PUBLIC OPINION AND THE
POLICY CHOICES OF STATE HIGH COURTS
By
Frederick Stewart Wood

A DISSERTATION
Submitted to
Michigan State University
in partial fulfillment of the requirements
for the degree of
DOCTOR OF PHILOSOPHY
Political Science
2011

ABSTRACT
PUBLIC OPINION AND THE POLICY CHOICES OF STATE HIGH COURTS
By
Frederick Stewart Wood
This project addresses the question of whether the judiciary is a representative
part of state government. I will examine whether state high court justices, as other
political actors, are influenced by constituency opinion in their decision-making process.
I contend that popular elections provide an incentive to hold the justices more
accountable to the electorate and comparing the impact of public opinion on the votes of
state supreme court justices from a diverse set of electoral arrangements will help to
understand the democratic nature of the judiciary. This dissertation will contribute to the
literature on the role of public opinion in the policy process, the effect of institutional
design on elite behavior, and the impact of judicial selection and retention reforms.

Copyright by
FREDERICK STEWART WOOD
2011

Dedicated to Charles, Donna, Samantha, Daisy, and Abigail Wood

iii

ACKNOWLEDGMENTS

In addition to my family, to whom I dedicate this dissertation, I have a number of
individuals to acknowledge. First, I have to recognize that this dissertation would not
exist at all without Dr. Saundra K. Schneider. The other members of my committee, Dr.
William Jacoby, Dr. Valentina Bali, and Dr. Mae Kuykendall, were also invaluable
advisors. I would also like to recognize some of my fellow former Spartans. Dan Lewis
was the best work spouse that any graduate student could have. I would also like to thank
Jeremy Duff, Carl Snook, Greg Robinson, Mike Crespin, Dave Dreyer, and all of the
other inhabitants of the 2nd floor of South Kedzie Hall for not going insane due to the
institutional yellow paint in the hallway. Finally, the standard disclaimer is that all errors
contained below are the fault of the author and not those credited above. That is certainly
true with this dissertation as well.

iv

TABLE OF CONTENTS

LIST OF TABLES ……………………………………………………………………... vii
LIST OF FIGURES …………………………………………………………………...... xi
CHAPTER 1
INTRODUCTION ………………………………………………………………………. 1
Courts as Policy Actors ………………………………………………………….. 5
Importance of State Courts ..…………………………………………………….. 7
CHAPTER 2
LITERATURE REVIEW ………...……………………………………………………. 10
Judicial Decision-Making ……………………………………………………… 19
Public Opinion and the Courts ……………………………………………….... 27
The Importance of Context …………………………………………………….. 30
Conclusion ……………………………………………………………………... 32
CHAPTER 3
DATA AND METHODS ……………………………………………………………… 34
Dependent Variable …………………...……………………………………….. 35
Measuring Public Opinion in the States ………………………………………... 37
Control Variables ………………………………………………………………. 43
Modeling Strategy ……………………………………………………………… 51
Conclusion ………………………………………………………………...…… 53
CHAPTER 4
DISCRIMINATION BASED ON SEXUAL ORIENTATION ………………………..55
Marriage Cases………………………………………………………………….. 58
Child Custody Cases …………………………………………………………… 60
Sodomy Cases………………………………………………………………..…. 61
Benefits Cases……………………………………………………………………62
Discrimination Cases…………………………………………………………… 62
Other Cases………………………………………...…………………………… 63
Bivariate Model Results ………………...……………………………………… 64
Multivariate Model Results ….…….…………………………………………… 64
Control Variables ………………………………………………………………. 68
Interactive Effect ..……………………………………………………………... 74
Conclusion ………………………………………………….……………...……76
APPENDIX A …………………………….………………………………………….….79

v

CHAPTER 5
GENDER DISCRIMINATION CASES ……………………..……..…………………. 91
Criminal Cases ….…………………………………………….………...……… 93
Abortion Cases …………………………………………………………………. 95
Family Cases …………………………………………………………………….96
Benefits Cases………………………………………………...………………… 97
Discrimination Cases…………………………………………………………….97
Bivariate Model Results ……………………………………….……………….. 99
Multivariate Model Results …………………………………………………….. 99
Control Variables ………………………………………………………………103
Interactive Effect ………………………………………………………...…….108
Conclusion …………………………………………………………………..... 109
APPENDIX B …………………………………………………………….……………112
Chapter 6
CONCLUSION ………………………………………………………………………. 124
Policy Differences ……………………………………………………………...126
Implications ………………………………………………………………….... 129
APPENDIX C………………... …………………………………………………….… 133
REFERENCES …………………………………………………………...................... 140

vi

LIST OF TABLES

Table 1: Methods of Retention for State High Courts….……...……… ………............. 45
Table 2: The number of sexual orientation discrimination cases by state, 1981–2004.... 57
Table 3: Percentage of observations by case type in sexual orientation models……..… 58
Table 4: Estimated coefficient for the Berry et al. measure of public preferences in cases
involving claims of discrimination based on sexual orientation……...……………...…. 65
Table 5: Estimated coefficient for the Brace et al. measure of public preferences in cases
involving claims of discrimination based on sexual orientation………………………....66
Table 6: Estimated coefficient for the Erikson et al. measure of public preferences in
cases involving claims of discrimination based on sexual orientation.…...………….… 67
Table 7: Estimated coefficient for the Norrander measure of public preferences in cases
involving claims of discrimination based on sexual orientation.…………………….…. 67
Table 8: Coefficient estimates for control variables based on statistical significance..... 69
Table 9: Predicted Probabilities of a vote favoring the rights claims of gay & lesbians
using the Berry et al. measure of public preferences…………………………………… 70
Table 10: Predicted probability of a pro-gay vote by case issue type …..…………...… 73
Table 11: A comparison of the percentage of observations correctly predicted for each
measure of public opinion excluding and including interaction terms with retention
methods…………………………………………………………………………….…… 76
Table 12: A comparison of the area under the Receiver-Operator Characteristic Curve. 76
Table 13: Logit estimates for a bivariate model including the Berry et al. measure of
public preferences in cases involving discrimination based upon sexual orientation….. 79
Table 14: Logit estimates for a bivariate model including the Brace et al. measure of
public preferences in cases involving discrimination based upon sexual orientation.…. 79
Table 15: Logit estimates for a bivariate model including the Erikson et al. measure of
public preferences in cases involving discrimination based upon sexual orientation….. 80
Table 16: Logit estimates for a bivariate model including the Norrander measure of
public preferences in cases involving discrimination based upon sexual orientation.…. 80
vii

Table 17: Logit estimates including the Berry et al. measure of public preferences in
cases involving discrimination based upon sexual orientation.….……………………... 81
Table 18: Logit estimates including the Berry et al. measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences......................................................................... 82
Table 19: Logit estimates including the Brace et al. measure of public preferences in
cases involving discrimination based upon sexual orientation.….……………………... 83
Table 20: Logit estimates including the Brace et al. measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences.…….………..………………………….……. 84
Table 21: Logit estimates including the Erikson et al. measure of public preferences in
cases involving discrimination based upon sexual orientation………………………..... 85
Table 22: Logit estimates including the Erikson et al. measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences.………………………………………….….... 86
Table 23: Logit estimates including the Norrander measure of public preferences in cases
involving discrimination based upon sexual orientation..………………...………….… 87
Table 24: Logit estimates including the Norrander measure of public preferences in cases
involving discrimination based upon sexual orientation with interactions between
retention method and public preferences………………………………………….....…. 88
Table 25: Percentage of observations by case type in gender discrimination cases……. 92
Table 26: The number of gender discrimination cases by state, 1970 – 2001………….. 93
Table 27: Estimated coefficient for the Berry et al. measure of public preferences in cases
involving claims of discrimination based on gender..……………………………….....100
Table 28: Estimated coefficient for the Brace et al. measure of public preferences in cases
involving claims of discrimination based on gender….……………………………..... 101
Table 29: Estimated coefficient for the Erikson et al. measure of public preferences in
cases involving claims of discrimination based on gender……………………………. 101
Table 30: Estimated coefficient for the Norrander measure of public preferences in cases
involving claims of discrimination based on gender…………………..…....……….... 102
Table 31: Coefficient estimates for control variables based on statistical significance. 104
viii

Table 32: Predicted probabilities of a vote favoring the removal of a gender based
distinction by retention method…………………..……………………………..…….. 105
Table 33: Predicted probabilities of a vote favoring the removal of a gender based
distinction by case type…………………………….………………………………..… 106
Table 34: A comparison of the percent of observations correctly predicted for each
measure of public opinion excluding and including interaction terms with retention
methods………………………………………………………………………………... 109
Table 35: A comparison of the area under the Receiver-Operator Characteristic Curve for
each measure of public opinion excluding and including interaction terms with retention
methods………………………………………………………………………………... 109
Table 36: Logit estimates for a bivariate model including the Berry et al. measure of
public preferences in cases involving discrimination based upon gender…………….. 112
Table 37: Logit estimates for a bivariate model including the Brace et al. measure of
public preferences in cases involving discrimination based upon gender……….….… 112
Table 38: Logit estimates for a bivariate model including the Erikson et al. measure of
public preferences in cases involving discrimination based upon gender……..…..….. 113
Table 39: Logit estimates for a bivariate model including the Norrander measure of
public preferences in cases involving discrimination based upon gender………....….. 113
Table 40: Logit estimates for the model including the Berry et al. measure of public
preferences in cases involving discrimination based upon gender….…………..…….. 114
Table 41: Logit estimates for the model including the Berry et al. measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences………………………..…………… 115
Table 42: Logit estimates for the model including the Brace et al. measure of public
preferences in cases involving discrimination based upon gender...……………....….. 116
Table 43: Logit estimates for the model including the Brace et al. measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences……..………………………..…….. 117
Table 44: Logit estimates for the model including the Erikson et al. measure of public
preferences in cases involving discrimination based upon gender………….……….... 118

ix

Table 45: Logit estimates for the model including the Erikson et al. measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences………………………………....….. 119
Table 46: Logit estimates for the model including the Norrander measure of public
preferences in cases involving discrimination based upon gender….……………..….. 120
Table 47: Logit estimates for the model including the Norrander measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences……………..…………………..….. 121
Table 48: Logit estimates for cases involving discrimination based upon sexual
orientation without a measure of public preferences………...………….…………….. 132
Table 49: Logit estimates for cases involving discrimination based upon gender without a
measure of public preferences…………………………………………….……….….. 134
Table 50: Percent of cases predicted correctly for models involving discrimination based
upon sexual orientation………………….…………………………………………….. 135
Table 51: Percent of cases predicted correctly for models involving gender
discrimination.……….…………………………………………………………..……. 135
Table 52: Logit results for gay rights cases including all four public opinion
measures …………………………………………………………………..…………... 136
Table 53: Logit results for gender discrimination cases including all four public opinion
measures …………………………………………………………………..…………... 137
Table 54: Correlation between public opinion measures in gay rights cases…………. 138
Table 55: Correlation between public opinion measures in gender discrimination
cases…………………………………………………………………………...………. 138

x

LIST OF FIGURES

Figure 1: Mean Predicted Probability with 95% CI, Berry et al…………..………….. 89
Figure 2: Mean Predicted Probability with 95% CI, Brace et al…………………...…. 89
Figure 3: Mean Predicted Probability with 95% CI, Erikson et ……………...……..... 90
Figure 4: Mean Predicted Probability with 95% CI, Norrander………………………. 90
Figure 5: Mean Predicted Probability with 95% CI, Brace et al.…...…………..…… 122
Figure 6: Mean Predicted Probability with 95% CI, Norrander.…………………...... 123

xi

Chapter 1 Introduction
Shortly after midnight on May 17th 2004, Marcia Kadish and Tanya McCloskey
became the first same-sex couple to obtain a valid marriage license and by 9:10 a.m. they
became the first same-sex couple to legally marry in the United States. They and
hundreds of other gay & lesbian couples were able to do so because of a Massachusetts
state high court decision that held the state's constitution did not allow for the disparate
treatment of same-sex couples seeking the protections, benefits, and obligations
emanating from marriage (Goodridge vs. Department of Public Health 2003). This
decision not only rekindled the debate about how government policy should treat samesex relationships, it also highlights the important role that state high courts can have on
the lives of individuals and within the policy-making process.
The Goodridge decision also presents the question as to why the Massachusetts
Supreme Judicial Court found a right to marriage for gays and lesbians after other courts
1

had explicitly refused to do so for decades. The Goodridge decision was not unanimous
and in the conversation between the majority, concurring, and dissenting opinions the
justices provided a hint that external considerations played a significant role in deciding
the case. Justice Cordy, in a dissenting opinion, believed that the majority had altered its
2

approach in applying a legal test due to a “perfect storm” of external pressures. Among

1

See Baker v. Nelson (1971), Baker v. State (1999), De Santo v. Barnsley (1984), Dean
v. District of Columbia (1995), Jones v. Hallahan (1973), and Singer v. Hara (1974).
2
“I fully appreciate the strength of the temptation to find this particular law
unconstitutional - there is much to be said for the argument that excluding gay and
lesbian couples from the benefits of civil marriage is cruelly unfair and hopelessly
outdated; the inability to marry has a profound impact on the personal lives of committed
gay and lesbian couples (and their children) to whom we are personally close (our
friends, neighbors, family members, classmates, and co-workers); and our resolution of
1

the reasons for the court to change the marriage laws of the state was that society now
had “a more fully developed understanding of the invidious quality of the discrimination”
(at 328). In a concurring opinion, Justice Greaney described the change in more direct
terms:
“We share a common humanity and participate together in the social contract that
is the foundation of our Commonwealth. Simple principles of decency dictate that we
extend to the plaintiffs, and to their new status, full acceptance, tolerance, and respect.
We should do so because it is the right thing to do” (at 349-350).
Justice Greaney continued by elaborating on his larger judicial philosophy that
allows judges to be influenced by external forces:
“The provisions of our Constitution are, and must be, adaptable to changing
circumstances and new societal phenomena, and, unless and until the people speak again
on a specific subject, conformable in their concepts of liberty and equality to what is fair,
right, and just” (at 350).
This description of how the court arrived at its decision raises the question of
whether justices always decide cases in this manner. Courts are often thought of as
separate from the democratic process. The republican design of the national and state
government with government power divided among separate institutions has contributed
to differing expectations for the behavior of policy makers. The legislative and executive
branches were designed with the expectation that they would represent or reflect the
views of the citizenry to varying degrees. However, in this three branch design, the
this issue takes place under the intense glare of national and international publicity.
Speaking metaphorically, these factors have combined to turn the case before us into a
"perfect storm" of a constitutional question. In my view, however, such factors make it
all the more imperative that we adhere precisely and scrupulously to the established
guideposts of our constitutional jurisprudence, a jurisprudence that makes the rational
basis test an extremely deferential one that focuses on the rationality, not the
persuasiveness, of the potential justifications for the classifications in the legislative
scheme. I trust that, once this particular "storm" clears, we will return to the rational
basis test as it has always been understood and applied.” at 363.

2

judiciary was expected to be removed from popular politics and instead established as the
umpire of the political process. The task of the courts was to ensure that the laws were
created in the appropriate manner within the bounds of government authority and
enforced without violating the rights of the people. In the course of performing this role,
courts have had to maintain their independence from the political branches and assert a
judicial power to abate the decisions of the same political branches (McCloskey 1960).
To maintain the power of judicial review, judges have attested that the source and
justification for this power is not political in nature, but rather grounded in reason and
logic. If the decisions of a court were to appear to be based on political concerns, like
public opinion, rather than the logical extension of the requirements of the law, then a
court’s exercise of judicial review would appear to encroach on the charge of the other
branches. This encroachment could negatively affect their legitimacy with the public and
those with the ability to enforce their decisions. Judges and courts who stray outside of
the bounds of what the public wants are risking diminishing their influence over policy.
Ignoring the fact that this policy change originated from a judicial body, scholars
of state politics could point to a number of factors to explain why Massachusetts had
changed its policy instead of the other states. One of the primary explanations would be
that compared to other states, Massachusetts has a number of liberal policies in place and
same-sex marriage is just an extension of the liberal citizen and elite opinion.
Alternatively, by focusing on the judicial source of the policy change, judicial decisionmaking scholars could offer contradictory explanations that focused internally on the
personnel on the court or the institutional features of the Supreme Judicial Court and the
contextual factors relied upon by state politics scholars would be mostly ignored. It is the

3

goal of this study to reconcile these two approaches and determine whether the external
forces that are often used to explain state policy influence judicial decision-making.
Democratic theory posits that a government composed of representatives elected
by the people will be faithful to the policy wants of the people and therefore enact
policies that pursue these wants. As a result, there should be evidence of congruence
between public opinion and policy. For decades, political scientists have found a link
between the public’s preferences and when policy is created or altered for institutions
other than judicial bodies (Achen 1978, Bartels 1991, Burstein 2003, Erikson 1978, Hill
et al. 1995, Lascher et al. 1996, Miller and Stokes 1963, Page and Shapiro 1983, Weber
and Shaffer 1972). Previous research focused on state governments has found that a
state’s overall policy priorities and the ideological direction of its policies are influenced
by public opinion (Erikson et al. 1993, Jacoby and Schneider 2001). Despite the
introductory illustration and numerous other instances, the public policy literature has
largely excluded the courts as policy-makers (Barlcay and Birkland 1998, but see Roch
and Howard 2008).
Because of the tendency to exclude courts in studies of policy formation, the
expectation of popular control has not been extended to the rulings of state high courts,
even though their decisions can have a profound effect on the availability of policy
alternatives to state governments and in many states, the public elects the justices. In
spite of calls from prominent scholars (Brace et al. 2000a, Caldeira 1991), to date there
has been little study of the impact of public opinion on the decision-making process of
state supreme court justices. In this dissertation, I propose to address this omission by

4

conceptualizing state high courts as policy-makers to evaluate the influence that public
opinion has on the decision-making process of individual justices.
Courts as Policy Actors
The common assumption is that courts are solely legal and not political
institutions and they only act to limit the actions of other actors or settle disputes between
actors. Judges are typically thought of as speaking in the language of the law and using
policy neutral legal principals derived from a philosophy of justice to arrive at their
decisions. These decisions are a result of simply applying the facts of a case to the
established law. This “Myth of Legality" is reinforced by the judges themselves with
various symbolic acts such as the wearing of robes, the elevation of their benches, and the
deference that is required when in a courtroom. The resulting “cult of the robe" has
become part of our American political culture (Scheb and Lyons 2000). Another feature
that discourages the inclusion of the judiciary in the policy making process is the reactive
nature of the judiciary. Judges can only act when a controversy is brought before the
court and the case meets the legal requirements of standing, jurisdiction, and an available
remedy.
State high courts can use the power of judicial review to participate in the policy
process and act as policy-makers themselves (Langer 2002). The judiciary can act as a
constraint on the policy process by restricting the alternatives available to policy
implementers. As an example, in Printz v. United States, the U.S. Supreme Court held
that the Brady Handgun Violence Prevention Act could not require local law enforcement
officials to perform background checks on handgun purchasers (Printz v. United States
1997). The national government cannot impose upon the sovereignty of a state and

5

compel it to perform those duties that it is not willing or able to do. Courts can also preapprove or suggest policy alternatives that would be constitutional to undertake. In
Swann v. Charlotte-Mecklenburg Board of Education, the U.S. Supreme Court outlined
which policies school districts could constitutionally employ to desegregate their student
population (Swann v. Charlotte-Mecklenburg Board of Education 1971).
Courts not only resolve policy disputes, they also act as agenda-setters and their
decisions can define the availability of policy alternatives (Dahl 1957, McCann 1994). A
court can use its rulings to redirect the attention of policy makers towards a specific
policy area or preapprove policy alternatives by pronouncing those that would be
constitutional to undertake. Courts may also alter the agenda setting process by allowing
groups who have been excluded by other institutions into the policy process (Cobb and
Elder 1983). Like most agenda-setters, the courts are limited in the frequency in which
they can define the policy agenda. Still, a court decision can spark a wide range of
reactions by the public and government, from minor clarifications in the law to the
proposal of constitutional amendments or ballot initiatives.
The judiciary participates in a wide array of policies. As an institution, courts are
probably most recognized for their involvement in the creation and implementation of
policy concerning criminal and civil justice procedures (Wasby 1993). Creating rules
like the Miranda rule and the exclusionary rule, the courts have regulated how members
of law enforcement agencies treat individuals suspected of committing criminal offences.
The reach of courts extends beyond the issues that they do not directly implement.
For example, state high courts in Vermont (Brigham v. State of Vermont 1997) and New
York (Campaign for Fiscal Equity v. State of New York 2003) have ruled that their state’s

6

scheme for financing public schools is unconstitutional and requires reform to make them
more equitable. Additionally, other policy makers can enlist the judiciary when
implementing a policy. When the Environmental Protection Agency decided to combat
pollution with the Superfund program, the agency resorted to litigation as a strategy of
regulatory enforcement (Church and Nakamura 1993).
Importance of State Courts
The state judiciary is important to study as a policy-making institution due to the
volume and types of cases that it decides. One estimate places almost 98% of the
nation’s litigation in state court systems (Hall 2008). State high courts decide more than
ten thousand cases a year (Devins 2010). Not only are state high courts responsible for
interpreting state constitutions, but they also decide much of the civil litigation that takes
place in the United States (Kritzer et al. 2007). The diversity of the types of cases that
come before state court systems is vast (Meeker 1986).

3

State high courts can use their judicial review powers to initiate policy change and
their judicial authority to ensure the implementation of the law. State courts have
increasingly become a source of constitutional rights for American citizens. In a famous
essay, Justice William Brennan called for state high courts to “be the guardians of our
liberties” when the Supreme Court was unwilling to do so (Brennan 1977 at 491). The
3

“the number of inmates in state prisons, where children go to school, whether land can
be used for highways, whether and when abortions may be performed, the protection of
various species, the need for motorcycle helmets and thousands of other issues ” (Thomas
and Hrebenar 1994 at 3 cited in Corbally, Sarah F. 2004. Filing of Amicus Curiae Briefs
in State Courts of Last Resort: 1960-2000 Justice System Journal)

“the final decision makers on most issues of commercial, property, family, inheritance,
tort and criminal law, as well as state constitutional issues and local governmental powers
and procedural issues” (Meeker, 1986, at 3)

7

Supreme Court itself subsequently proclaimed in a decision that “a state court is entirely
free to read its own State’s constitution more broadly that this Court reads the Federal
Constitution” (City of Mesquite v. Aladdin's Castle, Inc. 1982). Observers noted that the
number of cases in which state high courts considered constitutional questions had
increased dramatically between 1977 and 1986 and quickly doubled in the following teo
years (Hershkoff 1993). While the rights granted by state courts are not initially national
in their effect, they contribute to the rights environment that exists in the law and
decisions in one state can be used by courts as justification for expanding rights in other
states (Caldeira 1985). Additionally, compared to the United States Supreme Court, the
state courts have an added ability to influence constitutional law because it is free from
limitations on the types of cases that it hears (Hershkoff 2001). Another important
difference between state high courts and the federal Supreme Court is that state high
courts play a large role in shaping the common law that is used for civil torts (Sager
1978). Finally, state high courts are more integrated into the policy making process of
state governments than their federal counterpart. A number of state high courts are able
to intervene in cases before there is a formal appeal, some have a mandatory docket when
it comes to hearing constitutional challenges, others have the ability to offer advisory
opinions before a law is passed, and state courts have been more willing to hear disputes
between government officials (Linde 2005).
Even though state high courts are a policy making institution, previous research
on the effect of public opinion on judicial decision-making has focused on the United
States Supreme Court. Unfortunately, given the purposeful isolation of the Supreme
Court, it is difficult to conclude that their behavior is representative of the great number

8

of judges in the United States’ legal system. This analysis is needed because unlike
studies of the federal courts, there is a sufficient amount of variation in the selection
method of state supreme court justices to determine whether the form of retention
provides a linkage between the court and the public or if the nature of a court’s power
requires the courts to act in this manner to protect its efficacy. Given the diversity
contained within the states of their institutional design and mass attitudes, a comparative
analysis of judicial voting behavior will contribute to both the state policy and judicialdecision making literatures. This study will test one of the key findings of the state
policy literature within a new institutional setting, the state judiciary, to determine
whether the policy-making process employed by the courts is similar to that of the elected
institutions. This would allow state policy researchers to consider the judiciary in their
policy-making explanations. The study would contribute to the judicial literature by
incorporating political context into decision-making models.
This study will continue, in Chapter 2, with a survey of previous research on state
policy making and judicial decision-making, further demonstrating a need for this study.
The design of the study is motivated by the previous findings of state policy research.
Chapter 3 will discuss where the data used in this study comes from and how it will be
analyzed. Chapters 4 and 5 will provide an in-depth discussion and analysis of the legal
issues that courts are addressing that are relevant to gays and lesbians (Chapter 4) and
gender discrimination (Chapter 5). This study will conclude with Chapter 6, which will
discuss the findings, its implications, and directions for future research.

9

Chapter 2 Literature Review
An oft-used metaphor to describe the role of states in the federal system is as
4

laboratories of democracy. This expression is attractive because it provides a simple
reason as to why there is ample variation in the institutions and policies among the states.
Political scientists have focused on the study of the American states due to their rich
variation on institutional, socio-economic, and political characteristics (Dawson and
5

Robinson 1963). Under the rubric of federalism and a national government of limited
powers, state governments initially had great latitude to exert their police powers over the
general areas of health, safety, welfare, and morals. However, over time, the federal
government centralized many of the functions previously performed by the states.
Following the expansion of the federal government’s authority during the New Deal and
the Great Society, a period of devolution has returned some discretion to state
governments to address their policy problems (Conlan 1998). However, in some policy
areas such as homeland security, there are significant constraints placed upon state
governments. A review of policy subject to devolution concluded that devolution may
have been more political rhetoric than reality (Bowman 2002).
As some autonomy returned to state governments in the 1980’s, they were once
again able to serve as policy innovators. A number of reforms to state government, such
4

Justice Brandeis wrote “It is one of the happy incidents of the federal system that a
single courageous State may, if its citizens choose, serve as a laboratory; and try novel
social and economic experiments without risk to the rest of the country.” (New State Ice
Co. v. Liebmann 1932 at 311)
5
“The fifty states share a common institutional framework and general cultural
background, but they differ in certain aspects of economic and social structure, political
activity, and public policy. Therefore, they provide a large number of political and social
units in which some important variables can be held constant while others are varied.”
(Dawson and Robinson 1963, 265)
10

as amending constitutions, professionalizing legislatures and bureaucracies, increasing
gubernatorial authority, and the diversification of state revenues, have aided state
governments in their transition back to playing a more substantial role in governing
(Reeves 1990). As a result of these changes, states have been able to experiment with
their own solutions to policy problems in diverse areas such as economic development,
education policy, environmental regulation, and welfare reform (Bowman and Kearney
1986, Osborne 1988).
Scholars of state politics have for decades attempted to develop an explanation as
to why states have taken divergent paths to solving social problems. Broadly speaking,
the literature has found that social, economic, and political factors influence the content
of state policy (Gray 2008). The social and economic properties of the state that affect
6

policy include the size and composition of the state’s population , the geographic
features of the state, the economic makeup and conditions, and the presence of natural
resources. The political factors that influence the policies of the states include the
political party in control of the government, the level of electoral competition between
the two parties in general elections, the effectiveness of interest groups, the amount of
professionalization present in government, and the power of the governor within the state.
Early scholarship on state policy making focused on the socio-economic
differences among the states. Differing from the case study approach that political
scientists employed, economists such as Fabricant (1952) and Fisher (1961) utilized

6

Hero (1998) looks towards the differences in terms of social diversity and how the
ethnic and racial composition of a state influences the political processes, institutions, and
polices of a state. However, a later study by Barrilleaux (2006) concludes that racial
diversity is not as important of a factor to explain state policy liberalism as various forms
of party competition within the state and public opinion are.
11

government reports detailing the expenditures of state and local governments to
determine why expenditures varied by state. Fisher found that per capita income,
population density, and degree of urbanization explained “a considerable amount” of the
differences among states levels of expenditures (Fisher 1961 at 355). However, Fisher
did not attempt to include variables that represent the political differences of the states; he
excludes them because they are not easily quantified.
Political scientists responded to this literature by including economic variables in
their studies of the state political process (Peters 1972). For example, in their study of
state welfare policy Dawson and Robinson (1963) found that economic factors, such as
per capita income, had higher correlations with welfare policy outcomes than political
party competition. In a study of the effect of legislative apportionment on state policy,
Hofferbert writes that “structural characteristics…do not seem to go very far toward
explaining the kind of policies produced in the states” (Hofferbert 1996, at 82). These
are but two examples of a line of research in the 1960’s that produced similar conclusions
(Dye 1968, Sharkansky 1968).
Questions were raised about this body of work in respect to how the authors
utilize systems theory for their underlying theoretical framework and how the system
inputs and outputs are operationalized (Jacob and Lipsky 1968, Rakoff and Schaefer
1970). After changing the dependent variable analyzed to the redistributive impact of
revenues and expenditures, Fry and Winters (1970) found that political variables are
more powerful than socio-economic variables in explaining policy variation among the
states.

