Public Opinion in State Politics 9780804767972

Citation preview

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

0

Natl. Res. 0

Highways G

Education

Policy area

Govt. Admin.

0

Health Care

0

Corrections

0

Hospitals

0

Police

G

Parks

G

-0.5

State

0

Welfare

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

0

0 0

0

State policy priorities

80

0

60

8

0

~

• 0

0

0

0

0

0

8o

0 0 0

©

'o

0

0

0

08 0

0

0

40

0

0

oo @0 0

0

0

0 0

0 0

0

8

0

0 0 0

i 0

8

0

0 00

So

0

0

x-1

0

§

§

g8

'"!! @

0

0

~ 0

oo

0

8

0

20

8

0

8

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 0 0

0

0

0

0

0

0

0

0

State policy priorities

80

0

0

0

0

0

0

0

0

0

0

0

60 0 0 0 0

40

0

0

@

00

oo

0 0 0

0 0

0

Oo

0

0 0

0

0 0

0

20

0 0

0

0

0

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 0

0 0

State policy priorities

80

0

0

60

0 0

0 0

40

0

0

0 0

oO

0

0 0

il'

&

0

00

@

0

0 0

0

0

oo

0 0

8'l9

0 0

20

0

0

0

0

0 -0.1

0

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

0

100 O@

0

0

80

0

State policy priorities

0

0

0

60 0

oo

0

0

40 0 0

0

Q

0

0

0

0

20

0

0

0

0

0

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

0

0

0

0 0 0

80

Q

0

State policy priorities

'

~8

0

@

Oo

40

g8 0

80 o

0 o€!'

0

0

0

0 0

0

8,

Q

0

0

0

8

0

Q

0

0

0

@

0

Q

8

'

§0

0

'g, 8'0

00 0 0

~0,

0

60

0

0 0 0

8 Q

8

0

20

§

0

8

0 20

25

30

35

40

Collective goods interest groups

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

0 0 0

0

0

"'&g

80 State policy priorities

8

0 0

Oo 0

0 0 0 00

0

oo

~

~

0

80 o®

8

60

00

0

~§

0

0 0

~

0

8

::s

0

8

0

8 0 0 0 Q

0

0

rJl

0

8 8

0

80

40

8

0 0 0

8

0

0

8

8

0

20

§

0

8

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

CBS/NY Times

No Pooling

q

q

Q)

~

u

5

0

c:

Q)

~

~ 0

u

5

0