inﬂuences both policy and current opinion, public policy still plays a role in cementing public opinion in a state. Public laws contribute to the continuity in public opinion over time by providing a legitimation eﬀect. Not all laws will have equal eﬀect on public opinion.When public laws are such that their implementation is an ongoing news story, such as capital punishment sentencing, these laws will have a greater opportunity to inﬂuence public opinion. When the implementation of public laws is a more quiet, private activity, such as parental consent for abortions, the

50

Norrander and Wilcox

public may be less aware of the existence of the law, and the law will have little to no inﬂuence on public opinion.

Notes 1. Measures of demographic variables were constructed along the following lines. Education is number of years of schooling (vPS0606). Income (vPS0629) and age (vPS0604) were unchanged from their original format. Job status categories (vPS0613) were homemaker (codes 70 and 75), working (codes 10 though 18), unemployed (20, 40), retired (50), and student (80). Disabled is the excluded category. Marital status designates those currently married (code 1 in vPS0605) from all others.With a large number of respondents in the SNES, racial categories (vPS0631) were created for whites, blacks, Native Americans, and Asian Americans.The excluded category is “other” races. Latino designation includes all those with any Hispanic background (vPS0632 codes 1 through 4). Four religion categories were created from the denomination variable (vPS0630): no religion (codes 800, 801, 995), conservative Christians (i.e., evangelicals, Pentecostals, fundamentalists, and Mormons, codes 20, 120 through 128, 133 through 148, 180 through 219, 221, 222, 250 through 269, and 301), Jews (500 through 503), and Catholics (400). The excluded category represents mainly mainstream Protestants. Union membership includes anyone in the family setting (vPS0617). All attitude questions were recoded such that a high response would indicate a conservative opinion.To maintain more respondents in the analysis, the summary ideology variable (vPS0550) is used.This measure includes those who were reluctant to designate an ideological label in the ﬁrst question but did select a label in response to a second question. 2. The source of state laws on capital punishment is U.S. Department of Justice, Capital Punishment series. Abortion laws are taken from National Abortion and Reproductive Rights Action League, Who Decides: A State-by-State Review of Abortion Rights,Washington, DC:The NARAL Foundation, 1988.

References Abramowitz,Alan I., and Kyle L. Saunders. 1998.“Ideological Realignment in the U.S. Electorate.” Journal of Politics 60 (August): 634–652. Arceneaux, Kevin. 2001. “The ‘Gender Gap’ in State Legislative Representation: New Data to Tackle an Old Question.” Political Research Quarterly 54 (March): 143–160. Berger, Ronald J., W. Lawrence Neuman, and Patricia Searles. 1991. “The Social and Political Context of Rape Law Reform: An Aggregate Analysis.” Social Science Quarterly 72 ( June): 221–238. Berman, David R., and Lawrence L. Martin. 1992. “The New Approach to Economic Development: An Analysis of Innovativeness in the States.” Policy Studies Journal 20 (Number 1): 10–21. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson.

State Residency, State Laws, and Public Opinion 51 1998. “Measuring Citizen and Government Ideology in the American States, 1960–1993.” American Journal of Political Science 42 ( January): 337–348. Brace, Paul, Laura Langer, and Melinda Gann Hall. 2000. “Measuring the Preferences of State Supreme Court Justices.” Journal of Politics 62 (May): 387–413. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46 ( January): 173–189. Burden, Barry C., and Steven Greene. 2000.“Party Attachments and State Election Laws.” Political Research Quarterly 53 (1 March): 63–76. Call, Jack E., David Nice, and Susette M.Talarico. 1991.“An Analysis of State Rape Shield Laws.” Social Science Quarterly 72 (December): 774–788. Campbell, James E. 1992. “Forecasting the Presidential Vote in the States.” American Journal of Political Science 36 (May): 386–407. Cook, Elizabeth Adell,Ted. C. Jelen, and Clyde Wilcox. 1993a.“State Political Cultures and Public Opinion About Abortion.” Political Research Quarterly 46 (December): 771–782. ———. 1993b.“Catholicism and Abortion Attitudes in the American States:A Contextual Analysis.” Journal for the Scientiﬁc Study of Religion 32 (September): 223– 230. Elazar, Daniel. 1966. American Federalism:A View from the States. New York:Thomas Y. Crowell. Erikson, Robert S., John P. McIver, and Gerald C.Wright, Jr. 1987.“State Political Culture and Public Opinion.” American Political Science Review 81 (September): 797–813. Finkel, Steven E., and Howard A. Scarrow. 1985. “Party Identiﬁcation and Party Enrollment: The Diﬀerence and Consequence.” Journal of Politics 47 ( June): 620–642. Holbrook,Thomas M. 1991. “Presidential Elections in Space and Time.” American Journal of Political Science 35 (February): 91–109. Jackson, Robert A., and Thomas M. Carsey. 1999a. “Group Components of US Presidential Voting Across the States.” Political Behavior 21 ( June): 123–151. ———. 1999b. “Presidential Voting Across the American States.” American Politics Quarterly 27(October): 379–402. Johnson, Cathy Maries, and Kenneth J. Meier. 1990. “The Wages of Sin: Taxing America’s Legal Vices.” Western Political Quarterly 43 (September): 577–596. Jones, Bradford S., and Barbara Norrander. 1996. “The Reliability of Aggregated Public Opinion Measures.” American Journal of Political Science 40 (February): 295–309. Lowery, David, and Lee Sigelman. 1982. “Political Culture and State Public Policy: The Missing Link.” Western Political Quarterly 35 (September): 376–384. Meier, Kenneth J. 1992. “The Politics of Drug Abuse: Laws, Implementation, and Consequences.” Western Political Quarterly 45 (March): 41–70. Meier, Kenneth J., and Cathy M. Johnson. 1990. “The Politics of Demon Rum: Regulating Alcohol and Its Deleterious Consequences.” American Politics Quarterly 18 (Number 4): 404–429. Miller,Warren E., Donald R. Kinder, Steven J. Rosenstone, and the National Elec-

52

Norrander and Wilcox

tion Studies. 1999. American National Election Study: Pooled Senate Election Study, 1988, 1990, 1992 [Computer ﬁle]. 3rd ed. Ann Arbor, MI: University of Michigan, Center for Political Studies [producer]. Inter-university Consortium for Political and Social Research [distributor]. Mooney, Christopher Z., and Mei-Hsien Lee. 2000. “The Inﬂuence of Values on Consensus and Contentious Morality Policy: US Death Penalty Reform, 1956– 82.” Journal of Politics 62 (February): 223–239. National Abortion and Reproductive Rights Action League. 1988. Who Decides: A State-by-State Review of Abortion Rights. Washington, DC: The NARAL Foundation. Nice, David C. 1991.“The Impact of State Policies to Limit Debt Financing.” Publius 21 (Number 1): 69–82. ———. 1992.“The States and the Death Penalty.” Western Political Quarterly 45 (December): 1037–1048. Norrander, Barbara. 1989.“Explaining Cross-State Variation in Independent Identiﬁcation.” American Journal of Political Science 33 (May): 516–536. ———. 2000. “The Multi-Layered Impact of Public Opinion on Capital Punishment Implementation in the American States.” Political Research Quarterly 53 (December): 771–793. ———. 2001.“Measuring State Public Opinion with the Senate National Election Study.” State Politics & Policy Quarterly 1 (March): 113–127. Norrander, Barbara, and Clyde Wilcox. 1999. “Public Opinion and Policymaking in the States: The Case of Post-Roe Abortion Policy.” Policy Studies Journal 27 (December): 707–722. Pool, Ithiel de Sola, Robert P.Abelson, and Samuel L. Popkin. 1965. Candidates, Issues, and Strategies. Cambridge, MA: MIT Press. Shelley, Fred M., and J. Clark Archer. 1994.“Political Geography of Contemporary Aﬀairs: Some Geographic Aspects of the American Presidential Election of 1992.” Political Geography 13 (March): 137–159. Sullivan, John L., James E. Piereson, and George E. Marcus. 1978. “Ideological Constraint in the Mass Public: A Methodological Critique and Some New Findings.” American Journal of Political Science 23 (May): 233–249. Weber, Ronald E.,Anne H. Hopkins, Michael L. Mezey, and Frank Munger. 1972– 1973. “Computer Simulation of State Electorates.” Public Opinion Quarterly 36 (Winter): 49–65. Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship and Ideology with Survey Data.” Journal of Politics 47 ( June): 469–489.

Chapter

Mexican-American and Cuban-American Public Opinion: Diﬀerences at the State Level?

4

David L. Leal

Introduction Latinos are a pervasive topic of discussion in the new century. On radio and television, and in newspapers, magazines, and academic journals alike, an interest in the Latino community is increasingly audible and visible. Politicians, business executives, movie directors, and academics are all thinking about the implications of this growing population for their own professions. Many of these people are asking the same question—What do Latinos want?—so that their company, party, or organization can sell commercial or political goods to this group. Others worry about the political, cultural, and economic implications of immigration. Such attention both positive and negative is unlikely to dissipate anytime soon. A key question, however, is whether a “Latino community” exists in the United States. Concepts like “Latino” or “Hispanic” are social constructions that have not necessarily been adopted by the populations in question.A number of researchers are interested in whether “Latinos” think of themselves in such pan-ethnic terms, or whether they prefer nationalorigin identiﬁers such as Mexican American or Cuban American (de la Garza et al., 1992; Jones-Correa and Leal, 1996). In response to such concerns, a number of scholars have looked in more detail within the Latino label and found a number of important diﬀerences by national-origin group. For instance,Trueba (1999, 33) noted that, “we cannot trivialize the ethnic, social, racial, and economic diﬀer53

54

Leal

ences of Latino subgroups.” De la Garza and DeSipio (1994, 3) argued that the study of the aggregate Latino population: confuses rather than clariﬁes our understanding because of the characteristics that distinguish the national-origin groups thus subsumed (Bean and Tienda, 1987; de la Garza, Fraga, and Pachon, 1988; Pachon and DeSipio, 1988; Fuchs 1990). This approach fails to assess diﬀerences in political culture associated with the distinct socialization experiences within the United States and the countries of origin and neglects the link between those diﬀerences and political behavior.

A key event in this research agenda was the creation of the Latino National Political Survey (LNPS) in 1989–1990, which was speciﬁcally designed to compare the political and policy attitudes of Latinos of Mexican, Puerto Rican, and Cuban descent. The investigators (de la Garza et al., 1992, 13) found a number of important diﬀerences: What, then, does the LNPS tell us? Particularly among citizens, it illustrates that Mexicans, Puerto Ricans, and Cubans have a great deal in common, but they also diﬀer in signiﬁcant ways on important issues.Thus, there may be a Hispanic political community, but its parameters do not ﬁt any existing presuppositions.

In his review of the above volume, Domínguez (1994) made two similar points. First, he noted that,“On question after question, the circumstances and attitudes of Cubans place themselves outside any U.S. Latino referent that might apply to the Mexicans and Puerto Ricans” (354). He also discussed, however, “three credible scholarly claims about the appropriateness, necessity, and utility of the concept of Latino” (354). Due to such research, it is now increasingly clear that the aggregate term Latino must be used with caution. Many but not all scholars are now careful to avoid sweeping statements about Latino political behavior or public opinion, and much work is now focused on the experiences of particular national-origin groups. Recent population trends have also contributed to this scholarly trend.Although words like Latino and Hispanic are often used interchangeably with “Mexican American,” the number of non-Mexicanorigin Latinos has grown dramatically in recent decades. This is in large part due to substantial migration from Central America and the Caribbean. The next step in this research agenda is to delve more deeply into the validity of the national-origin group categories. Establishing the fact of Latino-Anglo opinion diﬀerences does not indicate whether the categories of Latino (or even Anglo) are meaningful; in a similar way, opinion variation according to Latino national-origin group does not indicate that such groups are coherent entities. It is possible that aggregate nationalorigin data mask a number of diﬀerences within such communities, perhaps to the point that it makes little sense to discuss “Mexican-American opinion” or “what the Cuban-American community wants.”This chapter

Mexican-American and Cuban-American Public Opinion 55

explores such diﬀerences by using a national survey of Latinos that allows for the state-level analysis of opinion, speciﬁcally with regard to Mexican Americans and Cuban Americans. There do exist some hints about the potential importance of variation within such groups. According to Los Angeles Times journalist Frank del Olmo (1987, 25),“I have found that polls attempting to gauge Latino sentiment on issues or candidates seem to be more reliable when they focus on a speciﬁc national group like Cuban Americans or Puerto Ricans, or on a particular community like San Antonio or East Los Angeles, than when they deal with Latinos as a national bloc.” This statement indicates that opinion diﬀerences might be expected among Mexican Americans in California and Texas—a proposition this chapter examines in detail. Others indications show that other Latino national-origin groups might not be unitary actors. Domínguez (1994, 354) noted that the LNPS excluded Puerto Ricans living in Puerto Rico and argued that, “it would have been valuable to compare the two Puerto Rican subpopulations.” Some comparative work on this population has taken place, particularly comparing the political participation rates of Puerto Ricans in the United States and on the island (all of whom are American citizens), but less work is found on public opinion.Although it does not make sense for this project to compare the attitudes of Cuban Americans in the United States with the citizens of Cuba, for instance, we can examine whether Cuban Americans in Florida express diﬀerent opinions than do those living elsewhere in the United States. More generally, scholars have noted the importance of studying variation within Latino national-origin groups. Keefe and Padilla (1987, 140), in discussing Latino and Anglo familism, wrote that, “The signiﬁcant diﬀerences within the Mexican American ethnic group demonstrate the importance of conducting subgroup analyses before making generalizations for the ethnic group as a whole.” The chapter presents public opinion data in both aggregate form and via regression analysis. The former points out diﬀerences according to state residence, but does not indicate the factors on which any such diﬀerences are based. If Mexican Americans in California are more likely than other Mexican Americans to support government health care programs, this might indicate a more liberal California population that believes in the utility of big government. On the other hand, it could reﬂect varying income levels by state, as respondents with lower incomes might be particularly interested in government-provided health beneﬁts. There are a number of reasons to expect state-level opinion variation within the Mexican-American population.This is due to the varying state political contexts in California and Texas during the 1990s. In that decade,

56

Leal

California saw an outpouring of anti-immigrant feeling that many believe was a thinly veiled attack on Latinos. According to Pantoja, Ramirez, and Segura (2001, 735),“The political climate for Latinos in California through much of the mid-1990s was described by some observers as hostile, and by nearly all as tense.” The policy changes sought by voters in this time period included the banning of state services for undocumented immigrants (Proposition 187), a requirement for state employees to report the undocumented to the INS (Proposition 187), the banning of aﬃrmative action (Proposition 209), and the banning of bilingual education (Proposition 227). Although some Latinos did support these measures, the clear majority was opposed. Scholars have noted the consequences of these events for Latino voter registration and turnout (Barreto and Woods, 2000; Pantoja, Ramirez, and Segura, 2001; Scott, 2000), but less is known about the ideological consequences.We do know that much of the subsequent increased Latino registration and voting activity in California was to the beneﬁt of the Democratic party. Previous Republican gains were erased as Democrats would largely sweep races for statewide oﬃces and reestablish clear control of the state legislature. Such events did not happen in all states, however. In Texas, which is often thought to be the most conservative of the states with signiﬁcant Hispanic populations, such measures were not seriously debated, let alone enacted into law. On the contrary, important (and conservative) political ﬁgures actively opposed such measures and prevented the California climate from infecting the Lone Star state.Although a variety of explanations have been advanced, such as the lack of a ballot initiative process in Texas or the large number of Latino elected oﬃcials at local levels, the key fact is that the anti-immigrant dog did not bark in Texas. For example, Pastor and Fernandez de Castro (1998, 189) noted that: In August 1995 Texas Governor George W. Bush warned prospective U.S. presidential candidates that he would not permit them to engage in immigrant bashing while campaigning in Texas . . . At the same time, California Governor Pete Wilson, having ridden an anti-Mexican issue, Proposition 187, to reelection in California, based part of his presidential bid on anti-immigrant rhetoric.

This state of aﬀairs may have led Mexican Americans in Texas to become more supportive (or less opposed) to the Republican party, which in turn may have led to more aﬃnity for (or less opposition toward) the generally conservative policy positions of this party. The argument is diﬀerent for Cuban Americans. For this population, much of the city of Miami functions as an “ethnic enclave.” Although

Mexican-American and Cuban-American Public Opinion 57

Cuban Americans reside in large numbers in other parts of the United States, such as New Jersey, the ethnic enclave experience is primarily found in south Florida. According to Moreno (1997, 211–212), “More than 500,000 immigrated to the United States under the Cuban Adjustment Act (Perez 1992, 85). Most Cuban exiles settled in Miami or returned to the city after ﬁrst settling in the North. Another large Cuban community of about 80,000 was established in the area around Union City, New Jersey, while smaller communities exist in Chicago and Los Angeles.” An ethnic enclave consists of more than just a large number of people from a speciﬁc national-origin group. This phrase is often discussed in terms of economics, as when Portes and Rumbaut (1990, 21) wrote that, “These areas of concentrated immigrant entrepreneurship are known as ethnic enclaves.”They noted the importance of a concentration of immigrant business expertise, access to capital, and access to labor.Without such features, the Cuban-American community might have become incorporated into the existing Anglo economic structure in Miami. This has important social and cultural implications. Moreno (1997, 213) discussed “the enclave’s most important overall feature: institutional completeness (Perez 1992, 90–91). Cubans in Miami can, if they wish, live out their lives within the community.” Population growth may be a necessary but not suﬃcient condition for the development of an “enclave,” however. Sometimes a crucial event is needed to act as a catalyst. In Florida, Portes and Stepick (1993, 35) discussed the economic and political changes caused by the Mariel events of 1980. It was not just the additional 125,000 Cubans, but the strongly negative reaction of the Anglo establishment, that led to signiﬁcant changes within the Cuban-American community. Before 1980, the latter was politically unorganized and economically integrating into the existing business structure.The wave of criticism of the Mariel newcomers, as well as the successful anti-bilingual referendum of that year, forced Cuban Americans to “turn their attention inward and confront their condition as a domestic ethnic minority” (33).This process of “reactive formation” (34) led to the creation of economic institutions parallel to existing Anglo institutions, a new determination to seize political power locally and statewide, and the creation of organizations such as the Cuban-American National Foundation (CANF).This helped to build and fortify the unique Miami enclave so often discussed today. Few have attempted to understand whether the political dynamics of this region have unique opinion consequences for Cuban Americans visà-vis those Cuban Americans living in other parts of the United States. The signiﬁcant role of the politicized exile community in Florida means

58

Leal

that Cuban-American politics in that region revolve around Castro and Cuba. Observers believe this has oriented the population toward the Republican party, as the latter has more consistently adopted anti-Castro and anti-communist stances. This support for the more conservative political party may also aﬀect how Cuban Americans in Florida view other policy issues. As the LNPS found, Cuban Americans are not particularly conservative as a group on many questions. Domínguez (1994, 355) noted that the Latino subgroup most supportive of abortion rights, most supportive of bilingual education, and least supportive of increases in defense spending was Cuban Americans. As the Republican party espouses more conservative views on these and many other issues, it is possible that support for Republicans may have led Cuban Americans in Florida to be relatively more favorable toward such views. Outside of Florida and the enclave, Cuban Americans may be less interested in Castro, the embargo, and the politics of the Republican party.As Welch and Sigelman (1993, 77) noted,“there are some indications that as they stay longer in the United States and migrate outside of Miami their Republican sympathies may fade” (Boswell and Curtis, 1983; Brischetto, 1987;Torres, 1988).

Latinos and the Political Future As the size and percentage of the nation’s nonwhite population grows, it is increasingly diﬃcult to understand American politics without reference to Latinos and African Americans. One key demographic trend is the declining share of the Anglo population. This is primarily due to the growing Latino population, which is reshaping the political and cultural landscape. As a result, the traditional and straightforward bifurcated blackwhite paradigm on racial questions is slowly changing into a more complex black-white-Hispanic perspective. The 2000 U.S. Census found that Latinos had become America’s largest minority group. Although the Census Bureau in 1999 estimated that Latinos would be 11.4 percent of the population in 2000, the census revealed that they were 12.5 percent—a ﬁgure Latinos were not projected to reach until 2005.This may understate the true Latino population, as the census is an actual count, not an estimate, and many Latinos are noncitizens with incentives to avoid contact with government oﬃcials.These ﬁgures for the ﬁrst time put Latinos ahead of African Americans, who constituted 12.1 percent of the population.The most recent data from the Bureau of the Census indicate that the Latino population has grown (in just three years) to 13.4 percent

Mexican-American and Cuban-American Public Opinion 59

Given the substantial levels of Latino immigration, as well as the relatively high birthrate of Latinos in the United States, this population will likely continue to grow. Anglos currently constitute 68 percent of the U.S. population, and in a number of states they are shifting from majority to plurality status. By 2050, at least one in every four Americans will likely be Latino, and by 2100, one in every three Americans may be Latino.“Minority” attitudes will therefore be of critical importance as the U.S. transitions to a society in which no single ethnic or racial group is the majority.

Latinos and Public Opinion The political science understanding of Latino public opinion is much less developed than that of Anglo public opinion. One key reason is the historic lack of interest in Latino opinion and the concomitant desultory eﬀorts to include Latino respondents in major polls.This was particularly true in the early decades of postwar public opinion polling, but this situation continued into the 1970s and 1980s and still exists today in some quarters. The most prominent polling project in political science, the American National Election Study (ANES), does not interview in Spanish, which creates a problematic Latino sample in a number of important ways. The ﬁrst attempt to collect national-level survey data on political attitudes for even a single Latino national-origin group was in 1977. The Chicano Survey (Arce, 1997) included 991 Mexican-American respondents, of whom 667 were citizens. The survey focused on respondents’ demographic characteristics, employment history, health issues, and social identity, with only a smaller cluster of questions examining political opinions, attitudes, and participation. The sample, representative of an estimated 90 percent of the Mexican-American population living in the United States at the time, was drawn from ﬁve states (California, Texas, New Mexico, Colorado, and Arizona) and the city of Chicago. It was not until 1989 that political scientists would attempt to survey a national sample of Latinos. This project, the Latino National Political Survey (LNPS), included respondents from the three largest nationalorigin groups: Mexican Americans, Puerto Ricans, and Cuban Americans. It gathered information on a wide range of political activities, preferences, and behaviors among Hispanics in the United States. Unlike the 1979 Chicano Survey, the focus of the LNPS was expressly political.The survey population, randomly selected from 40 standard metropolitan statistical areas (SMSAs) across the United States, was representative of 91 percent of the nation’s Latinos (de la Garza et al., 1992, 7).

60

Leal

These eﬀorts notwithstanding, by the end of the 1980s (the so-called Decade of the Hispanic), there existed only scattered data on Latino opinion. This largely consisted of uneven exit polls conducted by news organizations and occasional state and local surveys conducted by academics.1 This situation caused de la Garza (1985, 1) to lament: For whom did Latinos vote in the 1984 elections? What were their views on the issues and candidates in that election? Surprisingly, reliable answers to these and related questions are unavailable. No independent source—not the national press, not the nation’s leading public opinion pollsters, no one—systematically asked Latinos their views or monitored their participation in the 1984 election.

Why is such inattention by the ﬁeld of public opinion consequential? As de la Garza (1985, 4) further argued, “Polls can and do inﬂuence candidate selection and issue resolution. By not having their views regularly reported in polls, Latinos are eﬀectively excluded from inﬂuencing both of these outcomes. In a fundamental sense, because of the role that polls play in our political life, being excluded from them is tantamount to partial disenfranchisement.” Due to the lack of data, many scholars who are interested in comparatively studying Latino opinion have only been able to use surveys that encompass a speciﬁc population or locale (Cain and Kiewiet, 1987; de la Garza, 1985; de la Garza, Polinard, Wrinkle, and Longoria, 1991; de la Garza and Weaver, 1985; Lovrich, 1974; Miller, Polinard, and Wrinkle, 1984; Uhlaner, 1991). The most comprehensive eﬀort is found in the volume that resulted from the aforementioned LNPS, titled Latino Voices (de la Garza et al., 1992). It broadly concluded that, “On many key domestic issues, signiﬁcant majorities of each [Latino] group take the liberal position. On other issues, there is no consensus and, depending on the issue, Mexicans may be on the right, while Cubans and many Puerto Ricans are on the left of the nation’s current political spectrum. Thus, labels such as liberal or conservative do not adequately describe the complexity of any one group’s political views” (15). When scholars investigate opinion diﬀerences between Latinos and other groups, the source is typically aggregate survey data. Such information is only partially useful to social scientists, however, as many intervening variables are likely to aﬀect public opinion. Latinos systematically diﬀer from Anglos in terms of income, education, age, and other factors, and if there were Latino-Anglo opinion diﬀerences, perhaps the former would not express opinions diﬀerent than their socioeconomic status (SES) would suggest.The same theory would also apply to the analysis of Latino public opinion across the states. There may be diﬀerences within

Mexican-American and Cuban-American Public Opinion 61

particular groups according to state residence, but the basis of such diﬀerences might only be varying state SES. Although this is a secondary question for our chapter, which is only an initial exploration of the topic, it is nonetheless helpful in understanding whether any distinctive political orientations go beyond SES. To investigate state-level opinion variation requires a survey with nationally representative samples of Latino national-origin groups. The research design should include regression models with political and policy opinions as the dependent variables.The independent variables should include state residence, SES, and other demographic factors. In this way we can see if any unique attitudes are captured by the state variables. A long line of research has used a similar approach to determine whether Latino and African-American political participation diﬀers from that of Anglos. It is well established in political science that voting is correlated with factors such as education and age, and to some extent income. Many scholars have wanted to know whether minority political behavior is diﬀerent when such factors are taken into account.2 Political scientists have not often sought, however, to understand minority public opinion in a similar way. Among the few who have done so are Kinder and Sanders (1996), who studied whether and how there is a public opinion divide between Anglos and African Americans for multiple policy questions.They concluded that “blacks and whites are deeply divided over racial policy . . . The diﬀerences are enormous, quite unlike any other social cleavage, and cannot be explained by black-white diﬀerences in income or educational attainment or indeed anything else. The racial divide in political aspirations and demands is really racial” (9). Others who have tested for distinctiveness in Latino public opinion are de la Garza, Falcon, and Garcia (1996), Welch and Sigelman (1993), and Leal (2002a). In a similar manner, this chapter tests for state-level distinctiveness in Mexican-American and Cuban-American policy and political preferences in the aggregate and ceteris paribus.

Data and Models A. Data This chapter uses the 1999 Washington Post/Henry J. Kaiser Family Foundation/Harvard University National Survey on Latinos in America (NSLA). This is a nationally representative sample of 4,614 adult respondents, including 2,417 Latinos. It was conducted by telephone between June 30 and August 30, and includes residents of California, Florida, Illinois, New York,Texas, and the Washington, DC, metro area.

62

Leal

It is the most appropriate survey for the purposes of this chapter because it contains a large number of policy and political questions as well as a large Latino sample. Most national surveys of American political opinion do not include a suﬃcient Latino sample to conﬁdently assess this large and complex population.This chapter also requires a survey designed to sample respondents from speciﬁc national-origin groups, and the NSLA includes respondents of Mexican, Puerto Rican, Cuban, Caribbean, and Central/South American heritage. Some surveys are problematic because they do not use bilingual interviewers.This dramatically underrepresents Spanish-dominant respondents and generates a Latino sample that is biased in a number of important ways. In the Post/Kaiser/Harvard survey, 49 percent of the Latino respondents choose to be interviewed in Spanish, which indicates the importance of this option. The survey includes 818 Mexican Americans, 318 Puerto Ricans, 312 Cuban Americans, and 593 Central/South Americans, although only the Mexican-American and Cuban-American respondents will be included in this analysis. Of the Mexican-American respondents, 215 were from California, 222 were from Texas, and 381 were from other locations. Most of the Cuban-American respondents reside in Florida (263), but there are 49 Cuban Americans from other locales.The sampling margin of error is ±4 percent for Mexican Americans and ±6 percent for Cuban Americans.

B. Dependent Variables The political opinion questions are grouped into several related categories: (1) policies, (2) policies with speciﬁc relevance to Latinos, (3) moral (sometimes referred to as “social”) issues, (4) orientations toward government, and (5) opinions about political parties and ﬁgures.3 Although some research has examined Latino opinion on these questions, rarely do we see any comparisons across states or within national-origin groups. The ﬁrst set of policy questions involves health care and criminal justice issues. The former include HMO reform, government-provided health insurance, and handgun control.The policy questions of particular concern to Latinos are bilingual education and immigration policy. The next set of questions involves attitudes toward moral issues with political implications—speciﬁcally attitudes toward abortion, the death penalty, and assisted suicide, the three life-related issues addressed by Pope John Paul II’s Encyclical Letter Evangelium Vitae (Paul, 1995). In addition, the survey asks three questions that involve traditional morality: the acceptability of divorce, homosexual/lesbian sexual activity, and bearing children out of wedlock.4

Mexican-American and Cuban-American Public Opinion 63

Three attitudes about the U.S. government are also examined.The ﬁrst asks whether Washington or the individual should be responsible for solving problems, the second solicits opinions about the size of government, and the third asks about trust in government. Lastly, the chapter tests two questions about party support and two about presidential evaluations. The former ask about the respondent’s party aﬃliation and which party best looks out for the interests of Latinos. The latter ask whether the respondent supported presidential candidate Al Gore or George W. Bush and for an evaluation of outgoing president Bill Clinton. Although we know that Mexican Americans are more likely to support the Democratic party than are Cuban Americans, we might wonder if Cuban-American support for this party declines outside of Florida. Cuban-American support for George W. Bush might also be higher in Florida because of their strong support for Governor Jeb Bush (who won approximately 80 percent of the Cuban-American vote in his reelection campaign). It is also possible that Mexican-American support for George W. Bush may be greater in Texas, where he developed a reputation for working constructively with Latinos.

C. Independent Variables When Kinder and Sanders (1996) tested whether policy opinions diﬀered by race, they included variables for race, income, education, and occupational status.We followed their lead and include all of these as well as a number of others. In the Mexican-American model, the key independent variables are for Mexican Americans in California and in Texas. The other Mexican Americans in the sample are therefore the base case. In the CubanAmerican models, the key independent variable will be for Cuban Americans in Florida. Just as the minority political participation literature tests whether racial or ethnic diﬀerences remain after the inclusion of multiple control variables, these models test whether respondents from varying regional settings express unique opinions on policy and political questions. Those at the lower end of the SES spectrum should support government programs, particularly those designed to alleviate poverty.We therefore include variables for median family income and education level. A variable for gender is included, although the attitudes of women are complex. They are often considered to be more liberal than men, as they express more liberal opinions on social welfare programs and general ideology (Shapiro and Mahajan, 1986; Deitch, 1988; Erikson, Wright, and

64

Leal

McIver, 1993). They are also more conservative than men on traditional value issues, such as school prayer, sex education, drug use, alcohol laws, and pornography (Shapiro and Mahajan, 1986). Verba, Schlozman, and Brady (1995, 252) found that, “to the extent that there are opinion diﬀerences between the sexes, they tend to be more pronounced on issues like war and the use of violence than on what are often referred to as ‘women’s issues,’ ” such as the ERA and abortion. Also included are measures of age and martial status. Older Americans are more likely to express greater conservative ideological identiﬁcation (Erikson,Wright, and McIver, 1993). Survey evidence also shows married people are more politically conservative and more likely to vote Republican than are single people. Two variables measure Latino acculturation. Those who are more acculturated may possess less distinctive beliefs, as the process of “boundary reduction” (Yinger, 1985)5 moves them toward mainstream Anglo opinion. The ﬁrst variable is citizenship status, as citizens are more likely to be acculturated than non-citizens. Latino non-citizens are also less cynical toward government than are Latino citizens, and so might be more likely to support government programs. The second variable is language preference, speciﬁed as the language chosen by the respondent for the interview. Those who chose to answer the survey in English are likely to be more acculturated than those who preferred Spanish.6 For simplicity of presentation, the models use the same set of independent variables for each opinion regression. The regression models are therefore: Opinion = f{California,Texas, income, education, age, gender, marital status, employment status, citizenship, English language} Opinion = f{Florida, income, education, age, gender, marital status, employment status, citizenship, English language}

Findings A. Aggregate Data: Mexican Americans Table 4.1 presents the aggregate opinions of Mexican Americans in California,Texas, and elsewhere.The ﬁrst step is to establish consistent criteria for assessing diﬀerences across the states. Given the imprecise nature of polling data, it would be inappropriate to assume that all reported diﬀerences are meaningful. For example, the question on handgun control found that Californians are 4 percentage points more likely than Texans to favor making the pur-

Mexican-American and Cuban-American Public Opinion 65

chase of such guns more diﬃcult.Although this might conform to general expectations about the political cultures of California and Texas, the level of diﬀerence is not very large (73 percent versus 69 percent). Because it is better to provide conservative estimates, thereby preferring to risk underinterpreting than overinterpreting the data, this chapter only discusses aggregate opinion diﬀerences of 10 percentage points or more. The most noticeable characteristic of the data was the overall lack of opinion diﬀerences between Mexican Americans according to state residence.Although there are many diﬀerences less than 10 percentage points, they do not point in any coherent direction and are therefore likely the result of random noise. For instance, whereas Californians are 4 percentage points more likely than Texans (87 percent versus 83 percent) to favor government-provided health insurance,Texans are 6 percentage points (75 percent versus 69 percent) more likely to favor greater regulation of HMOs.Although it is possible to derive a relatively complex interpretation of this (and other) ﬁndings, the most cautious conclusion is to focus on the relatively high level of support for the role of the federal government in the provision of health care and the regulation of the health care industry. Of the 19 policy and political questions, there were only three instances of diﬀerences greater than 10 percentage points. The ﬁrst was the anticipated vote for Gore and Bush in the 2000 election. We might want to interpret the data cautiously, however, as the survey was conducted between 14 to 16 months before election day and neither candidate had secured their party’s nomination. Nevertheless, Mexican Americans in California (53 percent) were much more likely than Mexican Americans in Texas (35 percent) or elsewhere (39 percent) to support Gore. As Bush was the governor of Texas at the time of the survey, it is perhaps not a surprise to ﬁnd his support greatest among Tejanos, that is, Latino Texans. The second and third instances are found in the “social issues” questions that involve morality. For the question about whether it is acceptable to have children out of wedlock, we found that Californians (45 percent) and Texans (46 percent) are less accepting than are Mexican Americans in other states (55 percent). In addition, a similar pattern holds for the question of whether divorce is acceptable. The remaining policy questions—ranging from immigration to abortion to NAFTA to the death penalty—exhibit no noteworthy opinion diﬀerences by state.This also applies to questions of partisanship and trust in government. Overall, this lack of variation suggests that Mexican American may be a valid category. Although we might ﬁnd more diﬀerences if more states

66

Leal Table 4.1 Mexican-American Opinion Variation by State Question

California

Texas

Elsewhere

The following did not contain opinion differences of at least 10 percentage points: HMO Reform 1 = more regulation necessary 69% 75% 77% 0 = not necessary 31% 25% 23% Government Provided Health Insurance 1 = support 87% 83% 84% 0 = oppose 13% 17% 16% Guns 2 = Harder to buy 73% 69% 72% 1 = Current laws right 19% 25% 18% 0 = Easier to buy 8% 6% 10% Bilingual Education 1 = Favor instruction in native language 64% 65% 61% 0 = Favor instruction in English 36% 35% 39% Immigration 2 = Increase 26% 24% 32% 1 = Keep the same 55% 57% 59% 0 = Decrease 19% 19% 9% Abortion 4 = Legal in all cases 8% 9% 16% 3 = Legal in most cases 32% 33% 27% 2 = Illegal in some cases 25% 27% 26% 1 = Illegal in all cases 34% 31% 31% Assisted Suicide 1 = Support 45% 38% 43% 0 = Oppose 55% 62% 57% Death Penalty 1 = Favor 60% 57% 51% 0 = Oppose 40% 43% 49% Gay/Lesbian Sexual Activity 1 = Acceptable 26% 23% 30% 0 = Unacceptable 74% 77% 70% Washington solve problems 1 = Yes 68% 64% 65% 0 = No 32% 36% 35% (continued)

were analyzed, this comparative analysis (which includes the two states with the largest Mexican-American populations) provides some useful perspective on a consequential question. More research is clearly needed, but given the lack of political polls with Latino respondents located in diﬀerent states, progress may be slow.

Mexican-American and Cuban-American Public Opinion 67 Table 4.1 (continued) Question

California

Texas

Elsewhere

The following did not contain opinion differences of at least 10 percentage points: Size of Government 1 = Larger 77% 67% 71% 0 = Smaller 23% 33% 29% Trust Government 4 = Always 19% 20% 16% 3 = Most 25% 29% 31% 2 = Some 55% 48% 51% 1 = Never 1% 2% 1% Party Affiliation 2 = Democrat 60% 61% 55% 1 = Independent 11% 10% 10% 0 = Republican 29% 30% 35% Which Party Supports Latinos 2 = Democrat 45% 43% 42% 1 = No difference 41% 47% 43% 0 = Republican 14% 10% 14% Clinton Evaluation 1 = Approve 84% 89% 81% 0 = Disapprove 16% 12% 18% NAFTA 2 = Good 37% 46% 46% 1 = Unsure 49% 38% 41% 0 = Bad 15% 16% 14% The following contained opinion differences of at least 10 percentage points: Children out of Wedlock 1 = Acceptable 45% 46% 55% 0 = Unacceptable 55% 54% 45% Divorce 1 = Acceptable 50% 52% 60% 0 = Unacceptable 50% 48% 40% Gore vs. Bush (anticipated vote) 1 = Gore 53% 35% 39% 0 = Bush 47% 65% 61% Sources: 1999 Washington Post / Henry J. Kaiser Family Foundation / Harvard University National Survey on Latinos in America (NSLA).

In addition, if the data had revealed a large number of diﬀerences, this would add a new level of complexity to the job of those who wish to survey Latinos. It would call into question the utility of analyzing “MexicanAmerican” opinion, just as research recent has found that “Latino” is a problematic category of analysis.

68

Leal Table 4.2 Cuban-American Opinion Variation by State Question

Florida

The following did not contain opinion differences of at least 10 percentage points: HMO Reform 1 = more regulation necessary 77% 0 = not necessary 23% Government Provided Health Insurance 1 = support 88% 0 = oppose 12% Guns 2 = Harder to buy 75% 1 = Current laws right 14% 0 = Easier to buy 11% Immigration 2 = Increase 36% 1 = Keep the same 50% 0 = Decrease 13% Assisted Suicide 1 = Oppose 51% 0 = Favor 49% Washington solve problems 1 = Yes 63% 0 = No 37% Clinton Evaluation 1 = Approve 74% 0 = Disapprove 26%

Elsewhere

72% 28% 83% 17% 80% 16% 4% 36% 53% 11% 52% 48% 69% 31% 79% 21% (continued)

B. Aggregate Data: Cuban Americans The analysis of Cuban-American opinion, by contrast, arrived at almost the opposite conclusion. More times than not, Table 4.2 shows large diﬀerences between the opinions of Cuban Americans who live in Florida and those who live elsewhere. Of the 19 policy and political questions, there were opinion diﬀerences larger than 10 percentage points in twelve cases. Because this chapter chose a somewhat conservative standard for assessing diﬀerences, this result is more meaningful than if the chapter had decided on a single-digit standard. On the other hand, the relatively low number of respondents outside of Florida (49) means that just a small number of respondents can have a relatively strong inﬂuence on the percentage in favor or opposed to a particular policy. For the policy questions, we ﬁrst see that Floridians are more likely to

Table 4.2 (continued) Question

Florida

The following contained opinion differences of at least 10 percentage points: Bilingual Education 1 = Favor instruction in native language 58% 0 = Favor instruction in English 42% Abortion 4 = Legal in all cases 23% 3 = Legal in most cases 32% 2 = Illegal in some cases 24% 1 = Illegal in all cases 20% Death Penalty 1 = Favor 71% 0 = Oppose 29% Gay/Lesbian Sexual Activity 1 = Acceptable 25% 0 = Unacceptable 75% Children out of Wedlock 1 = Acceptable 51% 0 = Unacceptable 49% Divorce 1 = Acceptable 63% 0 = Unacceptable 37% Size of Government 1 = Larger 62% 0 = Smaller 38% Trust Government 4 = Always 19% 3 = Most 34% 2 = Some 44% 1 = Never 2% Party Affiliation 2 = Democrat 42% 1 = Independent 5% 0 = Republican 52% Which Party Supports Latinos 2 = Democrat 36% 1 = No difference 40% 0 = Republican 24% Gore vs. Bush (anticipated vote) 1 = Gore 34% 0 = Bush 66% U.S. Diplomatic Relations with Cuba 1 = Approve 22% 0 = Disapprove 78%

Elsewhere

47% 53% 33% 35% 17% 15% 54% 46% 57% 43% 63% 38% 77% 23% 51% 49% 6% 43% 51% 0% 58% 11% 32% 55% 38% 7% 51% 49% 43% 57%

Source: 1999 Washington Post / Henry J. Kaiser Family Foundation / Harvard University National Survey on Latinos in America (NSLA).

70

Leal

favor bilingual education (58 percent) than are Cuban Americans in other states (47 percent). Given the high levels of Spanish language use in Florida, among both Cuban Americans and other Latinos, it makes sense that support would be relatively high for an instructional approach that allows for some Spanish maintenance. Dade County was also the site of an early Spanish-English bilingual education program in 1961, spearheaded by migrants from Cuba. On the other hand, there were no large opinion diﬀerences for gun control, government regulation of HMOs, and government-provided health insurance. On a number of “social issues,” Cuban Americans in Florida express more conservative opinions than do Cuban Americans in other states. First, support for abortion rights is weaker among Floridians (33 percent in the highest support category) than other Cuban Americans (23 percent). Floridians are also 17 percentage points more favorable toward the death penalty, 32 percentage points less favorable toward gay and lesbian sexual activity, 12 percentage points less accepting of children out of wedlock, and 14 percentage points less accepting of divorce. The only social issue with little opinion variation was assisted suicide. Cuban Americans expressed diﬀerent opinions about the government depending on their state.Those in Florida were 11 percentage points more supportive of a large federal government and were 13 percentage points more likely to say they “always” trusted government. Very large diﬀerences are also evident in the partisanship questions. As might be predicted, Cuban Americans outside of Florida were 16 percentage points more likely to identify themselves as Democrats. It is therefore unsurprising that Cuban Americans in Florida were 14 percentage points more likely to believe that the Republican party was more likely than the Democratic party to support the interests of Latinos. Although there were no diﬀerences in the evaluation of President Clinton, Cuban Americans in Florida were 17 percentage points more supportive of George Bush than Al Gore.What might explain the diﬀerence between support for Clinton and his vice president? The crucial factor may be Jeb Bush, the governor of Florida and brother of the 2000 Republican presidential candidate. Lastly, there was a large diﬀerence of opinion about whether the United States should establish diplomatic relations with Cuba. While 78 percent of Cuban Americans outside of Florida approved, only 43 percent of those in Florida shared this view—a very large 35 percentage point diﬀerence. Taken together, these data indicate substantial state-level diﬀerences between Cuban Americans across a number of diﬀerent political and pol-

Mexican-American and Cuban-American Public Opinion 71

icy questions. This suggests that the category of “Cuban American” may mask substantial diﬀerences based on residency. Although researchers might be tempted to minimize such diﬀerences, a more cautious approach would at least control for Florida status in regression models. More research should also be conducted on this topic; hopefully future studies of Latino opinion will contain a larger number of Cuban-American respondents who live outside of Florida.

C. Regression Results: Mexican Americans The chapter also uses regression analysis to test for statistically signiﬁcant state opinion diﬀerences within the Mexican-American and CubanAmerican populations. The regressions themselves are not presented due to space considerations, as this would require the inclusion of the results from 38 regression models, but the results are described below. What does it mean if statistically signiﬁcant state variables are revealed? It means that the opinion diﬀerences between Latinos inside and outside of a state cannot be simply explained by the demographic diﬀerences between them. The meaning of such a ceteris paribus relationship is subject to debate, however, as there is no way to prove what might explain the relationship. One likely explanation is the political culture of the locales, particularly the unique political situations of California, Texas, and Florida. For Mexican Americans, there are few clear patterns in the regression results.This section broadly tested two hypotheses involving the political dynamics of California and Texas, but there was only weak evidence for either.The ﬁrst hypothesis was that the anti-immigrant political events of the 1990s in California would lead to a more Democratic and liberal Mexican-American population in that state. The second hypothesis was that the more inclusive Latino outreach strategy adopted by Governor George W. Bush, who was relatively popular among Tejanos, might move Mexican Americans in Texas toward the Republican camp and perhaps toward more conservative views. For partisanship, there were no state-level diﬀerences in party aﬃliation or opinions about which party has “more concern for Latinos.”We did see that Mexican Americans in Texas were less likely to anticipate voting for Gore (and therefore more likely to support their governor, George W. Bush), and that Mexican Americans in California were more likely to support Gore. On the other hand, Californians were not uniquely inclined to positively evaluate then-President Clinton, but Mexican Americans in Texas were more likely to give him positive reviews. Why this might be the case is unclear. There were few clear trends in the policy issue models. Californians

72

Leal

were less likely to support stronger HMO regulation, less likely to support immigration, and more likely to support the death penalty. This would appear to indicate a relatively conservative population, but it did not express unique opinions about government provided health insurance, gun control, bilingual education, abortion, and assisted suicide. Mexican Americans in Texas were particularly likely to support bilingual education, to oppose immigration, and to oppose abortion. This is a mix of liberal and conservative positions, and we must keep in mind that Mexican Americans in California and Texas did not have unique opinions on the other ﬁve policy questions. On the three social issue questions, Mexican Americans in both California and Texas were generally more likely to take conservative positions than were Mexican Americans living in other states. Lastly, no state population adopted a unique position on trust in government, the proper role of the government, and the right size of government.

D. Regression Results: Cuban Americans Concerning Cuban-American respondents, there are only a few instances of a statistically signiﬁcant Florida variable.This outcome suggests that many of the opinion diﬀerences noted above may be the result of demographic and SES variation between Cuban Americans in Florida and elsewhere. Of the 19 regressions, the Florida variable is statistically signiﬁcant four times. First, Cuban Americans in Florida were more likely to oppose the establishment of diplomatic relations with Cuban. Florida politics were dramatically changed by the sudden migration of Cubans after the Cuban Revolution, and the politically inﬂuential exile community has long promoted a strong anti-Castro stance. Cuban Americans who live outside of this unique environment may develop diﬀerent views, at least in degree if not in direction, which may be reﬂected by the signiﬁcant Florida variable. Cuban Americans in Florida are also more likely to believe that the Republican party has “more concern for Latinos” than does the Democratic party.This likely reﬂects the particularly strong support of the embargo by Cuban Americans in Florida, which ﬁnds relatively less support within the Democratic party. On the other hand, Cuban-American overall support for the parties is not inﬂuenced by state location. This suggests that demographic diﬀerences play a more important role in partisan aﬃliation. In addition, two policy questions are associated with state residence. Cuban Americans in Florida are less likely to support increased overall (not just Latino) immigration and more likely to favor the death penalty. Nevertheless, the Florida variable is statistically insigniﬁcant in the six

Mexican-American and Cuban-American Public Opinion 73

other policy models, the three “social issues” models, and the three models concerning the government.This does not suggest any clear and consistent ceteris paribus opinion diﬀerences according to state location.

Conclusions A growing number of people from the business, political, and academic worlds want to communicate with and understand the “Latino community,” but it is unclear whether such a construction exists. Many scholars now understand the importance of looking within this label and studying Latino respondents according to national-origin group status, as there may be signiﬁcant diﬀerences among those of Mexican, Cuban, Puerto Rican, Caribbean, Central American, and South American heritage. Less often asked, however, is whether these national-origin categories are as meaningful as we generally assume them to be. In physics, scientists have long sought to understand the component parts of matter. As time advances and technologies improve, the known parts have become smaller. In a similar manner, social scientists may need to put Latino groupings under the microscope to see if important variation exists within speciﬁc categories. In this chapter, we tested for state-level variation in the policy and political opinions of Mexican Americans and Cuban Americans to see whether respondents in these groups had similar orientations. If there were no important diﬀerences, then we would have greater conﬁdence that research discussing “Mexican-American” and “Cuban-American” opinion is using reliable categories. If there were many diﬀerences, however, then perhaps scholars would be better advised to separately examine respondents from diﬀerent locales.Although this would complicate the job of researching an already complex population, it would be consistent with eﬀorts to avoid the study of “Latinos” and instead look within the category. Using data from the 1999 National Survey on Latinos in America, this chapter found very few aggregate opinion diﬀerences between Mexican Americans in Texas, California, and elsewhere.This lack of variation suggests that “Mexican American” may be a valid category in terms of public opinion. If the data had revealed many diﬀerences, it would call into question the utility of reporting “Mexican-American” opinion. One caveat is that the analysis of additional states might reveal more diﬀerences, but more extensive state-level data are not available at this time. For Cuban Americans, however, there were many aggregate-level opinion diﬀerences.This is the case for policy questions,“social issues,” the size of government, partisanship, candidate evaluations, and opinions about the

74

Leal

Cuban embargo. This suggests that research ﬁndings on the political and policy opinions of “Cuban Americans” may mask substantial diﬀerences based on residency. For both Cuban Americans and Mexican Americans, the results from the regression models are less clear, but indicate that SES and other demographic variables likely play some mediating role. Last, these ﬁndings are only the beginning of a research agenda that looks within Latino national-origin group opinion. More research is clearly needed, and hopefully future Latino surveys will contain two features: samples that include respondents from the wide range of nationalorigin groups and samples that span the United States so that additional state-level comparisons can be made.

Notes 1. Although Valdez (1987, 193) in the mid-1980s identiﬁed 220 Latino public opinion data sources constructed at an estimated cost of over $13 million, he noted that “the quality of the data is so uneven that a considerable amount of eﬀort remains ahead.The truly adequate data resources are few.” 2. See, for instance Olsen, 1970;Verba and Nie, 1972;Antunes and Gaitz, 1975; Miller Gurin, Gurin, and Malanchuk, 198l; Wolﬁnger and Rosenstone, 1980; Shingles, 1981; Calvo and Rosenstone, 1989; Ellison and Gay, 1989; Uhlaner, Cain, and Kiewiet, 1989; Bobo and Gilliam, 1990;Verba, Schlozman, and Brady, 1995. 3. For the exact wording and coding of all dependent variables, as well as the aggregate responses by Latinos, African Americans, Anglos, and the Latino national-origin groups, see http://www.kﬀ.org/kaiserpolls/3023-index.cfm. This information is not included in the chapter because of space considerations. 4. All but the abortion question allowed two responses, so they are analyzed using logit.The four-response abortion question is examined using ordinal probit analysis. 5. For more on Latinos and acculturation, see Padilla (1980), Mendoza (1984), and Ramirez (1984). 6. One might question the inclusion of non-citizen respondents in the analysis, as they are not eligible to participate in electoral politics. On the other hand, they do participate in non-electoral political activities (Leal, 2002b; Uhlaner, Cain, and Kiewiet, 1989), likely respond to public opinion polls that inﬂuence politicians, and may act as local opinion leaders and thereby inﬂuence those who do vote.To delete such a large share of this population would provide an inaccurate view of Latino opinion. In any case, a dummy variable controls for non-citizens.

References Antunes, George, and Charles Gaitz. 1975. “Ethnicity and Participation: A Study of Mexican-Americans, Blacks and Whites.” American Journal of Sociology 80 (March): 1192–1211.

Mexican-American and Cuban-American Public Opinion 75 Arce, Carlos H. 1997. Mexican Origin People In The United States:The 1979 Chicano Survey [Computer ﬁle]. Ann Arbor, MI: University of Michigan, Survey Research Center. ICPSR ed. Inter-university Consortium for Political and Social Research [producer and distributor]. Barreto, Matt, and Nathan Woods. 2000. Voting Patterns and the Dramatic Growth of the Latino Electorate in Los Angeles County, 1994–1998. Claremont, CA: Tomás Rivera Policy Institute. Bean, Frank, and Marta Tienda. 1987. The Hispanic Population of the United States. New York: Russell Sage Foundation. Bobo, Lawrence, and Franklin Gilliam, Jr. 1990. “Race, Socio-Political Participation, and Black Empowerment.” American Political Science Review 84 ( June): 377–393. Boswell,Thomas, and James Curtis. 1983. The Cuban American Experience.Totowa, NJ: Rowman and Allanheld. Brischetto, Robert. 1987.“Latinos in the 1984 Election Exit Polls.” In Rodolfo O. de la Garza, ed., Ignored Voices: Public Opinion Polls and the Latino Community. Austin,TX: Center for Mexican American Studies. Cain, Bruce, and D. Roderick Kiewiet. 1987. “Latinos and the 1984 Election: A Comparative Perspective.” In Rodolfo de la Garza, ed., Ignored Voices: Public Opinion Polls and the Latino Community.Austin,TX: Center for Mexican American Studies. Calvo, Maria Antonia, and Steven J. Rosenstone. 1989. Hispanic Political Participation. San Antonio: Southwest Voter Research Institute. de la Garza, Rodolfo. 1985.“As American as Tamale Pie: Mexican-American Political Mobilization and the Loyalty Question.” In Walker Connor, ed., MexicanAmericans in Comparative Context. Washington, DC: The Urban Institute Press. de la Garza, Rodolfo, Louis DeSipio, F. Chris Garcia, John Garcia, and Angelo Falcon. 1992. Latino Voices: Mexican, Puerto Rican, and Cuban Perspectives on American Politics. Boulder, CO:Westview. de la Garza, Rodolfo, Angelo Falcon, and F. Chris Garcia. 1996. “Will the Real Americans Please Stand Up: Anglo and Mexican-American Support of Core American Political Values.” American Journal of Political Science 40 (May): 335– 351. de la Garza, Rodolfo, Luis Fraga, and Harry Pachon. 1988. “Toward a Shared Agenda.” Journal of State Government 61(2): 77–80. de la Garza, Rodolfo, and Janet Weaver. 1985.“Does Ethnicity Make a Diﬀerence: Chicano-Anglo Public Policy Perspectives in San Antonio.” Social Science Quarterly 66 (September): 576–586. de la Garza, Rodolfo, Jerry Polinard, Robert Wrinkle, and Tomas Longoria. 1991. “Understanding Intra-Ethnic Attitude Variations: Mexican Origin Population Views of Immigration.” Social Science Quarterly 72 ( June): 379–387. de la Garza, Rodolfo, and Louis DeSipio. 1994.“Overview:The Link Between Individuals and Electoral Institutions in Five Latino Neighborhoods.” In Rodolfo O. de la Garza, Martha Menchaca, and Louis DeSipio (eds.), Barrio Ballots: Latino Politics in the 1990 Elections. Boulder, CO:Westview, pp. 1–41. Deitch, Cynthia. 1988.“Sex Diﬀerences in Support for Government Spending.” In

76

Leal

Carol Mueller, ed., The Politics of the Gender Gap:The Social Construction of Political Inﬂuence. Newbury Park, CA: Sage. del Olmo, Frank. 1987. “A Journalist’s View of Latino Public Opinion Polls.” In Rodolfo de la Garza, ed., Ignored Voices: Public Opinion Polls and the Latino Community. Austin,TX: Center for Mexican American Studies. Domínguez, Jorge I. 1994. “Do ‘Latinos’ Exist?” Contemporary Sociology 23 (May): 354–356. Ellison, Christopher, and David Gay. 1989.“Black Political Participation Revisited: A Test of Compensatory, Ethnic Community, and Public Arena Models.” Social Science Quarterly 70 (March): 101–119. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. Cambridge: Cambridge University Press. Fuchs, Lawrence. 1990. The American Kaleidoscope: Race, Ethnicity, and the Civic Culture. Hanover, NH:Wesleyan University Press. Jones-Correa, Michael, and David Leal. 1996. “Becoming ‘Hispanic’: Secondary Pan-Ethnic Identiﬁcation Among Latin American-Origin Populations in the United States.” Hispanic Journal of Behavioral Sciences 18 (May): 214–254. Keefe, Susan, and Amado Padilla. 1987. Chicano Ethnicity.Albuquerque: University of New Mexico Press. Kinder, Donald, and Lynn Sanders. 1996. Divided By Color: Racial Politics and the Democratic Ideal. Chicago: University of Chicago Press. Leal, David L. 2002a.“Latino Political Opinion: Is It Unique?” Paper presented at the annual meeting of the American Political Science Association. Boston, MA, August 29–September 1. ———. 2002b. “Political Participation by Latino Non-Citizens in the United States.” British Journal of Political Science 32 (April): 353–370. Lovrich, Nicholas. 1974.“Diﬀering Priorities in an Urban Electorate: Service Priorities Among Anglo, Black and Mexican-American Voters.” Social Science Quarterly 55 (December): 704–717. Mendoza, Richard. 1984. “Acculturation and Sociocultural Variability.” In Joe Martinez and Richard Mendoza (eds.), Chicano Psychology, 2nd ed. Orlando: Academic Press. Miller, Lawrence, Jerry L. Polinard, and Robert Wrinkle. 1984. “Attitudes Toward Undocumented Workers: The Mexican-American Perspective.” Social Science Quarterly 65 ( June): 482–494. Miller, Arthur, Patricia Gurin, Gerald Gurin, and Oksana Malanchuk. 1981. “Group Consciousness and Political Participation.” American Journal of Political Science 25 (August): 494–511. Moreno, Dario. 1997. “The Cuban Model: Political Empowerment in Miami.” In F. Chris Garcia, ed., Pursing Power: Latinos and the Political System. South Bend: University of Notre Dame Press. Olsen, Marvin E. 1970.“Social and Political Participation of Blacks.” American Sociological Review 35 (August): 682–697. Pachon, Harry, and Louis DeSipio. 1988. The Latino Vote in 1988. NALEO Background Paper #7.Washington, DC: NALEO Educational Fund.