12

Wright, Erikson, and McIver (1987) address this contest between socioeconomic
and political explanations when attempting to determine the role of public opinion on
state policy. After describing how some literature interprets the significance of
socioeconomic variables as reflecting the demands of the public on inputs, Wright et al.
suggest that these variables are actually proxies for an underlying public opinion. If this
interpretation of why socioeconomic variables have such explanatory power in predicting
state policy outcomes is correct, then including a measure of public opinion should reveal
that public opinion is the actual determinant. An alternative explanation for the statistical
significance of socioeconomic variables is that they reflect nonpolitical economic
development (Dye 1968). State policy is the result of a predictable sequence; as a state
develops, it is then able to provide services to its citizens. If this explanation is correct,
then public opinion as a political variable should have little to no explanatory power.
Their analysis found that public opinion was a strong predictor of state policy and state
political processes and institutions matter.
Another source of variation among the states is the amount of policy
experimentation or innovation that they undertake. The success of states in fixing their
problems is what is meant by the metaphor of states as laboratories and can lead to other
states adopting the same policy. States experience problems to varying degrees and when
a problem is severe it provides a political opportunity for policy change (Berry and Berry
1994). Also states may have policy entrepreneurs, political actors that promote policy
ideas, which raise the probability of a state adopting a policy (Mintrom 1997). Policy
innovation is also influenced by the available resources within the state (Hill and
Leighley 1992). States with greater resources are able to take risks and afford policy

13

failures. The state’s political culture may also influence a state’s proclivity towards
innovation (Miller 1991).
In addition to the political context within a state, the relationships between the
states can influence their policies. States that share a border may also share common
problems that are better solved by cooperation. Interstate compacts, such as the Driver
License Compact, allow states to enter into agreements with each other (Zimmerman
2002). States may also adopt policies that have been adopted by nearby states. One
example of a policy that appears to spread by regional diffusion is the establishment of a
lottery within a state (Berry and Berry 1990). To avoid the losses that a state may incur
from its citizens participating in the lotteries of neighboring states, a state may choose to
create its own lottery. For other types of policies neighboring states may serve as
innovators that create knowledge and policy change is the result of social learning
(Mooney 2001a). For example, the economic development approach adopted by a state
is influenced by those employed by its neighbors (Saiz 2001).
A comparative focus on the variation of the state governments has uncovered
some marked differences in the effect of institutional contexts on the choices of policy
actors and the outcomes that follow. For example, in regards to fiscal policy, the
presence of ballot initiatives has been found to lower levels of spending by state
governments (Matsusaka 1995) and states with governors who possess greater power
over the state budget process are more likely to distribute benefits statewide (Barrilleaux
and Berkman 2003). These differences in institutional design have significant effects on
the content and effectiveness of government policy.

14

Common to each state is a democratic form of government. However, the exact
form of the government can vary in many ways. Another source of variation is whether
states have direct democracy measures in place to allow citizens to influence policy. The
initiative process allows citizens to write and enact legislation or amend the state’s
constitution (Bowler and Donovan 2008). Other states have an indirect initiative process
that allows the state legislature to become involved in the drafting process to varying
degrees. Through the initiative process citizens are able to enact government reforms and
address subjects like taxes, social issues, and environmental policy (Tolbert et al. 2001).
The initiative allows for interest groups that otherwise would not get access to the
political process to have an effect on policy-making (Boehmke 2005). Aside from the
immediate effect of an initiative, some scholars have found that the use of initiatives
increases voter turnout (Smith 2001, Tolbert et al. 2001) and levels of citizens’ political
knowledge (Smith 2002). The open nature of the initiative process has led some to
question whether direct democracy has a negative effect on minority groups (HaiderMarkel et al. 2007). Another potential effect of direct democracy is to alter the behavior
of elected officials depending on the policy under consideration (Gerber 1996, Lascher et
al. 1996).
A less direct method of citizen participation, but common to every state, is
through the electoral process. One of the reasons for states to have different policies is
that the composition of the citizenry is different. In an early study of southern state
policy towards blacks, V. O. Key (1949) found differences within what was previously
thought of as a single culture. Subsequent work extended the concept of culture to
describe the policy choices of states. Daniel Elazar (1966) created typologies of political

15

subcultures in the states along lines of moralistic, individualistic, and traditionalistic
cultures. Moralistic states prefer policies that seek to improve the general welfare.
Individualistic states prefer policies that emphasize a limited role of government.
Traditionalistic states prefer government to maintain the existing conditions within a
state. However, the utility of this measure as a proxy for public preferences is limited.
Schiltz and Rainey (1978, 414), in comparing Elazar’s political culture classifications
with survey data, conclude that a scale of political culture should not be considered
“accurate representations of reality.”
Survey data would be the direct manner to measure the policy preferences of a
state’s citizenry; however the cost to do so for 50 states has proven prohibitive. In an
attempt to create measures of public preferences, scholars have used census data and
national surveys to simulate public opinion for the states (Poole et al. 1965, Weber 1971).
The studies that used simulations found a strong link between policy and public opinion
(Weber and Shaffer 1972), but concerns were raised about the assumptions that were
required in computing the measures of opinion (Seidman 1975). Erikson (1976) was able
to use survey data from the 1930’s on the death penalty, child labor prohibitions, and the
right of women to serve on juries, to correlate the opinion of the states with the policy
that existed in those states, reaffirming the belief that public opinion can influence policy.
In an attempt to better study state politics, Wright, Erikson, and McIver (1985)
created a measure of opinion and partisanship. Previous studies on the opinion and
policy linkage required policy area specific measures of policy. The new approach would
seek to create overall measures of a state’s preferences. These measures would allow
researchers to compare all 50 states and the policy outputs studied would no longer be

16

limited to the subject area that was asked in a single survey question. In order to get a
representative sample, poll questions from many policy areas were coded to generate a
value on a liberal or conservative ideological dimension. The conceptualization of public
preferences would allow a researcher to examine the policy and opinion linkage as long
as it could be reasonably assumed that the policy outputs were part of the liberalconservative ideological dimension.
Using their aggregate public liberalism measure, Wright, Erikson, and McIver
(1987) examine its influence on an index of state policy liberalism. They used eight
policy issues to create an index based on a liberal-conservative ideological scale. They
concluded that state public opinion was the major determinant of the ideological direction
of state policy. This relationship was tested again in their text Statehouse Democracy
(Erikson et al. 1993) with similar result. In a review essay, Burstein (2003) concludes
that most scholars accept the conclusion that public opinion influences public policy.
This conclusion comes after decades of research into the question by numerous scholars
since the work of Erikson, Wright, and McIver (Erikson et al. 1993, Wright et al. 1987).
Scholars have continued the study of state policy in two manners. On the one hand, the
linkage between policy and opinion could be examined at the macro level by comparing a
measure of a state’s overall policy liberalism to a state’s overall opinion liberalism
(Barrilleaux 1997, Erikson et al. 1993, Gray et al. 2004, Wright et al. 1987). On the other
hand, one could examine a specific policy area to determine whether the linkage between
opinion and policy survives the aggregation of numerous policies. Studies of a state’s
abortion policy have repeatedly shown that conservative states are more likely to enact
restrictions on obtaining abortions (Camobreco and Barnello 2008, Norrander and

17

Wilcox 1999). The opinion and policy linkage has been found in diverse set of policies
such as the presence of environmental regulations (Hays et al. 1996), the level of welfare
benefits distributed by a state (Fording 1997, Volden 2002), and the strictness of
campaign finance regulations (Pippen et al. 2002). Haider-Markel and Kaufman (2006)
conclude that public opinion was significant for the adoption of hate crime laws including
sexual orientation as a protected group. The timing of passage of any hate crime law
was found to be influenced by public opinion by Grattet et al. (1998). In their study of
same sex marriage bans, Lewis and Oh (2008) found that the link between public opinion
and policy strengthened over time.
The state policy literature has repeatedly demonstrated that among the various
influences on state policy there is a link between public opinion and the policy outputs of
a state. However, this literature does not consider how the judiciaries of the states have
shaped policy. For example, the dependent variable in Statehouse Democracy, policy
liberalism, is an index composed of primarily legislative activities like spending and
regulation. The explanation offered for the linkage revolve around the state legislature as
the policy making actor and the two party system for keeping the elected government in
line with public opinion. According to Barclay and Birkland (1998), because policy
scholars view the judiciary in the traditionally legal sense of solving individual disputes
using agreed upon principles policymaking is outside of the court’s function. This
description of the judiciary limits the court’s importance only in areas of constitutional
decision-making, in cases where individuals go to the court as a last resort, or as
participants in implementing but not initiating policy change. However, public law

18

scholars have taken a different approach to the role of courts and treat the judiciary as a
political institution.
Judicial Decision-Making
Like other political actors, a wide range of theories have been applied to explain
how and why justices decide cases in the manner they do (Baum 2006, Gibson 1983).
Much of our understanding of the judicial making process stems from the exemplar of the
Supreme Court of the United States. While campaigning for the ratification of the
Constitution, Alexander Hamilton wrote in Federalist #78 that the future U. S. Supreme
Court would regard the document as the fundamental law of the United States. In those
instances when the Constitution did not speak directly to the controversy in question,
justices would be constrained from imposing their own will on the people by the law
7

itself. The need for a legal education to understand the controversy relative to the large
number of precedents and principles would also teach jurists a respect for the law and the
need for stability.
The judiciary, like any other institution, needs to have the support of the citizenry
and be viewed by them as a legitimate authority to be an effective governing body. As
Justice Frankfurter noted in Baker v. Carr: “The Court's authority - possessed of neither
the purse nor the sword - ultimately rests on sustained public confidence in its moral
sanction” (Baker v. Carr 1962 at 296). The Constitution does not provide the judicial

7

Hamilton Federalist #78 states: “To avoid an arbitrary discretion in the courts, it is
indispensable that they should be bound down by strict rules and precedents, which serve
to define and point out their duty in every particular case that comes before them; and it
will readily be conceived from the variety of controversies which grow out of the folly
and wickedness of mankind, that the records of those precedents must unavoidably swell
to a very considerable bulk, and must demand long and laborious study to acquire a
competent knowledge of them.”
19

branch with any formal mechanism to enforce its orders. The only authority the judiciary
8

holds is as the voice of the law. As a result, the judicial process was designed to
proceed in a manner that would foster trust and derive its power from its perceived
legitimacy. Judges have described their role in this system as ensuring that decisions are
the result of the application of the relevant facts of a controversy to the plain meaning of
a statute, precedent, or constitution. Written and published opinions of a court often
follow a pattern of reciting the facts of the case with the application of a legal standard to
the case at hand.
In the early twentieth century a number of legal commentators began to offer a
new perspective on the judicial process. Individuals such as Oliver Wendell Holmes, Jr.,
Jerome Frank, and Roscoe Pound spoke of the law as a temporary phenomenon. They
believed the outcomes of cases depended more on who was participating as the arbiter
than the controlling law. These scholars thought the application of social science
research and theory to studies of judicial decision-making would enhance the effort to use
litigation as an instrument for positive social change.
Jerome Frank (1930) wrote one of the first examinations into the decision-making
processes of judges. Frank concluded that when judges first examine a case, they identify
their desired outcome and then they publicly present the legal reasoning that justifies
8

Al Gore’s concession speech on December 13th 2000 provides an excellent example of
the Court’s success in being considered the voice of the law: “Over the library of one of
our great law schools is inscribed the motto, "Not under man but under God and law."
That's the ruling principle of American freedom, the source of our democratic liberties.
I've tried to make it my guide throughout this contest as it has guided America's
deliberations of all the complex issues of the past five weeks. Now the U.S. Supreme
Court has spoken. Let there be no doubt, while I strongly disagree with the court's
decision, I accept it. I accept the finality of this outcome which will be ratified next
Monday in the Electoral College. And tonight, for the sake of our unity of the people and
the strength of our democracy, I offer my concession.”
20

their preferred result. It is not difficult for a judge to support his or her chosen decision
given the large body of precedent. Frank describes the root of a judge’s decision as a
hunch. To explain the composition or source of these hunches, Frank (1949) later utilizes
Gestalt psychology, which posits that human responses are not an aggregation of smaller
distinct thoughts, but rather a single complete thought. The completeness of a thought is
the reason that individuals often find it difficult to explain their decision-making process.
The published opinion announcing the court’s decision is necessary to preempt the critics
of the decision and to justify the judge’s own actions. This public explanation reinforces
the apparent existence of the traditional legal model. The introduction of psychological
theories to this question was as far as Frank would go. Frank thought a lifetime of
experiences combined to form the personality of a justice. Frank believed that looking at
the judge’s background and personal characteristics, such as education, family history,
political affiliations, and their personality traits both temperament and intellectual, would
prove to be a useless exercise.
One of the earliest political science attempts at understanding the voting behavior
of justices was completed by C. Herman Pritchett (1948). His tome, The Roosevelt
Court, examined the changes to the law made by President Franklin Roosevelt’s
appointments to the Supreme Court. The first part of Pritchett’s analysis found that
justices routinely dissented in blocs. In examining these blocs, Pritchett found a clear
distinction between the justices. While this work mainly examined the patterns in which
justices joined the written opinions of their colleagues, it demonstrated that the product of
the Court depended upon its composition. In the second part, Pritchett analyzed the
direction of the votes of justices on cases involving claims of personal liberty. Pritchett

21

found that the separation of justices into blocs could be summarized on a liberal –
conservative dichotomy depending on the composition of the court for each term.
Examining judicial voting behavior in the backdrop of the New Deal made the effect
more visible. As new justices were elevated to the bench, many recent precedents were
overturned and the expansion of the federal government was allowed to continue.
Pritchett did not investigate the source of the differences between the two primary
blocs on the Court because of the influence that the composition of the Court has on the
positioning of a justice within a bloc. For example, Justice Black may have originally
been viewed as a liberal justice when he was the only Roosevelt appointee among the
larger number of conservative justices on the court. Yet, ten years later, when Roosevelt
had made eight other appointments to the Court, Justice Black may no longer be a part of
the liberal bloc relative to the new members. Pritchett’s work did not attempt to solve the
question of why justices vote the way they do, but it did set a number of scholars down
the path of investigating why justices vote in ideologically consistent patterns.
Inspired by the unanswered questions left by the work of Pritchett, Glendon
Schubert wrote his text The Judicial Mind (1965). Schubert framed his work as an
attempt to answer the question of what role a public official’s personal beliefs play in
their public acts. Methodologically, Schubert utilized scaling and dimensional analysis to
construct scores for the justice’s personal values and the case facts. From the
combination of values and case facts, Schubert was able to construct what he termed
ideal points. These ideal points would place a justice within an attitudinal space. The
ideal points work more along the lines of a pivot or tipping point. When the facts of a
case change and go beyond what a justice believes is allowable, the justice would be

22

expected to change his vote in the other direction. While Schubert went into greater
depth in arranging the votes of justices on an ideological scale, there was little
investigation into why the justices vote in the direction that they do. Schubert’s analysis
went beyond the blocs identified by Pritchett, but the scales offered did not present a
causal theory.
Some of the first research to examine the reasons behind an individual justice’s
vote utilized attitude theory (Spaeth 1972, Spaeth and Parker 1969). In their work,
Supreme Court Decision Making, Rohde and Spaeth (1976) combine attitude theory with
a rational choice framework to develop their explanation of how the justices of the U.S.
Supreme Court decide to vote. As with many other rational choice frameworks, Rohde
and Spaeth make a number of assumptions some of which were originally outlined in an
earlier work by Spaeth (1972). The first assumption is that the justices of a court are goal
oriented and they make decisions with the achievement of these goals as the primary
concern. The resulting vote should be the alternative that is closest to achieving that
goal. A second assumption is that the justices recognize the institutional structures within
which they operate. This assumes that the justices understand that there is the possibility
that their decision could potentially be either ignored by those charged with
implementing the policy contained in the decision, be circumvented by additional
legislation, overruled by the passage of a constitutional amendment, or impeached and
removed by Congress. The third and final assumption is that decisions are dependent on
the particulars of the cases before them. Put simply, a justice cannot vote to overturn the
death penalty in a case that does not raise it as an issue.

23

Rohde and Spaeth argue that preferences are composed of an individual’s beliefs,
attitudes, and values. These factors are constructs and as such cannot be observed. Their
function is to aid in the explanation of a justice’s personal policy preferences. The beliefs
of an individual combine to form an attitude. Rohde and Spaeth assert that attitudes can
be measured by the use of scaling techniques on the votes of the justices in similar cases.
Values are similar to attitudes, but are more general in nature. To measure values, Rohde
and Spaeth propose to examine the interrelatedness among attitudes to define values. In
doing so, they identified three major values, civil liberties, equality, and economic
activity, to account for 85% of the cases in their sample. Using the major values in
combination with two additional minor values, privacy and taxation, Rohde and Spaeth
were able to classify the justices according to an overall value system. To test their
theory they classified cases to identify the most likely basis for the Court’s decision and
then applied the scaling scores for the justices on that value. After using their theory to
predict cases, the authors conclude that justices base their decisions more on the situation
than the characteristics of the litigants.
The broader theory of attitudes rather than the law forming the basis of judicial
decision making was further developed in Spaeth’s later work with Jeffrey Segal (Segal
and Spaeth 1993, 2002). In this work, the authors provide evidence to assert that the
attitudinal model can explain the voting behavior of U.S. Supreme Court justices. Their
theory is explained succinctly in the following quote:
“[The Attitudinal] model holds that the Supreme Court decides disputes in light of
the facts of the case vis-à-vis the ideological attitudes and values of the justices.
Simply put, Rehnquist votes the way he does because he is extremely

24

conservative; Marshall voted the way he did because he was extremely liberal.”

9

In their text, Segal and Spaeth provide a number of tables, graphs, and examples to
demonstrate that the votes of the justices fit on the liberal – conservative dimension.
These voting patterns are also demonstrated to be highly correlated with the views of
public commentators in four of the nation’s leading newspapers at the time of their
confirmation (Segal and Cover 1989) and again their previous votes on similar cases.
While some scholars were critical of the empirical evidence for the model (see Baum
1994), others confirmed the central premise with courts other than the United States
Supreme Court. Scholars have reported evidence of the attitudinal model operating on
the United States Court of Appeals (Songer et al. 1994), state supreme courts (Brace et al.
2000b) and the Supreme Court of Canada (Ostberg et al. 2002).
To protect their authority it is believed that the judiciary has fostered “the myth of
legality” (Scheb and Lyons 2000). According to the myth of legality, judges base their
decisions on the plain meaning of the statute, original intentions of the drafters of the law,
and previous legal decisions. When two laws or decisions are in conflict, judges are to
balance the concerns of both in their decisions. The use of legal symbols such as black
robes, a high bench, and a secretive decision-making process were designed to bolster the
perception that justices use objective legal reasoning and not their own personal
preferences when reaching a decision.
In an effort to examine whether justices live up to the myth of legality, Spaeth and
Segal examined the voting behavior of United States Supreme Court justices in both
landmark and non-landmark cases and their progeny to determine whether the principle
9

Segal, Jeffrey A. and Harold J. Spaeth. The Supreme Court and the Attitudinal Model
Revisited. Cambridge, UK: Cambridge UP, 2002 at 86.
25

of stare decisis was followed later by members of the Supreme Court (Segal and Spaeth
1996, Spaeth and Segal 1999). When justices disagree with the outcome in a case, they
rarely change their voting preferences on subsequent cases and vote to uphold the
precedent that they originally dissented from. According to their analysis, only two
justices, Stewart and Powell, could be viewed as being influenced by precedent and this
is at a level of 33% of the votes in their analysis. Critics of this work have correctly
pointed out that there are numerous aspects to a law-based model of judicial decision
making and these factors may also be in conflict with stare decisis (Brisbin 1996).
Another empirical examination of the effect of the law on the decisions of justices
can be found in Richards and Kritzer’s (2002) identification of jurisprudential regimes.
According to the authors, a jurisprudential regime is created when a case determines the
factors that future cases will be decided upon and the amount of scrutiny that the court
will apply to a case. This concept appears to be simply just a focusing on precedent
under a different label.
For courts other than the United States Supreme Court, a concept that is consistent
with the ideal of the legal model is the issue of compliance. According to the legal
model, all lower courts should not only be bound by their precedents but also those of a
higher court. Compliance with higher courts is often assumed because the lower courts
follow the practice of citing precedents in their written opinions. To examine the validity
of this expectation Songer and Sheehan (1990) study the impact that two U.S. Supreme
Court decisions Miranda v. Arizona and New York Times v. Sullivan had on the areas of
criminal procedure and libel respectively. The authors find high levels of compliance
with both decisions, yet the impact of the decisions varied. Following Miranda, the

26

percentage of decisions supporting the criminal defendant remained virtually the same,
while support for libel defendants increased following the New York Times decision. The
limitations of this study are recognized by the authors themselves and the impact of the
personal preferences of the justices was not considered in their analysis.
This line of work demonstrates that the decision-making process of justices
should be treated similarly to other political actors. While justices frame their decisions
relative to legal principles or precedents, arriving at that decision is mostly an individual
and personal enterprise that has the potential to be influenced by external conditions. As
Gibson (1983) notes: “Comprehensive theory can best be developed through models that
incorporate influences stemming from various levels (e.g., group, institution,
environment) but that are ultimately focused on the individual. (pg. 8)”
Public Opinion and the Courts
Given the democratic nature of governance in the United States, political
scientists often explore the link between the positions of the citizenry and the actions of
public officials and institutions. The literature examining the relationship between public
opinion and judicial decision-making has produced conflicting results. Those who
suspect there is a relationship between the courts and the public point to the fact that the
decisions of the courts are often in line with the government or public opinion. In his
analysis, Barnum (1985) concluded that the U.S. Supreme Court is not as
countermajoritarian as it had been previously thought. In a number of policy areas, such
as school desegregation, access to contraception, interracial marriage, abortion, and
women's rights, a decision of the U.S. Supreme Court was supported by a majority, or
near majority, of the citizenry. However, there was also some evidence that the U.S.

27

Supreme Court ruled contrary to public opinion in cases involving school prayer and the
implementation of the death penalty. Barnum does not conclude that the Court is
following the wishes of the public when deciding these cases, as he only attempted to
answer the question of whether the Court has acted in the countermajoritarian fashion.
Marshall (1988) also found that in situations where a clear majority supports the issue
before the Supreme Court, approximately 60% of the time the Court’s decision agrees
with the public’s preference. Research in this vein does not assert nor find a causal link
between the results of a case and the public’s opinion on an issue. Furthermore, the
anecdotal nature of these analyses find as many instances of the U.S. Supreme Court
almost equally going against public opinion as they do with it. Finally, these studies do
not examine the votes of individual justices and accordingly do not aid in understanding
the decision making process.
In an effort to find a causal relationship between public opinion and the outputs of
the Supreme Court, Mishler and Sheehan (1993) used Stimson's (1991) Mood Indicator
in finding that the overall policy liberalism of the U.S. Supreme Court corresponds to
changes in the public mood and the effect was strongest after a 5-year lag. While being
careful to not assert too strong of a relationship between the public and the decisions of
the Court, Mishler and Sheehan assert that the Court does respond to changes in the
public mood and the relationship may be reciprocal in that the Court’s decisions
legitimize changes in public opinion towards policy. Further analysis has also shown that
the Court’s responsiveness depends on the policy area under consideration (Link 1995).
In their critique, Norpoth and Segal (1994) reanalyze the data used by Mishler and
Sheehan and contend that any effect of public opinion is likely a result of changes to the

28

composition of the Court. As the membership of the court is changed, the views of the
public are incorporated into this updated court. Norpoth and Segal also find fault with
the methodological approach taken by Mishler and Sheehan and suggest that introducing
a variable capturing the realignment of the U.S. Supreme Court due to Nixon’s
appointees renders the effect of public mood insignificant. Utilizing an alternative model
specification Stimson, MacKuen, and Erikson (1995) find the Court responding to
changes in public opinion sooner than Mishler and Sheehan, but believe it is not strong
enough to make a definitive claim. McGuire and Stimson (2004) employed an alternative
measure of the individual justices’ liberalism to reexamine the relationship between
public preferences and judicial decision-making. Their study concludes that there are
both the indirect effects described by Norpoth and Segal and that the direct relationship
between the public and the U.S. Supreme Court is stronger than was previously reported
by Mishler and Sheehan.
Flemming and Wood (1997) examine the individual justice’s response to changes
in the public mood. They concluded that individual justices were similar to each other in
their response to changes in the public mood across and between all issue areas.
However, similar to the findings of Stimson, MacKuen, and Erikson, Flemming and
Wood conclude that the reaction of the Court to changes in the public mood is quicker
than Mishler and Sheehan had originally speculated, but also that the impact is rather
small. Also, Flemming and Wood demonstrate that the responsiveness of the Court
varies among the justices and across issue areas. Given the relative isolation of the
federal judiciary from the electorate, it is not at all surprising that the effect of public
opinion on judicial decision-making is not substantial, if at all. The weak findings of a

29

small link between public opinion and the decisions of the U.S. Supreme Court are likely
due to changes of the personnel to the Court and not a conscious decision to keep the
Court in touch with the populace.
The Importance of Context
There is conflicting evidence for public opinion having an impact on U.S.
Supreme Court justices, but there have been a number of studies supporting this linkage
at lower court levels. The literature on judicial decision-making has primarily focused on
explaining the internal and ideological behavior of the voting decision, rather than the
10

influence of exogenous conditions and broader context.

