THE INFLUENCE CASE STRENGTH ON SUPREME COURT DECISION MAKING

By

Elizabeth Lane

A DISSERTATION

Submitted to

Michigan State University

in partial fulfillment of the requirements

for the degree of

Political Science – Doctor of Philosophy

2019

ABSTRACT

THE INFLUENCE CASE STRENGTH ON SUPREME COURT DECISION MAKING

By

Elizabeth Lane

The Supreme Court sits at the top of the United States legal system, and despite its po-

sition, researchers do not possess a systematic way to assess the contemporaneous influence

of law on judicial decision making. This is problematic, given research findings over the past

two decades that reveal law does impact judicial decision making. In this dissertation I ar-

gue that bench memoranda provide a solution to this issue. I use Justice Harry Blackmun’s

bench memoranda, authored by his law clerks to provide an evaluation of each party’s legal

argument based on the information provided from merits and amicus curiae briefs to create

a measure of case strength. This measure helps to distinguish the legal aspects of judicial

decision making to determine which litigant has a stronger case. With this measure I demon-

strate that Supreme Court justices can be constrained by law when it acts contrary to their

policy preferences. I also show that an attorney’s position that is more strongly supported

by law can help overcome a poor oral argument performance. Additionally, I examine how

law impacts other aspects of the decision making process, and show that justices’ bargaining

responses to the majority opinion are dependent on the how strongly the law supports the

majority’s position. Lastly, I demonstrate the richness of these bench memoranda data by

showing that the quality of amicus briefs significantly impacts the likelihood of success for

the position in which they advocate.

Copyright by
ELIZABETH LANE
2019

To my mom, I owe you everything

iv

ACKNOWLEDGEMENTS

There are a number of people who supported me on my journey through graduate school

and made the completion of this dissertation possible. On a personal note, I would not be

here today without my mom and dad. Their support has never wavered, and to them I am

deeply grateful. I would not have made it through this process without Adam and Lola.

Adam thank you for keeping me fed, cheering me up, and reminding me that it is okay and

necessary to put away the computer and enjoy life. Thank you to Lola for being the best

research assistant and writing companion I could ask for.

I am sincerely grateful to the National Science Foundation because without their financial

support this research would not have been possible (Award Number: 1728975). This support

allowed me to collect my data and have the enjoyable experience of mentoring some great

undergraduate students to help me complete my research including Anna Schroeder, Maddy

Broderick, Carly Regalado, Madeline Nelson, Steven Dyke, Tyler Hoguet, Alexis Smith, Eli

Pales, and Ashley Fernandez. I am also thankful for the financial support of the Department

of Political Science, the Graduate School, and the College of Social Science. Also, to Karen

Battin, Rhonda Burns, and Sarah Krause, who could always answer my questions about

anything related to the bureaucratic function of our large university and always kept the

train moving.

I was lucky to be surrounded by an exceptional group of peers during my time at MSU.

Each of them is motivated, driven, and has helped and challenged me to work harder.

Particularly, I would like to thank my fellow members of the American Politics Research

Group, Adam Enders, Marty Jordan, Miles Armaly, and Robert Lupton. You helped improve

my research and also made me a better member of the discipline by giving me the opportunity

to read your work as it developed.

I would like to thank Miles Armaly specifically, for

reaching out to talk to me about research, comprehensive exams, and being open to answer

v

any questions I had. You showed me what it meant to be a good mentor to other students

and I like to think that this has carried on throughout the American Politics students in

the department. Also, a special thank you to Ezra Brooks who has the patience of a saint.

You are always willing and able to answer or help me solve any coding question and always

keeping these sessions lighthearted even when I was frustrated.

The support I received from the amazing female graduate students in the department

is unparalleled. Each of you contributed to my success. Starting school with a majority-

female cohort is not typical, but I think each of us was better for it.

I saw how hard

each of you worked and it always motivated me to work even harder. To Emma Slonina,

Jamil Scott, Lora Di Blasi, and Jessica Schoenherr, our weekly WHiPS (Women Hustling in

Political Science) working groups were so helpful in maintaining accountability and getting

things done. And, even those weeks that were tough when maybe not as much got done, I

appreciated that someone was always there to talk. I will miss each of you so much and I

hope that we can continue to plan a yearly writing retreat. A special thank you to Jessica

Schoenherr. Thank you for choosing to study nine old people instead of just one! You have

made grad school, dare I say, a fun experience? You are a coauthor who has turned into a

great friend. Thank you for being a sounding board, talking through ideas, making me a

better writer, and still wanting to work with me.

My committee has been integral in my success as a graduate student. Each one helped

shape me into who I am as a scholar, frame the way I approach research, and I hope to one

day echo their success. Ian Ostrander, you are the newest member of the department but I

am so appreciative of the time you spent reading numerous drafts of fellowship proposals,

manuscripts sent out for review, and countless tips and words of wisdom you offered when

it came to being on the job market. Eric Juenke was the instructor for my first graduate

seminar. I’ve always appreciated your kindness and positivity, especially when it came to the

fall of my market year. I am also thankful for your voluntary participation at numerous talks

vi

and professionalization events, in which I learned so much about this very different profession

I am about to join. Cory Smidt is responsible for nearly all of my graduate education in

quantitative research methods. I appreciate you open door policy and willingness to always

answer any methods question, no matter how small or silly. I would always leave your office

with a page full of ideas of different things to try. I learned so much from you and I will

always remember to “be boldly wrong instead of meekly right.”

Lastly, but most certainly not least, thank you to my advisor, Ryan Black. Pages can fill

books with the debt of my gratitude to you. You and Tim Johnson made this dissertation

possible by sending me to the Library of Congress for seven weeks to collect conference note

data. In the mean time I slacked off a bit because I started examining these thick packets in

the middle of Justice Blackmun’s case files called bench memos, which was such an important

and significant experience in my education. Thank you for notebook pages filled with advice,

not only on research or academia, but life in general - always reminding us it is important

to work but it is just as important to enjoy life. Thank you for reading numerous drafts

of papers or job materials whether you were in cold and snowy Poland, or destined for a

warm tropical climate. Thank you for sharing your resources and leading by example how to

mentor undergraduate students. Your availability and willingness to mentor me can never

be repaid. I only hope I can use the tools you’ve given me to continue my research and help

my future students the way you have helped me and so many others. Thank you.

vii

TABLE OF CONTENTS

LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 1

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 The Law and Judicial Politics . . . . . . . . . . . . . . . . . . . . . . . . . .
1.2 The Benefit of Bench Memos . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.3 Bench Memo Content . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.4 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 2 DOES LAW CONSTRAIN OR POLICY PREVAIL? THE EFFECT
OF LITIGANT CASE STRENGTH ON U.S. SUPREME COURT
DECISION MAKING . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 The Law and Judicial Politics . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 The Benefit of Bench Memos . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3 Examining Case Strength . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.3.1 Methodology and Empirical Results . . . . . . . . . . . . . . . . . . .
2.4 Policy vs. Preferences . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4.1 Methodology and Empirical Results . . . . . . . . . . . . . . . . . . .
2.5 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 3 THE IMPACT OF CASE STRENGTH ON THE SUPREME COURT

JUDICIAL DECISION MAKING PROCESS, A REEXAMINATION .
3.1 The Influence of Oral Arguments on Case Outcomes . . . . . . . . . . . . . .
3.1.1 Data and Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.2 Method and Results
. . . . . . . . . . . . . . . . . . . . . . . . . . .
3.1.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Case Strength and Strategic Responses to Majority Opinions . . . . . . . . .
3.2.1 Data and Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.2 Method and Results
. . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2.3 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
3.3 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 4 QUALITY VERSUS QUANTITY: AMICUS CURIAE BRIEF IN-
FLUENCE AND DECISION MAKING ON THE UNITED STATES
SUPREME COURT . . . . . . . . . . . . . . . . . . . . . . . . . .
4.1 The Influence of Amicus Briefs at the Supreme Court . . . . . . . . . . . . .
4.2 Assessing Effective Versus Ineffective Amicus Briefs . . . . . . . . . . . . . .
4.3 Method and Results
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.4 Effective and Ineffective Amicus Briefs Influence on Supreme Court Case

Outcomes

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

x

xi

1
3
8
10
15

16
19
23
26
33
37
41
46

48
50
52
55
60
61
62
68
79
80

83
86
90
94

98

viii

4.5 Method and Results
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 100
4.6 Discussion and Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . 105

CHAPTER 5 CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 109

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 110
CHAPTER 1 APPENDIX . . . . . . . . . . . . . . . . . . . 111
CHAPTER 2 APPENDIX . . . . . . . . . . . . . . . . . . . 116

APPENDIX A
APPENDIX B

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 120

ix

LIST OF TABLES

Table 2.1: Precedent Vitality Coding . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table 2.2: Ordered Logistic Regression of Case Strength

. . . . . . . . . . . . . . .

Table 2.3: Control Variable Measurement . . . . . . . . . . . . . . . . . . . . . . . .

Table 2.4: Logistic Regression Estimates with Justice-Level Random Effects

. . . .

Table 3.1: Oral Argument Control Variable Measurement

. . . . . . . . . . . . . . .

Table 3.2: Logistic Regression Model of the Justices’ Propensity to Reverse . . . . .

Table 3.3: Contextual Control Variable Measurement . . . . . . . . . . . . . . . . . .

Table 3.4: Multinomial Logistic Regression Model of Justices’ Strategic Responses

to the Majority Opinion Author’s First Draft . . . . . . . . . . . . . . . .

Table 4.1: Multinomial Logistic Regression Model of Effective and Ineffective Amici .

30

34

40

42

55

57

65

69

94

Table 4.2: Logistic Regression Model of the Justices’ Propensity to Vote Liberally . . 102

Table B.1: Logistic Regression Model of the Likelihood of Bargaining . . . . . . . . . 117

Table B.2: Multinomial Logistic Regression Model of Justices’ Strategic Responses

to the Majority Opinion Author’s First Draft . . . . . . . . . . . . . . . . 118

Table B.3: Maltzman, Spriggs, and Wahlbeck’s Original Model

. . . . . . . . . . . . 119

x

LIST OF FIGURES

Figure 1.1: Bench Memo Content

. . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 1.2: Bench Memo Contentions Subsections . . . . . . . . . . . . . . . .

Figure 2.1: Bench Memo Content

. . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2.2: Likelihood of Case Strength Across Petitioner Vitality - The
left panel displays the likelihood of case strength operationalized as a
the five category scale of bench memo recommendations across the 95th
percentile of petitioner vitality while holding all other variables at their
mean and the dichotomous variable at its mode. The right panel depicts
the likelihood of case strength across the 95th percentile of petitioner
experience advantage while holding all other variables at their mean and
the dichotomous variable at its mode.

. . . . . . . . . . . . . . . . . . .

Figure 2.3: Likelihood of a Policy Vote Across Ideological Extremity - This
graph shows the likelihood of a policy vote across the 95th percentile
of ideological extremity. The dashed green line shows the likelihood
of a policy vote when law and policy align. The dashed purple line
displays the likelihood of a policy vote when the law is uncertain. The
dashed orange line represents the likelihood of a policy vote when law
and a justice’s policy preferences come into conflict. The matching solid
lines represent 95% confidence intervals. All variables are held at their
medians.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 2.4: Likelihood of a Policy Vote Across Case Complexity and Case
Salience - The left panel exhibits the results of the relationship between
complexity and case strength and the right panel shows the relationship
between case salience and case strength. The dashed green lines show
the likelihood of a policy vote when law and policy align. The dashed
purple lines display the likelihood of a policy vote when the law is un-
certain. The dashed orange lines represent the likelihood of a policy
vote when law and a justice’s policy preferences come into conflict. The
matching solid lines represent 95% confidence intervals. All variables
are held at their medians.

. . . . . . . . . . . . . . . . . . . . . . . . . .

11

12

24

35

43

45

xi

Figure 3.1: Likelihood of a Reversal Across Oral Argument Performance
- The above plot shows the likelihood of siding with the petitioner for
three levels of case strength across oral argument performance. The
green line represents the law supporting a strong affirmance or a favor-
able outcome for the respondent. The purple line is when the law is
uncertain and the case could go either way and the orange line repre-
sents cases when law favors the petitioner. The vertical bars are 95%
confidence intervals. All other continuous variables are held at their
means and categorical variables are held at their modes.

. . . . . . . . .

Figure 3.2: Likelihood of a Reversal Across Ideological Compatibility - The
plot displays the likelihood of a vote for the petitioner at three levels
of case strength across ideological compatibility. Vertical lines around
that estimates are 95% confidence intervals. All continuous variables
are held at their means and categorical variables are held at their modes.

Figure 3.3: Average Marginal Effect of Majority Coalition-Law Agreement
on Bargaining Response - Each point represents the average marginal
effect of majority coalition-law agreement on each bargaining response.
Vertical lines around that estimates are 95% confidence intervals. All
continuous variables are held at their means and dichotomous variables
are held at their modes.

. . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.4: Likelihood of Threatening and Conurring Responses Across
Majority Coalition-Law Agreement - The left panel displays the
likelihood of a threatening response given the different levels of majority-
law agreement. The right panel shows the likelihood of a concurrence
across the same range. Vertical lines around that estimates are 95%
confidence intervals. All continuous variables are held at their means
and dichotomous variables are held at their modes.

. . . . . . . . . . . .

Figure 3.5: Likelihood of Joining the Majority Across Majority Coalition-
Law Agreement - These are the predicted values of joining the ma-
jority opinion at each level of majority opinion-law agreement. Vertical
lines around that estimates are 95% confidence intervals. All continuous
variables are held at their means and dichotomous variables are held at
their modes.

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.6: Predicted Probabilities of Bargaining Responses Across Across
Majority Coalition-Law Agreement - The plots above show the
predicted probabilities of each bargaining response across each level of
majority-law agreement. Vertical lines around that estimates are 95%
confidence intervals. All continuous variables are held at their means
and dichotomous variables are held at their modes.

. . . . . . . . . . . .

58

60

71

73

75

76

xii

Figure 4.1: Average Marginal Effect of Multifiler on Amicus Type - This
plot shows the average marginal effect of multifiler on three types of
amicus briefs. Vertical lines represent 95% confidence intervals . . . . . .

Figure 4.2: Average Marginal Effect of OSG and State Government on
Amicus Type - The left plot shows the average marginal effect of an
amicus brief filed by the OSG on amicus type. The right plot shows the
average marginal effect of an amicus brief filed by a state government
on amicus type. The vertical lines are 95% confidence intervals.

. . . . .

96

97

Figure 4.3: Likelihood of a Liberal Vote Across MQ Scores Comparing Ef-
fective and Ineffective Amicus Briefs to Regular Amicus Briefs
- Each plot has the 95th percentile of the transformed Martin and Quinn
scores along the x-axis an d the likelihood of a liberal vote on the y-axis.
The orange line in each plot represents a single conservative or liberal
amicus brief that is neither effective no ineffective. Green lines repre-
sent an effective liberal or conservative amicus brief and purple lines
represent a single ineffective liberal or conservative brief. The vertical
lines around that estimates are 95% confidence intervals. All continuous
variables are held at their means and dichotomous variables are held at
their modes. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 101

xiii

CHAPTER 1

INTRODUCTION

On October 30, 1987, Charles Acevedo walked out of an apartment in Santa Ana, Cali-

fornia, placed a small paper bag in the trunk of his car, and drove away. Police, from a tip

received by a drug enforcement agent, had strong reason to believe the apartment Acevedo

had exited from contained a sizable amount of marijuana. On this basis, they subsequently

stopped Acevedo, opened his trunk, and located and opened the paper bag, which contained

marijuana. Acevedo was arrested and charged with possession with intent to distribute. Did

the police, in conducting a warrantless search of Acevedo’s trunk and the paper bag, violate

his Fourth Amendment right against unreasonable searches and seizures? In January 1991,

the U.S. Supreme Court sat to hear arguments in California v. Acevedo on this question.

This was not the first time the Court had taken up the topic. A number of exist-

ing precedents – some stretching back more than 60 years – offered potential ways to re-

solve the constitutionality of Acevedo’s search. On the pro-search side, for example, the

Court had previously held that probable cause was sufficient to allow a search both of a car

(Carroll v. U.S. [1925]) and of containers or packages found within the car (U.S. v. Ross

[1982]). Yet the Court had also held that personal luggage briefly coming in contact with

a trunk (U.S. v. Chadwick [1977]) as well as baggage actually being transported in it

(Arkansas v. Sanders [1979]) could not be searched without a warrant – though the car

itself could be.

Acevedo provides a poignant example of the difficulties facing appellate courts in a com-

mon law system. Precedents, the currency of such a system, are supposed to guide – and

perhaps constrain – case outcomes. If, however, the law is ambiguous, then the guidance

becomes more ambiguous and the constraint weaker. What is more, cases like Acevedo are

more the norm as opposed to the exception. Oftentimes the law is ambiguous and Supreme

1

Court justices and other appellate judges must simply choose a side to prevail even if cloaked

in uncertainty. Judges are regularly faced with difficult choices between competing prece-

dents and other aspects of law. On top of this, most appellate judges decide hundreds of

cases each year, so they must make these decisions quickly, and efficiently each term.