Mexican-American and Cuban-American Public Opinion 77 Padilla, Amado. 1980.“The Role of Cultural Awareness and Ethnic Loyalty in Acculturation.” In Amado Padilla, ed., Acculturation:Theory, Models and Some New Findings. Boulder, CO: American Association for the Advancement of Science. Pantoja, Adrian, Ricardo Ramirez, and Gary Segura. 2001. “Citizens by Choice, Voters by Necessity: Patterns in Political Mobilization by Naturalized Latinos.” Political Research Quarterly 54 (December): 729–750. Pastor, Robert, and Rafael Fernandez de Castro. 1998.“Replacing Paternalism with Partnership.” In Robert Pastor and Rafael Fernandez de Castro (eds.), The Controversial Pivot:The U.S. Congress and North America.Washington, DC: Brookings Institution Press. Paul, Pope John, II. 1995. The Gospel of Life:The Encyclical Letter on Abortion, Euthanasia, and the Death Penalty in Today’s World. New York:Times Books. Perez, Lisandro. 1992. “Cuban Miami.” In Guillermo Grenier and Alex Stepick (eds.), Miami Today. Miami: Florida International Press. Portes, Alejandro, and Rubén Rumbaut. 1990. Immigrant America:A Portrait. Berkeley: University of California Press. Portes, Alejandro, and Alex Stepick. 1993. City on the Edge: The Transformation of Miami. Berkeley:The University of California Press. Ramirez, Manuel. 1984.“Assessing and Understanding Biculturalism-Multiculturalism in Mexican-American Adults.” In Joe Martinez and Richard Mendoza (eds.), Chicano Psychology, 2nd ed. Orlando: Academic Press. Scott, Steve. 2000. “Competing for the New Majority Vote.” California Journal 31 ( January): 16–23. Shapiro, Robert, and Harpreet Mahajan. 1986.“Gender Diﬀerences in Policy Preferences: A Summary of Trends from the 1960s to the 1980s.” Public Opinion Quarterly 50 (Spring): 42–61. Shingles, Richard. 1981. “Black Consciousness and Political Participation: The Missing Link.” American Political Science Review 75 (March): 76–91. Torres, Maria de los Angeles. 1988. “From Exiles to Minorities: The Politics of Cuban-Americans.” In F. Chris Garcia, ed., Latinos and the Political System. Notre Dame: University of Notre Dame Press. Trueba, Enrique (Henry). 1999. Latinos Unidos: From Cultural Diversity to the Politics of Solidarity. Lanham, MD: Rowman & Littleﬁeld. Uhlaner, Carole J. 1991. “Perceived Discrimination and Prejudice and the Coalition Prospects of Blacks, Latinos, and Asian Americans.” In Byron O. Jackson and Michael B. Preston (eds.), Racial and Ethnic Politics in California. Berkeley: IGS. Uhlaner, Carole, Bruce Cain, and Roderick Kiewiet. 1989.“Political Participation of Ethnic Minorities in the 1980s.” Political Behavior 11 (September): 195–231. Valdez, Armando. 1987. “An Assessment of Data Resources on Latinos in the United States.” In Rodolfo de la Garza, ed., Ignored Voices: Public Opinion Polls and the Latino Community. Austin, TX: Center for Mexican American Studies. Verba, Sidney, Kay Lehman Schlozman, and Henry Brady. 1995. Voice and Equality: Civic Voluntarism in American Politics. Cambridge: Harvard University Press. Verba, Sidney, and Norman Nie. 1972. Participation in America: Political Democracy and Social Equality. Chicago: University of Chicago Press.

78

Leal

Welch, Susan, and Lee Sigelman. 1993. “The Politics of Hispanic Americans: Insights from National Surveys, 1980–1988.” Social Science Quarterly 74 (March): 76–94. Wolﬁnger, Raymond, and Steven Rosenstone. 1980. Who Votes? New Haven:Yale University Press. Yinger, Milton. 1985. “Assimilation in the United States: The MexicanAmericans.” In Walker Connor, ed., Mexican-Americans in Comparative Context. Washington, DC:The Urban Institute Press.

Chapter

Public Opinion in the States: Determinants of Legislative Job Performance

5

John A. Hamman

Introduction Public dissatisfaction with government deepened in the years following the Vietnam War and Watergate.Yet, some institutions fared worse in the eyes of the public than others.While substantial majorities of the American public at times approved of the president, as many as two-thirds disapproved of Congress (Patterson and Magleby, 1992).1 Many studies fault legislative professionalism for the low legislative job performance ratings (Cotter, 1986; Davidson and Oleszek, 1994; Durr, Gilmour, and Wolbrecht, 1997; Hibbing and Thiess-Morse, 1995; Jewell, 1982; Newkirk, 1979; Parker, 1997). Systematic study of the relationship between legislative professionalism and job performance ratings has been diﬃcult. At the national level, there is only one unit of analysis. The independent variable, legislative professionalism, does not vary. Stronger research designs are possible at the state level where levels of legislative professionalism vary by state. However, the unavailability of comparable data for a large sample of states has limited studies to one or a relatively few states (Cotter, 1986; Squire, 1993).2 More powerful comparative research designs at the state level have not yet been fully realized. This analysis extends the designs of previous studies of legislative professionalism and job performance in the states. Data from more than 120 polls taken in 13 states as well as two national surveys serve as the basis for 79

80

Hamman

assessing the impact of legislative professionalism on job performance ratings.After a review of the relevant studies of public opinion and state legislative job performance, data and methodological issues are considered, and the analysis is presented. The chapter concludes with a discussion of the implications of the ﬁndings.

Legislative Professionalism and Job Performance Professional state legislatures represent larger constituencies and are found in states with larger legislative districts. They also possess greater administrative capacity, have better compensated full-time legislators (Berkman, 1994; Squire, 1992; Thompson and Moncrief, 1992), and are found in more metropolitan, socially, and economically diverse states (Squire, 1993). The impetus for professional state legislatures generally is attributed to the U.S. Supreme Court’s landmark decision, Baker v. Carr (369 U.S. 186 [1962]), which set standards for how states conﬁgure legislative districts.This and subsequent court rulings resulted in the recruitment of legislators that better reﬂected the balance between rural, suburban, and urban interests. The new breed of legislators in many states enacted reforms that increased the frequency and duration of sessions, enlarged staﬀs, and raised the compensation of legislators. Political scientists measure legislative professionalism diﬀerently. A few studies focus on characteristics of the legislators themselves (Eliassen and Pedersen, 1978; King, 1981), most studies of American state legislatures focus on characteristics of the legislatures (Citizens Conference on State Legislatures, 1971; Grumm, 1971; Patterson, 1996).3 For instance, Squire’s (1993) professionalism scale compares three legislative characteristics: session length, staﬃng levels, and member compensation. The scale is characteristic of the other scales in the literature (Mooney, 1994) in that it indicates the extent to which state legislatures and their members mirror Congress.The scale’s values range from 0.016 to 0.625.4 King (2000) ﬁnds levels of legislative professionalism for two-thirds of state legislatures reach their highest levels by the 1980s. Our understanding of how legislative professionalism aﬀects job performance ratings comes from the ﬁndings of studies of Congress (Parker, 1977; Patterson and Caldeira, 1990), as well as from studies of single state legislatures (Patterson, Ripley, and Quinlan, 1992), compilations of state polls ( Jewell, 1982; Cotter, 1986), a mix of both approaches (Patterson, Hedlund, and Boynton, 1975), and surveys from a regional sample of states (Squire, 1993). This literature ﬁnds that legislative professionalism aﬀects job performance ratings. Depending on the level of government, these studies also

Public Opinion in the States 81

identify other important inﬂuences. At the national level, presidential job performance and economic conditions also aﬀect the public’s assessment of how well Congress is doing its job. Similarly, gubernatorial approval and state economic conditions aﬀect the job performance ratings of state legislatures. Characteristics of survey respondents such as partisanship, wealth, education, ethnicity, and gender have an eﬀect in some studies as well. It is unclear, though, why citizens respond negatively to legislatures with professional characteristics. Some studies embody systems theory concepts of state policy making and focus on policy outputs.5 This perspective presumes that knowledgeable citizens rate legislative performance according to how well policy coincides with their policy preferences.The ultimate eﬀectiveness of professional legislatures depends on the policies they make. If professional legislatures perform as intended by the reformers, then full-time professional legislatures assisted by more competent staﬀ will be more representative of statewide constituencies. This enables professional legislatures to enact policies that are more consistent with the preferences of their constituents. Finding that professional legislatures receive lower job performance ratings contradicts these assumptions and supports critics who argue that citizens are not well represented and served by professional legislatures. The critics argue that citizen preferences are pushed aside by special interests and career considerations. Policy performance and responsiveness suﬀer as a result and are reﬂected in lower job performance ratings (Rosenthal, 1998;Weber, 1999). Several studies assess how well the policies of professional legislatures conform to citizen preferences (Carmines, 1974; Hoﬀerbert, 1966; Karnig and Sigelman, 1975; Roeder, 1979). As it turns out, the ﬁndings are mixed although the study with the strongest research design found that the policy of professional legislatures is ideologically more representative (Maestas, 2000). But, even if studies conclusively find that the policies of professional legislatures either do or do not match citizen preferences, the analyses still have to wrestle with the fact that other research fairly conclusively shows that the public is not very knowledgeable of legislation and spends little time systematically following legislative politics (Delli Carpini and Keeter, 1996; Jewell, 1982; Squire, 1993). So, it is likely that legislative job performance ratings reﬂect other factors. Research showed that the public is particularly uncomfortable with the deliberative, conﬂictual, or democratic processes that build policy consensus within legislatures and between legislative and executive branches of government (Hibbing and Theiss-Morse, 1995). Professional legislatures have greater legislative capacity and potentially play a larger role in policy making.The processes of formulating and adopting policy, building coali-

82

Hamman

tions of support, and of passing legislation may leave the public feeling the institution is less responsive to them, particularly if the legislation conﬂicts with the agenda of other political elites. Rieselbach (1986) argues that one consequence of Congress becoming more assertive in policy making after reform in the 1970s ironically was a loss of the public’s sense that Congress is responsive to them. General unrest and lack of trust in government (Citrin, 1974) may manifest themselves in the form of public “revolts” such as tax reform and term limits (Lowery and Sigelman, 1981; Sears and Citrin, 1982). Karp (1995, 376) explains support for public backlashes such as term limits in the early 1990s “involved an expression of symbolic attitudes rather than an expression of immediate self-interest . . . the idea of limiting legislators’ terms is not so much a solution to a speciﬁc problem as it is a reﬂection of cynicism and anger.”These studies note that professional politicians, in particular, are criticized by reformers as being out of touch with constituents and mostly interested in promoting their own careers. In the case of term limits, Patterson (1996, 199) found that conservative business interests and Republican activists were able to tap into this vein of public discontent by arguing term limits would restore “a citizen legislature peopled by amateurs . . . and curtail political corruption.” Members of professional legislatures seem particularly prone to scandal as such controversies often arise from how they handle campaign ﬁnances. Understanding why respondents disfavor professional legislatures may require that we consider how media portrayal of legislative action contributes to public cynicism toward government and lower job performance ratings. How media coverage ﬁlters information about legislatures may be a key for understanding the legislature’s relatively poor standing with the public. In the states, media focus tends to be on the governor and the governor’s policy agenda. In contrast, coverage of the legislature tends to probe selﬁsh or political interest, maverick behavior, or malfeasance. Full-time legislatures pursuing political careers are the most likely to be perceived as corruption prone and unresponsive. Even if individual members manage their own press well, no one oﬃcial speaks well for the legislature. Indeed, individual members often campaign against their own institution. At best, a single leader speaks for a single house with individual members speaking only for themselves.This inevitably results in mixed signals and ambiguous messages being given to the public (Rosenthal, 1998). If public perceptions of the law-making process, where there is a tendency for the press to portray legislative actions as contentious or obstructionist, are what matters in evaluating a legislature’s job performance, then part of better understanding legislative job performance ratings requires

Public Opinion in the States 83

analyzing more of what state legislatures do—how they make policy. From this perspective, the likelihood of being exposed to undesirable aspects of the legislative process increases with the number of bills being passed and sent to the governor for enactment. Public support also erodes as legislatures pass legislation destined to be vetoed, or when legislatures subsequently override a governor’s veto. These types of actions pit parties and institutions against each other. Finding that conﬂict and discord among branches of state government contribute to public dissatisfaction with state legislatures would be consistent with ﬁndings about the impact of disharmony between Congress and the president.The public judges Congress more severely when major legislation is passed, or when Congress seeks to override presidential vetoes (Durr et al., 1997; Parker, 1977). In these ways actions of professional legislatures also tap into general public unrest and depress job performance assessments of state legislatures. This chapter assesses the impact of legislative professionalism on job performance ratings in two ways. First, longitudinal data from over 100 polls administered in 13 states is analyzed to show how job performance ratings decline as levels of professionalism in state legislatures reach their highest levels. The state polls also show the impact that job performance ratings of the governor and relative state unemployment have on legislative job performance ratings. Second, cross-sectional data from two national surveys test the impact of legislative professionalism on job performance ratings more systematically. Multivariate logistic regression analysis of the two ABC/Washington Post surveys ﬁnds evidence to support the proposition that legislative action is an important determinant of legislative job performance ratings.The analysis also ﬁnds that citizens in states adopting term limits in the early 1990s give their state legislatures lower job performance ratings.

Legislative Professionalism and Trends in Legislative Job Performance in 13 States Data collected by the National Network of State Polls make it possible to analyze trends in legislative job performance and investigate whether changes in levels of performance ratings correspond to the rising levels of legislative professionalism discussed above. Currently, there is a question asking respondents to rate the job performance of their state legislature in a total of 124 polls administered in 13 states over a period spanning 26 years.6 Because these data come from polls administered by diﬀerent polling organizations using diﬀerent question formats, the analysis measures legislative job performance as the percentage of positive responses

84

Hamman

divided by the sum of the percentage of positive responses and the percentage of negative responses (Beyle, Niemi, and Sigelman, 2002). Regressing this measure of legislative job performance on year (not shown) shows that legislative job performance ratings decline from the early 1980s through the early 1990s by about 1 percent per year when controlling for question format and polling ﬁrm. Further analysis shows job performance ratings trend lower for the states with the most professional legislatures. Five or more polls from 1980 through 2000 are available for seven states. Five of the states, Alabama, Florida, Kentucky, Mississippi, and New Jersey, have professional scores in the low to middle range of Squire’s professionalism scale. California and Illinois have considerably more professional state legislatures.The trends in the scores of these low to moderate and high professionalism states are plotted in Figure 5.1. Performance ratings decline for all the legislatures. But, the trends in the states with the two most professional legislatures are visibly lower. This is reﬂected in each state’s average legislative performance rating. For instance, California has the most professional legislature. Its score on Squire’s professionalism index is 0.62.The California legislature’s average job performance rating is 0.27.Alternatively, two of the least professional state legislatures, Mississippi and Alabama, both score 0.16 on the professionalism index.Their job performance rating averages are 0.32 and 0.30.A linear regression (not shown) of all the available 124 polls ﬁnds that ratings are 7 percent lower for the two most professional legislatures. The state polls clearly show the relationship between the level of legislative professionalism and legislative job performance is negative for these states over this time period. For most polls, there is either a question in the same poll or from another state poll in the same month that asks respondents about the governor’s job performance.7 A measure of relative state unemployment is also coded by lagging the diﬀerence between state and national unemployment by one month.8 Trends in both variables are also displayed in Figure 5.1 from 1980 to 2000.The trends replicate previous ﬁndings in the literature lending some conﬁdence in the overall results of the trend analysis. Presidential popularity and economic performance are important predictors of congressional job performance (Parker, 1977; Patterson and Caldeira, 1990) and gubernatorial job performance ratings and state economic conditions aﬀect public assessments of state legislatures (Squire, 1992).The job performance of governors and state legislatures is inversely related to relative state unemployment, as expected. Figure 5.1 shows that the public consistently rates legislatures lower than governors regardless of their level of professionalism. Based on these data, we can only speculate why this might be, but some studies suggest

Public Opinion in the States 85 1.2

1

0.8

0.6

0.4

0.2

0 1980 1981 1982 1983 1984 1985 1986 1987 1988 1989 1990 1991 1992 1993 1994 1995 1996 1997

Year Approval of Governor

Relative State Unemp.

Approval Low Prof. Legislatures

Approval of High Prof. Legislatures

Figure 5.1. Trends in Institutional Job Performance and State Relative Unemployment 1980–1998. Source: National Network of State Polls.

that media coverage plays a part. Rosenthal’s (1990) observation that state legislatures generally are a poor match for the governor in policy disputes might hold the key. In the public arena, the media tend to give the governor center stage and focus on the governor’s policy proposals while probing for petty self-interest and graft in the legislature. It is also interesting to note that during 1990 to 1994, when legislative job performance ratings were at their lowest levels in states for which there are polls, the term limit movement got under way. Only two states with relatively restrictive initiative qualiﬁcations, Arkansas and Illinois, failed to eventually place a term limit initiative on the ballot. In one additional state, Louisiana, the state legislature placed a referendum for term limits on the ballot and was approved by the electorate. Only two states, Mississippi and North Dakota, voted down term limits (Bowler and Donovan, 2004; Hamm and Moncrief, 1999). The conventional wisdom focuses on policy outputs and holds that low job performance ratings in the early 1990s heightened public dissatisfaction with the policies of state legislatures. The polls, in conjunction with previous research (Cotter, 1986; Jewell, 1982; Newkirk, 1979), indicate that performance ratings began declining as early as the late 1970s and contin-

86

Hamman

ued to decline until the beginning of the term limit movement in the early 1990s. At that time, citizens dissatisﬁed with the policies and programs of their state legislatures ﬁnally lashed out against their legislatures by approving term limits. On its face, this appears to be a plausible explanation for citizen support of term limits. But another explanation is more plausible when it is recognized that the public is unfamiliar with state legislatures and their policies. Karp (1995) argues that voter support of term limits reﬂected dissatisfaction with the political process and the sense that professional politicians are not responsive to their constituents’ interests (Abramson, 1983; Craig, 1993; Karp, 1995; Thompson and Moncrief, 1992; Will, 1992).9 From this perspective, voting for term limits in the states should be seen more as a symbolic protest. Supporters may also have been indirectly expressing their frustration with Congress whose job performance ratings had fallen to historically low levels, 17 percent, by April 1992 (Patterson and Magleby, 1992). The poll data show a fairly strong positive relationship between term limits and legislative professionalism (r = 0.58), but no relationship between the adoption of term limits and legislative job performance levels.10 The voters most likely to approve term limits were in the states with the most professional state legislatures. To summarize, the state polls show that job performance ratings declined as state legislatures became more professional.The public also rated the job performance of state legislatures lower in states with more professional legislatures. Economic performance and public assessment of the job performance of the state’s governor were related to legislative job performance as expected.The states in which citizens supported term limits had the most professional legislatures.Thus, there is fairly strong anecdotal evidence to suggest that an unanticipated result of reform may have been to create more professional institutions over the end of the twentieth century that are less eﬀective and responsive in the eyes of the people they represent. Further sorting out these relationships and reaching ﬁrmer conclusions requires a more systematic, multivariate model.The next section reports the ﬁndings of such an analysis using two national surveys administered at this time in the early 1990s.

Legislative Professionalism and Job Performance: Multivariate Analysis Using National Survey Data The cross-sectional analysis cannot model trends in legislative job performance. However, it can build on previous studies that examine legislative job performance at a single point in time but use samples from rela-

Public Opinion in the States 87

tively few numbers of states or omit variables of theoretical interest.11 Equivalent surveys for all 50 states would provide the most powerful research design, but none currently exist. However, there are two ABC/ Washington Post national polls from 1990 and 1991 with job performance and other important attitudinal and demographic control variables.12 These surveys were administered to people in practically all of the states and were given when legislative job performance ratings were at about their lowest level in most states. Using national surveys such as these avoids methodological issues encountered in similar research of other institutions that use in-person, clustered samples.13

Dependent Variable The dependent variable for the analysis, legislative job performance, is the respondent’s response to the question: do you approve or disapprove of the way your state legislature is doing its job? This format is typical of presidential and gubernatorial job performance questions used in previous studies and the levels of public approval are comparable.14 Fifty-eight percent of respondents approve of the job performance of their legislatures in 1990 and 52 percent approve in 1991.The weighted average for both surveys is 55 percent. Because the variable is dichotomous, the equation parameters will be estimated with logistic regression.

Independent Variables The independent variables that are of primary theoretical interest measure the variables identiﬁed in the preceding discussion and analysis. They pertain to the professional characteristics and legislative actions of state legislatures. Findings in the literature further point to a negative relationship between professional legislatures and job performance ratings. The preceding analysis of 124 state polls showed performance ratings decline after levels of professionalism reach their highest levels.The more the characteristics of legislatures resemble Congress, the lower job performance ratings are likely to be. In addition to professional characteristics of state legislatures such as session length, staﬀ size, compensation, and constituency size, the discussion also questioned whether low job performance ratings result from legislative actions and assertiveness in policy making.

Legislative Characteristics The analysis uses Squire’s legislative professionalism index to measure the extent to which characteristics of state legislatures mirror Congress. The survey data include the full range of values on Squire’s index. How-

88

Hamman

ever, diagnostic analysis shows that the distribution of the index’s values is not normal and that a polynomial form provides the best ﬁt of the equation.15 If legislative professionalization results in institutional and representational dynamics that depress job performance ratings, as argued above, then state legislatures that meet longer, are better staﬀed, and are better compensated will be rated lower by the public. Professional legislatures generally have larger legislative districts. The literature is not clear what the impact of constituency size on legislative performance will be. Squire (1993) hypothesized that larger districts result in less contact with constituents. The result is poorer representation and lower performance evaluations. Larger districts may also be politically and socially more diverse. This makes meeting constituent needs more diﬃcult. Conversely, others argue that politics are more divisive in smaller districts ( Jewell, 1982; Patterson, Ripley, and Quinlan, 1992). Smaller districts could enable more interaction between constituents and state legislators. Constituents would then become more aware of existing conﬂicts and controversies. Constituency size is measured as either the average population size of the districts of both houses of the legislature, or the single body in the case of Nebraska.

Legislative Action The analysis uses three measures of legislative action: the number of bills a legislature passes, the number of bills vetoed by the governor, and the number of vetoed bills reinstated by the legislature.The variable measuring the number of bills passed by the legislature provides a general indication of overall legislation action. As the amount of legislation increases, so do the opportunities for public exposure to the deliberative and democratic processes hypothesized to lower job performance ratings. Legislative action represented by the number of bills produced is the most visible “manifestation of the legislature at work” (Wahlke, Eulau, Buchanan, and Ferguson, 1962, 18).There is considerable variation in the level of state legislative action with some respondents residing in states passing hundreds of pieces of legislation over the time period covered by the 1990 and 1991 polls.There are also variables for the number of bills vetoed by the governor and the number of vetoes overridden by the legislature.The variables measuring the number of vetoes and veto overrides more directly assess legislative assertiveness and the extent to which legislatures directly confront governors.The number of vetoes occurring in the states varies considerably, but overriding the governor’s veto is a relatively rare event with many legislatures not reversing a single veto.

Public Opinion in the States 89

Control Variables The preceding trend analysis found states whose citizens were more likely to vote for term limits had more professional legislatures. This ﬁnding suggests that levels of political cynicism and unrest are higher in states with more professional legislatures. The surveys did not ask questions typically used to measure political cynicism and trust in government. However, a variable for whether the state adopted term limit provisions in the early 1990s is included to provide a general indication of public unrest. States that adopted term limits are expected to have lower job performance ratings. Other control variables that the literature predicts will aﬀect legislative job performance are also used.These variables include the respondent’s assessment of how well the governor is performing in oﬃce, performance of the state and local economy, whether the party the respondent identiﬁes with controls the state legislature, whether more than one party controls legislative and executive branches of government, state population, and demographic characteristics of the respondent. Four variables measure the respondent’s assessment of the governor’s job performance, national economic performance, and local economic performance, and the respondent’s partisanship. Cotter (1986), Patterson et al. (1992), and Squire (1993) ﬁnd that evaluations of the state executive inﬂuence how respondents view their state legislature. Governors provide an important cue for a public that is less familiar with the state legislature than they are with the governor and the policies of state government more generally. The equations include a measure for how respondents assess the governor’s job performance, which is worded the same as the dependent variable except it asks about the job performance of the governor. Prospective economic assessment variables account for the respondent’s assessment of the national and local economy. Past research indicates that prospective economic assessment is important, but not whether perceptions of local economic conditions, which are within the policy jurisdiction of state legislatures, or national economic conditions, which are not, aﬀect job performance ratings. It seems reasonable to expect people, when asked, to hold state legislatures more responsible for local conditions than for national economic conditions.16 Political cynicism and citizen unrest are typically blamed for adoption of term limits in fourteen states by 1994. If public unrest manifests itself in the passage of term limits, it may also be reﬂected by lower performance ratings. A dichotomous variable is given a value of one if a state adopted

90

Hamman

term limits between 1990 and 1994 and zero if it did not. It tests whether legislatures in term limit states have lower job performance ratings. There is also a variable measuring whether the state legislature is controlled by the same party as the respondent.Although respondents tend to give governors of their own party higher performance ratings, it is less clear how they will evaluate legislatures. In contrast to the governor, where the media and public can identify one fairly visible individual with their political party and the governorship, the partisanship of a legislature seldom can be characterized as neatly as all but one legislature is divided into two houses. So, it is likely that it is more diﬃcult to attach a partisan label to state legislatures because of their decentralized nature and the fact that the public is less aware of them. Also, the partisan control of the state legislature was split for about one-fourth of the respondents in the 1990 survey and almost a third in the 1991 poll. Not unexpectedly, studies report mixed ﬁndings. Patterson et al. (1992) ﬁnds a relationship for partisans of the same party controlling Congress and job performance ratings, but none for the state legislature. Patterson et al. (1975) and Squire (1993) ﬁnd partisans of the same party controlling their state legislature give it a higher performance rating. A variable measuring whether one party controls both the legislature and statehouse is also used in the analysis although there is not a consensus in the literature of how divided government aﬀects job performance ratings. On the one hand, the public may be more likely to approve of the job performance of the legislature under conditions of divided government for several reasons. Citizens may be more likely to approve of the legislature’s job performance if they believe the governor is blocking popular legislation.They may give higher marks to the legislature if it is controlled by their own party. Citizens may rate both branches more in terms of how they see overall government performance regardless of partisan control. If the public believes government is performing well, they will be more likely to approve of the legislature’s job performance than when they are displeased with government.They may even approve of the legislature’s job performance because divided government requires greater consensus to pass legislation and less unpopular legislation results. On the other hand, it is possible that the public does not perceive that state legislatures with divided party control perform diﬀerently from legislatures with uniﬁed partisan control. In this case, the analysis will ﬁnd that divided government has no eﬀect on legislative job performance ratings. Mooney (1995) and King (2000) argue that state population size is an important surrogate for a number of factors that likely contribute to public satisfaction with state government. Larger states generally face more

Public Opinion in the States 91

complex problems associated with urbanization and there is more likely to be alienation from the politics involved in building coalitions among more diverse political constituencies in larger states. Finally, four variables control for demographic characteristics of the respondents. Like the rest of the control variables, they are not the theoretical focus of the analysis but serve to better specify the model. The impact of these variables varies somewhat in past studies, also (Cotter, 1986; Jewell, 1982; Patterson et al., 1975; Patterson et al., 1992; and Squire, 1993).Age, income, and education either do not matter or are negative (Patterson et al., 1992; Squire, 1993). Nonwhite respondents and men also assess legislatures more negatively. Deﬁnitions for the variables are provided in the appendix to the chapter.

Results The equations are estimated with logistic regression because the dependent variable is dichotomous. Logistic regression coeﬃcients and robust standard errors for the independent variables in two of the equations are reported in Table 5.1. An initial equation was estimated (not shown) using Squire’s professionalism scale with the control variables that is comparable to models used in previous studies. It yields similar results.The sign of the coeﬃcient for professionalism is negative, but statistically signiﬁcant only at the 0.10 level. Like previous studies, this suggests that respondents in states with more professionalized legislatures are somewhat more likely to evaluate job performance negatively. However, diagnostics show that when a polynomial of the legislative professionalism scale is entered, it is statistically signiﬁcant and improves the ﬁt of the equation.17 In contrast to previous ﬁndings, this result shows that up to a point, respondents living in states with more professional legislatures evaluate job performance more favorably.The eﬀects then become negative, indicating that respondents in states with the most professionalized legislatures judge them less favorably. This ﬁnding about the functional form of the professionalism index is notable and diﬀers from previous ﬁndings where professionalism has less of an impact and is negative. Here the highest levels of professionalism equate with the negative impact. Equation 1 in Table 5.1 reports the results of the equation with the polynomial form of the professionalism index. All of the variables are entered in equation 2.The results are listed in the right column of Table 5.1 and are largely consistent with the analysis of the state polling data in the previous section. They show that when entering the variables for the legislative professionalism scale, other legislative characteristics, and legislative action, only the variable measuring the volume of legislation is statistically signiﬁcant. There is no evidence that public sup-

Table 5.1 Legislative Professionalism and Job Approval 1990–1991 (with Robust Standard Error Estimation)1 Variable

Legislative professionalism Legislative professionalism squared Constituency size Volume of legislation Legislation vetoed Vetoes overturned by legislature Gubernatorial job approval R’s perception of national economy R’s perception of local economy R’s party controls legislature Term limits Divided party control of government State population R’s income R’s age R’s gender R’s minority status Survey year Constant Model Chi square Degrees of freedom Sig. Correctly predicted N

Equation 1

4.06* (1.69) –5.39** (2.04) — — — — — — — — 1.38** (0.11) 0.02 (0.10) 0.33** (0.08) 0.32** (0.13) — –0.02 (0.11) –0.001 (0.001) –0.11** (0.03) –0.010** (0.003) –0.34** (0.11) 0.19 (0.15) –0.37** (0.12) 0.31 (0.41) 240.01 13 0.000 67.5 % 2,108

Equation 2

2.49* (1.86) –3.05 (2.24) –0.000006 (.000004) –0.0006** (0.0002) 0.001 (0.001) –0.008 (0.01) 1.40** (0.11) 0.01 (0.09) 0.35** (0.08) 0.36** (0.13) –0.27** (0.13) 0.09 (0.13) 0.004 (0.003) –0.11** (0.04) –0.01** (0.003) –0.36** (0.11) 0.17 (0.15) –0.30** (0.12) 0.68 (0.44) 261.7 18 0.000 67.7 % 2,108

Source: October 1990 and June 1991 ABC/Washington Post polls 1Logit coefficients (robust standard errors in parentheses) *p < 0.10, **p < 0.05

Public Opinion in the States 93 Table 5.2 The Probability of Approving State Legislative Performance for Different Model Specifications and Independent Variable Values Variable Values

Lowest Median Highest Mean 1990 Mean 1991

Legislative Characteristics1 (Not Shown)

Legislative Activity 2 (Not Shown)

Legislative Characteristics and Activity 3 (Table 5.1)

0.55 0.63 0.26

0.62 0.56 0.38

0.66 0.59 0.33

0.56 0.51

Source: ABC News/Washington Post poll, October 1990 and June 1991 1Probabilities reflect substitution of the lowest, median, and highest value of average district size, professionalism and professionalism squared term. Mean values are substituted for the other independent variables from the equation in column 1 of Table 5.1. 2Probabilities reflect the estimation of the same equation as equation 1 except variables for the number of vetoes, overrides, and bills passed are estimated instead of the professionalism variables.The lowest, median, and highest value of the statistically significant legislative activity variable, the number of bills passed, is then substituted into the equation while setting the values of the other variables to their mean value.This equation is not reported in Table 5.2. 3These probabilities reflect substitution of the lowest, median, and highest value of the average district size and the number of bills passed while substituting mean values for the other independent variables from the equation in column 2 of Table 5.2.

port erodes when more legislation is vetoed or when legislatures override vetoes.18 It is notable that when all of the measures of legislative characteristics and action are entered together in the equation, the coeﬃcients of legislative professionalism and constituency size are also statistically insigniﬁcant.This is consistent with the argument that the public disapproves of the political process itself, the legislature at work. It is not so much that policy is out of step with citizen preferences as it is that people are turned oﬀ by processes that produce higher volumes of legislation. Probabilities of the impact of legislative characteristics and action on legislative job performance ratings are listed for the estimates of three equations in Table 5.2. The probabilities in the left column of the table reﬂect substituting the lowest, median, and maximum value of the variables measuring legislative characteristics: average district size, professionalism, and professionalism squared. Mean values are substituted for all of the other independent variable values.When the minimum values are substituted, legislative job performance is 0.55.This increases to 0.63 for the median values. Job performance then decreases to 0.26 when the maximum values are substituted.The upward trend from the lowest to median

94

Hamman

values of legislative characteristics supports an argument that up to a point citizens rate professional legislatures, legislatures more like Congress, favorably. The ratings then decrease substantially for legislatures in states with the most professional characteristics.This is not the case, however, in the equations with legislative action variables. Job performance ratings decrease as legislative action or the combinations of both types of variables are included in the equation. For legislative action, job performance ratings decrease from 0.62 to 0.38. The likelihood of approving of the state legislature’s job performance falls from 0.66 to 0.33 when using estimates from the full equation. In each instance, the most substantial change in the probability of approving legislative job performance corresponds to the diﬀerence between the median and most professional legislature. Overall, the equations perform satisfactory and correctly predict over two-thirds of the cases. Many of the control variables are statistically signiﬁcant and have the expected relationship with legislative job performance providing further assurance about the speciﬁcation of each equation.The results also are largely consistent with the analysis of state polls in the previous section. In particular, public assessment of the governor’s job performance is strong and positive. Economic conditions are important. The relationship between local economic outlook and job performance is statistically signiﬁcant and positive. Stronger economic expectations translate into a greater probability of approving of the state legislature’s job performance. Legislatures in states with term limits have lower job performance ratings. Finally, the respondents were more likely to approve of the state legislature if it were controlled by their own political party.When controlling for all of the other variables, the volume of legislative action as measured by number of bills passed has a substantial impact on job performance ratings and best predicts job performance ratings.

Discussion The state polls add a longitudinal perspective to the analysis of legislative professionalism and legislative job performance.The polls show the expected negative impact of legislative professionalism on job performance ratings. Legislative job performance ratings decline in the 1980s and early 1990s after levels of legislative professionalism reach their highest levels. Moreover, the losses are greatest in states with the most professional legislatures.The polls also show that levels of gubernatorial job performance decline over this time period although citizens in any given state are more likely to ﬁnd the governor’s job performance acceptable than they are the performance of the

Public Opinion in the States 95

legislature.This is the case even for ratings in states with less professional legislatures.The analysis speculated that lower job performance ratings for state legislatures may reﬂect systematic diﬀerences in media coverage. Wellstaﬀed, deliberative legislative bodies meeting over longer sessions, passing more legislation, and having more legislation vetoed by the governor may be ampliﬁed by such diﬀerential treatment by the media. Unfortunately, a more deﬁnitive test was beyond the scope of the state poll analysis. The national-level survey data greatly extend the scope of previous studies and systematically test the impact of legislative professionalism on job performance with a cross-sectional design.The ﬁndings diverge from earlier studies in a couple of interesting ways. An equation similar to ones in previous studies but which models the functional form of the legislative professionalism scale shows that the probability of approving job performance increases for state legislatures scoring from low to median values on the professionalism scale. Up to a point, professional characteristics positively aﬀect job performance ratings. This ﬁnding appears to somewhat vindicate reformers who worked to enhance the state legislature’s role in policy making by raising the expertise and initiative of state legislatures to a level comparable to other actors in the policy-making process. It is not until there is a much closer resemblance between state legislatures and Congress in terms of session length, staﬃng, and careerism that the probability for approving job performance decreases. However, the fully speciﬁed equation shows only the number of bills passed by the legislature to be statistically signiﬁcant and the volume of legislation passed has a negative impact.This ﬁnding may point to public dissatisfaction with legislative action and politics more generally. The ﬁnding helps clarify what the public ﬁnds unsatisfactory with more professional legislatures. It also supports those who argue the public is turned oﬀ by high stakes, high ﬁnance, campaigning, and lobbying that is becoming increasingly characteristic of the most professional legislatures.19 The results resonate with congressional studies that ﬁnd the public is less likely to approve of Congress’ job performance when it passes more legislation or when it is seen as challenging the president’s policy initiatives. Finally, the ﬁnding that job performance ratings are also lower in states with term limits has some interesting implications. It lends some support to the argument that media portrayal of legislative action contributes to public cynicism toward government and lower job performance ratings. How media coverage ﬁlters information about legislatures may be a key for understanding the legislature’s relatively poor standing with the public. The relatively lower job performance ratings of legislatures in term limit states could be a reﬂection of higher levels of public unrest. Un-

96

Hamman

fortunately, the surveys do not contain the questions needed to test this idea directly. Hopefully, such data will become available in the future. In conclusion, reform seldom is guided by a comprehensive conceptual framework for governing. This analysis suggests that as legislatures and their politics become more professional, as their production of legislation increases, they reach a point where the public reacts negatively to them. During the 1980s and early 1990s citizens become less likely to approve of their state legislature or governor. Still, given the overall decline in job performance ratings and relative unpopularity with the governor, citizens with a more professional legislature are less likely to approve of its job performance and are more likely to approve of its terms being limited.Therefore, the growing professionalism of state legislatures appears to contribute to the decline in job performance ratings. In the ﬁnal analysis, the ﬁnding concerning legislative action and the volume of legislation helps to shed additional light on public dissatisfaction with professional legislatures.

Notes 1. The Gallop Poll shows that a decline in Congress’s job approval in March 2005 was more than for other measures of public satisfaction with government. Just 37 percent approved of the way Congress did its job compared with a 52 percent approval rating for the president. Gallop editor-in-chief, Frank Newport, identiﬁed arguing among senators and congressmen on television over social security reform and other issues as one reason why approval of Congress had fallen so low (USA Today,Tuesday, March 15 8A). 2. The problem limited samples have on the variance of legislative professionalism as well as other independent variables are discussed in the analysis section. 3. There are alternative deﬁnitions and usages of the term professionalism. For instance, King (1981) and Eliassen and Pedersen (1978) use it to refer to characteristics of the legislators themselves. Professionalism in this chapter refers to extent to which structural characteristics of state legislatures mirror those of Congress. 4. The scores of these four highest states range from 0.614 to 0.659 on Squire’s professionalism index. The scores of the ﬁve least professional states in Squire’s index range from scores ranging from 0.042 to 0.083. For more about the index see Squire (1992). 5. For instance see Anderson (2003) or Peters (2004) for models of public policy making where the public input depends in part on the policy output of government. 6. The National Network of State Polls is maintained by University of Kentucky Odum Institute for Research in Social Science. It archives state polling data for public use and can be accessed at http://survey.rgs.uky.edu/nnsp/. 7. These responses are corrected in the same way as legislative popularity.The percentage of positive responses divided by the sum of the percentage of positive responses and the percentage of negative responses. 8. The actual relative unemployment rate is multiplied by 10 for presentation

Public Opinion in the States 97 purposes in the graph. For example, a 6 percent rate of unemployment appears as 0.6. 9. By the 1980s, only about 25 percent of Americans respond that they trusted government to do what is right.This is down from almost 75 percent in the 1950s and 1960s. 10. Obviously it is more diﬃcult for states to adopt term limits that do not have the initiative. But this is the case even when whether or not the state has the initiative is taken into account. 11. For instance Squire’s (1993) multivariate analysis of legislative job performance approval that uses seven comparable state samples to test hypotheses about factors that inﬂuence public evaluations of state legislatures. National survey data can be used to test whether constrained variance in the professionalism variable aﬀects results. For instance, the most professionalized states (i.e., New York, Michigan, California, and Massachusetts) with scores from 0.615 to 0.659, are not in Squire’s study. New York and Michigan have scores more than twice the magnitude of Illinois and nearly three times as great as Iowa. So, only states with low to medium scores, with most in the medium range of the scale, are analyzed. This limits the scale’s variance and may potentially dampen the eﬀect. Also one state, South Dakota, accounts for most of the variation in the values of the index score measuring professionalism. South Dakota’s very low score, 0.083, contrasts sharply with the other scores of the other states (Iowa, Illinois, Minnesota, Missouri, Nebraska, South Dakota, and Wisconsin) that range from 0.186 to 0.302. 12. There are 1,006 respondents in the 1990 ABC/Washington Post poll and 1,511 in the 1991 poll. The sample weights are adjusted to reﬂect the diﬀerent sample sizes. Pooling data from national samples increases the degrees of freedom for analysis and allows more general tests in terms of spanning greater time periods. Pooling data may bias standard error estimates to the extent that state-speciﬁc factors correlate with other independent variables so robust standard errors are estimated. A dummy variable indicating the diﬀerent surveys is also entered in the equations (Stimson, 1985). 13. The national survey data are also attractive because they are randomly dialed. Random selection avoids problems of pooling in-person surveys that employ cluster sampling techniques (Waksberg, 1978). 14. This dichotomous measure is similar to the ordinal scale used by Cotter (1986), Patterson et al. (1992), and Squire (1993) to measure legislative job performance with values from 1 (strongly disapprove) to 4 (strongly approve). The earlier Patterson et al. (1975) study used an indicator of a seven-item scale to measure a more broadly construed concept of legislative support. 15. An exponential term (the natural log of the scale) and a polynomial (the professionalism variable plus a squared professionalism term) are estimated in place of the professionalism scale. Neither the natural log nor the exponential terms are statistically signiﬁcant. See Mooney’s (1994) analysis of professionalism scales for an in-depth analysis of Squire’s as well as other professionalism scales. Such precautions probably were not necessary in previous studies because the most professional legislatures were not included in their samples. However, the polynomial form is statistically signiﬁcant and improves the ﬁt of the equation. 16. Squire (1993) and Patterson et al. (1992) ﬁnd positive and no relationships

98

Hamman

respectively between retrospective economic assessments and state-level job eﬀectiveness while MacKuen, Erikson, and Stimson (1992) ﬁnd that prospective economic assessments aﬀect presidential popularity. 17. The analysis tested the functional form of the legislative characteristic and action variables estimating equations with logged, exponential, and polynomial forms of each. Alternative speciﬁcation only improved the ﬁt of the equation for transformations of the professionalism scale. 18. All of the legislative characteristic and action variables, with the exception of veto overrides, are correlated to varying extents.The interrelationships are not strong enough to construct a reliable and valid scale, but more professional legislatures pass more legislation. Respondents with legislatures that pass more legislation also generally have legislatures that have more legislation vetoed and reside in larger districts.The results here indicate that it is volume of legislation that best predicts legislative job performance levels. 19. See Hibbing and Theiss-Morse (1995) for ﬁndings regarding public intolerance to conﬂict and debate among governmental leaders and Durr et al. (1997) for public disapproval of Congress for debating and passing major legislation.

References Abramson, Paul R. 1983. Political Attitudes in America: Formation and Change. San Francisco:W. H. Freeman. Anderson, James E. 2003. Public Policymaking: An Introduction. Boston: Houghton Miﬄin. Berkman, Michael B. 1994. “State Legislators in Congress: Strategic Politicians, Professional Legislatures, and the Party Nexus.” American Journal of Political Science 38 (November): 1025–1055. Beyle,Thad, Richard G. Niemi, and Lee Sigelman. 2002.“Gubernatorial, Senatorial, and State-Level Presidential Job Approval Ratings:The U.S. Oﬃcials Job Approval Ratings ( JAR) Collection.” State Politics & Policy Quarterly 2 (Fall): 215–229. Bowler, Shaun and Todd Donovan. 2004.“The Initiative Process.” In Politics in the American States: A Comparative Analysis, ed.Virginia Gray and Russell Hanson. Washington, DC: CQ Press, 129–156. Carmines, Edward G. 1974.“The Mediating Inﬂuence of State Legislatures on the Linkage between Interparty Competitions and Welfare Policy.” American Political Science Review 68 (September): 1118–1124. Citizens Conference on State Legislatures. 1971. State Legislatures:An Evaluation of Their Eﬀectiveness. New York: Praeger. Citrin, Jack. 1974. “Comment: the Political Relevance of Trust in Government.” American Political Science Review 68 (September): 973–988. Council of State Governments. 1990. The Book of the States 1990–91. Lexington, KY: 28: 107–192. Council of State Governments. 1991. The Book of the States 1991–92. Lexington, KY: 29: 123–208. Cotter, Patrick. 1986. “Legislatures and Public Opinion.” State Government 59(1): 46–51.

Public Opinion in the States 99 Craig, Stephen C. 1993. The Malevolent Leaders: Popular Discontent in America. Boulder, CO:Westview. Davidson, Roger H., and Walter J. Oleszek. 1994. Congress and Its Members.Washington, DC: CQ Press. Delli Carpini, Michael X., and Scott Keeter. 1996. What Americans Know About Politics and Why It Matters? New Haven:Yale University Press. Durr, Robert H., John B. Gilmour, and Christina Wolbrecht. 1997.“Explaining Congressional Approval.” American Journal of Political Science 41 ( January): 175–207. Eliassen, Kjell A., and Mogens N. Pedersen. 1978.“Professionalism of Legislatures: Long-term Change in Political Recruitment in Denmark and Norway.” Comparative Studies in Society and History 20 (April): 286–318. Grumm, John G. 1971.“The Eﬀects of Legislative Structure on Legislative Performance.” In State and Urban Politics, ed. Richard I. Hoﬀerbert and Ira Sharkansky. Boston: Little, Brown. Hamm, Keith E. and Gary E. Moncrief. 1999. “Legislative Politics in the States.” In Politics in the American States: A Comparative Analysis, ed.Virginia Gray, Russell Hanson, and Herbert Jacob.Washington, DC: CQ Press, pp. 144–190. Hibbing, John R. and Elizabeth Theiss-Morse. 1995. Congress as Public Enemy: Public Attitudes toward American Political Institutions. New York: Cambridge University Press. Hoﬀerbert, Richard I. 1966. “The Relationship between Public Policy and Some Structural and Environmental Variables in the American States.” American Political Science Review 60 (March): 73–82. Jewell, Malcolm E. 1982. Representation in State Legislatures. Lexington: University Press of Kentucky. Karnig, Albert K., and Lee Sigelman. 1975. “State Legislative Reform and Public Policy: Another Look.” Western Political Quarterly 28 (September): 548–552. Karp, Jeﬀrey A. 1995. “Explaining Public Support for Legislative Term Limits.” Public Opinion Quarterly 59 (Autumn): 373–391. King, Anthony. 1981.“The Rise of the Career Politician in Britain and Its Consequences.” British Journal of Political Science 11 ( July 3): 249–285. King, James D. 2000. “Changes in Professionalism in the U.S. State Legislatures.” Legislative Studies Quarterly 25 (May): 327–343. Lowery, David, and Lee Sigelman. 1981.“Understanding the Tax Revolt: Eight Explanations.” American Political Science Review 75 (December): 963–974. MacKuen, Michael B., Robert S. Erikson, and James A. Stimson. 1992.“Peasants or Bankers? The American Electorate and the U.S. Economy.” American Political Science Review 86 (September): 597–611. Maestas, Cherie. 2000.“Professional Legislatures and Ambitious Politicians: Policy Responsiveness of State Institutions.” Legislative Studies Quarterly 25 (November): 663–690. Mooney, Christopher Z. 1994. “Measuring U.S. State Legislative Professionalism: An Evaluation of Five Indices.” State and Local Government Review 26 (Spring): 70–78. ———. 1995.“Citizens, Structures, and Sister States: Inﬂuences on State Legislative Professionalism.” Legislative Studies Quarterly 20 (February): 47–67.

100

Hamman

Newkirk, Glenn. 1979. “State Legislatures through the People’s Eyes.” State Legislatures 5: 8–9. Parker, Glenn R. 1977. “Some Themes in Congressional Unpopularity.” American Journal of Political Science 21 (February): 93–109. Patterson, Samuel C. 1996. “Legislative Politics in the States.” In Politics in the American States: A Comparative Analysis, ed. Virginia Gray and Herbert Jacob. Washington, DC: CQ Press, 159–206. Patterson, Samuel C. and Gregory A. Caldeira. 1990. “Standing Up for Congress: Variations in Public Esteem since the 1960s.” Legislative Studies Quarterly 15 (February): 25–47. Patterson, Samuel C., Ronald D. Hedlund, and G. Robert Boynton. 1975. Representatives and Represented. New York:Wiley. Patterson, Samuel C., Randall B. Ripley, and Stephen V. Quinlan, 1992. “Citizens’ Orientations toward Legislatures: Congress and the State Legislature.” Western Political Quarterly 45 ( June): 315–338. Patterson, Kelly D., and David B. Magleby. 1992. “The Polls—Poll Trends: Public Support for Congress.” Public Opinion Quarterly 56(4): 539–551. Peters, B. Guy. 2004. American Public Policy: Promise and Performance. Washington, DC: CQ Press. Rieselbach, Leroy N. 1986. Congressional Reform.Washington, DC: CQ Press. Roeder, Phillip W. 1979.“State Legislative Reform: Determinants and Policy Consequences.” American Politics Quarterly 7 ( January): 51–70. Rosenthal,Alan. 1990. Governors & Legislatures: Contending Powers.Washington, DC: CQ Press. ———. 1998. The Decline of Representative Democracy: Process, Participation, and Power in State Legislatures.Washington, DC CQ Press. Sears, David, and Jack Citrin. 1982. Tax Revolt: Something for Nothing in California. Cambridge, MA: Harvard University Press. Squire, Peverill. 1992. “Legislative Professionalism and Membership Diversity.” Legislative Studies Quarterly 17 (February): 69–79. ———. 1993. “Professionalization and Public Opinion of State Legislatures.” Journal of Politics 55 (May): 479–491. Stimson, James A. 1985.“Regression in Space and Time: A Statistical Essay.” American Journal of Political Science 29 (November): 914–947. Thompson, Joel A., and Gary F. Moncrief. 1992.“The Evolution of the State Legislature: Institutional Change and Legislative Careers.” In Changing Patterns in State Legislative Careers. ed. Gary F. Moncrief and Joel A.Thompson.Ann Arbor: University of Michigan Press. Waksberg, Martin P. 1978.“Sampling Methods for Random Digit Dialing.” Journal of the American Statistical Association 73 (March): 40–46. Wahlke, John C., Heinz Eulau,William Buchanan, and LeRoy C. Ferguson. 1962. The Legislative System: Explorations in Legislative Behavior. New York: John Wiley. Weber, Ronald E. 1999.“The Quality of State Legislative Representation: A Critical Assessment.” Journal of Politics 61 (August): 609–627. Will, George F. 1992. Restoration: Congress,Term Limits and the Recovery of Deliberative Democracy. New York: Free Press.

Public Opinion in the States 101

Appendix:Variable Deﬁnitions Dependent Variable Legislative Approval Independent Variables

0 if disapprove, 1 if approve. (Do you approve or disapprove of the way your state legislature is doing its job?)

Legislative Characteristics Professionalism

Squire’s (1993) professionalism index has values between 0 and 1, depending on how close the state legislature resembles Congress in terms of member compensation, staﬃng, and session length.

Constituency Size

This measure averages the population size of districts across the two houses in each state.

Divided Government

0 = uniﬁed partisan control of legislature and governor, 1 otherwise.

Legislative Activity Volume of Legislation*

Number of bills passed by the legislature.

Vetoes*

Number of bills vetoed in session.

Vetoes Overridden*

0 if disapprove, 1 if approve. (Do you approve or disapprove of the way your state’s governor is handling his or her job?)

Control Variables Gubernatorial Approval

0 if worse, 1 if same, and 2 if better. (How about the national economy where you live: is it getting better, getting worse, or staying the same?)

National Economy

0 if worse, 1 if same, and 2 if better. (How about the local economy where you live: is it getting better, getting worse, or staying the same?)

State Economy

This measure equals 1 if the state adopts terms limits between 1990 and 1994, 0 otherwise.

Term Limits

Number of bills vetoed by governor overridden in current session.

Legislature Same Party*

0 if own party not controlling legislature, 1 if own party controls both houses of legislature.

State Population

1992 state population.

Income

Six-point scale with 1 less than $8,000 and 6 more than $50,000.

Age

In years.

Gender

0 female, 1 male.

Minority

0 Caucasian, 1 minority.

Survey

1 October 1990 survey, 2 June 1991 survey.

*Compiled from 1990–1991 and 1991–1992 Book of the States.

Chapter

The State Economy, the National Economy, and Gubernatorial Popularity

6

Jeffrey E. Cohen and James D. King

Introduction Job approval ratings now are seemingly ubiquitous in American politics. Most research on job approval focuses on the president; we possess comparatively little research on gubernatorial approval, the topic of this chapter.As with presidents, job approval is important to governors. Job approval ratings provide governors with clues about their reelection prospects; research indicates that job approval aﬀects gubernatorial elections (Kenney and Rice, 1983; King, 2001; Svoboda, 1995). California Governor Gray Davis’s low approval ratings, arguably, emboldened his opponents in their successful recall eﬀorts in 2003. Approval also provides governors with a resource to motivate legislators and bureaucrats to accept their policy initiatives. Research ﬁnds that governors with higher approval levels receive more support from the legislature (Crew, 1998; Ferguson, 2003) and from bureaucrats (Dometrius, 2002). Job approval also provides a mechanism for the public to hold governors accountable.Through approval polls, the public can voice its support or opposition to the governor between elections, and governors may alter their policies and behavior in oﬃce in reaction to their approval level. But considerable debate exists over whether the public actually holds governors accountable in approval polls for policies and issues over which governors claim some responsibility, like jobs and unemployment. Speciﬁcally, governors take credit for the state economy when it is healthy, pointing to their actions and policies in oﬃce as evidence of their 102

The State Economy, the National Economy, and Gubernatorial Popularity 103

economic stewardship and publicizing their eﬀorts to improve the economic climate and performance of the state. Factory openings or expansions, announcements of a new employer coming to the state, and the like are often greeted with great fanfare, photo opportunities, and media events, all designed to create positive publicity for the incumbent governor’s economic leadership. And electoral challengers to sitting governors often run campaigns that target the incumbent’s economic policy failures. It is, as Grady observes,“good politics for the governor to be seen aggressively engaged in fostering [economic] growth” (1991, 109). Despite such an environment, considerable debate exists as to whether the public holds the governor accountable for the state’s economy. Crew, Branham,Weiher, and Bernick (2002), Hansen (1999a, 1999b), Howell and Vanderleeuw (1990), and Orth (2001) ﬁnd that state unemployment lowers gubernatorial approval. In contrast, Crew and Weiher (1996) ﬁnd no impact,Adams and Squire (2001) detect only spotty eﬀects of state unemployment on gubernatorial approval, and MacDonald and Sigelman (1999) unearth no relationship between state economic health and gubernatorial popularity.All told, the literature oﬀers a mixed picture of the eﬀect of the state economy on gubernatorial approval. Several factors may account for these divergent ﬁndings. First, studies use diﬀerent samples of states. Designs also diﬀer, with some using time series (Adams and Squire, 2001; Crew and Weiher, 1996; Crew et al., 2002; Hansen, 1999b), others conduct cross-sectional analyses with states as the units of analysis (MacDonald and Sigelman, 1999), while still others analyze individual level survey data (Cohen, 1983; Howell and Vanderleeuw, 1990; Orth, 2001). Studies also diﬀer in their model speciﬁcations. Some control for national economic conditions (Crew and Weiher, 1996; Crew et al., 2002; Hansen, 1999a, 1999b; Howell and Vanderleeuw, 1990; Orth, 2001), while others do not (Adams and Squire, 2001; MacDonald and Sigelman, 1999). Political variables, such as presidential approval, political predispositions, and time in oﬃce are also used inconsistently. Rarely is presidential popularity included (but see Crew and Weiher, 1996; Crew et al., 2002; Orth, 2001), although studies report that presidential approval aﬀects gubernatorial elections (e.g.,Atkeson and Partin, 1995; Carsey and Wright, 1998; King, 2001; Simon, 1989). Moreover, the theoretical linkage between state unemployment and gubernatorial popularity is not well speciﬁed. We revisit the issue of the impact of state economic conditions on gubernatorial approval. We use the Oﬃcial State Job Approval Ratings ( JAR) database (Beyle, Niemi, and Sigelman, 2002), which provides approximately 2,000 gubernatorial approval readings across all 50 states over 20 years.We also employ a more comprehensive set of control vari-

104

Cohen and King

ables to guard against spurious results. Plus, we develop a theory of accountability that recognizes that governors serve in a complex federal system in which national economic policies and performance aﬀect statelevel economic performance.The public, in our model, grades the governor not by the absolute level of the economy but by how well the state economy performs relative to the national economy. In the next section, we develop our theoretical model. Then, we introduce the data, discuss our variables, and present the data analysis.

State Economic Conditions and Gubernatorial Approval A large literature demonstrates that the American public holds its political leaders accountable for economic performance. Most of these studies focus on the president, and demonstrate that presidential approval cycles with the performance of the macroeconomy and with people’s attitudes about the national economy (Erikson, MacKuen, and Stimson, 2002; Gronke and Newman, 2003).We assume the public wants to hold the governor accountable for state economic performance, but the federal context complicates matters for the public. The ﬁrst complication is that state economies are to some degree integrated into the national economy. Although some evidence suggests that state economies are becoming more independent of national economic forces (Brace, 1993; Hendrick and Garand, 1991), there is little doubt that national economic conditions and policies aﬀect a state’s economy. Furthermore, state policy makers do not possess the necessary tools to inﬂuence all types of economic conditions. Most notably, state policy makers have little ability to manage a state’s inﬂation rate or money supply. Consequently, state policy makers possess only limited means of aﬀecting their state’s economic performance. We assume that the public takes into account these two limitations—the lack of state economic autonomy and state policy makers’ limited economic policy toolkit. Still, governors are the most visible political leaders in their states (Squire and Fastnow, 1994) and they often focus people’s attention on their ability to generate jobs for the state’s economy and to maintain a healthy job climate. Herzik (1991) found two-ﬁfths of governors mentioning economic development in their 1988 state-of-the-state addresses; only education was mentioned more frequently. Many states have established oﬃces to attract jobs and help employers, and state industrial policy initiatives often focus on attracting new employers (Brace, 1993; Grady 1991). When it comes to job retention or expansion, states often compete with each other. During economic expansion, employers must make decisions,

The State Economy, the National Economy, and Gubernatorial Popularity 105

not only about how many new employees to hire but where they will work. Such locational decisions foster jobs competition among states. This context leads us to hypothesize that state publics will hold governors accountable for the employment situation in the state, but the absolute level of unemployment will be less relevant to state voters than the level of state employment compared to the national unemployment level, the relative level of unemployment. Thus, in good times, when the absolute unemployment level is low, public support for the governor will erode if the state’s unemployment level supercedes the national level. And in bad times, when state unemployment is high, governors should receive an approval boost if the state level falls below the national level. This hypothesis assumes a somewhat sophisticated state electorate, which is contrary to much research on public opinion. However, research suggests that state publics reasonably discriminate between the state and national economies (Niemi, Bremer, and Heel, 1999) and that they are reasonably knowledgeable with regard to unemployment rates (Holbrook and Garand, 1996). Limited evidence of voters distinguishing between national and state economic conditions is found in King’s (1999) analysis showing Wyoming voters’ evaluating state economic performance more negatively at a time when the state’s economy was mired in recession during the national economic boom of the 1990s. Furthermore, voters in Wyoming acted on this distinction, as evaluation of state economic performance signiﬁcantly inﬂuenced vote choice for governor while evaluation of national economic performance did not. Finally, some research argues that the public is more responsive to changes in unemployment rates than to changes in other economic indicators (Conover, Feldman, and Knight, 1986; Haller and Norpoth, 1997).1 Two studies employ variables that resemble our relative unemployment concept. Hansen (1999b) uses a variable that measures the diﬀerence between national and state unemployment, but does not ﬁnd it to be statistically signiﬁcant.2 Crew et al. (2002) set the problem up slightly diﬀerently than we do, using the level of state unemployment and the diﬀerence between national and state unemployment. As we detail below, we too use the diﬀerence between state and national unemployment, but control for national unemployment rather than state unemployment. While Crew et al. (2002) ﬁnd that the level of state unemployment is consistently signiﬁcant, the diﬀerence variable is not. It is not clear why state publics would punish or reward the governor when the national economy is performing better or worse than the state economy. We formally deﬁne relative unemployment as the one month lagged state unemployment percentage minus the one month lagged national unemployment

106

Cohen and King

percentage. By applying a control for national unemployment, we can assess whether the public rewards (or punishes) the governor when state unemployment is lower (or higher) than national unemployment.3 We cannot easily enter both state and national at the same time because of their high correlation (r = 0.64, p < 0.0001). Doing so produces multicollinearity and considerable coeﬃcient instability.4 As our theoretical model focuses on the unemployment situation of the state relative to the nation, theory calls for controls for the national unemployment level, rather than for the state level.The national unemployment level serves as a base against which voters judge the state’s performance. We include national unemployment to fully specify the voters’ decision calculus.5

The Oﬃcial State Job Approval Data Set Data limitations have plagued research on gubernatorial approval.The paucity of data and the lack of comparability in data sets and designs may be one source of the diﬀerences in ﬁndings across studies. Data limitations, too, may force compromises on estimated models, leading to model misspeciﬁcation, which may result in spurious or suppressor eﬀects. We use the most comprehensive collection of data on gubernatorial approval yet gathered, the Oﬃcial State Job Approval Ratings ( JAR) project.6 This project collected state-level job approval ratings for governors from numerous survey organizations from 1947 through 2001. All told, the JAR project supplies us with approximately 2,000 readings of state-level gubernatorial popularity.We focus on the data from 1980 through 2001 because of diﬃculty in collecting data on some variables of interest for earlier years and because the overwhelming number of cases come from this period.7 There are several complications in using the JAR data. First, because the data come from a variety of data sources and polling organizations, data quality varies. For instance, sometimes only positive evaluations of the governor were available (Beyle, Niemi, and Sigelman, 2002) and the percentage of “don’t know” responses appears to vary because of the large variety of question wordings and other factors. In order to retain as many cases as possible, we need to make the data as comparable as possible and compensate when comparability is at issue.To correct for the variability in “don’t know” responses, we use the formula “Percent Positive/(Percent Positive + Percent Negative)” to measure gubernatorial approval.8 On average, governors received a 55 percent approval rating, which varies widely from a low of 10 to a high of 97, with a standard deviation of 16. Our dependent variable is this adjusted gubernatorial popularity score. Second, this data set combines numerous variations on the general

The State Economy, the National Economy, and Gubernatorial Popularity 107

approval question.9 Some questions presumably ask about gubernatorial job performance in manner similar to the Gallup Poll’s presidential approval question; others ask how much the respondent likes the governor. Exact question wordings are not available, but the response categories (e.g., approval versus disapproval, like versus dislike) are. Beyle et al. (2002) ﬁnd the likelihood of a negative rating rises as the number of response categories increases.Third, house eﬀects may be present because the data set combines polls across many ﬁrms. To illustrate the potential impact of these factors, we created dummy variables when 30 or more observations from a single polling ﬁrm exist, a total of 41 dummies, as well as 3 dummies for response categories (for three, four, or ﬁve response categories). A regression analysis (not shown) of gubernatorial approval on three response category dummies and polling ﬁrm dummies produced an R2 of 0.31. Each of the response category dummies is statistically signiﬁcant as are 25 of the ﬁrm dummies.10

Control Variables More than relative unemployment may aﬀect citizens’ approval of their governor.The literatures on gubernatorial approval and elections ﬁnd that state and national factors aﬀect gubernatorial approval and voting. From these literatures we identify relevant control variables, which we arrange into two sets—state factors and national factors.