Seeking out a new institutional

approach that explains “behavior taking the cultural contexts that gives it shape,
direction, and meaning” into consideration, Clayton and Gillman conclude that research
on judicial actors needs to acknowledge that “different contexts make it more or less
possible for individuals to act on different sets of beliefs” (1999, 3). The uncommon
nature of the United States Supreme Court’s institutional design makes generalization to
other judges untenable (Tate 1981). To avoid this problem, increased attention has been
given to judiciaries with different institutional designs operating within a variety of social
environments. The examination of multiple high courts alleviates the concerns that
generalizing beyond a single institution can create.
Kuklinski and Stanga (1979) examined the sentencing practices of California trial
courts following an initiative that proposed to decriminalize the personal use of
marijuana. Employing the vote supporting the initiative at the county level as an
10

According to McCann: “[g]iven that mobilization of the law in practice is highly
context dependant, it again stands to reason that adequate analysis of legal tactics should
identify those contextual variables most and least favorable to movement success” (1994
p. 11).
30

independent variable, the authors found that after the election the sentencing practices of
justices changed to become more attune to the desires of the county electorate. Gibson’s
(1980) examination of criminal trial courts in Iowa concluded that judges were perceptive
to the electoral environment at the county level. The trial judges in Iowa rode circuit by
traveling from county to county, and the sentencing behavior of a judge would vary
depending on their perception of the environment. Both of these articles stress the
electoral connection between the judges and the electorate.
Among the small number of studies on state high courts, Ragland (1995)
concluded that the decisions of the Texas Supreme Court were largely consistent with
public opinion. By matching statewide public opinion polls taken between 1978 and
1994 with decisions of the court, Ragland found that 76% percent of the court’s outcomes
corresponded with the position of the public. However, Ragland’s study does not focus
on the decision-making process, it instead attempts to determine whether the final
decision is in line with public preferences.
Brace and Boyea (2007) examine the vote of individual justices to reverse a death
sentence. Using a series of comparisons between elected and appointed justices, the
authors conclude that elected justices are less likely to overturn a death sentence than
appointed judges are. However, there is no evidence presented to the reader that the
differences in percentages meet conventional levels of statistical significance and the
study does not conduct a multivariate analysis.
In another study, Brace and Boyea (2008) conclude that public opinion only has a
direct effect in those states with popular elections. Their empirical analysis found that
public opinion towards the death penalty, as measured by Brace et al. (2002), had a

31

statistically significant effect on the vote of whether to overturn a death sentence for
those justices who were subject to regular elective formats, defined as partisan,
nonpartisan, or retention elections. However, the authors note that this finding stems
from a highly salient issue. Furthermore, courts are often asked in death penalty cases to
evaluate whether the trial court has followed the proper procedure rather than establish
policy that will affect future cases.
As Devins (2010) notes that there are multiple mechanisms in place that suggest
state high courts need to be concerned about the consequences of their decisions. In
regards to the public, in many states citizens have the ability to remove individual justices
in a future election or are able to reverse a decision through a direct democracy
mechanism. Fortunately, the justices of state high courts are in a unique position to
predict whether their decisions will incite a backlash (Sager 1985). One simple reason
that state justices are able to do so is that they have been exposed to how the state
operates. An analysis of the personal characteristics of justices serving on state high
courts in 1994 and 2000 revealed that over 60% of justices were either born or attended
law school in-state (Bonneau 2001).
Conclusion
Scholars studying policy change at the state level have identified numerous
influences on the overall direction of policy. Differences in the composition of the
citizenry, institutional design, the political environment have all been used to explain the
policy making process of state governments. However, most of this research does not
include the judiciary as an agent of policy change. As a result, it is unknown whether the

32

explanations offered to explain state policy change have an influence on the state
judiciaries.
As Devins (2010) notes that there are multiple mechanisms in place that suggest
state high courts need to be concerned about the consequences of their decisions. In
regards to the public, in many states citizens have the ability to remove individual justices
in a future election or are able to reverse a decision through a direct democracy
mechanism. Fortunately, the justices of state high courts are in a unique position to
predict whether their decisions will incite a backlash (Sager 1985). One simple reason
that state justices are able to do so is that they have been exposed to how the state
operates. An analysis of the personal characteristics of justices serving on state high
courts in 1994 and 2000 revealed that over 60% of justices were either born or attended
law school in-state (Bonneau 2001).
The previous research on the judicial decision-making process of the United
States Supreme Court demonstrates that justices do not make decisions based solely on
the law and judges are able to pursue their own policy preferences. However, at the same
time, judges can also be influenced by institutional and environmental factors. Public
opinion has been found to be one of these environmental factors. Early research on state
supreme courts, which are institutionally more diverse than the Unites States Supreme
Court, has found that public opinion may have an influence on the decision-making
process. Yet thus far, the only study has been on the individual appeals of those
convicted of the death penalty and not on cases that have public policy implications. This
dissertation will apply what Erikson, Wright, and McIver have called “the dominant
influence on policymaking in the American states” to the state judiciary (1993at 244).

33

Chapter 3 Data and Methods

In this chapter, I outline how previous research on the judicial decision-making
process of state high court justices informs the design of the empirical models. Then, I
describe how these concepts are measured and quantified in this analysis. Finally, I
discuss the factors that influence the design of the estimation strategy employed.
Previous research has validated the need for a comparative approach to the study
of the judiciary (Epp 1998). Research on the judicial decision-making process of state
supreme court justices has revealed that among the factors affecting a justice’s vote are
the institutional and contextual conditions in which they operate (Brace and Hall 1997).
As part of the laboratory of democracy, states have instituted an assortment of
institutional arrangements to balance accountability and legitimacy in the judicial branch
(Brace et al. 2001). From partisan elections to a life term, the full range of public
participation in the selection and retention of state high court justices exists. Along with
other differences, interested scholars have sufficient institutional variety to determine the
influence of public opinion on judicial decision-making and the resulting policy outputs.
The lack of debate over the responsiveness of state supreme court justices to
public opinion has been primarily due to the assumption that justices who face popular
elections must be, to some degree, responsive to the electorate in order to be re-elected.
However, this assumption has been largely untested. Previous studies of judicial voting
behavior have shown that justices subject to popular elections are more conservative in
their voting behavior (e.g. Brace and Hall 1993, 1997, Hall 1987, 1992, Hall and Brace
1992). It should be noted that these studies were performed on votes involving death
penalty appeals and built into an examination of a state's implementation of the death

34

penalty is the reality that a majority of the electorate in that state supports the policy
(Norrander 2000). I believe that examining a policy other than the death penalty, where
there are varying levels of support, will enhance our understanding of the influence of
public opinion on judicial decision-making and the policies that follow.
It is important to consider multiple policy areas when examining state policy
making (Gray 1974) and judicial decision-making (Epstein and Mershon 1996, Link
1995). As a result, I examine two broad policy areas where the judicial branch has
played a role in shaping state policy. Each policy area will receive its own empirical
analysis. I have chosen policy areas that would be salient to the public at large. The
policy areas were chosen to ensure that there would be a diverse number of cases across
time and space. These cases were selected because there is no national consensus on the
issues that are presented and the treatment of the groups is different across the United
States. Furthermore, these are two policy areas where the judiciary must work with the
other branches to create policy. Finally, I selected these two policy areas because there
are valid and sustained legal arguments on both sides of the issue that would allow the
justices to be open to considerations other than the law. In other words, there is an
opportunity for justices to engage in policy making instead of norm enforcement as with
death penalty appeals. If external and non-legal factors influence the judicial decisionmaking process, the two policy areas I have selected will allow for that influence to be
displayed.
Dependent Variable
The first analysis pertains to cases involving the treatment of gays and lesbians.
In Chapter 4, I discuss in more detail the types of policy questions that courts are asked to

35

answer. I expect that in states where the public is either more tolerant of homosexuality,
as measured by Brace et al., or more liberal generally, justices will be more likely to vote
to overturn a distinction based upon sexual orientation. The second analysis includes the
litigation of claims of gender discrimination. Specific cases are discussed in Chapter 5.
I expect that in states where the public is either more feminist, as measured by Brace et
al., or more liberal generally, justices will be more likely to vote to overturn a distinction
based upon gender. After examining each case opinion, the votes of the individual
justices were classified according to whether he or she voted to extend equal treatment to
the group claiming discrimination or to remove the legal distinction that denied equal
treatment. When the case syllabus did not explicitly list each individual justice
participating, a reference list of justices serving on a court was consulted (Langer
National Science Foundation CAREER Grant, SES-0092187).
The list of gay rights cases come from Daniel Pinello’s text Gay Rights and
American Law (2003). Pinello identifies cases concerning the rights of gays and lesbians
from 1981 – 2000 by searching the Westlaw database of published court opinions using a
series of keywords. The main search terms were “gay,” “homosexual,” “lesbian,” “same
sex,” “sexual orientation,” and “sexual preference.” After consulting with numerous gay
and lesbian rights interest groups about the results of his search, Pinello concluded that
“the odds for courts’ never referring to any of the six search keywords are negligible”
(Pinello 2003 at 165). I used the same procedure as Pinello to augment his list to the year
2004 by using the same search terms with the Lexis-Nexis database. A one-year overlap
between Pinello's list and the Lexis-Nexis search was used to confirm the reliability of
the process. All six of the cases identified by Pinello for state supreme courts in 2000

36

were also identified in the additional case collection. There were 94 cases identified for
the years 1980 - 2004.
The gender discrimination cases were obtained from a dataset provided by
Baldez, Epstein, and Walker (2006) from their article “Does the U.S. Constitution Need
an ERA?” The authors identified sex discrimination cases decided by state high courts
from 1950 – 2001 using a similar keyword search as Pinello. Baldez et al. also
“shepardized” their case list to look for previously unidentified cases that were cited by a
court in an already identified case. This process resulted in identifying 651 cases. Not all
of the cases will be included in the analysis as not all of the independent variables have
been measured back to 1950.
The 94 gay rights cases and 651 sex discrimination cases were coded using a
scheme similar to the State Supreme Court Data Project. Among the information
collected was the voting behavior of the individual justices. The dependent variable in
this analysis is a dichotomous indicator of the direction of a justice’s vote in cases where
discrimination was alleged. All of the votes were coded in a binary fashion to indicate
whether a justice voted in agreement with those seeking equal treatment. The
observations are pooled from a number of cases and individual justices may appear in the
dataset multiple times.
Measuring Public Opinion in the States
The primary independent variable for these analyses is an indicator of state-level
public opinion. However, developing measures of public preferences at the state level
has been a difficult enterprise, both generally and for specific policy issues. When the
high cost of a representative survey is multiplied by the 50 states, it becomes cost

37

prohibitive to generate a dataset similar to the American National Election Study or the
General Social Survey. Instead, scholars have had to develop alternative methods to
create measures of public preference at the state level.
One approach has been to perform computer simulations that combine
information from the national decennial Census and multiple national surveys to create
estimates of state-level public preferences (i.e. Weber 1971). This method is comprised
of multiple stages. First, national survey data is used to define the preferences of
predetermined voter types based upon demographic characteristics. Second, the Census
data is used to determine the number of individuals that appeared in each voter-type
residing in a state. Finally, the frequency of voter types and survey data are combined to
create an estimate that reflects the composition of the state and what the preferences of
those citizens are. The strategy behind this approach has recently been adapted to
incorporate Bayesian methods of parameter estimation (Park et al. 2006) and multilevel
modeling (Lax and Phillips 2009b).
Others have sought to acquire a sufficiently large random sample by aggregating
the responses from multiple surveys. This approach avoids the strong assumptions that
are required to simulate measures of public preferences (Seidman 1975). Wright,
Erikson, and McIver (1985), and later Erikson, Wright, and McIver (1993), pooled a
series of CBS/New York Times polls to estimate the self-reported political ideology of a
state’s citizenry. Although the polls were different, they used similar question wording
and random sampling within states. A benefit of this approach is that the aggregation of
these polls can be done for specific periods. For the models in this study, the measure
was computed by taking the average ideological identification over the period of the

38

sample. For sexual orientation cases, the years 1977 - 2003 were used and for gender
discrimination cases the years 1976 – 2001 were used to compute the ideological average.
The resulting variable has a range of values between -1 to 1, with increasing values
indicating a more liberal public opinion. This variable is static for the period of this
study.
Berry, Ringquist, Fording, and Hanson (1998) developed an alternative method to
creating state level estimates of public preferences because they believed that previous
attempts had ignored the potential for opinion to change over time. Berry et al. created
measures of citizen ideology by combining election returns with interest group ratings of
members of Congress in addition to the partisan composition of state government. The
use of interest group ratings allow for their measure to be comparable across states and
the inclusion of voting returns aids in creating a measure that reflects changes within the
citizenry. To create a similar measure for government ideology, Berry et al. focus on
ideology scores for the governor and major party delegations in the state’s legislative
bodies. These variables have a range of values from 0 to 100, with increasing values
indicating a more liberal populace or government. It should be noted that Berry et al. use
the term ideology in naming their measure and in an effort to distinguish among the
public opinion measures I will continue to employ their terminology when discussing
results of statistical analyses that include their measure. I am not using the term to refer
to how attitudes may be structured in an individual.
Norrander (2001) used the National Election Survey of the Senate (Miller et al.
1993) to generate state level estimates of public opinion. These surveys included a full
sample for each state that had a U.S. Senate election in 1988, 1990, or 1992. Another

39

advantage from using this data source is that the questions were identically worded.
These characteristics allow Norrander to create measures of state level public opinion for
various issues and the overall opinion of the state’s citizenry. The scale of values ranges
from 1 to 5, with higher numbers indicating a more conservative citizenry. I have
recoded the measure to be consistent with the other measures of public opinion by
reversing the direction to indicate a more liberal public as the measure increases.
Brace, Sims-Butler, Arceneaux, and Johnson (2002) extended the procedure of
pooling survey responses from many years with the goal of creating public opinion
measures for specific public attitudes. Brace et al. pooled the General Social Survey
from 1974-1998 and then disaggregated the responses at the state level. The resulting
indices capture the differences in opinion on specific policy issues such as abortion,
sexual orientation, and capital punishment, support for government spending on welfare
and the environment, and broader concepts such as tolerance, feminism, and religiosity.
Some of the indices were computed from a single question, while others were the result
of multiple questions. The Brace et al. measure for public preferences towards
homosexuals was derived from a respondent’s reaction to the question of when “sexual
relations between two adults of the same sex” is wrong. The resulting index ranges from
a 0 to 1 with higher numbers indicating more tolerance towards homosexuals. Their
index of public feminism was calculated from two questions. The first asked respondents
to evaluate the statement “Women should take care of running their homes and leave
running the country to men.” The second question asked the respondent to evaluate the
position that “most men are better suited emotionally for politics then women.” This

40

index has a range of 0 to 2, with increasing values indicating a higher amount of public
feminism.
There have been a number of articles debating the advantages and disadvantages
of the different measures. Berry et al. (Berry et al. 1998, 2006) argue that any measure of
public preferences should be temporally dynamic.

11

This assertion calls into question the

conclusion of others that intrastate public opinion is stable over time. Erikson, Wright,
and McIver (1993, 32) found “near-perfect stability” of state opinion and evidence of
change is mostly due to measurement error and subsequent study reaffirmed this
conclusion (McIver et al. 2001). Brace et al. (2004) reinforce this conclusion and argue
that the Berry et al. measure demonstrates movement because it is not based on surveys
but elite behavior that is more susceptible to change.
Erikson et al. (2007) and Brace et al. (2006) suggest that using the Berry et al.
measure to predict state policy would lead to misleading results due to validity problems
in the measure. In their study of interest group scores, Groseclose, Levitt, and Snyder
(1999) conclude that because interest groups use different bills on many issues, the scale
that is developed can include considerable measurement error and any use of these scores
need to adjust for movement in the underlying scale.
Another reason to include multiple measures of public preferences is that there is
disagreement concerning what the measures are truly capturing. Berry et al. originally
considered their measure to be state citizen ideology but they have subsequently found
that the measure is also highly correlated with a measure of policy mood or operational

11

“Nothing in the revised results…alters our conviction that citizen ideology varies
across states and over time, in ways that are relevant to our understanding of policy
choices made at the subnational level” (Berry et al. 2006 at 495).
41

ideology (Berry et al. 2007a). Berry et al. also argue that “scholars seeking to test
hypotheses about the impact of public opinion on public policy…should use a measure of
policy mood, not ideological self-identification” such as the Erikson et al. and Brace et al.
measures (Berry et al. 2007a at 127). However, Erikson et al. have argued that instead of
a state’s public mood, Berry et al. are actually capturing the state’s congressional
delegation’s roll call ideology (Erikson et al. 2007). Others also contend that the Berry et
al. measure reflects a concept other than public preferences and might be a measure of
elite opinion due to its non-survey source (Brace et al. 2004, 2006, Norrander 2001).
There is no consensus in the field about which measure is best (Berry et al. 2007a,
b, Brace et al. 2007, Erikson et al. 2007, Norrander 2007) and new measures continue to
be developed (Lax and Phillips 2009b, Park et al. 2006). Berry et al. (2007a, 127) have
gone as far to state that “no single measure adequately captures the concept of state
political ideology.” As a result, this dissertation will include both the specific issue
measure, from Brace et al., and the general measures of opinion, from Erikson et al.,
Berry et al., and Norrander. By estimating a separate model for each measure of public
attitudes, I will reduce the sensitivity of my analysis to the concerns of the validity of any
one specific public opinion measure when estimating their effect on judicial decisionmaking and any differences among the models may allow for greater explanation of how
the linkage works. Measures of public preferences that are computed from either
Bayesian or multilevel modeling will not be included due to their relatively recent
introduction and a lack of study commensurate with the measures discussed above.

42

Control Variables
The review of literature on judicial decision-making in chapter 2 above has
demonstrated that it is primarily an individual process. To control for the effect that
individual-level preferences have on the decision-making process, I have obtained
ideology scores for each justice sitting on state supreme courts from 1970 to 2004.
Traditionally, partisan identification was used to differentiate liberal and conservative
justices. Recognizing that the Democratic and Republican party labels have different
meanings across the United States led to the creation of an alternative measure of state
supreme court justice's ideology. These are available in the form of Party-Adjusted
Judge's Ideology (PAJID) scores (Brace et al. 2000b, Langer n.d.). Discrimination based
upon sexual orientation or gender are issues that can be identified as liberal and
conservative. I expect that justices who are more liberal in their ideology should be more
likely to vote to overturn a distinction based on sexual orientation or gender.
The differences in how justices are selected and retained in office have frequently
been found to influence judicial decision making behavior in a variety of situations
including death penalty appeals (Brace and Boyea 2008, Brace and Hall 1997),
challenges to abortion (Brace et al. 1999), campaign and election laws (Langer 2002),
workers’ compensation laws (Langer 2002), unemployment compensation laws (Langer
2002), and welfare benefits laws (Langer 2002). State governments have always been
free from federal influence when designing their judicial branch. Currently, only a small
number of states have followed the federal method and given justices a virtually life term

43

in office.

12

The most frequently used method of selection and retention is what is

referred to as the “Missouri Plan” or merit selection.

13

Under this system, typically

nominees are recommended to the political branches by a nonpartisan commission. The
commission is tasked with evaluating and recommending acceptable candidates to the
political branches for selection. Typically, a number of qualified candidates are
presented to the governor for selection. Depending on the specifics of the process, the
governor’s selection may be appointed to the court without the consent of the legislative
branch. If a justice seeks to remain on the court, then they run for retention in the form of
a plebiscitary election. With the approval of the public, the justice then remains on the
court for another term of office until he or she will face the electorate again.
Open elections are also used by the states to compose its judiciary. These
elections can be partisan, nonpartisan, or a combination of both. The purely partisan
states allow candidates for judicial offices to contest party primaries and if successful, the
nominees are placed on the general election ballot with partisan labels. The purely
nonpartisan states follow the usual primary and general election process, but candidates
appear on a nonpartisan section of the ballot. Two states, Ohio and Michigan, allow

12

For example, in Massachusetts the justices of the Supreme Judicial Court are
appointed by the governor and serve until they reach a mandatory retirement age of 70.
Similar methods are used in New Hampshire, where justices nominated by the governor
and approved by the executive council serve until they reach age 70, and in Rhode Island
where the justices nominated by the governor and confirmed by both the state house and
senate serve a life term.
13
Missouri was the first state to adopt this selection method in 1940, which was
endorsed by the American Bar Association in 1937. One commentator noted that this
adoption “deserves inclusion in any chronology of major events in the history of what has
been done to improve our courts.” (Elliott 1957)
44

political parties to hold a primary election to determine their candidates, but do not allow
the partisan label to be attached to the candidates on the general election ballot.

Table 1: Methods of Retention for State High Courts
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

Partisan Election
Retention Election
Retention Election
Nonpartisan Election
Retention Election
Retention Election
Elite Reappointment
Elite Reappointment
Retention Election
Nonpartisan Election
Elite Reappointment
Nonpartisan Election
Partisan Election
Retention Election
Retention Election
Retention Election
Nonpartisan Election
Partisan Election
Elite Reappointment
Retention Election
Life-Like Term
Nonpartisan Election
Nonpartisan Election
Nonpartisan Election
Retention Election

Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
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

45

Nonpartisan Election
Retention Election
Nonpartisan Election
Life-Like Term
Elite Reappointment
Partisan Election
Elite Reappointment
Nonpartisan Election
Nonpartisan Election
Nonpartisan Election
Retention Election
Nonpartisan Election
Partisan Election
Life-Like Term
Elite Reappointment
Retention Election
Retention Election
Partisan Election
Retention Election
Elite Reappointment
Elite Reappointment
Nonpartisan Election
Partisan Election
Nonpartisan Election
Retention Election

For some, the use of popular elections of judges has long been a controversial
method of judicial selection and an effort promoting reform continues today.

14

The

American Bar Association’s Commission on the 21st Century Judiciary espoused the
position that the popular selection of judges to state courts politicizes the judiciary and
reduces its legitimacy with the public (ABA 2003). Among the evidence the ABA relies
on is a 2001 survey commissioned by the Justice at Stake Campaign.

15

When asked

“[h]ow much influence do you think campaign contributions made to judges have on
their decisions,” only 5 percent of the respondents stated that they thought there was “no
influence at all.” Furthermore, 67 percent of respondents agreed with the statement:
“[i]ndividuals or groups who give money to judicial candidates often get favorable
treatment.” These figures led the commission to declare “[t]he time has come to
inoculate America’s courts against the toxic effects of money, partisanship, and narrow
interests” (ABA 2003). By removing judges from popular selection, the ABA believes
the public’s confidence and trust will be restored in the judiciary. The absence of
elections will remove the need for candidates to raise campaign funds, respond to the
questionnaires of interest groups, and participate in public debates and should result in
higher levels of public support for the judiciary. Merit selection has also gained the
endorsement of a number of other interest groups: the American Judicature Society, the
Brennan Center for Justice and the Constitution Project.

14

"Putting courts into politics and compelling judges to become politicians in many
jurisdictions has almost destroyed the traditional respect for the bench." (Pound 1962)
15
Survey results are available at
http://faircourts.org/files/JASNationalSurveyResults.pdf.
46

Recent reforms have been enacted with the desire to make judicial elections less
political. For example, in 2004 North Carolina held its first judicial election with a
number of new reforms in place. The selection and retention method used in North
Carolina changed from partisan to nonpartisan elections, added the option of public
financing for candidates, lowered limits on campaign contributions, and mandates the
publication of a voter guide to provide information about the candidates in appellate court
elections. Recently, other states have reformed the selection and retention methods by
removing partisan labels from the ballot or adopting the Merit Plan.

16

In this analysis, each state high court was coded for the type of retention method
that was used at the time the decision was made. The states that use a reappointment
system that incorporates elected officials rather than members of the public were placed
in a single category of elite reappointment.

17

In three states, Massachusettes, New

Hampshire, and Rhode Island, justices serve until they reach a mandatory retirement age
or die. These states have been classified as having a life-like term and are the baseline
category for the models below. Generally, justices who face popular elections to keep
their seats decide cases in a more conservative direction.

16

In addition to North Carolina, Alabama and Arkansas also switched to nonpartisan
elections in 2000 and 2001 respectively and South Carolina moved to a merit plan in
1996.
17
These states are: Connecticut (Gubernatorial reappointment and legislative
confirmation), Delaware (Gubernatorial reappointment from referral by judicial
nominating commission and Senate confirmation), Hawaii (reappointment by the judicial
nomination commission), Maine (Gubernatorial reappointment and Senate confirmation),
New Jersey (Gubernatorial reappointment and Senate confirmation), New York
(Gubernatorial reappointment from referral by judicial nominating commission and
Senate confirmation), South Carolina (legislative reappointment), Vermont (legislative
reappointment), and Virginia (legislative reappointment).
47

Because state high courts are reactive institutions and hear cases that have been
appealed to them rather than creating controversies to rule upon, not every case is
expected to be as equally significant to the policy making process. To evaluate the
expected importance of the case, I include a measure of whether a third party filed an
amicus curiae brief. The procedure of filing an amicus curiae brief allows an outside
party to provide the court with its position on the legal issues in the case (Comparato
2003). Caldeira and Wright (1988) describe the filing of an amicus brief as an avenue for
interest groups to inform the court of the significance of the decision beyond the
immediate effect to the litigants. I expect that as the number of amicus curiae briefs
increases, a justice will be more likely to vote to remove a distinction in the law due to
the significance of the decision to the policy process.
Today, most states have intermediate appellate courts in place to reduce the
workload of its high court while preserving the general right of a losing party to appeal an
adverse trial court decision. A potential side effect of this organizational feature is the
possibility that there is a difference in the cases that reach a high court depending on
whether the case was considered by an intermediate appellate court. A high court that
does not have an intermediate appellate court acting as a filter will hear cases that are less
controversial or routine (Canon and Jaros 1970). A study by Atkins and Glick (1976),
concludes that the inclusion of an intermediate appellate court in a state’s judicial system
alters the distribution of issues which come before the state’s high court. Previous
research on judicial decision-making behavior has found that dissent rates are higher in
those states with intermediate level appellate courts between themselves and trial courts
(Canon and Jaros 1970, Hall and Brace 1989). A dichotomous variable indicating the

48

presence of an intermediate appellate court in the state’s judicial system will be included
in the analyses. I expect that if statistically significant, the presence of an intermediate
appellate court would increase the probability of a justice to vote to overturn a law that
discriminates.
The level of political competition in a state will be included using the folded
Ranney Index (Ranney 1976). This index is based upon elections for legislative and
executive state offices and is composed of the proportion of success of the parties, the
duration of that success, and the frequency of divided government. The index has been
folded to remove the identification of which party is in control if there is no competition.
The folded measure now ranges from 0.5, indicating complete single party dominance, to
1.0, signifying complete competition, and the point at which dominance becomes
competition is 0.85. Research on voting in cases involving a death sentence has often
found that justices in politically competitive environments were more likely to vote in a
conservative manner (Brace and Hall 1997, Hall and Brace 1992). In states where there
is greater competition between the two parties, I would expect that a justice would be
more likely to vote to uphold the discriminatory regulation so as to not create a
potentially controversial vote that would be politicized.
I also include a measure state government ideology in recognition of the
interdependence of the judicial branch with the legislative and executive branches (Berry
et al. 1998). This variable is included to control for the preferences of the other branches
that would be charged with implementing or enforcing judicial decisions. I expect that
like public opinion, government ideology would have a positive effect on the vote of a
justice. Justices in states with more liberal governments should be more likely to vote to

49

overturn a discriminatory law because they would not face a state government unlikely to
enforce the decision.
The size of the legal community is measured by the Bureau of Economic Analysis
as the State Level Gross Domestic Product of Legal services. The BEA defines its
Standard Industrial Classification of Legal Services as “those establishments which are
headed by members of the bar and are engaged in offering legal advice or legal services.”
This variable is measured in millions of current U.S. dollars and has been lagged for one
year. This variable is included purely as a control to serve as an indication of the
availability of legal resources to potential litigants. I would expect that as the size of the
legal community increases, there would be a positive effect on the probability of
overturning a discriminatory law.
Criminal justice issues are among the most discussed topics in judicial campaigns
(Culver and Wold 1993, Goldberg et al. 2005, Goldberg and Sanchez 2004). Hall (1995)
found that a justice is more likely to vote to uphold a death sentence as the murder rate
increases. Additionally, in a study focusing on the effects of re-election success of
justices, Hall (2001) also concludes that the electoral success of incumbent justices in all
of the types of election were negatively influenced by increases in the murder rate. The
murder rate variable is taken from the Uniform Crime Reports compiled by the Federal
Bureau of Investigation and it represented the number of murders and non-negligent
manslaughter per 100,000 residents. This variable has been lagged by one year. There
are a number of criminal cases included in the sample for both policy areas. If the
murder rate has an effect, I expect that it would be in the negative direction.