This difficulty also filters down to researchers who seek to understand how and why

judges arrive at one choice versus another, which represents a foundational aspect of judicial

decision making. Researchers recognize a host of factors that influence judges’ decisions

including personality traits (Black et al. 2019; Epstein and Knight 2013; Hall 2018), policy

preferences (Segal and Spaeth 1993), case characteristics (Black and Owens 2009; Caldeira

and Wright 1988; Caldeira, Wright and Zorn 1999), and the law (Bailey and Maltzman

2008; Hansford and Spriggs 2006; Richards and Kritzer 2002). While scholars have reliable

measurements and indicators for most of these topics, systematically accounting for law and

legal influence has proven difficult.

To help aid scholarship and develop our understanding of judicial processes, in this disser-

tation I create, validate, and demonstrate the significance of a novel measure of case strength

from Justice Harry Blackmun’s bench memos. These data will enhance how scholars assess

the impact of law on judicial decision making on the U.S. Supreme Court, a topic that has

fascinated researchers for decades. My approach has a number of desirable characteristics

that significantly improve upon existing approaches. First, the data for my measure comes

directly from materials actually used by justices in cases. In particular, I analyze new and

unique archival data - bench memoranda, prepared for Justice Harry Blackmun by his clerks

prior to oral arguments. As I explain in greater detail below, bench memos provide a unique

perspective on the relative strengths of competing legal arguments in a case. Second, because

the bench memos are written immediately prior to oral argument, they represent a strictly

contemporaneous assessment of legal quality that does not depend upon how opinions are

subsequently “treated” by other courts or political actors.

2

These data will provide scholars with an unparalleled view of the relative strengths of the

legal arguments being made by each side immediately before the Supreme Court begins to

decide a case. Such data can be used to operationalize legal strength, a concept that existing

scholarship acknowledges as being important but difficult to measure. As such, my research

has the potential to impact decades worth of scholarship that has struggled to properly

measure law. To demonstrate the utility and necessity of this measure in judicial behavior

research, I address three questions throughout the dissertation. First, I address a novel, yet

significant question in the literature: What happens when law and ideology come in conflict

in a case? Next, I reevaluate existing research questions to demonstrate the importance

of incorporating the contemporaneous influence of law on multiple aspects of the judicial

process including: Does law impact Supreme Court justices’ decisions on the merits? Does

legal strength influence the opinion bargaining process? I also demonstrate the richness of

these data by examining what factors make an effective amicus brief.

In this introduction, I review existing approaches to incorporating law into the study of

judicial politics. I also provide background on bench memos including why they are useful

for capturing case strength and descriptive details on the specific data I am using.

1.1 The Law and Judicial Politics

Law or politics? This debate has been the main event in the field of judicial politics for

decades. Throughout its history, judicial politics has been muddled in controversy as to how

research on the courts should be approached. Epstein and Knight (2000) detail the history

of the discipline and the sharp divides that arose between “traditional” public law scholars

and behavioralists. These divisions were both methodological and theoretical; public law

researchers studied the influence of law on judicial behavior qualitatively, and behavioralists

deployed quantitative methods to document the influence of “politics,” which, depending on

the decade, has been called attitudes, ideology, or preferences.

The attitudinal approach, as it would eventually be coined (Segal and Spaeth 1993),

3

emerged as the prevailing framework for studying judicial politics. Near the turn of the

century the strategic model began to gain traction, which acknowledged the importance

of preferences in judicial outcomes, but unlike the attitudinal model, did not simply view

justices as “politicians in robes” (Epstein and Knight 2013, 14). Instead, the strategic ap-

proach allows for influence of non-attitudinal constraints, which scholarship has shown to

include things like institutional norms (Maltzman, Spriggs and Wahlbeck 2000), public opin-

ion (McGuire and Stimson 2004), and the law (George and Epstein 1992), to name only a

few.

Despite the fact that schoalrs moved past the attitudinal approach to recognize that

justices are influenced by more than policy preferences they struggled to adequately account

for legal influence. One of the difficulties that plagues researchers is that of observational

equivalence. While law is often thought of as a constraint on justices pursuing their policy

preferences, this is not always the case. Oftentimes, law and justices’ policy preferences align

and when this happens, that is law and attitudes predict the same behavior, it is impossible

for researchers to determine which factor is driving that behavior.

The earliest attempt to try and distinguish between law and policy preferences was

conducted by the pioneers of the attitudinal model, Segal and Spaeth (1996). They theorized

if justices truly respect precedent as they claim, voting behavior in cases that follow a

landmark decision should demonstrate respect for said decision, particularly by justices who

dissented in the original case. For example, Justices White and Rehnquist dissented in the

Court’s landmark decision in Roe v. Wade (1973). According to Segal and Spaeth’s (1996)

theory, White and Rehnquist should accept this decision and vote in favor of a woman’s

right to privacy in progeny cases like Planned Parenthood v. Danforth (1976) or Planned

Parenthood of Southeastern Pennsylvania v. Casey (1991).

Instead, they found that 90

percent of the votes cast were consistent with an attitudinal as opposed to legal account,

with only two out of fifteen justices showing even modest evidence of being influenced by

4

precedent (see also Spaeth and Segal (2001)). Subsequent work also identified a number

of concerns regarding Segal and Spaeth’s (1996) research design, which some argued biased

finding support for preferential votes (Brisbin 1996; Bailey and Maltzman 2008). To address

these concerns, scholars began to take different approaches to documenting law’s influence.

Three distinct but complementary lines of research are particularly noteworthy.

Richards and Kritzer (2002) were some of the first to attempt a new approach to dis-

entangling law and preferences. They introduced the concept jurisprudential regime theory

(JRT). According to Richards and Kritzer (2002), landmark cases establish new regimes,

which establish “which case factors are relevant for decision making” (305). Put differently,

Richards and Kritzer (2002) zone in on landmark cases that establish tests (e.g., the Lemon

Test) or rules. These tests provide a framework for what aspects of the case facts are relevant

for the justices’ decision. Because these relevant factors are established prior to the case the

Court is deciding, JRT provides another opportunity to falsify the legal model. Jurispruden-

tial regime theory (JRT) has been applied to a number of distinct areas (Luse et al. 2009;

Richards and Kritzer 2002; Kritzer and Richards 2003; Richards, Smith and Kritzer 2006),

but its usage is not without controversy. For one, JRT rests on the assertion that the Court’s

decisions result in “revolutionary” change. That is, once a new rule or test is established by

a Court opinion, the justices immediately apply this test in all future cases. Bartels and

O’Geen (2014) demonstrate that this is typically not the case. In fact, most decisions lead

to a more gradual change, and at times decisions that should lead to change actually result

in relative stability.

Beyond the theoretical concerns of JRT, a more practical limitation is that regimes are,

by design, specific to a narrow substantive issue area where the Court has established tests

Moreover, one also must know which regime is dominant at a particular point in time, which

often only becomes clear after the Court has issued several decisions in an issue area. As a

result, it becomes impossible to incorporate JRT-style legal influence into analyses that spans

5

across a host of issue areas or attempts to explain case outcomes in a contemporaneous (i.e.,

as they are decided) manner. This approach, then, would not be viable for trying to address

the Court’s dilemma in Acevedo, for example, without leaning heavily on what happened

after the decision was released.

A second approach comes from the collaborative work of Hansford and Spriggs (2006),

who argue that policy-minded justices “treat” existing precedents with a view towards not

just their own ideological preferences but also the previous treatment of that precedent – a

concept they label “vitality.” This approach teaches us much about how the Court develops

legal policy through the expansion, limitation, and even overruling of its own precedents

(Spriggs and Hansford 2001). Yet for all it teaches us about the impact of law, its practical

usefulness is somewhat lacking. Consider, again, the leading example this chapter, Acevdeo.

Adopting vitality as a predictor for what the Court would do is not impossible, but poses

a number of challenges. One could, for example, examine the table of authorities for the

parties’ briefs in a case to determine the overall vitality of Supreme Court precedents cited

by each side. But, as Hansford and Spriggs’ work tells us, citing a precedent is not the same

thing as interpreting it. Indeed, both petitioner and respondent briefs in Acevedo cite all four

of the most relevant precedents mentioned earlier. This means any variation between the

petitioner and respondent’s vitality score will be driven by peripheral precedents that one

brief happens to mention. Only a content analysis that determined how each brief “treated”

each precedent would yield an appropriate vitality measure. Such work is possible, but would

be difficult to practically (and reliably) accomplish across any substantial number of cases.

The third approach, is the most recent and sophisticated, offered by Bailey and Maltzman

(2008). The authors’ goal is to distinguish between the political and legal aspects of First

Amendment issues. To do this, the authors develop the concept of bridging observations,

which shared issues between the Supreme Court and legislative and executive branches. More

specifically, they are cases voted on by the justices and also commented on, or sometimes

6

voted on, by Members of Congress and/or the President. So long as elected officials’ positions

are less influenced by law than the votes of Supreme Court justices (2008, 373-374), then

Bailey and Maltzman are able to identify just who is influenced by law and how much legal

influence exists. The difficulty with this approach is being able to identify and incorporate

these bridging observations into their data. Politicians feasibly comment on hot topic First

Amendment issues that Bailey and Maltzman (2008) use, like freedom of speech, press, and

religion. However, these cases comprise less 7.65% of all of the Supreme Court’s cases from

1946 to the 2017 term (Epstein et al. 2019), which is to say the vast majority of the Court’s

cases are topics that are unlikely to warrant politicians’ attention. Additionally, this data

limitation is exacerbated by the fact that if the House or Senate are not voting on a more

mundane issue, they are even less likely to be making public statements on it. In short, the

scope of cases available to analyze is necessarily limited.

To be clear, I do not argue that we need to show that law “matters.” To the contrary,

the seminal works discussed make it difficult for one to plausibly claim that judges are little

more than legislators in robes. That being said, where the discipline continues to lag behind

is in providing researchers with measures that allow them to validly and reliably document

the impact of law when conducting their analyses.

An example illustrates. Johnson, Wahlbeck and Spriggs (2006) argue that an attorney’s

performance during oral argument influences his likelihood of winning a justice’s vote. To

show this relationship, Johnson and colleagues draw from the grades given to attorneys by

Justice Blackmun during oral argument. This finding was controversial, as it challenged the

conventional wisdom; to wit, Segal and Spaeth state that oral arguments do not “regularly,

or even infrequently, determine who wins and who loses” (2002, 280). And, despite John-

son, Wahlbeck and Sprggs’ (2006) thorough and careful empirical analysis, they themselves

concede that “it is possible that attorneys get higher grades in cases in which they have the

‘better’ legal position” (2006, 108, n24). Though they go on to argue for why they do not

7

believe this is true, absent being able to measure the quality of the underlying legal claim,

there is ample reason to be skeptical of the finding (see also Lax 2012; Budziak and Lempert

2015). In Chapter 2, I will examine this criticism using my novel measure of case strength.

Even a decade later this continues to be a problem confronting the literature. Black

et al. (2015), for example, show that justices are more likely to rule in favor of the party

whose brief contains less emotional language as it undermines their credibility. This could be

true, but it might also be the case that higher quality legal positions are easier to articulate

in briefs, thereby making it less imperative to invoke emotional language. The research I

propose here would fill a critical need in the literature and allow for both the testing of new

theories and claims about decision making and, perhaps just as importantly, re-testing a

variety of existing ones with more appropriate measures. Having outlined the intellectual

merit of measuring legal quality, I turn next to discussing how bench memos can provide

this information.

1.2 The Benefit of Bench Memos

How should scholars measure legal quality or case strength? In a world without resource

constraints, one might impanel a group of skilled attorneys, who would review all briefs in

all cases prior to being decided. These attorneys would assess the strength of each argument

made by a litigant and then, perhaps, offer an overall summary assessment of which side has

the “better” case. Individual idiosyncrasies would be offset by having multiple raters, and by

having the ratings performed before a case was decided to guard against possible post hoc

bias. And, of course, such ratings would exist for all cases ever decided.

Although reality necessarily falls short of the ideal, using bench memos as a tool to

rigorously assess case legal strength comes quite close to many aspects of a perfect measure.

In short, bench memos are crafted by some of the best and brightest young lawyers in the

country (Ward and Weiden 2006; Peppers 2006), who are tasked with both summarizing

and evaluating the totality of information and arguments put forward in a case before the

8

Supreme Court. They make a dispositional recommendation in a case and support that

justification with legal reasoning based on the arguments in the case.

Bench memoranda are written to prepare justices for oral arguments. They provide an

initial evaluation of the legal arguments of each party in a case (based on briefs on the mer-

its), as well as a description of amici, the justice’s and Court’s past treatment of precedent in

the relevant area of case law, and will also even mention Congressional intentions on statues,

and/or the legislative or the executive branch’s preferences in a case. As described, these

memos provide descriptions of multiple aspects of legal doctrine, important for truly deter-

mining the effect law has on judicial decisions (Bailey and Maltzman 2008). Additionally,

these memos are written by clerks prior to when the case is orally argued. This prevents

potential post hoc biases – conscious or otherwise – from being injected into the measure

based on coding being done weeks, months, or even years after the Court had issued its

decision in a case (Harvey and Woodruff 2013).

Although I am the first to draw upon bench memos, existing work shows how archival

materials – especially those generated by law clerks – can be of value to scholars. In the con-

text of Supreme Court agenda setting, for example, what little guidance the Court provides

into its decision making suggests that legal conflict is an important factor in its decision
to grant discretionary review (see, i.e., Supreme Court Rule #10).1 Yet coming up with

a reliable and valid measure of conflict alluded scholars and significantly limited scholarly

ability to systematically study agenda setting. Enter the role of certiorari memos, which are

the agenda-setting equivalent of bench memos: documents prepared by law clerks when a

case seeks Supreme Court review. Here, too, law clerks summarize the claims, provide an

assessment of their strength, and, in particular, comment on the extent to which a claimed

conflict is real or exaggerated. These data, gathered with extensive support from the NSF,
1Supreme Court Rule 10 dictates the compelling reasons a justice should be compelled to
grant review, including conflict amongst lower courts, a lower court incorrectly interpreted
precedent, or a new legal question.

9

have been a tremendous boon to scholars and used extensively in published research (e.g.,

Black and Owens 2009, 2011, 2012a; Black and Boyd 2012, 2013; Black, Boyd and Bryan

2014).

1.3 Bench Memo Content

Beginning in 1972, the format of Justice Blackmun’s bench memos took shape and re-

mained consistent until the end of his career. The average length of Justice Blackmun’s

bench memos is 29.5 pages. The length breakdown and a description of each primary section

that appears in every memo is shown in Figure 1.1.

Sometimes called the introduction, other times called the summary, each memo begins

with a short overview of the case. It typically begins with the primary questions the litigants

are asking the Court to answer followed by a brief description of the dispute. When the

Summary section is longer clerks will also preview their dispositional recommendation and

reasoning in this section. This section ranges from a half page to seven pages long, and

averages just over two and a half.

After the summary section is the facts section, also referred to as background, or facts

and proceedings below. Dependent on the complexity of the case, this section can be as

short as one page or extend up to 19 pages. The Summary section takes up a more detailed

recounting of the facts and the proceedings in the lower court. Highlights of this section

include summaries of lower court opinions and, in particular, dissents. These sections read

like the any Constitutional law textbook summary of the facts of the case.

10

Figure 1.1: Bench Memo Content

After reviewing what has taken place to date, the bench memo addresses the contentions

before the Court. As Figure 1.1 makes clear, the contentions section occupies the largest

amount of space in most memos, comprising just over 10 pages on average, but the longest

contentions section spans over 46 pages. In this section, the memo author summarizes the
claims made by both parties in a case and amici.2 This section typically is comprised of

multiple subsections, as shown in Figure 1.2. The petitioner’s and respondent’s contentions

sections are analogous to the arguments section in each litigant’s merits brief because they

provide justification as to why each party believes they should win. In Acevedo, for exam-
2It is also worth noting that this section reports arguments made from reply briefs. Such
data will be valuable as existing work focuses on arguments that appear in the parties’
initial merits briefs, which fails to consider the iterative nature of appellate brief writing
(e.g., Corley 2008; Black et al. 2015).

11

0.02.55.07.510.0SummaryFactsContentionsDiscussionConclusionQuestionsMiscellaneousSectionsAverage Section Length in Pagesple, the State of California argued that the Acevedo’s paper bag should be given the same

treatment as the car itself. As the memo author writes: “When resp[ondent] placed the bags

of drugs in the Honda, the bag became as mobile as the vehicle it was in...Given probable

cause to believe the vehicle contained a bag, and that bag contained illegal drugs, the officers

should have been able to seize the bag and search it without a warrant” (7).

Immediately following each litigant’s contentions is a description of the amicus briefs filed

on behalf of them. If the OSG files an amicus brief for either party they have their own

subsection separate from the rest of the briefs. As I explain in Chapter 3, clerks will provide

commentary on the usefulness of certain amicus briefs. They will even instruct Justice

Blackmun to spend considerable time reading them because they offer unique insights or

they are superior to the litigant’s brief.