Other State Factors and Gubernatorial Approval State Partisan and Ideological Composition The structural political context of the state may aﬀect the governor’s popularity. Strong evidence exists that presidential co-partisans’ and those whose ideological leanings correspond to the president’s are more likely to judge the president approvingly than disapprovingly (Bond and Fleisher, 2001).Agreement with the executive’s policies may account for this eﬀect. We expect a similar dynamic to occur at the state level. Everything else being equal, we expect governors to ﬁnd more support from members of their own party than from opposition party members and from people who hold similar ideological leanings than from people who hold divergent ones. Hence, we expect gubernatorial popularity to covary positively with the partisan or ideological advantage in the mass public. To measure partisan advantage, we use the updated Erikson-WrightMcIver (EWM) (1993) state partisanship data.11 Following EWM, we calculate the percentage of Democratic and Republican identiﬁers for subsets

108

Cohen and King

of years (1977–1984, 1985–1992, and 1993–2001). The subsets are long enough to protect against short-term variation that may result from sampling and other error, but still remain relatively close in time to the measurement of the governor’s popularity.We calculate partisan advantage as: (Percentage of Gubernatorial Party Identiﬁers) – (Percentage of Opposition Party Identiﬁers). We use ideological identiﬁcation to indicate agreement or disagreement with the governor’s policies, assuming that liberal identiﬁers are more likely to agree with the policies of Democrats than Republicans, while the opposite will hold true for conservative identiﬁers. Again, we use the updated EWM data and deﬁne net ideological advantage for Democratic governors as: (Percentage of Liberal Identiﬁers) – (Percentage of Conservative Identiﬁers) and the reverse for Republican governors. For independent governors, the partisan and ideological scores are set to “0.” The partisan advantage argument also implies that independent governors may have a harder time rallying public support than partisan governors, who possess a ready-made core of supporters.12 Thus, we add a dummy variable for whether the governor was an independent, expecting independents to have lower approval levels. Lastly, we include a Republican dummy variable to distinguish any other unmeasured partisan eﬀects.

Election Performance Governors may also possess a personal support coalition, which may cross party and ideological lines. Gubernatorial election campaigns provide opportunities for governors to build such a personal base of support that can translate into votes at reelection time (Adams and Kenny, 1989) and likely into positive assessments in opinion polls.To measure this personal support base we use the percentage of the total vote that the governor received in the last election.

Governing Diﬃculty: State Diversity and Size Some states may be harder to govern than others. Policies are likely to be more popular when large segments of the population are like-minded and homogeneous. As the population becomes more dissimilar, some segment of the population is likely to dislike the policy. Although severe political divisions may appear in states with few but deep socioeconomic distinctions, in general, more heterogeneous states are likely to possess greater levels of dissimilarity than homogeneous ones. We measure heterogeneity-homogeneity with two variables, an index of sociocultural diversity (Morgan and Wilson, 1990) and population size. As diversity and population size increase, we expect gubernatorial popularity to fall.13

The State Economy, the National Economy, and Gubernatorial Popularity 109

Length of Tenure Several studies detect a temporal pattern to gubernatorial approval (Beyle, Niemi, and Sigelman, 2002; Crew and Weiher, 1996; Hansen, 1999b), as do many studies of presidential popularity (Mueller, 1970; Stimson, 1976; Brace and Hinckley, 1991).14 Several diﬀerent theoretical accounts have been oﬀered for approval’s temporal path, including Mueller’s (1970) coalition of minorities’ notion, Stimson’s (1976) cycle of expectation and disillusionment, and Brace and Hinckley’s (1991) cycle of deﬂating expectations.We cannot resolve these issues here but we need to control for the possible eﬀect of time in oﬃce on approval. Accordingly, we test both the linear and cyclical models on gubernatorial popularity, measuring length of tenure as months in oﬃce.The linear model predicts a negative impact of tenure on approval. The cyclical model predicts the same, but further predicts that as tenure extends, approval will increase.To test this cyclical model, we use tenure in months squared, which we hypothesize to have a positive sign.

United Government Last, divided government may aﬀect public approval of the governor. When the opposition party holds at least one house of the legislature, tensions between the governor and the legislature are likely to rise. Open conﬂict between the executive and legislature may undermine public conﬁdence in gubernatorial leadership. But Nicholson, Segura, and Woods (2002) argue that presidents (executives) will garner higher popularity under divided than uniﬁed government. Under divided government, the executive and legislature can point the ﬁnger of blame at each other, making it diﬃcult for the public to assess blame. As a result, some people will blame the legislature and others the executive. But when government is united, the executive will bear the full brunt of public outcry and criticism, being unable to deﬂect blame onto a legislature of his party. Lowry, Alt, and Ferree (1998) ﬁnd a similar result in state elections, with a greater impact of economic factors under united government. Uniﬁed government is a dummy variable deﬁned as the governorship and both chambers of the state legislature controlled by one political party equal to 1, and 0 otherwise.

National Factors Presidential Eﬀects Despite the fact that the public seems easily able to recognize or recall their governor’s name (Squire and Fastnow, 1994), we should not infer that

110

Cohen and King

knowledge about the governor or state politics and policy runs very deep. Studies indicate that the public possesses only a limited amount of attention and interest in political aﬀairs (e.g., Delli Carpini and Keeter, 1996). Consequently, people may use heuristics, cues, or other shortcuts to organize their thinking about state-level politics. The president and the parties often are used as referents for people in organizing their political world.Voters who rely on such shortcut devices may associate all politicians of a particular party as part of one team, rewarding or blaming the entire set, rather than distinguishing between national and state politicians (Carsey and Wright, 1998). People also may be inclined to see parties in terms of the president and generalize their attitudes about the president to all politicians of his party. Through such a mechanism (Simon, Ostrom, and Marra, 1991), the president’s approval may aﬀect the approval level of governors of the president’s party, much as it has been found to inﬂuence gubernatorial voting decisions and elections (e.g., Atkeson and Partin, 1995; King, 2001; Simon, 1989). We measure this presidential eﬀect with the Gallup presidential approval measure, lagged one month and corrected for party of the governor. That is, when the president and governor share party aﬃliation, we use the Gallup approval reading. When they come from diﬀerent parties, we subtract approval from 100 (100 – approval). This equation suggests a process in which opposition party governors are helped when the president is unpopular, but harmed when he is popular. Last, we enter a dummy variable, indicating whether the governor comes from the president’s party to pick up any aspect of presidential-gubernatorial partisanship that the approval variable leaves unmeasured.

National Economy Our theoretical model implies that people distinguish between the state and national economy and that they understand the limited ability of state policy makers to aﬀect the state economy (unemployment). Despite the fact that people seem to distinguish between the two economies in reasonable ways (Niemi, Bremer, and Heel, 1999), they may still target the governor based on the condition of the national economy. They may do so in part because the governor is a handy target, being a relatively wellknown public ﬁgure, and people may feel a need to vent their discontent. Moreover, despite being aware of diﬀerences between the state and national economies, people may hold all public oﬃcials accountable for the health of the economy. With ﬁxed terms of oﬃce, often withholding approval is the only way that the public can signal displeasure with the way things are going during the interim between elections. To national

The State Economy, the National Economy, and Gubernatorial Popularity 111

unemployment, which we discussed above, we add the monthly national inﬂation rate, lagged one month, deﬁned as the percentage change in the consumer price index.15

Results Table 6.1 presents results of three estimations: OLS, robust regression, and ﬁxed eﬀects regression (ﬁxed on state). We use robust regression because the gubernatorial approval ratings are not necessarily randomly distributed over time and state. The state ﬁxed eﬀects control for other factors associated with the states that are unmeasured and may aﬀect levels of gubernatorial approval across the states.16 Both of these estimations essentially duplicate the OLS results, with only minor changes in statistical signiﬁcance and regression coeﬃcient magnitudes.17 Unless otherwise noted, our discussion focuses on the OLS results.18 Results indicate that gubernatorial approval is highly responsive to relative state unemployment and the eﬀect is strongly signiﬁcant, with p values of 0.000 or better in all estimations. The OLS and robust regression estimations ﬁnd that each percentage point spread between state and national unemployment leads to a 3.14 percent point shift in approval.The ﬁxed eﬀects estimation slightly dampens the eﬀect of relative unemployment to 2.54. Based on the OLS results, when state and national unemployment diverge by one standard deviation (1.4), we can expect a shift of about 4.4 points in gubernatorial approval. Governors in states with the best relative unemployment performance (–4.5) can expect to see a whopping 14.2 points added to their approval levels. In contrast, those serving in states lagging behind the nation at maximum levels (6) can expect an approval loss approaching 19 percent. National unemployment also aﬀects gubernatorial approval in expected ways. Each 1 percent shift in national unemployment leads to a 4.5 percentage point shift in gubernatorial approval.19 We can compare the impact of relative and national unemployment. Whereas one standard deviation rise in relative unemployment depresses gubernatorial approval about 4.4 points, a similar one standard deviation (1.2) rise in national unemployment dampens approval about 5.4 points, which is only marginally higher than the impact of relative unemployment. The combined impact of both forms of unemployment suggests that state publics are highly sensitive to the state’s job climate. Overall, this provides strong conﬁrmation of our hypothesis that voters grade the governor based on how well their state’s job climate fairs compared to the nation. Furthermore, relative unemployment aﬀects guber-

13.88

0.78 0.04 0.04 0.09 0.001 1.35 1.79 9.02 0.00005 0.78 0.04 7.34

–6.13 –0.13 –0.11 –0.38 0.004 4.83 –2.26 –59.50 0.00004 –0.89 0.33 104.08 2016 0.27 49.11 0.26 0.0000

0.24 0.02 0.36 0.03

se

OLS

–3.14 –0.07 –4.53 0.11

b

* R2 within = 0.22, R2 between = 0.46 Source: Official State Job Approval Ratings ( JAR) Dataset.

Relative unemployment Inflation National unemployment Presidential popularity Governor and president of the same party State partisanship State ideology Months Months squared Republican governor Independent governor Diversity Population Unified government Governor's electoral percentage Constant n R2/ Adj. R2 F /Prob. F Wald chi2/ Prob. chi2 Root MSE sigma_u sigma_e rho

Variable

0.000 0.001 0.007 0.000 0.001 0.000 0.207 0.000 0.438 0.259 0.000 0.000

0.000 0.000 0.000 0.001

p

13.88

0.0000

0.78 0.04 0.04 0.12 0.002 1.46 1.68 8.97 0.00006 0.77 0.05 7.65

0.24 0.02 0.36 0.03

SE

Robust Regression

–6.13 –0.13 –0.11 –0.38 0.004 4.83 –2.26 –59.50 0.00004 –0.89 0.33 104.08 2016 0.27 61.03

–3.14 –0.07 –4.53 0.11

b

0.000 0.002 0.012 0.001 0.006 0.001 0.180 0.000 0.482 0.250 0.000 0.000

0.000 0.000 0.000 0.001

p

Table 6.1 Impact of Relative Unemployment on Gubernatorial Popularity

6.34 12.52 0.20

573.30

–5.75 –0.09 –0.10 –0.48 .005 5.78 0.33 –72.03 0.0002 –0.37 0.24 124.93 2016 0.26

–2.54 –0.12 –4.68 0.10

b

0.0000

0.77 0.04 0.04 0.08 0.001 1.36 1.98 28.12 .0002 0.78 0.04 14.09

0.31 0.02 0.34 0.03

SE

0.000 0.025 0.013 0.000 0.000 0.000 0.867 0.010 0.310 0.638 0.000 0.000

0.000 0.000 0.000 0.001

p

Fixed (State) Effects Regression*

The State Economy, the National Economy, and Gubernatorial Popularity 113

natorial job approval controlling for other state and national factors, under varying model speciﬁcations, points to which we now turn.

Impact of Control Variables Eight of the control variables aﬀect gubernatorial approval as hypothesized, three do not reach statistical signiﬁcance (population size, united government, Independent status), and two (state ideology and state partisanship) have signs opposite what we hypothesized. Of the variables that signiﬁcantly aﬀect gubernatorial approval, they do so quite strongly with coeﬃcients signiﬁcant at the 0.001 level or better: None of the signiﬁcant variables is marginal in terms of statistical signiﬁcance. Gubernatorial tenure has curvilinear eﬀects, as hypothesized. The longer a governor stays in oﬃce, the lower his or her approval, up to a point, when approval begins to rise with tenure.These results suggest that each month in oﬃce leads to a drop of 0.39 in gubernatorial approval, counterbalanced by a rise of 0.004 from the squared term.The curvilinear model anticipates that after a year in oﬃce a governor will see a drop in approval of 3.95 points just as a function of being in oﬃce. At the end of one term, the net eﬀect of tenure is –8.45 approval points. For a governor to be reelected, much else must be working to the governor’s advantage. Only well into the second term does the eﬀect of service show a positive eﬀect on approval. At 96 months, two full years, the net eﬀect of tenure stands at 2.7. The net eﬀect of tenure only begins to turn positive at month 90, when the eﬀect of tenure nets the governor 0.2 approval points. As expected, more diverse states depress gubernatorial approval, but population size adds little over and above the eﬀects of diversity. Approval of the governor in the most diverse state will be about 9.3 points less than that of the governor of the least diverse state and about 4.7 points lower than a governor of a state of average diversity. Uniﬁed government, however, does not aﬀect gubernatorial approval.20 But governors who come to oﬃce with stronger election support can expect to see higher approval than those who enter oﬃce with less voter support. The results indicate that each percentage point in electoral support produces a 0.33 approval base for the governor. Thus, governors elected with a landslide of 60 percent will see approval that is about three points higher than those elected at a bare majority of 50.1 percent. Independent governors do not seem to see a signiﬁcant approval loss.The Republican governor dummy suggests an additional 4.8 point boost approval.21 Both state ideological and partisan advantage work opposite to expec-

114

Cohen and King

tations, however. Substantively, each 10 percentage point shift in partisan advantage produces a corresponding shift of 1.3 points in gubernatorial approval. Governors with the largest net partisan bases see approval about 4.6 points lower than a governor serving in a state with perfectly balanced partisanship. Governors serving in states most deﬁcient in their partisan base will see approval that is 3.1 points higher than governors in state with equally sized partisan electorates and 9 points higher than those in the most partisan advantaged states.22 State ideology has comparable impacts. Each 10 percentage point shift in a state’s ideological makeup shift gubernatorial approval about 1.1 points. A governor serving in a state with the greatest ideological advantage can expect to see an approval level some 3.2 points lower than a governor with an ideologically balanced state public. When the ideological balance tips to the maximum against the governor, approval is predicted to rise about 3.2 point above that for governors in balanced states. Why this seemingly counterintuitive eﬀect for partisan and ideological advantage in state publics? Perhaps minority party executives try harder to appeal to a broader set of voters because they have too few co-partisans (or co-ideologues) to rely on them exclusively. In contrast, majority governors do not have to seek public support from outside of the ranks of their co-partisans (or co-ideologues) in the mass public. Hence, their public style may be more intensely partisan or ideologically extreme or rigid. As a consequence these governors may more easily (or intentionally?) alienate citizens who are not members of their party or ideological temperament. Turning to national factors, governors of the president’s party can expect to see a 0.11 change in their approval with every percentage point change in presidential approval.Although highly statistically signiﬁcant, the substantive implications of presidential approval do not appear major.A one standard deviation change in presidential approval (11.2) translates into only about a 1.2 percentage point change in gubernatorial approval. A change from the lowest (11) to the highest level of presidential (82) approval produces a 7.8 point shift in gubernatorial approval, but such great swings in presidential approval are rare. Results also indicate that governors see a drop of about 6.13 approval points when they come from the president’s party, indicating that the presidency aﬀects gubernatorial approval beyond the swings in presidential approval. If we drop the presidentgovernor same party dummy from the estimation, the presidential approval variable fails to reach statistical signiﬁcance, also indicating a more complex relationship than the simple presidential approval spillover eﬀect. Finally inﬂation aﬀects gubernatorial approval. The regression coeﬃ-

The State Economy, the National Economy, and Gubernatorial Popularity 115

cient points to a –0.07 drop in approval for each percent point increase in inﬂation, which suggests that gubernatorial approval substantively is not highly sensitive to national inﬂation tides. Moreover, inﬂation moves very slightly from month to month, on average 1.4 percent in these data. Still, over a longer period of time, sustained inﬂationary spirals may cumulate into substantively signiﬁcant impacts on gubernatorial approval. In comparing the impact of the variables on gubernatorial popularity, the state-level factors, when signiﬁcant, seem to have greater impact than the national-level factors. Two of the national-level factors, presidential approval and inﬂation, have at best marginal substantive implications for gubernatorial approval, although national unemployment and being of the president’s party strongly aﬀect gubernatorial substantively.

Conclusion The literature on unemployment and gubernatorial approval is rife with controversy, with some studies ﬁnding state unemployment eﬀects (Barth and Ferguson, 2002; Crew et al., 2002; Hansen, 1999a, 1999b; Orth, 2001), but others failing to do so (Adams and Squire, 2001; Crew and Weiher, 1996). Our analysis reformulates the issue into one in which voters compare national and state unemployment levels. We argue that the federal structure of the political system, the integration of state economies into the national economy, and the limited economic policy tools available to governors presents barriers for citizens who want to hold their governor accountable for the state economy. Voters resolve this complex situation by holding the governor accountable for how well or poorly the state’s unemployment level fares compared to the national unemployment level. The actions of governors help the public in rendering such a comparative assessment. Governors often possess policy tools that may marginally aﬀect unemployment, making a focus on jobs substantively reasonable. They publicize their eﬀorts to bring jobs to the state, while electoral opponents often chastise the incumbent administration’s performance with regard to jobs. The focus on relative unemployment provides a way for citizens to grade their governor, while also recognizing the fact that the national economy and economic policy decisions by national policy makers will aﬀect the performance of the state’s economy. Under diﬀerent speciﬁcations, using diﬀerent control variables, and diﬀerent estimation assumptions, we ﬁnd that relative unemployment levels aﬀect gubernatorial approval as our model predicts. Our ﬁndings are robust to these alterations in estimation, and the magnitude of eﬀect is relatively stable across

116

Cohen and King

the various speciﬁcations. And our results support our theory, which assumes that state electorates are sophisticated enough to distinguish between the state and national economies when evaluating the governor’s performance in oﬃce, and accordingly, holds the governor accountable for the state economy, is borne out by the analysis. This study also demonstrates the beneﬁt of using a more comprehensive data collection on gubernatorial approval than used in past research. The JAR data set, which provides 2,000 cases for analysis, allows us to include a larger array of control variables than past research has been able to do. Our theoretical model and analysis suggests that the public rates governors based on the outcomes and policies over which they have come control or impact. One direction for future research is to extend to other areas over which governors possess some ability to aﬀect policy and performance, and which are often high priorities in gubernatorial agendas, such as education.

Notes 1. The large literature of presidential approval has grappled with assessing the relative impact of unemployment, inﬂation, and other economic conditions on presidential approval. Although no consensus exists over which economic factors mostly strongly aﬀect presidential approval, inﬂation at times appears to have the stronger eﬀect than unemployment in aggregate time series studies of presidential approval. See the detailed review of the literature in Gronke and Newman (2003). 2. This conclusion is based on an analysis that pools data from eight states. In a separate analysis focusing only on California, Hansen (1999a) ﬁnds the relative unemployment variable to be signiﬁcant. However, Hansen spends little time discussing the theoretical basis for using this variable. 3. It makes little diﬀerence whether we use lagged or non-lagged forms of these variables. Analysis, not shown, produces similar results. 4. State (national) income also highly correlates with state (national) unemployment and thus will also produce multicollinearity. 5. The state-national unemployment diﬀerence is correlated with state unemployment (r = 0.76, p < 0.0001), but is not related to national unemployment (r = –0.0001, p > 0.99). 6. This data were collected by Thad Beyle, Richard Niemi, and Lee Sigelman, and funded by the NSF (Grant No. SES-9974176).We thank them for allowing us to use the data. 7. Some 76 cases date prior to 1980, almost all from California and Iowa. Other cases are lost due to missing or incomplete data. For instance, we require monthly observations, but JAR poll data does not always provide the month of the poll. We also lost some cases when a presidential transition occurs, because we code the lagged presidential popularity variable as missing data during the transition month ( January) from one president to another. All told, we have complete

The State Economy, the National Economy, and Gubernatorial Popularity 117 and reliable data on 2,016 cases.The data are not evenly distributed across states, yet no subset of state dominates the data set either.The top ﬁve states (California, Connecticut, New York, Alabama, Florida), each with 5 percent or more of the cases, account for only 29.5 percent of cases. California, the leader, accounts for only 7 percent.The standard deviation of the percentage of cases per state is also quite tight at 1.6. Details can be obtained from the authors. 8. Beyle, Niemi, and Sigelman (2001) also recommend this correction. 9. The data set contains general approval as well as policy speciﬁc approval questions.We only use the general approval questions. 10. In several instances, more than one ﬁrm asked a gubernatorial evaluation question of a state’s citizens. We decided to keep both readings in the data set as separate data points because we have no assured way to combine them into one reading. 11. These data from John McIver’s web page http://socsci.colorado.edu/ ~mciverj/wip.html. 12. Moreover, the maverick behavior of many independents (e.g., Jesse Ventura) may alienate voters. Many independents came to oﬃce in three-way contests, and thus, rarely enjoyed majority vote support in winning the election. 13. Gronke (2001) notes that population size and diversity are conceptually distinguishable, but in this data set the two are highly correlated (r = 0.50, p = 0.000). 14. Kernell (1978) argues that time is atheoretic and tries to model it substantively. 15. National income cannot be added because of its high correlation with national unemployment, which will produce multicollinearity. Because our theoretical focus is on the importance of unemployment for gubernatorial approval, we retain the unemployment variables. We also experimented with economic perceptions data by entering monthly (lagged) consumer economic conﬁdence and monthly (lagged) consumer economic expectations from the University of Michigan’s Survey of Consumers. Results (not shown) indicate that both variables are statistically signiﬁcant and signed properly and do not detract from other variables in the estimation. Given the theoretical importance that economic perceptions have taken in recent studies of presidential approval, it would be of interest to incorporate economic perceptions data into studies of gubernatorial approval. Howell and Vanderleeuw (1990) and Orth (2001) use perceptual economic data in their studies from their surveys of Louisiana and Michigan voters, respectively. 16. The xtreg procedure in Stata 8.0 was used to estimate the ﬁxed eﬀects model. 17. Above we also noted complications in using the JAR data. Thus, we also ran numerous other estimations to ensure that our results are robust and that these identiﬁed data set complications do not aﬀect our ﬁndings. Hence, we add controls for polling ﬁrm, number of response categories, the response scale used, and governor into the OLS, robust regression, and state ﬁxed eﬀects estimations. Although overall model ﬁt usually improves with these additional control variables, in each case, relative unemployment and national unemployment retain their signiﬁcant eﬀects, although in some instances the impact of some control variables change. Details of these analyses can be obtained from the authors. 18. The modest R2 of 0.27 deserves comment.Variation in question wording

118

Cohen and King

and polling ﬁrms, which may use diﬀerent question wording, question placement, and sample frames, may induce a degree of sampling and other error when comparing across these surveys. In estimations not shown, when dummy variables are added for the number of response categories, the year of the survey, the polling ﬁrm, and the individual governor, the R2 approaches 0.75. The addition of these dummy variables, however, has little substantive impact on the substantive variables of interest. Details of these results may be obtained from the authors. 19. The strong correlation between national and state unemployment precludes us from arguing that national, as opposed to the state-level unemployment, is aﬀecting voters. 20. This variable becomes signiﬁcant with the correct sign when controls for governor dummies are applied. Details of these analyses can be obtained from the authors. 21. Both of these variables are unstable when other variables, such as gubernatorial dummies, are entered as controls because of the multicollinearity. Details of these analyses can be obtained from the authors. 22. With the added controls, such as governor, the ideology variable tends not to maintain statistical signiﬁcance. But the partisan advantage variable keeps its statistical signiﬁcance and unexpected sign no matter the controls that are used. Details can be obtained from the authors.

References Adams, Greg, and Peverill Squire. 2001. “A Note on the Dynamics and Idiosyncrasies of Gubernatorial Popularity.” State Politics & Policy Quarterly 1 (Winter):380–393. Adams, James D., and Lawrence W. Kenny. 1989. “The Retention of State Governors.” Public Choice 62 ( July): 1–13. Atkeson, Lonna Rae, and Randall W. Partin. 1995. “Economic and Referendum Voting: A Comparison of Gubernatorial and Senatorial Elections.” American Political Science Review 89 (March): 99–107. Barth, Jay, and Margaret R. Ferguson. 2002.“American Governors and Their Constituents:The Relationship between Gubernatorial Personality and Public Approval.” State Politics & Policy Quarterly 2 (Fall): 268–282. Beyle,Thad, Richard G. Niemi, and Lee Sigelman. 2002. “Gubernatorial, Senatorial, and State-Level Presidential Job Approval Ratings: The U.S. Oﬃcials Job Approval Ratings ( JAR) Collection.” State Politics & Policy Quarterly 2 (Fall): 215–229. Bond, Jon R., and Richard Fleisher. 2001. “The Polls: Partisanship and Presidential Performance Evaluations.” Presidential Studies Quarterly 31 (September): 529–541. Brace, Paul. 1993. State Government and Economic Performance. Baltimore, MD: Johns Hopkins University Press. Brace, Paul, and Barbara Hinckley. 1991.“The Structure of Presidential Approval: Constraints Within and Across Presidencies.” Journal of Politics 53 (November): 993–1017. Carsey,Thomas M., and Gerald C.Wright, Jr. 1998.“State and National Forces in

The State Economy, the National Economy, and Gubernatorial Popularity 119 Gubernatorial and Senatorial Elections.” American Journal of Political Science 42 ( July): 994–1002. Cohen, Jeﬀrey E. 1983. “Gubernatorial Popularity in Nine States.” American Politics Quarterly 11 (April): 219–235. Conover, Pamela Johnston, Stanley Feldman, and Kathleen Knight. 1986.“Judging Inﬂation and Unemployment:The Origins of Retrospective Evaluations.” Journal of Politics 48 (August): 565–588 Crew, Robert E., Jr. 1998. “Gubernatorial Leadership: Testing a Preliminary Model.” Social Science Journal 36 (1): 15–27. Crew, Robert E., Jr., David Branham, Gregory R. Weiher, and Ethan Bernick. 2002. “Political Events in a Model of Gubernatorial Approval.” State Politics & Policy Quarterly 2 (Fall): 283–297. Crew, Robert E., Jr., and Gregory R. Weiher. 1996. “Gubernatorial Popularity in Three States: A Preliminary Model.” Social Science Journal 33 (1): 39–55. Delli Carpini, Michael, and Scott Keeter. 1996. What Americans Know About Politics and Why It Matters. New Haven:Yale University Press. Dometrius, Nelson C. 2002. “Gubernatorial Approval and Administrative Inﬂuence.” State Politics & Policy Quarterly 2 (Fall): 251–267. Erikson, Robert S., Michael MacKuen, and James A. Stimson. 2002. The Macropolity. New York: Cambridge University Press. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Ferguson, Margaret Robertson. 2003. “Chief Executive Success in the Legislative Arena.” State Politics & Policy Quarterly 3 (Summer): 158–182. Grady, Dennis. 1991.“Managing the State Economy:The Governor’s Role in Policymaking.” In Gubernatorial Leadership and State Policy, eds., Eric B. Herzik and Brent W. Brown. New York: Greenwood. Gronke, Paul. 2001. The Electorate, the Campaign, and the Oﬃce: A Uniﬁed Approach to Senate and House Elections. Ann Arbor: University of Michigan Press. Gronke, Paul, and Brian Newman. 2003.“From FDR to Clinton, from Mueller to ?? A Field Essay on Presidential Approval.” Political Research Quarterly 56 (December): 501–512. Haller, H. Brandon, and Helmut Norpoth. 1997. “Reality Bites: News Exposure and Economic Opinion.” Public Opinion Quarterly 61: 555–575. Hansen, Susan B. 1999a.“Governors’ Job Performance Ratings and Unemployment: The Case of California.” State and Local Government Review 31 (Winter): 7–17. ———. 1999b. “‘Life is not Fair’: Governors’ Job Performance Ratings and State Economies.” Political Research Quarterly 52 (March): 167–188. Hendrick, Rebecca M., and James C. Garand. 1991.“Variation in State Economic Growth: Decomposing State, Regional, and National Eﬀects.” Journal of Politics 53 (November): 1093–1110. Herzik, Eric B. 1991. “Policy Agendas and Gubernatorial Leadership.” In Gubernatorial Leadership and State Policy, eds., Eric B. Herzik and Brent W. Brown. New York: Greenwood. Holbrook, Thomas, and James C. Garand. 1996. “Homo Economous? Economic

120

Cohen and King

Information and Economic Voting.” Political Research Quarterly 49 ( June): 351– 375. Howell, Susan E., and James M.Vanderleeuw. 1990. “Economic Eﬀects on State Governors.” American Politics Quarterly 18 (April): 158–168. Kenney, Patrick J., and Tom W. Rice. 1983. “Popularity and the Vote: The Gubernatorial Case.” American Politics Quarterly 11 (April): 237–241. Kernell, Samuel. 1978. “Explaining Presidential Popularity.” American Political Science Review 72 ( June): 506–522. King, James D. 1999.“Wyoming’s Gubernatorial Elections: Retrospective Voting in Good Times and Bad.” Comparative State Politics 20 (No. 4): 17–26. ———. 2001.“Incumbent Popularity and Vote Choice in Gubernatorial Elections.” Journal of Politics 63 (May): 585–597. Lowry, Robert C., James E. Alt, and Karen E. Ferree. 1998. “Fiscal Policy Outcomes and Electoral Accountability in American States.” American Political Science Review 92 (December): 759–774. MacDonald, Jason A., and Lee Sigelman. 1999. “Public Assessments of Gubernatorial Performance: A Comparative State Analysis.” American Politics Quarterly 27 (April): 201–215. Morgan, David R., and L. A.Wilson. 1990.“Diversity in the American States: Updating the Sullivan Index.” Publius 20 (No. 1): 71–81. Mueller, John E. 1970. “Presidential Popularity from Truman to Johnson.” American Political Science Review 64 (March): 18–34. Nicholson, Stephen P., Gary M. Segura, and Nathan D.Woods. 2002.“Presidential Approval and the Mixed Blessing of Divided Government.” Journal of Politics 64 (August): 701–720. Niemi, Richard G., John Bremer, and Michael Heel. 1999.“Determinants of State Economic Perceptions.” Political Behavior 21 ( June): 175–193. Orth, Deborah A. 2001. “Accountability in a Federal System: The Governor, the President, and Economic Expectations.” State Politics & Policy Quarterly 1 (Winter): 412–432. Simon, Dennis M. 1989. “Presidents, Governors, and Electoral Accountability.” Journal of Politics 51 (May): 286–304. Simon, Dennis M., Charles W. Ostrom, Jr., and Robin F. Marra. 1991. “The President, Referendum Voting, and Subnational Elections in the United States.” American Political Science Review 85 (December): 1177–1192. Squire, Peverill, and Christina Fastnow. 1994. “Comparing Gubernatorial and Senatorial Elections.” Political Research Quarterly 47 (September): 705–720. Stimson, James A. 1976. “Public Support for American Presidents: A Cyclical Model.” Public Opinion Quarterly 40 (1): 1–22. Svoboda, Craig T. 1995. “Retrospective Voting in Gubernatorial Elections: 1982 and 1986.” Political Research Quarterly 48 (March): 135–150.

Chapter

Ideological Cleavage, Political Competition, and Policy Making in the American States

7

Charles Barrilleaux

The most frequent treatment of the eﬀects of ideology on policy outputs in the states suggests a politics of agreement wherein the dominant ideological persuasion within any state elects a majority to the legislature and, as a direct consequence, gets the lion’s share of rewards from state government. These rewards may be public policies, judicial decisions, and other products of the state (see, e.g., Brace, Sims-Butler,Arceneaux, and Johnson, 2002; Erikson, Wright, and McIver, 1993; Langer, 2002; Norrander and Rivera, 2002). Even though that statement may be for the most part accurate, it is possible that the measure of ideology most often employed—the average liberalism or conservatism of a state’s citizenry1 —masks some important variation in public opinion that may exist. States whose citizens have an average ideological liberalism of 65 (on a scale of 0–100) with a standard deviation of 5 may exhibit diﬀerent politics than a state whose citizenry reports a mean score of 65 but with a standard deviation of 2. How this matters is unclear; policies may be easier to enact in the ﬁrst state and more conﬂictual in the latter, and this increasing conﬂict may temper the eﬀects of opinion liberalism on public policy outputs. In short, diﬀerences of opinion may matter. In this chapter I consider not only the tendency of opinion within states as an inﬂuence on public policies, but consider as well the diversity of opinion. Political scientists have largely settled the “politics versus economics” debate and have for nearly 20 years been developing increasingly nuanced 121

122

Barrilleaux

models of the eﬀects of variations in political institutions and political attitudes and behavior on public policies in the states. Perhaps the most frequently invoked explanation of the eﬀects of politics on policy outcomes is provided by Robert Erikson, Gerald Wright, and John McIver (1989). Building on their demonstration of the eﬀects of ideology (Wright, Erikson, and McIver, 1987) on public policies, they construct a model that elaborates the manner in which citizen preferences for liberal policy, expressed as mean liberal ideology, inﬂuence the ideological positions of state party elites and legislators.This selection of liberal legislators leads in turn to policy outcomes consistent with public preferences. In short, they describe a simple model of representation in which citizens more or less successfully identify and elect representatives whose views are consonant with their own, those legislators reward those citizens with policies that are consistent with voters’ ideological preferences, and parties oﬀer distinct choices for voters.Their research tells us that representative democracy, on average, works: citizens have ideological and policy preferences, they elect representatives who reﬂect those preferences, and those politicians produce policies that reﬂect citizen demands. Subsequent research largely validates their results; the link between opinion and policy has proved robust under a number of speciﬁcations and with diﬀerent data, and is now part of the canon of state-level policy research. Even though the Erikson,Wright, and McIver model (hereafter EWM model) is largely accepted, some questions about the nature of policy making in the states remain.Their model diﬀers from some treatments of public policy in the states in that it for the most part leaves out considerations of political competition, which were featured in most early policy outputs studies and retain theoretical currency. If competition matters, it suggests a slightly diﬀerent politics than is implicit in the “politics of agreement” that is suggested by the prevailing explanation for the eﬀects of ideology on policies. This research tests the eﬀects of three forms of political competition on state government policy making.

Political Competition, Political Parties, and Public Policy The relationship between political competition and public policy appears to be straightforward. The common wisdom, rooted in work by Schattschneider (1942), Key (1949), and others, holds that more competition between political parties leads politicians and parties to reward (through public policy) more groups. More inclusive policy typically contains more rewards for the have-nots and is expressed as more liberal public policy. The majority of quantitative research addressing the role of

Ideological Cleavage, Political Competition, and Policy Making 123

competition in policy determination concerns the eﬀects of party control and competition on policy. Some authors report evidence of substantial party-policy linkages (Barrilleaux, 1997; Barrilleaux, 2000; Barrilleaux, Holbrook, and Langer, 2002; Budge and Hoﬀerbert, 1990; Dye, 1984; Garand, 1985; Jennings, 1979; Plotnick and Winters, 1990; Wright and Schaﬀner, 2002), contrasting with reports of no signiﬁcant eﬀects of party competition or control on policy outputs (Dye, 1966; Ringquist, Hill, Leighley, Hinton-Anderson, 1998;Winters, 1976).2 The preponderance of evidence now suggests that parties make a diﬀerence under some conditions, but the attention placed on parties obscures consideration of other facets of political expression and competition. Concepts of political competition are often restricted to the competition between parties in elections. A number of features of American state politics, such as incumbency advantages and the former dominance of Democrats in the South and Republicans in parts of New England and the West, make electoral competition a weak measure of political competition (Chubb, 1987; Fiorina, 1994; King, 1989; Stonecash, 1987). Even if electoral competition were not diminished by institutional developments, a single indicator cannot adequately measure so complex a phenomenon. Consider, for example,V. O. Key’s description of the politics of some of the then one-party dominated southern states in Southern Politics (1949). Although the South was avowedly one-party during that time, dominated by Democrats, Key describes several states as hosting ﬁerce political competition between and among factions despite their being uncompetitive so far as two-party conﬂicts are concerned. Louisiana is probably the best documented among the southern states, although others had similar, but less ﬂamboyant, situations. In his account of Earl Long’s 1959 run for the Louisiana statehouse, A. J. Liebling described Louisiana’s politics as being “of an intensity and complexity that are matched . . . only in the republic of Lebanon” (Liebling, 1961, 18). Liebling described a political system that involves multiple factions, rampant side payments, and otherwise lessthan-standard arrangements, all of which suggest tumultuous and highly competitive politics. Despite all of this, a person basing his or her understanding of that state’s politics on standard measures of competition would learn that Louisiana’s politics in the era Liebling described were uncompetitive (see, e.g., Ranney, 1965). Surely politically relevant competition is manifest in arenas other than elections. Because we know that electoral competition aﬀects outcomes at least some of the time (see, e.g., Jennings, 1979), it follows that other forms of competition may also inﬂuence public policy outcomes. I employ a view of political competition drawn from Douglas Rae and

124

Barrilleaux

Michael Taylor’s research on political cleavages (1970). Rae and Taylor identify three dimensions of cleavage, or competition, within political systems: citizen ascriptive traits, citizen attitudes, and citizen behaviors. Ascriptive traits include race, religion, ethnicity, nationality, and other politically relevant personal characteristics that are determined at birth. Attitudinal cleavages are opinion based and may include ideology or preferences. Behavioral cleavages include voting or organizational membership (Rae and Taylor, 1970, 1). Rae and Taylor argue that each form of competition has meaningful independent eﬀects on governmental policies (1970, 36–44). Ascriptively based competition is, as the term implies, innate in a polity. It may spring from population diversity, as more diverse populations should contain more diﬀerences of opinion regarding appropriate government actions. Although factional conﬂicts are unsettling, one view holds that their ultimate eﬀect in a democracy is to expand the budget so that more individuals and groups receive rewards (see, e.g., Olson, 1982; Sullivan, 1973). In this line of reasoning, and assuming relatively open access to channels of political expression and demand as would be found in contemporary democracies, politicians respond to diverse demands with policies that encompass larger blocs of interests, leading states with more diverse populations to have more liberal public policies as a byproduct of the admittedly not directly observable competition for beneﬁts among contesting ascriptive groups. But from another view, diﬀerences breed illiberal public policies, especially when those ascriptive diﬀerences are between or among races (Hero and Tolbert, 1996). Hero and Tolbert argue that liberal public policies are most often found in states in which there is a “moralistic culture . . . characterized by a concern for the ‘commonwealth’ ” (867). This commonwealth notion, drawn from Daniel Elazar’s (1984) work on political culture, is argued to exist most strongly in states in which “individuals are less racially/ethnically divided, in more homogeneous contexts” and, in a statement that echoes Banﬁeld and Wilson’s ethos theory, they claim that “the commonwealth outlook is not necessarily extended to those of diﬀerent racial/ethnic backgrounds (Hero and Tolbert, 1996, 867).The gist of the Hero and Tolbert argument is akin to something tested by Plotnick and Winters (1985): that citizens seem to be more generous to people who are similar racially. Plotnick and Winters report no signiﬁcant results of their hypothesis test, which is undertaken using a fully speciﬁed model. Hero and Tolbert’s conclusion is drawn from evaluation of a series of models in which measures of public policy are regressed on two variables—racial diversity and white ethnic diversity.They ﬁnd racial diversity

Ideological Cleavage, Political Competition, and Policy Making 125

to be linked negatively to all of the measures of policy liberalism, which leads them to the conclusion that race is a driving factor in state policy decisions. Hero (1998) conducts a series of similar analyses and, from those results, expands the claim to argue that racial diﬀerences are the single most important factor inﬂuencing American state politics, and that racial diﬀerences breed antipathy rather than understanding as is suggested by the prior diversity hypothesis. Hero and Tolbert suggest that political competition in homogeneous societies is best viewed as a family feud: there are shared boundaries that proscribe arguments, and an underlying understanding about “who gets what.” The estimates presented here, which test the two hypotheses speciﬁed by Hero in his work with Tolbert (1996) and alone (1998), also contain controls for other plausible inﬂuences and as result may provide a more stringent test of the hypotheses. Hence we have two contending arguments about ascriptive diversity: that it leads to increased understanding, log rolling, and eventually to more generous policies; and the opposite, that it leads to disagreements and limits public spending. Attitudinal competition springs from competing ideas: as the abundance of opinions regarding a policy or other government act increases, so does attitudinal competition. Attitudinal competition may work to the detriment of democratic outcomes. Most notably, it may result in individuals or groups’ intensely held positions being given undue weight and reward by oﬃceholders (Dahl, 1956). In the more general case, the spread of opinion within a state should inﬂuence legislative outputs.That is, two states may be equally “liberal” as indicated by some measure of mean ideological positions among their citizens. However, the spread of opinion may be greater in one than another, that is, identical means can be calculated from distributions that are shaped quite diﬀerently. Presumably, legislators making decisions in a state in which the opinion distribution is clustered fairly tightly are confronted with a diﬀerent set of decisions than are legislators making decisions from one with widely disparate ideological positions. Presumably, a broader array of opinion forces legislators to consider more points of view than they might be forced to consider where there is greater opinion homogeneity, which in turn should lead to the production of policies that incorporate more views and are as result less clearly representative of the majority. Behavioral competition is that most typically considered in policy outputs research. It refers to acts such as organization membership, electoral competition, and other phenomena that require citizen action. The most frequently invoked indicator of behavioral competition is that between or among political parties or candidates. Where parties compete, policies

126

Barrilleaux

should be tailored to appeal to a broader spectrum of the populace and, as result, to be more closely attuned to public opinion.Where opinion leans toward the liberal end of the spectrum, this should result in more liberal policies (i.e., policies that are more redistributive in their intent) inasmuch as the poor are more likely to be included in politicians’ calculations of demand where have-nots exercise political voice (Barrilleaux, Holbrook, and Langer, 2002). Competition for oﬃce in the states may be conceived as competition between parties for seats in government, or it may be thought of irrespective of party strength and in light of electoral competition, which refers to how diﬃcult it is to gain election in a state. Interparty competition is comparatively easy to measure. Existing measures typically consider electoral support of parties, for example, percentage voting for one of the two parties, party control, or both (Bibby and Holbrook, 2003; Patterson and Caldeira, 1983).A clear distinction may be drawn between interparty competition and electoral competition, where the latter measures the diﬃculty of gaining election in a state (Holbrook and Van Dunk, 1993). The two have distinct and important eﬀects on policy making in the states (Barrilleaux, 1997). Each is important but neither captures the full import of competition within a state; the eﬀects of liberal party strength on public policies vary according to the competitiveness of elections within a state (Barrilleaux, Holbrook, and Langer, 2002). Electoral competition is conceived here diﬀerently, as the diversity of partisanship among citizens.The sorts of competition measured in previous research focuses on outcomes— either turnout or results. Here it focuses instead on inputs, the extent to which a state’s citizens show allegiance to diﬀerent parties. I expect states with more diverse party membership to enact more liberal public policies, all things being equal, because diﬀerences in partisan allegiance force elective oﬃcials to appeal to more people, and appealing to a more heterogeneous group should result in consideration of and reward to a diversity of interests.This should result in more policies being enacted, and with more policy enactments should come more liberal policy making.

Indicators of State-Level Competition I construct four measures of competition for use in this research. The measures are calculated following a formula for diversity provided by Stanley Lieberson (1969): p

AW = 1 – [ 6Yk2 V ] k-1

(1)

Ideological Cleavage, Political Competition, and Policy Making 127

Where AW = the diversity of the item in question, Yk = the proportion of the population that falls into a category within each of the variables, V = the number of variables, and p = the number of categories within all the variables (Sullivan, 1973). Calculation for units of analysis in which there is zero diversity on some qualitative trait yields a diversity score of 0. Units where there is total diversity yields a value of 1.0. The measure is interpreted as the likelihood that two subjects, sampled randomly with replacement, diﬀer on some category of interest. Lower scores represent lower population diversity.The Lieberson measure has been used in state politics in the past and the application is often described as the Sullivan index, following John Sullivan’s (1973) application of the method to measure diversity in state-level population and political characteristics. To illustrate, calculating diversity in ideology requires consideration of one variable with four categories for each state: proportion of the population that self-identify as conservative, middle-of-the-road, and liberal, respectively, and proportion that does not know.To calculate the value for Alabama, using data published in Wright, Erikson, and McIver (1985): Ideological DiversityAL = 1 – [((.155)2 + (.352)2 +(.361)2 + (.132)2/1)] = 0.71

(2)

This tells us that two Alabamans selected randomly with replacement have about a 70 percent chance of diﬀering ideologically. Note that the division by 1 reﬂects that only one characteristic is being considered here; if we had more than one variable to measure ideological diversity we would divide by the number of variables. So, for example, measuring a concept that requires more than one variable, like population diversity, which considers education, income, ethnicity, and other traits, requires that we divide by the number of variables considered to ensure that the index does not take on a value greater than 1. I measure the diversity of four characteristics of states: ideology, as illustrated above, population ethnicity and religion, race, and partisanship. Following Rae and Taylor’s formulation, ideological diversity measures attitudinal diversity, and population and racial diversity measure ascriptive diﬀerences. Diversity in partisanship captures the probability that two citizens diﬀer in their partisan allegiance. A fourth measure, the strength of organized interest groups in the states, is included in the test of competition as a further measure of behavioral competition. It is the sole competition measure that is not captured in the diversity framework (because suitable data are not available) and as a result strength of its coeﬃcients is

128

Barrilleaux

not readily comparable to those of the other measures, all of which are scaled identically. Values for each of the measures are displayed in Table 7.1, below. Data for the ideology measure are drawn from Erikson, Wright, and McIver (1989) and represent ideology in about 1980.3 Oklahoma, with a value of 0.696, is the median state. The most ideologically diverse states are Arkansas and Kentucky, where there is a 73 percent chance that two citizens will profess diﬀerent ideological leanings. The least ideologically diverse state is Idaho, where citizens diﬀer only about 63 percent of the time.The measure of ideological diversity is correlated weakly and negatively (–0.32) with Erikson, Wright, and McIver’s (1993) measure of ideological polarization in the states, which is the diﬀerence in the mean liberal and mean conservative scores for their ideology measure in each state. The diversity measure captures whether citizens in a state diﬀer in their ideological leanings but includes moderates along with liberals and conservatives and as result does not capture polarization. Population diversity is measured using indicators of social, economic, ethnic, and religious group membership, following Sullivan (1973). Speciﬁcally, the index contains information about population diversity in six areas—education, income, occupational status, housing, ethnicity, and religion (Sullivan 1973, 71). Categories for education include the proportion of the population with less than an eighth grade education, those with some high school, those with a high school education, those with some college, and those who are college graduates. Data were collected from published reports from the 1980 U.S. Census of Population; thus data presented here represent state-level population diversity circa 1980. Income categories are households with incomes under $8,000, incomes between $8,000 and $15,000, incomes between $15,000 and $25,000, and incomes greater than $25,000. Religious diversity is included in the measure as the proportion Catholic, the proportion Jewish, and the proportion Protestant, and other. As measured, the most diverse state is New York, where citizens diﬀer in these terms about 55 percent of the time.The least diverse state is Alabama, where ethnicity and religion diﬀer only about 38 percent of the time. Maryland is the median state, with a score of about 0.45. Sullivan (1973) links population diversity to more generous public spending for welfare, leading to the expectation that it will lead to more liberal policy enactments. Sullivan’s measure does not include information about race, and the addition of race reverses the signs of the observed relationships. Racial diversity is measured using 1980 U.S. Census of Population data that identify white, black, native American, and other races by state.4 Because some categories are so small, they are lumped into an

Table 7.1. State Ideological, Population, Racial, and Partisan Diversity Scores and Ranks, and Group Strength Scores Ideological Diversity State

Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana Nebraska Nevada New Hampshire New Jersey New Mexico New York N. Carolina N. Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island S. Carolina S. Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming

Population Diversity

Racial Diversity

Partisan Diversity

score

rank

score

rank

score

rank

score

rank

Groups

0.717

6*

0.386 0.397 0.439 0.375 0.482 0.426 0.491 0.442 0.458 0.406 0.479 0.391 0.478 0.410 0.420 0.412 0.406 0.459 0.438 0.435 0.496 0.439 0.439 0.388 0.428 0.420 0.431 0.435 0.452 0.500 0.456 0.525 0.392 0.446 0.442 0.390 0.418 0.466 0.497 0.392 0.424 0.386 0.455 0.381 0.458 0.415 0.421 0.386 0.454 0.394

48* 39 22 50 6 28 5 19 11 38 7 43 8 36 32 35 37 10 23 25 4 21 20 45 27 31 26 24 16 2 13 1 41 17 18 44 33 9 3 42 29 46 14 49 12 34 30 47 15 40

0.446 0.444 0.468 0.420 0.503 0.444 0.495 0.482 0.471 0.469 0.517 0.396 0.509 0.432 0.419 0.430 0.430 0.526 0.429 0.484 0.488 0.468 0.440 0.463 0.458 0.431 0.440 0.451 0.441 0.517 0.505 0.542 0.450 0.450 0.466 0.426 0.421 0.486 0.486 0.458 0.441 0.428 0.495 0.386 0.448 0.461 0.427 0.395 0.464 0.402

29* 31 17 45 7 30 8 14 15 16 3 48 5 36 46 39 38 2 40 13 10 18 35 21 24 37 34 25 32 4 6 1 26 27 19 43 44 12 11 23 33 41 9 50 28 22 42 49 20 47

0.547

48*

0.676 0.657 0.684 0.678 0.671 0.656 0.673 0.611

26 33 18 24 28 35 27 45

0.678 0.691 0.690 0.694 0.692 0.650 0.570 0.679 0.654 0.619 0.696 0.667 0.648 0.679 0.689 0.695 0.701 0.681 0.688 0.654 0.700 0.669 0.701 0.694 0.632 0.684 0.690 0.584 0.689 0.657 0.681 0.661 0.692 0.622 0.703 0.651 0.656 0.694 0.638

25 12 14 7 10 39 47 23 36 44 5 30 40 22 16 6 2 21 17 37 4 29 3 8 42 19 13 46 15 32 20 31 11 43 1 38 34 9 41

1.00 1.00 2.00 2.00 2.00 3.00 4.00 4.00 1.00 2.00 2.00 2.00 3.00 3.00 3.00 3.00 2.00 1.00 3.00 3.00 3.00 3.00 4.00 1.00 3.00 2.00 2.00 2.00 3.00 3.00 1.00 3.00 3.00 3.00 2.00 2.00 2.00 3.00 3.00 1.00 3.00 1.00 2.00 2.00 4.00 2.00 2.00 1.00 3.00 2.00

0.675 0.728 0.701 0.683 0.703 0.701 0.694 0.717

43 1 16 41 13 15 29 7

0.627 0.689 0.693 0.687 0.691 0.728 0.694 0.688 0.697 0.696 0.696 0.681 0.719 0.712 0.699 0.696 0.683 0.673 0.695 0.698 0.702 0.704 0.664 0.690 0.696 0.687 0.700 0.696 0.704 0.655 0.726 0.695 0.659 0.719 0.696 0.688 0.708 0.684 0.706

48 34 31 38 32 2 30 35 20 22 21 42 4 8 18 26 40 44 27 19 14 11 45 33 25 37 17 24 12 47 3 28 46 5 23 36 9 39 10

*For ranks, 1 is the most diverse. Sources: See notes to Appendix Table 7.1.

130

Barrilleaux

“other” category for purposes of this analysis. Also seen in Table 7.1, the most racially diverse states are spread across the country rather than being clustered in one region. New York is the most racially diverse state (0.542), followed by Louisiana (0.522), Hawaii (0.517), New Jersey (0.517), and Illinois (0.509). The least racially diverse state is Utah (0.386), followed closely by Idaho (0.396) and West Virginia (0.395). As discussed above, racial diversity is expected to be linked to less liberal public policies. Diversity in partisanship is measured using the EWM data as well, for the mid-1980s.The data contain information about respondents’ reports of allegiance to the Democratic or Republican parties, of being Independents, or of providing no response.The most diverse state in terms of partisan identiﬁcation is Virginia, where there is a greater than 70 percent chance that two people selected with replacement will diﬀer in their partisan allegiance. The least diverse state in terms of partisan aﬃliation is Alabama, where people diﬀer only about 55 percent of the time. The median state is Colorado (0.678). Consistent with prevailing theories of the eﬀect of interparty competition on policy outputs, I expect policies to become more liberal, that is, to oﬀer more rewards for voters, as diversity in partisan allegiances increases.The expectation is based on the expectation that politicians who are confronted with multiple demands will attempt to provide beneﬁts to a broader array of citizens and that this diversity of demand (as expressed through diverse partisans) will have as a by-product more policies that reward liberal demands. The measure of interest group inﬂuence is drawn from Thomas and Hrebenar (1999). In the mid-1980s they organized researchers in 50 states to amass information on the organization and function of organized interests in the states, and they use the information to array the states on a continuum of interest group inﬂuence. States scored 1 are those dominated by organized interests and states scored 4 have weak interest group inﬂuence. Hrebenar and Thomas’s classiﬁcation scheme arrays state interest groups systems on a continuum from “dominant” to “subordinate,” with states categorized as dominant hosting organized interests that in fact dominate state politics. Five states—Alabama, Florida, Nevada, South Carolina, and West Virginia—are in this category. No states are ranked in the “subordinate” category, but four—Minnesota, Rhode Island, South Dakota, and Vermont—are among those in the penultimate category, termed “complementary/subordinate” in which groups are viewed as the weakest extant vis-à-vis their success in inﬂuencing state government. Twenty-ﬁve states fall in the dominant/complementary category and 16 states are in the complementary category, where groups and the government are thought to exist in a system in which groups either work with

Ideological Cleavage, Political Competition, and Policy Making 131

government or are constrained by other forces (Thomas and Hrebenar, 1999, 135–136). The liberalism of state public policies is expected to increase as the dominance of organized interests wanes, as weaker interest systems should give greater voice to voters, ceteris paribus, and legislators should pay attention to those demands. Given the coding of the measure I expect a negative coeﬃcient.

Testing the Rae and Taylor Competition Model A simple cross-sectional model of policy liberalism is estimated, using Erikson, Wright, and McIver’s index of state party liberalism (SPL) in about 1980 as the dependent variable.5 The model is expressed: SPL = b0 + b1 population diversity – b2 racial diversity + b3 ideological diversity + b4 partisan diversity + b5 group influence

(3)

Results of the estimation are provided in Table 7.2.The model explains about 62 percent of the variation in the dependent variable (adj. R2 = 0.62) and is statistically signiﬁcant (F = 18.56, p = 0.000). The two measures of ascriptive diversity inﬂuence policy liberalism as expected. A one unit increase in population diversity increases SPL by 167.64 units (t = 5.30; p < 0.01). As expected, diversity in race reduces the liberalism of policies: a one unit increase lessens policy liberalism by 86.66 units (t = –2.88; p < 0.01). The measure of attitudinal competition, ideological diversity, has a positive inﬂuence on state policy liberalism, as was expected. Increasing ideological diversity by one unit increases SPL by 68.86 units (p < 0.05). Finally, the two measures of behavioral competition exert positive inﬂuences on SPL. Increasing the diversity of citizens’ partisan attachments by one unit increases SPL by nearly 30 units (t = 1.78; p < 0.05). The measure of groups is scaled diﬀerently than each of the other measures in the model—from 1 to 4 (see Appendix Table 7.1)—so yields smaller coeﬃcients.A one unit movement in the measure, which captures the extent to which groups are inﬂuential in a state’s politics, increases SPL by 1.29 units (t = 1.70; p < 0.05). Because four of the ﬁve measures are scaled identically, as diversity measures that range between 0 and 1, it is possible to compare the relative eﬀects of each variable on the dependent variable. Population diversity is by far the strongest inﬂuence on policy liberalism, with a coeﬃcient nearly twice that of racial diversity, which comes in second in the “race of

132

Barrilleaux Table 7.2 Influences of Ascriptive,Attitudinal, and Behavioral Competition on State Policy Liberalism Independent Variables

Intercept Population diversity Racial diversity Ideological diversity Partisan diversity Interest groups

b (sb )

–103.764 (28.00) 167.64 (31.62) –86.66 (3.12) 68.863 (32.67) 29.652 (16.79) 1.289 (0.76)

t

Prob t (one-tail)

–3.705

0.0005

5.302

0.000

–2.877

0.003

2.108

0.02

1.766

0.04

1.697

0.05

Adj. R2 = 0.62; SEE = 3.22; F = 18.75, prob. F = 0.000, N = 47 Sources: See notes to Appendix Table 7.1.

the variables.”This calls into question Hero’s (1998) contention that race is the single strongest force aﬀecting state politics; at least insofar as the liberalism of policies is concerned, race is a distant second to broader population diversity. More importantly, these results show ascriptive competition to exert a much stronger inﬂuence on policies than attitudinal or behavioral diﬀerences. The sole measure of attitudinal competition, ideological diversity, exerts a positive inﬂuence as expected and is fairly close to the magnitude of the racial diversity measure. The diversity in partisanship measure is much weaker. Together, this suggests that competition may be most important in the sorts of demands it creates rather than in its manifestations in political behavior.