50

Modeling Strategy
The parsimonious solution to estimating the influence of the various independent
and control variables on the dependent variable would be a single model with fixed
effects for each variable. However, the data in this dissertation has a multilevel structure.
There are three distinct levels contained within each observation. On the first level are
individual justices. These individual justices decide cases as a member of a larger panel
composed of all the justices on the court. For example, the Supreme Court of Michigan
has seven members that hear oral argument en banc and subsequently each justice casts a
vote to determine the outcome of the case. The second level of data contains information
on the specifics of the case itself. Each justice sitting on a case is presented with the
same facts and relevant laws. The third level contains information about the state’s
institutional structures and social environment. Beyond these levels, other groupings
could be established. Proceeding without consideration of the structure of the data
potentially ignores or violates assumptions that are necessary for inference. Maximum
Likelihood Estimation includes an assumption of how the errors are distributed and the
estimated parameters can be affected by residuals that have influence, resulting in
sensitivity to the misspecification of the error term.
Previous research on judicial-decision making has often ignored this detail of the
data structure and combined the votes of the justices together while disaggregating the
characteristics of the upper levels down to the individual level. When this approach is
taken the actions of individual justices are directly linked to higher level data and as a
result are assigned attributes that they do not possess as individuals (Heck and Thomas
2000). Zorn (2006) provides an example of how hypothesis testing and the conclusions

51

that follow are negatively affected by ignoring the presence of groups within data.
Furthermore, this approach often ignores information about the context in which the
decision-making behavior takes place.

18

As advised by Primo et al. (2007), robust standard errors will be incorporated
into the model to correct for the multi-leveled structure of the data. These clustered
standard errors will reduce concerns of non-exchangeability and other violations of
statistical assumptions associated with MLE (King 2001). Comparing MLE and GEE
model results, Zorn (2006) found that the decision of what unit to cluster the observations
on had more influence on the consistency of estimates than the estimation strategy used.
There are a number of options in determining on what level to cluster the standard errors
due to the structure of the dataset. Because the focus of analysis is the individual
justice’s voting behavior in different contexts, I will use robust standard errors clustered
at the individual justice level to improve the validity of hypothesis testing given this data
structure. Maximum Likelihood Estimation (MLE) will be used as the strategy to create
parameter estimates that predict the change to the probability of observing an outcome.

18

To address these concerns, multi-level modeling has been extended to a wide range of
political phenomena and there are a number of advantages to using this method (Gelman
and Hill 2007). Incorporating the structure of the data into the analysis will result in
greater efficiency and less bias while reducing the probability of falsely rejecting the null
hypothesis. Ignoring the hierarchical properties of the data leads to the violation of a
number of assumptions made when testing hypotheses (Burstein 1980).
While theoretically there are advantages of using a Multi-Level Model, for this
dissertation it is not a practical approach (Primo et al. 2007). Like the example in the
Primo et al. article, this model does not converge using modern computing software,
indicating that parameters cannot be generated. The model attempted in the Primo et al.
article had approximately 45,000 observations. In this study, one policy area has almost
700 observations and the second policy area has approximately 2,000. Theoretically, a
multi-level model would be the most appropriate estimation strategy, but it is not possible
in this study.
52

Another approach that has been used is the inclusion of a multiplicative
interaction term to allow the effect of one independent variable on the dependent variable
to vary within groups of other independent variables. Given recent work on the
significance testing when using interaction terms, additional attention will be given to
whether interaction terms are needed in the model and if so how they should be
interpreted (Ai and Norton 2003, Berry 1999, Berry et al. 2008, Berry and Rubin 2007,
Brambor et al. 2006, Braumoeller 2004, Kam and Franzese 2006). If there is an
interaction between public opinion and retention method, I expect that justices
Research on the appropriateness of interactions has developed some practical
suggestions on how to determine whether an interactive relationship exists. Berry and
Rubin (2007) use Monte Carlo methods to demonstrate that the practice of examining the
statistical significance of the product term’s coefficient is not a sufficient test for an
interaction between two independent variables. Instead, Berry and Rubin suggest
measures of model performance be examined to evaluate the appropriateness of including
interaction terms. One such method is to compare the area under the receiver-operator
characteristic (ROC) curve between the base model with interaction terms and the model
that included interaction terms. Berry and Rubin argue that among three possible fitbased estimators, the ROC-based estimator of the relationship was the most likely to
correctly identify that an interactive relationship existed and least likely to indicate an
interactive relationship where there was none.
Conclusion
In this chapter, I have described the variables and methods that are going to be
used in this analysis. The dependent variable of individual justice’s votes was obtained

53

by coding a list of gay rights and sexual discrimination cases from previous works.
Given the difficulty in measuring public preferences at the state level, I will perform the
analysis four times, each with a different measure of public opinion. This will ensure that
the conclusions made are not dependent on the choice of competing public opinion
measures. Two separate policy areas will be analyzed to ensure that the findings are not
dependent on the type of cases in the analysis. If the results indicate that there is an
opinion-policy linkage, then further analysis concerning whether elections provide that
linkage will be performed.

54

Chapter 4 Discrimination Based on Sexual Orientation
The regulation of private behavior in the consideration of the public good includes
a number of highly personal actions. Within the principle of a state’s inherent police
powers are the ability to regulate health, safety, welfare, and morals of its citizens. State
governments use this ability to create laws that prohibit specific physical acts and the
possession of obscene objects. Because of this capacity to draw distinctions between
acceptable and prohibited behavior, groups are established and future laws can be
developed that build upon or extend the original groupings.
Discrimination based upon sexual orientation has been rooted in the consideration
of homosexuality as an individual trait, such as a personal moral failing, criminal act, or
medical condition, and not a characteristic of a group of people (Slovenko 1985). As a
result of this characterization, discrimination against gays and lesbians has taken many
forms and is persuasive throughout society. One source for discrimination was the use of
religion to construct the criminal laws of governments. Sodomy was considered a sin and
a crime against nature to be punished by death.

19

In the United States, the punishment

for sodomy has decreased over time and it is no longer a crime. Another basis for
discriminating against gays and lesbians was the belief that homosexuality was an illness
that could be treated and cured. This perspective of homosexuals as mentally ill gave
further reason to exclude gays and lesbians from certain occupations like teaching.
Although homosexuality is no longer considered a mental disorder by the American

19

According to the Bible “If a man also lie with mankind, as he lieth with a woman,
both of them have committed an abomination: they shall surely be put to death; their
blood [shall be] upon them.” (Leviticus 20:13, King James Version of 1769)
55

Psychiatric Association, it is still considered to be a question of morality by many in
American society (Fiorina et al. 2011).
A person’s sexual orientation is used to publically discriminate directly against
individuals in employment, public accommodations, housing, credit, marriage. For
example, gays and lesbians are prohibited from serving openly in the national armed
forces. Until 1995, gays and lesbians were prohibited from holding positions in the
national government that required security clearance. Sexual orientation is not always
included among the protected personal characteristics like race, color, national origin,
religion, sex, age, or disability often are when prohibiting discrimination. Some
institutions like the U.S. Equal Employment Opportunity Commission include sexual
orientation as a form of conduct that does not adversely affect employee performance.
This has a practical result of not allowing an individual to be fired because of their sexual
orientation, but the exclusion of sexual orientation as a class provides less protection than
if the class were explicitly mentioned. Attempts to include sexual orientation as a class
have been proposed in bills dating back to 1974, and recently as the Employment NonDiscrimination Act (ENDA), but the law has not been passed. At the state level,
Wisconsin was the first state to pass an anti-discrimination law that included sexual
orientation in 1982. Since 1982, other state and local governments have included sexual
orientation in their laws, but the lack of a national law has created conditions that allow
varying levels of discrimination against gays and lesbians working in the United States.
I have constructed a dataset of all gay rights decisions made by state supreme
courts from 1981 - 2004. The cases for 1981 – 2000 were initially identified by Pinello
in his text, Gay Rights and American Law (2003). Using identical search terminology,

56

the case list was augmented to include cases from 1981 to 2004. Cases where the sexual
orientation of the parties was not the justification for discrimination or was merely
tangential to the claim presented were not included in this analysis. Table 2 displays the
distribution of these cases among the states.
Table 2: The number of sexual orientation discrimination cases by state, 1981 - 2004.
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
Total = 94

3
2
1
2
2
1
2
1
1
5
1
0
0
0
3
0
1
3
1
3
9
0
0
5
1

Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
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

1
2
0
1
2
0
5
0
2
4
3
0
2
3
0
3
1
2
1
4
2
3
2
3
1

The cases were grouped into six broad categories based on the legal and
substantive policy issues raised in the case. This will allow me to control for the effect of
any jurisprudential regimes, the set of related legal precedents in an area of the law

57

(Kritzer and Richards 2003). For example, it is expected that every legal challenge to the
constitutionality of a criminal prohibitions of sodomy will be a discussion of the
individual’s right to privacy. Table 3 displays the percentage that each case type
composes of the sample for each model.
There were 94 cases in the dataset resulting in approximately 700 individual
votes. Long (1997) suggests that this is a more than adequate number of observations for
maximum likelihood models.

Table 3: Percentage of observations by case type in sexual orientation models.

Berry et al.
Brace et al.
Erikson et al.
Norrander
Average

Child
Marriage Benefits Custody Sodomy Discrimination Other
1.88
11.54
49.93
12.12
20.20
4.33
1.78
11.89
50.52
12.48
18.87
4.46
1.77
10.47
50.29
12.39
20.65
4.42
1.88
11.54
49.93
12.12
20.20
4.33
1.83
11.36
50.17
12.28
19.98
4.39

Marriage Cases
According to reports by the General Accounting Office, marital status is a
determinant in receiving benefits, rights, or privileges in 1,138 federal statutory
provisions (GAO 1997, 2004). Many of the same benefits conferred by state and local
governments and private entities are similarly determined by marital status. States
discriminate against gays and lesbians by only allowing opposite sex partnerships to
obtain marriage licenses. Without a license, gay or lesbian commitment ceremonies have
no legal credibility and confer no additional rights onto the couple as a legally recognized
family.

58

The current debate over gay marriage was sparked by a Hawaii Supreme Court
decision, which reversed a lower court's finding that discrimination of same-sex couples
is allowed under the state constitution's equal protection clause (Baehr v. Lewin 1993).
The subsequent trial court decision that held there was no compelling state interest in
allowing such discrimination (Baehr v. Miike 1996) was effectively overturned in 1998
by a state constitutional amendment. This amendment reserved the power to grant samesex marriages to the legislature before the Hawaii Supreme Court could review the
decision.
One of the effects of the Hawaii decision was that it amplified the question of
same-sex marriage as part of the gay rights debate and the rights movement in general.
After the Hawaii decision, a number of states have ratified amendments to their state
constitutions that explicitly banned the recognition of all same-sex marriages. Of these
additional states, only Alaska did so in response to a lower state court decision (Brause v.
Bureau of Vital Statistics 1998). Some states also responded by passing laws stating
explicitly that marriage was to be between opposite sex partners. Between 1993 and
2004, 31 states passed legislation prohibiting same-sex marriages within their
jurisdiction. During the same period, 33 states passed legislation denying recognition of
same-sex marriages performed from outside of the state. These figures may have been
higher had a number of states not already explicitly defined marriage as being between a
man and a woman or passed legislation prohibiting same-sex marriage before the Hawaii
decision.
The first states that allowed for some form of same-sex marriage have done so at
the command of their state supreme courts. A 1999 Vermont State Supreme Court

59

decision ordered the state legislature to provide the equal benefits of marriage to samesex couples (Baker v. State 1999). This decision was followed by the 2003
Massachusetts Supreme Judicial Court decision Goodridge, which held that the state's
constitution does not allow for the disparate treatment of same-sex couples seeking the
protections, benefits, and obligations emanating from marriage. Following a request for
clarification from the Massachusetts Senate, the Supreme Judicial Court held that civil
unions would not fulfill the constitutional requirements of the Goodridge decision. On
April 20, 2005, Connecticut legislatively created civil unions for its citizens in same-sex
relationships. The legislation also defines marriage as being between opposite sex
partners. The passage of this law preempted a court challenge filed in the Connecticut
judicial system (Kerrigan & Mock v. Connecticut Dept. Of Public Health 2008).
Whether the threat of a negative court decision played a role in the passage of this
legislation is unknown at this time. In December 2006, New Jersey passed a law creating
civil unions to comply with a state high court decision (Lewis et al. v. Harris 2006). On
May 15, 2008, California ruled that the state’s statutory scheme, which limits the
designation of the legal term “marriage” to opposite sex couples, violated the equal
protection principles in the state constitution (In Re Marriage Cases 2008). A subsequent
public proposition overturned the California decision and the state high court affirmed the
proposition’s effect (Strauss v. Horton 2009). The Iowa Supreme Court also legalized
same sex marriage because of a high court decision (Varnum v. Brien 2009).
Child Custody Cases
In addition to regulating the entrance into a marriage, states also have control over
the aftereffects of a terminated marriage. Child custody disputes that arise from the

60

dissolution of a state sanctioned marriage are typically resolved in the family courts of
the states. Gay and lesbian parents that were previously part of a heterosexual marriage
have sought to ensure that their sexual orientation is not a negative influence when
determining child custody. For example, if a state had a statute against sodomy, then gay
and lesbian parents were assumed to be potentially subjecting their children to criminal
activity and therefore it was in the best interest of the child to grant custody to the
heterosexual parent. Additionally, this rationale was used to justify restrictive visitation
orders that would exclude the child from visiting a gay or lesbian parent while another
individual of the same sex was also in the residence, regardless of whether or not they
were in a sexual relationship.
Furthermore, for those gays and lesbians who seek to adopt or serve as foster
parents, the state government determines what criteria are to be considered relevant. In
the state of Florida, gays and lesbians are excluded from becoming parents by adoption,
though they may still serve as foster parents. Sexual orientation has also been a factor
when interpreting state law to determine whether a non-biological parent, or the partner
of an individual who has previously adopted a child, is eligible to adopt his or her
partner's children through what are commonly termed “second parent adoptions”.
Adoption laws also have to be interpreted to determine whether gays and lesbians are
eligible for other types of adoptions, such as stepparent or joint adoptions.
Sodomy Cases
Among the powers of the state are police powers, which can make the private
relations between individuals a crime. The definition of sodomy often criminalized
consensual same-sex relations, and the presence of these laws and any subsequent

61

convictions from these laws could be used to discriminate in other areas of the law like
child custody or employment. Constitutional challenges to a state's sodomy prohibitions
have been made on both equal protection and privacy claims. For example, in 2002, the
Arkansas Supreme Court held that the state's sodomy statute was unconstitutional under
the state constitution because it violated the fundamental right to privacy and it
discriminated against gays and lesbians (Jegley v. Picado 2002). The U. S. Supreme
Court invalidated the remaining 13 state sodomy statutes under similar grounds in 2003
(Lawrence v. Texas 2003).
Benefits Cases
Many states have instituted programs that allowed gays and lesbians to register as
domestic partners in lieu of civil marriage. However, while providing some legal
protections and benefits, domestic partnerships do not approach the level of benefits
offered to married couples. An example of a case with this type of issue is Snetsinger v.
Montana University System, where the Montana Supreme Court held that the state must
provide the domestic partners of university employees with the same benefits options that
it provided married couples (Snetsinger v. Montana University System 2004). In addition
to mandating the equal treatment of gays and lesbians, state supreme courts also had to
resolve challenges from taxpayers who petition the court to stop the extension of benefits
to non-married couples (City of Atlanta v. Mckinney 1995).
Discrimination Cases
In regards to discrimination, the state supreme courts have had to determine what
responsibilities the state government has to protect its gay and lesbian citizens from
violations of their civil rights and liberties. The Supreme Court of New Jersey had to

62

consider whether a private organization could expel one of its members for being gay
(Dale v. Boys Scouts of America 1999). An example of discrimination in the workplace
can be found in Madsen v. Erwin and Others where a writer for The Christian Science
Monitor was fired because she was a lesbian and had refused to seek treatment for her
sexual orientation through the healing of the Christian Science Church (Madsen v. Erwin
and Others 1985). Other activities where gays and lesbians have encountered disparate
treatment in the law include the process of changing a person’s legal name (In Re
Bicknell et al. 2002), receiving student housing (Levin et al. v. Yeshiva University et al.
2001), participating in public parades (Irish-American Gay, Lesbian, and Bisexual Group
of Boston v. City of Boston 1994), recognition of a student group by a public university
(Gay Activists Alliance v. The Board of Regents of the University of Oklahoma 1981), and
admission to a legal bar association (Florida Board of Bar Examiners: In Re N.R.S.
1981).
Other Cases
There were a few cases whose subject matter does not fit neatly into one of the
categories above and they were placed in a fifth category for analysis. For example, a
student group sought an injunction to prevent U.S. military organizations from recruiting
on campus because of the military’s prohibition against gays and lesbians from serving
openly in the military (Gay and Lesbian Law Students Association v. Board of Trustees
1995). In Crooke v. Gilden, a court had to consider whether a property contract was
invalid because the parties who entered into it were lesbians (Crooke v. Gilden 1992). In
Collins v. Faith School District, a court was asked to determine whether a teacher could
be fired for answering questions by students about the sexual activities of gays and

63

lesbians as part of a sex education lesson (Collins v. Faith School District 1998). These
cases were included in the analysis because they centered on the role of sexual orientation
in the law.
Bivariate Model Results
I first ran a set of models where the only independent variable was the measure of
public preferences. This will allow for the comparison of the estimated coefficients of
the bivariate and fully specified models to ensure that the effect of the public opinion is
consistent across models and not the result of an incorrectly specified model. For each of
the four models, the measure of public preferences was statistically significant and in the
expected direction. The results of each model can be found in Tables 13- 17 located in
the appendix to the chapter. The models correctly predicted between 55.91 and 63.13
percent of the observations. To assess how well the models correctly discriminates
between possible outcomes, the area under the Receiving Operating Characteristic Curve
was computed. The area under the ROC Curve for the four models ranges from 0.576 to
0.630 and none of the models reaches the 0.7 level that indicates an “acceptable” amount
of discrimination (Hosmer and Lemeshow 2000 at 162). These low values of the area
under the ROC Curve indicate that the models are only slightly better at predicting a
justice’s vote than flipping a coin. Although statistically significant, a measure of public
opinion alone is not sufficient to explain the voting behavior of high court justices.
Multivariate Model Results
Tables 4 – 7 display the estimated coefficients for only the public preferences
variable from the multivariable models. The full model results can be found in the
appendix. Overall, each of the four models correctly predicted approximately two-thirds

64

of the cases. Another measure of fit, the area under the receiver-operator characteristic
(ROC) curve, indicated that each model produced an “acceptable amount” of
discrimination when comparing the observed and predicted votes of the justices (Hosmer
and Lemeshow 2000).

Table 4: Estimated coefficient for the Berry et al. measure of public preferences in cases
involving claims of discrimination based on sexual orientation.

Independent Variable
Citizen Ideology

Coefficient
Estimate

Standard
Error

P-Value

0.040

0.011

0.000

Number of Observations
693
Area Under ROC Curve
0.743
Percent Correctly Predicted
67.39
Note: Standard Errors are clustered on the individual justice

As seen in Table 4, the Berry et al. measure of citizen ideology is statistically
significant and in the expected positive direction. Justices were more likely to vote to
remove a restriction based upon sexual orientation as the citizenry becomes more liberal.
As this model is specified, the probability that a justice votes to remove a restriction
based on sexual orientation increases by 0.6430 as we move from the minimum observed
value to the maximum observed value of public liberalism, holding all other variables
constant. Predicted probabilities with 95% confidence intervals were generated using the
CLARIFY program for the full range of possible values and can be seen in Figure 1
located in the appendix to this chapter (King et al. 2000).

65

Table 5 Estimated coefficient for the Brace et al. measure of public preferences in cases
involving claims of discrimination based on sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Acceptance of Homosexuality

4.824

2.163

0.026

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

673
0.731
64.49

Independent Variable

Table 5 contains the coefficient from the model utilizing the Brace et al. measure
of public preferences. This measure of public opinion, which is specifically the level of
tolerance towards homosexuals, is statistically significant and in the expected positive
direction. Judges in states that are more tolerant of homosexuals are more likely to vote
to remove a law that differentiates based upon sexual orientation. The probability that a
justice votes to remove a restriction increases by 0.4277 as we move from the minimum
observed value to the maximum observed value of tolerance towards homosexuals.
Predicted probabilities with 95% confidence intervals were generated using the
CLARIFY program for the full range of values and can be seen in Figure 2 (King et al.
2000).

66

Table 6: Estimated coefficient for the Erikson et al. measure of public preferences in
cases involving claims of discrimination based on sexual orientation.

Independent Variable
Public Opinion
Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

Coefficient
Estimate

Standard
Error

P-Value

7.969

2.038

0.000

678
0.7433
67.85

Table 6, using the measure developed by Erikson et al., also indicates that the
justices serving in states that are more liberal are more likely to vote in a manner that
would remove a restriction based upon sexual orientation. The probability that a justice
votes to remove a restriction increases by 0.5509 as we move from the minimum
observed value to the maximum observed value of public opinion. Predicted
probabilities with 95% confidence intervals were generated using the CLARIFY program
for the full range of values and can be seen in Figure 3 (King et al. 2000).

Table 7: Estimated coefficient for the Norrander measure of public preferences in cases
involving claims of discrimination based on sexual orientation.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

1.259

0.626

0.044

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

693
0.733
65.51

Independent Variable

67

Finally, the estimates presented in Table 7 indicate that as a state’s citizens
become more liberal, using the Norrander measure of public preferences, the justices in
that state are more likely to remove a restriction based upon sexual orientation. The
probability that a justice votes to remove a restriction increased by 0.2882 as we move
from the most conservative observed value to the most liberal observed value of
Norrander’s citizen liberalism measure. Predicted probabilities with 95% confidence
intervals were generated using the CLARIFY program for the full range of values and
can be seen in Figure 4 (King et al. 2000).
The measure of public preferences was statistically significant and in the expected
direction in all four models. The results demonstrate that the decisions of state high court
justices were influenced by the public’s preferences. In states that are more liberal,
justices are more likely to change the law to match the preferences of the public. Given
that a diverse set of measures of public preferences had a consistent effect on the
direction of a justice’s vote in gay rights cases, we can conclude that for this issue public
opinion can help explain why the laws concerning gays and lesbians are different among
the states. The models demonstrate that the rights of gays and lesbians are being
expanded in the states that are either more liberal or more tolerant of homosexuals. The
justices who vote to expand rights have confidence that the public will not be incline to
hold their votes against them in future elections, if elected, and will support the changes
in the law.
Control Variables
Table 8 displays the remaining variables for each of the separate models. For
ease of interpretation, only those coefficients with a p-value less than 0.10 are presented

68

in the table. The coefficients whose p-value is between 0.10 and 0.05 are marked with an
asterisk. The full results for each separate model are available in the appendix.

Table 8: Coefficient estimates for control variables based on statistical significance.

Control Variable

Berry et al. Brace et al. Erikson et al. Norrander

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

--1.563
1.491
---

--1.751
1.654
---

--2.160
1.959
0.979

--1.448
1.389
---

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

--0.921*
0.055
-2.058
-1.579
-1.346
---1.377

----0.096
-2.541
-1.186*
-----1.046*

----0.058*
-1.790*
-1.303*
-1.082*
-1.286

----0.051*
---1.356
-----1.148*

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-------0.873
-------6.096

--0.519
---0.835
----0.012
-5.974

-------0.654
-------6.422

--0.529
---1.023
----0.008*
-6.637

Note: Estimates denoted by * 0.1 > p > 0.05

The first group of control variables indicates the type of retention method that was
used by the state where the justice was seated. The excluded group was the indicator for

69

a life-like tenure. Across the four models, justices who were subject to nonpartisan or
retention elections to remain on the bench were more likely to vote to remove a
distinction in the law based upon sexual orientation than justices who serve a life-like
term. For each of the models, the predicted probability of a positive vote was higher in
those situations where a justice was subject to nonpartisan or retention elections than the
remaining methods. This is contrary to previous research that has found that justices who
face retention by popular elections vote in a more conservative direction. However, some
of this previous research does not take into account public preferences as a variable.
Once the public’s preferences are included, the reason for elections alone to influence
judicial behavior is controlled for.
Table 9: Predicted Probabilities of a vote favoring the rights claims of gay & lesbians
using the Berry et al. measure of public preferences.