Figure 1.2: Bench Memo Contentions Subsections

Having described the scope of the legal claims being made, the memo author next dis-

cusses the relative strength of these arguments. This section is one of the most important

12

01234Petitioner'sContentionsPetitioner'sAmiciRespondent'sContentionsRespondent'sAmiciSectionsAverage Section Length in Pagessections of the entire memo. It is where a clerk provides an analysis of both parties’ legal

arguments. The benefit of the bench memo, and discussion section in particular, over the

individual merits briefs is that the clerks will bring to light any factually incorrect or reach-

ing statements. In an effort to succeed before the Court, attorneys will sometimes overstate

the relevance of a particular case or try and twist the interpretation of a particular statute

ever so slightly. Take for example the case Bonelli Cattle Co. v. Arizona (1973), Justice

Blackmun’s clerk wrote: “I think that pet[itione]r has somewhat overstated the scope of the

Bonelli decision id. the Ariz[ona] courts. Federal, Indian, other federal patent land, Cali-
fornia, and Nevada issues are not properly before this Court (18).”3 Or, Gateway Coal Co.

v. United Mineworkers (1973), where Blackmun’s clerk calls out both the petitioner and

respondent for exaggerating the potential consequences of the Court’s decision. “Pet[itione]r

clearly overstates the dangers of chaos and Respondent clearly overstates the dangers of an
arbitrator deciding, safety questions (16).”4

In addition to correcting anything misstated by the parties, the clerks will also touch

on each legal question presented. They will discuss the pros and cons of each litigant’s

argument in light of past decisions by both the Court and Justice Blackmun, Congress’s

intent when they wrote particular statutes, and even interpreting the Constitution. All of

their discussion leads to a final recommendation. In the case of Acevedo, the memo author

found more strength in California’s aforementioned claim versus one offered by Acevedo’s

attorney, who argued that a closed container should carry a higher expectation of privacy,

especially once the automobile that is carrying it has been stopped (thereby eliminating

exigent need for a warrantless search). She wrote: “The police saw resp[ondent] place the
3Bench memorandum in Bonelli Cattle Co. v. Arizona (72-397). Located in the Supreme
Court Case Files (Box 173, Folder #8) of Justice Harry A. Blackmun in the Library of
Congress, Washington D.C.

4Bench memorandum in Gateway Coal Co. v. United Mineworkers (72-782). Located in
the Supreme Court Case Files (Box 175, Folder #2) of Justice Harry A. Blackmun in the
Library of Congress, Washington D.C.

13

brown paper bag in the trunk of his car...The police may search that area of the car and

any container within it which is likely to contain the object of their search. Here, the

police searched the trunk and the brown bag and discovered marijuana...I would admit the

evidence” (16-17). The Discussion section is the second longest portion of the memo, and as

shown by Figure 1.1, it averages just over eight pages.

Importantly, the bench memo provides a conclusion, which is generally around one page

long. This conclusion is not just a recommended disposition in a case (e.g., affirm or reverse),

but also a brief summary of the legal rationale to support that disposition. That is, what sort

of justification should the Court use in its majority opinion to reach a particular outcome?

In Acevedo, the clerk argued that a single rule should govern all automobile exceptions:

so long as the police have probable cause to search an area of the car, then that probable

cause extends to all containers located therein. This section, as I will explain in Chapter 1, is

crucial for creating a measure of case strength. This final recommendation is the culmination

of the complete and analyses of both litigants’ arguments. It represents the litigant that has

the stronger legal argument.

The majority of bench memos are also accompanied by questions written by the clerk

directed towards both respondent and petitioner for Justice Blackmun to ask during oral

arguments. Justice Blackmun would often scribble his own questions on these pages as well.

Unfortunately, the availability of these question section in many cases is limited. They are

hardly ever attached to the rest of the memo because Justice Blackmun actually would bring

this page with him to oral argument. As a result, they sometimes did not make it back into

the case folder. Yet, these pages are demonstrate the importance of bench memos to the

way Justice Blackmun, and many other chambers on the Court function (Johnson, Stras and

Black 2014; Peppers and Zorn 2008).

Finally, the miscellaneous section. This is not the name of an actual section in the

bench memos, but a name I gave to represent the miscellaneous supplementary materials

14

clerks would occasionally attach to the end of a memo. Most often this takes the form of

a written memo to provide new or updated information, sometimes a correction. In other

circumstances, it is copies of newspaper clippings, maps, or relevant case material that cannot

be accurately conveyed through text.

1.4 Discussion

In what follows, I will establish, validate, and apply my measure of case strength, a

systematic, contemporaneous evaluation of legal doctrinal strength that can be used across

multiple cases and terms from 1972 - 1993. With my measure of case strength and ancillary

data, researchers will finally be able to understand the impact of law on judicial decisions,

and its constraining, or encouraging, capacity for justices to pursue their preferred outcomes.

This measure will provide the most comprehensive view to date of the the Supreme Court

decision process, and fundamentally transform the way the way political scientists understand

case outcomes and their policy consequences.

A measure of legal influence that can be used across terms and issue areas has alluded

scholars for decades. My project fills this void and provides political scientists with a novel

measure of legal quality to assess the impact of law on the Supreme Court decision making.

This operationalization of legal influence will not only help answer fundamental questions of

how justices are constrained by law and propose new theories. It will also allow scholars to

reevaluate existing theories in light of this new measure.

15

CHAPTER 2

DOES LAW CONSTRAIN OR POLICY PREVAIL? THE EFFECT OF

LITIGANT CASE STRENGTH ON U.S. SUPREME COURT DECISION

MAKING

What could explain Justice Antonin Scalia, one of the most conservative voices on the

modern Court, abandoning his policy preferences to be the critical fifth liberal vote in an

important criminal rights case? Surprisingly, this is exactly where Scalia found himself

in during the 1990 term in the case Arizona v. Fulminante. While in prison, Fulminante

confessed to a confidential informant to his involvement unsolved murder in exchange for

protection from threatening inmates. This confession and a subsequent confession to the

informant’s wife were used to indict Fulminante and charge him with first degree murder.

The question brought to the Supreme Court to answer: were the confessions coerced and
therefore unconstitutional? 1

This was not the first time the Court had taken up the topic of coerced confessions. A

number of existing precedents – some stretching back more than 100 years – offered potential

ways to resolve the constitutionality of Fulminante’s confession. The Court previously held

that a confession was coerced when the police told a man if he did not confess they could not

protect him from an angry mob waiting to lynch him (Payne v. Arkansas [1958]). Further-

more, the Court ruled that for a confession to be voluntary it must not be extracted under

any threat of violence, nor obtained by promises from government officials (Brady v. U.S.

[1970]), and any involuntary confession is inadmissible (Bram v. U.S. [1897]). Overwhelming

precedent pointed towards voiding Fulminante’s confession. The law is often ambiguous, but

in this case, it appeared that there was ample evidence to support the respondent.

Scalia’s vote Fulminante and other cases like it are tossed up to be random noise in data,
1Bench memorandum in Arizona v. Fulminante (89-839). Located in the Supreme Court
Case Files (Box 565, Folder #3) of Justice Harry A. Blackmun in the Library of Congress,
Washington D.C.

16

oftentimes unexplainable even with a host of factors researchers recognize as influencing

judges’ decisions including personality traits (Black et al. 2019; Epstein and Knight 2013;

Hall 2018), policy preferences (Segal and Spaeth 1993), and case characteristics (Black and

Owens 2009; Caldeira and Wright 1988; Caldeira, Wright and Zorn 1999). Past research

recognizes the constraining capacity legal doctrine can exert on judges’ decisions (Bailey

and Maltzman 2008; Hansford and Spriggs 2006; Richards and Kritzer 2002), but what is

missing is a reliable and systematic method to account for the contemporaneous influence of

law on justices’ decision making. A measure like this would achieve one of the discipline’s

primary goals, it would allow researchers to disentangle law and legal influences from politics.

It would help explain Scalia’s vote in Fulminante, and thousands of other policy antithetical

votes on the Court instead of tossing them up to an unexplainable anomaly.

To better understand how judges arrive at their final decisions on the merits and how law

impacts this decision, I analyze new and unique archival data - bench memoranda, prepared

for Justice Harry Blackmun by his law clerks prior to oral arguments. Bench memoranda

provide a unique perspective on the relative strengths of competing arguments in a case.

They detail factual distinctions and sort out relevant precedents. Importantly, these memos

also contain a final recommendation on how the case should be decided after weighing each

party’s argument. These memoranda provide an initial merits evaluation, untarnished by

the bargaining process. I contend that this recommendation represents an assessment of the

strength of each party’s legal argument that can be used as an objective view of law to help

explain votes in cases like Fulminante.

With this measure I examine the factors that influence the likelihood of a justice voting

with his policy preferences. I demonstrate that law acts in both a constraining and liberating

capacity. When law and policy conflict, a justice is significantly less likely to pursue his policy

preferences. This relationship is conditional on a justice’s ideology. Ideologically moderate

justices like Powell, O’Connor, and Kennedy, are more likely to be constrained by the law

17

than their ideologically steadfast peers, like Scalia or Douglas. I also find that the Solicitor

General and the litigants’ resource advantages impact the likelihood of a policy vote.

My findings make two important contributions. First, using over 100,000 pages of origi-

nally collected bench memoranda I developed a measure of legal strength for the cases decided

between the 1972 - 1993 terms. This provides a systematic way to assess the contemporane-

ous influence of law on decision making. Works by Bailey and Maltzman (2008); Bartels and

O’Geen (2014); Hansford and Spriggs (2006); Richards and Kritzer (2002) demonstrate the

importance of law on judicial decision making. Yet, as I will discuss in further detail below,

most of these approaches rely on post hoc information to assess legal influence. Because

bench memos are written immediately prior to oral argument, they represent an assessment

of legal quality at the time of the decision. They do not rely on subsequent treatments,

shifts in treatments, or other extra-judicial actors to evaluate each party’s legal argument.

What is more, my method of assessing legal quality can be used systematically across all

issue areas for 22 terms of the Court; and using this measure I am able to demonstrate that

law can constrain justices enough to vote against their ideological predispositions. Insight

into this relationship is pivotal for understanding how the justices arrive at their decisions

and, more generally, how one of the most powerful branches of government operates.

Next, I contribute to the long line of research that demonstrates the value of archival

data in furthering researchers’ understanding of the judicial process. Although my work is

the first to draw upon bench memos, existing work shows how archival materials – especially

those generated by law clerks – can be of value to scholars. In the context of Supreme Court

agenda setting, for example, a reliable and valid measure of conflict alluded scholars and

significantly limited scholarly ability to systematically study agenda setting. Enter the role

of certiorari memos, which are the agenda-setting equivalent of bench memos: documents

prepared by law clerks when a case seeks Supreme Court review. Here, too, law clerks

summarize the claims, provide an assessment of their strengths, and, in particular, comment

18

on the extent to which a claimed conflict is real or exaggerated. These data, gathered

with extensive support from the NSF, have been a tremendous boon to scholars and used

extensively in published research (e.g., Black and Owens 2009, 2011, 2012a; Black and Boyd

2012, 2013; Black, Boyd and Bryan 2014).

In order to establish this link between legal case strength and the final recommendation

of each bench memo, I begin by detailing what a bench memo is and the advantages of using

this new data. I also discuss how scholars have captured legal influence and the impact of

law in the past. To support my assertion that these memos capture the quality and strength

of a case, I analyze the recommendation provided at the end of each memo to determine

if indicators traditionally related to a superior legal argument, such as attorney experience

and precedent vitality are related to this conclusion. Next, I demonstrate the importance

and utility of this measure by examining if how law impacts justices’ voting. Specifically, my

results reveal that law can act in both a liberating and constraining capacity for a justice

strategically pursuing his policy preferences (Epstein and Knight 2000).

2.1 The Law and Judicial Politics

Throughout its history, the field of judicial politics has been muddled in debate as to

how research on the courts should be approached. The law versus politics dispute has been

central to this controversy since the beginning. Originally, they were viewed as two mutually

exclusive influences on judges’ decision making. Depending on who you asked, judges were

principled decision makers only swayed by the rule of law, or, on the contrary, judges were

“politicians in robes” (Epstein and Knight 2013, 14) using symbols of law to disguise their true

political preferences. Research has since moved on from this rival approach, the terminology

has expanded - politics, attitudes, ideology, or preferences - but the relationship between law

and politics still plagues researchers.

Law is often ambiguous. The song of precedent seldom sings in harmonious tone, but

more commonly results in a cacophony of discordant voices. At the end of the day, one

19

of these voices will be chosen to prevail above the rest. This, of course, is the task facing

Supreme Court justices and other appellate judges. Such individuals must make difficult

choices among competing precedents; and they must do this efficiently. This difficulty also

filters down to researchers who seek to understand how and why judges arrive at one choice

versus another, which represents a foundational aspect of judicial decision making. The

ambiguity of law also makes it difficult to measure. Additionally, law is not diametrically

opposed to judges’ attitudes, and when they point in the same direction it is exceedingly

difficult to distinguish between the two. Not to mention, there are a host of other factors

that also influence judicial decision making, like institutional norms (Maltzman, Spriggs and

Wahlbeck 2000), public opinion (McGuire and Stimson 2004; Casillas, Enns and Wohlfarth

2011a), and personality traits (Black et al. 2019; Epstein and Knight 2013; Hall 2018).

Scholars seeking to disentangle the relative influence of legal considerations versus other

factors (such as those identified above) have been plagued by the problem of observational

equivalence. That is, if both law and attitudes predict the same behavior, then it is im-

possible to determine which is driving such behavior. Segal and Spaeth (1996) offered an

innovative and influential approach by examining how dissenting justices in landmark deci-

sions (such as Roe v. Wade [1973]) subsequently behaved in “progeny” cases (such as Planned

Parenthood v. Danforth [1976]). Did such justices continue to adhere to their preferences and

dissent, or did they accept the precedential value of the landmark case and join the majority?

By their accounting, fully 90 percent of the votes cast were consistent with an attitudinal

as opposed to legal account, with only two out of fifteen justices showing even modest evi-

dence of being influenced by precedent (see also Spaeth and Segal (2001)). Subsequent work

identified a number of concerns regarding Segal and Spaeth’s research design, particularly a

bias toward finding support for preferential votes (Brisbin 1996; Bailey and Maltzman 2008).

To address these concerns, scholars began to take different approaches to documenting law’s

influence. Three distinct but complementary lines of research are particularly noteworthy.

20

The first of these approaches comes from Richards and Kritzer (2002), who introduce the

idea of jurisprudential regimes. These regimes, according to Richards and Kritzer, establish

“which case factors are relevant for decision making” (305). Jurisprudential regimes further

provide an opportunity to falsify the legal model because the influence of these factors are

distinct “from their influence in cases decided prior to the establishment of the regime” (305-

306). Jurisprudential regime theory (JRT) has been applied to a number of distinct areas

(Luse et al. 2009; Richards and Kritzer 2002; Kritzer and Richards 2003; Richards, Smith

and Kritzer 2006), but its usage is not without controversy. Bartels and O’Geen (2014),

for example, argue not all of the Court’s outputs result in “revolutionary” JRT-like change,

other decisions lead to more gradual change, or even result in relative stability. Beyond

this concern, a more practical limitation of JRT is that regimes are, by design, specific to a

narrow substantive issue area where the Court has established tests (e.g., the Lemon Test).

Moreover, one must also know which regime is dominant at a particular point in time, which

often only becomes clear after the Court has issued several progeny decisions in the same

issue area. As a result, it becomes impossible to incorporate JRT-style legal influence into

analyses that span across a host of issue areas or attempts to explain case outcomes in a

contemporaneous (i.e., as they are decided) manner. This approach, then, would not be

viable for trying to address the Court’s dilemma in cases where the law is unclear without

leaning heavily on what happened after the decision was released.

A second approach comes from the collaborative work of Hansford and Spriggs (2006),

who argue that policy-minded justices “treat” existing precedents with a view towards not just

their own ideological preferences but also the previous treatment of that precedent – a concept

they label “vitality.” This approach teaches us much about how the Court develops legal

policy through the expansion, limitation, and even overruling of its own precedents (Spriggs

and Hansford 2001). Yet for all it teaches us about the impact of law, its practical usefulness

is somewhat lacking. In order to obtain the vitality estimate for each litigant’s argument one

21

would have to examine not only which cases the petitioner and respondent mention in their

briefs, but also analyze how they “treat” them. This is an extremely timeconsuming edeavour

when applied to briefs but feasible if applied to bench memos, which summarize the primary

contentions of briefs while cutting out superfluous information. While precedents are one

of the main attributes of law justices must consider it is not the only one. Congressional

intent, constitutional interpretation, and statutes and administrative laws all factor into

their decision-making process. Only considering past Supreme Court decisions leaves an

incompletel picture of law’s influence on judicial decision making.

The third approach, offered by Bailey and Maltzman (2008) is the most recent and

sophisticated yet and comes closes to truly disentangling law and preferences. This approach

very cleverly uses bridging observations – cases both voted on by the justices and also

commented or, less commonly, voted on by members of Congress and/or the President –

to identify law’s impact. So long as elected officials’ positions are less influenced by law

than the votes of Supreme Court justices (2008, 373-374), then Bailey and Maltzman are

able to identify just who is influenced by law and how much influence exists. Crucially,

however, this approach depends on being able to identify and incorporate bridging data for

non-judicial actors. While it is easy to think of politicians voicing opinions on controversial

topics such as equal protection or privacy, other issues like patent law will be more difficult

to find on a regular basis. Additionally, this data limitation is exacerbated by the fact that

if the House or Senate are not voting on a more mundane issue, they are even less likely

to be making public statements on it. In short, the scope of cases available to analyze is

necessarily limited.