An Elaborated Model The model presented above is fairly naive. Here, I test the competition model more rigorously by including additional variables that are presented elsewhere as an explanation for variations in state policy liberalism. The best-known model is that initially presented in Erikson,Wright, and McIver (1989), which explains state policy liberalism as a function of three independent variables—citizen ideology, legislative liberalism, and Democratic legislative strength—and explains over 80 percent of the variation in state policy liberalism. It contains no direct observation of political competition.

Ideological Cleavage, Political Competition, and Policy Making 133

Instead, competition is treated implicitly in the inclusion of a measure of Democratic legislative strength in state legislatures inasmuch as that indicates something about party control of the legislature, a by-product of political competition.The measure’s meaning is unclear: a party that has full control of the legislature on winning a series of landslide elections can behave much diﬀerently than one that is in full control on winning a number of closely contested races. Presumably, the former has more latitude to stray from public opinion than the latter.This contention has been explored elsewhere (Barrilleaux, 1997; Barrilleaux, 2000; Barrilleaux, Holbrook, and Langer, 2002), and the preponderance of evidence suggests that consideration of both the strength of parties and the competitiveness of election is warranted. Hence, I use a model that contains measures of both party strength and electoral competition as a baseline model for comparison.The model is expressed: SPL = b6 + b7 ideology + b8 legislative liberalism – b9 democratic legislative strength + b10 electoral competition

(4)

The ﬁrst three terms are drawn directly from Erikson, Wright, and McIver (1989) and the fourth term is a measure of the average competitiveness of legislative races in the states (Holbrook and Van Dunk, 1993). Thus the model expresses state policy liberalism as a function of shared ideological liberalism, the mean ideological position of state legislators, the percentage of Democrat legislators in each state, and the diﬃculty of winning election in each state. Estimates for the model are contained in column 1 of Table 7.3. The model explains about 82 percent of the variation in the dependent variable and is statistically signiﬁcant. The results are consistent with those reported in Erikson, Wright, and McIver (1989, 1993) and in Barrilleaux (1997). Each predictor enters the equation signiﬁcantly and in the expected direction. SPL increases with increases in electoral competition, ideological liberalism, and electoral competition, and declines with increases in the strength of Democrats in the state legislature. Adding each of the measures presented in equation 3 to the baseline model (Table 7.3, column 2) results in neither model standing fully to additional speciﬁcation. None of the “new” diversity measures enters the full-ﬂedged model signiﬁcantly, and the Democratic legislative strength measure also drops out. Ideological liberalism, legislative liberalism, and electoral competition continue to exert statistically signiﬁcant inﬂuences on SPL. In addition, the magnitudes of the coeﬃcients for liberalism and electoral competition change little between the two speciﬁcations. This

134

Barrilleaux Table 7.3 Baseline, Overidentified, and Reduced Models of State Policy Liberalism 1 Baseline Model

Independent Variables

Intercept

b (sb )

t (one-tail p)

2 Overidentified Model b (sb )

t (one-tail p)

–1.73 (1.93) 32.29 (7.77) 2.52 (0.74)

–0.90 (0.19) 4.16 (0.000) 3.39 (0.001)

13.44) (29.27 36.22 (10.12) 1.49 (0.92)

0.46 (0.33) 3.58 (0.000) 1.62 (0.06)

–7.35 (3.04) 0.10 (0.055) NA

–2.42 (0.01) 1.79 (0.04) NA

Racial diversity

NA

NA

Ideological diversity

NA

NA

Partisan diversity

NA

NA

Interest groups

NA

NA

–0.93 (4.69) 0.09 (0.05) 32.74 (33.52) –14.43 (28.74) –39.49 (33.34) 10.42 (16.40) 0.48 (0.61)

–0.20 (0.42) 1.54 (0.07) 0.98 (0.34) –0.50 (0.32) –1.19 (0.12) 0.64 (0.26) 0.79 (0.22)

Ideological liberalism Legislative liberalism Democratic legislative strength Electoral competition Population diversity

3 Reduced Model b (sb )

t (one-tail p)

32.27 (23.43) 38.67 (8.94) 1.44 (0.62)

1.38 (0.09) 4.33 (0.000) 2.35 (0.01)

NA

NA

0.12 (0.04) 21.59 (14.76) NA

2.76 (0.01) 1.46 (0.08) NA

–58.18 (27.16) NA

–2.14 (0.02) NA

NA

NA

Baseline Model Adj. R2 = 0.82; SEE = 2.58; F = 49.65, prob. F = 0.000, N = 45 Overidentified Model Adj. R2 = 0.82; SEE = 2.58; F = 22.95, prob. F = .000, N = 45 Reduced Model Adj. R2 = 0.83; SEE = 2.48; F = 43.98, prob. F = 0.000, N = 45 Sources: See notes to Appendix Table 7.1.

result, overall, is not especially surprising; the data set is limited inasmuch as it is small (N = 45) and a number of the measures have minimal variation. Given this, and on evaluation of the collinearity diagnostics, a trimmed model is presented in the third column of Table 7.3. Ultimately two of the measures of competition introduced in this chapter enter the model fruitfully: the measure of diversity of opinion and the measure of population diversity. The opinion diversity measure’s sign is reversed so that it now is negative, which may make sense in the presence of a control for mean opinion. It suggests that policy makers are attentive to both level and dispersion of opinion. The model displayed in column 3 makes some intuitive sense.The ideology of citizens and legislators matters, as does the competitiveness of a state’s politics. Further, the diversity of a state’s citizens—ascriptive competition—appears to have a lasting eﬀect on state politics, with more lib-

Ideological Cleavage, Political Competition, and Policy Making 135

Table 7.4 Model of State Policy Liberalism with the Conditional Effect of Ideological Diversity Independent Variables

Intercept Population diversity

Legislative liberalism Ideology Ideological diversity Ideology u ideological diversity Electoral competition

b (sb)

t

Prob t (one-tail)

–49.56 (48.11) 20.47 (14.29)

–1.03

0.15

1.43

0.08

1.79 (.62) –327.09 (189.07) 59.89 (66.63)

2.89

0.003

–1.72

0.05

0.90

0.19

1.93

0.03

2.80

0.004

525.57 (272.54) 0.12 (0.04)

Adj. R2 = 0.84; SEE = 0.40; F = 39.76, prob. F = 0.000, N = 45 Sources: See notes to Appendix Table 7.1.

eral policies being enacted in response to a presumed diversity of demand. However, the measures of racial diversity, diversity in partisanship, and interest group inﬂuence do not enter fruitfully into a fully speciﬁed model. This suggests a ﬁnal model speciﬁcation in which the relationship between ideology and policy outputs is expressed as being conditioned by the extent of ideological diversity within a state. This speciﬁcation is shown in Table 7.4.The model explains about 84 percent of the variation in the dependent variable, is statistically signiﬁcant, and provides a theoretically satisfying explanation for the eﬀect of ideology on policy making in the states.When the net eﬀect of ideology and ideological liberalism is calculated (525.57 + 59.89 – 327.09 = 258.37), it yields a positive but moderated eﬀect of citizen ideological liberalism on SPL. (Although these coeﬃcients appear to be enormous, it is important to remember that the values on the independent variables are very small, and that a “one-unit” change would be heroic.) This comports well with political reality, suggesting that mean liberalism’s eﬀects on the liberalism of policies are tempered by the extent of diﬀerence of opinion within a state. The measure

136

Barrilleaux

of population diversity persists as a signiﬁcant inﬂuence, albeit at a fairly generous threshold of statistical signiﬁcance. Again, this indicates a lasting eﬀect of ascriptively based competition on policy choices.

Conclusion These results provide additional evidence of the basic health of representative democracy in the states.The intent in this chapter was to test the eﬀects of varying forms of political competition on the ideological timbre of states’ public policies. Using the Erikson, Wright, and McIver (1989) model of state policy liberalism as a point of departure, these results suggest that the three forms of competition identiﬁed by Rae and Taylor (1970) matter. The population diversity measure taps ascriptive competition and survives fairly rigorous tests to remain in a model of state policy liberalism. Similarly, a measure of electoral competition, which taps into Rae and Taylor’s notion of behavioral competition, also remains in a model. Finally, attitudinal competition, in the form of a measure of the diversity of opinion within the states, has a meaningful eﬀect as a moderator of the inﬂuence of policy liberalism in the states. It suggests that legislators, even where faced with large majorities who embrace a liberal ideology, moderate their behavior to account for diﬀerences of opinion within polities. This provides a slightly more nuanced understanding of how ideology aﬀects decisions than has been available in the past. Erikson, Wright, and McIver (1993) present a model in which the eﬀects of ideology are not only pronounced, but in which they are not tempered by the competitiveness of elections or other moderating factors. Their model suggests that agreement, rather than competition, is largely what forms the basis of American states’ policy decisions. This research underlines the singular importance of ideology, but also shows that diﬀerences may moderate ideology’s eﬀects. In sum, it suggests that representatives pay attention to diﬀerences of opinion as well as to majority demands, and that speaks well for the state of state democracy. The explanatory improvement of this model over that of the Erikson, Wright, and McIver model is modest— there is about a 3 percent increase in explained variation.Thus, the earlier model continues to explain a large portion of the variation in state policy liberalism and these results amend that model. The discussion about population diversity that is central to this research forced consideration of the eﬀects of racial diversity on state policies. As noted above, Hero and Tolbert (1996) and Hero (1998) argue that race is the single most important inﬂuence on state policies and call for increased

Ideological Cleavage, Political Competition, and Policy Making 137

attention to race in the study of American state politics.The results of this research suggest that in this case race is not as important as those authors suggest: the measure of racial diversity fails to inﬂuence empirical models signiﬁcantly. The research conducted here does not replicate that conducted by Hero and Tolbert, nor is it intended to do so. It does call into question their argument that representation is conditioned on there being a shared set of cultural norms in a polity. Their argument is provocative and, although it does not bear up well in this research, deserves further consideration. Finally, the results reported here suggest that additional research addressing the eﬀects of diﬀering forms of diversity are in order. In particular, a more detailed causal map of how diversity interacts with levels of diﬀerent characteristics of states, for example the relationship between mean ideological liberalism and ideological diversity, is in order. Research completed in the wake of Wright, Erikson, and McIver’s (1985) dissemination of a measure of state-level ideology leaves little question that ideology has a meaningful eﬀect of “who gets what” from state governments. Appendix Table 7.1 Descriptive Statistics Variables

N

Minimum

Maximum

Mean

Standard Deviation

Democratic legislative strengthi Ideologya State policy liberalisma Ideological diversitya Racial diversityb Population diversityc Partisan diversitya Interest groups Ideology × ideological diversitya Electoral competitiond Interest group influencee Valid N

46 47 47 48 50 50 48 50 47 49 50 45

–0.69 –0.33 –9.25 0.63 0.39 0.38 0.55 1.00 –0.22 9.26 1

–0.01 –0.01 12.69 0.73 0.54 0.52 0.70 4.00 0.00 56.58 4

–0.3651 –0.1490 –0.0572 0.6946 0.4570 0.4331 0.6660 2.3600 –0.1030 39.0390 NA

0.19441 0.07901 5.95472 0.01873 0.03547 0.03618 0.03462 0.87505 0.05320 11.40219 NA

Data Sources: a Data provided by Gerald Wright; originally published in Wright, Erikson, McIver, 1985;Wright, 1987; Erikson et al. 1989, 1993. b Author’s calculations from data provided by U.S. Bureau of the Census, Decennial Census of the United States, 1980. c Author’s calculations from data provided by U.S. Bureau of the Census, Decennial Census of the United States, 1980. d Holbrook and Van Dunk, 1993. e Thomas and Hrebenar, 1990.

138

Barrilleaux

This research provides some evidence that the shape of the ideological distribution, along with the level of ideological liberalism, has some inﬂuence on policy adoptions.

Notes 1. There are numerous methods for measuring opinion using various data sets. Nonetheless, survey data-based measures (e.g., Erikson,Wright, and McIver, 1993; Norrander, 2001; Brace et al., 2002), measures that are based on extrapolations from demographic data (Plutzer and Berkman, 2002;Weber, Hopkins, Mezey, and Munger, 1972–73), and measures that are based on election outcomes (Berry, Ringquist, Fording, and Hanson, 1998; Holbrook-Provow and Poe, 1987) depict ideology as an average tendency with small consideration of the extent to which scores are scattered around that mean. 2. Erikson,Wright, and McIver (1989, 1993) add a novel result to this skein of research: they show a strong link between liberal party strength and policy, but it is a negative eﬀect, just the opposite of what would be expected theoretically. 3. Erikson, Wright, and McIver (1993) provide ideology and other data that “center” on the late 1980s but not an updated dependent variable, which they describe (Erikson,Wright, and McIver,1993, 75;Wright, Erikson, and McIver, 1987) as representing state policy in about 1980. I use the earlier ideology data because they are contemporaneous with the measurement of the dependent variable while the mid-1980s data actually follow the dependent variable temporally, which, strictly speaking, violates the logic of cause and eﬀect.This has no real eﬀect on the substantive conclusions or hypothesis tests presented here, however, as the EWM data from the two time periods provide near-indistinguishable results. 4. The measures of racial diversity diﬀer from those presented in Hero and Tolbert (1996).They include measures of Hispanic in their racial diversity indicator. Here, Hispanic populations are treated as ethnicity (and thus included in the population diversity measure) rather than racial diversity because persons with Hispanic backgrounds vary on racial characteristics. Using the Hero and Tolbert measures in the regressions reported in this chapter does not alter my conclusions regarding their eﬀects in models of state policy liberalism. 5. The Erikson,Wright, and McIver measure is “a grand index of state policy” (1993, 75) constructed using information about state policies in eight areas: education, Medicaid,Aid to Families with Dependent Children, consumer protection, criminal justice, legalized gambling, support for the Equal Rights Amendment, and tax progressivity (Erikson,Wright, and McIver 1993, 75–77).

References Barrilleaux, Charles. 1997. “A Test of the Independent Inﬂuences of Electoral Competition and Party Strength in a Model of State Policy-Making.” American Journal of Political Science 41 (October): 1462–1466.

Ideological Cleavage, Political Competition, and Policy Making 139 ———. 2000. “Party Strength, Party Change, and Policy-making in the American States.” Party Politics 6 ( January): 61–73. Barrilleaux, Charles, T. Holbrook, and L. Langer. 2002. “Electoral Competition, Legislative Balance, and American State Welfare Policy.” American Journal of Political Science 46 (April): 415–427. Berkman, Michael, and Eric Plutzer. 2002.“Crucibles of Democracy: Institutions, Race and Policy Responsiveness in America’s School Districts.” Presented at the annual meetings of the American Political Science Association, Boston, September. Berry,William D., E. Ringquist, R. Fording, and R. Hanson. 1998.“Measuring Citizen and Government Ideology in the American States, 1960–93.” American Journal of Political Science 42 ( January): 327–348. Bibby, John, and Thomas Holbrook. 2003.“Parties and Elections.” In Virginia Gray and Russell Hanson, eds., Politics in the American States: A Comparative Analysis, 8th ed.,Washington, DC: Congressional Quarterly, pp. 62–99. Brace, Paul, K. Sims-Butler, K. Arceneaux, and M. Johnson. 2002. “Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46 ( January): 173–189. Budge, Ian, and Richard I. Hoﬀerbert. 1990. “Mandates and Policy Outputs: U.S. Party Platforms and Federal Expenditures.” American Political Science Review 84 (March): 111–132. Chubb, John E. 1988.“Institutions, the Economy, and the Dynamics of State Elections.” American Political Science Review 82 (March): 133–154. Dahl, Robert A. 1956. A Preface to Democratic Theory. Chicago: University of Chicago Press. Dahl, Robert A., and Edward R. Tufte. 1973. Size and Democracy. Stanford, CA: Stanford University Press. Dye, Thomas R. 1966. Politics, Economics and the Public. Chicago: Rand-McNally. ———. 1984. “Party and Policy in the States.” Journal of Politics 46 (November): 1097–1116. Elazar, Daniel J. 1984. American Federalism: A View From the States, 3rd ed. New York: Harper & Row. Erikson, Robert S., Gerald C.Wright, Jr., and John P. McIver. 1989.“Political Parties, Public Opinion, and State Policy in the United States.” American Political Science Review 83 (September) :729–750. ———. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Fiorina, Morris P. 1994.“Divided Government in the American States:A Byproduct of Legislative Professionalism.” American Political Science Review 88 ( June): 304–316. Garand, James C. 1985. “Partisan Change and Shifting Expenditure Priorities in the American States, 1945–1978.” American Politics Quarterly 13 (October): 355–391. Hero, Rodney. 1998. Faces of Inequality: Social Diversity in American Politics. New York: Oxford University Press.

140

Barrilleaux

Hero, Rodney E., and Caroline J.Tolbert. 1996.“A Racial/Ethnic Diversity Interpretation of Politics and Policy in the States of the U.S.” American Journal of Political Science 40 (August): 851–871. Holbrook,Thomas M., and Emily Van Dunk. 1993.“Electoral Competition in the American States.” American Political Science Review 87 (December): 955–962. Holbrook-Provow,Thomas M., and Steven C. Poe. 1987. “Measuring State Political Ideology.” American Politics Quarterly 15 ( July): 399–416. Jennings, Edward T. 1979. “Competition, Constituencies, and Welfare Policies in the American States.” American Political Science Review 73 ( June): 414–429. Key,V. O., Jr. 1949. Southern Politics in State and Nation. New York: Knopf. King, James D. 1989. “Interparty Competition in the American States: An Examination of Index Components.” Western Political Quarterly 42 (March): 83–92. Langer, Laura. 2002. Judicial Review in State Supreme Courts: A Comparative Study. Albany: State University of New York Press. Lieberson, Stanley. 1969. “Measuring Population Diversity.” American Sociological Review 34 (December): 850–862. Liebling, A. J. 1961. The Earl of Louisiana. New York: Simon and Schuster. Morehouse, Sarah McCally. 1981. State Politics, Parties and Policy. New York: Holt, Rinehart and Winston. Norrander, Barbara. 2001.“Measuring State Public Opinion with the Senate National Election Study.” State Politics & Policy Quarterly 1 (March): 113–127. Norrander, Barbara and Sylvia M. Rivera. 2002.“Diversity in Partisanship and Ideology in the States.” Presented at the annual meetings of the Midwest Political Science Association, April. Olson, Mancur. 1982. The Rise and Decline of Nations. New Haven:Yale University Press. Patterson, Samuel, and Gregory Caldeira. 1984. “The Etiology of Partisan Competition.” American Political Science Review 78 (September): 691–707. Plotnick, Robert, and Richard Winters. 1985.“A Politico-Economic Theory of Income Redistribution.” American Political Science Review 79 ( June): 458–473. Rae, Douglas, and Michael Taylor. 1970. The Analysis of Political Cleavages. New Haven:Yale University Press. Ranney, Austin. 1965.“Parties in State Politics.” In Herbert Jacob and Kenneth N. Vines, eds., Politics in the American States. Boston: Little Brown, pp. 61–99. Ringquist, Evan, Kim Q. Hill, Jan E. Leighley, and Angela Hinton-Anderson. 1997. “Lower-Class Mobilization and Policy Linkage in the U.S. States: A Correction.” American Journal of Political Science 41 ( January): 339–344. Schattschneider, E. E. 1942. Party Government. New York: Rinehart. Stonecash, Jeﬀrey M. 1987. “Inter-Party Competition, Political Dialogue, and Public Policy.” Policy Studies Journal 16: 243–262. Sullivan, John L. 1973. “Political Correlates of Social, Economic, and Religious Diversity in the American States.” Journal of Politics 35 (February): 70–84. Thomas, Clive, and Ronald Hrebenar. 1990. “Interest Groups in the States.” In R. Albritton, V. Gray, and H. Jacob, eds., Politics in the American States, 5th ed. Boston: Little, Brown. Thomas, Clive and Ronald Hrebenar. 1999.“Interest Groups in the States.” In Vir-

Ideological Cleavage, Political Competition, and Policy Making 141 ginia Gray, Russell Hanson, and Herbert Jacob, eds., Politics in the American States: A Comparative Analysis, 7th ed. Washington, DC: Congressional Quarterly, pp. 113–143. Weber, Ronald E.,Anne H. Hopkins, Michael L. Mezey, and Frank Munger. 1972– 1973. “Computer Simulation of State Electorates.” Public Opinion Quarterly 36 (Winter): 49–65. Winters, Richard F. 1976. “Party Control and Policy Change.” American Journal of Political Science 20 (November): 596–636. Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship and Ideology with Survey Data.” Journal of Politics 47 ( June): 469–489. ———. 1987.“Public Opinion and Policy Liberalism in the American States.” American Journal of Political Science 31 (November): 980–1001. Wright, Gerald C., Jr., and Brian F. Schaﬀner. 2002. “The Inﬂuence of Party: Evidence from State Legislatures.” American Political Science Review 96 ( June): 367– 380.

Chapter

The Civil State:Trust, Polarization, and the Quality of State Government

8

Eric M. Uslaner

Does good government depend on good citizens? Robert Putnam (1993) made this connection in Making Democracy Work and others (including myself) have echoed this claim.Yet, it has proved diﬃcult to measure “good government” (much less to agree what it is) and the causal chain from a positive citizenry to governmental performance remains murky. Democracy depends on a participatory citizenry, to be sure. Representation depends on an alert citizenry. But what do citizens need to do to secure eﬀective government? Putnam’s link between good citizens and good government is encapsulated in his concept of social capital, which encompasses social networks, formal organizations, and norms of trust. Putnam (1993, 115) argues that in Northern Italy: choral societies and soccer teams and bird-watching clubs and Rotary clubs. Most citizens in these regions read eagerly about community aﬀairs in the daily press. They are engaged by public issues, but not by personalistic or patron-client politics. Inhabitants trust one another to act fairly and to obey the law. Leaders in those regions are relatively honest.

But in Southern Italy: Engagement in social and cultural associations is meager. . . . “Compromise” has only negative overtones. Laws (almost everyone agrees) are made to be broken, but fearing others’ lawlessness, people demand sterner discipline, nearly everyone feels powerless, exploited, and unhappy. . . . it is hardly surprising that representative government here is less than in more civic communities.

142

Trust, Polarization, and the Quality of State Government 143

Putnam’s argument is that an engaged citizenry becomes more trusting and cooperative and more likely to demand that their representatives place great emphasis on the public good rather than clientelism or patronage. In one respect, this is not a new argument: Lane (1959, 164–165) and Almond and Verba (1963, 3) suggested that “civic cooperation” was important for democratic government. Inglehart (1999) found a link between generalized trust and democratic government (as well as support for democracy). Knack and Keefer (1997), LaPorta, Lopez-Silanes, Scheifer, and Vishney (1997, 1999), and Uslaner (2002, ch. 8; 2004) ﬁnd that countries with high levels of trust have lower levels of corruption, better functioning bureaucracies, more eﬀective legal systems, lower rates of theft, and “better government” more generally. King, Zeckhauser, and Kim (2001) and Knack (2002) ﬁnd that American states with higher levels of “social capital” have more eﬀective governments. Eﬀective government in the states is measured by a study by the Government Performance Project (GPP) of Governing magazine and the Maxwell School of Citizenship at Syracuse University.The measure includes program information, a survey, and “interviews [with] budget oﬃcers, auditors, public managers, auditors, academics, and legislative aides in every state” (Knack, 2002, 775). Each state was “graded” on its performance in ﬁnancial management, capital management, human resources, “managing for results,” and information technology. I consider each of these measures in this chapter, as well as two other measures: the number of Ford Foundation/Kennedy School of Government (Harvard University) awards for innovation a state has won; and the level of corruption in a state, as perceived by political reporters in state capitals (Boylan and Long, 2001).The GPP measures, the Ford Foundation/Kennedy school awards, and the level of corruption are all measures of the quality of a state government. King, Zeckhauser, and Kim (2001) and Knack (2002) have shown that states with higher levels of social capital have better functioning governments, using most of these same measures of state eﬀectiveness. So what is new here? Two rationales for another look stand out. First, there is much conceptual confusion about social capital and why (or whether) it might aﬀect government performance. I shall argue that most of social capital is irrelevant to good government and the important aspect of social capital, generalized trust, has been poorly measured in other studies. Beyond social capital, I oﬀer a vision of the civil state as a polity where people trust others who are unlike themselves, where there is minimal political divisiveness, where leaders work with each other toward ﬁnding some common ground, and where the political environment militates against confrontation. This civil state has better, more eﬀective govern-

144

Uslaner

ment because there is less back-biting and at least some of its foundations lead to greater honesty in government and more participation toward the common good by ordinary citizens. Better theory and better measurement lead to ﬁrmer support for the link between trust and performance. Second, trust is important because it leads to a more cooperative, less polarized society and polity:The decline in congressional productivity over the past three to four decades has a direct linkage to falling levels of generalized trust; as we have less faith in each other, there are more ﬁlibusters and gridlock in the Congress and the ideological gap between the congressional parties has grown (Uslaner, 1993, ch. 6; 2001; 2002, 214).A less polarized polity is more productive (see also Binder, 1999). Knack (2002, 778) brieﬂy makes an argument about polarization, but his measures (divided government and racial heterogeneity) do not tap people’s preferences. I present models of good government in the civil state for the American states. The models emphasize political polarization, relying on measures of partisan polarization from national surveys and from state legislative voting. More trusting and less polarized states should have more eﬀective government.The evidence for trust is generally strong, but polarization eﬀects are more sporadic (and, in one case, even positive). I also consider other political variables and ﬁnd that high levels of party competition leading to divided government generally lead to less eﬀective government, as do strong party organizations. Traditionally, we think of the most “innovative” (and perhaps most “eﬀective”) state governments to be dominated by liberals, but instead I ﬁnd that eﬀectiveness in management is greatest in states where the dominant party is most conservative. “Social capital” matters mightily for good government, but not as Putnam’s argument would lead us to expect. It is trust, not civic engagement, that leads to good government. Trust is signiﬁcant for most of the measures of government quality, but it has more powerful eﬀects for corruption (see Uslaner, 2004). Corruption thrives where political competition is lacking—in one-party states and where there are strong party organizations—and where some groups (mainly whites) are better oﬀ than others. A vigilant public may guard against corruption, but as with government eﬀectiveness more generally, civic engagement is not enough. No measure of civic engagement leads to better government performance (see below). A particularly demanding form of participation—the share of the public making political speeches (ranging from 0.05 percent in New Jersey to 9 percent in Oregon and South Dakota)—does combat corruption—but this is hardly the sort of political participation that we might expect of ordinary citizens.

Trust, Polarization, and the Quality of State Government 145

Good Government and Good Citizens Good government means diﬀerent things to diﬀerent people and that is one reason it has proven so diﬃcult to measure. We normally think of eﬀective government as including strong service delivery, impartial bureaucrats and judges, and the honesty of government oﬃcials (Kaufmann, Kray, and Mastruzzi, 2003).These factors are terribly diﬃcult to measure. There are multiple measures of government corruption, including the indictment and conviction rates of political leaders and newspaper reporters’ subjective estimates of corruption (Boylan and Long, 2001), which is closest to international estimates of corruption by Transparency International. However, measures of overall quality have been elusive. The GPP measures are an attempt to provide some data on governmental eﬀectiveness. They are not measures of legislative productivity or policy gridlock in the states (which might be ideal).They are indicators of state capacity and bureaucratic functioning.The ﬁve measures and some of their components are (King, Zeckhauser, and Kim, 2001, 31–32; see also Knack, 2002, 775–776): • Financial management: accurate revenue estimates, state budget adopted on time, has long-term budget planning; accurate ﬁnancial reporting; management of goods and services; audited ﬁnancial statements. • Human resource management: clear personnel policies, can hire new employees quickly; merit pay; can state discipline and terminate unproductive employees? • Information technology management: how widely available is technology and how quickly can it be ordered; IT training; use of websites for citizen information. • Capital management: evaluation of planning process for construction; integration of capital and operating budget; planning for capital maintenance. • Managing for results: does state have a strategic plan and are citizens, businesses, unions, and other groups involved in planning? Eﬀectiveness of performance measures for policy makers. These are all measures of bureaucratic performance. Only “Managing for Results,” which includes citizen participation, includes any direct link between citizens and their government.The GPP project gave each state a letter grade—A, B, C, D, with pluses and minuses—on each category and King, Zeckhauser, and Kim (2001) translated these grades into numeric scores.The

146

Uslaner

ﬁve indicators were all positively related to each other—with correlations ranging from 0.71 for ﬁnancial and capital management to a low of 0.37 between capital and human resources management. Overall, the states with the highest grades were Missouri, Utah, Washington, and Virginia and those ranking at the bottom were Alabama, Hawaii, Connecticut, New York, and Rhode Island. The Ford Foundation/Kennedy School innovation awards (administered by the Kennedy School and the Council for Excellence in Government in Washington, DC) have been given annually since 1986 for public sector initiatives in areas including “customer focus” (meeting the demands of diverse constituencies); public-private competition; “performance benchmarking” (developing clear standards for evaluating governmental eﬀectiveness); citizen participation; consumer choice; and privatization (Altshuler, 1997; Sparrow, 2000, 81). These awards measure not only performance, but also doing things diﬀerently than before—suggesting that the roots of innovation may be diﬀerent from simple measures of performance such as the GPP indicators. These measures may not be ideal, but they are the best available on the quality of state government—with one exception, a measure of corruption by statehouse political reporters in 1999 reported by Boylan and Long (2001). Corruption is notoriously diﬃcult to measure: If done well, it should be unobservable (unless, of course, the public doesn’t care about it). The most widely used measures in cross-national research are reputational measures from Transparency International.1 Reputational measures are always controversial, but the Transparency International measures are widely accepted for their face validity.The Boylan-Long measures for 47 states have Rhode Island and Louisiana as the most corrupt and the Dakotas and Colorado as the most honest. An alternative measure of corruption is the share of public oﬃcials indicted or convicted (Meier and Holbrook, 1992).2 The most corrupt states in 1995 were Florida and Virginia and the least corrupt were New Hampshire and Vermont. The two measures are not identical—both Florida and Virginia rank 14th and 26th respectively on the Boylan-Long measure and the overall correlation between the two measures is just 0.259. The reporters’ measure seems to be the better one, since it has greater face validity. Prosecution indicators may reﬂect the personal priorities of prosecutors (Boylan and Long, 2001, 3–4)—and it may simply be more diﬃcult to gain an indictment and conviction in a heavily corrupt state.Thus, I rely on the reporters’ perception measure, which seems less troubled by endogeneity issues (such as whom to prosecute and whom to convict). I now turn to an examination of the role of trust in civic life, to other

Trust, Polarization, and the Quality of State Government 147

predictors of government performance, the results for the ﬁve GPP indicators and the Ford Foundation/Kennedy school awards, and then a theoretical and empirical discussion of corruption.

The Civil State and Good Government The civil state is a moderate state. Citizens in civil states trust people who are diﬀerent from each other. They seek to avoid confrontation and to seek common ground in decision making. People in polarized societies don’t trust folks who disagree with them. Diﬀerent views become a sign of heresy rather than simple disagreement. Fundamentalists, be they religious or political, are likely to trust only their own kind—and to favor confrontation over compromise. Good government requires a commitment to seeking common ground—and it will have diﬃculty ﬂourishing in a confrontational political environment. So where the public or the elites (legislative leaders) are highly polarized, it will be more diﬃcult to obtain eﬀective government. In states where we ﬁnd strong political party organizations, we may also ﬁnd fewer incentives for the two parties to cooperate with each other. Especially when diﬀerent parties control the legislative and executive branches—and the division between the parties is close (as we have seen in recent years in Congress), the opportunities for gridlock multiply and neither party has any incentive to cooperate with each other. The basis of good government, then, lies with more trusting, less polarized states, where party organizations are weak and the political opportunities for grandstanding and blocking legislation are minimal. Such foundations are partly societal (trust and polarization) and partly structural (party organizations and the partisan balance of power in a state).Together they amount to a simple lesson: It is diﬃcult to have good government when there is polarization and where the political environment provides few incentives for cooperation. Perhaps the most important part of the civil state is the level of trust. In such a state, high levels of people say that most people can be trusted (as opposed to agreeing that “you can’t be too careful in dealing with people”). Elsewhere I show that this question, which has been asked in a large number of national surveys since 1960, reﬂects a variety of trust that is not based on experience. Instead, it is based on a worldview of optimism and control:The world is a good place, it is going to get better, and I can help make it better. Whites who trust others are more tolerant of minorities, while in general people who are more willing to give of themselves through charity and volunteering time see interactions with people of diﬀerent backgrounds as opportunities rather than risks. Generalized trust

148

Uslaner

is faith in people who are diﬀerent from yourself (in contrast to faith only in your kind, particularized trust, see Uslaner, 2002, chs. 2–3). Generalized trusters look for common ground with people of diﬀerent backgrounds and views—so trust is the font of cooperation and compromise and is the opposite of polarization. Divided societies—through ideological polarization and especially from economic inequality (Uslaner, 2002, chs. 6, 8)— rank low on trust. Trusting societies have good government because their citizens put aside diﬀerences to work for common purposes—not because they are active in politics or even in civic aﬀairs. When people join civic groups, and especially in their informal social life, they connect with people with similar interests, backgrounds, and worldviews. Political participation usually polarizes people rather than building trust with opponents (Rosenblum, 1998, 48; Uslaner, 2002, chs. 2, 4). Political action may be important to make government more responsive to citizens, but there is no clear reason to believe that it leads to more eﬀective government. And there is even less reason to believe that participating in choral societies or bowling leagues translates into better government. Membership in civic groups and trust are not the same thing—and there is scant evidence that one leads to the other. It is trust in other people that matters for good government, not membership in civic groups or other participation (with one exception)—and not trust in government. So it is essential that we separate out the component parts of “social capital.” King, Zeckhauser, and Kim (2001) use Putnam’s (2000) state-level measure of social capital, which is a mixture of estimates of “trust,” civic engagement, turnout, and informal social connections such as visiting friends and entertaining people at home. This is a veritable dog’s breakfast (a bit of this, a bit of that), making it diﬃcult to isolate what aspect of social capital might lead to more eﬀective government. As I argue elsewhere (Uslaner, 2002, chs. 2, 4), there is little reason to expect—and no evidence to support—the argument that most forms of civic engagement and especially informal social connections might lead to more trust or better government. Knack (2002, 780) does separate out the component parts of social capital and ﬁnds little evidence that volunteering, informal socializing, or attending club meetings shapes any of the GPP measures of government eﬀectiveness and that trust has a weak eﬀect only on human resources management.Trust is a signiﬁcant predictor of a summary measure of government eﬀectiveness derived from adding the ﬁve components, but not of most individual measures. So what are we to make of these conﬂicting ﬁndings? We need not worry quite so much. Neither Knack nor King,

Trust, Polarization, and the Quality of State Government 149

Zeckhauser, and Kim actually test the impact of generalized trust. Their indicator of trust comes from the DDB Needham Life Style surveys, which have large samples that can be aggregated to the state level from 1975 to 1998. However, these surveys are not random samples nor do they ask directly about trust. Instead, the DDB Needham surveys asked whether people agreed that “most people are honest”—and honesty (using the same question) is only modestly correlated (tau – c = 0.345) with trust in the 1972 American National Election Study. I cannot say much about trust because I have not actually tested it. I have calculated state-level measures of generalized trust using a variety of surveys conducted over time and here use the estimates for the decade of the 1990s.3 While there may well be concern about deriving state-level estimates from national surveys such as the General Social Survey and the American National Election Studies—which are not designed to be representative of state populations—Brace, Sims-Butler, Arcenaux, and Johnson (2002) have shown that such estimates are very reliable. In Table 8.1 below, I present the estimated shares of trusters in each state in the 1990s used in this analysis. Beyond, or instead of, trust, good government is driven by state size, percent African American, and especially economic inequality (Knack, 2002, 779) and by legislative professionalism, the number of good government groups, the business environment, and neighboring states’ rankings (King, Zeckhauser, and Kim, 2001, 30). I considered each of these variables, but all fell to insigniﬁcance in models focusing on trust, polarization, legislative ideology, party organizational strength, and political competition.4 An eﬀective government is not one that is bickering all the time. And a confrontational political environment should not be conducive to quality government.The measures of quality are all based on objective assessments of performance and a conﬂictual environment makes objectivity diﬃcult to get. So I suggest that states with polarized publics or polarized legislatures should be less likely to have quality governments. My measure of public opinion polarization is the absolute diﬀerence between the share of Democrats in a state who are liberal and the proportion of Republicans who are conservative. These measures come from exit polls in the 1990s (see Erikson, Wright, and McIver, 1993) and were computed for me by Gerald C. Wright. Where most Democrats are liberals and most Republicans are conservative, there is less room for moderation or for compromise.The polarization measure is thus a variant on the well-known measures of public opinion at the state level in Erikson et al.:Wright estimated the share of liberals, moderates, and conservatives for Democratic, Republican, and Independent identiﬁers in each state. My measure of

150

Uslaner Table 8.1 Trust Estimates by State for the 1990s Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Illinois Indiana Iowa Kansas Kentucky Louisiana Maryland Massachusetts Michigan Minnesota Mississippi Missouri Montana

0.162 0.304 0.413 0.105 0.384 0.413 0.440 0.250 0.351 0.447 0.400 0.398 0.412 0.488 0.296 0.293 0.447 0.448 0.448 0.513 0.188 0.380 0.500

New Hampshire New Jersey New York North Carolina North Dakota Ohio Oklahoma Oregon Pennsylvania Rhode Island South Carolina South Dakota Tennessee Texas Utah Vermont Virginia Washington West Virginia Wisconsin Wyoming

0.630 0.305 0.362 0.223 0.594 0.333 0.290 0.453 0.436 0.316 0.259 0.516 0.266 0.279 0.560 0.558 0.382 0.413 0.263 0.521 0.417

Source: Uslaner and Brown (2005).

polarization is the absolute value of the share of Democratic identiﬁers calling themselves liberal minus the share of Republican identiﬁers calling themselves conservatives. My measure of elite polarization is more complex: It is based on NOMINATE scores derived from roll call data from the American states for 1999–2000 also provided by Wright (see note 3). The measure I employ is the average for both legislative chambers of the absolute diﬀerence between the squared (quadratic) mean NOMINATE scores for Democrats and Republicans. I average these measures for each party for the two chambers (one in Nebraska). I then square the NOMINATE diﬀerences and then take the absolute diﬀerence between the parties. Why such a complex measure? Most spatial models of party competition employ squared distances: Quadratic measures make larger diﬀerences more “extreme” and smaller gaps less “extreme.” I believe that this formulation better captures the polarization in legislative politics than a simple diﬀerence of party positions. This is my measure of legislative polariza-

Trust, Polarization, and the Quality of State Government 151

tion—and for both the mass and elite indicators, I expect negative relationships with government quality. I also employ the ideology of the majority party in the legislature from the same NOMINATE scores, with higher scores indicating greater liberalism. There are compelling arguments for hypotheses in either direction: Knack (2002, 774–775) argues that much of the existing literature on quality equates performance with innovation or liberalism. So we might expect that legislatures with more liberal majority parties would have higher levels of government quality. Alternatively, these speciﬁc measures of performance focus on management and in some cases a businessoriented approach to managing government (especially for human resource management, with its emphasis on merit pay and dismissing employees).This business orientation would support a negative relationship between legislative ideology (liberalism) and performance. Beyond these measures, I also expect that states with strong party organizations would be less likely to rank highly on these government performance measures. Strong party states, which have what Mayhew (1991) called “traditional party organizations,” place a heavy emphasis on patronage and on party control of the bureaucracy. “Scientiﬁc” management is an anathema to strong parties. I use Mayhew’s classiﬁcation of party systems from weak to very powerful party organizations. Divided control of the legislative and executive branches of government may lead to less productive government. When one party controls the legislative branch and another the executive branch in the U.S. Congress, budget deﬁcits spiral out of control (McCubbins, 1991). Overall, there seems to be little relationship between the enactment of major legislation and divided government (Binder, 1999; Mayhew, 1991), but there is a strong relationship between divided government and the ability of the executive to get what he or she wants, especially in periods when the party system is highly polarized (Conley, 2003). Divided government may lead to more diﬃculties when the two major parties are relatively equal in strength. This makes it easier for the minority party to obstruct the agenda of the majority party. In contrast, single-party government with an overwhelming majority for one party should have the least diﬃculty in enacting legislation. Preliminary estimations suggest that divided government per se does not shape the quality of government— however, divided government interacted with the level of party competition in a state should be more likely to lead to less eﬀective governments. Divided government might lead to greater compromise if each party is negotiating out of strength—as reﬂected in the dominance of one of the institutions. But where margins are tight, neither party wants to give any

152

Uslaner

ground to the other. Each party might prefer standing on principle than letting the other party claim credit for government accomplishments. Parties will use these accomplishments in the next campaign and even a small gain in a tightly balanced environment might lead to a change in the balance in power. So divided government together with a competitive political environment should lead to less eﬀective government. The ﬁve GPP measures are all indicators of a business-oriented approach to governing. This leads me to expect that polarization eﬀects might not be as great as in the Ford Foundation/Kennedy School Awards— which reward “innovation” rather than simply bureaucratic performance. Because innovation may reﬂect liberal rather than conservative ideologies (Knack, 2002, 774–775), the sign on legislative ideology may be positive rather than negative.

Good Government:The Models Because these six models are not independent, I estimated the regressions using Zellner’s (1962) seemingly unrelated equation (SUR) technique.5 I report the SUR estimates in Table 8.2. There is considerable, though not unconditional, support for the claim that civil states have better functioning governments in these regressions. Generalized trust is important for four of the six measures of state government performance. A low-trusting state that would otherwise receive a C on ﬁnancial management would receive a B if it became the highest trusting state; for human resource management, the grade would increase to B+, for information technology to B+/A–, and for capital management to A–.Trust does not matter for managing for results or for receiving awards from the Ford Foundation and the Kennedy School of Government. Mass polarization does not seem to have deleterious eﬀects on government performance. For human resources management, a more polarized environment seems to encourage higher rankings.The most polarized state would receive a ranking two grades higher than the state with the smallest partisan ideological division.The components of the human resources management index—including merit pay and terminating unproductive employees—may make this dimension of management an ideological magnet. States where the majority party tilts to the right are also likely to rate highly on human resources management (see below). However, polarization does matter mightily for Ford Foundation/Kennedy School Awards. The most polarized state will receive ﬁve fewer awards than the state with the smallest partisan ideological division among the

4.298** (1.67) 3.573 (2.912) –3.422*) (2.291 –1.157** (0.493) –0.452** (0.208) –1.019 (2.212) 8.858**** (1.501) 0.284 1.674

5.486*** (2.46) 5.908+ (2.527) –5.851*** (1.987) –0.819** (0.427) –0.146 (0.180) –4.974*** (1.918) 6.712**** (1.302) 0.436 1.451

Human Resources Management

6.543*** (2.82) 1.907 (2.629) –7.500**** (–3.63) –1.455**** (0.444) –0.500*** (0.187) –5.204*** (1.995) 8.220**** (1.354) 0.556 1.509

Information Technology Management

Entries are regression coefficients, standard errors under coefficients in parentheses. + Significant in opposite direction from hypothesized.All tests one-tailed except for constants. N = 44 **** p < 0.0001; *** p < 0.01; ** p < 0.05; * p < 0.10 Source: Uslaner and Brown (2005).

R2 Root mean square error

Constant

Divided government* competition

Traditional party organization

Legislative ideology

Legislative polarization

Mass polarization

Generalized trust

Independent Variable

Financial Management

6.986** (2.28) 1.541 (.44) –3.584* (2.734) –0.963* (0.587) –0.309 (0.248) –3.294 (2.639) 7.623*** (1.791) 0.258 1.996

Capital Management

Table 8.2 Seemingly Unrelated Regressions for Government Quality Measures

.879 (3.036) –3.988 (3.440) –3.624* (2.704) –1.532*** (0.581) –0.480** (0.245) –4.336** (2.610) 10.785**** (6.09) 0.307 1.975

Managing for Results

–1.068 (3.912) –16.843**** (4.432) –2.741 (3.484) 2.124 (0.748) 0.242 (0.316) –5.274* (3.363) 7.672*** (2.283) 0.378 2.544

Ford Foundation/ Kennedy Awards

154

Uslaner

public.6 Where the party constituencies are further apart, state oﬃcials will perceive greater political pressures and will ﬁnd it more diﬃcult to innovate. Although mass polarization does not seem to matter much, legislative polarization sends a very loud and clear symbol to the state bureaucracy: No factor matters as much as legislative polarization in shaping governmental eﬀectiveness. An otherwise average state with the lowest level of elite division will receive a ranking two grades higher on ﬁnancial management, capital management, and managing for results than a state with the highest level of conﬂict between legislative parties: An otherwise C performance becomes an A grade when elite polarization is at a minimum. For human resources management, which seems to thrive on mass polarization, is more negatively eﬀected by elite conﬂict: The least polarized state would receive an A, while the state with the most discord would get a low-end D. Perhaps most surprising is the powerful eﬀect on information technology, where two states similar in all but elite conﬂict would receive an A and an F, respectively. Is there really so much conﬂict in computers? One might not think so, but perhaps parties might worry that the opposition would use technology to promote their own reelection: High-level technology will not only allow state oﬃcials to communicate with each other, but also with constituents. E-government may mean much more than renewing a ﬁshing license online. It also allows legislators to reach out to their constituents and to promote themselves.Technology can thus become a key source of legislative conﬂict. Legislative polarization is, not surprisingly, connected to state innovation.The very strong impact of public discord may displace the impact of legislative conﬂict. It is not just polarization that matters a lot, but also the dominant ideology in a state’s legislature. On all ﬁve GPP measures, the more conservative a state is, the better functioning its government—which may not be so surprising, because the management criteria seem to be patterned on a business model. For innovation, however, a more liberal state majority party is important. The biggest impact for legislative ideology comes for information technology—an average grade change from C– to A as we move from the most liberal to the most conservative legislature. Here is further evidence for the polarization of technological innovation. Republican states seem to have made greater technological advances than Democratic states, so it is not surprising that IT management gets higher grades in states with conservative majority parties. Managing for results also is strongly aﬀected by a conservative ideology—once again increasing from a C– to an A over the range of legislative

Trust, Polarization, and the Quality of State Government 155

ideology. Financial management is two grades higher where the majority party is the most conservative, while the human resources management grade is just over one rank higher in such states. Given the power of both mass and elite polarization for human resources management, there may be only a limited role for ideology itself on this dimension. The Ford Foundation/Kennedy School awards are higher when the governing state legislative party is more liberal. The most liberal state will receive an additional 3.5 accolades. There are also strong eﬀects for three of the ﬁve GPP measures, but not for the Ford Foundation/Kennedy School innovation awards, for traditional party organizations. States with very powerful parties will have, on average, management scores two grades lower than those with weak parties for ﬁnancial management, information technology management, and managing for results. When a state has divided government and close political competition, it will score two grades lower than a state with uniﬁed government on human resources management, information technology, and managing for results—and it will receive two fewer Ford Foundation/Kennedy School awards. The civil state does indeed seem to encourage more eﬀective government. Across most of the measures, a trusting citizenry promotes better governance. So does a lower degree of polarization between the legislative parties—though partisan conﬂict among the public does not seem to matter as much—and, for human resource management, it even seems to promote higher rankings. Beyond greater trust and lower levels of elite polarization, strong parties and divided government with closely divided parties also lead to lower grades on eﬀectiveness.The most eﬀective government, then, seems to rest on a civil state that has high levels of generalized trust, lower amounts of partisan conﬂict among elites, and weaker party organizations. When one party has control of the legislative and executive branches, government is also more eﬀective—especially in contrast to divided government with high levels of party competition. Better management depends on a conservative majority party, but greater innovation reﬂects a liberal governing party. There is hardly uniform agreement on what constitutes good government, so it is not surprising that good “management” gets high scores in conservative states and “innovation” ranks more highly in liberal states. And because innovation may reﬂect policies more salient to the public, the greater link between Ford Foundation/Kennedy School awards to mass polarization makes sense. Because the bureaucracies respond directly to state legislatures, legislative polarization should matter more—and it does.

156

Uslaner

The Civil State and Honest Government One criterion for good government that gains widespread support is honesty. Corrupt governments rob the public purse and take away funds that could be used for more productive purposes (Mauro, 1997, 7)—and corrupt oﬃcials look out more for themselves than for the public weal. Corruption leads to ineﬀective government (Mauro, 1997, 5; LaPorta et al., 1998, 32)—indeed, corruption is sometimes used as an indicator of ineﬀective government (Knack and Keefer, 1997). Across countries, corruption leads to higher tax evasion; lower growth; fewer expenditures for the public sector, education, or transfer payments from the rich to the poor; lower levels of government responsiveness; more bureaucratic red tape; and a less eﬀective judiciary (Uslaner, 2004). Corruption is the scourge of good government. Corruption also depends on a civil state. Countries with high levels of trust have low levels of corruption.7 So as with other measures of good government, I expect that states with high levels of trust will have less corruption. But beyond this measure of the civil state, there are fewer reasons to expect that ideological polarization should generate greater honesty. And it doesn’t. Instead, I argue that beyond a trusting public, we need a competitive polity with weaker party organization, vigilant elites, as well as greater equality among people. Corrupt regimes thrive on inequality. It transfers resources from the poor to the rich (Onishi and Banerjee, 2001)— and depends on clientelistic relations between elites and masses. Where there is greater inequality, it is easier for leaders to exploit the public—and especially when there is great social and racial inequality, political leaders can pursue a divide and conquer strategy and spend more time enriching themselves than their constituents. The measure of racial inequality I use is the black/white poverty ratio in a state. Trust and inequality are intricately related: High levels of inequality lead to lower generalized trust.8 Both high levels of inequality and low generalized trust should contribute to a culture of corruption. A one-party state, and especially a state with strong parties, will likely lead to greater corruption. In one-party states, politicians don’t fear that someone is looking at their performance carefully—and they have little fear that they might be thrown out of oﬃce if they are found ethically wanting. A more vigilant citizenry can, in the words of former Virginia Lt. Governor Henry Howell, “keep the big boys honest.” I examined a variety of measures of political activity and include only one—the share of citizens in a state who make a political speech.This is a tiny fragment of the population, but it does point to an elite that is active in political

Trust, Polarization, and the Quality of State Government 157

Table 8.3 Model of Corruption Perceptions in the American States Independent Variable

Generalized trust Dominant party share Traditional party organization Black/white poverty ratio Make political speech Constant

Coefficient

Standard Error

t Ratio

–2.578*** 3.850** 0.232*** 0.435*** –15.063*** 2.900***

1.080 1.751 0.079 1.634 6.195 0.728

–2.39 2.20 2.98 2.66 –2.43 3.98

R2 = 0.733;Adjusted R2 = 0.678; RMSE = 0.589; N = 30 Source: Uslaner and Brown (2005).

aﬀairs—that may have access to the media.9 Traditional party organizations are the homes of old-line political machines, sources of patronage and dishonesty. I report the regression for perceptions of corruption in the states in Table 8.3 above. The corruption measure I use is the one that seems the most reliable: the reporters’ perceptions of the level of corruption (Boylan and Long, 2001). Each of the independent variables strongly shapes corruption levels in the 30 states for which there are complete data on all of the variables.The truncated sample does not seem unrepresentative of the 44 states in the earlier analyses, at least based on the one variable common to all estimations, trust.The mean share of trusting respondents in the 30 states in this model is 0.359 compared to 0.383 for the 44 states in the earlier analyses.10 Additionally, the reporters’ corruption mean for the full 47 states for which data are available has a mean of 3.484 and a standard deviation of 1.038, while the estimated sample has a mean of 3.578 and a standard deviation of 1.114 (with higher scores indicating greater corruption on a scale from 1.5 to 5.5). The omitted states, then, are slightly more trusting and a bit less corrupt—but the diﬀerences do not seem great. High levels of trust and greater black/white inequality lead to lower levels of corruption—although we ﬁnd more corruption in one-party states and in states with very powerful political parties.The index of corruption ranges from 1.5 (least) to 5.5 (most). Moving from the least to the most trusting state, corruption increases by 1.17—a change equivalent to the distance between Pennsylvania and Rhode Island. As the black/white poverty ratio increases, corruption rises by 1.25, the gap between Utah and Rhode Island.The most dominant party system has a corruption ranking 0.862 higher than the least and strong party states are 0.98 “more corrupt”—an increase equivalent from moving from Ohio to Rhode Island. As the share of citizens reporting making a political speech rises from less

158

Uslaner

than 1 percent to slightly more than 9 percent, perceived corruption falls by 1.3 percent, the diﬀerence between Maryland and Rhode Island.11

Reprise Honest government, like good government, rests on a civic population. But polarization, either in the legislature or in the executive, doesn’t matter so much for honesty as it does for performance and innovation. Both good government and honest government do rest on social capital—but not social capital as Putnam traditionally perceives it. Membership in civic organizations has no signiﬁcant eﬀects on good government or on honest government. Political participation, outside of the handful of people making political speeches in the corruption model, has no impact on any of the GPP measures, the Ford Foundation/Kennedy School awards, or corruption. There is also little evidence that institutional structures matter. Structural variables such as gubernatorial power or legislative professionalism (King, Zeckhauser, and Kim, 2001) fade to insigniﬁcance once trust and polarization are taken into account. Legislative professionalism does not even predict corruption (or the lack of it) when added to the model in Table 8.3.We might expect that “Interest Group Strength” in the states could lead to less eﬀective government—because interest groups ﬁght for their own interests against the common weal—or “good government” groups such as Common Cause might promote better government (King, Zeckhauser, and Kim, 2001). In my models above, the good and the bad eﬀects of interest groups wash out in the face of trust, polarization, and competition. Corruption is rather sticky. Across 52 countries, the correlation of a corruption perceptions measure in 1980–1983 by Business International (Monitola and Jackman, 2002) and the Transparency International indicator for 2003 is 0.867. There are no early measures for corruption in the American states, but it makes sense to expect an inverse relationship between present-day corruption and support for political reform in the past. A good measure of support for reform is the state-level vote for Robert LaFollette in 1924.The LaFollette vote in 1924 does clearly track reporters’ perceptions of corruption in the late 1990s (r 2 = 0.284). The LaFollette vote is also highly correlated with trust in the 1990s (r = 0.624) and moderately correlated with traditional party organizations (r = –0.263)—so the historical measure drops out of a multivariate estimation. But there is clear support for the link of corruption to a civic culture (and a civil state), reaching back to the 1920s. What you get out of government is what you put into it.You need a civic citizenry to get a civil state. Possibly, the argument might work the

Trust, Polarization, and the Quality of State Government 159

other way around—good government produces good citizens (Brehm and Rahn, 1997). But this seems less likely. Government management is out of sight for most people, so it seems unlikely that citizens learn to trust one another, and especially people who are diﬀerent from themselves, by judging ﬁnancial auditing or capital budgets. More generally, people use rather diﬀerent criteria for judging government performance and deciding whether to trust others (Uslaner, 2002, ch. 5). Government performance and honesty are important and its link to generalized trust, inequality, polarization, and political competition suggest that reform is more than an exercise in political engineering. Some states are simply better candidates for good government than others. Structural tinkering, such as a more professionalized legislature, may be as in George Bernard Shaw’s message about second marriages,“the triumph of hope over experience.”

Notes Acknowledgments. This research was supported by a grant from the Russell Sage Foundation and the Carnegie Corporation under the Social Dimensions of Inequality Project. Some of the data come from the Inter-University Consortium for Political and Social Research, which is not responsible for any of our interpretations. I am also grateful to the General Research Board, University of Maryland—College Park, for support on related projects and to M. Mitchell Brown for research assistance, to David C. King for both providing data and important bibliographic help, and to an anonymous reviewer for Stanford University Press—and the editor, Jeﬀrey Cohen. 1. For more on these measures see Uslaner, 2002. 2. These data have been updated by Richard F.Winters, who kindly provided them to me. 3. See Uslaner and Brown (2005) for a discussion of the data and a more comprehensive list of the pre-1990 surveys we used for generating trust for other decades. Fengshi Wu helped with the data aggregation at the beginning of the project and M. Mitchell Brown did most of the work.We used the following surveys for generating the trust estimates: General Social Survey or GSS (1990, 1991, 1993, 1994, 1996, and 1998), American National Election Study (1992, 1996, and 1998), the Washington Post trust in government survey (1995), the Pew Civic Engagement survey (1997), the New York Times Millennium survey (1999). We are grateful to Robert Putnam and John Robinson for making the state codes for the GSS available to us and to Richard Morin of the Washington Post for the 1995 survey and to Michael Kagay of the New York Times for his paper’s Millennium survey. 4. The GPP and Ford Foundation/Kennedy School measures as well as measures of party competition were kindly provided by David C. King. The divided government variables come from the 1994, 1996, and 1998 editions of the Book of the States (computed by me).The data on public opinion and legislative polariza-

160

Uslaner

tion were graciously provided by Gerald C. Wright and the data on racial inequality in poverty (below) came from Rodney Hero. Arthur Jones, Jr., of the United States Bureau of the Census provided the state level Gini indices (also see below). 5. Because the models are identical, the coeﬃcients will be identical to those of ordinary least squares (OLS). However, the standard errors will be incorrectly estimated using OLS.The correlations among the residuals range from moderate (0.273) to high (0.651), except for the Ford Foundation/Kennedy School awards measure (ranging from 0.058 to 0.293).The Breusch-Pagan test of independence has Chi square = 91.433, p < 0.00001, so OLS would be less eﬃcient. 6. The overall range of the Ford Foundation/Kennedy School awards is from 0 to 15. 7. And countries with low levels of corruption have high levels of trust. Uslaner (2004) sorts out the causal order by looking at changes in trust and corruption over time and ﬁnds that trust comes ﬁrst in the causal order. 8. Both over time in the United States and cross-nationally. See Uslaner (2002, chs. 6, 8). 9. The measure comes from the Roper Social and Political Trends Archive; the Roper survey asked national samples of 2,000 ten times a year for 21 years (1974– 1994) about levels of political activity. Hence there are suﬃcient cases to aggregate activities for the states by decade. Here I use the estimates for making political speeches in the 1990s. 10. The standard deviations, respectively, are 0.110 and 0.117.The 14 states excluded from this model are Alaska, Arizona, Kansas, Kentucky, Massachusetts, Montana, New Hampshire, New Jersey, North Dakota, Oklahoma, Rhode Island, Vermont,Wisconsin, and Wyoming. 11. As a long-time resident of Maryland, this seems less consequential than the data suggest.