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment
Life-Like Term

Berry et al.
0.375
0.699
0.683
0.471
0.327

Brace et al.
0.435
0.712
0.691
0.428
0.300

Erikson et al. Norrander
0.467
0.419
0.504
0.719
0.768
0.707
0.730
0.552
0.276
0.375

The variable measuring the ideology of the justice was not statistically significant
in any of the four models. Given that during the period of this study there was little
difference between Republicans and Democrats on the issue of gay rights, this finding is
not surprising. Previous research on judicial decision-making was based on cases
involving the death penalty, a partisan issue. Additionally, the rights of gays and lesbians
may be considered a morality policy and previous research has shown that public

70

officials defer to public opinion over their own personal policy preferences (Mooney and
Lee 1995).
In only one of the four models was the presence of an intermediate appellate court
approach statistical significance. In the model using the Berry et al. measure, the
coefficient was indicating that the presence of an intermediate appellate court increased
the probability that a justice would vote to reverse a pro-gay law compared to those who
had none. Holding all other variables constant in the model using the Berry et al.
measure, the predicted probability of a positive vote increases from 0.1929 to 0.3750, for
a change in the probability of 0.1821, with the addition of an intermediate appellate court.
The presence of an intermediate appellate court changes the composition of the high
court’s docket and the cases that are frivolous are disposed of before they arrive to the
high court.
In all four models, there was a positive and statistically significant estimate for the
number of amicus curiae briefs filed in a case. This indicates that as the number of
groups outside of the litigation participating in the case increases, a justice is more likely
to vote to remove a distinction based upon sexual orientation. By generating predicted
probabilities for a pro-gay vote as the number of amicus briefs filed in a case increases
from the minimum to the maximum observed value, we can see that there is a substantial
effect. The range of increases in the probability of a pro-gay vote was of 0.3141 for the
Norrander model up to 0.4674 for the model incorporating the Brace et al. measure. The
increased attention that outside briefs bring to a case can alert a justice that there is
support for a change in the law. The presence of amicus curiae briefs can also serve as a

71

measure of interest group involvement in the state as those that are interested in the
outcome of the case will file a brief.
The estimate indicating whether the state had banned the recognition of same sex
marriages did not reach the level of statistical significance in any of the four models.
In two of the model specifications, those using the Norrander and Brace et al. measures,
justices were more likely to vote to overturn a distinction based upon sexual orientation if
the legislature has previously removed the criminal prohibition against consensual samesex sodomy. The magnitude of this effect was to increase the probability of a pro-gay
vote by 0.1290 in the Brace et al. model and 0.1313 in the Norrander model.
In states where direct democracy mechanisms were in place, justices were less
likely to vote to remove a distinction based upon sexual orientation in all four of the
models. Given the small number of gays & lesbians in a state and the low levels of
tolerance towards homosexuals, this result indicates that justices in states with direct
democracy are not as willing to vote in a counter-majoritarian direction and risk having
their decision overturned by the people. Additionally, direct democracy mechanisms
provide an additional opportunity for a court’s decision to be politicized by other political
actors in the state. The magnitude of the effect was to decrease the probability of a progay vote by 0.1699 in the model using the Erikson et al. measure to 0.2132 in the model
using the Norrander measure.
The estimate for the level of political competition in a state, the Ranney
Competition Index, was statistically significant and in the negative direction in all four of
the models. As specified in this model, justices are less likely to vote to overturn a legal
distinction based upon sexual orientation when they are in states that have higher levels

72

of political competition between the two major parties. The justices may be reluctant to
do so because it would provide the political branches a decision to attack the judicial
branch with and politicize the court’s activities. The predicted probability of a pro-gay
vote decreased by 0.4523 for the Berry et al. model and 0.4965 for the Erikson et al.
model as the value of political competition increased from the minimum observed value
to the maximum value.
There is also a pattern among the models concerning the subject matter of the
case. Table 10 displays the predicted probabilities for each case type. Across the four
models, justices are least likely to cast a positive vote in cases that involve a challenge to
marriage laws. Justices are also slightly less inclined to vote positively in cases that
involve either sexual relations or child custody. Using the terminology from the
psychological theory that the Attitudinal Model is based upon, the justices are influenced
by the situation that they encounter the object. There was an observed difference
between cases that involved the private relationships of the litigants and those that
involved public discrimination.
Table 10: Predicted probability of a pro-gay vote by case issue type.
Berry et al.

Brace et al.

Marriage

0.0573

0.0470

0.1910

0.1081

Sodomy

0.3224

0.3847

0.2959

0.3250

Discrimination

0.3750

0.4349

0.3729

0.4190

Benefits

0.4467

0.4544

0.4196

0.4434

Child Custody

0.3679

0.4182

0.3310

0.3724

Other

0.6976

0.6717

0.6197

0.6515

73

Erikson et al. Norrander

Government ideology was statistically significant in two of the four models. In
the models using the Brace et al. and Norrander measures, a justice was more likely to
vote to remove the distinction in the law as the state’s government became more liberal.
Generated predicted probabilities show an increase in the likelihood of a pro-gay vote of
0.2635 for the Brace et al. model and 0.1889 for the Norrander model as the value of
government ideology moves from the minimum observed value to the maximum.
Justices in states where the right to privacy is in the state’s constitution were not
more likely to vote to remove a distinction in the law based upon sexual orientation.
Although it was hypothesized that a right to privacy present in a state would provide
justices with a legal justification to overturn a discriminatory law, there are other legal
issues involved in these types of cases. In none of the models did the state’s murder rate
influence the voting behavior of the justices. While the state’s murder rate has an
influence on criminal justice appeals, there was no statistical connection with cases
involving a discrimination claim by gays and lesbians. The size of the legal community
in a state did not have an influence upon the justice’s voting behavior. The variation
among the states of legal resources available to litigants did not influence how cases were
disposed of.
Interactive Effect
The control variables indicating the method of retention that a state employs
suggests that in comparison to a life-like term and holding all other variables constant,
the justices who face nonpartisan or retention elections are more likely to vote in a progay manner. This finding suggests that the use of interaction terms between the public

74

opinion measure and the type of retention method employed may provide us with
additional insight into how the linkage operates.
Each of the models above indicates that justices are influenced by public
preferences when casting their vote in cases about sexual orientation. The next step
would be to narrow in on the mechanism that provides this linkage between the two.
Those who advocate for a democratically selected judiciary claim that the threat of
popular elections will force justices to consider the public’s opinion when acting.
To address this question, a second set of models for each of the measures of
public opinion were estimated with an interaction term between public opinion and the
method of retention the state used. If judicial elections have their intended effect and
result in a judiciary that is concerned with the public’s preferences, then the influence of
public opinion should vary with the retention method in place.
According to Berry and Rubin (2007), when considering whether an interaction
term should be included measures of goodness of fit can be examined to evaluate whether
there is an interactive effect. Compared to the non-interactive models, those with
interactions terms saw improved predictive abilities. However, the size of this
improvement is substantively small. Tables 11 and 12 contain the changes in fit – based
evaluations for the interactive models compared to the original models.

75

Table 11: A comparison of the percentage of observations correctly predicted for each
measure of public opinion excluding and including interaction terms with retention
methods.

Public Opinon Measure

No Interaction

Interaction

Improvement

Berry et al.

67.39

67.39

0.002

Brace et al.

64.49

65.68

1.189

Erikson et al.

67.85

69.03

1.180

Norrander

65.51

68.40

2.886

Table 12: A comparison of the area under the Receiver-Operator Characteristic Curve

Public Opinon Measure

No Interaction

Interaction

Improvement

Berry et al.

0.743

0.752

0.009

Brace et al.

0.731

0.731

0.000

Erikson et al.

0.743

0.751

0.007

Norrander

0.733

0.739

0.005

The goodness of fit measures in Tables 11 and 12 lead me to conclude that there
is no interactive effect between public opinion and the judicial selection methods used by
a state. There is almost zero improvement in the area under the ROC curve and there is
little improvement in the percentage of cases correctly predicted for some of the models.
Conclusion
The analysis presented in this chapter leads me to conclude that justices are
influenced by public preferences when voting in cases challenging a distinction in the law
based on sexual orientation. A justice becomes more likely to vote to remove a
distinction in the law based upon sexual orientation when the state is more liberal in its

76

policy preferences, holding all other variables constant as specified in the four models.
That the finding was consistent across all four models using different measures of public
preferences for bivariate and multivariate models suggests that the finding is robust to
any measurement errors or bias that may be present.
A comparison of estimates derived from models with and without interaction
terms between the measure of public opinion and the method of retention employed by
the states does not indicate that there is an interactive relationship. The absence of an
interactive effect suggests that justices on state high courts are not forced in line with the
public’s preferences because of electoral pressures. Justices in gay rights cases are
influenced by public preferences regardless of how they are retained on the bench.
The results of the analysis suggest that the decision-making process of state high
court justices is similar to the policy making process described in the study of state
policy. Justices in this analysis were influenced by a state’s public preferences,
competition between political parties, the presence of direct democracy measures, and the
participation of interest groups in addition to the legal characteristics of the cases. These
results suggest that potential litigants should consider the political environment when
deciding whether to appeal an adverse decision to a high court.

77

APPENDIX A

78

Table 13: Logit estimates for a bivariate model including the Berry et al. measure of
public preferences in cases involving discrimination based upon sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology (Berry et al.)

0.025

0.005

0.000

Constant

-0.925

0.256

0.000

Number of Observations

723

Area Under ROC Curve

0.63

Percent Correctly Predicted

62.66

NOTE: Standard errors are clustered on the individual justice.

Table 14: Logit estimates for a bivariate model including the Brace et al. measure of
public preferences in cases involving discrimination based upon sexual orientation.

Coefficient Standard
Estimate
Error
P-Value
Acceptance of Homosexuality (Brace et al.)

2.884

0.984

0.003

Constant

-0.292

0.218

0.181

Number of Observations

690

Area Under ROC Curve

0.576

Percent Correctly Predicted

56.81

NOTE: Standard errors are clustered on the individual justice.

79

Table 15: Logit estimates for a bivariate model including the Erikson et al. measure of
public preferences in cases involving discrimination based upon sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Public Opinion

3.356

0.788

0.000

Constant

0.729

0.153

0.000

Number of Observations

689

Area Under ROC Curve

0.624

Percent Correctly Predicted

63.13

NOTE: Standard errors are clustered on the individual justice.

Table 16: Logit estimates for a bivariate model including the Norrander measure of
public preferences in cases involving discrimination based upon sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

1.023

0.339

0.003

Constant

3.914

1.208

0.001

Number of Observations

728

Area Under ROC Curve

0.586

Percent Correctly Predicted

55.91

NOTE: Standard errors are clustered on the individual justice.

80

Table 17: Logit estimates including the Berry et al. measure of public preferences in
cases involving discrimination based upon sexual orientation.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

0.040

0.011

0.000

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.211
1.563
1.491
0.606

0.515
0.429
0.424
0.417

0.682
0.000
0.000
0.146

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.003
0.921
0.055
-2.058
-1.579
-1.346
-1.050
-1.377

0.005
0.545
0.028
1.045
0.688
0.635
0.691
0.618

0.505
0.091
0.047
0.049
0.022
0.034
0.129
0.026

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.151
0.345
0.176
-0.873
0.000
0.011
-0.003
-6.096

0.204
0.240
0.323
0.250
0.000
0.034
0.005
1.339

0.461
0.150
0.587
0.000
0.951
0.743
0.615
0.000

Constant

3.142

1.636

0.055

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

693
0.743
67.39

Independent Variable

NOTE: Standard errors are clustered on the individual justice.

81

Table 18: Logit estimates including the Berry et al. measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology
Citizen Ideology X Partisan Election
Citizen Ideology X Retention Election
Citizen Ideology X Elite Reappointment
Citizen Ideology X Nonpartisan Election

0.020
0.044
0.024
0.068
0.009

0.025
0.033
0.028
0.027
0.027

0.417
0.188
0.394
0.011
0.737

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

-2.094
0.553
-0.094
-3.725

1.944
1.766
1.806
1.777

0.281
0.754
0.959
0.036

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.002
0.810
0.069
-2.350
-1.532
-1.324
-0.902
-1.329

0.005
0.556
0.031
1.227
0.716
0.657
0.716
0.633

0.660
0.145
0.028
0.055
0.032
0.044
0.208
0.036

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.146
0.262
0.224
-0.828
0.000
0.006
-0.004
-5.823

0.221
0.272
0.332
0.273
0.000
0.035
0.005
1.382

0.508
0.336
0.501
0.002
0.895
0.870
0.434
0.000

Constant

4.526

2.220

0.041

Independent Variable

Number of Observations
693
Area Under ROC Curve
0.752
Percent Correctly Predicted
67.39
Note: Standard Errors are clustered on the individual justice
82

Table 19: Logit estimates including the Brace et al. measure of public preferences in
cases involving discrimination based upon sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Acceptance of Homosexuality

4.824

2.163

0.026

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.586
1.751
1.654
0.559

0.670
0.536
0.501
0.490

0.382
0.001
0.001
0.254

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.005
0.483
0.096
-2.541
-1.186
-0.978
-0.899
-1.046

0.005
0.612
0.033
1.125
0.667
0.617
0.668
0.596

0.281
0.430
0.004
0.024
0.076
0.113
0.178
0.079

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.105
0.519
-0.512
-0.835
0.000
-0.023
0.012
-5.974

0.213
0.256
0.418
0.276
0.000
0.032
0.004
1.449

0.621
0.042
0.221
0.002
0.757
0.479
0.008
0.000

Constant

3.252

1.708

0.057

Independent Variable

Number of Observations
673
Area Under ROC Curve
0.731
Percent Correctly Predicted
64.49
Note: Standard Errors are clustered on the individual justice

83

Table 20: Logit estimates including the Brace et al. measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences.
Coefficient Standard
Estimate
Error
P-Value

Independent Variable
Acceptance of Homosexuality
Acceptance of Homosexuality X Partisan Election
Acceptance of Homosexuality X Nonpartisan Election
Acceptance of Homosexuality X Retention Election
Acceptance of Homosexuality X Elite Reappointment

18.605
-9.881
-16.365
-12.469
-15.882

12.720
14.145
13.585
13.482
12.166

0.144
0.485
0.228
0.355
0.192

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

4.395
6.273
5.545
5.297

3.528
3.756
3.742
3.485

0.213
0.095
0.138
0.128

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.006
0.544
0.105
-2.623
-1.119
-0.907
-0.850
-0.929

0.005
0.619
0.033
1.172
0.669
0.619
0.677
0.593

0.284
0.380
0.001
0.025
0.095
0.143
0.209
0.117

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.089
0.642
-0.358
-0.950
0.000
-0.025
0.012
-5.924

0.223
0.325
0.503
0.326
0.000
0.037
0.004
1.477

0.690
0.048
0.477
0.004
0.885
0.501
0.008
0.000

Constant

-1.080

3.930

0.784

Number of Observations
673
Area Under ROC Curve
0.731
Percent Correctly Predicted
65.68
Note: Standard Errors are clustered on the individual justice

84

Table 21: Logit estimates including the Erikson et al. measure of public preferences in
cases involving discrimination based upon sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Public Opinion

4.263

1.273

0.001

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.182
1.659
1.538
0.672

0.523
0.448
0.442
0.429

0.728
0.000
0.000
0.117

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.004
0.237
0.074
-1.932
-1.355
-1.008
-0.813
-1.192

0.005
0.562
0.032
1.104
0.712
0.641
0.714
0.628

0.374
0.673
0.020
0.080
0.057
0.116
0.255
0.058

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.114
0.472
-0.298
-0.848
0.000
-0.035
0.005
-6.797

0.205
0.244
0.360
0.260
0.000
0.032
0.005
1.437

0.576
0.053
0.408
0.001
0.722
0.272
0.312
0.000

Constant

6.427

1.649

0.000

Independent Variable

Number of Observations
678
Area Under ROC Curve
0.741
Percent Correctly Predicted
65.49
Note: Standard Errors are clustered on the individual justice

85

Table 22: Logit estimates including the Erikson et al. measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences.

Coefficient
Estimate

Standard
Error

P-Value

2.983
18.527
-0.471
0.143
1.343

1.310
2.670
-0.160
0.050
0.440

0.189
0.008
0.876
0.963
0.658

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

3.954
1.393
1.475
0.200

2.710
2.480
2.300
0.180

0.007
0.013
0.022
0.854

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.003
0.058
0.087
-2.232
-1.252
-0.967
-0.768
-1.193

0.530
0.090
2.280
-1.740
-1.680
-1.480
-1.060
-1.890

0.595
0.927
0.023
0.081
0.092
0.140
0.287
0.059

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.126
0.411
-0.333
-0.746
0.000
-0.042
0.007
-5.469

-0.510
1.570
-0.890
-2.410
0.370
-1.210
1.490
-3.810

0.613
0.116
0.372
0.016
0.709
0.225
0.137
0.000

Constant

5.303

3.090

0.002

Independent Variable
Public Opinion
Public Opinion X Partisan Election
Public Opinion X Nonpartisan Election
Public Opinion X Retention Election
Public Opinion X Elite Reappointment

Number of Observations
659
Area Under ROC Curve
0.743
Percent Correctly Predicted
65.71
Note: Standard Errors are clustered on the individual justice
86

Table 23: Logit estimates including the Norrander measure of public preferences in
cases involving discrimination based upon sexual orientation.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

1.259

0.626

0.044

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.184
1.448
1.389
0.717

0.581
0.455
0.449
0.446

0.752
0.001
0.002
0.107

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.006
0.428
0.051
-1.379
-1.356
-0.952
-0.853
-1.148

0.005
0.576
0.029
1.030
0.678
0.627
0.670
0.607

0.200
0.457
0.074
0.180
0.046
0.129
0.203
0.059

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.136
0.529
-0.005
-1.023
0.000
-0.016
0.008
-6.637

0.213
0.245
0.325
0.254
0.000
0.032
0.004
1.362

0.522
0.031
0.988
0.000
0.835
0.620
0.054
0.000

Constant

9.741

2.667

0.000

Independent Variable

Number of Observations
693
Area Under ROC Curve
0.733
Percent Correctly Predicted
65.51
Note: Standard Errors are clustered on the individual justice

87

Table 24: Logit estimates including the Norrander measure of public preferences in
cases involving discrimination based upon sexual orientation with interactions between
retention method and public preferences.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology
Citizen Ideology X Partisan Election
Citizen Ideology X Nonpartisan Election
Citizen Ideology X Retention Election
Citizen Ideology X Elite Reappointment

4.299
-1.518
-2.765
-4.013
-2.067

2.144
3.077
2.330
2.184
2.254

0.045
0.622
0.235
0.066
0.359

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

-3.642
-7.183
-11.725
-5.577

11.254
7.705
7.153
7.443

0.746
0.351
0.101
0.454

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.005
0.281
0.067
-2.012
-1.604
-1.213
-1.068
-1.346

0.005
0.605
0.032
1.143
0.730
0.680
0.721
0.657

0.328
0.642
0.039
0.078
0.028
0.075
0.139
0.040

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.019
0.625
0.066
-1.220
0.000
-0.017
0.009
-7.115

0.225
0.283
0.352
0.288
0.000
0.036
0.004
1.376

0.931
0.027
0.852
0.000
0.963
0.640
0.052
0.000

Constant

20.183

7.247

0.005

Independent Variable

Number of Observations
693
Area Under ROC Curve
0.739
Percent Correctly Predicted
68.40
Note: Standard Errors are clustered on the individual justice

88

Figure 1

Mean Predicted Probability with 95% CI

0

.5

1

No Interaction Term

0

20

40
60
Citizen Ideology (Berry et al.)

80

100

Figure 2

Mean Predicted Probability with 95% CI

0

.5

1

No Interaction Term

0

.2

.4
.6
.8
Acceptance of Homosexuality (Brace et al.)

89

1

Figure 3

Mean Predicted Probability with 95% CI

0

.5

1

No Interaction Term

-1

-.5

0
Public Opinion (Erikson et al.)

.5

1

Figure 4

Mean Predicted Probability with 95% CI

0

.5

1

No Interaction Term

1

2

3
Citizen Ideology (Norrander)

90

4

5

Chapter 5 Gender Discrimination Cases
Both men and women seeking to address disparities in the law that result in
differential treatment can file gender discrimination claims. This treatment is often a
reflection of social conventions that held women as the center of the domestic, and not
public, sphere.

20

Among these conventions was the belief that women were in a constant

condition of guardianship by males. Economic survivorship was regulated under the
practice of coverture, where the liberty to contract and other property rights of females
were subject to the approval of a male head of household. Additionally, differences in
biology are frequently used as a justification for the differential treatment of men and
women in the law.

21

In some situations, males assert that the demarcation between the sexes restricts
the liberty of men. Biological differences have provided women with a legal advantage
in situations involving child custody and at times are used to grant rights to females
earlier than males. For example, in many jurisdictions females reached the age of
majority before males because it was thought that females tended to mature physically,
emotionally, and mentally before males. The analysis in this chapter examines claims

20

“Man is, or should be, women’s protector and defender. The natural and proper
timidity and delicacy which belongs to the female sex evidently unfits it for many of the
occupations of civil life. The constitution of the family organization, which is founded in
the divine ordinance, as well as the nature of things, indicates the domestic sphere as that
which properly belongs to the domain and functions of womanhood.” Justice Joseph
Bradley (Bradwell v. Illinois 1873 at 140-142).
21
The two sexes differ in the structure of body, in the functions to be performed by each,
in the amount of physical strength, in the capacity for long-continued labor, particularly
when done standing, the influence of vigorous health upon the future well-being of the
race, the self-reliance which enables one to assert full rights, and in the capacity to
maintain the struggle for subsistence.” Justice David J. Brewer (Muller v. Oregon 1908
at 422).
91

from both men and women. Before I discuss the results, I will describe the policy areas
and legal questions that incorporate claims of gender discrimination.
In a study by Baldez, Epstein, and Martin (2006) they employed a technique
similar to Pinello’s to create a list of cases addressing claims of gender discrimination. I
used their list of cases to create a dataset of votes. The issues in these cases include
abortion, family matters, employment discrimination, disparate treatment within the
criminal justice system, and benefit eligibility.
In the dataset, the number of votes analyzed ranges from 1,934 votes in the model
utilizing the Erikson et al. measure of public opinion to 2,320 votes in the models using
the Berry et al. and Norrander measures of public preferences. Table 25 displays the
percentage that each case type composes of the sample for each model. Table 26 displays
the distribution of cases by state for the period of 1970 - 2001.
Table 25: Percentage of observations by case type in gender discrimination cases.
Model
Berry et al.
Brace et al.
Erikson et al.
Norrander
Average

Abortion
5.99
6.19
6.41
5.99
6.15

Crime
33.45
32.27
32.37
33.45
32.88

Discrimination Family Benefits
13.92
39.22
7.41
14.34
39.51
7.69
14.27
39.56
7.39
13.92
39.22
7.41
14.11
39.38
7.48

92

Table 26: The number of gender discrimination cases by state, 1970 – 2001.
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
Total = 606

21
7
3
22
14
14
13
6
17
34
6
7
16
12
8
7
2
51
5
20
25
8
17
19
17

Montana
Nebraska
Nevada
New Hampshire
New Jersey
New Mexico
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

14
5
8
6
10
5
20
8
7
0
8
3
19
10
12
4
9
8
17
7
8
19
10
11
7

Criminal Cases
A large proportion of the votes in this analysis are on appeals from criminal court
trials. Gender differences existed in most aspects of the criminal justice process
including the definition of crimes, jury selection, and sentencing. Criminal cases are
often appealed because of the large incentive to have a conviction overturned and the
ability to combine multiple claims into a single appeal.

93

The text of criminal statutes defining sexual crimes were frequently written using
gendered language and as a result, these statutes were frequently challenged based on
denying the equal protection of the law. For example, in State vs. Meloon the Supreme
Court of New Hampshire was asked to determine whether a law that prohibited a male
from having nonconsensual sex with a female was unconstitutional because the same law
did not apply to females who forced a male to have nonconsensual sex (State v. Meloon
1976). Additionally in some states, prostitution statutes were written to specifically
prohibit women from engaging in an action and as a result, women could be charged with
a crime that a male could not (Wilson v. Indiana 1972).
The exclusion of women in public life included serving on trial court juries. The
degree of exclusion varied among the states. In Louisiana, Article 7, Section 41 of the
Louisiana Constitution provides: "[N]o woman shall be drawn for jury service unless she
shall have previously filed with the clerk of the District Court a written declaration of her
desire to be subject to such service." The observed result of this option was that the
presence of women on grand and trial juries was virtually nonexistent. In Hoyt v. Florida
the United States Supreme Court held:
"Despite the enlightened emancipation of women from the restrictions and protections of
bygone years, and their entry into many parts of community life formerly considered to
be reserved to men, woman is still regarded as the center of home and family life. We
cannot say that it is constitutionally impermissible for a State, acting in pursuit of the
general welfare, to conclude that a woman should be relieved from the civic duty of jury
service unless she herself determines that such service is consistent with her own special
responsibilities." (Hoyt v. Florida 1961 at 61 - 62)
During jury selection, the prosecution and the defense both have the ability to
remove a potential juror from the panel by exercising a peremptory challenge. The
attorney does not have to provide a reason for excluding the potential juror when a

94

challenge is used. In a previous ruling the U. S. Supreme Court ruled that the use of
peremptory challenges to prevent a group of individuals from serving on a jury subverted
the equal protection of the law (Batson v. Kentucky 1986). One example of such a case in
relation to gender discrimination is Illinois v. Blackwell. During jury selection for a
murder trial, the prosecution used 15 of its 17 peremptory challenges to remove women
from the venire panel. The defendant appealed his conviction on the grounds that the
gender bias exhibited by the prosecutor resulted in a jury that was less sympathetic
towards the effect that a conviction and a death sentence would have on his mother
(Illinois v. Blackwell 1996).
Another source of discrimination between men and women were the differences
in sentences for criminal convictions. In Commonwealth v. Daniel, Daisy Douglas and
Richard Johnson were both convicted of aggravated robbery and conspiracy
(Commonwealth v. Daniel 1968). Without any disparities in their criminal records or
circumstances, Mr. Johnson received a sentence of from four to ten years, while Ms.
Douglas was given an indeterminate sentence. The trial judge in this case had no
discretion in sentencing Ms. Douglas to a specific length of time because of a state law
that mandated convicted women to be sentenced to an indeterminate sentence but could
specify Mr. Johnson’s sentence.
Abortion Cases
Abortion cases often include claims of gender discrimination. Prior to the U.S.
Supreme Court decision in Roe v. Wade, states were able to determine whether abortion
was allowed through its criminal code (Roe v. Wade 1973). In this study, there are cases
that deal with the question of whether the state can prohibit or regulate the performance

95

of abortion (People v. Belous 1969). Prior to Roe, doctors who had performed abortions
were either charged with a crime or had their medical licenses revoked (Kudish v. Board
of Registration in Medicine 1969). In addition to the larger question of the legality of
abortion, there were other cases that regulate when the procedure could be performed.
The most common case involved challenges to the requirement that a minor obtain
parental consent before receiving an abortion (Ex Parte: Anonymous 1988). Courts also
had to consider when during a pregnancy an abortion could be legally performed
(Simopoulos v. Commonwealth of Virginia 1981). After abortion had been
decriminalized nationally, courts then had to determine whether states were required to
pay for abortion as part of their public health programs (Committee to Defend
Reproductive Rights et al. v. Myers 1981).
Family Cases
Cases in this category focus on the issues surrounding divorce, child custody, and
adoption. The standards used in child custody proceedings today have moved away from
utilizing an explicitly gendered legal doctrine and instead focus on the outcome of a case
being “in the best interests of the child”. Prior to this change, it was typical for the law to
openly prefer that children be placed with their mothers with all other factors being equal
and fathers would frequently challenge this sort of presumption (J. B. v. A. B. 1978).
Biological fathers who were not married to the mother of the child often had no parental
rights. This lack of legal standing would allow the mother of a child to give a child up
for adoption without consulting the biological father (In the Matter of the Adoption of
Baby Boy L. 1982) and prohibit the unwed father from having legal standing to
participate in litigation on behalf of the child (A v. X, Y, & Z 1982).

96

Other family questions that involved gender discrimination claims include
whether it was permissible to require a parent to pay child support for a female child until
she obtained the age of 18 and a male child until he obtained the age of 21 (Stanton v.
Stanton 1974). Additional aspects of divorce, such as the obligations of alimony, are also
included in this category (Rand v. Rand 1977).
Benefits Cases
This category of cases involves situations where the law treated men and women
differently in respect to how wills were to be settled during the probate process and
whether individuals would be eligible for survivors benefits. Some states had laws that
required a woman to rewrite her will after she altered her marital status (Parker v. Hall
1978). Other aspects of probate law that involved gender included whether widows had
an automatic right to a portion of a deceased husband’s estate if there was no will
(Ransome v. Ransome 1981). In addition to the probate process, cases have also
considered the acceptability of using gender in the determination of benefits in a variety
of programs such as retirement, unemployment, or workman’s compensation programs.
Frequently when these types of programs are established, it is assumed that the female is
not the primary source of support for families and this could affect the eligibility for or
level of benefits received (Da Rosa v. Carol Cable Company 1979).
Discrimination Cases
Cases classified as focusing on discrimination claims are those cases that
challenge laws that explicitly treat males and females differently but not in respect to the
other categories of Crime, Abortion, Family, or Benefits. Included in these types of cases
are instances where licensing requirements prohibited women from joining a profession.