It is clear from these past approaches that law matters. Using originally collected data

I provide a fourth approach that captures which litigant has a superior legal argument ac-

cording to law. This approach can be used systematically across all issue areas, incorporates

all aspects of law and legal influence, and uses relevant information that was available at the

22

time of the decision.

2.2 The Benefit of Bench Memos

Capturing the strength each litigant’s legal argument in a case is difficult task. The

law is complex system of rules, statutes, and past court decisions and in order to efficiently

and effectively complete this endeavor a deep understanding of the American legal system

is required. In a world without resource constraints, one might impanel a group of skilled

attorneys who would review all briefs in all cases prior to being decided. These attorneys

would assess the strength of each argument made by a litigant and then, perhaps, offer an

overall summary assessment of which side has the “better” case. Individual idiosyncrasies

would be offset by random assignment and by having multiple raters. Ratings would be

performed before a case was decided to guard against possible post hoc bias. And, of course,

such ratings would exist for all cases ever decided.

Although reality necessarily falls short of the ideal, using bench memos as a tool to

rigorously assess legal quality comes quite close to many aspects of a dream measure. In

short, bench memos are crafted by some of the best and brightest young lawyers in the

country (Ward and Weiden 2006; Peppers 2006). Generally, they all graduated in the top

of their class from elite law schools and have experience clerking for lower federal court

judges. Bench memos evaluate the strength of each argument, offer a summary of their

overall assessment, and report which side has a better case. All of this is done prior to oral

arguments based on the information provided in merits and amicus briefs.

I am using bench memos to asses case strength and while at the Supreme Court they serve

this purpose they are also used to expedite the decision making process. Between litigants’

merits briefs, reply briefs, supplemental briefs, and amicus briefs, the amount of information

for one case can become overwhelming and Supreme Court justices are busy. Think of the

briefs and other information provided by litigants in a case as a hot and boring baseball game

in mid-July. Bench memos act as the Sports Center highlight reel, providing a comprehensive

23

yet concise overview of the case, while prioritizing the most important information. Justice

Ruth Bader Ginsburg calls bench memos a “guided tour on the briefs,” because in addition

to breaking down the legal arguments and highlighting important information, they detail

which briefs to read, to skip, and to skim (Peppers 2006, 199).

Figure 2.1: Bench Memo Content

The structure and length of memos vary by justice, they all include some of the same

basic information. Figure 2.1 depicts the breakdown of Justice Blackmun’s memos, which

24

0.02.55.07.510.0SummaryFactsContentionsDiscussionConclusionQuestionsMiscellaneousSectionsAverage Section Length in Pagesaverage 29.5 pages.2 Each memo begins with a very brief, typically just over two pages,

summary of the case. The facts section details what has taken place to date, including

the lower court proceedings and the basic fact pattern of the case. As Figure 2.1 shows,

the contentions section occupies the largest amount of space in most memos, comprising

about 35 percent of their length (just over 10 pages). This section typically includes the

claims of each party and a discussion of their legal reasoning, as well as any unique or

valuable information offered by amici.

In Fulminante, for example, the State of Arizona

argued that the “totality of the circumstance” fail to demonstrate coercion and even if they

did the second confession to the informant’s wife was given voluntarily. Alternatively, the

respondent argues that the informant was in a position of power and therefore the situations

was inherently coercive. Following a description of the scope of the legal claims being made,

the discussion section provides a justification or rationale for their recommendation and how

the Court should arrive at their final decision. The final section, and key to this paper, is

the recommendation. This section is quite small relative to the total length of the memo

because it is typically a short paragraph recapping their reasoning with a simple dispositional

recommendation.

To demonstrate the benefit and efficiency of bench memos, consider Justice Blackmun’s

memos, which on average, contain 30 unique case citations. The contentions sections of his

bench memos mirror the argument section in both of the litigants’ briefs on the merits. The

contentions section is the largest section of the bench memo, and includes 48 case citations

on average (this includes repeated citations). Yet, this is less than half of the number of

citations in the argument section of a single litigant’s brief (Schoenherr and Black 2019).

In addition to consolidating important information for justices, bench memos have the

added benefit of also vetting information and legal claims. Sometimes attorneys can overstate

the relevance of a particular precedent to their case, or misstate the similarities (differences)
2This figure was created with all memos with tables of content from the 1972 - 1993

terms.

25

in an attempt to bolster their claims. Regardless if this is done purposefully or because of

lack of experience and knowledge, bench memos give clerks the opportunity to address and

correct these claims. For example, in McDonald v. City of West Branch, Michigan et al.

(1983), Blackmun’s clerk McIntosh, in a rather blunt manner, points out the issues with

the respondent’s primary argument and goes on to detail how an amicus is somewhat more

successful:

Resp[ondents]s’ attempt to distinguish Alexander and Barrentine is not so much
unpersuasive as unintelligible. The EEAC’s attempt is much more credible, but
not very compelling...3

Corrections like these, along with all of the factual content, citations, analysis, and

evaluations are encapsulated in a bench memo’s final dispositional recommendation. That

is to say the final recommendation is based on a myriad of legal considerations, and boils

them down to a single recommendation. As a result, this recommended disposition favors

the party that has a stronger legal position. Stated differently, the final recommendation

supports the litigant with a higher quality legal argument. Therefore, I contend that this

recommendation can be used as an indicator of legal quality.

2.3 Examining Case Strength

Before I can examine the influence of case strength on case outcomes, I must verify that

bench memo recommendations favor the litigant with a higher legal quality, or a stronger

legal argument. Bench memo data were collected by taking a photograph of each page of

every bench memo from Justice Harry Blackmun’s archives at the Library of Congress. These

photos were then processed using optical recognition software to convert the photographs to
3Bench memorandum in McDonald v. City of West Branch, Michigan et al. (83-219).
Located in the Supreme Court Case Files (Box 406, Folder #3) of Justice Harry A. Blackmun
in the Library of Congress, Washington D.C.

26

raw text files. Recommendations were extracted from the summary and/or recommendation

sections of each memo.

Not all recommendations in favor of the petitioner or respondent are created equal. Some

memos are consistent in their recommendation and will advise reversal in the summary

section and again at the end of the memo in the conclusion. Yet others are more wishy

washy, they will indicate that it is a “tough” or “close” case. Some memos will even provide

justification to support the petitioner and respondent because the law is unclear. In creating

a measure of case strength there is a theoretical justification to think that a consistent reversal

or affirmance recommendation is substantively different than an ambivalent recommendation.

That is, a petitioner’s legal argument or case strength can not be as strong if the respondent

could also win supported by solid legal reasoning. To account for this variation I extracted

the summary and recommendation sections from each bench memo and hand coded the

certainty surrounding the recommendation on a five-point scale. One represents a strong

affirmance and two represents a weak affirmance. Contrastingly, strong reversals take on

the value of five and weak reversals are coded as four. Three represents cases dealing with

questions that are so new and the law is underdeveloped, or the law is so ambiguous that
the clerk can not make a single, clear, recommendation.4,5

To ensure that clerks are not crafting bench memos based on what they think Justice

Blackmun wants to hear, I control for petitioner ideological compatibility. If bench memos

represent the legal merit and quality of a litigant’s position, they should not reflect Justice

Blackmun’s preferences. To estimate Blackmun’s preference in a case, I follow past work

4To see examples of the coding scheme please refer to the Appendix.
5Recommendations at the end of bench memos are very specific. Clerks will detail com-
ponents of a case to be affirmed, reversed, and remanded to the lower court. Yet, for use as
the dependent variable I do not have a theoretical reason to think that a reverse and remand
favoring the petitioner is different from a straight reversal also favoring the petitioner. A
reverse and remand does not signal uncertainty in the petitioner’s legal reasoning. Using
the caseDisposition variable from the Supreme Court Database, I follow past research and
collapse affirm and deny/dismiss as outcomes favoring the respondent and all other outcomes
as favoring the petitioner (there are no certifications outcomes in my data).

27

used to estimate the ideology of attorneys before the Court (Johnson, Wahlbeck and Spriggs

2006; Bailey, Kamoie and Maltzman 2005). First, I determine if the petitioner represents

the liberal or conservative position in a case by examining the ideological direction of the

lower court decision (Spaeth et al. 2018). Next, I use Martin and Quinn (2002) (MQ)

scores to determine Justice Blackmun’s ideology each term. Higher values indicate more

conservative preferences. During this period, 1979 and 1993, Blackmun had drifted more

liberal than his previous positions on the Court relative to his colleagues and his MQ score

ranged from -1.94 to -.01 (Epstein, Martin, Quinn and Segal 2007; Martin and Quinn 2002;

Owens and Wedeking 2012). Finally, if the lower court disposition is liberal, the petitioner

represents the conservative position in a case and this variable is coded as Blackmun’s MQ

score. Alternatively, if the lower court decision was conservative, the petitioner represents the

liberal position and this variable is coded as the negative value of his MQ score. If clerks are

in fact writing bench memos in accordance with their boss’s preferences, petitioner ideological

compatibility should be a significant and positive predictor of law clerk’s recommendations;

however, since I argue that bench memos are an assessment of case strength, I do not expect

this to have a significant coefficient.

The treatment of precedents is a primary component of assessing the strength of each

party’s legal argument (Hansford and Spriggs 2006). If a litigant is citing an overturned case

or a precedent that has been repeatedly negatively cited by the Court’s majority opinions,

their argument will not be strong. Yet if they are positively citing cases that the Court

has continuously upheld, their argument will be based on powerful precedent. To quantify

the treatment of precedent I rely on Hansford and Spriggs’s (2006) precedent vitality data

and Shepard’s Citations coding scheme.

I coded the treatment of the precedent used in

the petitioner and respondent’s subsections of the contentions section in the bench memo.

According to Shepard’s, negative treatments are cited as being “distinguished," “criticized,"

“limited," “questioned," and “overruled;" and a positive treatment applies to any precedent

28

that is “followed." Following this scheme, I assign -1 to all negatively treated cases and 1 to

positively treated cases. If the precedent is not legally interpreted or treated (i.e., included in

a string citation) it receives a 0 (Hansford and Spriggs 2006). Table 2.1 displays an example

of each type of treatment. Say, for example, these are all of the treatments for a petitioner’s

argument in a case. The treatment of each reference is multiplied by the vitality score for
the term it was referenced and then summed (e.g., (−1 ∗ 3) + (0 ∗ 1) + (1 ∗ 39) = 36). This
sum is the petitioner’s vitality score. This is done for the petitioner side and respondent side
of each case in the sample.6

Given that case strength is coded as petitioner relief, I expect petitioner vitality to be

positively related to legal quality and respondent vitality negatively related to legal quality.

When the petitioner has a stronger legal argument, they are more likely to be favored in

the final recommendation. That is, if the petitioner’s argument is supported by cases that

are being cited based on the way the Court has treated them in the past, the bench memo

recommendation should reflect this strength and favor the petitioner. Conversely, if the
6Precedent vitality has been validated by subsequent research (Black and Spriggs 2013;
Hansford, Spriggs and Stenger 2013), and therefore offers the best way to quantify legal
reasoning used by each party in the memos, however it does have some drawbacks. The
vitality dataset only includes orally argued cases decided by the Court dating back to 1946.
Therefore, any precedent cited in the bench memos disposed of without an oral argument
between 1946 and the bench memo term, decided prior to 1946, or not decided by the
Supreme Court, do not contribute to a litigant’s vitality score, regardless of treatment. I
coded a random sample of 80 of Justice Blackmun’s bench memos between 1979 and 1993,
which yielded 1,340 total case references. Of these, 667 are unique references. After merging
with the vitality data (and Supreme Court database, see Spaeth et al. (2018)) I am left with
1,182 total case references (88%) across 60 bench memos or cases. The random sample is
drawn from the 1979 - 1993 time frame due to the availability of other independent variables.
This small data loss is due to a lack of vitality score for at least one party to a case. That is,
all of the cases referenced or treated were (1) not orally argued before the Supreme Court,
(2) were orally argued prior to 1946, (3) were decided by a lower federal court or state
supreme court. While this data loss could create a bias, the loss of lower federal court and
state supreme court cases are inconsequential since the Supreme Court is not bound by thier
precedent. Additionally, Hansford and Spriggs (2006, 53) demonstrate that as precedents
age they are less likely to receive any treatment at all, therefore once more cases are coded,
data loss due to older cases being cited should be insignificant.

29

Case

Referenced
Oregon v.

Elstad

Maryland

v.

Louisiana

Swann v.
Charlotte-
Mecklenburg

Table 2.1: Precedent Vitality Coding

Reference
Context

This case is distinguishable from El-
stad for two reasons. First, the sec-
ond warnings were not adequate even
standing alone. Had Eagan been given
the standard Miranda warnings before
his second statement, this would be a
closer case. Second, considering the
two sets of warnings together, their
combined effect was inadequate.

The Act pre-empts any state actions
that impair either directly or indi-
rectly FERC’s power to regulate whole-
sale sales.
See, e.g., Maryland v.
Louisiana. 451 U.S. 725 (1981).

v.

Swann

In
Charlotte-
of Ed., 402
Mecklenburg Bd.
U.S. 1 (1971), the Court rejected the
argument that the equity powers of
federal district courts had been limited
by Title IV of the Civil Rights Act
of 1964, holding that there was no
suggestion in the statutory language of
an intention "to restrict those powers
or withdraw from courts their historic
equitable remedial powers." Id., at 17.
The same is true here.

Treatment Vitality

-1

0

1

3

1

39

TOTAL:

36

respondent’s argument is rooted in precedent with stronger vitality, I would expect the

recommendation to favor the respondent.

The majority of the information in bench memos comes from briefs, namely litigants’

briefs on the merits. It is important to consider how the attorneys that author these briefs

can influence legal quality. As McGuire (1995) demonstrates, litigation experience before

the Supreme Court significantly increases a party’s likelihood of success because experienced

attorneys are more likely to meet the justices’ informational needs. This is also reflected

in their certiorari success (Biskupic, Roberts and Shiffman 2014) and oral argument perfor-

30

mance (Johnson, Wahlbeck and Spriggs 2006), and brief writing skills (Black et al. 2015).

Their experience and immense resources make them more prepared and knowledgeable (Black

and Owens 2012a; Songer, Sheehan and Haire 1999), their arguments are viewed as more

credible (Johnson, Wahlbeck and Spriggs 2006; McGuire 1995), and their writing more con-

vincing (Black et al. 2015). For these reasons, I argue that experienced attorneys are better

able to situate their case in the corpus of existing case law in a common law system, which

establishes a stronger legal argument. To capture this experience, I use the number of previ-

ous appearances before the Supreme Court at oral argument for the petitioner and subtract

the previous oral argument experience for the respondent. I expect attorney experience to

be positively associated with legal quality. Positive values of this variable indicate that the

petitioner’s attorney has more experience, thus a higher quality legal argument.

I also include an indicator for amicus brief advantage - the difference in amicus briefs

filed on behalf of the petitioner and respondent. At the agenda setting stage, amicus briefs

convey the importance of the legal issues in a case (Caldeira and Wright 1988; Caldeira,

Wright and Zorn 1999; Black and Owens 2009). At the merits stage amicus briefs reiterate

litigants’ arguments, but also contribute unique information and legal reasoning (Spriggs

and Wahlbeck 1997a). What is more, justices find that amici make effective and sound legal

arguments and will borrow language from briefs for majority opinions (Collins, Corley and

Hamner 2015). Because of their ability to reinforce litigants’ arguments and offer unique

insight on legal issues, I argue that amicus briefs contribute to an increase in legal quality.

The amicus advantage variable subtracts the number of amicus briefs filed on behalf of the

respondent from the number of amicus briefs filed on behalf of the petitioner.

I expect

amicus advantage to be positively related to legal quality.

There are, of course, limitations to my findings, like past approaches to measuring law.

First, the random sample relied on to examine legal quality is limited. Second, I recognize

that any measurement is going to have potential drawbacks, and my estimate of case strength

31

is no different. For example, because law clerks author these memos, one might worry that

any biases possessed by a law clerk would also be folded into any measure that uses bench

memos as a data source. There are two variants to this concern. First, if a law clerk has

his own agenda that he seeks to push upon his justice, then this will undoubtedly appear

in a bench memo, which would be a key tool used to influence said justice. Second, and at

the opposite end of the spectrum, if a law clerk is fully loyal to his employing justice, then

the bench memo will be slanted to whatever ideological biases that justice has toward the

subject matter.

Ultimately, I do not believe either version of this concern sinks my approach. Although it

is undoubtedly true that law clerks have preferences about how cases are decided, it is quite

different to say that they systematically try to sway how their justice votes at every single

opportunity. The vast majority of Supreme Court law clerks have been vetted not just by a

justice’s trusted advisors (including former law clerks and the justice him or herself), but also

generally by a lower circuit judge, for whom they also clerked. To wit, justices are more likely

to pick clerks who have previously clerked for a lower court judge who is ideologically aligned

with themselves (Ditslear and Baum 2001). And, the risks to such behavior are significant.