References Almond, Gabriel, and Sidney Verba. 1963. The Civic Culture. Princeton: Princeton University Press. Altschuler, Alan A. 1997. “Public Innovation and Political Incentives.” Unpublished paper, John F. Kennedy School of Government, Harvard University, at http://www.innovations.harvard.edu/research/papers/altshuler.pdf (March 4, 2004). Binder, Sarah A. 1999.“The Dynamics of American Gridlock, 1947–96.” American Political Science Review 93 (September): 519–534. Boylan, Richard T., and Cheryl X. Long. 2001. “A Survey of State House Reporters’ Perception of Public Corruption.” Unpublished manuscript, Department of Economics,Washington University in St. Louis. Brace, Paul, Kellie Sims-Butler, Kevin Arcenaux, and Martin Johnson. 2002.“Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46 ( January): 173–189. Brehm, John, and Wendy Rahn. 1997. “Individual Level Evidence for the Causes

Trust, Polarization, and the Quality of State Government 161 and Consequences of Social Capital.” American Journal of Political Science 41 ( July): 888–1023. Conley, Richard S. 2003. The Presidency, Congress, and Divided Government. College Station,TX:Texas A&M Press. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Inglehart, Ronald. 1999.“Trust,Well-Being and Democracy.” In Mark Warren, ed., Democracy and Trust. Cambridge: Cambridge University Press. Kaufmann, Daniel, Aart Kray, and Massimo Mastruzzi. 2003. “Governance Matters: Governance Indicators for 1996–2002.” Washington, DC: The World Bank, at (February 24, 2004). King, David C., Richard J. Zeckhauser, and Mark T. Kim. 2001. “Explaining the Management Performance of U.S. States.” Unpublished manuscript, John F. Kennedy School of Government, Harvard University. Knack, Stephen. 2002. “Social Capital and the Quality of Government: Evidence from the States.” American Journal of Political Science 46 (October):772–787. Knack, Stephen, and Philip Keefer. 1997.“Does Social Capital Have An Economic Payoﬀ? A Cross-Country Investigation.” Quarterly Journal of Economics 112 (November): 1251–1288. LaPorta, Rafael, Florencio Lopez-Silanes,Andrei Schleifer, and Robert W.Vishney. 1997.“Trust in Large Organizations.” American Economic Review Papers and Proceedings 87 (May): 333–38. ———. 1998. “The Quality of Government. “ Unpublished manuscript, Harvard University. ———. 1999. “The Quality of Government.” Journal of Law, Economics, and Organization 15 (March): 222–279. Lane, Robert E. 1959. Political Life. New York: Free Press. Mayhew, David R. 1991. Placing Parties in American Politics. Princeton: Princeton University Press. Mauro, Paolo 1997. “Why Worry About Corruption?” Washington: International Monetary Fund. McCubbins, Mathew D. 1991.“Government on Law-Away: Federal Spending and Deﬁcits Under Divided Party Control.” In Gary W. Cox and Samuel Kernell, eds., The Politics of Divided Government. Boulder, CO:Westview. Meier, Kenneth J., and Thomas M. Holbrook. 1992. “‘I Seen My Opportunities and I Took ‘Em:’ Political Corruption in the United States.” Journal of Politics 54 (February): 135–155. Montinola, Gabriella R., and Robert W. Jackman. 2002.“Sources of Corruption:A Cross-Country Study.” British Journal of Political Science 32 ( January): 117–170. Onishi, Norimitsu with Neela Banerjee. 2001.“Chad’s Wait for Its Oil Riches May Be Long.” New York Times (May 16), (May 17, 2001). Putnam, Robert D. 1993. Making Democracy Work: Civic Traditions in Modern Italy. Princeton: Princeton University Press. ———. 2000. Bowling Alone:The Collapse and Revival of American Community. New York: Simon and Schuster.

162

Uslaner

Rosenblum, Nancy L. 1998. Membership and Morals. Princeton: Princeton University Press. Sparrow, Malcolm K. 2000. The Regulatory Craft.Washington, DC: Brookings. Uslaner, Eric M. 1993. The Decline of Comity in Congress. Ann Arbor, MI: University of Michigan Press. ———. 2000.“Is the Senate More Civil Than the House? “ In Burdett Loomis, ed., Esteemed Colleagues: Civility and Deliberation in the Senate. Washington, DC: Brookings Institution, 32–56. ———. 2002. The Moral Foundations of Trust. New York: Cambridge University Press. ———. 2004. “Trust and Corruption.” In Johann Graf Lambsdorf, Markus Taube, and Matthias Schramm, eds., Corruption and the New Institutional Economics. London: Routledge. Uslaner, Eric M., and M. Mitchell Brown. 2005.“Trust, Inequality, and Civic Engagement.” American Politics Research 33. Zellner, Arnold. 1962. “Estimators of Seemingly Unrelated Regressions: Some Exact Finite Sample Results.” Journal of the American Statistical Association 58 ( June): 977–992.

Chapter

Public Opinion and Policy Making in the Culture Wars: Is There a Connection Between Opinion and State Policy on Gay and Lesbian Issues?

9

Donald P. Haider-Markel and Matthew S. Kaufman

Introduction Government responsiveness to public preferences is a key tenant of democracy. But state government, often referred to as the invisible layer of government, is sometimes thought to ignore the opinions of an electorate whose attention is focused on local and national level politics (Roeder, 1994). Perhaps for this reason the inﬂuence of public opinion on statelevel policy adoption in the United States has not received much attention (but see Erikson, Wright, and McIver, 1989, 1993; Lowery, Gray, and Hager, 1989; Roeder, 1994; Weber and Shafer, 1972). Furthermore, only recently have researchers examined the inﬂuence of opinion on policy using measures of opinion on speciﬁc policy issues, such as abortion (Norrander and Wilcox, 1999). Early state-level research on the relationship between public opinion and policy made use of national opinion data to generate “simulations or dummy variables to obtain state opinion estimates” of opinion (Roeder, 1994, 37). By the 1980s researchers had developed more reliable measures of citizen partisanship and ideology (Erikson, Wright, and McIver, 1989, 1993). From that point researchers have developed far more valid measures of state-level opinion, including measures of state opinion on speciﬁc issues (Brace, Sims-Butler, Arceneaux, and Johnson, 2002; Norrander and 163

164

Haider-Markel and Kaufman

Wilcox, 1999), but it is clear that additional research is needed to understand when and how public opinion inﬂuences state policy adoption. We contribute to this growing ﬁeld by examining congruence between public opinion and policies impacting the gay and lesbian community. In particular we use event history analysis to explore the inﬂuence of public opinion on the state repeal of laws banning sodomy, state adoption of hate crime laws that include sexual orientation, and the laws banning the recognition of same-sex marriages.We also examine the inﬂuence of public opinion on a broad set of gay and lesbian related policies as a policy index. Thus, our chapter is the ﬁrst systematic study of the relationship between opinion and policy across a variety of gay-related policy issues and the ﬁrst to make use of a variety of gay-related state-level opinion measures. Our chapter begins with a brief overview of previous public opinion research, the inﬂuence of opinion on policy, and the circumstances under which we should expect public opinion to inﬂuence policy adoption. Next, we discuss the merits of examining gay and lesbian issues as a means for understanding the inﬂuence of public opinion on policy.We then outline our strategy of analysis and methodology. Finally, we explain the results of our multivariate analysis and discuss the implications of our ﬁndings.

The Inﬂuence of Public Opinion on Policy Adoption Representative democracy is based on the premise that government policies should reﬂect the preferences of citizens. In sum, elected oﬃcials depend on public support for reelection, and because oﬃcials desire to be reelected, they should be responsive to the electorate’s preferences (Mayhew, 1974).Thus, we should observe empirical connections between public opinion and policy (Burstein, 2003; Erikson et al., 1989, 1993; Page and Shapiro, 1983). For example, in a classic study Page and Shapiro (1983, 189) analyzed 357 instances of opinion change in the United States from 1935 and 1979 and found that “opinion changes are important causes of policy change.” Likewise, Hartley and Russett (1992) explored public opinion and changes in military spending and found a signiﬁcant relationship between opinion and policy. More recently, Monroe (1998) examined policy outcomes on over 500 issues between 1980 and 1993 and found that policy reﬂected public opinion in 55 percent of the cases.These studies are illustrative, but they focused on federal policy adoption and do not tell us whether opinion and policy are congruent at the subnational level.

Public Opinion and Policy Making in the Culture Wars 165

Early research on state policy-opinion congruence used simulation techniques to estimate state public opinion (Monroe 1998). Researchers uncovered a clear link between these simulated measures and policy (Cook, 1977; Edwards and Sharkansky, 1978; Schumaker and Getter, 1977; Weber and Shafer, 1972). As state-level data has become more abundant, recent studies have employed more direct measures of state-level opinion (Brace et al., 2002). For example, Lowery et al., (1989) found policyopinion congruence on tax progressivity. Additional studies at the state level have used direct measures of public opinion on speciﬁc issues and found policy-opinion congruence (Brace et al., 2003; Erickson et al., 1989, 1993; Hill and Hinton-Anderson, 1995; Kane, 2003; Norrander and Wilcox, 1999). Nevertheless, some researchers have raised questions regarding the inﬂuence of opinion on policy. They argue that gerrymandered districts (Cain, 1984), a decline in voter turnout (Burnham, 1982), and the domination of incumbents in national and state races have limited voter control of elected oﬃcials (Mayhew, 1974). Subsequently, these authors suggest we should expect less congruence between opinion and policy. Other researchers note that the link between opinion and policy is often inconsistent. For example, Gerber (1996, 100) argues that legislators are tied too closely to the parochial interests of their districts, creating the possibility that state-level “policy may diverge substantially and systematically from [aggregated] citizen preferences.” In addition, a lack of citizen knowledge, interest, and consistent preferences over time can also reduce the inﬂuence of public opinion on policy (Dye, 1987, 326). Monroe’s (1998) ﬁndings support these arguments and also suggest that lack of congruence over time might result from increased ideological conﬂict among elites and an increase in divided government (see also Lindaman and Haider-Markel, 2002). These arguments should not be taken as a refutation of a link between policy and public opinion. Rather, they should be taken as an indication that the inﬂuence of opinion on policy may be more likely under some circumstances versus others. For example, citizens are more likely to care about and be informed on issues that are highly salient in the media and are a topic of discussion among political elites. Thus, in cases of high issue salience, the link between public opinion and policy-making outcomes should be greater (Carmines and Stimson, 1989; Gormley, 1986; Hinkley, 1983; Lindaman and Haider-Markel, 2002; Monroe, 1998; Page and Shapiro, 1983). Likewise, Mooney and Lee (2000) provide evidence that policy makers are more responsive to public opinion when opinion is relatively divided (contentious) versus when it is more lopsided (consensus) on

166

Haider-Markel and Kaufman

salient issues such as the death penalty. Thus, if the issue is salient, and the public is relatively evenly split on an issue, the conditions should be ripe for public opinion to most strongly shape public policy.

The Salience of Gay and Lesbian Policy As a subset of morality policy, gay and lesbian policy is an especially appropriate area for examining potential links between public opinion and policy (Haider-Markel and Meier, 1996; Meier, 1999; Mooney, 2000). Mooney and Lee (1995, 600) consider “morality policies to be those which seek to regulate social norms or which evoke strong moral responses from citizens for some other reason. . . . Morality policy raises questions that instigate debate over ﬁrst principles, resulting in uncompromising clashes of values.” Morality policy, being highly salient and often being framed in terms easy for citizens to understand, is therefore responsive to public opinion (Haider-Markel and Meier, 1996; Lindaman and Haider-Markel, 2002; Mooney and Lee, 2000; Norrander and Wilcox, 1999). Furthermore, gay and lesbian issues easily ﬁt these criteria, making gay-related policy one where we should expect to ﬁnd a relatively strong linkage between opinion and policy (Haider-Markel, 1999; Lewis and Edelson, 2000). Indeed, researchers have demonstrated that decisions by elected oﬃcials on highly salient and noncomplex issues like morality policy generally and gay and lesbian policy speciﬁcally should be more likely to reﬂect constituency opinion (Carmines and Stimson, 1989; Haider-Markel, 1999; Tatalovich and Daynes, 1988, 218).And at the aggregate level, citizen preferences should be better reﬂected in policy (Kane, 2003; Lewis, 2003; Lewis and Edelson, 2000). Given that this is a relatively easy case, if we are unable to ﬁnd an inﬂuence of public opinion on gay and lesbian policy, we would have grounds to more strongly question whether opinion plays as much of a role in state policy making as democratic theory would suggest.

Methodology and Strategy of Analysis To determine whether public opinion inﬂuences state gay and lesbian policy we develop a general model of state policy adoption popularized by Berry and Berry (1999).We argue that the likelihood a state will adopt a speciﬁc policy in any given year is a function of three broad factors. First, adoption is driven by the motivation of public oﬃcials to adopt the policy. Motivation is typically driven by elite preferences, the salience of the issue, and the extent of the problem. Second, there are resources and

Public Opinion and Policy Making in the Culture Wars 167

obstacles to policy adoption internal to a state, such as public opinion, the ﬁnancial resources of a state, socioeconomic conditions, and the presence of policy entrepreneurs.And third, policy adoption is inﬂuenced by external diﬀusion eﬀects, such as the adoption of the policy by other nearby or regional states, and potentially, actions by the national government. Using data from the American states we create three separate models measuring the likelihood of the following events: (1) state repeal of laws banning sodomy, (2) state adoption of hate crime laws that include sexual orientation, and (3) state adoption of laws banning same-sex marriage. Each of these policy areas has been a key concern for gay activists and their opponents over the past 15 years and has become increasingly salient since the 1980s (Haider-Markel, 2000; Kane, 2003; Lewis, 2003; Smith and Haider-Markel, 2002). The coeﬃcients for each of these models are estimated using logistic regression in an event history analysis.The dependent variable for each is a dichotomous variable coded 1 if the state adopts the policy. If a state does adopt the policy, it is removed from the risk set. The data for each model begins at the ﬁrst year any state adopts or repeals the given policy. In addition we estimate a fourth model to assess whether opinion is linked to policy in a broader manner. Here we use an index of gay-related policies to indicate the extent to which state policies are friendly toward the gay and lesbian community. The index can range from –100 to 100 (see Table 9.1).The coeﬃcients for this model are estimated using ordinary least squares (OLS) regression.

Dependent Variables Data for the ﬁrst dependent variable, repeal of sodomy laws, is from Haider-Markel (2000) and the National Gay and Lesbian Task Force.The data set begins in 1961, when the ﬁrst state (Illinois) repealed its law criminalizing sodomy between consenting adults, and ends in 2002, when 14 states still maintained sodomy laws (see Table 9.1).1 States that repealed laws criminalizing sodomy between opposite-sex adults but maintained laws criminalizing sodomy between same-sex adults were not counted as having repealed their anti-sodomy laws. Data on state hate crime laws begins in 1984. In that year California was the ﬁrst state to pass hate crime legislation that included sexual orientation. The data set ends in 2002 when 28 states had hate crime laws protecting gays and lesbians (see Table 9.1).2 These laws fall under three broad categories: laws prohibiting intimidation or interference with civil rights, laws that create separate bias-motivation crimes, and laws that provide penalty enhancement provisions (Haider-Markel, 1998; Jenness and

Table 9.1 Selected Gay- and Lesbian-Related Policies in the American States State

AL AK AZ AR CA CO CT DE FL GA HI ID IL IN IA KS KY LA ME MD MA MI MN MS MO MT NE NV NH NJ NM NY NC ND OH OK OR PA RI SC SD TN TX UT VT VA WA WV WI WY

Scores on Gay-Friendly Policies

–82 –40 –25 –65 54 –35 81 –10 –67 –39 42 –80 –5 –20 –11 –82 –22 –61 2 4 64 –19 24 –90 –40 –37 –27 8 66 70 11 6 –80 –35 –4 –92 34 –33 68 –70 –37 –30 –18 –77 97 –87 9 –35 60 5

Hate Crime Law

Ban on Same-Sex Marriage

Law Sodomy (repeal)

0 0 95 0 84 0 87 95 91 0 01 0 90 0 90 0 98 97 87 91 96 0 88 0 99 0 97 89 90 90 0 00 0 0 0 0 89 02 91 0 0 00 91 0 90 0 93 0 88 0

96* 96 96 97 00 00 0 96 97 96 94 96 96 97 98 96 98 96# 97 0 0 96 97 96* 96 97 00 02 0 0 0 0 96 97 0 96 0 96 0 96 96 96 0 95 0 97 98 0 0 0

0 78 75 02 75 71 69 73 0 98 72 0 61 76 76 0 92 0 75 99 02 0 01 0 0 96 77 93 73 75 75 75 0 77 72 0 71 80 98 0 76 96 82 0 77 0 75 76 83 77

Notes: Existing laws are indicated by the year passed; states without a law are indicated by a 0; hate crime laws are listed only for those states that include sexual orientation in their law; # indicates nonbinding legislative resolution; sodomy laws may have been repealed by the state legislature or overturned in court–in some cases both actions occurred. Source: Haider-Markel (2000) and the National Gay and Lesbian Task Force.

Public Opinion and Policy Making in the Culture Wars 169

Grattet, 2001). Any state with a law covering one or more of these areas was counted as having a hate crime law. States with hate crime laws that exclude sexual orientation were not counted. Our third dependent variable captures laws banning the recognition of same-sex marriages. States began adopting laws deﬁning marriage as between one man and one woman in response to a Hawaii Supreme Court decision that threatened to allow marriage between same-sex partners (Haider-Markel, 2001). Between 1994 and 2002, 35 states and the federal government passed laws, executive orders, or issued nonbinding legislative resolutions banning same-sex marriage (see Table 9.1). States adopted these policies rapidly with 24 states adopting laws between 1996 and 1998.3 Although the gay and lesbian movement has been divided on pursuing legalized same-sex marriage, activists were uniformly opposed to policies banning same-sex marriage (Smith and Haider-Markel, 2002). Also, since June 2003, when the U.S. Supreme Court struck down sodomy laws, and November 2003, when the Supreme Judicial Court of Massachusetts ruled that the state could not prevent same-sex couples from being issued marriage licenses, a new wave of legislation and constitutional amendments have been introduced across the states. But prior to 2003, the 1990s wave of same-sex marriage bans had clearly ended (Kirkpatrick, 2004). The ﬁnal dependent variable comes from a Keen (2001). Keen performed an analysis of laws that aﬀect gays and lesbians in all 50 states. She assigned a score to each state on a scale ranging from –100 to +100. States earned or lost points based on existing laws. Positive points were assigned to states for pro-gay laws, including bans on discrimination in housing and employment, laws allowing same-sex couples to share domestic partner beneﬁts, and for laws that recognize gay parental, adoption, and foster care rights. Negative points were assigned for anti-gay laws, including laws banning homosexual sodomy, same-sex marriage, and explicitly preventing the recognition of gay parental rights. In addition to considering state laws, the study looked at executive orders, court rulings, local laws, the harshness of penalties, and the strength of legal protections. State scores range from a low of –92 (Oklahoma) to a high of +97 (Vermont). Thus, this measure captures of a snapshot of gay-related policies in the states at one point in time (see Table 9.1). We conduct cross-sectional analysis of this index because it provides us with a broader means by which to assess the potential linkage between opinion and policy. Our measures of speciﬁc policy are relatively narrow and therefore any relationship between opinion and policy may be obscured by the idiosyncratic nature of policy adoption across the states. For example, Minnesota had a law banning sodomy until 2001 even though

170

Haider-Markel and Kaufman

the states’ policies are generally gay-friendly. Our index of policy should be less sensitive to these outlying cases. However, the index is limited relative to the more speciﬁc measures of gay-related policy simply because it only captures policy at one point in time.

Independent Variables Based on the internal and external determinants model of policy adoption, our models incorporate variables measuring the following: (1) party competition, (2) government ideology, (3) presence of Protestant fundamentalists, (4) gay and lesbian interest group mobilization, and (5) regional adoption of the same policy. The ﬁrst two variables represent motivating factors for public oﬃcials, the second two represent resources and obstacles to policy adoption, and the ﬁnal measures diﬀusion eﬀects. In addition to these variables, each model will include a measure of public opinion on gay-related issues. The level of party competition in a state can inﬂuence the likelihood of policy adoption. If elected oﬃcials want to keep their jobs in a competitive party system, they will pursue issues and positions that appeal to the greatest number of constituents. Thus, greater party competition could make opinion-policy congruence more likely (Hill and Hinton-Anderson, 1995; Mooney and Lee, 1995, 617). Researchers suggest that is phenomenon leads to a more activist government with more progressive policies (HaiderMarkel and Meier, 1996, 338; Nice, 1988, 1992; Skogan, 1990, 393). However, when public opinion is clearly in favor of a policy, such as it is with bans on same-sex marriage (Lewis, 2003), we anticipate party competition will increase the likelihood of policy adoption—a nonprogressive outcome. In other areas, such as sodomy and hate crime laws, where opinion tends to be more divided, electoral competition should lead to the adoption of more progressive laws that beneﬁt gays and lesbians.We use the Holbrook and Van Dunk (1993) measure of party competition to test these hypotheses.4 Traditionally studies of state policy adoption include measures of partisan control of government. However, these measures fail to capture the nuances of ideological divergence within the parties across the country. For example, southern Democrats tend to be more conservative than their northern counterparts (Berry, Ringquist, Fording, and Hanson, 1998). Likewise, although Republicans, because of their focus on traditional family values, tend to be less tolerant of homosexuality (Haider-Markel, 2001, 14; Layman and Carmines, 1997), it is conservatives, both Republicans and Democrats, who tend to be least tolerant of homosexuals and less supportive of gay and lesbian civil rights (Yang, 2001). Therefore, we expect

Public Opinion and Policy Making in the Culture Wars 171

that more liberal state governments are more likely to adopt gay-friendly policies.We measure government ideology with the government ideology scores created by Berry et al. (1998).5 The scores are a composite of the ideology of a state’s governor and the legislature.The scores range from 0 to 100 with 0 representing the most conservative position and 100 representing the most liberal position. The prevalence of Protestant fundamentalists in a state is also likely to aﬀect the adoption of gay and lesbian policy (Button, Rienzo, and Wald, 1997; Haider-Markel, 2001; Kane, 2003; Nice, 1988; Wald, Button, and Rienzo, 1996). Protestant fundamentalist denominations have viewed homosexuality as immoral and as a threat to traditional values (see Diamond, 1995; Melton, 1991). Research has shown that larger Protestant fundamentalists populations make the adoption of gay civil rights policies less likely (Haider-Markel and Meier, 1996; Wald, Button, and Rienzo, 1996), the decriminalization of sodomy less likely (Kane, 2003), and the adoption of same-sex marriage bans more likely (Haider-Markel, 2001). Protestant fundamentalist populations might inﬂuence policy simply by having their beliefs represented in state opinion polls or through the political mobilization of their adherents in election and lobbying campaigns (Haider-Markel and Meier, 1996, 337).Thus, we hypothesize that a larger Protestant fundamentalist population will reduce the likelihood of state adoption of gay-friendly policies, but increase the likelihood of state adoption of policies that negatively impact gays and lesbians. We capture Protestant fundamentalists by including a measure of a state’s population that belongs to conservative Protestant denominations.6 Like religious conservatives, the gay and lesbian population can also inﬂuence state policy adoption.The gay population can serve as potential members for gay interest groups and a signiﬁcant grassroots resource in lobbying eﬀorts (Smith and Haider-Markel, 2002; Wald, Button, and Rienzo, 1996). One measure of potential gay interest group resources is the number of potential members a group can draw on.The larger the potential membership, the larger actual group membership should be. Groups with larger membership levels have greater potential inﬂuence in the policy-making process (Thomas and Hrebenar, 1996, 147). Our measure of gay and lesbian interest group mobilization is measured by the number of state residents per 100,000 state population belonging to the Human Rights Campaign and the National Gay and Lesbian Task Force.These are the two largest gay and lesbian groups in the country and they have been lobbying on gay issues for more than 20 years (Smith and Haider-Markel, 2002). Strong gay and lesbian interest group mobilization should increase the

172

Haider-Markel and Kaufman

probability of sodomy decriminalization, improve the likelihood of inclusion of sexual orientation in state hate crime laws, and result in a higher ranking gay-related policy index. Mobilization should also decrease the probability that a state will adopt same-sex marriage bans. Conditions external to a state may also inﬂuence policy adoption. A considerable body of research suggests that states are sensitive to the policy actions of the national government as well as to the policy actions of other states in the region or simply neighboring states (Berry and Berry, 1990; Eyestone, 1977; Gray, 1973, 1994; Lowry, 1992; Mooney, 2001; Mooney and Lee, 1995). On gay policy speciﬁcally, Kane (2003) found that gay and lesbian activists seeking the decriminalization of sodomy targeted states based on the success of decriminalization eﬀorts in surrounding states, and Haider-Markel (2001) found that the adoption of same-sex marriage bans was inﬂuenced by adoption of these policies by other states in the same region. We measure this external determinant on an annual basis by including a variable measuring the percentage of adjacent states that have adopted the policy being considered in a speciﬁc model. Diﬀusion should be positively related to policy adoption.7 The ﬁnal variable to be included in each model is a measure of statelevel public opinion. We consider three measures for their possible eﬀect on policy adoption: (1) public acceptance of gay sex, (2) public acceptance of gays in the workplace, and (3) a general measure of citizen political ideology. Our ﬁrst measure, public acceptance of homosexual sexual relations between adults, is based on data compiled by Lewis (2003) from 36 surveys and 56,000 respondents. Lewis (2003) developed an index that ranges from 0 to 1—the higher the score the lower the disapproval of legal homosexual sex.The measure has clear face validity as an indication of public support for sodomy laws.We expect it to be positively related to state repeal of sodomy laws. We also include this measure of public opinion in our model predicting state scores on the gay-friendly policy index. Sodomy laws, and widespread support for them, have often been used as a rationalization for laws and court rulings that discriminated against gays and lesbians (Haider-Markel, 2000; Kane, 2003; Lewis, 2003). Therefore, states where support for sodomy laws is strong should be more likely to have policies that negatively impact gays and lesbians overall. Further, because national polls until the mid-1990s suggested that 65 to 70 percent of adults believed that sexual relations between two adults of the same sex are always wrong (The Pew Research Center 2003, 17), we might expect the inﬂuence of public opinion on this issue to only have marginal inﬂuence on the repeal of sodomy bans. However, the gay-friendly policy index,

Public Opinion and Policy Making in the Culture Wars 173

which is a snapshot of state policy in 2001, coincides with divided public acceptance of homosexual sexual relations. In fact, by 2001 a bare majority of adults believed that sexual relations between two adults of the same sex are always wrong (The Pew Research Center 2003, 17). Thus, in the case of the policy index we expect a signiﬁcant positive association between public acceptance of homosexual sexual relations and state scores on the policy index. The second measure, public acceptance of gays in the workplace, is based on combining data from several Gallup Poll questions asking respondents about their acceptance of the hiring of gays as doctors, salespersons, elementary teachers, clergy, and members of the armed forces (Lewis and Edelson, 2000). Respondents were able to note disapproval for hiring gays in these occupations.Thus, this composite scale has face validity as a direct measure of support for bias against gays in hiring. Because hate crime laws are meant to decrease bias motivated crimes (HaiderMarkel, 1998), support for the laws should be contingent on a population believing bias against the protected group is wrong.Thus, higher levels of support for discrimination in hiring should reduce support for the inclusion of sexual orientation in hate crime laws.We expect public acceptance of gays in the workplace will increase the likelihood of state adoption of a hate crime law. Furthermore, because national polls show divided but majority support for including sexual orientation in hate crime laws and for banning sexual orientation discrimination in employment (HaiderMarkel, 2003), we expect that public opinion on this issue will have relatively greater inﬂuence on policy. Our ﬁnal measure of public opinion captures citizen political ideology. Again, because liberals tend to be more supportive of gay civil rights (Yang, 2001), states with more liberal citizens should be more likely to adopt gay-friendly policies.Therefore, in our model of same-sex marriage bans we include a measure of citizen ideology that was developed by Berry et al. (1998) and ranges from 0 to 100, with 0 being the most conservative position and 100 being the most liberal position.We expect that states with more liberal citizens should be less likely to adopt bans on same-sex marriage.

Results and Discussion Table 9.2 presents the results of the logistic regression model for sodomy law repeal.The table displays results for two models. For each policy area we estimate a model without a measure of public opinion (column 1) and a model with a measure of public opinion (column 2).

174

Haider-Markel and Kaufman Table 9.2 State Sodomy Law Repeal Independent Variables

Party competition

Model without Opinion Measure

Model with Opinion Measure

0.028 (0.020) 0.009 (0.008) –0.047* (0.025) –0.001 (0.001) 1.017*** (0.354) —

Constant

–4.764*** (1.096)

0.026 (0.020) 0.008 (0.008) –0.040 (0.025) –0.001 (0.001) 1.025*** (0.352) 0.019 (0.020) –5.550*** (1.344)

Log likelihood Pseudo R-square Chi square Prob. Chi square Percent predicted correctly Number of cases

311.198 0.10 27.058 0.000 97.2 1334

310.263 0.10 27.993 0.000 97.2 1334

Government ideology Protestant fundamentalists Gay interest groups Neighboring states Acceptance of gay sex

Note: Coefficients are logistic regression estimates; Significance levels in two-tailed test: ***< 0.01, **< 0.05, *< 0.10

The variable measuring Protestant fundamentalist population performs as expected, showing a negative inﬂuence on sodomy law repeal. In the second model the variable is not signiﬁcant at the 0.10 level. Consistent with previous research (Kane, 2003), and our expectations, the neighboring state adoption variable has a signiﬁcant positive inﬂuence on sodomy law repeal in both models. As Kane (2003) suggested, the repeal of a sodomy law in one state may lead activists to pursue repeal in surrounding states. Although the variables for party competition and government ideology are not signiﬁcant, their signs are in the hypothesized directions. Our measure of public opinion toward same-sex relations is not statistically signiﬁcant in the model nor does it improve the overall ﬁt of the model. Because public opinion on homosexual sodomy between consenting adults has generally been consensual, with large majorities responding that this behavior is always wrong, we expected public opinion to have relatively less inﬂuence in this policy area. Our results conﬁrm our suspi-

Public Opinion and Policy Making in the Culture Wars 175 Table 9.3 Adoption of Hate Crime Law Including Sexual Orientation Independent Variables

Party competition

Model without Opinion Measure

Constant

–4.928*** (1.360)

0.007 (0.023) 0.012 (0.010) –0.001 (0.025) –0.001 (0.001) –0.056 (0.089) 0.103*** (0.037) –4.285*** (1.303)

Log likelihood Pseudo R-square Chi square Prob. Chi square Percent predicted correctly Number of cases

217.891 0.09 16.485 0.006 96.6 698

211.052 0.13 23.913 0.001 96.6 698

Government ideology Protestant fundamentalists Gay interest groups Neighboring states Gays in the workplace

0.017 (0.023) 0.014 (0.010) –0.032 (0.027) 0.002* (0.001) –0.039 (0.093) —

Model with Opinion Measure

Note: Coefficients are logistic regression estimates; significance levels in two-tailed test: ***< 0.01, **< 0.05, *< 0.10

cions and suggest that sodomy decriminalization has been fought outside the arena of public opinion. Table 9.3 shows results from our analysis of hate crime law adoption. Again, most of our variables do not reach traditional thresholds of statistical signiﬁcance, making it diﬃcult to draw conclusions from the model’s results. However, gay interest group mobilization has a positive inﬂuence on policy adoption in the ﬁrst model.This result was expected and is similar to the ﬁndings in other work on hate crime policy (Haider-Markel, 1998). Contrary to expectations, the coeﬃcient changes sign when our opinion measure is added; however, the variable is not statistically signiﬁcant after adding the public opinion variable in the second model. The addition of our opinion measure substantially improved the overall ﬁt of the model and the opinion variable is signiﬁcant.The results suggest that the more accepting a state’s population is of gays and lesbians in the workplace the more likely a state is to adopt a hate crime law that includes

176

Haider-Markel and Kaufman Table 9.4 Adoption of Same-Sex Marriage Bans Independent Variables

Party competition

Model without Opinion Measure

Model with Opinion Measure

0.002 (0.019) –0.001 (0.009) 0.034** (0.017) –0.003** (0.001) 1.433*** (0.558) —

Constant

–2.211** (1.063)

0.006 (0.020) 0.002 (0.010) 0.030* (0.019) –0.002 (0.002) 1.505*** (0.573) –0.015 (0.024) –1.872 (1.177)

Log likelihood Pseudo R-square Chi square Prob. Chi square Percent predicted correctly Number of cases

189.558 0.142 21.429 0.001 87 269

189.142 0.145 21.842 0.001 87 269

Government ideology Protestant fundamentalists Gay interest groups Neighboring states Citizen ideology

Note: Coefficients are logistic regression estimates; significance levels in two-tailed test: ***< 0.01, **< 0.05, *< 0.10

sexual orientation. This ﬁnding is consistent with our expectations, especially because support for hate crime laws has been relatively contentious. Table 9.4 presents the results of our logistic regression analysis of state adoption of same-sex marriage bans. The results suggest that party competition and government ideology may have little or no inﬂuence on same-sex marriage bans. The measure of Protestant fundamentalists has a signiﬁcant positive inﬂuence on the likelihood of adoption. Likewise, gay interest group mobilization has a signiﬁcant negative inﬂuence on the likelihood of adoption. And as more neighboring states adopt a same-sex marriage ban, the likelihood of a state adopting a ban increases. When we include the measure of public opinion in column 2 the model ﬁt statistics improve marginally, but the opinion variable itself is not statistically signiﬁcant. The limited power of opinion to explain policy adoption in this model is consistent with Lewis (2003), who tested Mooney and Lee’s (2000) hypothesis that opinion matters less when con-

Public Opinion and Policy Making in the Culture Wars 177 Table 9.5 Determinants of State Scores on a Gay-Friendly Policy Index Independent Variables

Party competition Government ideology Protestant fundamentalists Gay interest groups Acceptance of gay sex R-square Adjusted R-square F-score Significance

Model without Opinion Measure

Model with Opinion Measure

1.178** (0.481) 0.459** (0.208) –1.238*** (0.426) 0.080** (0.032) —

0.807* (0.471) 0.408** (0.196) –0.985** (0.411) 0.045 (0.033) 1.536***

0.542 0.501 13.297 0.000

0.606 0.561 13.549 0.000

Note: Coefficients are OLS Regression estimates; significance levels in two-tailed test: ***< 0.01, **< 0.05, *< 0.10

sensus exists on morality policy, as it does with same-sex marriage bans. The rapid speed with which same-sex marriage bans were adopted may also indicate the inﬂuence of political elites and policy entrepreneurs (Haider-Markel, 2001). Participation by these actors can limit citizen inﬂuence on policy adoption. Table 9.5 displays the results of our OLS regression model predicting state scores on the gay-friendly policy index. Overall the models ﬁt the data well, explaining over half of the variation in the dependent variable. Our measures of party competition, Protestant fundamentalists, and gay interest groups are statistically signiﬁcant in the directions expected. However, after adding our public opinion measure, acceptance of homosexual sex, party competition, and interest group mobilization are no longer statistically signiﬁcant. In both cases the coeﬃcients remain in the hypothesized direction. Including our opinion measure signiﬁcantly improves the model ﬁt statistical by improving the R-square and F-score. The results suggest that higher acceptance of homosexual sex is associated with higher scores on the gay policy index. Because the policy index captures a variety of policies relating to gays and lesbians, including those that we examined individually above, we expected public opinion to have a signiﬁcant positive inﬂuence, and the results conform to our expectation.

178

Haider-Markel and Kaufman

Conclusions This chapter sought to examine the inﬂuence of public opinion on state policy in the area of gay and lesbian policy. Although a considerable body of research has examined state-level opinion and scholars have increasingly addressed policy making on gay issues, few researchers have explored both of these areas, and no existing studies have examined the inﬂuence of opinion on gay-related policy across a variety of policy areas or with a series of opinion measures. We address this research gap by examining the congruence of policy and opinion at the state level. Our analysis of state repeal of laws banning homosexual sodomy, adoption of hate crime laws including sexual orientation, adoption of same-sex marriage bans, and state scores on a gay policy index suggest that the inﬂuence of public opinion on policy varies depending on the distribution of opinion—if opinion is divided (contentious) or lopsided (consensus). When public opinion is contentious, its inﬂuence on policy is more likely, such as in the case of hate crime laws including sexual orientation. However, if opinion is consensual, then policy making is more likely to be driven by entrepreneurs, and is likely to occur rapidly, such as in the case of same-sex marriage bans (Haider-Markel, 2001; Lewis, 2003). Opinion was also of little importance when courts played a major policy-making role, such as in the case of sodomy law repeal. When comparing congruence between public opinion and a broadbased index of gay policies, we found that opinion was positively associated with gay-friendly policies. We anticipated this result simply because the index better reﬂected overall state policy and was not subject to the idiosyncrasies of policy adoption in speciﬁc issue areas. Our analysis of the index provides perhaps the most convincing set of ﬁndings regarding the link between opinion and policy, and leads us to conclude that elected oﬃcials respond to citizen preferences on this battlefront in the culture wars. However, our results also indicate that morality policy theory, which suggests there will always be a strong congruence between opinion and policy on moral issues, should be revised as outlined by Mooney and Lee (2000) to incorporate the idea that opinion is most likely to inﬂuence policy if citizens are divided. But we do not want to suggest that public preferences will not be congruent with policy when opinions are in consensus.We only suggest that because opinion is lopsided, other factors in the political process will be more important predictors of policy process outcomes. Future researchers should explore the nature of the opinion/policy relationship in other issue areas where public preferences are contentious on some speciﬁc polices, but consensual on others.

Public Opinion and Policy Making in the Culture Wars 179

Notes 1. In June 2003 the U.S. Supreme Court declared all laws criminalizing private, consensual sodomy between consenting adults unconstitutional in Lawrence v.Texas.Thus, 2002 is a reasonable point to end the data set. 2. Data are from Smith and Haider-Markel (2002) and the National Gay and Lesbian Task Force. 3. Data are from the National Gay and Lesbian Task Force and Haider-Markel (2001). 4. Holbrook and Van Dunk (1993) do not provide an estimate for Louisiana. Our comparison of this measure with other measures of party competition allows us to estimate Louisiana’s score at 17.07. 5. Data for 1999 comes from the Berry et al. (1998) data update on the ICPSR website, and data for 2000 to 2002 are extrapolated from earlier years. 6. Data for 1970 are from Johnson, Picard, Quinn (1974), 1980 from Quinn et al. (1982), and 1990 from Bradley et al. (1992). Intervening years were interpolated and years beyond 1990 were extrapolated based on existing data. 7. We do not include this variable in our model predicting overall gayfriendly policy scores simply because the index is measured at only one point in time.

References Berry, Frances, and William D. Berry. 1990.“State Lottery Adoptions as Policy Innovations: An Event History Analysis.” American Political Science Review 84(2): 395–415. ———. 1999. “Innovation and Diﬀusion Models In Policy Research.” In Paul E. Sabatier, ed., Theories of the Policy Process, Boulder, CO:Westview. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 1960–93.” American Journal of Political Science 42(1): 327–348. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46(1): 173–189. Bradley, Martin B., Norman M. Green, Dale E. Johnson, Mac Lynn, and Lou McNeil. 1992. Churches and Church Membership in the United States, 1990. Atlanta: Glenmary Research Center. Burnham, Walter Dean. 1982. The Current Crisis in American Politics. New York: Oxford University Press. Burstein, Paul. 2003. “The Impact of Public Opinion on Public Policy: A Review and an Agenda.” Political Research Quarterly 56(1): 29–40. Button, James W., Barbara A. Rienzo, and Kenneth D.Wald. 1997. Private Lives, Public Conﬂicts: Battles Over Gay Rights in American Communities. Washington, DC: CQ Press. Cain, Bruce E. 1984. The Reapportionment Puzzle. Berkley: University of California Press.

180

Haider-Markel and Kaufman

Carmines, Edward G., and James A. Stimson. 1989. Issue Evolution: Race and the Transformation of American Politics. Princeton, NJ: Princeton University Press. Cook, Beverly B. 1977. “Public Opinion and Federal Judicial Policy.” American Journal of Political Science 21(3): 567–600. Diamond, Sara.1995. Roads to Dominion: Right-Wing Movements and Political Power in the United States. New York: Guilford. Dye, Thomas R. 1987. Understanding Public Policy. Englewood Cliﬀs, NJ: Prentice Hall. Edwards, George, and Ira Sharkansky. 1978. The Policy Predicament. San Francisco: W. H. Freeman. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. ———. 1989. “Political Parties, Public Opinion, and State Policy in the United States.” American Political Science Review 83(3): 729–750. Eyestone, Robert. 1977.“Confusion, Diﬀusion, and Innovation.” American Political Science Review 71(2): 441–447. Gerber, Elisabeth R. 1996. “Legislative Response to the Threat of Popular Initiatives.” American Journal of Political Science 40(1): 99–128. Goggin, Malcolm L., Ann O. Bowman, James P. Lester, and William T. Gormley. 1986.“Regulatory Issue Networks in a Federal System.” Polity 18 (4): 595–620. Gray,Virginia. 1973. “Innovation in the States: A Diﬀusion Study.” American Political Science Review 67(4): 1174–1185. ———. 1994. “Competition, Emulation, and Policy Innovation.” In Lawrence C. Dodd and Calvin Jillson, eds., New Perspectives on American Politics.Washington, DC: CQ Press. Haider-Markel, Donald P. 1998. “The Politics of Social Regulatory Policy: State and Federal Hate Crime Policy and Implementation Eﬀort.” Political Research Quarterly 51(1): 69–88. ———. 1999. “Redistributing Values in Congress: Interest Group Inﬂuence Under Sub-Optimal Conditions.” Political Research Quarterly 52(1): 113–144. ———. 2000.“Lesbian and Gay Politics in the States: Interest Groups, Electoral Politics, and Public Policy.” In Craig A. Rimmerman, Kenneth D.Wald, and Clyde Wilcox, eds., The Politics of Gay Rights. Chicago: University of Chicago Press. ———. 2001.“Policy Diﬀusion as a Geographical Expansion of the Scope of Political Conﬂict: Same-Sex Marriage Bans in the 1990s.” State Politics and Policy Quarterly 1(1): 5–26. ———. 2003.“Gay Rights.” In Robert Singh, ed., Governing America:The Politics of a Divided Democracy. New York: Oxford University Press. Haider-Markel, Donald P., and Kenneth J. Meier. 1996. “The Politics of Gay and Lesbian Rights: Expanding the Scope of the Conﬂict.” The Journal of Politics 58(2): 332–349. Hartley,Thomas, and Bruce Russett. 1992.“Public Opinion and the Common Defense:Who Governs Military Spending in the United States?” American Political Science Review 86(4): 905–915. Hill, Kim Quaile, and Angela Hinton-Anderson. 1995. “Pathways of Representa-

Public Opinion and Policy Making in the Culture Wars 181 tion:A Causal Analysis of Public-Opinion Policy Linkages.” American Journal of Political Science 39(4): 924–935. Hinkley, Barbara. 1983. Stability and Change in Congress. New York: Harper & Row. Holbrook,Thomas M., and Emily Van Dunk. 1993.“Electoral Competition in the American States.” American Political Science Review 87(4): 955–962. Jenness,Valerie, and Ryken Grattet. 2001. Making Hate a Crime: From Social Movement Concept to Law Enforcement Practice. New York: Russell Sage Foundation. Johnson, D.W., P. Picard, and Bernard Quinn. 1974. Churches and Church Membership in the United States.Washington, DC: Glenmary Research Center. Kane, Melinda D. 2003. “Social Movement Policy Success: Decriminalizing State Sodomy Laws, 1969–1998.” Mobilization:An International Journal 8(3): 313–334. Keen, Lisa. 2001.“By the Books: D.C. Ranks Second in Survey of States Treatment of Gays.” Washington Blade 9 November. Kirkpatrick, David D. 2004.“Conservatives Use Gay Union as Rallying Cry,” New York Times, 8 February. 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(3): 751–777. Lewis, Gregory B. 2003. “Contentious and Consensus Gay Rights Issues: Public Opinion and State Laws on Discrimination and Same-Sex Marriage.” Paper presented at the annual meeting of the American Association for Public Opinion Research, Nashville,TN, May. Lewis, Gregory B., and Jonathan L. Edelson. 2000.“DOMA and ENDA: Congress Votes on Gay Rights.” In Kenneth D. Wald, Craig A. Rimmerman, and Clyde Wilcox, eds., The Politics of Gay Rights. Chicago: University of Chicago Press. Lindaman, Kara, and Donald P. Haider-Markel. 2002. “Issue Evolution, Political Parties, and the Culture Wars.” Political Research Quarterly 55(1): 91–110. Lowry,William. 1992. The Dimensions of Federalism: State Governments and Pollution Control Policies. Durham, NC: Duke University Press. Lowery, David,Virginia Gray, and Gregory Hager. 1989.“Public Opinion and Policy Change in the American States.” American Politics Quarterly 17(1): 3–31. Mayhew, David R. 1974. Congress: The Electoral Connection. New Haven:Yale University Press. Meier, Kenneth J. 1999. “Drugs, Sex, Rock, and Roll: A Theory of Morality Politics.” Policy Studies Journal 27(4): 681–695. Melton, J. Gordon. 1991. The Churches Speak On: Homosexuality; Oﬃcial Statements from Religious Bodies and Ecumenical Organizations. Detroit: Gale Research. Monroe, Alan D. 1998. “Public Opinion and Public Policy, 1980–1993.” Public Opinion Quarterly 62(1): 6–28. Mooney, Christopher Z. 2001.“Modeling Regional Eﬀects on State Policy Diﬀusion.” Political Research Quarterly 54(1): 103–24. ———. 2000.“The Decline of Federalism and the Rise of Morality-Policy Conﬂict in the United States.” Publius (Winter-Spring): 171–188. Mooney, Christopher Z., and Mei-Hsien Lee. 1995. “Legislating Morality in the American States:The Case of Pre-Roe Abortion Regulation Reform.” American Journal of Political Science 39(3): 599–627.

182

Haider-Markel and Kaufman

Mooney, Christopher Z., and Mei-Hsien Lee. 2000. “The Inﬂuence of Values on Consensus and Contentious Morality Policy: U.S. Death Penalty Reform, 1956–82.” Journal of Politics 62(1): 223–239. Nice, David C. 1992.“The States and the Death Penalty.” Western Political Quarterly 45(4): 1037–1048. ———. 1988. “State Deregulation of Intimate Behavior.” Social Science Quarterly 69(1): 203–211. Norrander, Barbara. 2001.“Measuring State Public Opinion with the Senate National Election Study.” State Politics and Policy Quarterly 1(1): 111–125. Norrander, Barbara, and Clyde Wilcox. 1999. “Public Opinion and Policymaking in the States: The Case of Post-Roe Abortion Policy.” Policy Studies Journal 27(4): 707–722. Page, Benjamin I., and Robert Y. Shapiro. 1983.“Eﬀects of Public Opinion on Policy.” American Political Science Review 77(1): 175–190. The Pew Research Center. 2003. Republicans Uniﬁed, Democrats Split On Gay Marriage: Religious Beliefs Underpin Opposition to Homosexuality, November 18.Washington, DC:The Pew Research Center. Quinn, B., M. B. Bradley, N. M. Green, Jr., D. E. Johnson, M. Lynn, and L. McNeil. 1982. Churches and Church Membership in the United States, 1980. Atlanta: Glenmary Research Center. Roeder, Phillip W. 1994. Policy Opinion and Policy Leadership in the American States. Tuscaloosa:The University of Alabama Press. Schumaker, Paul D., and Russell W. Getter. 1977. “Responsiveness Bias in FiftyOne American Communities.” American Journal of Political Science 21(2): 247– 281. Skogan, Wesley. 1990. “Crime and Punishment.” In Virginia Gray, Herbert Jacob, and Robert Albritton, eds., Politics in the American States, 5th ed., Glenview, IL: Scott, Foresman/Little Brown. Smith, Raymond A., and Donald P. Haider-Markel. 2002. Gay and Lesbian Americans and Political Participation. Denver: ABC-CLIO Publishers. Tatalovich, Raymond, and Byron W. Daynes.1988.“Conclusion: Social Regulatory Policymaking.” In Raymond Tatalovich and Byron W. Daynes, eds., Social Regulatory Policymaking—Moral Controversies in American Politics. Boulder, CO: Westview. Thomas, Clive S., and Ronald J. Hrebenar. 2004. “Interest Groups in the States.” In Virginia Gray and Russell L. Hanson, eds., Politics in the American States, 8th ed.Washington, DC: CQ Press, pp. 100–128. Wald, Kenneth D., James W. Button, and Barbara A. Rienzo. 1996.“The Politics of Gay Rights in American Communities: Explaining Antidiscrimination Ordinances and Policies.” American Journal of Political Science 40(4): 1152–1178. Weber, Ronald E., and William R. Shafer. 1972. “Public Opinion and American State Policy-Making.” Midwest Journal of Political Science 16(4): 683–699. Yang, Alan. 2001. The 2000 National Election Study and Gay and Lesbian Rights: Support for Equality Grows.Washington, DC: National Gay and Lesbian Task Force.

Chapter

10

Citizen Inﬂuences on State Policy Priorities: The Interplay of Public Opinion and Interest Groups

Saundra K. Schneider and William G. Jacoby

Political scientists generally believe that public opinion aﬀects public policy within the American states. Although such a connection may seem obvious, ﬁrm empirical evidence for the existence of this relationship has only been presented quite recently. Early studies of state policy making emphasized economic factors as the predominant determinants of governmental activity (Dawson and Robinson, 1963; Dye, 1966; Sharkansky and Hoﬀerbert, 1969). Once variables like per capita income, gross state product, and employment levels were taken into account, political characteristics like party composition of state legislatures, gubernatorial partisanship, and political culture seemed to have no eﬀect.These results were replicated a number of times and widely accepted by state politics scholars (Hoﬀerbert, 1974). Of course, political scientists were extremely uncomfortable about the sizable body of empirical evidence that showed politics was irrelevant to governmental decision making within the states.Therefore, research continued on this topic with dogged determination to show otherwise (Erikson, 1976; Pool, Abelson, and Popkin, 1965;Weber, Hopkins, Mezey, and Munger, 1972–1973; Weber and Shafer, 1972). The major breakthrough occurred with the pioneering work by Wright, Erikson, and McIver (1985). They used survey data, aggregated to the state level, in order to develop measures of citizen partisanship and ideology within each state. They provide compelling evidence that these measures are 183

184

Schneider and Jacoby

eﬀective gauges of mass orientations, and, more important, they show that these aspects of public opinion have a strong inﬂuence on state policy liberalism. Indeed, once public opinion is taken into account, economic factors play only a secondary role as determinants of policy (Erikson,Wright, and McIver, 1989, 1993). The primacy of public opinion was immediately and joyously accepted by political scientists, and it has become the new conventional wisdom within this ﬁeld. Recent work has been aimed at extending and reﬁning the Wright et al. variables, as well as developing alternative measures of citizens’ opinions at the state level (Berry, Ringquist, Fording, and Hanson, 1998). Regardless of speciﬁc operationalizations, mass orientations have been shown to aﬀect policy across a broad array of substantive areas (Wright, Erikson, and McIver, 1987; Erikson, Wright, and McIver, 1993; Lowery, Gray, and Hager, 1989; Burstein, 2003; Gray, Lowery, Fellowes, and McAtee, 2004). We wholeheartedly agree that the recent work represents a highly signiﬁcant step to understanding the state policy-making process. But, we also believe that further work is necessary in at least two directions. First, it is important to identify where and how public opinion impinges on the policy-making process. Second, the speciﬁcation of inﬂuences on policies must be expanded to incorporate another set of critical actors in the political process—interest groups. This chapter constitutes a ﬁrst step toward those objectives.

Measuring State Policy A perusal of the literature reveals three dominant approaches to measuring public policy within the states. Many researchers use program expenditures. Another set of scholars relies on speciﬁc policy adoptions. A third group employs composite variables constructed from multiple indicators.We believe that all of these approaches are problematic, for diﬀerent reasons.

Previous Approaches First, let us consider the research that uses policy expenditures as the dependent variable. Studies that take this approach generally focus on spending (usually adjusted by state population or program recipients) within a single policy area (e.g., Barrilleaux, Holbrook, and Langer, 2002). This measurement strategy has a number of advantages: Spending data are readily available; expenditures are tangible indicators of governmental eﬀorts; the dollar ﬁgures constitute continuous, interval-level measure-

Citizen Inﬂuences on State Policy Priorities 185

ment; and spending levels are immediately comparable across states and time. Probably for these reasons, program expenditures have been used to examine policy making across a wide variety of speciﬁc areas, including welfare, health care, education, and environmental protection (Lewis and Maruna, 1999; Ringquist, 1993; Rom, 1999). Despite their ubiquity in the ﬁeld, there is a serious drawback to the use of policy expenditures. Spending within any given policy area is invariably aﬀected by state spending across other policies, which are not under immediate investigation. Unlike the federal government, the states have balanced budget requirements.Therefore, spending increases on one program are invariably oﬀset by decreases in other programs, and vice versa (Garand, 1985, 1988; Garand and Hendrick, 1991). Empirical analyses that use expenditures on a single program as the dependent variable fail to take this rather basic feature of state policy-making into account. This creates an omitted variable bias in the estimates provided by statistical models. And the severity of this bias is determined by governmental decisions with respect to other policies—those that are not included in the statistical analysis. The second approach employs program enactments as the main indicator of governmental activity. This general tradition has encompassed several speciﬁc approaches. Some studies use innovation/diﬀusion models to examine the timing and sequence of state-level program adoptions (Eyestone, 1977; Gray, 1973; Walker, 1969). Others employ event history models to predict when each state will take a concrete policy action (Berry and Berry 1990). Regardless of the speciﬁc methodology, the hallmark of this research tradition is a focus on discrete actions by state governmental decision makers, rather than the degree or level of eﬀort or the amount of resources expended toward the achievement of particular goals. The problem with program enactments is not really a matter of their validity as an indicator of policy making; rather it involves their connection to public opinion. Stated simply, most citizens do not have the knowledge, interest, or expertise to exert a direct inﬂuence on the speciﬁc actions of governmental decision makers. The empirical evidence has repeatedly and conclusively demonstrated low levels of political sophistication throughout the American electorate (e.g., Zaller, 1992). If that is the case, then it is diﬃcult to specify how citizens’ orientations would directly aﬀect elite actions. We are not questioning the general correlation between state-level opinion and policy. However, given the cognitive limitations that exist within the mass public, we believe the relationship between opinion and program enactments to be spurious and due to the impact of public opinion on other elements of the policy process.

186

Schneider and Jacoby

A third general approach is to use data reduction techniques such as factor analysis to summarize many diﬀerent policy indicators simultaneously. Although their exact results diﬀer somewhat, the studies in this tradition all demonstrate that a systematic, coherent framework underlies the ways that states address the myriad array of pressures confronting them. The early pioneering studies typically produced two separate dimensions of state policy making (Hopkins and Weber, 1976; Sharkansky and Hoﬀerbert, 1969). More recent analyses have generated unidimensional representations of state policy liberalism (Klingman and Lammers, 1984; Wright, Erikson, and McIver, 1987).The strengths of this analytic strategy have been widely recognized and the resultant measures have been employed fairly frequently in the recent literature (Hill, Leighley, and Hinton-Anderson 1995; Lascher, Hagen, and Rochlin, 1996). There are some potentially serious problems in the multiple-indicator policy variables. For one thing, they typically incorporate an ad hoc array of governmental activities (e.g., program expenditures, legislative provisions, program adoptions, tax progressivity, and so on). Hence, they may confound several diﬀerent aspects of the public policy-making process. Furthermore, the composite variables usually combine data from several time points, spanning periods from 7 to 15 years long.This is problematic because policy considerations almost certainly change over time.Any such temporal variability is lost when the data are combined into a single summary index. For these reasons, it is impossible to say exactly which, if any, speciﬁc aspects of the policy process are represented in the ﬁnal summary measures that are developed in the composite policy studies.

An Alternative Approach: Policy Priorities Because of the problems in the previous work, we believe that it is necessary to develop an alternative approach for assessing the impact of public opinion on state policy making. Speciﬁcally, we argue that the appropriate dependent variable for this purpose is state policy priorities—the relative allocation of governmental resources across all of the program areas within which states are active. Policy priorities are a clear manifestation of the institutional commitments of state governments ( Jacoby and Schneider, 2001). They operationalize the “governmental decision agendas” within the respective states (Kingdon, 1995); that is, the relative salience that state-level public oﬃcials accord to various social and political issues (Baumgartner and Jones, 1993). Policy priorities occupy a central position in the state decision-making process. They are the prime target of those who would inﬂuence government (Raimondo, 1996; Winters, 1999). In eﬀect, priorities serve as a

Citizen Inﬂuences on State Policy Priorities 187

“bridge” between public demands and governmental services.The notion of relative policy priorities is fully consistent with the idea of public opinion as a “blunt instrument,” which aﬀects the general direction of state activity without becoming involved in speciﬁc programmatic details. Therefore, we believe this is the appropriate place to look for the impact of citizens’ attitudes on public policy in the states. The problem, of course, is to measure policy priorities. In doing so, it is useful to focus on program expenditures. But, for the reasons discussed earlier, spending within a single issue area is problematic. Many researchers agree that expenditures alone cannot be used to measure policy outputs or impacts (Erikson, Wright, and McIver, 1989; Hanson, 1983; Hoﬀerbert, 1974;Walker, 1969). However, there is a scholarly consensus that spending levels across areas provide the clearest, most unambiguous, indicators of governmental commitments to address various societal problems (Garand, 1985; Garand and Hendrick, 1991; Raimondo, 1996; Ringquist and Garand, 1999). But, how can this be done within a single variable? The raw data for our measure of policy priorities will consist of yearly state general expenditures in 10 program areas: education, welfare, hospitals, health, highways, policy, corrections, parks, natural resources, and government administration. These program areas represent virtually the full range of substantive concerns that typically confront state governments.We are interested in the states’ relative priorities across the diﬀerent policy areas, that is, how the states divide up their available pools of resources. Our analysis does not seek to explain how much states spend on the diﬀerent programs.Therefore, the yearly policy-speciﬁc values within each state are expressed as proportions of the total policy expenditures for that state in that year across all 10 categories. The policy priorities measure will be based on a geometric representation, often called a “spatial proximity model,” of the state spending data. The basic idea behind this model is very simple.The 50 states and the 10 policy areas are shown as two sets of points located along a common continuum.The relative positions of the points are determined by the empirical expenditure values. Speciﬁcally, states I’s spending on policy A is inversely proportional to the distance between the point representing I and the point representing A: As spending increases, this distance gets smaller and vice versa.Thus, state points will tend to be located close to the points representing policies for which their relative spending levels are high and far from the points representing policies where their relative spending levels are low. The overall spatial proximity model for the state spending data will consist of two distinct sets of points for each year included in the analysis:

188

Schneider and Jacoby

One set of 50 state points and a second set of 10 policy points. States with similar spending proﬁles will have points that are located close to each other along the dimension; states with markedly diﬀerent spending priorities will have larger distances between their points.A similar distance rule applies to the policies themselves: Policies that receive similar proportional allocations in state budgets will be represented by points that fall close to each other along the dimension. Policies that exhibit contrasting spending patterns (i.e., high relative expenditures in one policy area coincide with small expenditures in another) will be shown as widely separated points. Many speciﬁc procedures—usually called “unfolding techniques”— have been developed to estimate the parameters of the spatial proximity model (e.g., Cox and Cox, 2000). Our analysis uses a metric, least-squares unfolding method developed by Keith Poole (1984).The overall approach is called “unfolding” because, according to the geometry of the model, a state’s proﬁle of spending values can be obtained by “folding” the unidimensional continuum at the location of the state’s point (Coombs, 1964). The scaling task is the opposite of this process: We begin with the folded versions of the dimension (i.e., the input data values) and seek to “unfold” them simultaneously across all states, in order to estimate the dimension itself (i.e., the relative positions of the state and policy points). The method is “metric” because it assumes that the input data are measured at the interval level or higher (many unfolding techniques only assume ordinal or even nominal measurement levels). The method is “least-squares” because its immediate analytic objective is to ﬁnd the set of state and policy point locations such that the squared errors between distances and data values are minimized.