97

In The Maryland State Board of Barber Examiners v. Kuhn, a statute that prohibited a
female cosmetologist from cutting the hair of a male customer, while licensed barbers
could cut the hair of both sexes, was ruled unconstitutional (The Maryland State Board of
Barber Examiners v. Kuhn 1973). Another example of such a law includes ordinance
that prohibited females from working as bartenders unless they themselves or their
spouse owned the establishment. The New Jersey Supreme Court held that the law was
unreasonably discriminatory and found it unconstitutional (Paterson Tavern & Grill
Owners Association v. The Borough of Hawthorne 1970). In professions where licenses
were not required, the state could prohibit a woman from working by making the
underlying activity illegal. A Las Vegas, Nevada law prohibited wrestlers from grappling
with opponents of the opposite sex. The way the law was written has the effect of
prohibiting any wrestling entertainment that would involve a man and a woman
physically touching. The Nevada Supreme Court upheld this law because the restriction
applied to both genders equally (Oueilhe v. Lovell 1977).
Other questions in this category include the participation of females on male
sports teams (Haas v. South Bend Community School Corp., et al. 1972), the regulation
of who could purchase alcoholic beverages (State of Oklahoma v. County Beverage
License #Abl-78-145 of Mcmar General Stores 1982), and the legality of single sex
private clubs (Burning Tree Club, Inc. v. Bainum 1985). This category also contains
appeals where gender discrimination claims were alleged against an employer (Peper v.
Princeton 1978).

98

Bivariate Model Results
I first ran a set of models where the only independent variable was the measure of
public preferences. Of the four models, only the measure of public opinion by Berry et
al. was statistically significant and in the expected direction. The results for all four
models are in Tables 36– 39 located in the appendix to the chapter. The bivariate models
correctly predicted between 55.33 and 59.14 percent of the observations correctly. To
assess how well the models correctly discriminate between possible outcomes, the area
under the Receiving Operating Characteristic Curve was computed. The area under the
ROC Curve for the four models ranges from 0.482 to 0.562 and none of the models
reaches the 0.7 level that indicates an “acceptable” amount of discrimination (Hosmer
and Lemeshow 2000 at 162). These low values of the area under the ROC Curve indicate
that the models are only slightly better at predicting a justice’s vote than flipping a coin.
A measure of public opinion alone is not sufficient to explain the voting behavior of high
court justices in cases involving gender discrimination.
Multivariate Model Results
Results for each of the models are presented in Tables 27 – 30. Assessing the
overall performance of the models, the range of percentage of cases correctly predicted
was from 63.5% to 65.0%. For all four models, the area under the ROC curve was
approximately 0.7, which is the lower bound for “acceptable discrimination” comparing
the observed and predicted votes of the justices (Hosmer and Lemeshow 2000).
The coefficient for the Berry et al. measure of public preferences was at the
margins of statistical significance with a two-tailed p-value of 0.065. The coefficient was
in the positive direction, indicating that justices in states that were more liberal were

99

more likely to vote to remove a distinction based upon gender. However, since the pvalue was outside the conventional level of 0.05, I cannot reject the null hypothesis that
there is no effect of public preferences on judicial voting behavior in gender
discrimination cases as this model was specified.

Table 27: Estimated coefficient for the Berry et al. measure of public preferences in
cases involving claims of discrimination based on gender.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

0.009

0.005

0.065

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

2320
0.702
63.75

Independent Variable

In its separate model, the coefficient for the Brace et al. measure of public
attitudes towards was statistically significant with a p-value of 0.013 and in the negative
direction. This result indicates that justices in states that were more supportive of gender
quality were less likely to vote to overturn a distinction based upon gender. The
probability that a justice votes to remove a restriction decreases by 0.2462 as we move
from the minimum observed value to the maximum observed value of public feminism.
Predicted probabilities with 95% confidence intervals were generated using the
CLARIFY program for the full range of values and can be seen in Figure 5 (King et al.
2000).

100

Table 28: Estimated coefficient for the Brace et al. measure of public preferences in
cases involving claims of discrimination based on gender.
Coefficient
Estimate

Standard
Error

P-Value

Public Feminism

-1.333

0.539

0.013

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

2197
0.707
63.77

Independent Variable

In its separate model, the estimate for the Erikson et al. measure of public opinion
was not statistically significant with a p-value of 0.374. This result does not allow us to
reject the null hypothesis that there is no relationship between public preferences and
judicial behavior.

Table 29: Estimated coefficient for the Erikson et al. measure of public preferences in
cases involving claims of discrimination based on gender.
Coefficient
Estimate

Standard
Error

P-Value

Public Opinion

-0.523

0.588

0.374

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

1934
0.696
63.50

Independent Variable

In a fourth model, the coefficient estimate for the Norrander measure of public
preferences was statistically significant. However, the coefficient was negative,
indicating that justices from states that were more liberal were less likely to vote to

101

overturn a distinction based upon gender. The probability that a justice votes to remove a
restriction decreased by 0.2380 as we move from the most conservative observed value to
the most liberal observed value of Norrander’s citizen liberalism measure. This result has
the same direction as the Brace et al. measure. Predicted probabilities with 95%
confidence intervals were generated using the CLARIFY program for the full range of
values and can be seen in Figure 6 (King et al. 2000).

Table 30: Estimated coefficient for the Norrander measure of public preferences in cases
involving claims of discrimination based on gender.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

-1.346

0.378

0.000

Number of Observations
Area Under ROC Curve
Percent Correctly Predicted

2320
0.708
65.00

Independent Variable

Unlike the analysis of gay rights cases in Chapter 4, there is no consistent pattern
among the four measures of public preferences in regards to the effect of public opinion
on judicial behavior. The two models, Brace et al. and Norrander, with a statistically
significant measure have a coefficient that is in opposite direction than hypothesized.
These two models indicate that judges are negatively influenced by public opinion and
vote contrary to the preferences of the public. However, the Berry et al. and Erikson et
al. measures were not statistically significant. The lack of consistency across the four
models of gender discrimination cases suggests that there are practical differences among
the different measures of public preferences.

102

Control Variables
Table 31 contains the coefficients for the remaining variables across the four
separate models. Only those estimates which could be considered statistically significant
are listed. The coefficients whose p-value is between 0.10 and 0.05 are marked with an
asterisk. The complete results for the separate models can be found in the appendix to
this chapter.
The first set of control variables measure the method of retention that a state
utilizes. The results indicate that there are statistically significant differences between
justices who face different methods of retention. These indicators are a series of dummy
variables and to ease interpretation of the effect of the different methods I have generated
predicted probabilities for each measure of public preferences holding the other variables
constant. Examining the predicted probabilities across the four models shows that
justices who are retained by partisan elections and elite reappointment have a higher
probability of voting for gender equality in the law than judges that sit for a life-like term.

103

Table 31: Coefficient estimates for control variables based on statistical significance.
Berry et
al.

Brace et
al.

Erikson et
al.

Norrander

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.892
--0.492*
0.732

------0.900

0.871
----0.883

------0.488*

Justice Ideology
Intermediate Appellate
Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.007

0.008

0.011

0.009

--0.083
-1.365
-1.480
-0.444
---

0.362*
0.076
-1.351
-1.382
-----

--0.096
-1.808
-1.544
-0.829
-0.474

--0.090
-1.390
-1.470
-0.381*
---

---

0.367

0.281*

0.467

---

---

---

---

0.275
---

0.287
---

-----

0.367
---

---0.006

-----

-----

-----

---

---

---

---

Control Variable

Equal Rights Amendment
Right to Privacy in the
Constitution
Availability of Direct
Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition
Index

Note: Estimates denoted by * 0.1 > p > 0.05

The individual justice’s ideology scores were statistically significant and in the
expected direction across all four of the models. As a judge’s ideology score becomes
more liberal the judge is more likely to vote in favor of removing gender discrimination
in the law. The size of the effect ranges from an increase of 0.1567 in the Berry et al.

104

model to 0.2477 for the Erikson et al. model as the value of ideology increases from the
minimum to maximum observed value in the models. These results are consistent with
the Attitudinal Model of judicial decision-making discussed in Chapter 3. The equal
treatment of men and women is a policy issue that is identified with the liberalconservative split in ideology. As a result, these are cases where the justices would have
an opportunity to use their ideology to determine how to vote.

Table 32: Predicted probabilities of a vote favoring the removal of a gender based
distinction by retention method.

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment
Life-Like Term

Berry et al.
0.565
0.354
0.465
0.525
0.347

Brace et al.
0.505
0.356
0.480
0.621
0.400

Erikson et al. Norrander
0.510
0.487
0.333
0.326
0.386
0.450
0.513
0.543
0.304
0.421

The number of amicus curiae briefs filed in an appeal was statistically significant
and in the positive direction in all four of the models. This indicates that the more salient
a case is to outside participants, a judge is more likely to vote to remove the gender
distinction in the law. This is another indication that external forces, such as interest
groups, can influence a justice’s decision.
A series of dummy variables for the policy issue in the case demonstrated that
justices were less likely to vote to reverse gender discrimination in some issues than
others. In each of the models, the excluded baseline category were those cases that
involved benefits claims. Compared only to benefits cases in all four of the models
justices were less likely to vote to overturn cases involving abortion or crimes. In two of

105

the four models, justices were also less likely to vote to remove a gender distinction in
discrimination cases than benefits cases. Finally, in only one of the four models, a justice
was less likely to vote to remove a gender distinction in a family case compared to a case
involving benefits. In reference to the Attitudinal Model, these results demonstrate how a
justice’s voting behavior is influenced by the factual situation in which the justice
encounters the issue.
Table 33: Predicted probabilities of a vote favoring the removal of a gender based
distinction by case type.
Berry et al.

Brace et al.

Erikson et al. Norrander

Abortion

0.3305

0.2444

0.2695

0.2477

Crime

0.3055

0.2387

0.3245

0.2332

Discrimination

0.5536

0.4941

0.4955

0.4745

Family

0.6033

0.5124

0.5835

0.5140

Benefits

0.6590

0.5553

0.6924

0.5694

In two of the four models the variable capturing whether a state had ratified the
Equal Rights Amendment was statistically significant and in the positive direction. The
change in predicted probability was an increase of 0.0919 for the Brace et al. model and
0.1156 for the Norrander model. This indicates that justices were more likely to vote to
remove a gender distinction in the law after their state had voted to change the U.S.
Constitution. For these two models, the justices were either responding to a signal sent
by the public and political institutions or they were attempting to change the case law in
anticipation of a shift in constitutional law.
The availability of direct democracy was statistically significant in three of the
four models. A justice was more likely to vote to remove a distinction based upon gender

106

in states with direct democracy. This magnitude of the effect ranges from an increase in
the probability to change the law of 0.0666 for the model using the Berry et al. measure
to 0.0913 for the model using the Norrander measure. Justices may believe that if the
public disagrees with the decision then it would be possible for the public to reverse the
decision with a referendum. This threat of an additional institution to check the decisions
of the high courts limits the exercise of their powers to change the law.
Government ideology was statistically significant in only one of the four models.
In the model using the Berry et al. measure, a justice was less likely to vote to remove the
distinction in the law as the state’s government became more liberal. There was a
decrease in the probability of a vote to remove the gender distinction of 0.1413 as the
value of government ideology changed from the minimum observed value to the
maximum value as the Berry et al. model was specified. The contrary direction of this
finding is similar to the results concerning public preferences. For the other three
models, the ideology of the government was not statistically significant and it appears
that state high court justices make their decisions independent from the preferences of the
elected government.
The inclusion of a right to privacy in a state’s constitution was not statistically
significant in any of the four models. As with the gay rights models, there are other legal
issues that justices may use to decide questions of gender discrimination. Also not
statistically significant in any of the four models were the control variables measuring the
size of the legal community, murder rate, or the level of political competition in a state.
As with the gay rights models, those filing gender discrimination appeals were not
affected by the differences in the size of the legal community in the state and were able to

107

obtain sufficient legal counsel. The murder rate of the states did not influence the justices
as the public would not be likely to link the two issues together in evaluating the
performance of the court. The lack of statistical significance for political competition
indicates that the justices are deciding cases without concern that their decisions in
gender discrimination cases would be politicized. The presence of an intermediate
appellate court in the state where the case was decided did not have a statistically
significant influence on the judge’s vote. In one model, utilizing the Brace et al.
measure, it was marginally significant (p = 0.065) and in the positive direction. This
would indicate that a judge in a state with an intermediate appellate court is more likely
to agree with a claim of gender discrimination. A judge would be more likely to vote in
this manner because the intermediate appellate court would filter out all of the absurd or
easily decided cases and leave only the cases that have the most merit to the claim to the
highest court.
Interactive Effect
As with the analysis of gay rights cases in the previous chapter, the method of
retention that a state employed had a statistically significant influence on the votes of the
justices and there was evidence that public opinion was influential in two of the four
models. I repeated the process outlined in chapter 4 to determine whether the effect of
the public opinion is influenced by retention method. Tables 34 – 35 contain
comparisons of fit for the models without and with interaction terms between the measure
of public preference and the method a state used to retain its high court justices.

108

Table 34: A comparison of the percent of observations correctly predicted for each
measure of public opinion excluding and including interaction terms with retention
methods.
Public Opinon Measure

No Interaction

Interaction

Improvement

Berry et al.

63.750

65.260

1.510

Brace et al.

63.770

65.910

2.140

Erikson et al.

63.500

63.550

0.050

Norrander

65.000

65.860

0.860

Table 35: A comparison of the area under the Receiver-Operator Characteristic Curve for
each measure of public opinion excluding and including interaction terms with retention
methods.

Public Opinon Measure

No Interaction

Interaction

Improvement

Berry et al.

0.702

0.718

0.016

Brace et al.

0.707

0.715

0.008

Erikson et al.

0.696

0.704

0.007

Norrander

0.708

0.722

0.014

Similar to the models of discrimination based on sexual orientation in Chapter 4, there is
no large improvement in the models that include interaction terms for the models of
gender discrimination cases. The small amount of improvement may stem from the
inclusion of additional variables. I would again conclude that there is no evidence that
the influence of public opinion is a result of the method of retention a justice faces.
Conclusion
The analysis presented in this chapter is very different that in the previous
chapter. The findings lead me to question whether justices may be influenced by public

109

preferences when voting in cases challenging a distinction in the law based on gender.
There was not consistent evidence for the effect as in the previous chapter. Two of the
four models displayed a statistically significant effect on judicial decision-making, but
the effect was in the opposite direction as was expected. Instead of judges in liberal
states being more likely to vote to remove the distinction in the law based upon gender,
the justices were less likely to do so. The inverse direction of the relationship combined
with the inconsistent effect among the four measures of public preferences raises
questions about the nature of the relationship. A comparison of estimates derived from
models with and without interaction terms between the measure of public opinion and the
method of retention employed by the states does not indicate that there is an interactive
relationship.
The different result among the four measures of public opinion reaffirms the
decision that multiple measures of public opinion should be used. The Norrander and
Brace measures produced different findings compared to the Berry et al. and Erikson et
al. measures indicating that there are differences in what the measures are capturing.
This chapter also reaffirms the conclusion that the decision-making process of justices
varies by the policy area under consideration.

110

APPENDIX B

111

Table 36: Logit estimates for a bivariate model including the Berry et al. measure of
public preferences in cases involving discrimination based upon gender.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

0.013

0.004

0.001

Constant

-0.915

0.193

0.000

Number of Observations

2942

Area Under ROC Curve

0.5622

Percent Correctly Predicted

59.14

NOTE: Standard errors are clustered on the individual justice.

Table 37: Logit estimates for a bivariate model including the Brace et al. measure of
public preferences in cases involving discrimination based upon gender.

Coefficient Standard
Estimate
Error
P-Value
Public Feminism

-0.428

0.352

0.223

Constant

0.253

0.518

0.626

Number of Observations

2824

Area Under ROC Curve

0.482

Percent Correctly Predicted

58.64

NOTE: Standard errors are clustered on the individual justice.

112

Table 38: Logit estimates for a bivariate model including the Erikson et al. measure of
public preferences in cases involving discrimination based upon gender.

Coefficient
Estimate

Standard
Error

P-Value

Public Opinion

0.524

0.349

0.134

Constant

-0.153

0.067

0.023

Number of Observations

2158

Area Under ROC Curve

0.507

Percent Correctly Predicted

55.33

NOTE: Standard errors are clustered on the individual justice.

Table 39: Logit estimates for a bivariate model including the Norrander measure of
public preferences in cases involving discrimination based upon gender.

Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

-0.054

0.232

0.816

Constant

-0.548

0.812

0.500

Number of Observations

2986

Area Under ROC Curve

0.503

Percent Correctly Predicted

58.77

NOTE: Standard errors are clustered on the individual justice.

113

Table 40: Logit estimates for the model including the Berry et al. measure of public
preferences in cases involving discrimination based upon gender.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

0.009

0.005

0.065

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.892
0.031
0.492
0.732

0.295
0.275
0.281
0.289

0.002
0.909
0.080
0.011

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.007
0.209
0.083
-1.365
-1.480
-0.444
-0.240

0.003
0.180
0.036
0.275
0.197
0.219
0.167

0.008
0.247
0.023
0.000
0.000
0.043
0.151

Equal Rights Amendment
Right to Privacy in the
Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

0.214

0.135

0.112

-0.064
0.275
0.000
0.022
-0.006
-0.049

0.157
0.138
0.000
0.018
0.003
0.648

0.684
0.046
0.651
0.228
0.033
0.940

Constant

-0.968

0.620

0.119

Independent Variable

Number of Observations
2320
Area Under ROC Curve
0.702
Percent Correctly Predicted
63.75
Note: Standard Errors are clustered on the individual justice

114

Table 41: Logit estimates for the model including the Berry et al. measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology
Citizen Ideology X Partisan Election
Citizen Ideology X Nonpartisan
Election
Citizen Ideology X Retention Election
Citizen Ideology X Elite
Reappointment

0.020
-0.021

0.012
0.010

0.085
0.035

0.039
0.011

0.011
0.012

0.001
0.333

0.065

0.012

0.000

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

-1.507
-1.292
-3.250
-0.774

0.625
0.682
0.701
0.700

0.016
0.058
0.000
0.269

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.007
0.162
0.080
-1.405
-1.630
-0.473
-0.274

0.003
0.186
0.038
0.274
0.196
0.219
0.165

0.008
0.386
0.038
0.000
0.000
0.031
0.096

Equal Rights Amendment
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate

-0.012
0.039
0.496
0.000
0.004

0.142
0.162
0.140
0.000
0.019

0.933
0.810
0.000
0.742
0.823

Government Ideology
Ranney Competition Index
Constant

-0.007
-0.311
1.623

0.003
0.651
0.839

0.024
0.632
0.053

Independent Variable

Number of Observations
2320
Area Under ROC Curve
0.718
Percent Correctly Predicted
65.26
Note: Standard Errors are clustered on the individual justice

115

Table 42: Logit estimates for the model including the Brace et al. measure of public
preferences in cases involving discrimination based upon gender.
Coefficient
Estimate

Standard
Error

P-Value

Public Feminism

-1.333

0.539

0.013

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.425
-0.187
0.325
0.900

0.351
0.289
0.285
0.307

0.226
0.517
0.254
0.003

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.008
0.362
0.076
-1.351
-1.382
-0.246
-0.173

0.003
0.196
0.033
0.290
0.202
0.221
0.168

0.001
0.065
0.022
0.000
0.000
0.265
0.303

Equal Rights Amendment
Right to Privacy in the
Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

0.367

0.142

0.010

0.036
0.287
0.000
0.012
0.000
0.188

0.163
0.140
0.000
0.018
0.002
0.721

0.826
0.040
0.758
0.511
0.858
0.794

Constant

0.702

0.852

0.410

Independent Variable

Number of Observations
2197
Area Under ROC Curve
0.707
Correctly Predicted
63.77
Note: Standard Errors are clustered on the individual justice

116

Table 43: Logit estimates for the model including the Brace et al. measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences.
Coefficient
Estimate

Standard
Error

P-Value

Public Feminism
Public Feminism X Partisan Election
Public Feminism X Nonpartisan
Election
Public Feminism X Retention Election
Public Feminism X Elite
Reappointment

9.145
-11.589

5.710
5.668

0.109
0.041

-9.233
-12.379

5.874
5.629

0.116
0.028

-7.218

5.992

0.228

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

17.942
14.292
19.317
12.050

8.759
9.077
8.715
9.255

0.041
0.115
0.027
0.193

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.008
0.272
0.078
-1.343
-1.344
-0.233
-0.160

0.003
0.207
0.034
0.288
0.198
0.218
0.166

0.002
0.189
0.022
0.000
0.000
0.286
0.336

Equal Rights Amendment
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index
Constant

0.548
-0.159
0.251
0.000
0.035
0.000
0.364
-15.847

0.181
0.184
0.142
0.000
0.021
0.002
0.687
8.830

0.002
0.387
0.077
0.683
0.098
0.832
0.596
0.073

Independent Variable

Number of Observations
2197
Area Under ROC Curve
0.715
Percent Correctly Predicted
65.91
Note: Standard Errors are clustered on the individual justice

117

Table 44: Logit estimates for the model including the Erikson et al. measure of public
preferences in cases involving discrimination based upon gender.
Coefficient
Estimate

Standard
Error

P-Value

Public Opinion

-0.523

0.588

0.374

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.871
0.135
0.366
0.883

0.322
0.310
0.305
0.311

0.007
0.662
0.230
0.005

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.011
0.291
0.096
-1.808
-1.544
-0.829
-0.474

0.003
0.203
0.042
0.321
0.226
0.247
0.196

0.000
0.152
0.022
0.000
0.000
0.001
0.016

Equal Rights Amendment
Right to Privacy in the
Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

0.281

0.144

0.051

0.148
0.182
0.000
0.010
-0.004
-0.770

0.166
0.128
0.000
0.018
0.002
0.656

0.374
0.154
0.858
0.578
0.132
0.240

Constant

-0.147

0.629

0.815

Independent Variable

Number of Observations
1934
Area Under ROC Curve
0.696
Percent Correctly Predicted
63.50
Note: Standard Errors are clustered on the individual justice

118

Table 45: Logit estimates for the model including the Erikson et al. measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences.
Coefficient
Estimate

Standard
Error

P-Value

Public Opinion
Public Opinion X Partisan Election
Public Opinion X Nonpartisan Election
Public Opinion X Retention Election
Public Opinion X Elite Reappointment

2.948
-4.668
-4.699
0.410
-5.618

2.899
3.259
3.005
3.118
3.134

0.309
0.152
0.118
0.895
0.073

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.624
-0.071
0.851
0.735

0.466
0.324
0.325
0.331

0.180
0.826
0.009
0.026

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.010
0.314
0.084
-1.674
-1.532
-0.737
-0.434

0.003
0.210
0.042
0.322
0.221
0.247
0.191

0.000
0.134
0.044
0.000
0.000
0.003
0.023

Equal Rights Amendment
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index
Constant

0.127
0.210
0.290
0.000
0.004
-0.004
-1.135
0.178

0.146
0.171
0.136
0.000
0.018
0.002
0.650
0.628

0.383
0.221
0.033
0.636
0.812
0.071
0.081
0.777

Independent Variable

Number of Observations
1934
Area Under ROC Curve
0.704
Percent Correctly Predicted
63.55
Note: Standard Errors are clustered on the individual justice

119

Table 46: Logit estimates for the model including the Norrander measure of public
preferences in cases involving discrimination based upon gender.
Coefficient
Estimate

Standard
Error

P-Value

Citizen Ideology

-1.346

0.378

0.000

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.264
-0.409
0.118
0.488

0.324
0.289
0.296
0.294

0.414
0.157
0.689
0.097

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.009
0.201
0.090
-1.390
-1.470
-0.381
-0.224

0.003
0.183
0.041
0.283
0.194
0.214
0.166

0.000
0.272
0.028
0.000
0.000
0.075
0.178

Equal Rights Amendment
Right to Privacy in the
Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

0.467

0.147

0.002

-0.086
0.367
0.000
-0.004
0.002
0.739

0.151
0.132
0.000
0.017
0.002
0.677

0.570
0.006
0.504
0.833
0.438
0.275

Constant

-6.070

1.656

0.000

Independent Variable

Number of Observations
2320
Area Under ROC Curve
0.708
Percent Correctly Predicted
65.00
Note: Standard Errors are clustered on the individual
justice

120

Table 47: Logit estimates for the model including the Norrander measure of public
preferences in cases involving discrimination based upon gender with interactions
between retention method and public preferences.