Alienating one’s justice would likely undermine the substantial professional benefits a former

clerk stands to gain (e.g., lucrative private practice, opportunities in academia, etc.). Finally,

as an objective manner, recent research utilizing campaign contributions reveals that many

law clerks are not particularly extreme in their politics (Bonica et al. 2016). Additionally,

clerks are very transparent and indicate when their personal preferences pull the in one

direction but the law says otherwise, and often provide a recommendation supporting their

personal preferences and another they believe is correct according to the law.

32

2.3.1 Methodology and Empirical Results

To estimate this model, I employ an ordered logistic regression model. This model is

appropriate given the ordinal nature of the dependent variable. I fail to reject the null of

the Brant Test, which means that the parallel regression assumption holds and an ordered

logistic regression is preferred to a multinomial logistic regression. The results are shown in

Table 2.2 below.

Petitioner vitality is in the positive expected direction and significant. This suggests

that the treatment of precedent by the petitioner is an important factor in determining

case strength. The petitioner must cite cases according to their vitality in order to achieve

an advantage. To better understand this relationship the predicted probabilities are shown

in the left panel of Figure 2.2. The x-axis represents petitioner vitality. Negative values

suggest the petitioner is positively citing precedent that is outdated or has been overturned,

or he is negatively citing vital cases. Larger positive values suggest that the petitioner is

citing precedent accurately according to the Court’s recent decisions. A one unit increase

of petitioner vitality leads to an .33 percentage point increase in the likelihood of receiving
a strong reversal. As petitioner vitality increases from the minimum value of −16 to the
maximum value of 55 the likelihood of receiving a strong reversal recommendation increases
by 24 percentage points or a 56 percent change.7 Petitioner vitality has a slightly negative

relationship on the likelihood of receiving a weak reversal recommendation. This may be

due to the lack of small sample size, or because the law weakly favors the petitioner’s legal

argument but this is not represented by existing precedent but instead by other types of law.
7Figure 2.2 shows that even if the petitioner positively cites “bad law” he has a nearly
50% chance of success. This is because the Supreme Court is reversal prone and reverses
two-thirds of the cases that it hears (Epstein et al. 2019).

33

Table 2.2: Ordered Logistic Regression of Case Strength

Petitioner Vitality

Respondent Vitality

Petitioner Experience Advantage

Petitioner Amicus Advantage

Lower Court Disagreement

Petitioner Ideological Compatibility

Strong Affirmance | Weak Affirmance
Weak Affirmance | Uncertain
Uncertain | Weak Reversal
Weak Reversal | Strong Affirmance

Observations
AIC
Residual Deviance
Standard errors in parentheses
* denotes p< 0.05 for two-tailed test

0.0144∗
(0.008)
−0.003
(0.005)
0.030∗
(0.013)
0.154
(0.110)
−0.551
(0.558)
0.489
(0.288)
−1.642
(0.769)
−0.948
(0.747)
−0.796
(0.744)
0.057
(0.739)
65
187.419
167.419

As expected, there is a negative slope on weak and strong affirm recommendations sug-

gesting that as the petitioner cites more vital cases positively or non-vital cases negatively

the likelihood of the respondent receiving a favorable recommendation decreases. There is

no relationship between uncertainty and petitioner vitality.

In these cases the law is so

close that the law clerk can not make a sound recommendation, which is likely reflected the
citation patterns of precedent supporting each litigant.8

8This is the least common type of recommendation and in my small sample there is only

a single occurrence of uncertainty.

34

Figure 2.2: Likelihood of Case Strength Across Petitioner Vitality - The left panel displays
the likelihood of case strength operationalized as a the five category scale of bench memo recommen-
dations across the 95th percentile of petitioner vitality while holding all other variables at their mean
and the dichotomous variable at its mode. The right panel depicts the likelihood of case strength
across the 95th percentile of petitioner experience advantage while holding all other variables at
their mean and the dichotomous variable at its mode.

To the right of the petitioner vitality panel in Figure 2.2 are the predicted values for

the other significant predictor of case strength, petitioner experience advantage. Negative

values on the x-axis represent that the respondent’s attorney has more experience before

the Supreme Court than the petitioner’s attorney. A value of zero would suggest that the

attorneys are equally matched, and positive values mean that the petitioner’s attorney has

argued more cases before the Supreme Court than the respondent. As expected, as the peti-

tioner’s attorney gains an experiential advantage the more likely the petitioner has a stronger

case as represented by the dotted green line representing a strong reversal recommendation.

More specifically, moving from the most disadvantaged to the most advantaged experienced

35

0.00.10.20.30.40.50.60.70.80.91.0−16−641424344454Petitioner VitalityLikelihood of Recommendation0.00.10.20.30.40.50.60.70.80.91.0−19−67203346Petitioner Experience AdvantageLikelihood of RecommendationStrong AffirmWeak AffirmUncertainWeak ReverseStrong Reverseattorney arguing for the petitioner increases the likelihood of a strong case (strong reversal

recommendation) by 44 percentage points, a 129 percent change. As expected, there is also

a negative slope for strong and weak affirm recommendations. As the petitioner’s attorney

gains an experience advantage the respondent is less likely to have a stronger case. Moving

from the most advantaged respondent to the most advantaged petitioner the likelihood of a

strong affirm recommendation decreases from 27% to 5%, a 22 percentage point decrease, or

a 82 percent change.

Returning to the table, respondent vitality is in the expected direction but fails to have

a significant effect on case strength. This may be due to the fact that the petitioner has the

first mover advantage. They get to choose how to frame the issue in the case (Wedeking

2010) and therefore also select the relevant case law to cite. The respondent is then forced

to respond in ways that do not yield as vital of an argument.

Also in the expected direction, amicus advantage does not predict of legal quality. This

may be due to the heterogeneity of parties that file amicus briefs, where quantity does

not necessarily equal quality (Box-Steffensmeier, Christenson and Hitt 2013). While some

amicus groups are large with extensive resources to craft well researched briefs, others are

not. This may lead to diverse treatment or trust of this information, as some of it likely

contributes to legal quality, while other amicus groups do not.

Lastly, the control for Justice Blackmun’s ideology is indistinguishable from zero. This

indicates that bench memos are not a simple reflection of Blackmun’s policy preferences.

This coupled with the other results of this model suggest that bench memos do in fact

capture what I theorize to be case strength.

I can now use this to understand how case

strength impacts Supreme Court decisions. Do justices consider case strength when placing

their votes; do litigants with stronger legal arguments in have a higher chance of success

before the Court; and can justices who are predisposed to vote against them be swayed by

the strength of their argument?

36

2.4 Policy vs. Preferences

To determine if justices’ votes are influenced by law and the quality of a litigant’s legal

argument I examine a sample 2,927 individual justice-votes across 754 cases argued between
1972-1993 terms.9 My dependent variable is policy vote, a dichotomous indicator that takes

the value of 1 when a justice votes for the litigant he is ideologically aligned with and 0 when

he places a counter-policy vote. This variable was coded using the Supreme Court Database

(Epstein et al. 2019). If the lower court decision was liberal and a conservative justice votes

to reverse, the dependent variable is coded as 1 because he voted in favor of a petitioner

who is arguing a conservative position.

If the same conservative justice voted to affirm,

the dependent variable is coded as 0. Similarly, if the lower court ruling is conservative

and a liberal justice votes to reverse the dependent variable takes on a value of 1, and

if he votes to affirm it takes on the value of 0.

I follow Johnson, Wahlbeck and Spriggs

(2006) and exclude Blackmun’s votes from the dependent variable because of the reliance on

his archival materials to develop the indicator of legal quality. Justice Blackmun, and only

Justice Blackmun, read the bench memos his clerks wrote, which in turn, could also influence

his final vote. I aim to prevent any endogeneity by eliminating Blackmun from the model.

It will also make it more difficult to find an effect for law in the analysis, yet the presence of

a significant effect lends credence to the fact that bench memos are actually capturing the

litigant with the superior legal argument.

My primary independent variable is law-policy conflict. My data and measure of case

strength allow me to overcome the hurdle of observational equivalence and tease out what

suggests should be the final outcome from the ruling a justice is predisposed to favor. Since

the dependent variable is whether a justice placed a policy vote I recoded my five point

ordinal measure of case strength to a five point ordinal measure of law-policy conflict. This
9I merge my measure with Johnson, Wahlbeck and Spriggs’s (2006) data to obtain certain

independent variables.

37

variable takes on a value of 2 when law or the final recommendation at the end of a bench

memo is strong (affirmance or reversal) and conflicts with a justice’s ideological preferences.

For a conservative justice, this conflict occurs when the lower court decision was conservative

and the law strongly supports a reversal, as was the case for Scalia in Fulminante, or if the

lower court made a liberal decision and the law strongly calls for an affirmance. This variable

takes on a value of 1 under the same circumstances but when the law is not as strong (weak

reversal or affirmance). When law is unclear there is not necessarily a conflict between the

law and policy preferences so this variable is coded as 0. Alternatively, this variable is coded

as -2 when law and policy converge and a justice is free to vote in line with his attitudes. This

occurs for a conservative justice when the lower court rules conservatively and law strongly

suggests an affirmance is in order; or when the lower court hands down a liberal ruling and

law strongly favors a reversal. When the law moderately or weakly favors an outcome that

aligns with a justice’s policy preferences law-policy conflict is coded as -1. Similar to the

agenda stage (Caldeira, Wright and Zorn 1999; Black and Owens 2009), I expect justices

to be constrained when law and policy goals diverge and as a result, less likely to follow

their preferences. When law and policy coincide, justices are free to vote as they desire with

superior law and legal reasoning to support their choice.

Justices’ ideological preferences are obviously consequential in their likelihood of placing

a policy vote. Specifically, I expect that the ideological extremity of a justice’s preferences

influences the ability of law to act as a constraint. That is, justices with more entrenched

ideological views, like Scalia or Douglas, are more likely to vote with their preferences,

compared to their moderate colleagues, like O’Connor or Kennedy, the perennial “swing

votes” on the Court (Martin, Quinn and Epstein 2004; Dwyer 2018). To account for this

relationship I include ideological extremity coded as the absolute value of a justice’s Martin

and Quinn (2002) score. I also include a multiplicative term between ideological extremity

and my primary independent variable of interest, law policy conflict. This is because justices

38

who are ideologically extreme will be less easily swayed by law when it conflicts with their

preferences.

In addition to examining which justices are constrained by law, I investigate if certain

cases allow this relationship to persist. Justices are influenced by the public opinion, and alter

their behavior in salient cases when they know the public is watching (Unah and Hancock

2006; Wahlbeck 2006; Lax and Cameron 2007; Casillas, Enns and Wohlfarth 2011a; Black,

Sorenson and Johnson 2013). There are a number of methods to measure the salience of a

case (Epstein and Segal 2000; Collins and Cooper 2012; Black, Sorenson and Johnson 2013;

Clark, Lax and Rice 2015), theoretically Clark, Lax and Rice (2015) is the best option for

this model. Their measure estimates salience prior to the Supreme Court hearing a case for

oral argument. While Epstein and Segal (2000) and Collins and Cooper (2012) are both

valid and reliable measures, they measure salience based on newspaper coverage of the final
opinion, and are therefore are endogenous to what I am trying to measure.10 Clark, Lax and

Rice (2015) latent trait of salience based on coverage in the New York Times, Washington

Post, and Los Angeles Times is included in the model and also interacted with law-policy

conflict. As past research demonstrates, justices more likely to vote their preferences in

salient, often controversial cases, and therefore less likely to be impacted by law.

10Black, Sorenson and Johnson’s (2013) measure would be ideal because it captures a

justice-level measure case salience, but it is not available until the 1979 term.

39

Table 2.3: Control Variable Measurement

Variable
OSG Policy Match

Measurement
1 if the OSG argues in the direction of a justice’s policy
preferences, 0 otherwise

OSG Policy Mismatch

-1 if the OSG argues in the direction of a justice’s policy
preferences, 0 otherwise

U.S. Government Policy
Match

1 if the U.S. Government argues in the direction of a
justice’s policy preferences, 0 otherwise

U.S. Government Policy
Mismatch

-1 if the U.S. Government argues in the direction of a
justice’s policy preferences, 0 otherwise

Case Complexity

Case Complexity × Law-
Policy Mismatch

Factor analysis of the number of legal provisions in a
case and the number of issues involved according to
the Supreme Court Database (Johnson, Wahlbeck and
Spriggs 2006)

Factor analysis of the number of legal provisions in a
case and the number of issues involved according to the
Supreme Court Database × Law-Policy Conflict

Reversal Prone

1 if justice prefers to side with the petitioner, 0 otherwise

Resource Advantage of Pre-
ferred Litigant

Resource advantage of the litigant a justice prefers to
side with − resource advantage of opponent (Collins Jr
2004, 2007)

Amicus Advantage of Pre-
ferred Litigant

Number of amicus briefs filed for the litigant a justice
prefers − number of amicus briefs filed for their oppo-
nent (Collins 2008)

I also include a host of other control variables in the model. The variables and their

measurement are listed in Table 2.3. I control for whether the Office of the Solicitor General

(OSG) is arguing for or against a justice’s preferences. The OSG, often referred to as the

“Tenth Justice,” is extremely successful before the Court and has the ability to garner votes

from ideologically distant justices (Bailey, Kamoie and Maltzman 2005; Wohlfarth 2009;

Black and Owens 2012b). Generally speaking, the U.S. government is successful before the

Court (Johnson, Wahlbeck and Spriggs 2006), so I also control for whether a federal attorney

is arguing for or against a justice’s preferences. Case complexity may impact a justice’s

40

likelihood to place a policy vote. The law in complex cases covering a host of issues or legal

questions may be difficult to navigate, so instead of trying to, justices may revert to their

underlying preferences. I also interact this with law-policy conflict expect that even when

law and policy conflict, justices will increasingly vote with their preferences case grows in

complexity. Next, I control for amicus and litigant resource advantages. Litigants with more

resources can hire more skilled attorneys to write better briefs, and argue a stronger case

before the Court (McGuire 1995; Songer, Sheehan and Haire 1999; Johnson, Wahlbeck and

Spriggs 2006), which in return makes it easier for a justice to rule in their favor. Similarly,

amicus briefs offer unique perspectives and legal arguments to a case (Hansford 2004) and

influence justices’ final votes (Collins 2008), so this provides alternative avenues to rule for

a party that may not be included in their brief. Lastly, I control for the fact that justices

are reversal prone given that two-thirds of the cases that come before the Supreme Court

are reversed (Epstein et al. 2019).

2.4.1 Methodology and Empirical Results

My dependent variable, the likelihood of a policy vote, is dichotomous so I employ a

logistic regression. The data are hierarchical with case level observations nested within each

justice, with significant variation at level two. According to Clark and Linzer (2014), “random

effects produce superior coefficient estimates when there are few units or observations per

unit” (407). Given that for some justices the total number of observations is low, I follow

Clark and Linzer (2014) and employ random effects.

The results of the logistic regression model with justice-level random effects on justices’

likelihood of a policy-vote are shown in Table 3.2. Because of the non-linear nature of the

model, I focus on the predicted values to address the results rather than strictly focusing on

the coefficients. I begin with Figure 2.3, which addresses the likelihood of a policy vote across

ideological extremism dependent on whether or not the law supports a justice’s preferences.

41

Table 2.4: Logistic Regression Estimates with Justice-Level Random Effects

Law-Policy Conflict

Ideological Extremity
Ideological Extremity × Law-Policy Conflict

Early Salience
Early Salience × Law-Policy Conflict

OSG Policy Match

OSG Policy Conflict

U.S. Government Policy Match

U.S. Government Policy Conflict

Case Complexity
Case Complexity × Law-Policy Conflict

Reversal Prone

Resource Advantage of Preferred Litigant

Amicus Advantage of Preferred Litigant

Constant

σ

ρ

Observations
AIC
BIC
Log Likelihood
Standard errors in parentheses
* denotes p< 0.05 for two-tailed test

−0.306∗
(0.039)
0.080
(0.056)
0.018
(0.015)
0.089
(0.066)
−0.036
(0.038)
0.284∗
(0.144)
−0.177
(0.146)
0.021
(0.151)
0.630∗
(0.144)
0.031
(0.106)
0.048
(0.065)
0.453∗
(0.096)
0.048∗
(0.010)
0.017
(0.020)
0.317
(0.188)
0.382
(0.103)
0.043
(0.022)

2927
3486.211
3581.919
−1727.105

42

Figure 2.3: Likelihood of a Policy Vote Across Ideological Extremity - This graph shows the
likelihood of a policy vote across the 95th percentile of ideological extremity. The dashed green line
shows the likelihood of a policy vote when law and policy align. The dashed purple line displays the
likelihood of a policy vote when the law is uncertain. The dashed orange line represents the likelihood
of a policy vote when law and a justice’s policy preferences come into conflict. The matching solid
lines represent 95% confidence intervals. All variables are held at their medians.

I theorized that when law and policy conflict justices should be less likely to follow their

preferences and that this effect should be stronger on ideologically moderate justices. The

data bear this out.
significantly less likely to vote their preferences. This is shown in Figure 2.3.11

I find that when law and ideology are in opposition, all justices are

Despite overlapping confidence intervals, all three lines are statistically different from one
11To aid with data visualization I only include strong law-policy agreement/conflict and

uncertain law.