Empirical Estimates of State Policy Priorities The spatial proximity model could be estimated for any years in which the relevant spending data are available. However, because of data limitations for some of the other variables in our analysis, we only use policy priorities for the years 1998, 1999, and 2000.1 During this time period, the unidimensional spatial proximity model provides an excellent representation of the state-level spending data.The squared correlation between the scaled interpoint distances (i.e., between state and policy points) and the input expenditure ﬁgures is extremely high at 0.958. Hence, the scaled array of state and policy points mirrors the empirical data almost perfectly. Figures 10.1 and 10.2 provide graphical representations of the scaled point locations.2 Figure 10.1 shows a dot plot of the policy points, while Figure 10.2 shows a similar display for the state points.The point locations exhibit virtually no movement whatsoever during this time period.

Citizen Inﬂuences on State Policy Priorities 189

Natl. Res. Highways Education

State

Policy area

Welfare Govt. Admin. Health Care Corrections Hospitals Police Parks

-0.5

0.0

0.5

1.0

1.5

Mean policy score, 1998-2000

WY KS OK NV ID ND UT IN MT WI IA NM SD MI VT GA NC AR CO AK WA WV NE VA AZ MN TX AL MS MO DE KY FL OH OR LA NJ HA SC TN MD CA IL NH PA ME MA RI CT NY 0.50

0.52

0.54

Mean state score, 1998-2000

Figure 10.1. (left) Mean Policy Score, 1998–2000 Figure 10.2. (right) Mean State Score, 1998–2000

Therefore, the graphical displays only show the mean locations (across the 3 years) for each policy and state point, respectively.The information in the dot plots is easily interpreted ( Jacoby 1997).To ﬁnd the scale value for any particular policy (or state), one merely locates the policy (or state) along the vertical axis, scans horizontally along the dotted line to the point location within the display, and then scans vertically down to the scale value located along the horizontal axis. In this manner, the two dot plots completely summarize all of the information provided by the scaling analysis. Examination of the dot plot in Figure 10.1 shows that most of the 10 policy areas fall within two strongly contrasting groups:The scaled points near the left end of the continuum are designed to deal with the needs and demands of speciﬁc state constituencies, ranging from the neediest strata (the poor, who beneﬁt from welfare and health care) to distinctive occupational groups (e.g., police oﬃcers and governmental employees who beneﬁt from law enforcement and administrative expenditures,

190

Schneider and Jacoby

respectively). The points located at the other side of the continuum represent policies that ostensibly beneﬁt all of society rather than narrowly deﬁned segments of the population—education, highways, and natural resources. Thus, the empirical evidence shows that state policy priorities range along a dimension from particularized beneﬁts to collective goods ( Jacoby and Schneider, 2001). Note that there is one potentially anomalous placement:The point representing spending on parks is located among particularized beneﬁts rather than collective goods. We have no ready explanation for this. But, we do not believe that it compromises the basic bipolar nature of the priorities continuum. It is simply the case that states with higher levels of spending for parks and recreational services also place greater emphasis on constituency-speciﬁc beneﬁts. Figure 10.2 shows the dot plot for the states. Note that the state points are located along the same dimension as the policy points.Therefore, the points that fall near the right side of the plot represent states that spend more on education, highways, and natural resources—in other words, collective goods. The points closer to the left side of the plot are states that devote larger portions of their budgets to particularized beneﬁts (e.g., welfare, health care, law enforcement, and so on). Points near the middle of the plotting region correspond to states that balance their spending more evenly across the two contrasting sets of policy areas.Thus, there is deﬁnitely variability in state policy priorities. And, the observed diﬀerences conform to a systematic, readily interpretable contrast between two types of governmental commitments: States that spend more on particularized beneﬁts inevitably spend less on collective goods, and vice versa. The main task for the remainder of the chapter is to explain this observed variability in state policy priorities. Speciﬁcally, we will try to determine the role of state public opinion as an inﬂuence on patterns of state program expenditures.

Initial Results: Bivariate Relationships The ﬁrst step in assessing the impact of citizens’ attitudes on policy priorities is to operationalize public opinion. Within the ﬁeld of state politics, researchers have taken several diﬀerent approaches over the years. These can be roughly divided into three sets: political culture/regional measures, survey-based measures, and estimates obtained from elections and oﬃce-holder characteristics. Let us consider brieﬂy the bivariate relationships between each of these public opinion measures and the state policy priorities variable.

Citizen Inﬂuences on State Policy Priorities 191

Political Culture and Region Scholarly conceptions of state political culture have been heavily inﬂuenced by the work of Daniel Elazar (1966, 1984). He divided states into three groups based on the predominant ethnic and religious values of the earliest settlers within each state’s borders. In moralist states, government plays an active role in promoting the common good. Individualist states emphasize the marketplace with government policies as a means of facilitating individual rewards. Traditionalist states have relatively elitist cultures aimed at maintaining existing social and economic strata. Sharkansky (1967) later reﬁned Elazar’s original categorical distinctions. He argues that cultural diﬀerences form a continuum ranging from moralist to traditionalist orientations, with individualistic cultures falling in between the two extremes. Figures 10.3 and 10.4 show the empirical relationships between the two related measures of political culture and the policy priorities variable. Consider the box plots in Figure 10.3. Individualistic states show extremely wide variation in policy priorities; the scores assigned to these states cover almost the full range of the policy priorities scale. The range of priority scores for traditionalist states falls almost completely within that for moralistic states. Both of the latter two categories overlap the upper half of the distribution for individualistic states.Thus, Elazar’s political culture categories do not diﬀerentiate the states according to their policy priorities. Figure 10.4 shows the scatterplot of state policy priorities versus Sharkansky’s political culture scores, along with an OLS regression line ﬁtted to the data. Stated simply, there is no relationship between the two variables.The cloud of points in the plot shows no discernible pattern, and the slope of the regression line is almost perfectly ﬂat. In contrast to subjective and possibly time-bound conceptions of political culture, other analysts simply use regional classiﬁcations as a crude indicator of diﬀerences in predominant opinions within states (Erikson, Wright, and McIver, 1993).Therefore, Figure 10.5 shows the distribution of state policy priorities within the four regions of the United States. One feature is immediately apparent. Northeastern states fall near the lower end of the priorities scale, indicating that they place a greater emphasis on particularized beneﬁts than do the other states. Beyond this, however, there are no clear patterns of regional distinctiveness in policy priorities. Taken together, these results show that state political culture, at least as it is conceptualized by Elazar and his followers, has no impact on policy priorities—nor do regional diﬀerences provide much leverage beyond the general distinctiveness of Northeastern states. It is important to emphasize

100

State policy priority scores

80

60

40

20

0

Indiv

Moral

Trad

Elazar political culture categories

Figure 10.3. State Policy Priorities and Elazar Political Culture Categories

100

State policy priorities

80

60

40

20

0 2

4

6

8

Sharkansky political culture scores

Figure 10.4. State Policy Priorities and Sharkansky Political Culture Scores

Citizen Inﬂuences on State Policy Priorities 193

100

State policy priorities

80

60

40

20

0

Midwest

Northeast

South

West

Region

Figure 10.5. State Policy Priorities and Region

that region and culture are more or less permanent characteristics of the states. Therefore, by deﬁnition, they cannot be used to account for any temporal variability in state policy priorities. Admittedly, the states do not change much during the short time period covered by our investigation. But, this problem still highlights the fact we need more direct measures of public opinion. And, it is to those that we now turn.

Survey-Based Measures The recent, rapid expansion of the survey research industry has enabled more direct measurement of citizen orientations within the states. The most prominent work in this area has been carried out by Wright, Erikson, and McIver (1985); also see Erikson,Wright, and McIver (1993).They use information drawn from state-level CBS News/New York Times surveys to obtain yearly readings of mass partisanship and ideology within the states. For each year, within each state, the former is deﬁned as the proportion of survey respondents who identify themselves as Republicans minus the proportion that call themselves Democrats.3 Similarly, ideology is deﬁned as the proportion who call themselves conservative minus the proportion that call themselves liberal. For our analysis, we use the 1998 through 2000 values of the state electorate partisanship and ideology variables. Figures 10.6 and 10.7 show the scatterplots of policy priorities against

100

State policy priorities

80

60

40

20

0 -0.3

-0.2

-0.1

0.0

0.1

0.2

0.3

Electorate partisanship

Figure 10.6. State Policy Priorities and Electoral Partisanship

100

State policy priorities

80

60

40

20

0 -0.1

0.0

0.1

0.2

0.3

0.4

Electorate ideology

Figure 10.7. State Policy Priorities and Electoral Ideology

Citizen Inﬂuences on State Policy Priorities 195

mass partisanship and ideology, respectively. The scatterplots also include OLS regression lines ﬁtted to the respective bivariate data sets. The evidence shows that these variables are deﬁnitely related to each other in a direction that corresponds to reasonable prior expectations. States with more Republican identiﬁers or conservative electorates exhibit higher priority scores, meaning that they place relatively a strong emphasis on collective goods. In contrast, states with more Democratic identiﬁers or more liberal citizens show low policy priority scores, indicating that they focus more directly on beneﬁts for particular groups. This is exactly what we would expect given the two parties’ platforms and the issue stands associated with liberal and conservative ideologies.Thus, in contrast to the results for political culture, these displays conﬁrm that public opinion is clearly related to the allocation of governmental expenditures within the states.

A Measure Based on Officeholder Characteristics Erikson et al. (1993) measure citizen ideology using individual respondents’ self-descriptions. However, this may be problematic because a massive amount of evidence suggests that sizable segments of the mass public do not deal very eﬀectively with abstract, ideological terms, like “liberal” and “conservative” (Converse, 1964; Jacoby, 1995). If that is the case, then it is not entirely clear what ideological identiﬁcations are really measuring. And, for that reason, we turn to a second measure of state-level citizen ideology. Berry, Ringquist, Fording, and Hanson (1998) use interest group ratings of, and popular votes received by, congressional incumbents and challengers, aggregated within each state and year. Their variable is coded so that larger values correspond to more conservative state electorates.4 Figure 10.8 shows the relationship between the Berry et al. (1998) ideology measure and the policy priorities scale. Just as with the Wright et al. (1985, 1987) measures, there is a clear relationship between the two variables. States with liberal electorates emphasize particularized beneﬁts, while conservative states focus on collective goods. What is distinctive here is that this relationship continues to exist with a behaviorally based ideology measure. Although the results are consistent, we believe that this represents a diﬀerent phenomenon than the identiﬁcation-based measure used by Erikson et al. (1993). Erikson et al. (1993) and Berry et al. (1998) make an important contribution by providing relatively direct measures of state public opinion. When operationalized in those ways, public opinion is clearly related to state policy priorities. However, the wide spread of points around the regression lines in the scatterplots for these variables also shows that the rela-

196

Schneider and Jacoby

100

State policy priorities

80

60

40

20

0 20

40

60

80

Citizen ideology

Figure 10.8. State Policy Priorities and Citizen Ideology

tionships are far from deterministic in nature.The correlation (Pearson’s r ) between state partisanship and policy priorities is only 0.413, while that for the Erikson et al.(1993) mass ideology measure is 0.458.The correlation for the Berry et al. (1998) citizen ideology variable is somewhat larger, but still only moderate in size, at 0.570.Thus, we argue that it is important to consider additional factors in order to obtain a more complete depiction of the inﬂuences on state policy priorities.

The Role of Interest Groups Along with public opinion, interest groups are universally believed to be an inﬂuence on governmental decision making.They are central ﬁgures in modern empirical democratic theory. For example, pluralist conceptions of power maintain that politics involves ongoing competition among groups (Dahl, 1961;Truman, 1951). Modern societies comprise many distinct interests, and the resultant existence of multiple groups guarantees the need for bargaining, accommodation, and compromise in the policymaking process. From this relatively benevolent perspective, interest

Citizen Inﬂuences on State Policy Priorities 197

groups serve as communication channels and focal points for distinctive points of view within the political system. A less positive interpretation holds that interest groups subvert the policy process to their own ends. Critical theories of democratic elitism contend that groups promote existing inequalities and stiﬂe the preferences of ordinary citizens (Schattschneider, 1960). A related perspective holds that powerful groups control the governmental agenda, thereby preventing important issues from ever coming up for consideration (Bachrach and Baratz, 1962).The role of interest groups is particularly insidious because “iron triangles” develop into relatively permanent linkages between private concerns and governmental decision makers (Lowi, 1964). Once again, the needs and interests of the general public are simply pushed aside in the face of the pressure groups and lobbyists who are more directly and constantly involved in the process. Regardless whether one takes a positive or negative view of interest groups, it is simply impossible to deny their ubiquity in public policy making. Given this fact, it is surprising that interest groups have seldom been incorporated into analyses of public opinion and the policy process. This is a serious deﬁciency because the resultant underspeciﬁed models probably distort the eﬀects of the hypothesized inﬂuences on governmental activity. Speciﬁcally, the measures of public opinion that are included in the models may be picking up some of the eﬀects that should be attributed to the interest groups that are omitted from the same models. At least part of the reason for this problem stems from the lack of data pertaining to state-level interest groups. However, this situation has been addressed by Virginia Gray, David Lowery, and their colleagues (2004). These researchers have collected information on the interest group communities within the states.Their work shows that state lobbying environments aﬀect a number of important phenomena, including the sizes of state governments and intrastate economic growth. In earlier work, we have shown that interest group concentrations aﬀected state policy priorities in the early 1990s ( Jacoby and Schneider, 2001).Thus, ample empirical evidence backs up beliefs about the importance of interest groups. For present purposes, we use some additional data recently collected and made available to us by Gray and Lowery (1996, 1998).They identify organized groups registered to lobby within each state and categorize those groups into 26 categories based on the substantive focus of their lobbying eﬀorts. We focus on the subsets of those categories that pertain to particularized beneﬁts and collective goods. The former include civil rights, government employees, health care, insurance, the legal profession, police and ﬁre protection, welfare, and women’s groups. Collective goods

100

State policy priorities

80

60

40

20

0 20

25

30

35

40

Collective goods interest groups

Figure 10.9. State Policy Priorities and Collective Goods Interest Groups 100

State policy priorities

80

60

40

20

0 25

30

35

40

Particularized benefit interest groups

Figure 10.10. State Policy Priorities and Particularized Beneﬁt Interest Groups

Citizen Inﬂuences on State Policy Priorities 199

include agriculture, communications, education, environmental groups, good government groups, natural resources, transportation, and utilities. We use this information to create two variables:The proportion of registered lobbying groups within a state devoted to particularized and collective goods, respectively. This information is available on a yearly basis for 1997 through 1999; however, the states’ values do not change at all within this time period. We do recognize that the proportion of groups active on speciﬁc kinds of issues does not translate directly into political power. Nevertheless, larger numbers of groups within a substantive area will increase the opportunity for the articulation of those interests, and, hence, the degree to which governmental decision makers are exposed to particular points of view.Therefore, we believe that these variables can function as reasonable indicators of interest group pressure on the policy-making process— the degree to which they are trying to move the states toward one of the two ends of the policy priorities continuum. Figures 10.9 and 10.10 show the relationship between the two interest group variables and state policy priorities.The results show that the concentration of interest groups in a particular substantive area corresponds to larger governmental expenditures on the corresponding policies. The connection is stronger for collective goods than for particularized beneﬁts: The bivariate correlations are 0.503 and –0.243, respectively. States with larger proportions of their interest group population focused on collective goods also devote greater portions of their overall resources to those kinds of policies.A similar relationship holds for groups and policies in the particularized beneﬁts area. This provides clear preliminary evidence that interest groups should, in fact, be incorporated into any model representing the determinants of state policy priorities.

Combined Eﬀects of Public Opinion and Interest The discussion so far has looked at the eﬀects of separate variables on state policy priorities. In order to obtain a relatively accurate assessment of their inﬂuence, however, it is necessary to combine these variables into a single, more comprehensive, model. Speciﬁcally, we will use the policy priorities variable as the dependent variable, expressed as a function of measures representing three types of inﬂuences. Two variables represent public opinion: Wright, Erikson, and McIver’s (1985) mass partisanship variable and Berry et al.’s (1998) citizen ideology indicator.5 Second, interest groups are represented by the two variables discussed earlier:The percentages of groups within each state devoted to particularized beneﬁts

200

Schneider and Jacoby

and collective goods, respectively. Third, region is operationalized as dummy variables representing northeastern, southern, and western states. Midwestern states are omitted as the reference category. We include the latter variables primarily as controls to pick up the residual eﬀects of socioeconomic diﬀerences that exist across the states. The data constitute measures on all 50 states across a 3-year period.The dependent variable values cover 1998 through 2000; the independent variables are lagged one year to provide an unambiguous causal direction for their inﬂuences. Given the cross-sectional, time-series nature of the data set, the model’s parameters are estimated using panel-corrected standard errors and allowing for ﬁrst-order autoregressive patterns in the residuals.

Direct Influences on State Policy Priorities Table 10.1 shows the Prais-Winston estimates from the regression analysis. The overall model ﬁt is quite good, with an R2 of 0.776. Notice also that the autoregression parameter is very large (rho = 0.679) indicating that there is a systematic linear pattern among the disturbances. This is not surprising at all, given the incremental nature of state policy making. Policy priorities do not change very much over time. So, the inertial pattern that apparently exists among the residuals is to be expected. Turning to the separate independent variables, public opinion does inﬂuence policy priorities, but the magnitude of its eﬀect is rather tepid at best.The coeﬃcient for mass partisanship is barely signiﬁcant at the 0.05 level with a directional test. But, using the same decision criterion for the hypothesis test, the coeﬃcient for citizen ideology does not achieve statistical signiﬁcance. Thus, Democratic electorates move a state toward more spending on particularized beneﬁts while a Republican citizenry leans in the other direction, toward spending on collective goods. Citizens’ liberal-conservative orientations have no such eﬀect. In contrast to the public opinion variables, the two interest group measures are both statistically signiﬁcant in the expected directions. Exactly as anticipated, a higher concentration of lobbyists working for collective goods moves the allocation of state resources toward those kinds of policies. And, of course, a larger percentage of lobbying groups aimed at particularized beneﬁts has the opposite eﬀect—leading states to spend more money on policies like welfare, health care, and law enforcement (i.e., those that fall on the particularized beneﬁts side of the policy priorities continuum). Comparing these two types of groups, the impact of collective goods lobbying is much more pronounced than that for particularized beneﬁts: The absolute value of the coeﬃcient for the former is

Citizen Inﬂuences on State Policy Priorities 201 Table 10.1 The Impact of Public Opinion, Interest Groups, and Region on State Policy Priorities

Public Opinion: Electorate partisanship Citizen ideology Interest Groups: Groups working on collective goods Groups working on particularized benefits Region: Northeastern states Southern states Western states Intercept R2 Rho (autoregression)

Prais-Winston Coefficient Estimates

Panel-Corrected Standard Errors

8.510* 0.136

5.132 0.126

1.161*

0.244

–0.245*

0.057

–28.136* –4.905* –5.929* 50.058 0.776 0.679

4.921 1.757 0.952

*Coefficient is statistically different from zero at the 0.05 level. One-sided tests are used for the public opinion and interest group variables.Two-sided tests are used for the region variables. Note: The dependent variable consists of the policy priority score for each state, for 1998, 1999, and 2000.The public opinion and interest group variables are lagged one year.The total number of observations is 150.

more than four times larger than that for the latter, and the diﬀerence is statistically signiﬁcant. Finally, the three dummy variables for region all have negative, statistically signiﬁcant coeﬃcients.This shows that midwestern states (i.e., those in the omitted reference category) place a higher priority on collective goods than do states in the rest of the nation.The diﬀerence is particularly pronounced when it comes to northeastern states:The coeﬃcient for the latter is more than ﬁve times larger than that for southern states and over four times larger than that for western states. Clearly, particularized beneﬁts are a more important governmental priority in the northeast than in anyplace else across the country.

The Indirect Effect of Public Opinion The results so far suggest that interest groups overwhelm public opinion in the determination of state policy commitments. However, we suspect

202

Schneider and Jacoby

that such a conclusion might be a bit premature. A number of theories assert that interest groups provide voice and channels of access for citizens within society. For example, the social capital literature shows that participation in formal organizations enhances democratic activity within a political system (Putnam, 2000). Another line of research focuses on increasing public involvement in issues, much of which is channeled through policyoriented groups (Heclo, 1978). Finally, the general theory of pluralism holds that interest groups are the main linkages between citizens and government (Dahl 1961; Truman, 1951). According to all of these theoretical perspectives, public opinion may inﬂuence the group environment within a state, and thereby exert an indirect eﬀect on policy priorities. In order to test this possibility, we regress the percentages of interest groups within a state devoted to collective goods and to particularized beneﬁts on the public opinion and region variables used above. The results are presented Table 10.2. Once again, we use Prais-Winston estimates with panel-correlated standard errors because the yearly information on the common pool of states yields nonindependent observations within the data set. The two separate regression models both ﬁt the data very well, and, just as before, there is clear evidence of temporal patterning in the disturbances. But, for present purposes, the most important results in the table involve the coeﬃcients on the public opinion variables. In the equation for collective goods, citizen ideology and partisanship both have signiﬁcant eﬀects. The direction of inﬂuence is consistent in each case, with more conservative and Republican states corresponding to higher percentages of groups devoted to collective goods. In contrast, the coeﬃcients for these same variables in the particularized beneﬁts equation are very small and not signiﬁcantly diﬀerent from zero. This result suggests that mass orientations have no discernible eﬀect on the formation of groups that address narrow concerns within their respective states. Finally, interest group concentrations vary quite a bit across regions. However, geographic diﬀerences are not the main focus of this analysis, and the current data provide no explanation for why they exist in the ﬁrst place.Therefore, we merely present the results without further comment. Our analysis shows that while interest groups circumscribe the impact of public opinion on policy priorities they certainly do not eliminate it. In fact, the empirical results explicate our understanding of mass inﬂuence by showing that public opinion has the greatest eﬀect when it is channeled through groups. So, for example, electorate partisanship has a signiﬁcant impact. This is reasonable because political parties are the main points of reference connecting citizens and the political system (Campbell, Converse,

Citizen Inﬂuences on State Policy Priorities 203 Table 10.2 The Impact of Public Opinion and Region on the Proportion of Interest Groups Within a State Working on Collective Goods and Particularized Benefits Dependent Variable

Public Opinion: Electorate partisanship Citizen ideology Region: Northeastern states Southern states Western states Intercept R2 Rho (autoregression)

Collective Goods

Particularized Benefits

2.810* (1.629) 0.102* (0.028)

–1.726 (1.648) –0.002 (0.011)

–0.432 (0.662) –2.351* (0.179) 4.086* (0.218)

0.197 (0.113) 0.817* (0.272) –3.073* (0.076)

31.587 0.919 0.745

33.233 0.967 0.804

* Coeﬃcient is statistically different from zero, 0.05 level. Onesided tests are used for the public opinion variables, and two-sided tests for the region variables. Note: The dependent variable for each regression consists of the proportion of interest groups within a state (i.e., groups registered to lobby in that state) working for either collective goods or particularized benefits. Data cover the years 1997–1999, so the total number of observations is 150.

Miller, and Stokes, 1960). In contrast, citizen ideology involves more abstract and nebulous concepts; therefore, it is much more diﬃcult for citizens to make the appropriate connections (Converse, 1964; Jacoby, 2002).Viewed in this light, the absence of a direct ideological inﬂuence on policy priorities is understandable. At the same time, both elements of public opinion, ideology and partisanship, aﬀect the formation of groups in the ﬁrst place. Interestingly, this relationship only exists among groups aimed at broad policy objectives. Thus, public inﬂuence is oriented toward organizations working on issues that have far-reaching and highly visible consequences, such as education and highways. Particularized beneﬁts, by deﬁnition, channel the rewards of government toward fairly narrow targets and these seem to be largely outside the domain of public concern.

204

Schneider and Jacoby

Conclusions In this chapter we have attempted to reﬁne the scholarly understanding of the role of public opinion in state politics. We believe our analysis makes two speciﬁc contributions. First, we have identiﬁed a likely access point where external actors—public opinion and interest groups—gain entry into the governmental system. Policy priorities serve this function very nicely because they represent the relative allocation of resources, but bypass speciﬁc details of governmental programs and administration.This is precisely where we would expect citizens’ voices to have the loudest resonance, and it is certainly a point in the policy-making process that self-motivated groups would try to inﬂuence. After all, nothing can be done unless the resources are provided to address speciﬁc conditions and social problems. The second contribution is to embed public opinion within a broader model of policy making than has previously been employed in the state politics literature. Recent empirical analyses of the relationship between public opinion and state policy have largely ignored the presence and possible eﬀects of interest groups (Gray et al., 2004, constitute a prominent exception to this generalization). This broad omission of a class of explanatory factors is particularly surprising given that interest groups are a central feature in all modern theories of political power. In conclusion, our analysis demonstrates that public opinion and interest groups are both factors aﬀecting policy priorities in the American states. One interpretation of these results would be optimism about popular inﬂuence over government. But, we believe some caution is still necessary. On the one hand, we suspect that policy priorities constitute the main point of access for the mass public, at least under most circumstances. On the other hand, lobbying activities permeate the entire governmental process. Therefore, further research is deﬁnitely necessary to disentangle the roles of public opinion and interest groups and to assess their respective net eﬀects on policy making at the state level.

Notes 1. The full policy priorities data set covers the time period from 1982 through 2000. The scaling approach that we use to estimate the parameters of the spatial proximity model provides separate sets of policy and state points for each year.The point locations are estimated separately within each year. But the estimates are obtained from an iterative process that uses the preceding year’s estimates as a starting point. This is fully consistent with the substantive nature of state decision-

Citizen Inﬂuences on State Policy Priorities 205 making processes, wherein the previous year’s budget is used as the baseline for spending allocations in any given year. 2. The actual yearly scale values for policies and states are available from the authors. 3. Note that this coding is reversed from the original scheme.Wright and his colleagues subtracted the proportion of Republicans (or conservatives) from the proportion of Democrats (or liberals) within each state.We reversed this simply to make the correlations with the policy priorities variable positive rather than negative. 4. Once again, this is a reversal of the original coding used by Berry and his colleagues, which gives higher scores to more liberal states. 5. We also tried an alternative speciﬁcation, by including the Wright, Erikson, and McIver (1985) electorate ideology measure in the equation (both along with, and as a replacement for, the Berry et al. [1998] variable). However, after controlling for the other independent variables, it is completely unrelated to policy priorities.Therefore, we dropped it from the model that we report here.

References Bachrach, Peter, and Morton Baratz. 1962. “Two Faces of Power.” American Political Science Review 56(4): 947–952. Barrilleaux, Charles,Thomas Holbrook, and Laura Langer. 2002.“Electoral Competition, Legislative Balance, and American State Welfare Policy.” American Journal of Political Science 46(2): 415–427. Baumgartner, Frank R., and Bryan D. Jones. 1993. Agendas and Instability in American Politics. Chicago: University of Chicago Press. Berry, Frances, and William D. Berry. 1990.“State Lottery Adoptions as Policy Innovations: An Event History Analysis.” American Political Science Review 84(2): 395–415. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 1960–93.” American Journal of Political Science 42(1): 327–348. Burstein, Paul. 2003. “The Impact of Public Opinion on Public Policy: A Review and an Agenda.” Political Research Quarterly 56(1): 29–40. Campbell, Angus, Philip Converse, Warren Miller, and Donald Stokes. 1960. The American Voter. New York:Wiley. Coombs, Clyde H. 1964. A Theory of Data. New York:Wiley. Converse, Philip E. 1964. “The Nature of Belief Systems in Mass Publics.” In David E. Apter, ed., Ideology and Discontent. New York:The Free Press 1964, pp. 206–261. Cox, Trevor, and Michael A. A. Cox. 2000. Multidimensional Scaling, 2nd ed. New York: Chapman and Hall/CRC. Dahl, Robert A. 1961. Who Governs? New Haven:Yale University Press. Dawson, Richard E., and James A Robinson. 1963.“Inter-Party Competition, Economic Variables, and Welfare Policies in the American States.” Journal of Politics 25(2): 265–289.

206

Schneider and Jacoby

Dye, Thomas. 1966. Politics, Economics, and the Public: Policy Outcomes in the American States. Chicago: Rand-McNally. Elazar, Daniel. 1966. American Federalism:A View from the States. New York:Thomas Y. Crowell. ———. 1984. American Federalism: A View from the States. 3rd ed. New York: Harper and Row. Erikson, Robert S. 1976. “The Relationship Between Public Opinion and State Policy: A New Look Based on Some Forgotten Data.” American Journal of Political Science 20(1): 25–36. Erikson, Robert S., Gerald C.Wright, Jr., and John P. McIver. 1989.“Political Parties, Public Opinion, and State Policy in the United States.” American Political Science Review 83(3): 729–750. ———. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Eyestone, Robert. 1977.“Confusion, Diﬀusion, and Innovation.” American Political Science Review 71(2): 441–447. Garand, James C. 1985. “Partisan Change and Shifting Expenditure Priorities in the American States, 1945–1978.” American Politics Quarterly 14(4): 355–391. ———. 1988. “Explaining Government Growth in the U.S. States.” American Political Science Review 82(3): 837–849. Garand, James C., and Rebecca M. Hendrick. 1991.“Expenditure Tradeoﬀs in the American States: A Longitudinal Test, 1948–1984.” Western Political Quarterly 44(4): 915–940. Gray,Virginia. 1973. “Innovation in the States: A Diﬀusion Study.” American Political Science Review 67(4): 1174–1185. Gray, Virginia, and David Lowery. 1988. “Interest Group Politics and Economic Growth in the U.S. States.” American Political Science Review 82(1): 109–131. ———. 1996. The Population Ecology of Interest Representation: Lobbying Communities in the American States. Ann Arbor:The University of Michigan Press. Gray,Virginia, David Lowery, Matthew Fellowes, and Andrea McAtee. 2004.“Public Opinion, Public Policy, and Organized Interests in the American States.” Political Research Quarterly. 57 (September): 411-420. Hanson, Russell L. 1983. “The ‘Content’ of Welfare Policy: The States and Aid to Families with Dependent Children.” Journal of Politics 45(3): 771–785. Heclo, Hugh. 1978. “Issue Networks and the Executive Establishment.” In Anthony King, ed., The New American Political System.Washington, DC: American Enterprise Institute. Hill, Jim Q., Jan Leighley, and Angela Hinton-Anderson. 1995.“Lower-Class Mobilization and Policy Linkage in the U.S. States.” American Journal of Political Science 39(1): 75–86. Hoﬀerbert, Richard I. 1974. The Study of Public Policy. New York: Bobbs-Merrill. Hopkins, Anne H., and Ronald E.Weber. 1976.“Dimensions of Public Policies in the American States.” Polity 8(3): 475–489. Jacoby,William G. 1995.“The Structure of Ideological Thinking in the American Electorate.” American Journal of Political Science 39(2): 314–335.

Citizen Inﬂuences on State Policy Priorities 207 ———. 1997. Statistical Graphics for Univariate and Bivariate Data. Beverly Hills, CA: Sage. ———. 2002. “Core Values and Political Attitudes.” In Barbara Norrander and Clyde Wilcox, eds., Understanding Public Opinion, 2nd ed.Washington, DC: CQ Press, pp. 177–202. Jacoby, William G., and Saundra K. Schneider. 2001. “Variability in State Policy Priorities: An Empirical Analysis.” Journal of Politics 63(2): 544–548. Kingdon, John W. 1995. Agendas,Alternatives, and Public Policies, 2nd ed. Boston: Little, Brown. Klingman, David, and William W. Lammers. 1984. “The ‘General Policy Liberalism’ Factor in American State Politics.” American Journal of Political Science 28(3): 598–610. Lascher, Edward L., Jr., Michael G. Hagen, and Steven A. Rochlin. 1996.“Gun Behind the Door? Ballot Initiatives, State Policies and Public Opinion.” Journal of Politics 58(3): 760–775. Lewis, Dan A., and Shadd Maruna. 1999. “The Politics of Education.” In Virginia Gray, Russell L. Hanson, and Herbert Jacob, eds., Politics in the American States, 7th ed.Washington, DC: CQ Press. Lowery, David,Virginia Gray, and Gregory Hager. 1989.“Public Opinion and Policy Change in the American States.” American Politics Quarterly 17(1): 3–31. Lowi,Theodore. 1964.“American Business, Public Policy, Case Studies and Political Theory.” World Politics 16(4): 677–715. Pool, Ithiel de Sola, Robert P.Abelson, and Samuel L. Popkin. 1965. Candidates, Issues, and Strategies. Cambridge, MA: MIT Press. Poole, Keith T. 1984. “Least Squares, Metric, Unidimensional Unfolding.” Psychometrika 49(3): 311–323. Putnam, Robert D. 2000. Bowling Alone:The Collapse and Revival of American Community. New York: Simon and Schuster. Raimondo, Henry J. 1996. “State Budgeting: Problems, Choices, and Money.” In Carl E. Van Horn, ed., The State of The States, 3rd ed. Washington, DC: CQ Press. Ringquist, Evan. J. 1993. Environmental Protection at the State Level. Armonk, NY: M. E. Sharpe. Ringquist, Evan J., and James C. Garand. 1999. “Policy Change in the American States.” In Ronald E.Weber and Paul Brace, eds., American State and Local Politics: Directions for the 21st Century. New York: Chatham House. Rom, Mark. 1999.“Transforming State Health and Welfare Programs.” In Virginia Gray, Russell Hanson, and Herbert Jacob, eds., Politics in the American States, 7th ed.Washington, DC: CQ Press. Schattschneider, E. E. 1960. The Semisovereign People. New York: Holt, Rinehart and Winston. Sharkansky, Ira. 1967. “Government Expenditures and Public Policies in the American States.” American Political Science Review 61(4): 1066–1077. Sharkansky, Ira, and Richard I. Hoﬀerbert. 1969. “Dimensions of State Politics, Economics, and Public Policy.” American Political Science Review 63(3): 867–879.

208

Schneider and Jacoby

Truman, David. 1951. The Governmental Process. New York: Alfred A. Knopf. Walker, Jack. 1969.“Innovation in the States: A Diﬀusion Study.” American Political Science Review 67(4): 1174–1185. Weber, Ronald E.,Anne H. Hopkins, Michael L. Mezey, and Frank Munger. 1972– 1973. “Computer Simulation of State Electorates.” Public Opinion Quarterly 36(1): 49–65. Weber, Ronald E., and William R. Shafer. 1972. “Public Opinion and American State Policy-Making.” Midwest Journal of Political Science 16(4): 683–699. Winters, Richard F. 1991.“The Politics of Taxing and Spending.” In Virginia Gray, Russell L. Hanson, and Herbert Jacob, eds., Politics In the American States.Washington, DC: CQ Press, pp. 304–348. Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship and Ideology with Survey Data.” Journal of Politics 47(4): 469– 489. ———. 1987.“Public Opinion and Policy Liberalism in the American States.” American Journal of Political Science 31(3): 980–1007. Zaller, John R. 1992. The Nature and Origins of Mass Opinion. New York: Cambridge University Press.

Chapter

11

State-Level Opinions from National Surveys: Poststratiﬁcation Using Multilevel Logistic Regression

David K. Park, Andrew Gelman, Joseph Bafumi

1. Introduction One of the ﬁrst projects to simulate state-level opinions using national data was undertaken by Pool, Abelson, and Popkin (1965). When the three MIT scholars began their work, computer simulation was only about 20 years old. It was primarily used by engineers who tackled military problems, bridge designs, ﬂight characteristics of new aircraft, and work assignment rules in factories.1 Pool et al. aggregated 64 survey data sets of national respondents from 1952 to 1960.They then used poll, voting, and census data to design 480 voter types based on such factors as income, religion, party, population density, region, race, and sex.2 Diﬀerences across states were not attributed to state-level factors but to difﬀerences in the proportion of voter types. With this assumption and the data in hand, they determined the percent of each voter type who held an opinion of interest, weighted it according to the number of each type in each state, and aggregated to state-level results. Their “best-ﬁt” simulation did quite well with respect to vote choice in 1960.Their estimates diﬀered by only 2.5 percent (in median state error) from the stateby-state election results. Weber, Hopkins, Mezey, and Munger (1972–1973) undertook a similar project. Their paper took issue with research that emphasized socioeconomic variables as determinates of policy output in the American states. Weber et al. attributed those results to invalid measures of state opinion and instead proposed to estimate state-level opinion in much the 209

210

Pa r k , G e l m a n , a n d Ba f u m i

same way as Pool et al. (1965). But considering a wider range of models and settling on a method including 960 categories of voter types, they were able to estimate and report state-level opinions on favorability toward the death penalty and toward teacher unionization.They note,“The voter-type approach to creating synthetic electorates could also be employed to calculate such electorates for congressional districts, metropolitan areas, counties or cities” (Weber et al., 1972–1973, 565). In a followup article, Weber and Shafer (1972) tested for congruence between state-level opinions and state-policy outputs.They found opinion to be a more important determinant of policy than other factors. Erikson (1976) provided further evidence in support of these conclusions. He cleverly used forgotten large sample opinion data from the 1930s and showed that state policy reﬂected the opinions of their populace for capital punishment, child labor, and female juror laws. Wright, Erikson, and McIver (1987) and Erikson,Wright, and McIver (1993) extended this analysis to generate state-level measures of partisanship and ideology and tested for representation across policy outcomes that fall on the left/right continuum. Their study was in part prompted by the deﬁciencies they found in Weber et al. (1972–1973) and Weber and Shafer (1972). Erikson et al. (1993) found Weber et al.’s method to inadequately capture variation in opinion across states (apart from socioeconomic factors) because Weber et al. deﬁned voter types largely based on socioeconomic demographic variables. They showed high congruence between state policy and state opinion even in the face of socioeconomic variables that had been deemed such powerful predictors. Their partisanship and ideology scores have become standard predictors in many state-level regression models (i.e., Hill and Leighley, 1992; Hill, Leighley, and Hinton-Anderson, 1995). Brace, Sims-Butler, Arceneaux, and Johnson (2002) sought to use the same methodology as Erikson et al. (1993) to estimate state-level attitudes for tolerance, racial integration, abortion, religiosity, homosexuality, feminism, environmentalism, welfare, and capital punishment. They collapsed General Social Survey (GSS) data from 1974 to 1998 for these items (some of which are composite), obtained adequate sample sizes for most states, applied appropriate weights, and calculated frequencies.3 As with Erikson et al., the method requires strong stability over time and reliability in the scores. Otherwise, it cannot be said that welfare opinions in New York, for example, are the same from 1974 to 1998. However, as Table 11.1 indicates, Erikson et al. and Brace et al. (2002) aggregated over 13 and 22 years, respectively, to produce state-level public opinion for less populous states, such as Montana,Vermont,Wyoming, and so on.4 This simple aggregation creates a signiﬁcant trade-oﬀ when producing state-level public opinion, namely, the researcher must focus on

State-Level Opinions from National Surveys 211 Table 11.1 Years, Surveys, and Sample Size to Produce State-Level Opinion Study

Erikson,Wright, McIver Brace, Sims-Butler, Arceneaux, Johnson Multilevel Model (1988) Multilevel Model (1992)

Years

Surveys

N

13 25 9 days 9 months

122 22 7 18

167,460 35,000* 13,544 24,072

* Approximate

state-level opinions that do not vary over time, or else, accept that one is implicitly estimating time-averaged opinions.5 In order to overcome the limitations of Erikson et al. (1993) and Brace et al. (2002), we construct a multilevel logistic regression model for a binary response variable conditional on poststratiﬁcation cells to estimate state-level opinions from national surveys.This approach combines the modeling often used in small-area estimation with the population information used in poststratiﬁcation (Gelman and Little, 1997), which is the standard method for adjusting for nonresponse in political polls (Voss, Gelman, and King, 1995). Section 2 of this chapter presents the model—both poststratiﬁcation and multilevel logistic regression. Section 3 presents the data used in the analysis—the national surveys and U.S. Census data. Section 4 evaluates the model by comparing state-level estimates of candidate choice with actual election outcomes for 1988 and 1992. Section 5 evaluates the model by comparing state-level estimates of partisanship and ideology with Erikson, Wright, and McIver. Section 6 considers further research to improve and test the model. Section 7 discusses the implications of the results.

2. Model 6 2.1 Poststratification The standard practice for weighting in pubic opinion polls is based entirely (or primarily) on poststratiﬁcation, which is generally referred to any estimation scheme that adjusts to population totals.There is a fundamental diﬃculty in setting up poststratiﬁcation categories. It is desirable to divide up the population into many smaller categories, but if the number of respondents is small, it is diﬃcult to accurately estimate the average response within each category. One can improve eﬃciency of estimation by ﬁtting a multilevel model. Consider subsets of the population deﬁned by R categorical variables, where the rth variable has Jr levels, for a total of 3Rr=1 Jr = categories (cells), which are labeled j = 1, ..., J. Assume that Nj, the number of individuals

212

Pa r k , G e l m a n , a n d Ba f u m i

in the population in category j, is known for all j. Let y be a binary response of interest. For each j, let nj and Nj be the number of individuals in category j in the sample and the population, respectively. The R variables should include all information used to construct survey weights, as well as other variables that might be informative about y. For example, the population of adults in the 50 states plus the District of Columbia are categorized by R = 5 variables: state of residence, sex, ethnicity, age, and education, with ( J1, ..., J5 ) = (51 × 2 × 2 × 4 × 4). The J = 3,264 categories range from “Alabama, Male, Not Black, 18–29, Not High School Graduate” to “Wyoming, Female, Black, 65 and over, College Graduate,” and the U.S. Census, provides good estimates of Nj in each of these categories.7 For example, in 1990 there were 66,177 adults who lived in Alabama, were male, not black, between the ages 18–29, and did not have a high school diploma. We shall consider adult population estimates (summing over all 3,264 categories) and also estimates within individual states (separately summing over the 64 categories for each state).

2.2 Regression Modeling in the Context of Poststratification One can set up a logistic regression model for the probability Sj of a “yes” for respondent in category j: logit(Sj ) = XjE where X is a matrix of indicator variables (such as age, education, gender and sex), and Xj is the jth row of X. If a uniform prior distribution on E is assumed, then Bayesian inference (for diﬀerent choices of X) under this model corresponds closely to various classical weighting schemes (Gelman and Little, 1997).

2.3 Multilevel Logistic Regression Model 8 The multilevel model allows for partial pooling across the cells by modeling exchangeable batches of coeﬃcients. The model then can be written in the standard form of a multilevel logistic regression as: yi ~ Binomial(Si ) logit(pi ) = (XE)i E ~ N(0, 6E) 2 Jkmind ~ N(0, Vm), k = 1, ..., Km.

State-Level Opinions from National Surveys 213

We write the vector E as D, J1...,Jm, where D is a subvector of unpooled coeﬃcients and 6E–1 is a diagonal matrix with 0 for each element of D, followed by Vm-2 for each element of Jm, for each m. Each Jm, for m = 1,...,M, is a subvector of coeﬃcients (Jkm) to which we ﬁt a multilevel model.We use the notation pi, for the probability corresponding to the unit i, as distinguished from Sj, the aggregate probability corresponding to the category j. A constant term has been included as part of the unmodeled coeﬃcients D, and so we can give each Jkm a prior mean of 0 with no loss of generality.The group-level standard deviations Vm are given independent noninformative prior distributions: Vm ~ Uniform(0, 100), m = 1, ..., M This essentially noninformative prior distribution allows each Vm to be estimated from the data.This can be contrasted to two extremes that correspond to classical analyses. Setting Vm to 0 corresponds to excluding a set of variables, that is, complete pooling; setting Vm to f corresponds to a noninformative prior distribution on the Jkm parameters, that is, no pooling.

2.4 Estimates Under the Model To obtain the quantities of interest, we used the following strategy (Gelman et al., 2003, section 16.6): 1. Perform Bayesian inference for the regression coeﬃcients, E, and the hyperparameters, Vm, given the data y. 2. For each of the J categories of person in the population, compute pj = logit-1(XE)j.This is done for all categories j including those, such as black male, college graduate, 18–29, Wyoming, that are not represented in the sample. 3. Compute inferences for the population quantities by summing Njpj’s. The model is estimated via Bayesian MCMC (Markov chain Monte Carlo) methods simulation, using WinBUGS (Spiegelhalter,Thomas, and Best, 2000) as called from R. In fact, these simulations are used to compute uncertainties and standard errors.

3. Data 3.1 National Survey Data 3.1.1 1988 Survey Data All respondents from seven pre-election national tracking polls conducted by CBS News/New York Times during the nine days preceding the 1988 U.S.

214

Pa r k , G e l m a n , a n d Ba f u m i

presidential election were used (N = 13,544).9 The outcome variable is support for the Republican presidential candidate where yi = 1 is assigned to supporters of Bush, yi = 0 to supporters of Dukakis, and NA for respondents who expressed “other” or were missing (we follow the standard practice and count respondents who “lean” toward one of the candidates as full supporters).10 Predictor variables include sex, ethnicity, age, and education (respondents were excluded if sex, ethnicity, age, or education were missing). Even though no data were included from Hawaii and Alaska, they are included in the model.11 Also, it is common practice to exclude Washington, DC, because (1) its voting preferences are so diﬀerent from the other states that a model that ﬁt the 50 states would not ﬁt the District of Columbia and (2) it would unduly inﬂuence the results for the other states. However, the District of Columbia is included, and in order to mitigate against such problems, the District of Columbia is given its own region code.

3.1.2 1992 Survey Data All respondents from 18 national surveys, conducted by CBS / NY Times and ABC/Washington Post from March to November 1992, were used (N = 24,072). However, due to the strong showing of a third-party candidate, Ross Perot, three subsets of the overall survey were taken to estimate support for Bush among the two major party candidates: ﬁrst subset included all respondents who were registered or expected to vote (N = 24,072); second subset included all respondents who expected to vote for either the Republican or Democratic presidential candidate (N = 18,460); and the third subset included all respondents who expected to vote for Bush or Clinton (N = 15,497). For the third subset of the data, and similar to 1988, yi = 1 is assigned to supporters of Bush and yi = 0 to supporters of Clinton, and NA for respondents who were missing.12 Again, respondents were excluded if sex, ethnicity, age, or education were missing. Even though no data were included from Hawaii and Alaska, they are again included in the model, and the District of Columbia is again given its own region code.

3.2 National Census Data Survey organizations create weights on the following variables (Voss, Gelman, and King, 1995): Census Region: Sex: Ethnicity: Age: Education:

Northwest, South, North Central,West Female, Male Black, Non-black 18–29, 30–44, 45–64, 65+ Not High School Grad, High School Grad, Some College, College Grad

State-Level Opinions from National Surveys 215

This includes all main eﬀects plus the interactions of “sex × ethnicity” and “age × education.” Sex, ethnicity, and their interaction are included as predictors with unmodeled coeﬃcients in the multilevel logistic regression model, and as mentioned previously, respondents with nonresponse in any of the predictor variables are excluded. The model goes beyond the traditional analysis by survey organizations by including indicators for the 50 states (plus the District of Columbia) clustered into ﬁve batches corresponding to the four census regions plus a separate region for the District of Columbia. In order to improve the estimates, the model is run with the average vote for the Republican presidential candidate for the past three presidential elections for the 50 states (plus the District of Columbia) as a group-level predictor for the state coeﬃcients.13 In addition, “sex × ethnicity,” education, age, and “age × education” are treated as separate batches of coeﬃcients in the multilevel model.The performance of the model is checked by comparing estimates to the actual presidential returns for each state (plus the District of Columbia). In order to poststratify on all the variables listed above, along with the state, we need the joint population distribution of the demographic variables within each state: that is, population totals Nj for each of the 2 × 2 × 4 × 4 × 51 cells of “sex × ethnicity × age × education × state.” As an approximation to that distribution the Census of Population and Housing, 1990 is used: Subject Tape File (SSTF) 6, Education in the United States. SSTF 6 contains the joint population distribution of the demographic variables within each state weighted to represent the total population. SSTF 6 contains sample data weighted to represent the total population. In addition, the ﬁle contains 100 percent counts and unweighted sample counts for total persons and total housing units.

4 Results The method can be applied for any yes/no survey response. We use presidential choice because it can be compared with the actual election outcomes.14 For 1988, we compare the multilevel model with three other models: CBS / New York Times, no pooling, and complete pooling. For 1992, we compare the multilevel model with pooled surveys from CBS/ New York Times and ABC/Washington Post.We expect the multilevel model to perform the best because of the ﬂexibility of the multilevel logistic regression model and because poststratiﬁcation uses the population numbers Nj (Gelman and Little, 1997).15 Survey organizations assign weights to each respondent as the inverse of the probability of selection, modiﬁed by a series of ratio estimates.The

216

Pa r k , G e l m a n , a n d Ba f u m i

ﬁrst stage estimate is essentially a non-interview adjustment within geographic region. For each region, an adjustment is made to approximate the number of adults in that region. In the next stage, ethnicity by sex is the ratio estimate characteristics, and the ﬁnal stage is age by education. On occasion, because of a small number of sample cases in some cells at the ﬁnal stage, some educational groups are collapsed within age categories.To obtain state-level estimates we perform weighted averages within each state, using the weights provided by the survey organization. The “no pooling model” estimates use demographic variables and also indicators for the states, regions, with no multilevel framework.The “complete pooling model” estimates use only age, education, ethnicity, and sex, with state and region eﬀects set to zero. This model allows the average responses within states and regions to diﬀer only because of demographic variables.16 For the no pooling, complete pooling, and multilevel logistic models, we ﬁt the regression models to the survey data, obtain a posterior simulation draws for each coeﬃcient, and reweight based on the 1990 Census of Population and Housing data to obtain the poststratiﬁed estimates for the proportion of voters in each state (including the District of Columbia, Alaska, and Hawaii) who supported Bush for President for 1988 and 1992.

4.1 1988 Election Results 17 Table 11.2 gives the mean national popular vote and mean absolute error of the states. At the national level, the ﬁve methods oﬀer very similar results. The actual result was 53.9 percent support for Bush, the four models produced results that ranged from 54.8 to 58.2 percent. The real eﬃciency gained from the model-based estimates occurs in estimating individual states. The reduction in the mean absolute error of the states from 5.4 percent (CBS/NYT ) to 3.7 percent (Multilevel I) can be attributable to the poststratiﬁcation and multilevel modeling. Furthermore, Table 11.3 shows that the uncertainty (average width of the 50 percent interval) from Multilevel I is relatively small and with slightly less than 24 of the 51 states estimates falling inside the 50 percent intervals. Table 11.3 presents the same estimates as Table 11.2, but includes states with sample sizes less than 100.Table 11.3 clearly illustrates the gain from the multilevel models for small population estimation. Again, at the national level, the ﬁve models oﬀer similar results (for states with sample size less than 100, the mean national popular vote was 53.9 percent and the ﬁve models produced estimates ranging from 50.7 to 58.7 percent). However, the reduction in the mean absolute error of the states falls from 10.3 percent (CBS/NYT ) to 4.9 (Multilevel I) to 2.1 percent (Multilevel II) and

State-Level Opinions from National Surveys 217

(a)

(b)

(c)

(d)

Figure 11.1. Election Results by State vs. (a) CBS/NY Times Weighted Means and Posterior Median Estimate for (b) No Pooling, (c) Complete Pooling, and (d) Multilevel Model for State, Region, and Previous Presidential Vote Eﬀects.

uncertainty from 0.227 (No Pooling) to 0.085 (Multilevel I) to 0.075 (Multilevel II) and with 8 out of 15 state estimates falling inside the 50 percent intervals for Multilevel I and 9 out of 15 states for Multilevel II. Figure 11.1 plots, by state, the actual election returns (vertical axis) versus the CBS/New York Times estimates and the posterior medians (horizontal axis) for the three models (+ indicates states with sample size larger than 100 and n indicates states with sample size smaller than 100. As the results from Table 11.4 indicate, the multilevel model reduces variance, and thus estimation error, especially for states with sample sizes less than 100.

218

Pa r k , G e l m a n , a n d Ba f u m i Table 11.2 Summary Statistics for CBS/New York Times, No Pooling, Complete Pooling, and Multilevel*

Summary

Mean of national popular vote Mean absolute error of states compared to actual election estimate Average width of 50% interval Number of states contained in 50% interval

Actual Result

CBS/NYT

No Pooling

Complete Pooling

Multilevel Model I

Multilevel Model II**

0.539

0.568

0.548

0.582

0.539

0.549

—

0.054

0.068

0.059

0.037

0.031

—

—

(0.112)

(0.012)

(0.053)

(0.048)

—

—

38/51

4/51

24/51

24/51

*Summaries given are the estimated mean of the 50-state vote (plus the District of Columbia) for Bush; the average width of the 50 percent intervals for the states (plus District of Columbia); and the number of the 50 states (plus District of Columbia) whose true values fall within the 50 percent intervals. **Includes the average vote share for the Republican presidential candidate for the previous three presidential elections for the 50 states (plus District of Columbia).

State-Level Opinions from National Surveys 219 Table 11.3 Summary Statistics for CBS/New York Times, No Pooling, Complete Pooling, and Multilevel for States with Sample Sizes Less than 100* Summary

Sample size*** < 100 Mean of national popular vote Mean absolute error of states compared to actual election estimate Average width of 50% interval Number of states contained in 50% interval

Actual Result

CBS/NYT

No Pooling

Complete Pooling

Multilevel Model I

Multilevel Model II**

0.531

0.520

0.509

0.587

0.507

0.529

—

0.103

0.139

0.075

0.049

0.021

—

—

(0.227)

(0.014)

(0.085)

(0.075)

—

—

8/15

2/15

8/15

9/15

*Summaries given are the estimated mean of the 50-state vote (plus the District of Columbia) for Bush; the average width of the 50 percent intervals for the states (plus District of Columbia); and the number of the 50 states (plus District of Columbia) whose true values fall within the 50 percent intervals. ** Includes the average vote share for the Republican presidential candidate for the previous three presidential elections for the 50 states (plus District of Columbia). ***States with sample size <100 are AK, DE, DC, HI, ID, ME, MT, NV, NH, ND, RI, SD, UT,VT, WY.

220

Pa r k , G e l m a n , a n d Ba f u m i

For example, in Nevada (N = 32), 59 percent supported Bush in the 1988 election, CBS/New York Times estimated 51 percent while the multilevel model estimated 59 percent; Vermont (N = 12), 51 percent supported Bush, CBS/New York Times estimated 80 percent while the multilevel model estimated 55 percent.

4.2 1992 Election Results 18 Table 11.4 gives the mean national popular support vote and the mean absolute error of the states. At the national level, the actual result was 46.5 percent support for Bush, while the weighted survey estimated 43.3 percent and the multilevel model 45.5 percent. As mentioned previously, the eﬃciency gained from the model occurs in estimating individual states. The reduction in the mean absolute error of the states falls from 7.9 percent for the weighted survey to 4.9 percent for the multilevel model.The average width of the 50 percent interval from the multilevel is 0.036 with 51 states (including the District of Columbia) falling inside the interval. Figure 11.2 plots, by state, the actual election returns (vertical axis) versus the CBS/New York Times and ABC/Washington Post estimates and the posterior medians (horizontal axis). Again, (+ indicates states with sample size larger than 100 and n indicates states with sample size smaller than 100). For example, in Delaware (N = 49), 45 percent supported Bush in 1992 while the national surveys estimated 35 percent and the multilevel model estimated 48 percent; Idaho (N = 66) Bush received 66 percent, national survey 36 percent and multilevel model 52 percent; Nevada (N = 66) Bush

Table 11.4 Summary Statistics for Weighted Survey Estimate and Multilevel Model* Summary

Mean of national popular vote Mean absolute error of states compared to actual election estimate Average width of 50% interval Number of states contained in 50% interval

Actual Result

National Surveys

Multilevel **

0.465

0.433

0.455

— —

0.073 —

0.049 (0.036)

—

—

51/51

*Summaries given are the estimated mean of the 50-state vote (plus the District of Columbia) for Bush; the average width of the 50 percent intervals for the states (plus District of Columbia); and the number of the 50 states (plus District of Columbia) whose true values fall within the 50 percent intervals. ** Includes the average vote share for the Republican presidential candidate for the previous three presidential elections for the 50 states (plus District of Columbia).

State-Level Opinions from National Surveys 221

(a)

(b)

Figure 11.2. 1992 Election Results by State vs. (a) Survey Weighted Means and (b) Posterior Median Estimate for Multilevel Model for State, Region, and Previous Presidential Vote Eﬀects.

received 48 percent, national survey 37 percent and multilevel model 48 percent; and Rhode Island (N = 56) Bush received 38 percent, national survey 26 percent and multilevel model 36 percent.

4.3 Partisanship 19 Table 11.5 gives the mean ideology estimates of the states. The multilevel model estimates the mean Republican identiﬁcation to be approximately 4.0 percent higher than Erikson et al.’s (1993) estimates. Recall that Erikson et al.’s estimates come from pooled surveys that span from 1976 to 1988; and the multilevel estimates are from 1988.Therefore, as Erikson et al. state, “[i]n terms of party identiﬁcation, the electorate did indeed change. As recently as the Carter presidency, the Democratic Party dominated the Republicans in terms of identiﬁcation, with an average edge of 14.2 percentage points in the states. By Reagan’s second term, this lead had shrunk to an average edge of only 2.8 points. Clearly the Republican party had made major gains” (Erikson et al. 1993, 30). Figure 11.3 shows the plots of Erikson et al. (horizontal axis) versus the multilevel model (vertical axis) for the Republican, Independent, and Democrat and the diﬀerence between Republican-Democrat selfidentiﬁers (+ indicates non-Southern states and n indicates Southern). The Republican graph clearly shows the expected Republican bias for the multilevel results. For example, Mississippi increased their Republican identiﬁcation from 27 to 40 percent, South Carolina from 28 to 37 per-

Table 11.5 Mean Partisanship Estimates for Erikson,Wright, and McIver and Multilevel Model and Their Correlations Mean Partisanship

Republican Democrat Independent

Erikson,Wright & McIver

Multilevel Model*

Correlations

0.29 0.37 0.34

0.33 (0.04) 0.39 (0.03) 0.29 (0.04)

0.72 0.66 0.69

* Does not add to 1.0 due to rounding.

(a)

(b)

(c)

(d)

Figure 11.3. Partisanship Self-Identiﬁcation in the United States: Erikson,Wright, and McIver vs. Multilevel Model: (a) Republican, (b) Independent, (c) Democrat, and (d) Republican-Democrat Diﬀerence.

State-Level Opinions from National Surveys 223

cent,Tennessee from 27 to 40 percent and Virginia from 29 to 42 percent. Therefore, relying on Erikson et al.’s scores of partisanship for 1988 would not capture the dramatic shift toward the Republican party for many of the states.

4.4 Ideology 20 Unlike the results for Republican identiﬁcation, the mean conservative scores for the multilevel model are relatively close to Erikson et al. As Table 11.6 shows, the multilevel model estimates 36 percent of the population identify as conservative, while Erikson et al. estimate 35 percent. Again, at the national level, this ﬁts with our expectations regarding the change (or lack thereof) in conservatism during 1976 to 1988. As Erikson et al. write: [m]ean ideology was quite stable over the 13 year period. In other words, the popular notion that the electorate became more conservative during the Reagan presidency is not supported by CBS / NYT survey responses on ideological selfidentiﬁcation. While the electorate certainly preferred the conservative label under Reagan, it was no more conservative than before. For example, the lead of conservative identiﬁers over liberals was virtually the same during Reagan’s second term as during the Carter presidency!” (Erikson et al., 1993, 30)

Figure 11.4 shows the plots of Erikson et al. (horizontal axis) versus the multilevel model (vertical axis) for conservative, moderate, liberal, and diﬀerence between conservative-liberal self-identiﬁcation (+ indicates non-Southern states and n indicates Southern). The conservative results show the Erikson et al. and multilevel model estimates to be relatively close (clustering around the 45-degree line). However, not all states remained the same. According to the multilevel estimates, several states were estimated as more conservative, such as Alabama from 41 to 50 percent,Arkansas from 37 to 45 percent, and Connecticut from 29 to 41 percent. Other states were estimated as less conservative, such as Arizona from

Table 11.6 Mean Ideology Estimates for Erikson,Wright, and McIver and Multilevel Model and Their Correlations Mean Ideology

Erikson,Wright, and McIver

Multilevel Model

Correlations

Conservative Liberal Moderate

0.35 0.21 0.44

0.36 (0.03) 0.19 (0.03) 0.45 (0.03)

0.54 0.71 0.21

224

Pa r k , G e l m a n , a n d Ba f u m i

(a)

(b)

(c)

(d)

Figure 11.4. Ideology Self-Identiﬁcation in the United States: Erikson,Wright, and McIver vs. Multilevel Model: (a) Conservative, (b) Moderate, (c) Liberal, and (d) Conservative-Liberal Diﬀerence.