Independent Variable

Coefficient
Estimate

Standard
Error

P-Value

-1.805
1.291

1.105
1.217

0.102
0.289

Citizen Ideology
Citizen Ideology X Partisan Election
Citizen Ideology X Nonpartisan
Election
Citizen Ideology X Retention Election
Citizen Ideology X Elite
Reappointment

-1.141
1.879

1.213
1.219

0.347
0.123

-1.198

1.276

0.348

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

5.089
-4.574
6.661
-3.670

4.173
4.093
4.070
4.334

0.223
0.264
0.102
0.397

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.009
0.195
0.085
-1.425
-1.643
-0.418
-0.276

0.003
0.192
0.044
0.288
0.192
0.212
0.160

0.000
0.308
0.052
0.000
0.000
0.049
0.085

Equal Rights Amendment
Right to Privacy in the Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index
Constant

0.306
-0.034
0.491
0.000
-0.028
0.002
0.728
-7.363

0.155
0.167
0.143
0.000
0.019
0.002
0.699
3.721

0.048
0.837
0.001
0.494
0.135
0.367
0.298
0.048

Number of Observations
2320
Area Under ROC Curve
0.722
Percent Correctly Predicted
65.86
Note: Standard Errors are clustered on the individual justice

121

Figure 5

Mean Predicted Probability with 95% CI

0

.5

1

No Interaction Term

0

.5

1
Public Feminism (Brace et al.)

122

1.5

2

Figure 6

Mean Predicted Probability with 95% CI

0

.5

1

No Interaction Term

1

2

3
Citizen Ideology (Norrander)

123

4

5

Chapter 6 Conclusion
Studies of non-judicial state policy-making processes have repeatedly
demonstrated that public opinion is a major determinant of a state’s policy and studies of
judicial-decision making that ignore this finding are potentially missing a key
explanation. Those who would leave out public opinion as an explanatory variable are
making the assumption that the justices of high courts are not concerned about how their
decisions are received by the public or other policy makers in state government.
One way to demonstrate that the addition of public opinion variables should be
included in judicial decision-making models is to compare models with and without the
measures. This comparison of models with and without a measure of public preferences
demonstrates that models with public opinion included are slightly better at predicting
judicial votes. Tables 50 and 51, located in the appendix to chapter 6 below, compares
the percent of cases correctly predicted for models that found a statistically significant
effect of a public opinion measure. In four of the six models, adding a measure of public
opinion resulted in an increase in the percentage of cases correctly predicted by the
model. The two models that did not see an improvement were those that used the issue
specific measure of public preferences developed by Brace et al. The inconsistency of
the results again demonstrates that the debate over which measurement of public opinion
scholars should use needs further examination. Which measure of public opinion should
be used in future work is not a question that is addressed in this dissertation. However,
given the differences among the measures as demonstrated by the gender discrimination
analysis, it is clear that these measures have some subtle differences in what component
of public opinion they are capturing. With new measures of public opinion continuing to

124

be developed, increased consideration will have to be paid by all scholars of state policy
as to how public opinion is operationalized.
Another approach to determining the role of public opinion considering the
different outcomes from the four measures is to treat them as measuring different
components of public preferences. If the four indicators of public opinion included in
this analysis are indeed capturing different aspects of the central concept, then it should
be possible to include all four in a single equation and perform post-estimation tests to
determine the influence of multiple variables on the decision of a justice. The results of
these models are available in Tables 52 and 53 in the appendix to this chapter. The four
public opinion variables are highly correlated with each other, with a range of values for
pairwise correlations from 0.73 to 0.82 in the dataset of gay rights cases and 0.37 to 0.75
in the gender discrimination dataset. A Wald test can be used to test constraints placed
on the variables of a model (Long 1997). A likelihood ratio test is not appropriate in this
situation because the model is estimated with robust standard errors to account for the
repeated observation of justices in the sample. The Wald test will determine whether all
four of the public opinion variables are simultaneously equal to zero. For the gay rights
model, the Wald test resulted in a chi-square value of 14.83 with a p-value of 0.0051. In
the gender discrimination model, the Wald test of all four variables being jointly equal to
zero resulted in a chi-square value of 27.76 with a p-value of 0.000. These Wald tests
demonstrate that the hypothesis that the effects of all four measures of public opinion are
simultaneously equal to zero can be rejected at the 0.01 level for both models. One
difficulty with the Wald test is that it does not determine the positive or negative
direction of the relationship. Both models contained a public opinion measure that was in

125

a different direction than the others. In the gay rights model, the Norrander measure was
in the negative direction, although it was not statistically significant. In the gender
discrimination model, the Brace et al. and Norrander measures of public opinion were in
the negative direction, with only the Norrander measure reaching the level of statistical
significance. Given the high amount of correlation among the four variables,
multicollinearity may be the cause of direction change and changes to the statistical
significance of variables. These tests further demonstrate that public preferences are a
necessary component when explaining state supreme court judicial decision-making, just
as it is when describing the state policy making process.
Policy Differences
The different findings between the gay rights and gender discrimination models
confirms that the subject of the policy matters in explanations of the decision-making
process of state high courts. This is consistent with previous research on the United
States Supreme Court. There are several reasons that can be offered to understand why
justices were uniformly influenced by public opinion in cases involving gay rights, but
were not similarly influenced in gender discrimination cases. The question of gay rights
has been classified as a morality policy (Mooney 2001b). Mooney and Lee have
described morality policies as a type of policy that creates political conditions that are
ideal for democratic responsiveness (Mooney and Lee 2000). The gender discrimination
dataset does contain cases that involve abortion, another morality policy, but these cases
comprise less than 10% of the sample and do not provide enough variation to allow for a
complete analysis separate from the other gender discrimination cases.

126

The outcome that only morality policies were affected by public opinion may be a
result that justices vote differently in cases that involve issues of greater salience to the
public. Hall’s study of the Louisiana Supreme Court found that justices altered their
behavior in death penalty appeals in order to avoid conflicting with the preferred position
of their constituency (Hall 1987). Gay rights and gender discrimination are two policy
areas that involve similar legal issues, but they have distinct political differences. As the
types of cases that the court considers become less political, the court may be less likely
to consider public opinion in its decision-making process. Since state high courts
consider a wide range of cases as part of its docket, public opinion may not be an
influence on every case. Brace and Hall’s examination of state high court dockets reveals
that from 1995 to 1997 all but 13 state high courts heard more cases involving civil torts
than criminal appeals (Brace and Hall 2005). In their examination of state policy towards
gays and lesbians using their own issue specific measures of public opinion, Lax and
Phillips (Lax and Phillips 2009a) conclude that states were more responsive when the
public held the issue to be salient. Agreement between public opinion and policy was
also more likely to occur when the issue was highly salient. The findings of this
dissertation are in line with those of Lax and Phillips.
Another explanation that could explain the difference in the findings between the
gay rights and gender discrimination models is the composition of that opinion within the
states. Mooney and Lee (2000) conclude that the responsiveness of policy makers to
public opinion varies depending on the degree of dispersion within the opinion. For
issues where the public is divided, policy makers are more responsive to the public. Gay
rights as a morality policy is more likely to be divided than issues of gender inequality. If

127

the law is currently in line with the public’s preferences, then justices would not be
influenced by public opinion to change the law because they are already congruent.
The difference in the direction of the effect of public opinion between the two
policy areas could be attributed to the group that was claiming discrimination. With the
gender discrimination cases there were a number of claims made by men. It would be
difficult for men to claim that the judiciary is the only institution that could provide relief.
Compared to gays and lesbians, women have substantially more political power. As an
example, although it was ultimately unsuccessful, the Equal Rights Amendment was
ratified by 35 of the 50 states. Conversely, the wave of same-sex marriage prohibitions
that were passed in 30 states during the 1990’s and 2000’s could be cited as evidence that
gays and lesbians lack political power.
The ideology measure of the individual justices was statistically significant only
for the gender discrimination cases and not in any of the models for gay rights claims.
This finding extends the findings of previous research on the United States Supreme
Court that the effect of individual ideology varies on the types of cases analyzed. In this
study, both models examined discrimination claims, but in one chapter gays and lesbians
were the object and in the other chapter was either men or women. Only a few of the
cases in the gay and lesbian models were filed by a heterosexual victim claiming
discrimination. The difference of the effect of ideology highlights another possible
explanation that in situations where there is no division based upon ideology, judges are
likely to be influenced by public opinion. The lack of an ideological presumption forces
a justice to look for other factors to make a decision. With the gender discrimination
cases there already existed an ideological division with respect to women’s rights.

128

However, with gay rights claims, there was no division as both liberals and conservatives
did not favor equal treatment.
The inconsistent results found in Chapter 5, with 2 models finding a statistically
significant relationship and 2 models not, suggest that there are potential differences
between the four measures of public opinion that were employed in this paper. These
differences may not have been observed in the analysis of gay rights cases due to the
large magnitude of the effect found in morality policies.
Implications
The results of the dissertation also raise a number of practical political questions.
What should litigants consider when deciding to litigate as a strategy to produce policy
change? Given that there was no evidence that elected judges are more responsive to the
public than non-elected judges, what does this research say about the controversy of
judicial selection and retention? Why would all judges, regardless of retention method,
be influenced by public opinion as in the gay rights cases? Should courts be treated
differently compared to other policy makers?
This study demonstrates an additional mechanism for public opinion to influence
policy. For those individuals and groups who seek to bring about political change,
concentrating resources on changing public opinion may benefit a group’s litigation
strategy as a side effect. Additionally, those who seek to litigate as a strategy to bring
about social change could engage in “forum shopping” to determine where might be the
best state to file a suit. Another factor that a group might want to consider is that the
ideological composition of the high court may not be as important as the public’s
preferences in a state if the policy is a morality issue.

129

The findings of this study also speak to the controversy over whether states
should elect their judges. Those who oppose judicial elections claim that the electoral
pressures associated with elections creates the potential for judges to make their decisions
based upon a desire to be re-elected. This study finds that the effect of public opinion is
not dependent on the method of retention a justice will face. The lack of an interactive
effect suggests that justices are responsive to public opinion not because of any electoral
pressures. This finding raises questions as to whether the negative costs that popular
elections may bring to an institution are necessary given that elections do not appear to
make justices more responsive to the public. Incorporating public opinion into the
judicial decision-making process is also likely because of the ineffectiveness of judicial
elections to produce changes to the composition of a court. The re-election rate of state
high court judges is comparable to that of Congress (Hall 2001). If elections are not
necessary to link the courts with the public and they are not effective at producing
substantial turnover on the court, then those that oppose reforms to judicial selection will
have to make a more normative argument.
I believe that all judges, regardless of their retention method, are possibly
responsive to public preferences due to the nature of their judicial power. Courts lack
explicit enforcement powers and as such require the other political branches to implement
their decisions. If public opinion is on the side of a court decision, then the implementers
of that decision will face similar pressures by the public to follow the decision. If a court
does not adjust the law to the preferences of the people, it risks appearing out of touch
with the public and enforcing the laws of a previous generation. Excluding measures of
public preferences in models of judicial decision-making could result in an incomplete

130

explanation of how courts participate in the policy process at the state level. Without
including a measure of public opinion, judicial scholars are implicitly assuming that the
court is not affected by the need to maintain its legitimacy. Since a court relies on its
legitimacy in order to be an effective policy making institution, it would be
counterintuitive for a court to purposefully diminish its standing with the public.
This dissertation demonstrates that the political environment matters in state
judicial decision-making. In comparing the models with gay rights and gender
discrimination claims, the justices on state supreme courts act differently when
confronted with morality policies. The justices were also influenced by the presence of
direct democracy mechanisms. When there was an ideological dimension to the policy
issue they were deciding, the justices voted according to their personal preferences.
When there was ideological unity, justices deferred to public opinion. Judges on state
high courts are contributing to the overall policy environment and are being influenced
by many of the same social and political factors that studies of state policy have
determined to matter. Because of the reality that justices behave like other political
actors when making decisions, the legal nature of their public duties should not exclude
policy studies from incorporating court decisions from studies of the state policy
environment.

131

APPENDIX C

132

Table 48: Logit estimates for cases involving discrimination based upon sexual
orientation without a measure of public preferences.

Independent Variable

Coefficient
Estimate

Standard
Error

P-Value

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

-0.387
1.002
1.019
0.484

0.523
0.442
0.441
0.442

0.459
0.023
0.021
0.274

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.008
0.488
0.053
-1.619
-1.509
-1.183
-1.016
-1.303

0.005
0.581
0.029
1.045
0.695
0.635
0.698
0.618

0.093
0.401
0.070
0.121
0.030
0.063
0.146
0.035

Same-Sex Marriage Ban
Legislative Repeal of Sodomy
Right to Privacy in the
Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

-0.293
0.492

0.205
0.241

0.153
0.041

0.225
-0.923
0.000
-0.041
0.010
-6.314

0.314
0.254
0.000
0.031
0.004
1.358

0.474
0.000
0.314
0.183
0.014
0.000

Constant

5.374

1.618

0.001

Number of Observations
693
Area Under ROC Curve
0.728
Percent Correctly Predicted
65.08
Note: Standard Errors are clustered on the individual justice

133

Table 49: Logit estimates for cases involving discrimination based upon gender without
a measure of public preferences.

Independent Variable

Coefficient
Estimate

Standard
Error

P-Value

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.766
-0.060
0.409
0.712

0.287
0.274
0.275
0.287

0.008
0.826
0.137
0.013

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.008
0.239
0.084
-1.357
-1.485
-0.410
-0.224

0.002
0.180
0.037
0.277
0.197
0.215
0.165

0.002
0.185
0.025
0.000
0.000
0.057
0.176

Equal Rights Amendment
Right to Privacy in the
Constitution
Availability of Direct Democracy
Legal Community
Murder Rate
Government Ideology
Ranney Competition Index

0.245

0.135

0.069

-0.091
0.285
0.000
0.011
-0.002
0.032

0.154
0.138
0.000
0.017
0.002
0.651

0.555
0.038
0.937
0.515
0.279
0.961

Constant

-0.718

0.598

0.230

Number of Observations
2320
Area Under ROC Curve
0.700
Percent Correctly Predicted
64.310
Note: Standard Errors are clustered on the individual justice

134

Table 50: Percent of cases predicted correctly for models involving discrimination based
upon sexual orientation.

Public Opinion
No Public Opinion
Berry et al.
Brace et al.
Erikson et al.
Norrander

Percent Correctly
Predicted
65.08
67.39
64.49
67.85
65.51

Improvement
2.31
-0.59
2.77
0.43

Table 51: Percent of cases predicted correctly for models involving gender
discrimination.

Public Opinion
No Public Opinion
Brace et al.
Norrander

Percent
Correctly
Predicted
64.31
63.77
65.00

Improvement
-0.54
0.69

135

Table 52: Logit results for gay rights cases including all four public opinion measures.
Coefficient Standard
Estimate
Error
P-Value

Independent Variable
Citizen Ideology (Berry et al.)
Public Opinion (Erikson et al.)
Citizen Ideology (Norrander)

0.035
0.459
1.867
-0.156

0.013
2.530
1.367
0.769

0.007
0.856
0.172
0.839

Partisan Election
Nonpartisan Election
Retention Election
Elite Reappointment

0.443
1.853
1.740
0.625

0.683
0.522
0.497
0.469

0.516
0.000
0.000
0.183

Justice Ideology
Intermediate Appellate Court
Number of Amici Briefs
Same-Sex Marriage
Sodomy Prohibitions
Discrimination
Benefits
Child Custody

0.001
0.675
0.088
-2.689
-1.456
-1.275
-0.916
-1.336

0.005
0.555
0.031
1.112
0.700
0.648
0.697
0.624

0.783
0.224
0.004
0.016
0.038
0.049
0.189
0.032

Same-Sex Marriage Ban
0.055
0.234
Legislative Repeal of Sodomy
0.327
0.255
Right to Privacy in the Constitution
-0.297
0.439
Availability of Direct Democracy
-0.788
0.282
Legal Community
0.000
0.000
Murder Rate
-0.005
0.035
Government Ideology
-0.002
0.006
Ranney Competition Index
-5.805
1.432
Constant
2.748
3.382
Number of Observations
659
Area Under ROC Curve
0.749
Percent Correctly Predicted
66.46
Note: Standard Errors are clustered on the individual justice

0.814
0.201
0.498
0.005
0.972
0.892
0.759
0.000
0.416

Acceptance of Homosexuality (Brace et al.)

136

Table 53: Logit results for gender discrimination cases including all four public opinion
measures.

Independent Variables

Coefficient Standard
PEstimates
Error
Value

Citizen Ideology (Berry et al.)
Public Feminism (Brace et al.)
Public Opinion (Erikson et al.)
Citizen Ideology (Norrander)

0.023
-0.131
0.231
-1.661

0.006
0.509
0.625
0.444

0.000
0.797
0.711
0.000

Justice Ideology
Partisan Elections
Nonpartisan Elections
Retention Elections
Elite Reappointment

0.011
1.085
0.783
0.992
1.122

0.003
0.350
0.314
0.315
0.305

0.000
0.002
0.013
0.002
0.000

Intermediate Appellate Court
Number of Amici Briefs
Abortion
Criminal
Discrimination
Family

0.308
0.088
-1.504
-1.209
-0.729
-0.550

0.191
0.040
0.299
0.194
0.214
0.168

0.107
0.028
0.000
0.000
0.001
0.001

Equal Rights Amendment
0.670
0.137
Right to Privacy in the
Constitution
0.122
0.166
Availability of Direct Democracy
0.022
0.122
Legal Community
0.000
0.000
Murder Rate
-0.016
0.017
Government Ideology
-0.009
0.003
Ranney Competition Index
-0.168
0.648
Constant
-7.676
2.228
Area Under ROC Curve
0.676
Percent Correctly Predicted
65.25
Number of Observations
2279
Note: Standard Errors are clustered on the individual
justice

0.000

137

0.462
0.858
0.544
0.349
0.002
0.796
0.001

Table 54: Correlation between public opinion measures in gay rights cases.

Berry et al. Brace et al.
Berry et al.
1.000
Brace et al.
0.725
1.000
Erikson et al.
0.785
0.754
Norrander
0.764
0.818
Number of Observations = 659

Erikson et al.

Norrander

1.000
0.732

1.000

Table 55: Correlation between public opinion measures in gender discrimination cases.

Berry et al. Brace et al.
Berry et al.
1.000
Brace et al.
0.423
1.000
Erikson et al.
0.642
0.371
Norrander
0.752
0.703
Number of Observations = 2279

138

Erikson et al.

Norrander

1.000
0.631

1.000

REFERENCES

139

REFERENCES
ABA. 2003 "Justice in Jeopardy: Report of the American Bar Association Commission
on the 21st Century Judiciary." Chicago, Illinois: American Bar Association.
Achen, Cristopher H. "Measuring Representation." American Journal of Political Science
22, no. 3 (1978): 475 - 510.
Ai, Chunrong, and Edward C. Norton. "Interaction Terms in Logit and Probit Models."
Economics Letters 80 (2003): 123-129.
Atkins, Burton M., and Henry R. Glick. "Environmental and Structural Variables as
Determinates of Issues in State Courts of Last Resort." American Journal of
Politcal Science 20, no. 1 (1976): 97 - 115.
Baehr v. Lewin, 852 P.2d 44 (1993).
Baehr v. Miike, 910 P. 2d 112 (1996).
Baker v. Carr, 369 U.S. 186 (1962).
Baker v. State, 744 A.2d 864 (1999).
Baldez, Lisa, Lee Epstein, and Andrew D. Martin. "Does the U.S. Constitution Need an
Era?" Journal of Legal Studies 35, no. 1 (2006): 243-283.
Barlcay, Scott W., and Thomas Birkland. "Law Policymaking, and the Policy Process:
Closing the Gaps." Policy Studies Journal 26, no. 2 (1998): 227-244.
Barnum, David G. "The Supreme Court and Public Opinion: Judicial Decision Making in
the Post-New Deal Period." The Journal of Politics 47, no. 2 (1985): 652 - 666.
Barrilleaux, Charles. "A Test of the Independent Influences of Electoral Competition and
Party Strength in a Model of State Policy-Making." American Journal of Politcal
Science 41, no. 4 (1997): 1462-1466.

140

———. "Ideological Cleavage, Political Competition, and Policy Making in the
American States." In Public Opinion in State Politics, edited by Jeffrey E. Cohen.
Stanford, California: Stanford University Press, 2006.
Barrilleaux, Charles, and Michael Berkman. "Do Governors Matter? Budgeting Rules
and the Politics of State Policymaking." Political Research Quarterly 56, no. 4
(2003): 409-417.
Bartels, Larry M. "Constituency Opinion and Congressional Policy Making: The Reagan
Defense Build Up." American Political Science Review 85, no. 2 (1991): 457 474.
Batson v. Kentucky, 476 U.S. 79 (1986).
Baum, Lawrence. "The Supreme Court and the Attitudinal Model Symposium." Law and
Courts 4, no. 1 (1994): 3 – 5.
———. Judges and Their Audiences: A Perspective on Judicial Behavior. Princeton, NJ:
Princeton University Press, 2006.
Berry, Frances Stokes, and William D. Berry. "State Lottery Adoptions as Policy
Innovation: An Event History Analysis." American Political Science Review 84,
no. 2 (1990): 395-416.
Berry, William , Evan Ringquist, Richard Fording, and Russell Hanson. "Measuring
Citizen and Government Ideology in the American States, 1960 - 1993."
American Journal of Political Science 42, no. 1 (1998): 327 - 348.
Berry, William D. 1999 "Testing for Interaction in Models with Binary Dependent
Variables."
Berry, William D., and Frances Stokes Berry. "The Politics of Tax Increases in the
States." American Journal of Politcal Science 38, no. 3 (1994): 855-859.
Berry, William D., Jacqueline H. R. DeMeritt, and Justin Esarey. 2008 "Estminating
Binary Dependent Variable Models under Conditions of Specification
Uncertainty."

141

Berry, William D., Evan J. Ringquist, Richard C. Fording, and Russell L. Hanson.
"Commentary On "Correction: Does State Political Ideology Change over
Time?"" The Political Research Quarterly 59, no. 3 (2006): 495.
———. "The Measurement and Stability of State Citizen Ideology." State Politics &
Policy Quarterly 7, no. 2 (2007a): 111-132.
———. "A Rejoinder: The Measurement and Stability of State Citizen Ideology." State
Politics & Policy Quarterly 7, no. 2 (2007b): 160-166.
Berry, William D., and Jacqueline H. Rubin. 2007 "How to Test for Interaction in Models
with Binary Dependent Variables."
Boehmke, Fredrick. Thw Indirect Effect of Direct Democracy: How Institutions Shape
Interest Group Systems. Columbus, OH: Ohio State University Press, 2005.
Bonneau, Chris W. "The Composition of State Supreme Courts 2000." Judicature 85, no.
1 (2001): 26-31.
Bowler, Shaun, and Todd Donovan. "The Initiative Process." In Politics in the American
States: A Comparative Analysis, edited by Virgina Gray and Russell L. Hanson.
Washington, D.C.: CQ Press, 2008.
Bowman, Ann O'M, and Richard C. Kearney. The Resurgence of the States. Englewood
Cliffs, N.J.: Prentice-Hall, 1986.
Bowman, Ann O'M. "American Federalism on the Horizon." Publius 32, no. 2 (2002): 322.
Brace, Paul, Kevin Arceneaux, Martin Johnson, and Stacy G. Ulbig. "Does State Political
Ideology Change over Time?" Political Research Quarterly 57, no. 4 (2004): 529540.
———. "Rejoiner to Berry, Ringquist, Fording, and Hanson "Comment"." The Political
Research Quarterly 59, no. 4 (2006): 655-655.
———. "Reply To "The Measurement and Stabilty of State Citizen Ideology"." State
Politics & Policy Quarterly 7, no. 2 (2007): 133-140.

142

Brace, Paul, and Brent D. Boyea. "Judicial Selection Methods and Capital Punishment in
the American States." In Running for Judge: The Rising Political, Financial, and
Legal Stakes of Judicial Elections, edited by Matthew J. Streb, 186 - 203. New
York: New York University Press, 2007.
———. "State Public Opinion, the Death Penalty, and the Practice of Electing Judges."
American Journal of Politcal Science 52, no. 2 (2008): 360 - 372.
Brace, Paul, and Melinda Gann Hall. "Integrated Models of Judicial Dissent." The
Journal of Politics 55, no. 4 (1993): 914-935.
———. "The Interplay of Preferences, Case Facts, Context, and Rules in the Politics of
Judicial Choice." The Journal of Politics 59, no. 4 (1997): 1206-1231.
Brace, Paul, Melinda Gann Hall, and Laura Langer. "Judicial Choice and the Politics of
Abortion: Institutions, Context, and the Autonomy of Courts." Albany Law
Review, no. 62 (1999): 1265 - 1302.
———. "Placing State Supreme Courts in State Politics." Paper presented at the Annual
Meeting of the Midwest Political Science Association, Chicago, IL, April 27-30
2000 2000a.
———. "Placing State Supreme Courts in State Politics." State Politics & Policy
Quarterly 1, no. 1 (2001): 81-108.
Brace, Paul, Laura Langer, and Melinda Gann Hall. "Measuring the Preferences of State
Supreme Court Judges." Journal of Politics 62, no. 2 (2000b): 387-413.
Brace, Paul R., and Melinda Gann Hall. "Is Judicial Federalism Essential to Democracy?
State Courts in the Federal System." In The Judicial Branch, edited by Kermit L.
Hall and Kevin T. McGuire, 174 - 199. New York: Oxford University Press,
2005.
Brace, Paul, Kellie Sims-Butler, Kevin Arceneaux, and Martin Johnson. "Public Opinion
in the American States: New Perspectives Using National Survey Data."
American Journal of Political Science 46, no. 1 (2002): 173 - 189.
Bradwell v. Illinois, 83 U.S. 130 (1873).

143

Brambor, Thomas, William Roberts Clark, and Matt Golder. "Understanding Interaction
Models: Improving Empirical Analyses." Political Analysis 14, no. 1 (2006): 6382.
Braumoeller, Bear F. "Hypothesis Testing and Multiplicative Interaction Terms."
International Organization 58, no. 4 (2004): 807-820.
Brause v. Bureau of Vital Statistics, 1998 WL 88743 (1998).
Brennan, William J. "State Constitutions and the Protection of Individual Rights."
Harvard Law Review 90 (1977): 489.
Brigham v. State of Vermont, 692 A.2d 384 (1997).
Brisbin, Richard. "Slaying the Dragon: Segal, Spaeth and the Function of Law in
Supreme Court Decision Making." American Journal of Politcal Science 40, no. 4
(1996): 1004 - 1017.
Burning Tree Club, Inc. v. Bainum, 501 A.2d 817 (1985).
Burstein, Leigh. "The Analysis of Multilevel Data in Educational Research and
Evaluation." Review of Research in Education 8 (1980): 158-233.
Burstein, Paul. "The Impact of Public Opinion on Public Policy: A Review and an
Agenda." Political Research Quarterly 56 (2003): 29 - 40.
Caldeira, Gregory A. "The Transmission of Legal Precedent: A Study of State Supreme
Courts." American Political Science Review 79, no. 1 (1985): 178-194.
———. "Courts and Public Opinion." In The American Courts : A Critical Assessment,
edited by John Boatner Gates and Charles A. Johnson, x, 534 p. Washington,
D.C.: CQ Press, 1991.
Caldeira, Gregory A., and John R. Wright. "Organized Interests and Agenda Setting in
the U.S. Supreme Court." The American Political Science Review 82, no. 4
(1988): 1109 - 1127.

144

Camobreco, John F., and Michelle A. Barnello. "Democratic Responsiveness and Policy
Shock: The Case of State Abortion Policy." State Politics & Policy Quarterly 8,
no. 1 (2008): 48-65.
Campaign for Fiscal Equity v. State of New York, 801 N.E.2d 326 (2003).
Canon, Bradley C., and Dean Jaros. "External Variables, Institutional Structure & Dissent
on State Supreme Courts." Polity 3, no. 2 (1970): 175-200.
Church, Thomas W., and Robert T. Nakamura. Cleaning up the Mess : Implementation
Strategies in Superfund. Washington, D.C.: Brookings Institution, 1993.
City of Atlanta v. Mckinney, 454 S.E.2d 517 (1995).
City of Mesquite v. Aladdin's Castle, Inc., 463 U.S. 1032 (1982).
Clayton, Cornell W., and Howard Gillman. Supreme Court Decision-Making : New
Institutionalist Approaches. Chicago: University of Chicago Press, 1999.
Cobb, Roger W., and Charles D. Elder. Participation and American Politics: The
Dynamics of Agenda Building. 2nd ed. Baltimore: The Johns Hopkins University
Press, 1983.
Collins v. Faith School District, 574 N.W.2d 889 (1998).
Committee to Defend Reproductive Rights et al. v. Myers, 625 P.2d 779 (1981).
Commonwealth v. Daniel, 243 A.2d 400 (1968).
Comparato, Scott A. Amici Curiae and Strategic Behavior in State Supreme Courts.
Westport, Connecticut: Praeger, 2003.
Conlan, Timothy J. From New Federalism to Devolution : Twenty-Five Years of
Intergovernmental Reform. Washington, D.C.: Brookings Institution Press, 1998.
Crooke v. Gilden, 414 S.E.2d 645 (1992).