43

0.40.50.60.70.80.90.00.51.01.52.02.53.03.54.04.5Ideological ExtremityLikelihood of a Policy VoteStrong Law−Policy AgreementStrong Law−Policy ConflictUncertain Lawanother. Additionally, uncertain law and strong law-policy conflict slopes are statistically

different from strong law-policy agreement. As shown by the gentle slope of the green line,

when law and policy are in agreement all justices are significantly more likely to vote with

their policy preferences regardless of their ideological extremity. This differs from cases where

law is ambiguous and in cases when law and policy are in disagreement where there is an

upward slope suggesting that moderate justices are less likely to vote with their preferences

than ideologically extreme justices. When law and policy conflict, a moderate justice has

a coin-flip chance of following his ideology over law, 54%, however when law and ideology

work together, this increases to 77%. This is a 23 percentage point increase or a 43% change.

Even a justice with the most entrenched beliefs only maintains a 62% chance of voting with

his preferences when law and policy diverge. The likelihood of a policy vote increases to 79%

when law and policy agree, a 17 percentage point increase.

Figure 2.4 displays the two other variables interacted with case strength. The left panel of

exhibits the results of the relationship between complexity and case strength. I hypothesized

that as complexity increases the likelihood for a policy vote should also increase. I do not

find support for this hypothesis. When law and policy conflict there is a slight increase

across the range of case complexity, which is consistent with my hypothesis, however this

increase is not statistically significant. What is more, when law and policy are in agreement

there is a slight negative slope implying that despite the fact that law supports a justice’s

preferences his likelihood of voting with these preferences decreases as a case grown more

complex. This relationship, however, is not statistically significant.

44

Figure 2.4: Likelihood of a Policy Vote Across Case Complexity and Case Salience -
The left panel exhibits the results of the relationship between complexity and case strength and the
right panel shows the relationship between case salience and case strength. The dashed green lines
show the likelihood of a policy vote when law and policy align. The dashed purple lines display the
likelihood of a policy vote when the law is uncertain. The dashed orange lines represent the likelihood
of a policy vote when law and a justice’s policy preferences come into conflict. The matching solid
lines represent 95% confidence intervals. All variables are held at their medians.

Similarly, I do not find support for my case salience hypotheses. The coefficient on

salience is positive as expected, but it does not have a significant effect on the likelihood of

policy vote, regardless of the position of the law. The right panels of Figure 2.4 displays

the relationship between case salience and, law-policy conflict, and the likelihood of a policy

vote. As salience increases there is no significant increase in the likelihood of a policy vote

regardless of legal constraint. The only differences that persist are between each level of

conflict or agreement across case salience.

I do find that when a justice’s preferred litigant gains a greater resource advantage over

45

0.40.50.60.70.80.9−0.50−0.250.000.250.500.751.00Case ComplexityLikelihood of a Policy Vote0.50.60.70.80.9−0.7−0.10.51.11.72.32.9Case SalienceLikelihood of a Policy VoteStrong Law−Policy AgreementStrong Law−Policy ConflictUncertaintytheir opposition that justice’s likelihood of placing a policy vote significantly increases. Ad-

ditionally, I find that when the Solicitor General argues in-line with a justice’s policy pref-

erences the likelihood of a policy vote significantly increases. Yet peculiarly, when the U.S.

government argues against a justice’s policy preferences he is more likely to vote with his

preferences.

2.5 Discussion and Conclusion

In this paper, I set out to create a new indicator of legal quality that could be used

systematically across all cases that span multiple terms regardless of issue area. I found this

estimate to be related to other indicators of quality and moved on to determine if legal quality

impacts case outcomes. Importantly, I demonstrate the ability to overcome the long faced

challenge of observational equivalence by disentangling a legal-vote from a policy-based vote

and find that law can significantly constrain a justice’s propensity to vote his true preference

in a case, but this constraining effect is conditional on how polarized a justice’s attitudes

are.

My findings contribute to the field’s growing understanding that law and legal consider-

ations impact justices’ decision making (Bartels 2009; Bartels and O’Geen 2014; Bailey and

Maltzman 2008; Black and Owens 2009; Hansford and Spriggs 2006; Richards and Kritzer

2002). This is a normatively important topic. As research has shown, most people are legal

realists and recognize that justice’s preferences influence their decision making, yet they still

believe that justices make principled decisions and are not simply legislators in robes (Gibson

et al. 2011). These findings demonstrate that justices’ decisions, while influenced by their

preferences, are principled and rely on law and legal reasoning.

As for the concern about clerks who have completely integrated their justice’s thinking

into their work, this version is, to be fair, more plausible. That being said, if this were true,

these recommendations would not be able to predict other justices’ votes on the merits, and

they do. As shown by the case strength model in Table 3.2, case strength is a significant

46

predictor of justices’ votes to reverse.

Despite these challenges, bench memos offer a new and unique opportunity to use the

very same material justices used to evaluate cases. This systematic indicator of legal quality

has many applications beyond the way it was used in this chapter, as I will demonstrate in

Chapter 3.

47

CHAPTER 3

THE IMPACT OF CASE STRENGTH ON THE SUPREME COURT
JUDICIAL DECISION MAKING PROCESS, A REEXAMINATION

There are a host of factors that researchers theorize to influence Supreme Court justices’

decision making: policy preferences, personality traits, public opinion, and legal factors. A

definitive consensus agrees that policy preferences are the primary determinant of justices’

decision making. Researchers have reached this consensus by using systematic and reliable

measures like Judicial Common Space (JCS) (Bailey 2007; Epstein, Martin, Segal and West-

erland 2007) and Martin and Quinn scores (Martin and Quinn 2002). Measures also exist to

reliably capture personality (Black et al. 2019; Hall 2018), and public opinion’s relationship

with the Supreme Court (Casillas, Enns and Wohlfarth 2011b; Durr, Martin and Wolbrecht

2000). Legal factors however, such as declaring a law unconstitutional at the lower court,

a dissent at the lower court, or the position of the Office of the Solicitor General are typi-

cally used as a proxies for law, or in some studies, legal influences are simply ignored. This

was the accepted practice for decades because judicial politics scholars lacked a measure

that captured the contemporaneous influence of law on judicial decision-making. America’s

legal system is complex. It consists of rules, statutes, orders, and precedents from numer-

ous branches and levels of government. To incorporate these numerous factors into a single

measure is an arduous task.

In Chapter 1 of this dissertation, I establish a measure of case strength, which provides

a contemporaneous and systematic measure of law. It is able to capture all of the relevant

discrete, yet related components of law because it originates from bench memoranda. These

memos are crafted by Supreme Court law clerks to summarize all relevant case facts, analyze

the relevant case law cited in litigant and amicus briefs, and ultimately choose which party

has the strongest case, or superior legal argument. This measure provides a method for

48

researchers to control for and examine how law influences or alters important relationships

in judicial politics across the majority of orally argued cases for over twenty terms of the

Supreme Court.

With this new measure it is also crucial to reexamine existing theories. It can be used to

reevaluate past research when a measure of law was unavailable and confirm that existing

conclusions hold in light of including a measure of law. In fact, some scholars have even called

for the incorporation of the quality or strength of a litigant’s legal arguments because they

suspect it to be a significant confounder in judicial politics research (Budziak and Lempert

2015). They were unable to test this theory, however, because a measure of legal quality did

not exist at the time they conducted their research.

I am not arguing that the incorporation of a contemporaneous measure of law into past

and future models of judicial behavior will completely change our understanding of how

Supreme Court justices make their decisions. Research is definitive that justices’ policy

preferences are the primary determinant of their behavior, but that is not to say that law

does not matter. The discipline is removed from the time when judges were simply viewed

as legislators in robes. Many scholars have shown the significance that different aspects

of law have on judicial decision making (Bailey and Maltzman 2008; Bartels and O’Geen

2014; Hansford and Spriggs 2006; Richards and Kritzer 2002). Yet, those studies focused

on demonstrating law’s impact, they did not extend their applications to other aspects of

the decision making process that law also effects. This was partially because they did not

have the ability to do so, they were restricted to a certain type of case, a particular area

of law, or the work would be too arduous to actually achieve. With this measure we can

now examine a multitude of relationships and control for or examine the effect of law like

we would ideology.

In what follows, I reevaluate two pieces of past research that examine different stages

of the Supreme Court decision-making process to determine how or if law impacts this

49

process. First, I reexamine Johnson, Wahlbeck and Spriggs (2006) . The authors originally

find that attorneys’ oral argument performance is positively related to their success on the

merits.

I find that this relationship still persists but I also find that a litigant with an

abysmal oral argument performance can still obtain a positive outcome before the Court if

law supports their legal argument over their opponents. Next, I examine how law impacts the

opinion bargaining process by reevaluating work by Maltzman, Spriggs and Wahlbeck (2000).

In their analysis of strategic response to majority opinion drafts the authors found that

justices are more likely to bargain when they are ideologically distant from the opinion author

and more likely to write or or join a separate opinion, concurrence, or dissent if they are

ideologically distant from the majority coalition. I find that law does influence this process.

The strength at which the majority’s opinion is supported by law is negatively related to a

justice’s likelihood to respond with a statement of joining or writing a concurrence.

3.1 The Influence of Oral Arguments on Case Outcomes

The Supreme Court is known for its secrecy. For many years the Court was a black box

- cases would go in and decisions would come out but the in between as to how the justices

arrived at these final decisions was unknown. Oral arguments were the peephole into this

black box. They provided a glimpse into the Supreme Court decision-making process that

gained the attention of journalists, Court watchers, and the public alike. Unlike the general

public, as the field of judicial politics gained understanding of the Court, particularly by

the aid of archival materials, researchers seemingly ignored oral arguments. Admittedly

the justices themselves claimed that oral arguments were insignificant. Justice John Harlan

referred to them as “little more than a traditionally tolerated part of the appellate process”

(Harlan 1955; Roberts Jr 2005) and Justice Antonin Scalia called them a “dog and pony

show” (Black, Johnson and Wedeking 2012).

The conventional wisdom that oral arguments were insignificant held for decades because

researchers lacked a way to assess the independent impact of these proceeding until Johnson,

50

Wahlbeck and Spriggs (2006) collected Justice Harry Blackmun’s oral argument notes, which

included individual grades for each attorney that argued before the Court. They demon-

strated that these grades are a reliable and valid measure of oral argument performance and

then went on to show that attorneys’ oral argument performance impacts the justices final

decisions. They showed that the conventional wisdom was incorrect. Even the justices who

claimed oral arguments were ineffectual, were in fact influenced by them.

Despite this research and progeny research (Black, Johnson and Wedeking 2012; Black,

Sorenson and Johnson 2013; Gleason 2019; Jacobi and Schweers 2017; Johnson and Spriggs

2007; Johnson et al. 2009; Johnson, Black and Wedeking 2009; Ringsmuth, Bryan and John-

son 2013; Malphurs 2010) the finding that attorneys’ oral argument performance significantly

impacts case outcomes is still met with skepticism. In 2015, Budziak and Lempert published

an article introducing simultaneous sensitivity analysis to the political science literature.

They applied this method to Johnson, Wahlbeck and Spriggs’s (2006) oral argument work

to demonstrate that the relationship between oral argument performance and case outcomes

is sensitive to a confounder. They postulate the missing confounder is “legal quality.” They

define legal quality as the “party with the more viable legal position,” and claim, “a party

who has the better oral argument is more likely to be defending a position of higher legal

quality as reflected in records, briefs, and memoranda” (Budziak and Lempert 2015, 13,

17-18). Yet, all the authors are able to do is theorize because at that time they did not have

a measure that would adequately capture all of the legal factors to determine which party

had a stronger legal argument. Their desired “legal quality” confounder is analogous to my

measure of case strength, which I will use to reevaluate Johnson, Wahlbeck and Spriggs’s

(2006) findings and test their confounder hypothesis.

51

3.1.1 Data and Measures

To test the relationship case strength exerts on Supreme Court case outcomes in lieu

of attorneys’ oral argument performances, my dependent variable is a justice’s propensity
to reverse a lower court’s decision between the 1972 - 1993 terms.1

If the Court voted

to reverse, this variable is coded as 1, and 0 if they voted to affirm (Epstein et al. 2019).

Justice Blackmun’s votes are excluded from the dependent variable and analysis in general

because of the reliance on his archives for both oral argument grades and case strength.

Johnson and his colleagues excluded Blackmun’s votes from their original analysis to avoid

any endogeneity that might arise. Justice Blackmun could feasibly assign higher grades to

attorneys he anticipated voting for. Similarly, Justice Blackmun, and only Justice Blackmun,

read the bench memos his clerks wrote, which in turn, could also influence his final vote.

By eliminating Blackmun from the model to prevent this endogeneity issue, it will make it

more difficult to find an effect for case strength. Yet, the presence of a significant effect lends

credence to the fact that bench memos are actually capturing case strength and not simply

Justice Blackmun’s preferences.

Johnson, Wahlbeck and Spriggs’s (2006) primary independent variable is oral argument

grades. Justice Blackmun recorded grades in his oral argument notes, sometimes accompa-

nied with a brief description of the attorney. These grades were found to be related to other

indicators of attorney quality like litigating experience and elite law school attendance. This

variable takes both attorneys’ oral argument grades and subtracts the respondent’s from the

petitioners. Therefore, larger values on this variable suggest that the petitioner performed

better at oral argument. Based on their past findings, oral argument grades should be posi-

tively related to case outcomes, but according to Budziak and Lempert (2015), there should
1Johnson, Wahlbeck and Spriggs’s (2006) original analysis included the 1970 - 1993 terms.
For this replication I only used cases for which a measure of case strength is available, which
excludes the 1970 and 1971 terms. Blackmun’s bench memos early in his career were evolving
and inconsistent, therefore to incorporate legal quality, this analysis examines 1972 - 1993.

52

be no relationship after accounting for case strength.

The authors also control for justices’ ideological compatibility with attorneys using Mar-

tin and Quinn (MQ) scores (Martin and Quinn 2002). Generally, negative values of MQ

scores represent liberal voting behavior and positive values represent a conservative voting

history. If the attorney argues for the liberal position, determined by the direction of the

lower court’s decision, this variable is the negative value of the justice’s MQ score, and if the

attorney makes a conservative argument a justice’s MQ score is used so that higher values

on this variable indicate higher ideological compatibility. Ideological compatibility and oral

argument grades are also interacted in the model. Johnson, Wahlbeck and Spriggs’s (2006)

final results reveal that even if a justice is ideologically distant from an attorney’s position

in a case, a strong oral argument performance can still increase their likelihood of receiving

a favorable vote.

The original model controls for case complexity. Complexity is the result of a factor

analysis of the number of legal provisions in a case. The authors also interact complexity

with oral argument grades with the expectation that oral arguments may be increasingly

important in complex cases. Litigants have limited space to make their arguments in briefs,

and if a case is exceedingly complex, they may need to clarify content that was unclear or

absent from their brief. Therefore, an attorney’s oral argument performance would reflect

their ability to answer the justices’ questions. The additional control variables included are

listed in Table 3.1. For all but the last variable listed, there is an appellant and appellee

version of the variable included in the model.

Additionally, I include case strength in the model. Just as its coding as a dependent

variable in the first chapter, this is an ordinal variable coded as 5 if the law strongly supports

the petitioner and overturning the lower court’s ruling, 4 the law also supports a reversal but

the law clerk was less certain in his recommendation, 3 if the case could go either way, 2 if the

law supports an affirmance but there is some hesitation or uncertainty, and 1 if an affirmance

53

is strongly supported. If the petitioner’s contentions are more supported by precedent and

other aspects of the law, their likelihood for success should exceed the respondent’s.

I also include two interaction terms including case strength. First, I interact case strength

and oral argument grade. Budziak and Lempert (2015) suggest that oral argument perfor-

mance is sensitive to, or may be conditioned by, case strength and that the litigant with legal

reasoning more firmly grounded in precedent has an easier time explaining their argument

in briefs, and finds it easier to justify his claims before the justices at oral arguments. This

suggests that oral argument performance should matter less for litigants with superior legal

quality because they will already have an advantage. I also interact legal quality with ideo-

logical compatibility. Past research demonstrates that the law matters (Richards and Kritzer

2002; Bailey and Maltzman 2008; Hansford and Spriggs 2006), and that legal factors can

alter a justice’s behavior regardless of his ideological preferences (Black and Owens 2009).

I believe that legal quality can condition a justice’s policy preferences. That is, a litigant

with a superior legal argument is more likely to receive a justice’s vote, regardless of their

ideological compatibility.

54

Table 3.1: Oral Argument Control Variable Measurement

Variable
U.S. Attorney

Solicitor General

Measurement
1 if non-Solicitor General U.S. attorney argues the case
for the appellant/appellee.

1 if the Assistant Solicitor General or actual Solicitor
General argues the case for the appellant/appellee.

Washington Elite Attorney

1 if non-federal government attorney has a Washington
DC address argues for the appellant/appellee.

Law Professor Attorney

1 if attorney is listed as a law school professor in the
Martindale Hubbell directory in the year the year the
case was argued or has a university address argues for
the appellant/appellee.

Former Clerk Attorney

1 if attorney served as a clerk on the Supreme Court be-
tween 1921 and 1991 argues for the appellant/appellee.

Elite Law School Attorney

1 if the attorney attended an elite law school (Harvard,
Yale, Columbia, Stanford, Chicago, Berkleley, Michigan,
and Northwestern) argues for the appellant/appellee.

Difference in Litigating Ex-
perience

Log of the difference between petitioner and respon-
dent’s litigating experience.