37 to 29 percent and Idaho from 43 to 34 percent. Furthermore, as the liberal graph shows, the estimate of liberal self-identiﬁcation decreased. For example, Nevada estimated liberalism went from 29 to 20 percent, West Virginia from 23 to 14 percent, and Oregon from 25 to 17 percent.Again, relying on Erikson et al.’s pooled estimates would produce a diﬀerent ideological picture of the American states.

State-Level Opinions from National Surveys 225

5. Conclusion In the past 20 years, both historical and institutional scholars have pushed the discipline to reconnect the study of institutions and policy making (and policy outcomes) to address the broader questions raised by democratic theory. Public opinion scholars have tended to narrowly focus on individual opinion formation and change.As Erikson et al. (1993) state, “[a]n unfortunate by-product of this emphasis on the individual over the aggregate has been the disconnection of the study of mass behavior from the larger body of democratic theory” (252). When public opinion scholars have emphasized the aggregate over the individual, they have tended to focus on the “national” electorate and the presidency. Even though the House and Senate represents a third of the federal system (and some would argue the most important institution in regards to understanding issues of representation and democratic theory), there has been a dearth of research examining the congruence, or lack thereof, between the “state” or “congressional” electorate and their representatives. The few statistical studies of political representation, beginning with the classic analysis of Miller and Stokes (1963), have found high degrees of representation between constituents’ policy preferences and their representatives’ policy choices in Congress (for example, Achen, 1978; Bartels, 1991, 2002; Erikson et al., 1993; Page, Shapiro, Gronke, and Rosenberg, 1984; Stimson, MacKuen, and Erikson, 1995). However, these previous works were limited by the small sample size, either at the state or congressional district level. The method presented here does not completely solve the problem, but it does allow the researcher to potentially produce more reliable estimates of constituency preferences than previous methods.

Notes 1. Pool et al. (1965) received the support of the Democratic party and the Kennedy campaign to employ computer simulation for the purposes of bolstering their electoral odds. Highlighting the controversy surrounding the project, Theodore Sorenson would insist that no information from the simulations was used to craft Kennedy’s campaign message (Sorenson, 1965). 2. The project was also the basis for a satirical novel by Eugene Burdick titled The 480. 3. State codes for respondents can be purchased from the GSS. 4. Erikson et al. (1993) do not include Alaska and Hawaii, and Brace et al. (2002) do not include Hawaii, Idaho, Maine, Nebraska, Nevada, and New Mexico in their analysis.

226

Pa r k , G e l m a n , a n d Ba f u m i

5. See Brace et al.’s (2002) discussion of stability of the state-level measures. 6. We follow Gelman and Little’s (1997) notation for poststratiﬁcation and the multilevel logistic regression. 7. Census of Population and Housing, 1990: Subject Summary Tape File (SSTF) 6, Education in the United States [machine-readable data ﬁles] prepared by the Bureau of the Census.Washington:The Bureau [producer and distributor], 1993. 8. For introduction to multilevel models see Chapters 5 and 13 of Gelman, Carlin, Stern, and Rubin (2003). Also see the following: Bryk and Raudenbush (1992), Gill (2002), Goldstein (1987), Heck,Thomas, and Thomas (2000), Kreft and de Leeuw (1998), Leyland and Goldstein (2001), and Reise and Duan (2001). 9. ICPSR numbers 9152, 9153, 9154, 9155, 9156, 9157, 9158. 10. There are 1,978 respondents who expressed “other” or were missing.WinBUGS, when encountering missing data on an outcome variable in a regressiontype setting, will make multiple imputations. See Jackman (2000) for a more detailed explanation of how WinBUGS handles missing data on the outcome variable. 11. WinBUGS will estimate parameters for these two states, using the Bayesian multilevel model to get estimates in the absence of any direct data. 12. There are 2,717 respondents who expressed “other” or were missing. 13. It would be slightly preferable to adjust these for home-state and homeregion eﬀects. 14. Jackman and Rivers (2001) use a dynamic Bayesian multilevel model to produce forecasts of the Electoral College outcome by aggregating state-level forecast. 15. In order to improve our election-forecasting model, we can adjust the Nj’s to represent registered or likely voters instead of the voting age population as a whole. 16. The “multilevel logistic model” (Multilevel I) estimates use demographic variables, with the 50 states (plus the District of Columbia) eﬀects, and ﬁve region eﬀects within a multilevel framework. Multilevel II includes all the variables used in Multilevel I, but adds the average vote share for the Republican presidential candidate for the previous three presidential elections for the 50 states, plus the District of Columbia. 17. Results are based on two chains each running for 3,000 iterations, with a 2,000 iteration burn-in. The chains showed convergence as measured by the method of Gelman and Rubin (1992). 18. Results are based on two chains each running for 5,000 iterations, with a 4,000 iteration burn-in. The chains showed convergence as measured by the method of Gelman and Rubin (1992). 19. Results are based on two chains each running for 3,000 iterations, with a 2,000 iteration burn-in. The chains showed convergence as measured by the method of Gelman and Rubin (1992). 20. Results are based on two chains each running for 3,000 iterations, with a 2,000 iteration burn-in. The chains showed convergence as measured by the method of Gelman and Rubin (1992).

State-Level Opinions from National Surveys 227

References Achen, Christopher H. 1978.“Measuring Representation.” American Journal of Political Science 22 (August): 475–510. Bartels, Larry M. 1991.“Constituency Opinion and Congressional Policy Making: The Reagan Defense Build Up.” American Political Science Review 85 ( June): 457–474. ———. 2002.“Beyond the Running Tally: Partisan Bias in Political Perceptions.” Political Behavior 24 ( June): 117–150. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46 ( January): 173–189. Bryk, Anthony S., and Stephen W. Raudenbush. 2001. Hierarchical Linear Models: Applications and Data Analysis Methods, 2nd ed.Thousand Oaks, CA: Sage. Census of Population and Housing, 1990: Subject Summary Tape File (SSTF) 6, Education in the United States [machine-readable data ﬁles] prepared by the Bureau of the Census. Washington: The Bureau [producer and distributor], 1993. Erikson, Robert S. 1976. “The Relationship between Public Opinion and State Policy: A New Look Based on Some Forgotten Data.” American Journal of Political Science 20 ( January): 25–36. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Gelman, Andrew, John B. Carlin, Hal S. Stern, and Donald B. Rubin. 2003. Bayesian Data Analysis. New York: Chapman and Hall/CRC. Gelman, Andrew, and Thomas C. Little. 1997. “Poststratiﬁcation into Many Categories Using Hierarchical Logistic Regression.” Survey Methodologist 23 (September): 127–135. Gelman, Andrew, and Donald Rubin. 1992. “Inference from Iterative Simulation Using Multiple Sequences” (with discussion). Statistical Science 7: 457 – 511. Gill, Jeﬀ. 2002. Bayesian Methods: A Social and Behavioral Sciences Approach. Boca Raton, FL: Chapman and Hall/CRC. Goldstein, Harvey. 1987.“Multilevel Covariance Component Models.” Biometrika 74(xx):430–431. Heck, Ronald H., Scott Thomas, and Loring Thomas. 2000. Introduction to Multilevel Modeling Techniques. Mahwah, NJ: Lawrence Erlbaum Associates. Hill, Kim Quaile, and Jan E. Leighley. 1992. “The Policy Consequences of Class Bias in State Electorates.” American Journal of Political Science 36(2): 351–365. Hill, Kim, Jan E. Leighley, and Angela Hinton-Anderson. 1995.“Lower Class Mobilization and Policy Linkage in the United States.” American Journal of Political Science 39(1): 75–86. Jackman, Simon. 2000. “Estimation and Inference Are Missing Data Problems: Unifying Social Science Statistics via Bayesian Simulation.” Political Analysis 8(4): 307–322. Jackman, Simon, and Douglas Rivers. 2001.“State Level Election Forecasting during Election 2000 via Dynamic Bayesian Hierarchical Modeling.” Paper pre-

228

Pa r k , G e l m a n , a n d Ba f u m i

sented to the annual meeting of the American Political Science Association, San Francisco. Kreft, Ita G., and Jan de Leeuw. 1998. Introducing Multilevel Modeling. Thousand Oaks, CA: Sage. Leyland,A. H., and H. Goldstein. 2001. Multilevel Modelling of Health Statistics. New York: John Wiley. Miller,Warren E., and Donald E. Stokes. 1963. “Constituency Inﬂuence in Congress.” American Political Science Review 57 (March): 45–56. Page, Benjamin I., Robert Y. Shapiro, Paul W. Gronke, and Robert S. Rosenberg. 1984. “Constituency, Party, and Representation in Congress.” Public Opinion Quarterly 48 (Winter): 741–756. Pool, Ithiel de Sola, Robert P.Abelson, and Samuel L. Popkin. 1965. Candidates, Issues, and Strategies. Cambridge, MA: MIT Press. Reise, Steven P., and Naihua Duan. 2001. Multilevel Models: A Special Issue of Multivariate Behavioral Research. Mahwah, NJ: Lawrence Erlbaum Associates. Sorenson,Theodore C. 1965. Kennedy. New York: Perennial Library. Spiegelhalter, David, Andrew Thomas, and Nicky Best. 2000. WinBUGS 1.3:Windows Bayesian Inference Using Gibbs Sampling. Cambridge, UK: MRC Biostatistics Unit. Stimson, James A., Michael B. MacKuen, and Robert S. Erikson. 1995. “Dynamic Representation.” American Political Science Review 89 (September): 543–565. Voss, D. S., Andrew Gelman, and Gary King. 1995.“Pre-election Survey Methodology: Details from Nine Polling Organizations, 1988 and 1992.” Public Opinion Quarterly 59: 98–132. Weber, Ronald E.,Anne H. Hopkins, Michael L. Mezey, and Frank Munger. 1972– 1973. “Computer Simulation of State Electorates.” Public Opinion Quarterly 36 (Spring): 49–65. Weber, Ronald E., and William R. Shafer. 1972. “Public Opinion and American State Policy-Making.” Midwest Journal of Political Science 16(4): 683–699. Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1987.“Public Opinion and Policy Liberalism in the American States.” American Journal of Political Science 31(3): 980–1007.

Chapter

12

Public Opinion in the States: A Quarter Century of Change and Stability

Robert S. Erikson, Gerald C. Wright, and John P. McIver

Introduction More than a decade ago, our book Statehouse Democracy (Erikson,Wright, and McIver, 1993) reported detailed depictions of the states in terms of citizens’ partisan and ideological identiﬁcations.That eﬀort was motivated by the then lack of any valid and reliable measures of public opinion in the states. As a result, explanations of policy diﬀerences among the states relied on census-collected indicators of wealth, urbanization, or population characteristics and a handful of “political” variables like interparty competition, legislative professionalism, and the governors’ constitutional powers.Without measures of public opinion, citizen preferences had no systematic place in accounts of the variations in policy among the states.The main ﬁnding of Statehouse Democracy is the very strong—indeed dominant—role for public opinion in the overall liberalism-conservatism of state policy. Our major eﬀort rested on a set of measures derived from pooling CBS News/New York Times polls taken from 1976 to 1988. Most of our work was necessarily restricted to cross-sectional analyses of public opinion and policy in the states. Eﬀorts at longitudinal analysis of those data confronted two problems. First, dividing the state samples into smaller overtime units like years yields small and highly unreliable estimates. Second, there appeared to be little or no temporal variation in state ideology, which is the prime explanatory inﬂuence on state policy. Because these 229

230

Erikson, Wright, and McIver

data are now almost two decades old, more than suﬃcient time has passed for an update of our ﬁndings.That is the purpose of this chapter. The passage of time presents a problem and also opportunities that this analysis seeks to address. The problem is that the data are dated, and we and others using our measures necessarily need to be concerned that things have changed. Our analyses in working with the initial data collection (1976–1988) led us to the view that state ideology was quite stable, but that partisanship seemed to be undergoing slow but systematic changes.1 We concluded for our purposes then that the changes in partisanship were not great enough to undermine conclusions based on our cross-sectional measures. However, we now need to consider the possibility that the opinion proﬁles of the states have changed signiﬁcantly so that use of the old measures to explain current policy may not yield accurate results. Hence, the ﬁrst task of this update is simply to provide a new set of estimates of state ideology and partisanship. Fortunately, CBS News and the New York Times have continued their polling partnership, providing us with the raw data for our examination of opinion change in the states. Understandably, scholars have argued for longitudinal analyses of public opinion and policy (Berry, Ringquist, Fording, Hanson, 1998; Gray, 1976; Lowery, Gray, and Hager, 1989; Norrander, 2001). Politics and policy are not static, and if policies change, we would like to be in a position to say whether these changes are a result of shifts in public preferences.We now have adequate time series on opinions in the states to assess our earlier conclusion about the stability of state ideology compared to state partisanship, and to go further and describe the nature and causes of those changes that have occurred. We ﬁnd that state ideology has been remarkably stable and that partisanship continued the changes that were under way and visible in the initial series.We ﬁnd compelling evidence that the changes in state partisanship have been wrought by a convergence of party with ideology; that is, a stable set of ideological preferences have driven changes in partisanship among the 50 state electorates. First, let us refresh ourselves regarding the nature of state public opinion in the 1970s and 1980s for our report in Statehouse Democracy. Back then, the nation (and therefore most states) identiﬁed decidedly as more Democratic than Republican in partisanship but decidedly more conservative than liberal in ideology. Interestingly, there existed virtually no state-level correlation between the two forms of identiﬁcation.The liberalism or conservatism of the state public had virtually no zero-order predictive power in terms of accounting for state partisanship. Only when the ideological tendencies of the state parties were taken into account did ideology begin to predict par-

Public Opinion in the States 231

tisanship. In eﬀect, state partisanship was a function of the relative ideological proximities of the state parties to the voters as measured by the relative ideological positions of the state electorates and state parties. State partisanship was locally determined—a marker for the relative success of the state Democratic and Republican parties at representing public opinion. Our longitudinal perspective allows us to document changes in state opinion as well as to look from the states’ viewpoint into the widely held perception that politics has become more ideological—more “polarized” as the cliché would have it. The recent close presidential elections make this discussion especially acute at the state level, as analysts typically see a vast political gulf between “red” (Republican) and “blue” (Democratic) states (but see Fiorina, 2004). This chapter examines state-level ideology and partisanship over time, using over a quarter century of data from CBS News/New York Times polls.What to do we ﬁnd with this extended time horizon? • In our earlier analysis, state variation in ideological identiﬁcation was found to be highly stable over the 1976–1988 period.This pattern has persisted, post-1988. The most liberal states in the 1970s and 1980s continue to be the most liberal states today. • In our earlier analysis, state party identiﬁcation showed some signs of systematic change toward the Republican party.That movement continued and even accelerated for some states while reversing course in others. Much, but not all, of the Republican gains have occurred with the southern states abandoning their one-time Democratic allegiance. • In the 1976–1988 period, state-level partisanship and ideology were not strongly correlated, and in fact the correlation for the data averaged over that 13-year period is slightly negative. Over time, this correlation has become decidedly positive. It is now true that the most liberal states are the most Democratic and the most conservative states are the most Republican. • We have evidence regarding what caused what. Did partisanship begin to aﬀect ideology or the other way around? Or could both processes be at work? We ﬁnd that state ideology drives state partisanship with no trace of the reverse process. The evidence suggests that citizens are shifting their partisanship relative to more uniform perceptions of the political parties’ ideologies and issue positions. • Measured by the standard technique of comparing variances over time, states have not become more diverse in their partisan or ideological preferences. However, polarization has increased in the sense

232

Erikson, Wright, and McIver

that partisanship and ideology are now more aligned so that they reinforce each other at the state level.

The Data Our data consists of the responses to CBS News/New York Times polls, 1976–2003. Our set includes 394 surveys, which yields over 430,000 respondents, more than the current population of Pittsburgh, PA. This is a huge increase over the 167,000 respondents used in Statehouse Democracy. One approach would be to divide these data into 28 annual time segments. However, the N’s per state are often so small that annual estimates of state opinion are unreliable. We discovered that a fruitful way is to divide the data into seven periods.This provides estimates of acceptable reliability and the temporal units are substantively meaningful.While the natural way to divide 28 years by seven is to simply do seven 4-year segments, we instead divided them by presidency. For this chapter, we discard 1976 data (Ford presidency) and divide by Carter (1977–1980) through GW Bush’s ﬁrst term (2001–2003), our one presidency with only three years of data.2 In between, we have Reagan I (1981–1984), Reagan II (1985–1988), GHW Bush (1989–1992), Clinton I (1993–1996), and Clinton II (1996–2000).3 For much of our analysis we examine opinion change in the states across presidential administrations. However, for a revised set of cross-sectional estimates of state opinion, it is probably wiser to aggregate over a longer time frame. Thus, we combine the second Clinton administration and the ﬁrst 3 years of George W. Bush as to provide a set of contemporary measures of state partisanship and ideology.These are provided in the appendix and are available in electronic form.4 Our approach of tracking trends across presidential administrations makes substantive sense. Presidents are often the focal point of politics for many Americans and presidential approval and support for the national government are intimately intertwined with the administrations of individual presidencies.The electoral successes of parties and candidates rise and fall with the public’s understanding and evaluation of presidents. It would not be surprising to see regional and local assessments associated with the administrations of national leaders. Table 12.1 illustrates the individual-level trends over the seven presidencies.The mean ideological identiﬁcation (shown for the national public—not the mean of state means) tilts in the conservative direction and is remarkably stable. Party identiﬁcation tilts Democratic, but has been moving more Republican over the seven presidencies. Figure 12.1 graphs these tendencies; it displays the patterns of change in mean national partisanship and ideology over the seven presidencies.

Public Opinion in the States 233 Table 12.1 Mean Ideological and Partisan Identification over Seven Presidencies, CBS News/New York Times Individual-Level Data Ideology a

Carter Reagan I Reagan II Bush 41 Clinton I Clinton II Bush 43 Total (1977–2003)

Party b

Mean

N

Mean

N

–11.79 –12.94 –13.16 –11.21 –15.10 –11.72 –11.37 –12.64

36,609 43,554 55,250 70,007 89,677 88,232 40,156 423,485

14.40 8.98 4.29 2.74 4.20 5.19 2.10 5.32

40,345 45,685 64,675 72,080 93,648 92,813 42,761 452,007

-20

-10

0

10

20

Source: CBS News/New York Times polls a –1 = Conservative, 0 = Moderate, +1 = Liberal b –1 = Republican, 0=Independent, +1=Democrat

Carter

Reagan I

Reagan II

Bush41

Party Identification

Clinton I

Clinton II

Bush43

Ideological Identification

NOTE: Partisanship = %Democratic – %Republican Ideology = % Liberal – % Conservative

Figure 12.1. National Partisanship by Presidential Administration. Source: CBS News/New York Times polls.

Table 12.1 also shows the total N’s of the individual-level data per presidency. Divided by 48 (states) yields the average N’s per state, which have a wide variation due to varying population size. In analyzing the state means, we need to adjust for the reliability of the data. Assuming the equivalent of simple random samples, this is straightforward (see Erikson,

234

Erikson, Wright, and McIver

Wright, and McIver, 1993, Chapter 2, for details). For ideology, the estimated reliability per presidency varies from 0.80 to 0.89. For partisanship, the estimated reliability varies from 0.91 to 0.96.

Stability and Change We have seen that the national mean of ideological identiﬁcation is quite stable over time. The nation was slightly conservative in 1976 and remains so today. Although a plurality of Americans declare themselves to be moderate or alternatively middle of the road, conservatives outnumber liberals by about 3–2 with the precise ratio from state to state aﬀecting the degree to which conservatives dominate. In the vast majority of cases, the mean of state electorates is to the right of center.5 Figure 12.1 also illustrates the greater instability of partisanship. Party identiﬁcation has been moving Republican in terms of the overall state mean. Here we examine if this change is a result of uniform changes across the states, or whether some states changed quite a bit and others less so. Initial answers are provided visually in Figure 12.2 and Figure 12.3. Figure 12.2 displays the state ideology as measured during the Carter years with states’ contemporary ideology from the early GW Bush administration.6 The states show considerable stability, even with the readings two decades apart.There is no consistent pattern of departures from the line of equality, or no change. Notice that those with the largest apparent changes tend to be the smaller states that have larger sampling errors.As we will see, the true ideological stability of the states is even greater, once this sampling error is taken into account. Figure 12.3 contrasts the partisan leanings of the states in the late 1970s with those of the ﬁrst few years of the twenty-ﬁrst millennium. Here we see unmistakable shifts. Almost the entire scatterplot has dropped, which signals increased Republican identiﬁcation. The states that have changed the most are from the South; notice the substantial distances of Louisiana, Georgia,Alabama, and Mississippi from their earlier partisanship. However, almost all the states are more Republican than they were in the Carter years. The South led the charge, but that is not the whole story. We will have more to say on the topic below. An improved perspective on stability and change can be drawn once we consider random measurement error inherent in these survey-based estimates. Table 12.2 (top panel) shows the pattern of reliability-corrected overtime correlations for ideology. For ideology, over-time imputed correlations are high—ranging from 0.69 to an impossible 1.01 (where the estimated product of the reliabilities of ideology measuring within each administration slightly exceeds the observed correlation between administration-

10 State Ideology 2001–2003 -20 -10 0

VT DE

CT WY

CA

CO MD NM PA ILMI ME MN OR WA FL MT WV AZ WIOH VANH MO IA NC GAKS IN KY TX ND OK UT ID SC MS

NY NJ

RI

NE TN LA AL AR

-30

SD

MA

-40

-30

-20 -10 State Ideology 1977–1980

0

10

State ideology = %Liberal – %Conservative

State Partisanship 2001–2003 -30 -20 -10 0 10 20 30 40

50

Figure 12.2. Stability of State Ideology. Source: CBS News/New York Times polls.

DE RI MA WV MD NM KY

NY CT CA MN PA ILMI WA VT NJ WI NC OR FL OK IA OH SD ME MO TN VA CO WY IN SC NH MT TX MS ND AZ KS

AR LA

GA AL

NE

ID UT

-20

-10

0 10 20 30 State Partisanship 1977–1980

40

50

State partisanship = %Democratic – %Republican

Figure 12.3. Stability of State Partisanship. Source: CBS News/New York Times polls.

236

Erikson, Wright, and McIver

speciﬁc ideology). State-level ideological diﬀerences persist over the last quarter of a century. States may change their ideological positions over time, but the quarter century of data we have do not reveal any signiﬁcant systematic shifts. Table 12.2 (bottom panel) shows the pattern of reliability-corrected over-time correlations for party identiﬁcation.The correlations range further for partisanship than ideology, from a low of 0.56 (between the Carter and Reagan I eras and the contemporary period of George W. Bush) to a perfect 1.00 between Clinton I and Clinton II. Clearly there has been more movement of relative state partisanship than of relative state ideology.This reﬂects what we see graphically in Figure 12.3. A few states like Iowa,Vermont, and Pennsylvania remained quite stable while others, particularly those in the Deep South, have changed dramatically. Table 12.3 summarizes the two sets of over-time correlations as mean correlations for the various time gaps between presidencies—from one term to six. Clearly, the correlations decline with the time gap.Those periods close to one another are most similar. Periods farthest apart are least alike.We can ﬁt an AR1 model to these correlations (meaning that earlier history is of no predictive value once the previous period’s score is known), with the assumption of a steady rate of correlation decay per the number of periods t as r t. The ideology data roughly ﬁts a pattern where the correlation declines at a rate of 0.97t where t is the gap between readings. For party, the ﬁt is roughly with a rate of 0.95t. Although theses rates of decay seem quite similar, they produce signiﬁcant diﬀerences in the clustering of states over time.They are also distinct enough so that state partisanship will shift considerably more than state ideology over extended periods. For example, after 5 periods (20 years), the correlation between ideology at time 1 and time 7 is an estimated 0.86 while the correlation between partisanship scores over the same time will only be 0.77 in this model. After 10 periods (40 years), the correlations between distance measures of ideology will be 0.73 but partisanship will correlate at less than 0.60. Even with these very low rates of decay, after 60 years, only 40 percent of the variance in ideology will be explained by its initial starting point and barely 20 percent of contemporary partisanship will be a product of early partisan leanings. Realignments in the past have reset the bar in less time than 15 periods; indeed the probability of realignment may be a function of increasing detachment from the partisan and ideological anchors that initially deﬁne a political era. In summary, this analysis clearly shows a remarkable stability to ideological preferences in the states. States do not make sudden liberal or conservative turns. Even partisan change is a gradual process.

Public Opinion in the States 237 Table 12.2 Reliability-Corrected Over-Time Correlations of State Ideological Identification and State Party Identification, 1977–2003, by Presidential Administration Carter

Reagan I

Reagan II

Bush 41

Clinton I

Clinton II

State Ideological Identification Carter Reagan I Reagan II Bush 41 Clinton I Clinton II Bush 43

0.88 0.96 0.89 0.85 0.79 0.79

0.90 0.86 0.82 0.76 0.69

0.94 1.00 0.91 0.81

0.93 0.94 0.91

1.00 0.91

1.01

State Party Identification Carter Reagan I Reagan II Bush 41 Clinton I Clinton II Bush 43

0.92 0.90 0.78 0.74 0.72 0.56

0.98 0.93 0.83 0.86 0.56

0.96 0.91 0.92 0.79

0.92 0.91 0.73

1.00 0.92

0.93

N = 48;Alaska, Hawaii, and District of Columbia excluded. Source: CBS News / New York Times polls

Table 12.3 Mean Reliability-Corrected Over-Time Correlations by Time Gap in Number of Presidential Administrations Ideology Party

1

2

3

4

5

6

0.94 0.95

0.93 0.91

0.88 0.82

0.81 0.80

0.74 0.64

0.79 0.56

Source: CBS News/New York Times polls

Some may think that this message of stability deﬁes what we know about elections in the states, where Democrats and Republicans often oscillate in power.These temporary electoral shifts in voting outcomes do not represent shifts in ideology and (in the short term) not even in partisanship.7 Rather, shifts in state-level voting may represent shifting policy demands due to one party being in power long enough to shift policy in its favor and away from the state’s median voter (Erikson, MacKuen, and Stimson, 2002; Stimson, 1991). For example, if the Republicans hold power in a conservative state long enough, they will push policy suﬃciently in a conservative direction that the state’s voters elect the Democrats to power to provide an ideolog-

238

Erikson, Wright, and McIver

ical correction.8 Alternatively, some electoral change is performance rather than policy based as when incumbents are punished for a bad economy or scandals. Strong evidence shows short-term national forces intrude in state politics adding to short-term electoral change that does indicate shifts in state partisanship or ideology (Carsey and Wright, 1998).

The Polarization of the State Electorates It is approaching common knowledge that the United States is becoming increasingly polarized in terms of the party-ideology connection. Liberal Republicans and conservative Democrats were once commonplace, but that is no longer true today. Here we can document the statelevel manifestation of this process. Table 12.4 displays the cross-sectional correlations between state ideology and partisanship. During the Carter presidency, state liberalism and state Democratic partisanship were weakly correlated, at 0.06. Since the Carter years, this correlation has increased in non-monotonic ﬁts and starts, to where it surged to 0.66 during the GW Bush presidency. We can see these changes graphically by comparing Figure 12.4A and 12.4B. During the late 1970s, there is little obvious connection between the partisanship of American states and their ideological preferences.Yet in the contemporary period we can clearly observe the political polarization in the form of alignment of ideology and partisanship.The liberal states of the Northeast are the most democratic.As the southern states have abandoned the Democratic party, we can see the integration of conservative ideology and Republican party support in the southern and mountain west states. Our over-time data allows some testing of which caused which.That is, does state partisanship drive ideology or vice versa? To ﬁnd out, we regress current values of the selected dependent variable on lagged values of this variable plus lagged values of the presumed causal variable. With seven presidencies, we can set the “current values” to be those for each of six presidencies. Table 12.5 (top panel) shows the results predicting current values of ideological identiﬁcation. Table 12.5 (bottom panel) does the same for party identiﬁcation.9 Table 12.5 (top panel) shows the results for ideology as the dependent Table 12.4 Reliability-Corrected Cross-Sectional Correlations Between State Ideological Identification and State Party Identification Carter 0.06

Reagan I 0.30

Reagan II 0.06

Bush 41 0.12

Clinton I 0.41

Clinton II 0.36

Bush 43 0.66

A: Jimmy Carter Admin. (1977–1980)

B: George W. Bush Admin. (2001–2003)

LA

30

OK

NC MS

20

TX

KY

MD DE

WV NM

DE MA

TN MN FLWY WACA VA MO CT NY OR AZ ME MI WI SD OH NJ IL INNE MT PA CO VT IA ND NH UT KS

RI RI

AR MD MA NY CT CA MN KY HI PA IL MI NJ VT WA NCWIFL OR OK IA OH ME SDTN GA MO VA CO WY AL IN NHMT MSSC TX ND AZ WV

LA

-10

0

10

SC

ID

-20

State Partisanship %Democratic – %Republican

40

GA AL AR

NM

NE AK KS

-30

ID UT

-30

-20

-10

0

10

-30

-20

-10

0

10

State Ideology %Liberal – %Conservative

Figure 12.4. State Partisanship by Ideology. Source: CBS News/New York Times polls.

Table 12.5 Reliability-Corrected Regressions of State Ideology and State Partisanship on Lagged State Ideology and Lagged State Partisanship State Ideology Reagan I Reagan II Bush 41 Clinton I Clinton II Bush 43 Mean State Partisanship Reagan I Reagan II Bush 41 Clinton I Clinton II Bush 43 Mean

Constant

Ideology t-1

Party t-1

R2

–0.05** 0.02 0.05** –0.05*** 0.04*** 0.01

0.71*** 0.96*** 1.17*** 0.90*** 1.05*** 1.01***

–0.04 –0.25*** 0.04 –0.02 –0.05 –0.04

0.78 0.99 0.89 0.87 0.97 1.02

0.96

–0.06

0.92

0.00

–0.05* –0.02 –0.05** 0.04*** 0.03* 0.03 0.00

0.07 0.04 –0.17 0.20* 0.07 0.37*** 0.1

0.90*** 0.80*** 1.10*** 0.80*** 1.06*** 0.85***

0.85 0.96 0.94 0.87 1.00 0.94

0.92

0.93

* = significant at 0.05; ** = significant at 0.01; *** = significant at 0.001, using two-tailed tests. Lag interval = 4-year band (one presidential term). Source: CBS News/New York Times polls

240

Erikson, Wright, and McIver

variable. These results are decisive. In ﬁve of the six regressions, the estimated impact of party on ideology is actually negative and in one case (mysteriously) statistically signiﬁcantly so.Thus, there is no statistical argument that state-level partisanship inﬂuences state-level ideology. Table 12.5 (bottom panel) shows the results for party identiﬁcation as the dependent variable. These results suggest that ideology moves partisanship. Interestingly, the equations suggest not a steady impact but rather two periods where the impact of ideology was particularly strong: These were the ﬁrst Clinton presidency and again during the GW Bush presidency. In each case a change of party in the White House signaled a partisan shift in response to ideology.These results sum to the quite substantial alignment of state ideology and partisanship that we see in comparing Figures 12.4A and 12.4B. The inﬂuence of ideology on party can most clearly be seen by regressing GW Bush partisanship on ideology and partisanship from the Carter presidency. The results of this exercise are shown in Table 12.6. Whether comparing the unstandardized coeﬃcients, the standardized coeﬃcients (betas), or t-values, the greater of the two “eﬀects” is ideology! Thus, to predict whether a state is Democratic or Republican in the twenty-ﬁrst century based on information from the Carter years, it is actually more informative to know its ideological preferences than its partisanship. Meanwhile, if we regress GW Bush era ideology on ideology and partisanship from the Carter presidency (not shown in a table), the partisanship coeﬃcient is small, insigniﬁcant, and negative.The lagged ideology coeﬃcient is a sturdy 0.75. The growing correlation between partisanship and ideology is clearly driven by ideology. This is as if the electorate has been re-sorting as Democrats and Republicans on the basis of their ideological leanings. In Statehouse Democracy (Erikson, Wright, and McIver, 1993), we observed a pattern whereby state partisanship was governed by the ideological preferences of state electorates relative to those of the state parties. It may be that state parties continue to be important signals to their electorates about what the parties stand for and the state parties have changed. The nature of this change would be that Democratic parties across the states have become more consistently liberal and the Republican parties more consistently conservative.Voters would be forming their partisan loyalties by the same process, still in reaction to the state parties.Another possibility is that the national parties have become more salient and supplanted the state parties as references for what the political parties stand for. Disentangling the causal factors here is beyond the scope of this analysis, but we can conclude that party identiﬁcations are more aligned and consistent with popular images of a liberal Democratic party and conservative Republican

Public Opinion in the States 241 Table 12.6. Reliability-Corrected Regressions of State Partisanship 2001–2003 (GW Bush) on State Partisanship and Ideology 1977–1980 (Carter)

Coefficient Standard error Standardized coefficient

Constant

Ideology (Carter Years)

Party ID (Carter Years)

0.04 0.03 –

0.75 0.14*** 0.61

0.48 0.10*** 0.52

R2

0.66

Source: CBS News/New York Times polls

Table 12.7. (Non-Reliability Corrected) Correlations Between Ideological and Party Identification by Region, for Six Presidencies Carter

North (N = 35) South, Border (N = 14)

Reagan I

Reagan II

Bush 43

Clinton I

Clinton II

Bush 41

0.63

0.82

0.70

0.78

0.75

0.80

0.85

0.16

–0.09

0.20

0.15

0.00

–0.05

0.31

South and Border states include the 11 former Confederate states plus Kentucky, Oklahoma, and West Virginia. Source: CBS News/New York Times polls

party. As the ideological clariﬁcation of the parties penetrates state political debates, we would expect partisanship to become more central in many of the states where the partisan divisions have been relatively muted. Much of the change that has occurred has been in the South. If we analyze the non-south alone, the movement has been more modest than for the nation as a whole.Table 12.7 shows this. Outside the South, ideology and partisanship were correlated even during the Carter years. For these non-southern states, the increase in the correlation has been gradual but signiﬁcant, given its already modest level in the 1970s. The political changes in the South have made a larger contribution to the heightened polarization of the state electorates than the rest of the nation combined.

The Southern Story—An Expanded Telling 10 All of the southern states have slowly migrated away from the Democrats, a process V. O. Key called “secular realignment” (Key, 1959).The general tendency is of a transformation from single (Democratic) party dominance into party competitive states.11 The larger states of the south are shown in Figure 12.5. In this ﬁgure, Georgia, a state that as late as the end of the 1970s was dominated by

242 .4

Erikson, Wright, and McIver

GA

NC TX

-.1

0

State Partisanship .1 .2 .3

AL

Carter

Reagan I

Reagan II

Bush 41

Clinton I

Clinton II

Bush 43

Presidential Administration

NC

GA TX AL

-.4

State Ideology -.2

0

Figure 12.5. The Changing South: Partisanship from Carter to Bush 43. Source: CBS News/New York Times polls.

Carter

Reagan I

Reagan II

Bush 41

Clinton I

Clinton II

Bush 43

Presidential Administration Figure 12.6. The Un-Changing South: Ideology from Carter to Bush 43. Source: CBS News/New York Times polls.

southern Democrats, is essentially a state with equal numbers of Democrats and Republicans today. Texas, once home to and controlled by Lyndon Johnson Democrats, is now a majority Republican state. This trend is found throughout the south.A more complex graph would show the same trends in Louisiana, South Carolina, Mississippi, and Kentucky.12

Public Opinion in the States 243

But the southern states’ march toward the Republican party is not the only story that may be told about the last quarter century. Figure 12.6 presents a very diﬀerent picture regarding the ideological identiﬁcation of southern electorates, one quite consistent with our story. Here the reader will see the minor ﬂuctuations in ideology over time. States move a few percentage points from one period to the next but never go far and always return to where they start. The region began the last two decades of the twentieth century as quite conservative and begins the twenty-ﬁrst century with the same basic political leanings.

Looking Outside the South Other state and regional changes are less obvious than the shift in southern partisanship. Nonetheless, there are changes going on that should be recognized for they have important consequences for both state and national politics. A surprisingly common pattern in these data that runs counter to the southern secular realignment is a trend toward the Republican party during the Reagan years that is followed by a shift toward the Democrats during the Clinton administrations.This is not a pattern of random ﬂuctuation over time but a fairly systematic shift across a number of state electorates. Figure 12.7 illustrates this pattern of change. California of the 1970s is a solidly Democratic state but it slips signiﬁcantly toward the Republican party throughout the Reagan-Bush years. By the early 1990s it appears to be “in play” with an electorate equally balanced between Republicans and Democrats.Yet today it is again a solidly Democratic state. California is not unique. Figure 12.7 displays similar transitions for Connecticut, New Jersey, and Michigan.13 Although an explanation for this pattern is not found in the display, we surmise the successes and popularity of the two-term presidents who dominated the decades of the 1980s and 1990s did sway the public’s view of the two major parties. And the pull toward greater consistency between ideology and partisanship outside the south is just as clear here as it is within the south. These diﬀerent historical shifts in southern and non-southern states partisanship are hidden in Figure 12.1 but they magnify state polarization. The over-time pattern in partisanship that we observed in Figure 12.1 is actually the product of three distinct trends that are averaged in that ﬁgure to produce what looks like only modest changes in partisanship over the last several administrations. After 1990, some states—dominated by those in the south—continued their migration toward the Republican party. But during this same time, other states that earlier appeared to be shifting in the same direction, reversed course and moved back toward the Demo-

244

.1

CA CT NJ MI

-.1

0

State Partisanship

.2

.3

Erikson, Wright, and McIver

Carter

Reagan I

Reagan II

Bush 41

Clinton I

Clinton II

Bush 43

Presidential Administration

Figure 12.7. Partisan Change over Time: Reclaiming Democratic Voters. Source: CBS News/New York Times polls.

cratic party.These two trends (overlapping with those states that remained fairly stable during the last 12 years) helped bring about the conﬂuence of ideology and partisanship producing the liberal Democratic states and conservative Republican states we observe today.

Implications for Electoral Politics: The Contemporary Battleground The contemporary polarization of American politics is widely recognized in popular commentary about the diﬀerences between “red” and “blue” America. But national polarization is the result of a diverse set of state-level changes. State electorates do not all dance in the same way to the beat of the same drummer. In fact, the presidential coalitions of 2004 have arrived where they are by quite varied paths as the states’ partisanship has diﬀerentially transformed to align with the increasingly clear ideological identities of the parties.We can see this by grouping the states by the popular red/blue divide and tracing the partisan trajectories of these groups of states have followed to their current patterns. Here we deﬁne red states as those that Bush carried by 5 percent or more and the blue states as those in which he lost by the same margin in the 2004 election. The remaining competitive states are purple. These patterns of partisanship are shown in Figure 12.8. One thing is evident from the ﬁgures: partisanship of the red/blue dis-

-5

Mean Partisanship %Dem - % Repub 0 5 10 15

20

Public Opinion in the States 245

Carter

Reagan I Reagan II

Bush41

Clinton I

Clinton II

Bush43

Red states(>55% Bush) Blue states (>55% Kerry) Purple states (>45%Kerry & <55% Bush)

Figure 12.8. Partisan Histories of 2004 Red and Blue States. Source: CBS News/ New York Times polls.

tinction in 2004 makes good sense: the red states show distinct Republican leanings, blue states tilt Democratic, and in the competitive purple states are the parties are near parity in terms of party identiﬁers.That Bush did better in 2004 among the states with more Republicans is not surprising. If it were otherwise, we would have to doubt the validity of our measures. But notice that through the Carter and early Reagan years the 2004 proBush red states were distinctly Democratic. They were more Democratic than the 2004 Kerry states and a lot more Democratic than they are today. Another dynamic revealed in Figure 12.8 is the character of partisan change among these three groups of states. For the 1970s and 1980s all three groups underwent a steady and substantial growth in Republican partisans. Then in the Clinton years and since, they have splintered with the red states continuing their Republican conversion and the blue states moving back toward, but not quite reaching, their earlier levels of Democratic advantage among party identiﬁers.The purple states tend to mirror movements of the red states, but just at a level closer to an even partisan balance. The dynamics of ideology and partisanship among the American state electorates helps us to understand the future of national campaigns within

246

Erikson, Wright, and McIver

a state-centered Electoral College system.The tight battle lines of the 2000 and 2004 presidential elections have focused attention on the distinction between the “red” (Republican) and “blue” Democratic states in terms of presidential voting. The image is of a stretching of the division of states where, based on their current partisanship and ideology, some vote invariably Republican and others invariably Democratic, with few up for grabs. As Fiorina (2004) has pointed out, this imagery of growing polarization as growing variance is clearly wrong when applied to state-level presidential election results. Speciﬁcally, the observed variance of percent Democratic (or Republican) of the two-party vote has not been increasing over recent elections. In terms of voting for president, the various states may be coming closer together rather than moving apart. What about state partisanship and ideology? We can observe the variance of state scores over time in the CBS News/New York Times data set. Further, we can adjust for sample reliability to estimate the true state variances. These adjusted variances are reported in Table 12.8. As the table shows, there exists no discernable trend in the variance of states in terms of partisanship or ideology. The stable variance of state ideology means that state diﬀerences in ideological preferences are not stretching over the years. And as state partisanship changes to accommodate the new inﬂuence of ideology, it too does not stretch.The depiction of states as getting more apart in terms of ideological or party preference is simply wrong. But we must remember that in a very diﬀerent and fundamental way the states are polarizing. In an earlier era, a very conservative state was just as likely to be a Democratic or a Republican state. A very liberal state was just as likely to be a Democratic or a Republican state. No more. Today ideology and partisanship generally reinforce each other, so that states are either Democratic and liberal or Republican and conservative.As we have seen, it is ideology that has driven partisanship rather than the other way around. What makes state-level voting in contemporary presidential elections appear as polarizing is that ideology has greater important and predictive power in state presidential voting patterns, but also, with the aligning movement of state partisanship to ideology. The result is that rather than partisanship and ideology being cross-currents in any given election, they are reinforcing. Jimmy Carter, in part because he was a southerner, could win as a Democrat in 1976 in a campaign the blurred ideological diﬀerences. But by 1992, state electorates reacted to another southerner, Bill Clinton in terms of the ideological preferences. We can see this in Table 12.9. Here we show the reliability-corrected regressions of presidential

Public Opinion in the States 247 Table 12.8 (Reliability-Adjusted) Variance of State Ideological Identification and Party Identification, by Presidencies Variance (Reliability Adjusted) Administration

Carter Reagan I Reagan II Bush 41 Clinton I Clinton II Bush 43

Ideology

78.1 50.7 46.0 72.3 66.9 75.7 72.8

Party Identification

155.1 99.4 146.1 129.3 102.3 120.7 127.3

Source: CBS News/New York Times polls

Table 12.9 (Reliability-Adjusted) Regression of Presidential Vote on State Ideology and Partisanship, 1976–2004 Election

Constant

1976 1980 1984 1988 1992 1996 2000 2004

47.9 45.2 45.2 51.1 58.5 60.2 55.5 55.7

Ideology

0.13 0.44*** 0.53*** 0.48*** 0.49*** 0.57*** 0.64*** 0.71***

Partisanship

0.41*** 0.44*** 0.23*** 0.17** 0.32*** 0.36*** 0.43*** 0.30***

R2

0.70 0.69 0.66 0.59 0.80 0.82 0.86 0.86

** p < 0.01, *** p < 0.001.Analysis excludes Nevada and includes Alaska and Hawaii for the 2000 and 2004 elections. Source: CBS News/New York Times polls

vote on contemporaneous state ideology and partisanship.14 The pattern of coeﬃcients suggests something of a sea-change in the character of presidential elections.The 1976 election was primarily about partisanship with state ideology not even making statistical signiﬁcance. Since that time, however, the electoral contours of American presidential politics have been transformed. Notice that the impact of state ideology grows monotonically with each presidential election so that state ideology, which in 1976 was almost invisible, becomes the dominant factor determining the presidential vote in the states. By 2004 diﬀerences in state ideological predispositions have over twice the impact on the vote as equivalent diﬀerences in partisanship.

248

Erikson, Wright, and McIver

We can also see that as partisanship and ideology aligned over the last quarter century, the predictability of the states’ presidential vote has increased, from R2’s in the 0.7 range in the 1970s and 1980s to 0.86 in the current period. This very predictability may have contributed to perceptions of a polarized nation. Rather than the states being further apart on either ideology or partisanship, they are further apart on the combination, and as a result, few states are “in play” in the presidential elections. In the 1970s many states, especially in the South, could vote Democratic if the campaigns appealed to values consistent with and that evoked those identiﬁcations, or could swing Republican if the issues of the campaigns turned on ideological preferences. A lot of states and a many more voters were cross-pressured between partisanship and ideology.That is much less the case today.With ideology and partisanship reinforcing each other, the states are more ﬁxed in how they will vote. For many, short of some cataclysmic change, how they will go in 2008 can be predicted today. A handful of states are now pretty evenly balanced in terms of partisanship and ideology and they have achieved a great deal of attention as the “battleground states.” The alignment of ideology and partisanship limit the options of campaigners because they pretty well deﬁne American national politics in terms of placement on a single liberal-conservative continuum.

Conclusion First, let us review what we have found in adding 15 years of polls to the initial estimates of state ideology and partisanship we used in Statehouse Democracy.The basic pattern is quite clear.The ideological preferences of the state electorates have been essentially static for the last quarter century.There has been almost no change in the order of the relative liberalism-conservatism of the states, and even further, we are surprised at the relative stability of ideology in the aggregate. One might understandably suspect with the onslaught liberalism has sustained from the right that citizens might shy away from the “L” word. But that has not been the case. Although self-identiﬁed liberals are outnumbered by those identifying themselves as conservatives, the ratio of liberals to conservatives is pretty much what it has been. State ideology has not undergone much change. The big changes have been in the gradual movement of state partisanship to align with state ideology.The South has been transformed from a one-party region to one in which the parties are about almost perfectly balanced in terms of partisanship. However, that movement toward the Republican party should continue unless the driving impact of ideology on state partisanship is broken.This is because the southern states remain

Dem.

33.3 33.1 39.7 21.8 32.5 32.8 25.3 24.8 34.7 32.2 24.7 43.0 28.7 35.7 29.0 42.7 30.8 29.7 29.8 26.8 16.1 30.6 27.2

State

Alabama Alaska Arizona Arkansas California Colorado Connecticut Delaware Florida Georgia Hawaii Idaho Illinois Indiana Iowa Kansas Kentucky Louisiana Maine Maryland Massachusetts Michigan Minnesota

29.6 42.0 29.2 36.0 26.5 37.5 39.7 37.5 28.0 30.3 36.3 35.8 33.0 33.4 40.0 29.2 23.5 25.2 44.4 25.0 50.0 33.8 36.9

Ind. 37.1 24.9 31.2 42.2 4.0 29.6 35.0 37.7 37.3 37.4 38.9 21.2 38.3 30.9 31.0 28.1 45.7 45.1 25.8 48.2 33.9 35.6 35.9

Rep. -3.8 8.1 8.5 -20.3 -8.5 3.2 -9.8 -12.8 -2.7 -5.2 -14.2 21.8 -9.6 4.9 -1.9 14.6 -14.9 -15.4 4.0 -21.4 -17.8 -5.0 -8.7

Mean 2,876 405 2,612 1,736 15,685 2,664 1,928 467 9,148 4,671 457 886 6,825 4,110 2,231 2,002 2,814 2,735 1,187 3,420 3,725 5,996 3,473

N 16.5 21.5 20.6 16.6 25.3 22.7 24.5 23.4 20.8 18.9 27.3 16.7 22.3 18.6 19.6 16.2 18.9 17.0 23.3 23.8 27.8 21.0 21.3

Lib. 43.9 41.4 45.9 42.8 46.7 46.0 50.0 49.4 47.3 43.8 45.4 44.8 48.5 43.5 45.4 45.4 44.7 42.3 45.0 48.0 47.8 47.1 49.2

Mod. 39.6 37.1 33.5 40.6 27.9 31.3 25.5 27.1 31.9 37.4 27.3 38.4 29.2 37.9 35.0 38.3 36.4 40.7 31.7 28.2 24.3 31.9 29.5

Con. -23.1 -15.6 -13.0 -24.0 -2.6 -8.6 -0.9 -3.7 -11.1 -18.5 0.0 -21.7 -6.9 -19.2 -15.4 -22.1 -17.5 -23.7 -8.4 -4.4 3.5 -10.9 -8.2

Mean

Appendix Table 12.1 Current Estimates for State Partisanship and Ideology ( from CBS/New York Times polls, 1996–2003) 2,737 391 2,515 1,649 14,956 2,516 1,850 431 8,652 4,460 447 861 6,606 3,933 2,105 1,879 2,619 2,541 1,108 3,177 3,487 5,882 3,359

N

250

Erikson, Wright, and McIver

more Democratic than we would predict on the basis of state ideology. What could this mean for presidential elections? Perhaps not much since Republicans are already winning most of the southern states. However, even more solid southern identiﬁcation with the Republican party would mean that there is little possibility for Democratic presidential victories to be fashioned out of winning in liberal coastal states while picking oﬀ a few southern states for an Electoral College majority. Underlying this is evidence that ideology has come to the fore in American national politics. Partisan cleavages provide a bit of electoral stickiness, but they appear to follow the fundamental diﬀerences between contemporary liberals and conservatives. We should anticipate, providing the parties do not undertake radical shifts in their overall strategies, that the battleground of future presidential elections will be deﬁned by the middle of the ideological spectrum, weighted, of course, by the Electoral College shares of the states. Our ﬁndings also raise some questions.We do not fully understand the processes of change in state partisanship, particularly the contributions of conversion versus generational replacement. Similarly, we have not examined the extent to which the sort of ideological alignment of the state electorates is reﬂected within the states. We would suspect a number of ramiﬁcations for state politics to follow from tighter ﬁts between general issues preferences and party identiﬁcations within the states.We examined mass polarization in Statehouse Democracy (1993, 39–42) and found a good deal of variation across the states. Predictably, in the southern states we found the weakest associations between ideology and party identiﬁcation. How and whether this has changed with the changing partisanship of the states needs to be investigated. That will be one of many avenues to be explored as we try to understand the full ramiﬁcations of the heightened importance of ideology and party polarization for state politics and policy as well as for politics on the national scene.

Notes Acknowledgments. Aggregation of the CBS News/New York Times polling data reported here was originally funded by the National Science Foundation. The data were obtained through the Interuniversity Consortium for Political and Social Research at the University of Michigan and the Roper Center for Public Opinion Research at the University of Connecticut.An early version of this chapter was originally presented at the 4th annual State Politics and Policy Conference May 1, 2004, Cuyahoga Falls, Ohio. The analysis and interpretations are the responsibility of the authors alone. 1. Erikson,Wright and McIver, 1993, 30–38.

Public Opinion in the States 251 2. The data collection for this project ends with 2003. However, we continue to collect data for 2004 and onward. 3. The CBS News/New York Times polling covered 48 states until 1995 when Alaska and Hawaii were added to the sampling frame. Most of the temporal analysis presented here is on the basis of 48.We note where Alaska and Hawaii are included in the statistical analysis. 4. These tables, as well as yearly estimates of state partisanship and ideology are available under “Data” at http://php.indiana.edu/wright1/.With these data one can generate estimates of partisanship and ideology for any combination of years, or even states. 5. Nonetheless, all parties compete for moderates and independents while working to secure and energize their base. 6. Just as we did in our very ﬁrst analysis of these state public preferences (Wright, Erikson, and McIver, 1985), for this particular graph we omit the state of Nevada. Erroneously, we believe, Nevada scored as a liberal in the early CBS/New York Times polls. Later Nevada samples look much more conservative but were still unusually erratic through the 1980s. 7. In the back and forth exchange of control within many states, we have the sequencing of maintaining and deviating elections (Campbell, Converse, Miller, Stokes, 1960, 531–533). A maintaining election “is one in which the patterns of partisan attachments prevailing in the preceding period persists and is the primary inﬂuence on forces governing the vote.” In contrast, Campbell and colleagues deﬁne a “deviating election” as an election where “the basic division of partisan loyalties is not seriously disturbed but the . . . forces on the vote are such as to bring about the defeat of the majority party.” 8. The stability of state public opinion has been the subject of some debate. In 1987, we argued that our measure of ideology based on late 1970s and early 1980s polls was stable,“conceivably, a reasonable reﬂection of state ideology throughout, say, most of the post-World War II era” (Wright, Erikson, and McIver, 1987). In Statehouse Democracy, we presented a similar portrait based on a slightly longer time frame (1976–1988): “the relative positions of the states in terms of ideological identiﬁcation appear to be almost perfectly stable” (1993, 37). In contrast, Berry et al. (1998) insisted that citizen ideology exhibits temporal variability and oﬀer an estimate of annual citizen ideology. Brace et al. (2002) compared our original results with the Berry et al. (1998) estimates and conclude that ideology is stable for most states. Here, we reiterate our sense that ideology is essentially stable. Much of the movement one can see in graphs such as Figure 12.2 above is the illusion of measurement error. 9. In each case we conduct the exercise while adjusting for reliability—using the “eivreg” program within STATA. 10. We are certainly not the only scholars or even the ﬁrst to comment on the changing South. But we have the opportunity here to clearly document the stateby-state change, noting that each state is not simply a clone of all others in the region but rather that each is unique.The states of the South diﬀer both in the degree of their Democratic party support at the start of the era and the extent to which they have transformed over the course of the last quarter century.

252

Erikson, Wright, and McIver

11. The process is well documented as having begun about a decade earlier with the success of the Goldwater campaign’s southern strategy. See Phillips (1969) for early speculation on the rise of the Republican party in the South. 12. The exception to the rule may be Florida. Most of the change in Florida comes early and it appears to have stabilized following the Reagan years. 13. A number of other states also follow this pattern but complicate the ﬁgure. 14. To achieve greater reliability in our measures we extended the time periods deﬁning “contemporaneous” state partisanship and ideology to include the previous two presidential administrations and then the administration following the election or a 12-year running average over the election.The two closest time periods were used to estimate opinion for the 1976 and 2004 regressions.

References Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 1960–1993.” American Journal of Political Science 42 ( January): 337–348. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46 ( January): 173–189. Campbell, Angus, Philip E. Converse, Warren E. Miller, and Donald E. Stokes. 1960. The American Voter. New York: John Wiley. Carsey, Thomas W., and Gerald C. Wright. 1998. “State and National Factors in Gubernatorial and Senate Elections.” American Journal of Political Science 42 ( July): 994–1002. Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Erikson, Robert S., Michael B. MacKuen, and James A. Stimson. 2002. The Macro Polity. New York: Cambridge University Press. Fiorina, Morris P. 2004. Culture War? The Myth of a Polarized America. New York: Pearson Longman. Gray, Virginia. 1976. “Models of Comparative State Politics: A Comparison of Cross-Sectional and Time Series Analysis.” American Journal of Political Science 20 (May): 235–256. Key,V. O. 1959. “Secular Realignment and the Party System.” Journal of Politics 21: 198–210. Lowery, David, Virginia Gray, and Gregory Hager. 1989. “Public Opinion and Policy Change in the American States.” American Politics Quarterly 17 ( January): 3–31. Norrander, Barbara. 2001.“Measuring State Public Opinion with the Senate National Election Study.” State Politics & Policy Quarterly 1(1): 111–125. Phillips, Kevin P. 1969. The Emerging Republican Majority. New Rochelle, NY: Arlington House. Stimson, James A. 1991. Public Opinion in America: Moods, Cycles and Swings. Boulder, CO:Westview.

Public Opinion in the States 253 Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. 1985. “Measuring State Partisanship and Ideology with Survey Data.” Journal of Politics 47 ( June): 469–489. ———. 1987.“Public Opinion and Policy Liberalism in the American States.” American Journal of Political Science 31 (November): 980–1001.

Chapter

13

Conclusions:Where We Have Been, Where Should We Go

Jeffrey E. Cohen

The chapters in Public Opinion in State Politics extend our understanding of the role of public opinion in democratic polities. Collectively, the chapters in this volume address three interrelated questions, all of which bear on the larger issues of the role of public opinion in democratic processes: (1) how do we measure state-level public opinion, (2) what are the inﬂuences on state level public opinion, and (3) what is the impact of state level public opinion on politics and policy making in the states? Furthermore, the Erikson, Wright, McIver chapter (Chapter 12) deals with dynamic properties of state opinion, a new and important direction for state public opinion research. While advancing our knowledge along all these fronts, the chapters herein also reﬁne old questions, raise new questions, and suggest new directions for future research.They also reveal frustrations and barriers in studying state-level public opinion and its role in democratic politics and policy making.

Measuring State Public Opinion The history of collecting state-level public opinion data has been one of compromise. Lacking large-scale surveys across the states, scholars have relied on essentially three data collection techniques—using demographics as surrogates, simulation, and pooling across existing surveys. Each poses its own limitations and forces compromises, which in the end, aﬀect 254

Conclusions 255

theory building and our understanding of the place of public opinion in state politics and policy making.

Demographic as Surrogate for Opinion The “demographic as surrogate” approach assumes that people of the same demographic category, or social placement, think alike. Obviously, demography aﬀects one’s political and social outlook. Demography structures one’s social circumstances, quality of life, and opportunities, which in many circumstance will aﬀect how a person views the political world, the policies that one prefers, and one’s attitudes about political leaders. But demography is not determinative; it is only one of several inﬂuences on a person’s political opinions. Although perhaps a minority view, some wealthy people support policies that will redistribute economic resources, to take just one example. Also, political opinions can be ﬂeeting and change in short order, whereas demographics are sometimes constant across a person’s lifetime (e.g., race, gender) or slow to change. A third criticism of the “demographic as surrogate” approach is that we may not possess data on all important or relevant demographic attributes to the political issue at hand. Finally, diﬀerent demographic characteristics may be important for understanding diﬀerent social and political attitudes. Economic class may speak to one’s attitudes about issues with an economic consequence, but may be irrelevant to issues without an obvious or direct economic dimension, such as abortion. The major limitations of the “demographic as surrogate” approach, which requires that one assume a close correspondence between demographic attributes and political opinion, lead scholars to search for other methods to measure opinion at the state level. Such eﬀorts, again, felt the constraint of not being able to run large surveys across the states.

Simulation Simulation proved a promising alternative to the “demographic as surrogate” approach, in that it combined information on the attitudes of diﬀerent demographic groups with the distribution of the groups within a state (Weber and Shafer, 1972; Weber, Hopkins, Mezey, and Munger, 1972–1973).The limited number of demographic categories used in early simulations, a function in part of computing issues, as well as the statistical methodologies used in the simulation, posed barriers to the development of the simulation approach (Seidman, 1973). Furthermore, some felt the assumption that demographic groups across states will hold similar opinion about an issue was untenable. Nothing short of actual state-level data, it seemed, would satisfy the scholarly community. After a short wave

256

Cohen

of enthusiasm in the 1960s and early 1970s, simulation of state opinion dried up. Before long, comparative state studies also waned as a growth subﬁeld; the lack of state-level public opinion being one of the factors that dampened activity in the subﬁeld.