145

Culver, John, and John Wold. "Judicial Reform in California." In Judicial Reform in the
States, edited by Anthony Champagne and Judith Haydel, 139 - 162. Lanham,
Maryland: University Press of America, 1993.
Da Rosa v. Carol Cable Company, 397 A.2d 506 (1979).
Dahl, Robert A. "Decisionmaking in a Democracy: The Supreme Court as a National
Policy-Maker." Journal of Public Law 6 (1957): 279-295.
Dale v. Boys Scouts of America, 734 A.2d 1196 (1999).
Dawson, Richard E., and James A. Robinson. "Inter-Party Competition, Economic
Variables, and Welfare Policies in the American States." The Journal of Politics
25, no. 2 (1963): 265-289.
Devins, Neal. "How State Supreme Courts Take Consequences into Account: Toward a
State-Centered Understanding of State Constitutionalism." Stanford Law Review
62 (2010): 1629-1693.
Dye, Thomas R. . Politics, Economics and the Public: Policy Outcomes in the States.
Chicago, IL: Rand McNally, 1968.
Elazar, Daniel Judah. American Federalism; a View from the States. New York,:
Crowell, 1966.
Elliott, Shelden D. "Improvements in Judicial Administration, 1906 - 1956." In The
Administration of Justice in Retrospect, edited by Arthur L. Harding, 42 - 57.
Dallas, Texas: Southern Methodist University Press, 1957.
Epp, Charles R. The Rights Revolution : Lawyers, Activists, and Supreme Courts in
Comparative Perspective. Chicago: University of Chicago Press, 1998.
Epstein, Lee, and Carol Mershon. "Measuring Political Preferences." American Journal
of Political Science 40, no. 1 (1996): 261 - 294.
Erikson, Robert S. "The Relationship between Public Opinion and State Policy: A New
Look Based on Some Forgotten Data." American Journal of Political Science 20,
no. 1 (1976): 25-36.

146

———. "Constituency Opinion and Congressional Behavior: A Reexamination of the
Miller-Stokes Representation Data." American Journal of Political Science 22,
no. 3 (1978): 511 - 523.
Erikson, Robert S., Gerald C. Wright, and John P. McIver. Statehouse Democracy. New
York: Cambridge UP, 1993.
———. "Measuring the Public's Ideological Preferences in the 50 States: Survey
Responses Versus Roll Call Data." State Politics & Policy Quarterly 7, no. 2
(2007): 141-151.
Ex Parte: Anonymous, 531 So. 2d 901 (1988).
Fabricant, Solomon. The Trend of Government Activity in the United States since 1900.
New York: National Bureau of Economic Research, 1952.
Fiorina, Morris P., Samuel J. Abrams, and Jeremy C. Pope. Culture War? Boston, MA:
Longman, 2011.
Fisher, Glenn W. "Determinants of State and Local Government Expenditures: A
Preliminary Analysis." National Tax Journal 14, no. 4 (1961): 349-355.
Flemming, Roy B., and B. Dan Wood. "The Public and the Supreme Court: Individual
Justice Responsiveness to American Policy Moods." American Journal of
Political Science 41, no. 2 (1997): 468-498.
Florida Board of Bar Examiners: In Re N.R.S., 403 So. 2d 1315 (1981).
Fording, Richard C. "The Conditional Effect of Violence as a Political Tactic: Mass
Insurgency, Welfare Generosity, and Electoral Context in the American States."
American Journal of Politcal Science 41, no. 1 (1997): 1-29.
Frank, Jerome. Law and the Modern Mind. New York: Brentano's, 1930.
———. Courts on Trial. Princeton, N.J.: Princeton University Press, 1949.
Fry, Brian R., and Richard F. Winters. "The Politics of Redistribution." American
Political Science Review 64, no. 2 (1970): 508-522.
147

GAO. 1997 "Defense of Marriage Act." US General Accounting Office: Office of the
General Counsel.
———. 2004 "Defense of Marriage Act: Update to Prior Report." US General
Accounting Office.
Gay Activists Alliance v. The Board of Regents of the University of Oklahoma, 638 P.2d
1116 (1981).
Gay and Lesbian Law Students Association v. Board of Trustees, 673 A.2d 484 (1995).
Gelman, Andrew, and Jennifer Hill. Data Analysis Using Regression and
Multilevel/Hierarchical Models. New York: Cambridge University Press, 2007.
Gerber, Elisabeth R. "Legislative Response to the Threat of the Popular Initiative."
American Journal of Political Science. 40, no. 1 (1996): 99-128.
Gibson, James L. "Environmental Constraints on the Behavior of Judges: A
Representational Model of Judicial Decision Making." Law & Society Review 14,
no. 2 (1980): 343-370.
———. "From Simplicity to Complexity: The Development of Theory in the Study of
Judicial Behavior." Political Behavior 5, no. 1 (1983): 7 - 49.
Goldberg, Deborah, Sarah Samis, Edwin Bender, and Rachel Weiss. 2005 "The New
Politics of Judicial Elections 2004." edited by Jesse Rutledge. Washington, D.C.:
Justice at Stake Campaign.
Goldberg, Deborah, and Samantha Sanchez. 2004 "The New Politics of Judicial Elections
2002." edited by Bert Brandenberg. Washington, D.C.: Justice at Stake
Campaign.
Goodridge vs. Department of Public Health, 798 N.E.2d 941 (2003).
Grattet, Ryken, Valerie Jenness, and Theodore R. Curry. "The Homogenization and
Differentiation of Hate Crime Law in the United States, 1978 to 1995: Innovation
and Diffusion in the Criminalization of Bigotry." American Sociological Review
63, no. 2 (1998): 286-307.

148

Gray, Virgina. "The Socioeconomic and Political Contexts of States." In Politics in the
American States: A Comparative Analysis, edited by Virgina Gray. Washington,
D.C.: CQ Press, 2008.
Gray, Virginia. "Expenditures and Innovation as Dimensions Of "Progressivism": A Note
on the American States." American Journal of Political Science 18, no. 4 (1974):
693-699.
Gray, Virginia, David Lowery, Matthew Fellowes, and Andrea McAtee. "Public Opinion,
Public Policy, and Organized Interests in the American States." Political
Research Quarterly 57, no. 3 (2004): 411-420.
Groseclose, Tim, Steven D. Levitt, and Jr. James M. Snyder. "Comparing Interest Group
Scores across Time and Chambers: Adjusted Ada Scires for the U.S. Congress."
The American Political Science Review 93, no. 1 (1999): 33-50.
Haas v. South Bend Community School Corp., et al., 289 N.E.2d 495 (1972).
Haider-Markel, Donald P., and Matthew S. Kaufman. "Public Opinion and Policy
Making in the Culture Wars: Is There a Connection between Opinion and State
Policy on Gay and Lesbian Issues?" In Public Opinion in State Politics, edited by
Jeffrey E. Cohen. Stanford, California: Stanford University Press, 2006.
Haider-Markel, Donald P., Alana Querze, and Kara Lindaman. "Lose, Win, or Draw?: A
Reexamination of Direct Democracy and Minority Rights." Political Research
Quarterly 60, no. 2 (2007): 304-314.
Hall, Melinda Gann. "Constituent Influence in State Supreme Courts: Conceptual Notes
and a Case Study." The Journal of Politics 49, no. 4 (1987): 1117-1124.
———. "Electoral Politics and Strategic Voting in State Supreme Courts." The Journal
of Politics 54, no. 2 (1992): 427-446.
———. "Justices as Representatives: Elections and Judicial Politics in the American
States." American Politics Quarterly 23, no. 4 (1995): 485-503.
———. "State Supreme Courts in American Democracy: Probing the Myths of Judicial
Reform." American Political Science Review 95, no. 2 (2001): 315-330.

149

———. "State Courts: Politics and the Judicial Process." In Politics in the American
States: A Comparative Analysis, edited by Virgina Gray and Russell L. Hanson,
229-255. Washington, D.C.: CQ Press, 2008.
Hall, Melinda Gann, and Paul Brace. "Order in the Courts: A Neo-Institutional Approach
to Judicial Consensus." The Western Political Quarterly 42, no. 3 (1989): 391407.
———. "Toward an Integrated Model of Judicial Voting Behavior." American Politics
Quarterly 20, no. 2 (1992): 147 - 168.
Hays, Scott P., Michael Esler, and Carol E. Hays. "Environmental Commitment among
the States: Integrating Alternative Approaches to State Environmental Policy."
Publius 26, no. 2 (1996): 41-58.
Heck, Ronald H., and Scott L. Thomas. An Introduction to Multilevel Modeling
Techniques. Edited by George A. Marcoulides, Quantitative Methodology Series.
Mahwah, NJ: Lawrence Erlbaum Associates, 2000.
Hero, Rodney E. Faces of Inequality: Social Diversity in American Politics. New York:
Oxford University Press, Inc., 1998.
Hershkoff, Helen. "State Constitutions: A National Perspective." Widener Journal of
Public Law 3, no. 1 (1993): 7.
———. "State Courts and The "Passive Virtues": Rethinking the Judicial Function."
Harvard Law Review 114 (2001): 1833.
Hill, Kim Quaile, Jan Leighley, and Angela Hinton Anderson. "Pathways of
Representation: A Causal Analysis of Public Opinion-Public Policy Linkages."
American Journal of Political Science 39, no. 4 (1995): 924 - 935.
Hill, Kim Quaile, and Jan E. Leighley. "The Policy Consequences of Class Bias in State
Electorates." American Journal of Politcal Science 36, no. 2 (1992): 351-365.
Hofferbert, Richard I. "The Relation between Public Policy and Some Structural and
Environmental Variables in the American States." American Political Science
Review 60, no. 1 (1996): 73-82.

150

Hosmer, David W., and Stanley Lemeshow. Applied Logistic Regression. New York:
John Wiley & Sons, Inc., 2000.
Hoyt v. Florida, 368 U.S. 57 (1961).
Illinois v. Blackwell, 665 N.E.2d 782 (1996).
In Re Bicknell et al., 771 N.E.2d 846 (2002).
In Re Marriage Cases, 08 C.D.O.S. 5820 (2008).
In the Matter of the Adoption of Baby Boy L., 643 P.2d 168 (1982).
Irish-American Gay, Lesbian, and Bisexual Group of Boston v. City of Boston, 636
N.E.2d 1293 (1994).
J. B. v. A. B., 242 S.E.2d 248 (1978).
Jacob, Herbert, and Michael Lipsky. "Outputs, Structure, and Power: An Assessment of
Changes in the Study of State and Local Politics." Journal of Politics 30, no. 2
(1968): 510-538.
Jacoby, William G., and Saundra K. Schneider. "Variability in State Policy Priorities: An
Empirical Analysis." The Journal of Politics 63 (2001): 544-568.
Jegley v. Picado 80 S.W. 3d 332 (2002).
Kam, Cindy D., and Robert J. Franzese. 2006 "Modeling and Interpreting Interactive
Hypotheses in Regression Analysis."
Kerrigan & Mock v. Connecticut Dept. Of Public Health, 957 A.2d 407 (2008).
Key, V. O. Southern Politics in State and Nation. 1st ed. New York,: A. A. Knopf, 1949.

151

King, Gary. "Proper Nouns and Methodological Propriety: Pooling Dyads in
International Relations Data." International Organization 55, no. 2 (2001): 497507.
King, Gary, Michael Tomz, and Jason Whittenberg. "Making the Most of Statistical
Analyses: Improving Interpretation and Presentation." American Journal of
Political Science 44, no. 2 (2000): 347 - 361.
Kritzer, Herbert M., Paul Brace, Melinda Gann Hall, and Brent T. Boyea. "The Business
of State Supreme Courts, Revisited." Journal of Empirical Legal Studies 4, no. 2
(2007): 427-439.
Kritzer, Herbert M., and Mark J. Richards. "Jurisprudential Regimes and Supreme Court
Decisionmaking: The Lemon Regime and Establishment Clause Cases." Law &
Society Review 37, no. 4 (2003): 827-840.
Kudish v. Board of Registration in Medicine, 248 N.E.2d 264 (1969).
Kuklinski, James H., and John E. Stanga. "Political Participation and Government
Responsiveness: The Behavior of California Superior Courts." The American
Political Science Review 73, no. 4 (1979): 1090-1099.
Langer, Laura. Judicial Review in State Supreme Courts : A Comparative Study, Suny
Series in American Constitutionalism. Albany: State University of New York
Press, 2002.
———. n.d. "Multiple Actors and Competing Risks in the Policymaking (Unmaking
Game) of Judicial Review."
———. National Science Foundation CAREER Grant, SES-0092187 "Multiple Actors
and Competing Risks in the Policymaking (Unmaking Game) of Judicial
Review."
Lascher, Edward, Michael Hagan, and Steven Rochlin. "Gun Behind the Door? Ballot
Initiatives, State Policies, and Public Opinion." The Journal of Politics 58 (1996):
806 - 818.
Lawrence v. Texas, 539 U.S. 558 (2003).

152

Lax, Jeffrey R., and Justin H. Phillips. "Gay Rights in the States: Public Opinion and
Policy Responsiveness." American Political Science Review 103, no. 3 (2009a):
367-386.
———. "How Should We Estimate Public Opinion in the States?" American Journal of
Political Science. 53, no. 1 (2009b): 107-121.
Levin et al. v. Yeshiva University et al., 754 N.E.2d 1099 (2001).
Lewis et al. v. Harris, 908 A.2d 196 (2006).
Lewis, Gregory B., and Seong Soo Oh. "Public Opinion and State Action on Same-Sex
Marriage." State and Local Government Review 40, no. 1 (2008): 42-53.
Linde, Hans A. "The State and the Federal Courts in Governance: Vive La Difference!"
William and Mary Law Review 46, no. 4 (2005): 1273-1288.
Link, Michael W. "Tracking Public Mood in the Supreme Court: Cross-Time Analysis of
Criminal Procedure and Civil Rights Cases." Political Research Quarterly 48, no.
1 (1995): 61 - 78.
Long, J. Scott. Regression Models for Categorical and Limited Dependent Variables.
Thousand Oaks, CA: Sage Publications, 1997.
Madsen v. Erwin and Others, 481 N.E.2d 1160 (1985).
Marshall, Thomas R. Public Opinion and the Supreme Court. Boston: Allen & Unwin,
1988.
The Maryland State Board of Barber Examiners v. Kuhn, 312 A.2d 216 (1973).
Matsusaka, John G. "Fiscal Effects of the Voter Initiative: Evidence from the Last 30
Years." The Journal of Political Economy 103, no. 3 (1995): 587-623.
McCann, Michael W. Rights at Work : Pay Equity Reform and the Politics of Legal
Mobilization, Language and Legal Discourse. Chicago, IL: University of Chicago
Press, 1994.

153

McCloskey, Robert G. The American Supreme Court. Chicago, IL: The University of
Chicago Press, 1960.
McGuire, Kevin T., and James A. Stimson. "The Least Dangerous Branch Revisited:
New Evidence on Supreme Court Responsiveness to Public Preferences." The
Journal of Politics 66, no. 6 (2004): 1018 - 1035.
McIver, John P., Robert S. Erikson, and Gerald Wright. 2001 "Public Opinion and Public
Policy in Temporal Perspective: A View from the States." In Annual Meeting of
the Midwest Political Science Association. Chicago, IL.
Meeker, James W. State Supreme Court Litigants and Their Disputes. New York:
Garland Publishing, 1986.
Miller, David Young. "The Impact of Political Culture on Patterns of State and Local
Government Expenditures." Publius 21, no. 2 (1991): 83-100.
Miller, Warren E., D. R. Kinder, and S. J. Rosenstone. 1993 "American National
Election Study: Pooled Senate Election Study, 1988, 1990, 1992." Inter-university
Consortium for Political and Social Research.
Miller, Warren E., and Donald E. Stokes. "Constituency Influence in Congress."
American
Political Science Review 57, no. 1 (1963): 45 - 56.
Mintrom, Michael. "Policy Entrepreneurs and the Diffusion of Innovation." American
Journal of Politcal Science 41, no. 3 (1997): 738-770.
Mishler, William, and Reginald S. Sheehan. "The Supreme Court as a
Countermajoritarian Institution? The Impact of Public Opinion on Supreme Court
Decisions." American Political Science Review 87, no. 1 (1993): 87-101.
Mooney, Christopher Z. "Modeling Regional Effects on State Policy Diffusion." Political
Research Quarterly 54, no. 1 (2001a): 103-124.
———. The Public Clash of Private Values : The Politics of Morality Policy. New York:
Chatham House Publishers, 2001b.

154

Mooney, Christopher Z., and Mei-Hsien Lee. "Legislative Morality in the American
States: The Case of Pre-Roe Abortion Regulation Reform." American Journal of
Political Science 39, no. 3 (1995): 599-627.
———. "The Influence of Values on Consensus and Contentious Morality Policy: U.S.
Death Penalty Reform, 1956-82." Journal of Politics 62, no. 1 (2000): 223-239.
Muller v. Oregon, 208 U.S. 412 (1908).
New State Ice Co. v. Liebmann, 285 U.S. 262 (1932).
Norpoth, Hemut, and Jeffrey S. Segal. "Popular Influence on Supreme Court Decisions."
American Political Science Review 88, no. 3 (1994): 711-716.
Norrander, Barbara. "The Multi-Layered Impact of Public Opinion on Capital
Punishment Implementation in the American States." Political Research
Quarterly 53, no. 4 (2000): 771-793.
———. "Measuring State Public Opinion with the Senate National Election Study." State
Politics & Policy Quarterly 1, no. 1 (2001): 111 - 125.
———. "Comment: Choosing among Indicators of State Public Opinion." State Politics
& Policy Quarterly 7, no. 2 (2007): 152-159.
Norrander, Barbara, and Clyde Wilcox. "Public Opinion and Policymaking in the States:
The Case of Post-Roe Abortion Policy." Policy Studies Journal 27, no. 4 (1999):
707-722.
Osborne, David. Laboratories of Democracy. Boston, Mass.: Harvard Business School
Press, 1988.
Ostberg, C.L., Matthew E. Wetstein, and Craig R. Ducat. "Attitudinal Dimensions of
Supreme Court Decision Making in Canada: The Lamer Court, 1991-1995."
Political Research Quarterly 55, no. 1 (2002): 235 - 256.
Oueilhe v. Lovell, 560 P.2d 1348 (1977).

155

Page, Benjamin I., and Robert Y. Shapiro. "Effects of Public Opinion on Policy." The
American Political Science Review 77, no. 1 (1983): 175-190.
Park, David K., Andrew Gelman, and Joseph Bafumi. "State-Level Opinions from
National Surveys: Poststratification Using Multilevel Logistic Regression." In
Public Opinion in State Politics, edited by Jeffrey E. Cohen. Stanford, CA:
Stanford University Press, 2006.
Parker v. Hall, 362 So. 2d 875 (1978).
Paterson Tavern & Grill Owners Association v. The Borough of Hawthorne, 270 A.2d
628 (1970).
People v. Belous, 458 P.2d 194 (1969).
Peper v. Princeton, 389 A.2d 465 (1978).
Peters, B. Guy. "Economic and Political Effects on the Development of Social
Expenditures in France, Sweden and the United Kingdom." Midwest Journal of
Political Science 16, no. 2 (1972): 225-238.
Pinello, Daniel R. Gay Rights and American Law. Cambridge, UK: Cambridge UP, 2003.
Pippen, John, Shaun Bowler, and Todd Donovan. "Election Reform and Direct
Democracy: Campaign Finance Regulations in the American States." American
Politics Research 30, no. 6 (2002): 559-582.
Poole, Ithiel de Sola, Robert P. Abelson, and Samuel L. Popkin. Candidates, Issues, and
Strategies. Cambridge, MA: MIT Press, 1965.
Pound, Roscoe. "The Causes of Popular Dissatisfaction with the Administration of
Justice." Journal of the American Judicature Society 46 (1962): 55 – 66.
Primo, David M., Matthew L. Jacobsmeier, and Jeffrey Milyo. "Estimating the Impact of
State Policies and Institutions with Mixed-Level Data." State Politics & Policy
Quarterly 7, no. 4 (2007): 446 - 459.
Printz v. United States, 521 U.S. 898 (1997).
156

Pritchett, C. Herman. The Roosevelt Court: A Study in Judicial Politics and Values,
1937-1947. New York,: Macmillan Co., 1948.
Ragland, Ruth Ann Vaughan. "Linkages between the Texas Supreme Court and Public
Opinion." PhD, University of North Texas, 1995.
Rakoff, Stuart H., and Guenther F. Schaefer. "Politics, Policy, and Political Science:
Theoretical Alternatives." Politics and Society 1 no. 1 (1970): 51-77.
Rand v. Rand, 374 A.2d 900 (1977).
Ranney, Austin. "Parties in State Politics." In Politics in the American States, edited by
Herbert Jacob and Kenneth Vines. Boston: Little, Brown & Co., 1976.
Ransome v. Ransome, 401 So. 2d 746 (1981).
Reeves, Mavis Mann. "The States as Polities: Reformed, Reinvigorated, Resourceful."
The ANNALS of the American Academy of Political and Social Science 509, no. 1
(1990): 83-93.
Richards, Mark J., and Herbert M. Kritzer. "Jurisprudential Regimes in Supreme Court
Decision Making." The American Political Science Review 96, no. 2 (2002): 305320.
Roch, Christine H., and Robert M. Howard. "State Policy Innovation in Perspective:
Courts, Legislatures, and Education Finance Reform." Political Research
Quarterly 61, no. 2 (2008): 333-344.
Roe v. Wade, 410 U.S. 113 (1973).
Rohde, David W., and Harold J. Spaeth. Supreme Court Decision Making. San Francisco:
W. H. Freeman, 1976.
Sager, Lawrence Gene. "The Legal Status of Underenforced Constitutional Norms."
Harvard Law Review 91, no. 6 (1978): 1212-1264.
———. "Foreward: State Courts and the Strategic Space between the Norms and Rules
of Constitutional Law." Texas Law Review 63, no. 6/7 (1985): 959-976.
157

Saiz, Martin. "Politics and Economic Development: Why Governments Adopt Different
Stategies to Induce Economic Growth." Policy Studies Journal 29, no. 2 (2001):
203 - 214.
Scheb, John M., II, and William Lyons. "The Myth of Legality and Public Evaluation of
the Supreme Court." Social Science Quarterly 81, no. 4 (2000): 928 - 939.
Schiltz, Timothy D., and R. Lee Rainey. "The Geographic Distribution of Elazar's
Political Subcultures among the Mass Population: A Research Note." The Western
Political Quarterly 31, no. 3 (1978): 410-415.
Schubert, Glendon A. The Judicial Mind : The Attitudes and Ideologies of Supreme Court
Justices, 1946-1963. Evanston, Ill.: Northwestern University Press, 1965.
Segal, Jeffrey A. , and Harold J. Spaeth. "Norms, Dragons, and Stare Decisis: A
Response." American Journal of Political Science 40, no. 4 (1996): 1064-1082.
Segal, Jeffrey A., and Albert D. Cover. "Ideological Values and the Votes of U.S.
Supreme Court Justices." The American Political Science Review 83, no. 2
(1989): 557-565.
Segal, Jeffrey Allan, and Harold J. Spaeth. The Supreme Court and the Attitudinal Model.
New York: Cambridge University Press, 1993.
———. The Supreme Court and the Attitudinal Model Revisited. Cambridge ; New York:
Cambridge University Press, 2002.
Seidman, David. "Simulation of Public Opinion: A Caveat." The Public Opinion
Quarterly 39, no. 3 (1975): 331-342.
Sharkansky, Ira. Spending in the American States. Chicago, IL: Rand McNally, 1968.
Simopoulos v. Commonwealth of Virginia, 277 S.E.2d 194 (1981).
Slovenko, Ralph. "The Homosexual and Society: A Historical Perspective." University of
Dayton Law Review 10, no. 3 (1985): 445-457.

158

Smith, Mark A. "The Contingent Effects of Ballot Initiatives and Candidate Races on
Turnout." American Journal of Political Science 45, no. 3 (2001): 700-706.
———. "Ballot Initiatives and the Democratic Citizen." Journal of Politics 64, no. 3
(2002): 892 - 903.
Snetsinger v. Montana University System, 104 P.3d 445 (2004).
Songer, Donald R., Jeffrey A. Segal, and Charles M. Cameron. "The Hierarchy of
Justice: Testing a Principal-Agent Model of Supreme Court-Circuit Court
Interactions." American Journal of Political Science 38, no. 3 (1994): 673-696.
Songer, Donald R., and Reginald S. Sheehan. "Supreme Court Impact on Compliance and
Outcomes: Miranda and New York Times in the United States Courts of
Appeals." The Western Political Quarterly 43, no. 2 (1990): 297-316.
Spaeth, Harold J. An Introduction to Supreme Court Decision Making. Rev. and enl. ed.
San Francisco,: Chandler Pub. Co., 1972.
Spaeth, Harold J., and Douglas R. Parker. "Effects of Attitude toward Situation Upon
Attitude toward Object." The Journal of Psychology 73 (1969): 173 - 182.
Spaeth, Harold J., and Jeffrey A. Segal. Majority Rule or Minority Will : Adherence to
Precedent on the U.S. Supreme Court. Cambridge ; New York: Cambridge
University Press, 1999.
Stanton v. Stanton, 517 P.2d 1010 (1974).
State of Oklahoma v. County Beverage License #Abl-78-145 of Mcmar General Stores,
652 P.2d 292 (1982).
State v. Meloon, 366 A.2d 1176 (1976).
Stimson, James A. Public Opinion in America : Moods, Cycles, and Swings. Boulder:
Westview Press, 1991.
Stimson, James A., Michael B. Mackuen, and Robert S. Erikson. "Dynamic
Representation." American Political Science Review 89, no. 3 (1995): 543 - 565.
159

Strauss v. Horton, 46 Cal.4th 364 (2009).
Swann v. Charlotte-Mecklenburg Board of Education, 402 U.S. 1 (1971).
Tate, C. Neal. "Personal Attribute Models of the Voting Behavior of U.S. Supreme Court
Justices: Liberalism in Civil Liberties and Economics Decisions, 1946-1978." The
American Political Science Review 75, no. 2 (1981): 355-367.
Tolbert, Caroline J., John A. Grummel, and Daniel A. Smith. "The Effects of Ballot
Initiatives on Voter Turnout in the American States." American Politics Research
29, no. 6 (2001): 625 - 648.
A v. X, Y, & Z, 641 P.2d 1222 (1982).
Varnum v. Brien, 763 N.W.2d 862 (2009).
Volden, Craig. "The Politics of Competitive Federalism: A Race to the Bottom in
Welfare Benefits?" American Journal of Politcal Science 46, no. 2 (2002): 352363.
Wasby, Stephen L. The Supreme Court in the Federal Judicial System. 4th ed. Chicago:
Nelson-Hall, 1993.
Weber, Ronald E. Public Policy Preferences in the States. Bloomington, Indiana:
Institute of Public Administration, Indiana University, 1971.
Weber, Ronald E., and William R. Shaffer. "Public Opinion and American State PolicyMaking." Midwest Journal of Political Science 16, no. 4 (1972): 683-699.
Wilson v. Indiana, 278 N.E.2d 569 (1972).
Wright, Gerald C., Robert S. Erikson, and John P. McIver. "Measuring State Partisanship
and Ideology with Survey Data." The Journal of Politics 47, no. 2 (1985): 469489.
Wright, Gerald C., Jr., Robert S. Erikson, and John P. McIver. "Public Opinion and
Policy Liberalism in the American States." American Journal of Political Science
31, no. 4 (1987): 980-1001.
160

Zimmerman, Joseph F. Interstate Cooperation: Compacts and Administrative
Agreements. Westport, CT: Praeger Publishers, 2002.
Zorn, Christopher. "Comparing GEE and Robust Standard Errors for Conditionally
Dependent Data." The Political Research Quarterly 59, no. 3 (2006): 329 - 341.

161