3.1.2 Method and Results

The dependent variable is dichotomous, reverse or affirm.

I use a logistic regression

model with robust standard errors clustered on justice (Johnson, Wahlbeck and Spriggs

2006). Robust standard errors are included because I am modeling individual justice votes

across multiple cases, which may lead to correlated errors by justice.

The results of the logistic regression are shown in Table 3.2. The middle column of the
table displays Johnson, Wahlbeck and Spriggs’s (2006) original model.2 The right column

displays the case strength model, which is the same as the original model with the addition of
2This is their original model using the same sample of data as the case strength model

from 1972-1993 to allow for easier comparison.

55

case strength and the interaction terms between case strength and oral argument grade, and

case strength and ideological compatibility. Coefficients produced by maximum likelihood

models are not useful for interpreting effect size but the results of the case strength model

support Johnson, Wahlbeck and Spriggs’s (2006) original findings. The coefficient direction

and significance remain the same in the case strength model for all of the variables included

in the original model.

With regard to model fit, the Akaike and Bayseian information criterion have smaller

values for the case strength model, which suggest that this model is a better fit of the data.

Percent correctly predicted is also improved by the case strength model. Another measure of

fit is the area under the receiving operating curve (ROC). This is a discrimination measure

that describes how well the model does at distinguishing the reverses (dependent variable

equals 1) from the affirms (dependent variable equals 1) based on the parameters included

in the model. The original model has a ROC value of .746. The case strength model is

significantly better (p < .05) at discriminating between reverse and affirm outcomes with

a ROC value of .773. To further discuss the model results I use plots of predicted values

instead of the table of coefficients.

56

Table 3.2: Logistic Regression Model of the Justices’ Propensity to Reverse

Original Model Case Strength Model

0.255∗
(0.032)
−0.033∗
(0.007)
−0.100∗
(0.014
0.463∗
(0.037)
0.463∗
(0.070)
0.041∗
(0.007)
0.064
(0.102)
0.023
(0.123)
0.929∗
(0.189)
−0.896∗
(0.133)
0.013
(0.110)
−0.121
(0.155)
0.571∗
(0.152)
0.250
(0.172)
−0.653∗
(0.215)
1.919∗
(0.311)
−0.355∗
(0.114)
−0.289
(0.281)
−0.052
(0.113)
0.102
(0.082)
−0.117∗
(0.019)
−0.541∗
(0.107)

0.352∗
(0.031)
0.217∗
(0.056)
0.055∗
(0.007)
0.180
(0.101)
−0.112
(0.121)
1.000∗
(0..180)
−0.953∗
(0.135)
0.200
(0.113)
0.076
(0.155)
0.556∗
(0.157)
0.240
(0.158)
−0.745∗
(0.185)
−1.893∗
(0.289)
−0.243∗
(0.117)
−0.359
(0.270)
0.046
(0.109)
−0.040
(0.086)
−0.104∗
(0.019)
0.251∗
(0.042)

Case Strength

Case Strength * Ideological Compatibility

Case Strength * Oral Argument Grade

Ideological Compatibility

Oral Argument Grade

Ideological Compatibility * Oral Argument Grade

Case Complexity

Case Complexity * Oral Argument Grade

U.S. Appellant

U.S. Appellee

SG Appellant

SG Applee

Washington Elite Appellant

Washington Elite Appellee

Law Professor Appellant

Law Professor Applee

Clerk Appellant

Clerk Appellee

Elite Law School Appellant

Elite Law School Applee

Difference in Litigating Experience

Constant

N
AIC
BIC
Log Likelihood
Percent Correctly Predicted
Robust standard errors in parentheses
* denotes p< 0.05 for two-tailed test

2262
2660.891
2741.027
−1316.445
70.11%

2262
2567.978
2648.114
−1269.989
72.33%

57

Figure 3.1: Likelihood of a Reversal Across Oral Argument Performance - The above plot
shows the likelihood of siding with the petitioner for three levels of case strength across oral argument
performance. The green line represents the law supporting a strong affirmance or a favorable outcome
for the respondent. The purple line is when the law is uncertain and the case could go either way and
the orange line represents cases when law favors the petitioner. The vertical bars are 95% confidence
intervals. All other continuous variables are held at their means and categorical variables are held
at their modes.

Figure 3.1 displays the likelihood of a justice voting for the petitioner at three differ-

ent levels of case strength (strong affirmance, uncertain, and strong reversal) across oral
argument grades.3 Recall that oral argument grades are coded as the difference between

the petitioner and respondent’s grades, and higher values on this variable indicate a better

performance at oral argument by the petitioner. The orange line demonstrates that even

if the petitioner’s oral argument performance is downright awful, if he has a stronger legal

3I include three for ease of interpretation and visual clarity.

58

0.00.10.20.30.40.50.60.70.80.91.0−2−1012Oral Argument PerformanceLikelihood of Siding with the PetitionerStrong AffirmanceStrong ReversalUncertainargument to fall back on, his likelihood of reversal, or winning his case, is 49%. This means

a petitioner’s attorney can make a complete fool of himself at oral argument and still have

a 50-50 chance of a favorable outcome. That being said, in an average case that has equally

matched oral advocates (oral argument performance equal to 0), if the petitioner has a su-

perior legal argument it increases their likelihood of success by 24 percentage points, a 54

percent change. Even when the petitioner runs circles around the respondent during oral

arguments and has a very good performance and is still favored, his odds of success increase

by 17 percentage points, a statistically significant difference.

This finding lends some support to Budziak and Lempert’s (2015) supposition that there

is another factor, likely related to the legal strength of a case missing from Johnson, Wahlbeck

and Spriggs’s (2006) original analysis. This is not to say that oral argument performance

does not matter. Oral argument performance is a positive and significant predictor of case

outcomes. However, case strength can lead to a successful outcome for the petitioner even

when oral argument performance is poor, and only the most exceptional oral argument

performance can make up for an inferior legal argument.

Next, Figure 3.2 displays the relationship between case strength and ideological compati-

bility. I theorized that legal quality can help overcome ideological differences, and I find some

support for this hypothesis. When a justice and the petitioner are at opposite ends of the

ideological spectrum and the petitioner has a weak legal argument, his chances of winning

are less than 12%. Yet, if the same petitioner were to have a strong legal argument, his

chances of winning surpass 28%. While this is only a modest increase, these are ideologically

extreme justices. Of the fifteen justices in the data, only Douglas, Marshall, and Rehnquist

are this extreme. Take for example Chief Justice Burger, a staunch conservative throughout

his career with a compatibility score that ranged between -1.25 and -1 for liberal litigants.

When a liberal litigant petitioned the Court Justice Burger could be persuaded to vote in

their favor if they possessed a stronger legal argument. This suggests that the law can have

59

a constraining effect on justices’ decision on the merits. That being said, when a more ide-

ologically extreme justice (i.e., Douglas, Marshall, and Rehnquist) and the petitioner in a

case are ideologically congruent, the strength of a litigant’s case is insignificant to their final

decision. This is consistent with the results from the first chapter that finds that ideological

extreme justices that also finds that law does not influence their final decision on the merits.

Figure 3.2: Likelihood of a Reversal Across Ideological Compatibility - The plot displays the
likelihood of a vote for the petitioner at three levels of case strength across ideological compatibility.
Vertical lines around that estimates are 95% confidence intervals. All continuous variables are held
at their means and categorical variables are held at their modes.

3.1.3 Discussion

The above findings demonstrate the importance of incorporating a measure of law such

as case strength into models of judicial decision making. An attorney whose argument has

the strong support of law is significantly more likely to succeed before the Court even if

their opponent performs better during oral arguments. Additionally, law has the ability to

constrain most justices from voting with their policy preferences. These findings do lend

60

0.10.20.30.40.50.60.70.80.9−3.25−2.75−2.25−1.75−0.75−1.25−0.75−0.250.250.751.251.752.252.753.253.754.25Ideological CompatibilityLikelihood of Siding with the PetitionerStrong AffirmanceStrong ReversalUncertainsupport to Budziak and Lempert’s (2015) hypothesis, but it does not discount Johnson,

Wahlbeck and Spriggs’s (2006) original findings. Oral argument performance still has a

positive and significant effect on success before the Court. This means that skeptics of

the importance and significants oral arguments, and even the justices themselves, are still

influenced by these proceedings.

3.2 Case Strength and Strategic Responses to Majority Opinions

After oral arguments justices retire to their conference room where many decisions must

be made before the public and litigants learn of the Court’s final decision. First, the justices

place their initial vote in order of seniority starting with the chief justice. Then, the chief or

most senior justice in the majority coalition assigns the opinion. Although they placed their

initial votes, their decisions are not set in stone. Once the opinion author completes the

first draft of the opinion, he circulates it to the other justices, and they must decide how to

respond. As Maltzman, Spriggs and Wahlbeck (2000) demonstrate, a justice can generally

respond in one of seven ways: he can choose to (1) join the majority opinion, (2) issue a

wait statement, (3) make a suggestion, (4) make a threat (5) state intentions to write a

separate opinion, (6) join or circulate a concurrence, or (7) join or circulate a dissent. Many

have examined how legal factors or discrete aspects of law influence the final decision on

the merits (Bailey and Maltzman 2008; Bartels and O’Geen 2015; Hansford 2004; Richards

and Kritzer 2002), however law’s influence at earlier stages of the decision making process,

like opinion bargaining, are still unsettled. While Maltzman, Spriggs and Wahlbeck (2000)

control for policy preferences, strategic interactions, and contextual controls like when the

case was heard, its salience, and who was assigned to author the opinion, they fail to account

for legal reasons this bargaining may or may not take place.

There are theoretical reasons to believe that law influences the opinion bargaining process.

Justices are seekers of their preferred policy position (Epstein and Knight 1998, 2000), but

policy preferences likely play out on the poles of opinion bargaining. That is, for an average

61

case if a justice prefers to affirm (reverse) and the majority is reversing (affirming), it is

unlikely justices will find themselves in the majority aside for some legal reason or he is on

the fence. If for some reason he is in the majority and still prefers the to affirm (reverse),

his likely response will be to sign-on or circulate a dissent. On the other hand, a justice

that agrees with the disposition of the majority opinion but disagrees with certain legal

aspects or reasoning behind the disposition less-drastic responses would be appropriate such

as issue a wait statement, offer a suggestion, or articulate a threat to gain leverage for his

bargaining. He may even circulate or join a concurrence if his policy preferences lead him to

dispositionally agree with the majority but disagree as to the legal justification.

In this section I will reexamine Chapter 3: Strategic Response to Draft Opinions from

Maltzman, Spriggs and Wahlbeck’s (2000) book “Crafting Law on the Supreme Court: The

Collegial Game” to determine if, and how, law and legal consideration impact the opinion

bargaining process. After demonstrating that the chief justice strategically assigns majority

opinions the authors seek to explain the bargaining that occurs after the first draft is circu-

lated. They find the the ideological distance between the bargaining justice and the opinion

author and coalition mean are significant predictors of the bargaining response. Ideological

preferences should still predict bargaining, but law and opinion it supports should also in-

fluence this process. That is, justices do not just bargain over political policy but they also

bargain over legal policy.

3.2.1 Data and Measures

Maltzman, Spriggs and Wahlbeck (2000) examine justices’ responses to majority opinion

drafts using data from the Burger Court (1969-1986) collected from the archives of Justice
William Brennan.4 They model strategic responses to opinions in two ways, the first is by
4To include law and legal reasoning in the model, I will start my analysis with the 1972
term. My data come from Justice Harry Blackmun’s bench memos and although Justice
Blackmun was appointed to the Court only one year after Chief Justice Burger, his bench

62

using a simple dichotomous indicator coded as 0 if a justice joined the majority opinion and

1 if they chose to bargain in some way (1). This modeling strategy lacks the nuances of

opinion bargaining. As Maltzman, Spriggs and Wahlbeck (2000) acknowledge, this depen-

dent variable fails to distinguish between bargaining options, which is where I expect the law
and the strength of a case to influence the justices’ decisions.5 Therefore, I will focus my

attention on the second model. In this model the dependent variable is a categorical variable

of members of the conference majority coalitions’ first responses to majority opinion author’s

draft. As previously stated, the categories are: (1) justice joins majority opinion, (2) justice

issues wait statement, (3) justice makes a suggestion, (4) justice articulate a threat, (5) jus-

tice signals he will join or circulate a separate opinion, (6) justice circulates or joins draft

of concurring opinion, or (7) justice joins draft of dissenting opinion. Due to the categorical

nature of this dependent variable, I follow the authors and employ a multinomial logistic

regression model. This model estimates the likelihood of each outcome relative to a baseline

category. I select the “join majority” category as the baseline since that is the only option

that does not involve any bargaining.

The authors’ primary independent variables of interest represent justices’ policy consid-

erations. The first is the issue-specific ideological distance between the majority opinion

author and each justice of the original coalition for every case. This variable is the absolute

value of the difference between a justice’s percentage of liberal votes in a case-specific issue

area (issue areas from Epstein et al. (2019)) and the majority opinion author’s percentage

of liberal votes in the same issue area. Also of interest is the coalition distance, which is

the ideological distance between the mean of the conference majority coalition (excluding

the author) and each justice. These ideology scores are also specific to each of the Supreme

Court Database’s twelve issue areas (Epstein et al. 2019). For both of these variables, as

memos did not become consistent in their form or purpose until the 1972 term.

5The original dichotomous dependent variable model and the same model with the addi-

tion of my measure of case strength is available in Table B.1 in the appendix.

63

the ideological distance decreases, the less likely the bargaining justice will circulate a wait

statement, submit a suggestion, make a threat, or will write or signal a concurrence or

dissent.

The next set of key independent variables are used to examine strategic interactions

between the justices. The first is the winning margin, which captures the number of votes

needed to form a winning coalition. This is calculated by subtracting the number of justices

who voted with the majority opinion assigner during conference from the number of justices

necessary to form the winning majority. When the winning margin is large, which means

the majority coalition is small and needs to gain or maintain justices, a justice should be

less likely to circulate a draft of a concurrence or dissent, but more likely to circulate a

wait statement, suggestion, threat, or signal he will write a separate opinion. The other

strategic variable captures the repeated collegial game the justices engage in. A justice may

be more hesitant to lend his support to a majority opinion if the author has not cooperated

with him in the past. Therefore, a justice is more likely to engage in all bargaining options

over signing the majority opinion if the author has not cooperated in the past. To measure

this variable the authors calculated the percentage of the separate opinions written by the

bargaining justice that were joined by the majority opinion author in the previous term

and then regressed this percentage on ideological distance in order to ensure this variable

is not a product of shared preferences. The residual of this model is then used to represent

cooperation in the current models.

The authors also control for numerous contextual factors. For expediency, these are listed

in the table below in Table 3.3, along with a description of how they are measured.

64

Table 3.3: Contextual Control Variable Measurement

Variable
Political Salience

Legal Salience

Case Complexity

End of Term

Workload

Measurement
Number of amicus briefs over or under than the average
number of amicus briefs for that term.

1 if the Court struck down a law as unconstitutional,
overturned, or altered precedent.

Factor score from a factor analysis of the number of
issues raised by the case, the number of legal provisions
relevant to a case, and the number of opinions released
in a case.

Number of days when the opinion was assigned until
July 1, which is traditionally the end of the Court’s term.

Number of majority and separate opinions the justice
was working on the day the first draft of the majority
opinion was circulated.

Chief Justice

1 if the justice is the chief justice, 0 otherwise.

Freshman Author

1 if the justice was in their first or second term, 0 oth-
erwise.

Expertise

Number of dissents or concurrences a justice wrote in a
specific issue area divided by the number of similar-issue
cases that made it to the Court in the preceding term.

I argue that a decision to join the majority unconditionally or write a dissent is primarily

determined by a justice’s policy preferences or their ideological distance from the majority

opinion coalition. If a justice were compelled to vote with the majority at conference it is

likely because he preferred the dispositional case outcome selected by the majority. Once

the disposition is selected subsequent opinion decisions are related to the legal reasoning

and justification to support the disposition. Therefore, any response by a member of the

majority coalition other than to join the opinion are due, in part, to the strength of the law

or legal reasoning supporting the majority’s decision.

To account the relationship between law and the majority opinion’s disposition I recode

my case strength measure from Chapter One of this dissertation. This measure is derived

65

from Justice Harry Blackmun’s bench memoranda. These memos are written after a case

is granted review but prior to oral arguments. They contain a summary and analysis of

the case at hand and make a final recommendation. This recommendation is twofold. It

is dispositional in the sense that it says reverse or affirm, but it is also supported with

the legal reasoning to support how the law leads to that decision. I use the dispositional

recommendation to craft my measure of case strength. It is an ordinal measure coded one

to five where one represents the law strongly support for an affirmance and five represents

that the law strongly favors a reversal. To determine if the majority opinion position is

supported by law, create a majority-law agreement variable. This variable takes on a value
of 5 if the majority disposition is strongly supported by the law.6 That is, if the majority

writes an opinion to reverse (affirm) and the law strongly favors a reversal (affirmance). This

variable takes on the value of 4 when the law weakly supports the majority’s position. Like

the original case strength variable, when the law neither supports nor opposes a particular

position, this variable is coded as 3. It takes on a value of 2 when the law weakly opposes

the majority’s position and 1 when it is strongly in conflict with the majority’s decision. A

value of 1 represents when law suggests a reverse (affirm) and the majority affirms (reverses)

the lower court’s decision.