Pooling Across Surveys With the innovative work of Erikson,Wright, and McIver (1993), pooling across many surveys to create large samples of respondents within states became the state of the art in estimating state-level opinion by the late 1980s.The commercialization of public opinion polling in the 1970s allowed the pooling technique. News organizations found that polls could be newsworthy; thus, many established their own polling operations in the 1970s and 1980s. Unlike academic polls, like the American National Election Studies (ANES) at the University of Michigan or the General Social Survey (GSS) of the University of Chicago, the commercial polls went into the ﬁeld often. Still, the pooling method has its own limitations. Despite the fact of the frequency of news organization and other nonacademic polls, news and economic factors, not theory, drove the questions posed to respondents. For the most part, such questions were timely as opposed to timeless, and few questions are repeated frequently enough to generate reliable statelevel estimates through pooling. Erikson,Wright, and McIver, consequentially, focused on two global political attitudes that the commercial ﬁrms asked respondents about repeatedly—party identiﬁcation and ideological identiﬁcation. News organizations also routinely asked respondents about presidential approval, but pooling on this question illustrates other inherent limitations of this methodology. Presidential approval often swings wildly and widely in short time frames. Pooling over extended time periods would wash out this short-term approval dynamic. Other than providing a sense of crossstate diﬀerences in presidential approval, the loss of the dynamic element drastically restricts the usefulness of a “pooled” presidential approval item. The pooling approach then seems best suited for attitudes and opinions that move quite slowly over time. Brace and his colleagues (2002, 2004) built on the Erikson,Wright, and McIver pooling methodology to construct state-level indicators of opinion on policy issues. Unlike Erikson, Wright, and McIver, who rely on commercial polls, Brace et al. (2002, 2004) turned to the GSS, an academic survey, which goes into the ﬁeld at least once a year for the past 30 years and frequently asked respondents about general policy issues with questions that are repeated enough to build state-level estimates. Using the

Conclusions 257

pooling methodology, Brace, Sims-Butler, Arceneaux, and Johnson (2002) generate nine speciﬁc policy indicators of state opinion—tolerance, racial integration, abortion, religiosity, homosexuality, feminism, environmentalism, welfare, and capital punishment. Brace and Johnson, in this volume (Chapter 2) and Uslaner (Chapter 8) extend this methodology to two other opinions, conﬁdence in the federal government and citizen trust. With such data at hand, we can compare the responsiveness of policy makers to these speciﬁc opinion indictors versus the more global partisanship and ideology indicators of Erikson, Wright, and McIver. In addressing this question at the national level,Wlezien (2004) ﬁnds that speciﬁc opinion inﬂuences some policies, notably defense and welfare, while global opinion aﬀects other policy areas, such as health, education, and the environment.Wlezien conjectures that speciﬁc policy attitudes aﬀect policies that are important and visible to the public. For policy areas of lesser importance to the public, global attitudes aﬀect policy decisions. Possibly the same process may work at the state level. Both the Erikson-Wright-McIver and Brace et al. pooling method, as noted, suﬀer from requiring relatively long time frames in order to collect enough cases per state to generate reliable estimates of state opinion. Consequently, both have been relegated to studying opinions that are relatively stable or that only shift slowly. Short-term dynamics are invisible, not so much due to the methodology employed, but due the infrequency with which relevant questions are posed to respondents. As Cohen and King demonstrate in the chapter on gubernatorial approval, short-time opinion dynamics also reveal much about state public opinion. Another problem with the pooling methodology is that national polls that are pooled typically do not use states as primary sampling units. This may aﬀect the validity and reliability of state opinion estimates, although Erikson, Wright, and McIver and Brace and colleagues painstakingly address these issues, and at least for the opinions that they look at, the state-level estimates appear reliable and valid.

Alternative Approaches Several approaches have been used, with varying success, to overcome the limitation of the pooling method in generating state-level opinion data. Using exit polls, Cohen and Barrilleaux (1993) are able to estimate state-level abortion opinion, albeit for only the time point when the exit poll was administered. In this book, Haider-Markel and Kaufman (Chapter 9) use a similar approach to measure state public opinion on gay and lesbian issues, relying on Lewis’s (2003) survey across 36 states. Barbara Norrander has made the greatest use of the ANES Senate

258

Cohen

Election Study (Norrander, 2000, 2001), including her chapter in this volume with Wilcox (also Norrander and Wilcox, 1999).The ANES’s Senate Election study used states as its primary sampling units, and over three election cycles (1988–1990–1992), polled residents in each state twice over the three elections. When pooled across the election cycle, reliable samples across all 50 states exist ( Jones and Norrander, 1996). Unlike most commercial polls used in other pooling exercises, respondents are asked a wide variety of political opinion questions. Furthermore, the Senate Election Study, because of the political focus of the surveys, assesses opinion across a wider variety of items that the GSS polls, which have interests beyond politics. Still, the great limitations of the ANES Senate Election Study is that such a data collection was conducted but once and to generate national coverage requires at a minimum two election cycles. Furthermore, these data have now become somewhat dated; even if many opinions are slow to change, the rapid population change in the United States over the past 15 years since the Senate Election studies may render opinion estimates obsolete as current indicators of state opinion. Last, the Senate Election Studies asked respondents about issues that all reached the national agenda. Some issues of importance to state agendas fail to rise onto the national agenda, such as gambling in the 1980s and 1990s. Berry, Ringquist, Fording, and Hanson (1998) oﬀer a technique that they suggests enables one to tap into the dynamic aspect of “citizen ideology,” in contrast to the static variables of Erikson,Wright, and McIver’s pooling method. Berry et al. use interest group ratings of candidates for congressional oﬃce and the percentage of the vote that competing candidates received to measure what they call citizen ideology.While we should expect some correspondence between the ideological preferences of voters and elected candidates, such a correspondence will not necessarily be “perfect.” People vote for candidates for reasons beyond a candidate’s ideological stance, for example, partisanship. Moreover, the reliance by Berry et al. on congressional elections allows incumbency eﬀects to enter into calculations of citizen ideology. In as much as incumbency aﬀords electoral advantages, another source of measurement error will be built into their indicator. Again, however, we see creative scholars searching for better ways of measuring state opinion and having to compromise due to data and resource limitations. Also, as exempliﬁed by Park, Gelman, and Bafumi in this volume (Chapter 11 and 2003), we may witness a revival of simulation (also see Ardoin and Garand, 2003, for a way to simulate congressional-district-level opinion and Berkman and Plutzer, 2004, for school district opinion). For a variety of reasons, after only a small number of attempts in the 1960s and early

Conclusions 259

1970s, simulation ceased as a major approach for estimating state-level opinion. But armed with enhanced computing power and new statistical techniques, Park, Gelman, and Bafumi have reopened the simulation door, demonstrating not only its feasibility, but also its potential eﬀectiveness in capturing short-term opinion dynamics, something that the pooling methodologies have a much harder time accomplishing. Finally, the JAR project of Beyle, Niemi, and Sigelman (2002; Cohen and King, in this volume) points in another direction.The JAR project collected the results of state-based polls across the state, conducted by a long list of polling organizations. As news organizations and other commercial ventures have found a commercial use for state polls and as polls has become less costly, due in large measure to the replacement of face-to-face with telephone interviewing, the volume of state polls has exploded. Here we have a collection of independently generated samples of state-based opinion, something that state public opinion scholars have long sought. Yet limitations inhere in the JAR approach. The JAR project collected data on approval of only three types of political ﬁgures—the president, governor, and U.S. senators.Two of these oﬃcials hold national, not state oﬃce. Furthermore, house eﬀects, question wording diﬀerences, and spottiness in the timing of state polls, raise issues of comparability across these approval polls. However, as pollsters query state publics about more and more topics, the potential for collecting data beyond the three approval sets of the JAR project stands as a strong possibility.

What to Measure Besides sampling and related issues reviewed above, the history of state public opinion research and the chapters in this volume also raise the question of what aspects of state public opinion one wants to measure. For the most part, past research has focused on opinions that are relatively stable over time, and thus, amenable to the increasingly common pooling methodology.Yet not all opinions that we want to know something about are stable enough over periods long to employ the pooling methodology, as the approval data from the JAR project and Cohen and King’s chapter in this book remind us.

Global versus Specific Political and Policy Preference Not surprisingly, due to both theoretical interests and data availability, most research on state public opinion has focused on global political attitudes, like partisanship and ideological identiﬁcation (Erikson, Wright,

260

Cohen

McIver, 1993; Berry et al., 1998) or attitudes on speciﬁc public policies (Brace et al., 2002; Cohen and Barrilleaux, 1993; Norrander, 2000, 2001). One can point to democratic theory to justify such an emphasis—we want to know if policy makers are responsive to the citizens’ policy preferences, and if so, under what conditions policy makers are responsive to global versus speciﬁc opinions (Wlezien, 2004). Subtle diﬀerences in democracy exist if policy makers are more responsive to global versus speciﬁc opinion. First, a larger segment of the public may have a voice if leaders tend to respond more to patterns in global opinions. As survey research informs us, “don’t know” rates are smaller, especially for partisanship, that for most speciﬁc opinions. In as much as “don’t know” responders diﬀer systematically from those with opinions, policy outcomes might also diﬀer. Second, policy makers possess greater opportunities to steer speciﬁc than global opinion. On speciﬁc issues, politicians may use rhetorical devices to frame an issue and take advantage of the generally low level of public information to inﬂuence public thinking on the issue. Publicity campaigns and “crafted talk” ( Jacobs and Shapiro, 2000) may also be employed. In contrast, global attitudes, especially partisanship, are rooted in deeper socialization processes and thus may be more resistant to shortterm change than opinion on speciﬁc issues, which are more easily redeﬁned, although the dynamic partisanship studies indicate that shortterm movement in party identiﬁcation. However, the distinction between speciﬁc attitudes and global attitudes, especially ideological orientation, may not be as stark as argued in the previous paragraph. Page and Shapiro’s (1992) rational public notion suggests that aggregate opinion at the national level tends to shift only slowly, incrementally, and in response to identiﬁable factors. Presumably the same holds at the state level. Furthermore, from Stimson’s (1999) public mood concept, speciﬁc issues positions and global ideology may be tightly linked, at least if we do not deﬁne ideology in the mass public as ideological sophistication, but rather liberal-conservative preferences.The study of state public opinion opens up the possibility of addressing these types of questions of democratic responsiveness.

Citizen System-Level Opinions Brace and Johnson (Chapter 2) and Uslaner (Chapter 8), in this volume, also point out that there are global attitudes besides partisanship and ideology, for instance those that focus on system-level properties, like government conﬁdence and generalized trust (trust in other people). Brace and Johnson ﬁnd that the penetration of the federal government into a

Conclusions 261

state aﬀects the level of conﬁdence in the national government. Although the amount of land that the federal government owns in a state suppresses levels of conﬁdence, the number of federal workers boosts trust. One wonders how conﬁdence in the federal government aﬀects conﬁdence in state government. Does conﬁdence at the two levels reinforce each other or are there trade-oﬀs, with higher federal conﬁdence undercutting conﬁdence in state government? Answering such questions is important to the debate on the proper level of government, whether or not it is “better” to locate government closer to the citizens, in smaller and local units. Unfortunately, the GSS government conﬁdence questions that Brace and Johnson do not ask about state government. Uslaner ﬁnds that when people trust other people, a concept related to Putnam’s (2000) idea of social capital, that government performance, the ability to eﬀectively deliver services, improves. This may be because such trust creates an environment that fosters greater cooperation between citizens and the government and lends greater legitimacy to government action than when people are distrustful of others. Other research by Uslaner (2002) looks at the sources of social trust. One wonders whether government performance, including levels of corruption (as Uslaner himself identiﬁes) may also foster higher levels of social and civic trust or mediating attitudes that encourage such trust. In other words, social and civic trust may interact with government performance.The development of dynamic measures of social trust and government performance, in the context of comparative state studies, may help sort out some of the causal properties that link trust and government performance, as well as tie to the point of Norrander and Wilcox (Chapter 3) and Leal (Chapter 4), in this volume, that state of residence may aﬀect public opinion, just as public opinion may aﬀect the behavior of political leaders and the policies that governments produce. In other words, each state is a distinguishable political system, with diﬀerent histories and political fault lines, and may socialize citizens to politics somewhat diﬀerently.

Attitudes Toward Leaders and Policy Performance The state public opinion literature has emphasized the linkage between opinions on issues and global political attitudes. Less attention has been paid to public evaluations of political leaders, which provides an alternative route of democratic linkage to the opinion-policy connection. Research shows that popularity predicts electoral prospects for gubernatorial candidates (Kenney and Rice, 1983; King, 2001) and may also aﬀect gubernatorial success with the state legislature (Ferguson, 2003).Thus, job

262

Cohen

approval and popularity provide democratic linkage through the election of oﬃce holders as well as by mediating the ability of governors to see their policy initiative enacted into law. But job approval and popularity also provide governors and other politicians with information on their level of public support between elections. Politicians may alter their behavior in oﬃce in response to trends in approval, thus promoting leader responsiveness between elections. But compared to the modest literature that focuses on the opinionpolicy linkage, the literature on leader evaluation at the state level is even more modest. In part, the relative lack of attention to leader evaluations stems from the paucity of data. As Cohen and King’s review of the literature on gubernatorial popularity in this book presents, past research has been restricted to comparisons of at best a handful of states (e.g., Cohen, 1983; Crew, 1998; Hansen, 1999a, 1999b) or one-shot, 50-state crosssectional comparisons (MacDonald and Sigelman, 1999). The JAR data set (Beyle, Niemi, and Sigelman, 2002) rectiﬁes this situation, but not completely. First, the JAR data set has compiled job approval and popularity questions across a multitude of survey organizations that diﬀer in sampling procedures, interview instruments, and questions, all of which may present issues of comparability. Second, the resulting data sets are somewhat spotty. As Cohen and King report, from a pooled, crosssectional time series design perspective, gubernatorial job approval readings exist for only about 10 percent of possible data points.Third, besides the gubernatorial approval component of the JAR data set, the rest of the data focuses on political leaders who are not purely state oﬃce holders (the president and U.S. senators). Perhaps the most glaring deﬁciency of the available job approval data at the state level is that once we go to oﬃces beyond the governor, very little data exists. But state government in the United States is not gubernatorial government, much as government at the national level is not presidential government. Other constitutionally empowered branches of government exist, yet we know very little about public evaluations of the legislature and the judiciary. In this book, Hamman’s study (Chapter 5) on state legislative approval illustrates the frustrations of conducting research on such an important topic. Hamman could ﬁnd only 124 polls across 13 states that ask about legislative job performance, and much of his analysis turns to an individual-level investigation that pools two national polls, which just happen to ask about the state legislature. Finally, as Barrilleaux (Chapter 7) demonstrates, we need to think beyond central tendency in creating measures of aggregate opinion. States with similar mean opinion may vary with regard to dispersion or diversity of opin-

Conclusions 263

ion around that mean. From a democratic theory perspective, it is relatively easy to represent a political unit when people hold similar opinions. But when opinion is spread, faithful representation becomes more problematic. For instance, consider an opinion distribution in which the public is evenly divided into two camps that are quite far apart (i.e., very liberal versus very conservative). Mean opinion may appear moderate, but few of the state’s citizens may hold moderate opinions. A representative who pursues a moderate policy course may oﬀend both the very liberal and the very conservative on the policy in question. The distribution of opinion may be as important as its central tendency in understanding democratic linkages.

New Data Collection Directions This review of the literature points out several major limitations of state-level public opinion research. First, we need cross-state data that also allows one to measure the dynamic properties of the opinions in question. From a substantive perspective, we still lack data on a large number of issues and job performance evaluations for policy makers and institutions besides the governor. In recent years, massive academic polling projects, like the 2000 and 2004 National Annenberg Election Studies (NAES) of the Annenberg Public Policy Center of the University of Pennsylvania, oﬀer some promise, especially for detecting short-term dynamics of some global and speciﬁc policy opinions. In 2000, NAES polled over 79,000 respondents across over 100,000 interviews. In 2004 a comparable number of respondents were polled. But the NAES design focuses exclusively on national politics associated with the presidential election campaigns, and thus, may not oﬀer enough items of interest to the state public opinion subﬁeld. The Holy Grail for state public opinion research is still 50 state surveys built especially for the concerns of state politics research, administered with enough frequency to capture some of the dynamics of state-level opinion. Using traditional survey research methods, such a project would be too expensive. But Internet polls, such as those of Knowledge Networks and Harris Interactive, oﬀer an alternative that might be practical and aﬀordable.We are still in the infancy of using the Internet to collect public opinion data.

Inﬂuences on State Public Opinion Less attention has been paid to the question of what inﬂuences statelevel opinion, in part because of the sheer diﬃculty in collecting statelevel opinion data, as reviewed in the previous pages. The answer to this

264

Cohen

question has profound implications for our understanding of state public opinion’s role, but also of the nature of states as political units and the impact of political and other institutions on citizens more broadly. We can identify two major sources of state-level opinion. First, state opinion may be merely a function of the social, economic, and other circumstances in which a person lives. From this perspective state of residence matters little. People with similar life circumstances in one state, say Alabama, will hold essentially the same political outlook as “identical” residents in other states, for instance, New York or California.A contrary perspective suggests that state residence independently inﬂuences one’s political opinions and outlook. Hence, “identical” people who reside in diﬀerent states will hold somewhat diﬀerent opinions.There is some support for the “state matters” perspective. Erikson, Wright, and McIver (1993) ﬁnd that state residence accounts for some of the variance in public opinion. In this volume, Norrander and Wilcox make a similar point. If state matters, what is it about state residence that aﬀects public opinion? First, the institutions and policies of a state may be strong enough to aﬀect the way that the public thinks about a range of matters political. A state’s political (and social and economic) institutions, including its public policies may be strong enough to direct behavior, and over time, alter public opinion. Here we may think of institutional strength as the ability of political institutions and the news media to penetrate into the daily life of its citizenry, borrowing roughly from the political development literature. Other psychological processes may also be at work, for instance, local pride, akin to patriotism, which opens up the citizen to the inﬂuence of a state’s political institutions and policies. Second, through population movement, people may move into a state because they ﬁnd the political climate congenial. Rather than the socialization processes alluded to above, self-selection processes may account for the impact of state residence on public opinion. But here, the motivation and source of immigration may be critical. Some may immigrate to a state for purely economic reasons, such as job availability, or to leave past circumstances behind, rather than because they ﬁnd the “lifestyle,” including a state’s politics and policies, congenial. Immigrants who feel forced to leave their homes may be more resistant to the “political culture” of their new state residence.They may also import with them their political and cultural heritage, which may alter the state if enough members of their group also immigrate. How much impact did the movement of southern blacks from the 1920s through 1940s have on northern politics? How much did their new residence in northern states change the political behavior and opinion of these southern blacks? Population dynamics and movements cut both ways.They can be a source

Conclusions 265

of state inﬂuence over public opinion, but also open up the state to new forces that transform state politics and policy. The study of public opinion at the state level can help illuminate these cross-cutting patterns, but to make much progress on this front we probably need better data than that which already exists. Yet answering the question of whether and how state residence aﬀects public opinion reveals much about the nature of states as political units in the United States, and bears on questions of how citizens related to their governments at all levels and on political development. In the latter senses, such comparative state studies have much in common with the comparative politics and political development subﬁelds.

Impact of State Public Opinion on State Politics and Policy Last, in this conclusion I turn to more substantive questions, speciﬁcally the role of public opinion in democratic politics and policy making.The general point that emerges from the studies in this volume is that state public opinion aﬀects governance and policy making in the states. Such a blanket and general statement is none too useful, however. We need to specify the extent of opinion impact and the conditions that promote or reduce public opinion’s impact. Several of the studies in this volume deal with these more speciﬁc questions and identify important conditioning factors. As such they lay a foundation for future research on the question of the importance of public opinion in state politics and policy making, which as I have been arguing throughout is part of the large question of the role of public opinion in democratic politics and policy making. Eric Uslaner (Chapter 8) suggests that the social trust climate of opinion may aﬀect the performance of government—its ability to eﬃciently and eﬀectively implement and deliver public policies.That the climate of trust may aﬀect interpersonal relationships among citizens, as well as the relationship between the citizen and the state, is often forgotten or missing in studies that look at the linkage between public opinion and the polity.The lack of social trust may increase the costs of policy implementation and delivery to governments. Gaining the cooperation of citizens may be more diﬃcult than when citizens are more trusting of each other and social and political institutions. Lack of social trust may create a barrier to the eﬀective use of bureaucratic knowledge, information, and expertise. Policy breakdowns may ﬂow from such a context, which may fuel discontent with incumbents in oﬃce and lead to their electoral replacement. Social and other forms of trust may be an important mediating variable in the linkage between the public and its government.

266

Cohen

Haider-Markel and Kaufman (Chapter 9) deal with a more classical perspective, asking whether the policy opinions of citizens inﬂuence the shape of public opinion. They look speciﬁcally at gay-related policies, a social policy area that has risen to the top of the public agenda in recent years, and thus one in which one would expect to ﬁnd strong public impacts on resulting policy. Not so surprisingly, Haider-Markel and Kaufman ﬁnd that public opinion toward gays and lesbians aﬀects some related policies but not others.They argue that the shape of public opinion conditions the impact of public opinion on policy. Speciﬁcally, when opinion is contentious or divided, public opinion will more strongly aﬀect policy than when it is lopsided or consensual. When opinion is lopsided, HaiderMarkel and Kaufman argue that the inﬂuence of other factors, like policy entrepreneurs will grow, although lopsided opinion will constrict the range of acceptable policies that can be pursued.When opinion is divided, policy makers have to work harder to mobilize opinion in their behalf in policy debates. Obviously, the insights of their chapter open up a host of possibilities for future research. Schneider and Jacoby (Chapter 10), too, look at a classic question of public opinion impacts, the relative inﬂuence between public opinion and interest groups in policy making. Past research has often couched this issue as a trade-oﬀ—when public opinion inﬂuence is weak, interest group inﬂuence is strong, and vice versa. Phrasing the question this way implies that public opinion and interest groups are always in opposition on issues. This may be too simplistic of an understanding of the relationship between interest groups and public opinion. Schneider and Jacoby ﬁnd that sometimes public opinion and interest groups reinforce each other and that interest groups may be the vehicle of public opinion expression in some instances.This is presumably because the broad ideological and partisan opinion environment provides fertile ground for the formation of groups compatible with that environment. Thus, group formation and mobilization is in part a product of the opinion climate. But opinion will only aﬀect policies of broad public concern. On more particularistic issues, public opinion registers little inﬂuence on policy outcomes. It would be of interest to couple these insights with Haider-Markel and Kaufman’s concerning the shape of public opinion.

Conclusion State politics is a fertile area in which to study many important questions of democratic theory and processes. This book emphasizes the role

Conclusions 267

of public opinion in state politics and policy. Although an old and venerable topic, the dearth of data limited progress in the ﬁeld. As the chapters in this book demonstrate, new methodologies from pooling to simulation and new data collection enterprises, which take advantage of the explosion in polling (e. g., the JAR project), have allowed researchers to address questions more pointedly than was the case as little as two decades ago. But as these studies also reveal, many roadblocks in studying state public opinion and its place in state politics still exist.We lack data on many important political opinions, and current methods of estimating state public opinion all possess certain limitations. Hopefully the research in this book will stimulate new eﬀorts on the question of state public opinion.

References Ardoin, Philip J., and James C. Garand. 2003. “Measuring Constituency Ideology in U. S. House Districts: A Top-Down Simulation Approach.” Journal of Politics 65 (November): 1165–1189. Berkman, Michael, and Eric Plutzer. Forthcoming. Ten Thousand Democracies: Politics and Public Opinion in America’s School Districts. Washington, DC: Georgetown University Press. Berkman, Michael, and Eric Plutzer. 2004. “School Budgets and Local Teacher’s Unions:The Mediating Role of Local Institutions.” Paper presented at the annual meeting of the American Political Science Association Hilton Chicago and the Palmer House Hilton, Chicago, IL. Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson. 1998. “Measuring Citizen and Government Ideology in the American States, 1960–93.” American Journal of Political Science 42(1): 327–348. Beyle,Thad, Richard G. Niemi, and Lee Sigelman. 2002. “Gubernatorial, Senatorial, and State-Level Presidential Job Approval Ratings: The U.S. Oﬃcials Job Approval Ratings ( JAR) Collection.” State Politics and Policy Quarterly 2 (Fall): 215–229. Brace, Paul, Kevin Arceneaux, Martin Johnson, and Stacy G. Ulbig. 2004. “Does State Ideology Change over Time?” Political Research Quarterly 57 (December): 529–540. Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. 2002. “Public Opinion in the American States: New Perspectives Using National Survey Data.” American Journal of Political Science 46 ( January): 173–189. Cohen, Jeﬀrey E. 1983. “Gubernatorial Popularity in Nine States.” American Politics Quarterly 11 (April): 219–235. Cohen Jeﬀrey E., and Charles Barrilleaux. 1993.“Public Opinion, Interest Groups, and Public Policy Making: Abortion Policy in the American States.” In Malcolm L. Goggin, ed., Understanding The New Politics of Abortion. New York: Sage. Crew, Robert E., Jr. 1998. “Gubernatorial Leadership: Testing a Preliminary Model.” Social Science Journal 36(1): 15–27.

268

Cohen

Erikson, Robert S., Gerald C. Wright, Jr., and John P. McIver. 1993. Statehouse Democracy: Public Opinion and Policy in the American States. New York: Cambridge University Press. Ferguson, Margaret Robertson. 2003. “Chief Executive Success in the Legislative Arena.” State Politics & Policy Quarterly 3 (Summer): 158–182. Hansen, Susan B. 1999a. “Governors’ Job Performance Ratings and Unemployment:The Case of California.” State and Local Government Review 31 (Winter): 7–17. ———. 1999b. “‘Life is not Fair’: Governors’ Job Performance Ratings and State Economies.” Political Research Quarterly 52 (March): 167–188. Jacobs, Lawrence R., and Robert Y. Shapiro. 2000. Politicians Don’t Pander: Political Manipulation and the Loss of Democratic Responsiveness. Chicago: University of Chicago Press. Jones, Bradford S., and Barbara Norrander. 1996. “The Reliability of Aggregated Public Opinion Measures.” American Journal of Political Science 40 (February): 295–309. Kenney, Patrick J., and Tom W. Rice. 1983. “Popularity and the Vote: The Gubernatorial Case.” American Politics Quarterly 11 (April):237–241. King, James D. 2001. “Incumbent Popularity and Vote Choice in Gubernatorial Elections.” Journal of Politics 63 (May): 585–597. Lewis, Gregory B. 2003. “Contentious and Consensus Gay Rights Issues: Public Opinion and State Laws on Discrimination and Same-Sex Marriage.” Paper presented at the annual meeting of the American Association for Public Opinion Research, May, Nashville,TN. MacDonald, Jason A., and Lee Sigelman. 1999. “Public Assessments of Gubernatorial Performance: A Comparative State Analysis.” American Politics Quarterly 27 (April): 201–215. Norrander, Barbara. 2000.“The Multi-Layered Impact of Public Opinion on Capital Punishment Implementation in the American States.” Political Research Quarterly 53 (December): 771–793. ———. 2001.“Measuring State Public Opinion with the Senate National Election Study.” State Politics & Policy Quarterly 1 (March): 113–127. Norrander, Barbara, and Clyde Wilcox. 1999. “Public Opinion and Policymaking in the States: The Case of Post-Roe Abortion Policy.” Policy Studies Journal 27 (December): 707–722. Page, Benjamin I., and Robert Y. Shapiro. 1992. The Rational Public: Fifty Years of Trends in Americans’ Policy Preferences. Chicago: University of Chicago Press. Park, David K., Joseph Bafumi, and Andrew Gelman. 2003. “Bayesian Multilevel Estimation with Poststratiﬁcation: State-Level Estimates from National Polls.” Political Analysis 12 (Autumn): 375–385. Putnam, Robert D. 2000. Bowling Alone:The Collapse and Revival of American Community. New York: Simon and Schuster. Seidman, David. 1973. “Simulation of Public Opinion: A Caveat.” Public Opinion Quarterly 39 (Autumn): 331–342. Stimson, James A. 1999. Public Opinion in America: Moods, Cycles, and Swings, 2nd ed. Boulder, CO:Westview.

Conclusions 269 Uslaner, Eric M. 2002. The Moral Foundations of Trust. New York: Cambridge University Press. Weber, Ronald E.,Anne H. Hopkins, Michael L. Mezey, and Frank Munger. 1972– 1973. “Computer Simulation of State Electorates.” Public Opinion Quarterly 36 (Spring): 49–65. Weber, Ronald E., and William R. Shafer. 1972. “Public Opinion and American State Policy-Making.” Midwest Journal of Political Science 16(4): 683–699. Wlezien, Christopher. 2004. “Patterns of Representation: Dynamics of Public Preferences and Policy.” Journal of Politics 66 (February): 1–24.

Index

Abelson, Robert P., 8, 39, 183, 209, 209n1, 210 Abramowitz,Alan I., 39 Abramson, Paul R., 22, 86 Achen, Christopher H., 225 Adams, Greg, 103, 115 Adams, James D., 108 Aldrich, John H., 22 Almond, Gabriel, 143 Alt, James E., 109 Altschuler,Alan A., 146 Anderson, Christopher J., 21 Anderson, James E., 81n5 Anderson, Jennifer L., 8 Antunes, George, 61n2 Arce, Carlos H., 59 Arceneaux, Kevin, 11, 23, 30, 39, 121, 149, 163, 210, 211, 211n5, 237n8, 256, 257 Archer, J. Clark, 38 Ardoin, Phillip J., 9, 258 Atkeson, Lonna Rae, 103, 110 Atkinson, Michael M., 21 Bachrach, Peter, 197 Bafumi, Joseph, xiii, 9, 11, 14, 258, 259 Baird,Vanessa A., 21, 29 Baker v. Carr, 80 Banerjee, Neela, 156

Banﬁeld, Edward, 124 Baratz, Morton, 197 Bardwell, Kedron, 8 Barreto, Matt, 56 Barrilleaux, Charles, 13, 123, 126, 133, 184, 257, 260, 262 Bartels, Larry M., 225 Barth, Jay, 8, 115 Baumgartner, Frank R., 186 Bean, Frank, 54 Berger, Ronald J., 38 Berkman, Michael, 9, 80, 121n1, 258 Berman, David R., 38 Bernick, Ethan, 8, 103 Berry, Frances, 166, 172, 185 Berry,William D., 38, 121n1, 166, 170–173, 171n5, 183, 185, 195, 196, 199, 199n5, 230, 237n8, 258, 260 Best, Nicky, 213 Beyle,Thad, 7, 8, 84, 103, 106, 106n6, 106n8, 107, 109, 259, 262 Bibby, John, 126 Binder, Sarah A., 144, 151 Bobo, Lawrence, 61n2 Bond, Jon R., 107 Boswell,Thomas, 58 Bowler, Shaun, 20, 85

272

Index

Boylan, Richard T., 143, 145, 146, 157 Boynton, G. Robert, 80 Brace, Paul, xiii, 3, 6, 7, 11, 22, 23, 27, 30, 38, 39, 104, 109, 121, 121n1, 149, 163, 165, 210, 211, 211n5, 237n8, 256, 257, 260, 261 Bradley, Martin B., 171n6 Brady, Henry, 61n2, 64 Branham, David, 8, 103 Brehm, John, 159 Bremer, John, 105, 110 Brischetto, Robert, 58 Brown, M. Mitchell, 149n3 Bryk,Anthony S., 212n8 Buchanan,William, 88 Budge, Ian, 123 Burden, Barry C., 39 Burdick, Eugene, 209n2 Burnham,Walter Dean, 165 Burstein, Paul, 164, 184 Bush, George H.W., 214, 216, 220, 232, 243 Bush, George W., 56, 63, 65, 70, 71, 232, 234, 236, 238, 240, 244, 245 Bush, Jeb, 63, 70 Business International, 158 Button, James W., 171 Cain, Bruce, 60, 61n2, 64n6, 165 Caldeira,Gregory A., 21, 29, 80, 84, 126 Call, Jack E., 38 Calvo, Maria Antonia, 61n2 Campbell,Angus, 202, 203, 237n7 Campbell, James E., 38 Caraley, Demetrios, 28 Carlin, John B., 212n8 Carmines, Edward G., 25, 81, 165, 166, 170 Carsey,Thomas M., 38, 103, 110, 238 Carter, Jimmy, 221, 222, 232, 234, 236, 238, 240, 241, 246 Cawley, L. MacGregor, 28 Chicano Survey, 59 Chubb, John E., 123 Citrin, Jack, 19, 20, 22

Clarke, Harold D., 21 Clinton, Bill, 63, 70, 71, 214, 232, 236, 240, 243, 245, 246 Cohen, Jeﬀrey E., 8, 11–13, 103, 257, 259, 260, 262 Coleman,William D., 21 Conley, Richard S., 151 Conover, Pamela Johnston, 105 Converse, Philip E., 195, 202,203, 237n7 Cook, Beverly B., 165 Cook, Elizabeth Adell, 39, 40 Coombs, Clyde H., 188 Corruption, 144, 145, 146, 156–159 Cotter, Patrick, 79, 80, 85, 87n14, 89, 91 Cox, Michael A.A., 188 Cox,Trevor, 188 Craig, Stephen C., 86 Crew, Robert E., Jr., 8, 102, 103, 105, 109, 115, 262 Cuban-American National Foundation, 57 Cuban-Americans, 12, 53–63, 68–74. See also Latinos Curtis, James, 58 Dahl, Robert A., 125, 196, 202 Davidson, Roger H., 79 Davis, Gray, 102 Dawson, Richard E., 5, 183 Daynes, Byron W., 166 de Castro, Rafael Fernandez, 56 De la Garza, Rodolfo, 53, 54, 59–61 de Leeuw, Jan, 212n8 Deitch, Cynthia, 63 Del Olmo, Frank, 55 Delli Carpini, Michael X., 81, 110 Democratic theory, 4, 14, 163, 225, 254, 260–263, DeSipio, Louis, 53,54 Diamond, Sara, 171 Dionne, E. J., 19 Divided government, 90, 109, 113, 114, 151, 152

Index 273 Dometrius, Nelson C., 8, 102 Domínguez, Jorge I., 54, 55, 58 Donovan,Todd, 20, 85 Duan, Naihua, 212n8 Dukakis, Michael, 214 Durr, Robert H., 79, 83, 95n19 Dye,Thomas R., 5, 123, 165, 183 Edelson, Jonathan L., 166, 173, Edwards, George, 165 Elazar, Daniel, 5n1, 40, 124, 191 Eliassen, Kjell A., 80, 80n3 Ellison, Christopher, 61n2 Erikson, Robert S., xiii, 3, 4, 6, 7, 11, 13–15, 25, 30, 38, 40, 43, 63, 64, 89n16, 104, 107, 121, 121n1, 122, 123n2, 127, 128, 128n3, 131–133, 131n5, 136, 137, 149, 163–165, 183, 184, 186, 187, 191, 193, 195, 196, 199, 199n5, 210, 210n4, 211, 221, 223–225, 229, 230n1, 233, 234, 234n6, 237, 237n8, 240, 250, 254, 256–260, 264 Eulau, Heinz, 88 Evangelium Vitae (Encyclical Letter), 62 Eyestone, Robert, 172, 185 Falcon,Angelo, 61 Fastnow, Christina, 104, 109 Federalism, 11, 12, 19, 20, 21, 27, 28, 31, 32 Feldman, Stanley, 105 Fellowes, Matthew, 184, 204 Ferguson, LeRoy C., 88 Ferguson, Margaret R., 8, 102, 115, 261 Ferree, Karen E., 109 Finkel, Steven E., 39 Fiorina, Morris P., 21n1, 123, 231, 246 Fleisher, Richard, 107 Ford Foundation/Kennedy School innovations awards, 146, 147, 152, 154, 155 Ford, Gerald, 232 Fording, Richard C., 38, 121n1, 170, 184, 195, 196, 199n5, 230, 237n8, 258, 260 Fraga, Luis, 54

Fuchs, Lawrence, 54 Furman v. Georgia, 46 Gaitz, Charles, 61n2 Garand, James C., 9, 104, 105, 123, 185, 187, 258 Garcia, F. Chris, 61 Gay, David, 61n2 Gays and Lesbians, 163–178, government policies and laws, 166–179, 266 Gelman,Andrew, xiii, 9, 11, 14, 211, 211n6, 212, 212n8, 214, 215, 258, 259 Gerber, Elisabeth R., 165 Getter, Russell W., 165 Gibson, James L., 21, 29 Gill, Jeﬀ, 212n8 Gilliam, Franklin Jr., 61n2 Gilmour, John B., 79 Globetti, Suzanne, 22, 32 Goldstein, Harvey, 212n8 Gore,Al, 63, 65, 70, 71 Gormley,William T., 165 Government eﬀectiveness, 142–159, 260, 261, 265 Government performance. See government eﬀectiveness Grady, Dennis, 103, 104 Grattet, Ryken, 167, 169 Gray,Virginia, 163, 172, 184, 185, 197, 204, 230 Greene, Stephen, 39 Gronke, Paul, 104, 105n1, 108n13, 225 Grumm, John G., 80 Guillory, Christine A., 21 Gurin, Gerald, 61n2 Gurin, Patricia, 61n2 Hagen, Michael G., 186 Hager, Gregory, 163, 184, 230 Haider-Markel, Donald P., 13, 14, 165–167, 167n2, 167n3, 169–173, 175, 177, 178, 257, 266 Hall, Melinda Gann, 38

274

Index

Haller, H. Brandon, 105 Hamm, Keith E., 85 Hamman, John, 10, 12, 262 Hansen, Susan B., 103, 105, 105n2, 109, 115, 262 Hanson, Russell L., 38, 121n1, 170, 184, 187, 195, 196, 199, 199n5, 230, 237n8, 258, 260 Harris Interactive, 263 Hartley,Thomas, 164 Heck, Ronald H., 212n8 Heclo, Hugh, 202 Hedlund, Ronald D., 80 Heel, Michael, 105, 110 Hendrick, Rebecca M., 104, 185, 187 Hero, Rodney E., 28, 124, 125, 128n4, 132, 136, 137, 149n4 Herzik, Eric B., 104 Hetherington, Marc J., 20, 22, 32 Hibbing, John R., 19–21, 23, 79, 81, 95n19 Hill, Kim Q., 123, 165, 170, 186, 210 Hinckley, Barbara, 22, 109, 165 Hinton-Anderson,Angela, 123, 165, 170, 186 Hispanics. See Latinos Hoﬀerbert, Richard I., 5, 81, 123, 183, 186, 187, 210 Holbrook,Thomas M., 38, 105, 123, 126, 133, 146, 170, 170n4, 184 Holbrook-Provow,Thomas M., 121n1 Hopkins,Anne H., 8, 39, 121n1, 183, 186, 209, 210, 255 Howell, Henry, 156 Howell, Susan E., 103, 111n15 Hrebenar, Ronald, 130, 131, 171 Inglehart, Ronald, 143 Interest groups, 14, 39, 158, 170, 171, 175, 176, 177, 184, 195–197, 199–204, 258, 266 Jackman, Robert W., 21, 158 Jackman, Simon, 213n10, 215n14 Jackson, Robert A., 38

Jacobs, Lawrence R., 260 Jacoby,William G., 14, 186, 189, 190, 195, 197, 203, 266 Jelen,Ted C., 39, 40 Jenness,Valerie, 167, 169 Jennings, Edward T., 123 Jewell, Malcolm E., xiii, 79–81, 85, 88, 91 Jewett,Aubrey, xiii Johnson, Cathy M., 38 Johnson, D.W., 171n6 Johnson, Lyndon, 242 Johnson, Martin, 6, 11, 23, 30, 39, 121, 149, 163, 210, 211, 211n5, 237n8, 256, 257, 260, 261 Jones, Bradford S., 7, 25, 38–40, 258 Jones, Bryan D., 186 Jones-Correa, Michael, 53 Kagay, Michael, 149n3 Kane, Melinda, 165–167, 171, 172, 174 Kang, Insun, 8 Karnig,Albert K., 81 Karp, Jeﬀrey A., 82, 86 Kaufman, Matthew, 13, 14, 257, 266 Kaufmann, Daniel, 145 Keefe, Susan, 55 Keefer, Philip, 143, 156 Keen, Lisa, 169 Keeter, Scott, 81, 110 Kennedy, John F., 209 Kenney, Patrick J., 102, 261 Kenny, Lawrence W., 108 Kernell, Samuel, 109n14 Kerry, John, 245 Key,V. O., 5, 122, 123, 241 Kiewiet, D. Roderick, 60, 61n2, 64n6 Kim, Mark T., 143, 145, 148, 149, 158 Kinder, Donald R., 40, 61, 63 King,Anthony, 80, 80n3 King, David C., 143, 145, 148, 149, 149n4, 158

Index 275 King, Gary, 211, 214 King, James D., 8, 11–13, 80, 90, 102, 103, 105, 110, 123, 257, 259, 261, 262 Kingdon, John W., 186 Kirkpatrick, David D., 169 Klingman, David, 186 Knack, Stephen, 143–145, 148, 149, 151, 152, 156 Knight, Kathleen, 105 Knowledge Networks, 263 Kornberg,Allan, 21 Kray,Aart, 145 Kreft, Ita G., 212n8 Lammers,William W., 186 Lane, Robert E., 143 Langer, Laura, 38, 121, 123, 126, 133, 184 LaPorta, Rafael, 143, 156 Lascher, Edward L. Jr., 186 Latino National Political Survey, 54, 55, 58– 60 Latinos, 12, 53–74 Layman, Geoﬀrey C., 170 Leal, David L., 9, 12, 53, 61, 64n6, 261 Leduc, Lawrence, 21 Lee, Mei-Hsien, 39, 165, 166, 170, 172, 176, 178 Legislative professionalism, 79–84, 86–89, 91, 93–95, 149, 158 Leighley, Jan E., 123, 210 Lewis, Dan A., 185 Lewis, Gregory B., 166, 167, 170, 172, 173, 176, 178, 257 Lewis, Neil A., 28 Lewis,Thomas J., 21 Leyland,A. H., 212n8 Lieberson, Stanley, 126, 127 Liebling,A. J., 123 Lindaman, Kara, 165, 166 Little,Thomas C., 211n6, 212, 215 Long, Cheryl X., 143, 145, 146, 157

Long, Earl, 123 Longoria,Tomas, 60 Lopez-Silanes, Florencio, 143 Lovrich, Nicholas, 60 Lowery, David, 40, 82, 163, 165, 184, 197, 204, 230 Lowi,Theodore, 197 Lowry, Robert C., 109 Lowry,William, 172 MacDonald, Jason A., 103, 262 MacKuen, Michael B., 4, 89n16, 104, 225, 237 Madison, James, 19 Maestas, Cherie, 81 Magleby, David B., 79, 86 Mahajan, Harpreet, 63, 64 Malanchuk, Oksana, 61n2 Marcus, George E., 39 Marra, Robin F., 110 Martin, Lawrence L., 38 Martorano, Nancy, 20, 23n2 Maruna, Shadd, 185 Mastruzzi, Massimo 145, Mauro, Paolo, 156 Mayhew, David R., 151, 164, 165 McAtee,Anrdea 184, 204 McCubbins, Mathew D., 151 McIver, John P., xiii, 3, 6, 7, 11, 13–15, 25, 30, 38, 40, 43, 63, 64, 107, 107n11, 121, 121n1, 122, 123n2, 127, 128, 128n3, 131– 133, 131n5, 136, 137, 149, 163, 164, 183, 184, 186, 187, 191, 193, 195, 196, 199, 199n5, 210, 211, 221, 223–225, 210n4, 229, 230n1, 233, 234, 234n6, 237n8, 240, 250, 254, 256–260, 264 Meier, Kenneth J., 38, 146, 166, 170, 171 Melton, J. Gordon, 171 Mendoza, Richard, 64n5 Mexican-Americans, 12, 53–56, 59–67, 71– 74. See also Latinos

276

Index

Mezey, Michael L., 8, 39, 121n1, 183, 209, 210, 255 Miller,Arthur H., 19–21, 62n2 Miller,Warren E., 40, 202, 203, 225, 237n7 Moncrief, Gary E., 80, 85, 86 Monroe,Alan D., 164, 165 Montinola, Gabriella R., 158 Mooney, Christopher Z., 39, 80, 88n15, 90, 165, 166, 170, 172, 176, 178 Moreno, Dario, 57 Morgan, David R., 108 Morin, Richard, 149n3 Mueller, John E., 109 Munger, Frank, 8, 39, 121n1, 183, 209, 210, 255 National Annenberg Election Study, 263 National Gay and Lesbian Task Force, 167 National Network of State Polls, 83, 83n6, 84 National Survey on Latinos in America, 61, 62, 73 Neuman,W. Lawrence, 38 Neustadt, Richard E., 22 Newkirk, Glenn, 79, 85 Newman, Brian, 101, 105n1 Newmark,Adam J., 8 Newport, Frank, 79n1 Nice, David C., 38, 170, 171 Nicholson, Stephen P., 109 Nie, Norman, 61n2 Niemi, Richard G., 7, 8, 84, 103, 105, 106, 106n6, 106n8, 109, 110, 259, 262 Norpoth, Helmut, 105 Norrander, Barbara, 3, 7, 9, 11–13, 25, 38– 40, 46, 49, 121, 121n1, 163–166, 230, 257, 258, 260, 261, 264 Nugent, John D., 20 O’Brien, Robert M., 25n7 Oﬃcial State Job Approval Ratings, 103, 106, 106n6, 106n7, 107, 111n17, 116,

Oleszek,Walter J., 79 Olsen, Marvin E., 61n2 Olson, Mancur, 124 Onishi, Norimitsu, 156 Orth, Deborah A., 103, 111n15, 115 Ostrom, Charles W. Jr., 110 Pachon, Harry, 54 Padilla,Amado, 54, 64n5 Page, Benjamin I., 4, 164, 165, 225, 260 Pantoja,Adrian, 56 Park, David K., xiii, 9, 11, 14, 258, 259 Parker, Glenn R., 79, 80, 83, 84 Partin, Randall W., 103, 110 Party competition, 123–126, 170, 177 Patterson, Kelly D., 79, 86 Patterson, Samuel C., 79, 80, 82, 84, 87n14, 88–91, 89n16, 126 Paul, Pope John, II, 62 Pedersen, Mogens N., 80, 80n3 Perez, Lisandro,57 Perot, H. Ross, 214 Peters, B. Guy, 81n5 Phillips, Kevin P., 241n11 Picard, P., 171n6 Piereson, James E., 39 Plotnick, Robert, 123, 124 Plutzer, Eric, 9, 121n1, 258 Poe, Steven C., 121n1 Polinard, Jerry, 60 Pool, Ithiel de Sola, 8, 39, 183, 209, 209n1, 210 Poole, Keith T., 188 Popkin, Samuel L., 8, 39, 183, 209, 209n1, 210 Portes,Alejandro, 57 Powell, Lynda W., 8 Powell, Richard J., 8 Public ethos, 124 Puerto Ricans, 54, 55, 59, 60, 62. See also Latinos

Index 277 Putnam, Robert D., 142, 144, 148, 149n3, 158, 202, 261 Quinlan, Stephen V., 80, 88 Quinn, Bernard, 171n6 Rae, Douglas, 123, 124, 127, 136 Rahn,Wendy, 159 Raimondo, Henry J., 186, 187 Ramirez, Manuel, 64n5 Ramirez, Ricardo, 56 Raudenbush, Stephen W., 212n8 Reagan, Ronald, 28, 221, 222, 232, 243 Reise, Steven P., 212n8 Religion, 40, 171, 174, 175, 176, 177 Rice,Tom W., 102, 261 Rienzo, Barbara A., 171 Rieselbach, Leroy N., 82 Ringquist, Evan J., 38, 123, 121n1, 170, 184, 185, 187, 195, 196, 199, 230, 237n8, 258, 260 Ripley, Randall B., 80, 88 Rivera, Sylvia M., 121 Rivers, Douglas, 215n14 Robinson, James A., 5, 183 Robinson, John, 149n3 Rochlin, Steven A., 186 Roe v.Wade, 47 Roeder, Phillip W., 81, 163 Rohde, David W., 22 Rom, Mark, 185 Rosenberg, Robert S., 225 Rosenblum, Nancy L., 148 Rosenstone, Steven J., 40, 61n2 Rosenthal,Alan, 81, 82, 85 Rubin, Donald B., 212n8 Rumbaut, Rubén, 57 Russett, Bruce, 164 Sagebrush Rebellion, 28 Sanders, Lynn, 61, 63

Saunders, Kyle L., 39 Scarrow, Howard A., 39 Schaﬀner, Brian, 123 Schattschneider, E. E., 122, 197 Scheifer,Andrei, 143 Schlozman, Kay Lehman, 61n2, 64 Schneider, Saundra K., 14, 186, 190, 197, 266 Schumaker, Paul D., 165 Scott, Steve, 56 Searles, Patricia, 38 Sears, David O., 9, 82 Segura, Gary, 56, 109 Seidman, David, 9, 255 Shafer,William R., 8, 15, 163, 165, 183, 210, 255 Shapiro, Robert Y., 4, 63, 64, 164, 165, 225, 260 Sharkansky, Ira, 6, 165, 183, 186, 191 Shaw, George Bernard, 159 Shelley, Fred M., 38 Shingles, Richard, 61n2 Sigelman, Lee, 7, 8, 40, 58, 61, 81, 82, 84, 103, 106, 106n6, 106n8, 109, 259, 262 Simon, Dennis M., 103, 110 Sims-Butler, Kellie, 11, 23, 39, 121, 149, 163, 210, 211, 211n5, 237n8, 256, 257 Skogan,Wesley, 170 Smith, Raymond A., 167, 169, 171 Sparrow, Malcolm K., 146 Spiegelhalter, David, 213 Squire, Peverill, 79–81, 80n4, 84, 87n11, 87n14, 88–91, 89n16, 103, 104, 109, 115 State policy, measurement of, 14, 184–190 State public opinion and: abortion, 11, 39–41, 43, 46, 47, 49, 62, 64, 65, 70, 72, 255, 257 aid to blacks/minorities, 38, 40, 43, 49 capital punishment, 38–41, 43, 46, 47, 49, 72 combining state polls, 7, 13, 20, 38, 40, 41, 257–259

278

Index

State public opinion and: (continued) conﬁdence and trust in government, 20, 22–24, 30–34, 63, 65, 70, 72, 79, 82, 89, 257, 261 culture or region, 11, 27, 40, 46, 71, 191, 193 demographics, 5, 8, 9, 20, 28–30, 32, 33, 39–41, 43, 46, 58–61, 63, 71, 72, 74, 255 economy, 12, 13, 89, 94, 104–106, 110– 111, 114–116 gay and lesbians, 14, 62, 70, 165, 172–178, gubernatorial approval, 7–10, 12, 103– 116, 257, 262 ideological leanings, 6, 13, 15, 29, 30, 32, 38, 39, 41, 43, 49, 63, 64, 71, 72, 107, 108, 113, 114, 130, 132, 133, 135, 149, 150, 152, 154, 163, 173, 183, 193, 195, 196, 199, 200, 202, 203, 210, 221–223, 229–250, 256–260, 266 Latinos, 12, 53–74 legislative approval, 8, 10, 12, 79–96, 262 national surveys, 6, 7, 20, 23–27, 40–41, 61, 62, 210, 211, 213–225, 229–250, 254, 256, 257 opinion diversity, 13, 121–137, 262–263 partisanship, 6, 15, 29, 30, 32, 38, 39, 41, 43, 49, 63, 65, 70–73, 90, 107, 108, 113, 114, 130, 149, 150, 152, 154, 163, 183 193, 195, 196, 199, 200, 202, 203, 210, 221, 222, 229–250, 256, 257, 259, 266 presidential approval, 7, 10, 103, 109, 110, 113, 256 simulations, 8–10, 14, 15, 39, 163, 209– 225, 255, 256, 258–260 social capital and generalized trust, 13, 142–159, 260, 261, 265 state laws, 46–50, state residency, 38–47, 49, 55, 56, 261, 264, 265 welfare. See aid to minorities State public opinion as a dependent variable, 11–13, 263–265

State public opinion as an independent variable, 13–15, 164–166, 265, 266 States as analytic units, 4 Stepick,Alex, 57 Stern, Hal S., 212n8 Stewart, Marianne C., 21 Stimson, James A., 4, 87n12, 89n16, 104, 109, 165, 166, 225, 237, 260 Stokes, Donald E., 22, 23, 202, 203, 225, 237n7 Stonecash, Jeﬀrey M., 123 Sullivan, John L., 39, 124, 127, 128 Svoboda, Craig T., 102 Talarico, Susette M., 38 Tatalovich, Raymond, 166 Taylor, Michael, 123, 124, 127, 136 Term limits, 85, 86, 87, 90, 94, 95 Theiss-Morse, Elizabeth, 19–21, 23, 79, 81, 95n19 Thomas,Andrew, 213 Thomas, Clive, 130, 131, 171 Thomas, Loring, 212n8 Thomas, Scott, 212n8 Thompson, Joel, 80, 86 Tienda, Marta, 54 Tolbert, Caroline J., 124, 126, 128n4, 136, 137 Torres, Maria de los Angeles, 58 Transparency International, 145, 146, 158 Trueba, Enrique (Henry), 53 Truman, David, 196, 202 Uhlaner, Carole, 60, 61n2, 64n6 Ulbig, Stacy G., 20, 23n2, 30, 256 Uslaner, Eric M., 6, 13, 143, 144, 146n1, 148, 149n3, 156, 156n7, 156n8, 159, 261, 265 Valdez,Armando, 60n1 Valentino, Nicholas A., 9 Van Dunk, Emily, 126, 133, 170, 170n4 Vanderleeuw, James M., 103, 111n15

Index 279 Ventura, Jesse, 108n12 Verba, Sidney, 61n2, 64, 143 Vishney, Robert W., 143 Voss, D. S., 211, 214 Wade, Michelle, 20, 23n2 Wahlke, John C., 88 Waksberg, Martin P., 87n13 Wald, Kenneth D., 171 Walker, Jack, 185, 187 Watt, James, 28 Weaver, Janet, 60 Weber, Ronald E. 8, 15, 39, 81, 121n1, 163, 183, 186, 209, 210, 255 Weiher, Gregory R., 8, 103, 109, 115 Welch, Susan, 58, 61 Wilcox, Clyde, 7, 9, 11–13, 39, 40, 163–166, 258, 261, 264 Will, George F., 86 Wilson, James Q., 124 Wilson, L.A., 108 Wilson, Pete, 56 Winters, Richard F., 123, 124, 146n2, 186

Wlezien, Christopher, 257, 260 Wolbrecht, Christina, 79 Wolﬁnger, Raymond, 61n2 Woods, Nathan, 56, 109 Wright, Gerald C. Jr., xiii, 3, 6, 7, 11, 13–15, 25, 30, 38, 40, 43, 63, 64, 103, 107, 110, 121–123, 121n1, 123n2, 127, 128, 128n3, 131–133, 131n5, 136, 137, 149, 149n4, 150, 163, 164, 183, 184, 196, 187, 191, 193, 195, 196, 199, 199n5, 210, 210n4, 211, 221, 223–225, 229, 230n1, 233, 234, 234n6, 237n8, 238, 240, 250, 254, 256–260, 264 Wrinkle, Robert, 60 Wu, Fengshi, 149n3 Yang,Alan, 170, 173 Yinger, Milton, 64 Zaller, John R., 185 Zeckhauser, Richard J., 143, 145, 148, 149, 158 Zeller, Richard A., 25 Zellner,Arnold, 152

Public Opinion in State Politics

Public Opinion

Crystallizing public opinion

Crystallizing public opinion

Public Opinion Polling in a Globalized World

Public Opinion and Constitutional Controversy

Public Opinion and Constitutional Controversy

Tides of Consent: How Public Opinion Shapes American Politics

Degrees of Democracy: Politics, Public Opinion, and Policy

Counting the Public In: Presidents, Public Opinion, and Foreign Policy

Opinion

Public Opinion and American Foreign Policy

The Gallup Poll: Public Opinion 2010

Why Welfare States Persist: The Importance of Public Opinion in Democracies (Studies in Communication, Media, and Public Opinion)

Policy and Politics in State Budgeting (Bureaucracies, Public Administration, and Public Policy)

Collective Behavior and Public Opinion: Rapid Shifts in Opinion and Communication

Selling the Korean War: Propaganda, Politics, and Public Opinion in the United States, 1950-1953

Public Opinion and Internationalized Governance (Beliefs in Government , Vol 2)

Polling for democracy: public opinion and political liberalization in Mexico

Bending Opinion: Essays on Persuasion in the Public Domain

Public Opinion and Internationalized Governance (Beliefs in Government , Vol 2)

Public Opinion and Changing Identities in the Early Modern Netherlands

Statehouse Democracy: Public Opinion and Policy in the American States

Dangerous Frames: How Ideas about Race and Gender Shape Public Opinion (Studies in Communication, Media, and Public Opinion)

Public Opinion, the First Ladyship, and Hillary Rodham Clinton (Women in American Politics, 2)

Selling the Korean War: Propaganda, Politics, and Public Opinion in the United States, 1950-1953

The Nature and Origins of Mass Opinion (Cambridge Studies in Public Opinion and Political Psychology)

The Nature and Origins of Mass Opinion (Cambridge Studies in Public Opinion and Political Psychology)

Publics, Politics and Power: Remaking the Public in Public Services

Public Philosophy: Essays on Morality in Politics

Public Philosophy: Essays on Morality in Politics

Public Opinion in State Politics

Public Opinion

Crystallizing public opinion

Crystallizing public opinion

Public Opinion Polling in a Globalized World

Public Opinion and Constitutional Controversy

Public Opinion and Constitutional Controversy

Tides of Consent: How Public Opinion Shapes American Politics

Degrees of Democracy: Politics, Public Opinion, and Policy

Counting the Public In: Presidents, Public Opinion, and Foreign Policy

Opinion

Public Opinion and American Foreign Policy

The Gallup Poll: Public Opinion 2010

Why Welfare States Persist: The Importance of Public Opinion in Democracies (Studies in Communication, Media, and Public Opinion)

Policy and Politics in State Budgeting (Bureaucracies, Public Administration, and Public Policy)

Collective Behavior and Public Opinion: Rapid Shifts in Opinion and Communication

Selling the Korean War: Propaganda, Politics, and Public Opinion in the United States, 1950-1953

Public Opinion and Internationalized Governance (Beliefs in Government , Vol 2)

Polling for democracy: public opinion and political liberalization in Mexico

Bending Opinion: Essays on Persuasion in the Public Domain

Public Opinion and Internationalized Governance (Beliefs in Government , Vol 2)

Public Opinion and Changing Identities in the Early Modern Netherlands

Statehouse Democracy: Public Opinion and Policy in the American States

Dangerous Frames: How Ideas about Race and Gender Shape Public Opinion (Studies in Communication, Media, and Public Opinion)

Public Opinion, the First Ladyship, and Hillary Rodham Clinton (Women in American Politics, 2)

Selling the Korean War: Propaganda, Politics, and Public Opinion in the United States, 1950-1953

The Nature and Origins of Mass Opinion (Cambridge Studies in Public Opinion and Political Psychology)

The Nature and Origins of Mass Opinion (Cambridge Studies in Public Opinion and Political Psychology)

Publics, Politics and Power: Remaking the Public in Public Services

Public Philosophy: Essays on Morality in Politics

Public Philosophy: Essays on Morality in Politics

Recommend Documents