When the law is clear and strongly supports the majority’s position there should be less

to bargain over as far as the content of the opinion. Justices’ archival materials reveal,

justice’s suggestions can be as simple as asking for a word change or omission or inclusion
of a citation.7 Therefore, when a justice agrees with the majority’s disposition and the law

strongly supports the same outcome, a join or offering a simple suggestion are the most likely
6Majority disposition is determined by the caseDisposition variable from the Supreme
Court Database.Epstein et al. (2019) All dispositions excluding 2 (affirm) and 9 (DIG) are
coded as relief for the petitioner and 2 and 9 are coded as relief for the respondent.

7See, for example, Justice Sandra Day O’Connor’s memo on February 12 ,1985, to Justice
William Brennan in the case Winston and Davis v. Lee (83-1334). She asks him to omit
the Tennessee v. Garner citation on page five of his opinion. If he complied she would be
happy to join.

66

responses. When the law is less clear or not as strongly in favor of the majority’s position

this may give rise to apprehension. Therefore, a justice may issue a wait statement to see

what other opinions circulate or even circulate or join a concurrence. Therefore, when the

law weakly favors the majority’s position, this means there could be other possible ways to

justify the Court’s legal position on an issue. This could result in the same dispositional

outcome but for different reasons. When law and the majority opinion are in agreement the

following should hold:

Law-Majority Strong Agreement: If the law strongly supports the majority

opinion, a justice should be more likely to join the majority, or make a suggestion.

A justice should be less likely to issue a wait statement, articulate a threat, signal

intentions for a separate opinion, or state intentions to author or join a concurring

or dissenting opinion.

Law-Majority Weak Agreement:

If the law weakly supports the majority

opinion, a justice should be more likely to circulate a wait statement or make a

suggestion, signal intentions for a separate opinion, or state intentions to author

or join a concurring opinion. A justice should be less likely to articulate a threat

or state intentions to author or join a dissenting opinion.

Another scenario is when the law is unclear. In these instances the law could feasibly

support the petitioner or respondent. This could have a number of effects on the justices.

Those who are indecisive may be more likely to issue a wait statement to see how each opinion

develops. Most likely, ambiguous law creates indecision amongst the justices and increases

the likelihood of separate opinions. As a result, those who joined the original majority opinion

may opt to write a concurrence or signal intentions for some type of separate opinion.

Law-Majority Uncertain: If the law is ambiguous and feasibly supports the

petitioner or the respondent, the likelihood of them responding with a wait state-

67

ment, signaling with intentions for a separate opinion, or to join or circulate a

concurrence should increase.

Finally, the majority’s position may not always be in agreement with the position the law

supports. However, the justices in the data already voted with the majority at conference

indicating that they prefer this disposition. It is possible that through further scrutiny and

discussion a justice could be compelled to switch their vote because they feel that the law

constrains their preferences or because the initial draft of the majority opinion was not what

they had in mind. If this occurs a justice has a number of possible options, they could wait

to see how the opinions develop, threaten to write a separate opinion, of just abandon their

initial vote for the minority. Weak and strong conflict between the law and the majority

opinion should result in similar bargaining responses but with strong conflict leading to a

higher likelihood because the law is less aligned with the majority’s position.

Law-Majority Weak Conflict: If the law weakly conflicts with the majority

opinion, a justice should be more likely to issue a wait statement, signal to write

a separate opinion, or declare intentions to circulate or join a dissent. He should

be less likely to join the majority opinion or write a concurring opinion.

Law-Majority Strong Conflict: If the law strongly conflicts with the majority

opinion, a justice should be more likely to issue a wait statement, threaten to

write a separate opinion, or declare intentions to circulate or join a dissent. He

should be less likely to join the majority opinion or write a concurring opinion.

3.2.2 Method and Results

The dependent variable is nominal and made up of the seven ways in which a justice

can respond to the majority opinion author’s first draft: join, write wait statement, make a

suggestion, make a threat, signal to write separate opinion, or state intentions to write or

68

Table 3.4: Multinomial Logistic Regression Model of Justices’ Strategic Responses to the
Majority Opinion Author’s First Draft

Legal Consideration
Majority Coalition-Law Agreement

Policy Preferences
Author Distance

Coalition Distance

Strategic Interaction
Winning Margin

Cooperation

Contextual Controls
Political Salience

Legal Salience

Case Complexity

End of Term

Workload

Chief Justice

Freshman Author

Expertise

Constant

N
Log Likelihood
AIC
BIC
Standard errors in parentheses
∗ p < 0.05

Wait
−0.029
(0.041)

0.021∗
(0.005)
0.010
(0.014)
−0.145∗
(0.045)
−2.964∗
(0.973)

0.141
(0.097)
0.064
(0.267)
−0.115∗
(0.043)
0.002∗
(0.001)
0.001
(0.024)
−0.595∗
(0.180)
−0.301
(0.207)
−0.063
(0.047)
−4.203∗
(0.670)
7563
−6028.889
12075.777
12138.157

Suggestion Threat Separate Concur Dissent

0.177∗ −0.014
(0.045)
(0.045)

−0.144∗
(0.050)

0.050
(0.044)

0.035∗
(0.005)
0.010∗
(0.004)

0.042∗
(0.007)
−0.022
(0.013)
0.200∗
(0.056)
−1.689
(1.394)

0.018∗
0.026∗
(0.005)
(0.005)
0.042∗
0.003
(0.007)
(0.007)
−0.159∗ −0.057
0.063
(0.056)
(0.038)
(0.054)
−2.583∗ −2.663∗
−1.520
(0.843)
(1.065)
(0.890)
−0.021
−0.068
0.188
0.146
(0.121)
(0.054)
(0.097)
(0.096)
0.569∗
−0.303
0.153
0.210
(0.366)
(0.180)
(0.243)
(0.317)
0.148∗
0.085
0.077
0.050
(0.100)
(0.043)
(0.079)
(0.044)
0.004∗
−0.001
0.001
0.077
(0.001)
(0.043)
(0.001)
(0.001)
0.031∗
−0.003
−0.006
−0.013
(0.025)
(0.016)
(0.010)
(0.018)
−0.876∗
−0.656∗ −0.639∗
0.158
(0.172)
(0.107)
(0.104)
(0.146)
0.543∗ −0.065
−1.160∗
0.359
(0.131)
(0.255)
(0.261)
(0.510)
0.099∗
0.137∗
0.053
0.036
(0.037)
(0.037)
(0.052)
(0.056)
−6.115∗ −3.330∗ −3.179∗ −5.313∗
(0.580)
(0.481)

(0.338)

(0.538)

−0.004
(0.036)

0.027∗
(0.004)
−0.003
(0.012)

0.071
(0.044)
−2.176∗
(0.508)
0.148∗
(0.075)
0.109
(0.142)
0.239∗
(0.065)
0.003
(0.001)
0.011
(0.029)
−0.811∗
(0.181)
0.030
(0.308)
−0.044
(0.054)
−4.945∗
(0.535)

69

join concurring or dissenting opinion. To model this type of variable I utilize a multinomial

logistic regression model. Multinomial logistic models estimate n-1 binary models for each

category in the dependent variable excluding one category as the baseline. Joining the

majority opinion is the baseline for this model. Additionally, like in the oral arguments

model, Justice Blackmun’s strategic behavior is excluded from this model in order to obtain

conservative estimates that are not simply a result of Blackmun following the advice of his

law clerks.

Each column in Table 3.4 represents one of the categories of the multinomial logistic

regression model, excluding the baseline category. The coefficients from maximum likelihood

are not useful for interpreting the effects so I will begin with Figure 3.3, which shows the

average marginal effect of majority coalition-law agreement for each response type. More

specifically each point estimate demonstrates the effect of moving from strong conflict to

strong agreement between law and the majority opinion for each response type.

70

Figure 3.3: Average Marginal Effect of Majority Coalition-Law Agreement on Bargaining
Response - Each point represents the average marginal effect of majority coalition-law agreement
on each bargaining response. Vertical lines around that estimates are 95% confidence intervals. All
continuous variables are held at their means and dichotomous variables are held at their modes.

As expected, as law-majority agreement becomes strong the likelihood of join increases.

Similarly the likelihood of wait and suggestion decreases when the majority opinion is

strongly supported by law, none of these relationships are statistically significant. Unex-

pectedly, as the support of the law for the majority opinion grows stronger the likelihood

of a threat significantly increases. The likelihood of writing a separate opinion decreases as

expected, but law-majority agreement does not have a significant effect on justices writing

a separate opinion. It does however, exert a significant and negative effect on the likelihood

of writing a concurrence. Strong majority-coalition and law agreement has an insignificant

effect on the likelihood of a dissent.

To further examine the relationship between law majority coalition agreement and the

71

−0.010−0.0050.0000.0050.010JoinWaitSuggestionThreatSeparateConcurrenceDissentOutcomeAverage Marginal Effectlikelihood of a justice responding with a threat, the relationship is plotted in the left panel

of Figure 3.4. The x-axis shows the five levels of law majority coalition agreement and the

y-axis is the likelihood of a justice responding with a threat. When there is strong conflict

between the law and the majority’s position the likelihood of a threat is 0.01. This likelihood

increases to .02 when there is strong agreement. This unexpected finding may be a product

of the terms included in the analysis. My data begin in 1972 and extend through 1986,

which encompasses nearly the entire Burger Court. Chief Justice Burger succeeded Chief

Justice Warren, whose leadership guided the country through significant liberal decisions in

individual and civil rights; and many of the liberal justices on the Court who served with

Warren were determined to maintain the significant changes they helped decide. As a result,

Justices Brennan and Marshall would join the conservative majority in order to try and

control the opinion content, but if it drifted too far from their reach they would be forced to

dissent (Woodward and Armstrong 1979).

72

Figure 3.4: Likelihood of Threatening and Conurring Responses Across Majority
Coalition-Law Agreement - The left panel displays the likelihood of a threatening response
given the different levels of majority-law agreement. The right panel shows the likelihood of a con-
currence across the same range. Vertical lines around that estimates are 95% confidence intervals.
All continuous variables are held at their means and dichotomous variables are held at their modes.

The right panel displays the relationship between the likelihood for a concurrence and ma-

jority coalition-law agreement. As law’s support for the majority’s position grows stronger,

the likelihood of a concurrence decreases from 0.04 to 0.02. This is consistent with Pamela Cor-

ley’s (2010) work. Corley writes, “Concurrences provide justices with discretionary oppor-

tunities to voice their legal perspectives” (2010, 8). Specifically, she finds that ideological

compatibility between a justice and the majority opinion author has virtually no effect on a

justice’s likelihood to author a concurrence. This is because a concurrence is the inability to

reach the same legal justification for the same disposition. When the law strongly supports

the majority’s decision this means that the law clearly points in one direction, thus creating

a consensus among the judges.

Multinomial regression models can be difficult to understand and interpret the results, so

to provide further insight I created a dichotomous indicator of each level of majority coalition-

73

0.0100.0150.0200.0250.030StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementMajority Coalition−Law AgreementLikelihood of a ThreatLikelihood of a ThreatAcross Majority Coalition−Law Agreement0.010.020.030.040.050.06StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementMajority Coalition−Law AgreementLikelihood of a ConcurrenceLikelihood of a ConcurrenceAcross Majority Coalition−Law Agreementlaw agreement to examine the predicted probabilities of each level of majority coalition-law

agreement exerts on each type of bargaining response. These are shown in Figure 3.5 and
3.6.8

Figure 3.5 shows the likelihood a justice joins the majority opinion across the different

levels of majority coalition-law agreement. I hypothesized above, when law and legal rea-

soning support the majority opinion, justices should be more likely to join the majority.

This influence should be greatest when there is strong agreement. I find support for this

hypothesis. When the law and majority opinion are in agreement at any level, a justice is

significantly more likely to join the majority opinion than when the law is uncertain. There

is not a statistical difference in the likelihood of joining the majority opinion when the law

supports the majority’s position and when it is in disagreement. As previously discussed,

this represents justices who will vote with the majority regardless of law and legal reasoning

because their policy preferences compel them to do so.

8The coefficient table for this model can be found in the appendix, Table B.2.

74

Figure 3.5: Likelihood of Joining the Majority Across Majority Coalition-Law Agreement
- These are the predicted values of joining the majority opinion at each level of majority opinion-
law agreement. Vertical lines around that estimates are 95% confidence intervals. All continuous
variables are held at their means and dichotomous variables are held at their modes.

75

0.700.750.800.85StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihoodJoins the Majority Opinion(a) Likelihood of a Wait Statement

(b) Likelihood of a Suggestion

(c) Likelihood of a Threat

(d) Likelihood of a Separate Opinion

(e) Likelihood of a Concurrence

(f) Likelihood of a Dissenting Opinion

Figure 3.6: Predicted Probabilities of Bargaining Responses Across Across Majority Coalition-
Law Agreement - The plots above show the predicted probabilities of each bargaining response across
each level of majority-law agreement. Vertical lines around that estimates are 95% confidence intervals. All
continuous variables are held at their means and dichotomous variables are held at their modes.

76

0.0000.0250.0500.0750.1000.125StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihood of Issuing a Wait Statement0.0000.0250.0500.0750.1000.125StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihood of Making a Suggestion0.0000.0250.0500.0750.1000.125StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihood of Articulating a Threat0.0000.0250.0500.0750.1000.125StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihood of a Separate Opinion0.0000.0250.0500.0750.1000.125StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihood of Circulating or Joining a Concurrence0.0000.0250.0500.0750.1000.125StrongConflictWeakConflictUncertainWeakAgreementStrongAgreementLaw and Majority Coalition RelationshipLikelihood of Circulatig or Joining a DissentFigure 3.6 displays the predicted values for the six other response types. The first plot

in the figure displays the likelihood of issuing a wait statement across the different levels

of majority coalition-law agreement. I hypothesized that strong conflict, weak conflict, un-

certainty, and weak agreement should all result in a higher likelihood of a wait statement

because there is some level of disagreement or wavering support between law and the major-

ity’s position, and strong agreement should result is a lower likelihood of a wait statement. I

find that that when the majority’s position is strongly supported by law, a justice that orig-

inally joined the majority at conference is significantly less likely to issue a wait statement

than when the law is uncertain. However, I find that when law and the majority’s position

disagree slightly, a justice is significantly less likely to issue a wait statement than any other

level of law-majority agreement or disagreement.

The second plot in the top row of Figure 3.6 depicts the likelihood of a justice responding

with a suggestion. I theorized that suggestions should be most likely when the majority’s

opinion is well supported by law. These suggestions can be simple word choice changes to

more substantial changes related to the narrowness of the opinion, yet if the justice already

voted for the majority’s position and the law supports this position they should be less likely

to want to change their position and more likely to want to work to improve the opinion and

make it satisfactory. I do not find support for this hypothesis. The likelihood of making a

suggestion across the range of legal support for the majority’s opinion does not significantly

differ. The lone exception is when there is weak agreement and a is justice significantly less

likely to make a suggestion than strong or weak conflict. When there is only weak support

for the majority’s position a justice may be more likely to try and improve the opinion in

another way other than a suggestion.

The final plot in the top row of the figure is the likelihood of a justice responding by

articulating a threat. First, I suggested that justices should be less likely to articulate a

threat when the law squarely aligns with the majority’s position. I do not find support for

77

this hypothesis, and in fact, I find the opposite to be true. Any level of legal support of the

majority opinion makes a justice significantly more likely to make a threat than any level of

conflict. This result echoes the results discussed above, and again, may be a product of the

time period being analyzed.

Moving to the second row of Figure 3.6, I find support for my hypotheses on writing

separate opinions. I hypothesized that uncertainty and ambiguity in the law will increase

the likelihood of writing separate opinions and I find that legal uncertainty significantly

increases the likelihood of a justice responding by signaling to write a separate opinion

compared to legal agreement with the majority opinion. This suggests that legal support for

the majority opinion helps to maintain that coalition from divulging into separate opinions.

Relatedly, I find that strong legal support of the majority opinion significantly reduces the

likelihood of a justice authoring a concurrence relative to strong conflict.

Lastly, I do not find support for my hypotheses on dissenting opinion responses. I theo-

rized that legal agreement should make dissents less likely and uncertainty or conflict should

make them more likely. The plot reveals that the likelihood of a justice joining or circulating

a dissent is the same regardless if the law conflicts with the majority opinion or agrees with

the majority opinion, however uncertainty in the law leads to a significantly lower chance of

a justice joining or circulating a dissent. This could be due to the fact that the uncertainty

and ambiguity leads a justice to revert to his policy preferences. He already voted to join

the majority opinion so why switch positions?

The results of the rest of the independent variables in the model remain the same as
the original model run on the same sample.9 Author ideological distance remains a signif-

icant and positive predictor of all bargaining options relative to joining the majority draft.

Similarly, the chief justice is significantly less likely to bargain than join. There is an im-

provement of fit in the model that incorporates legal considerations. Specifically the AIC

decreases from 12095.422 to 12075.777 and the BIC experiences a similar reduction from

9The table of results from this model can be found Table B.3 in the Appendix.

78

12157.802 to 12138.157. According to Long (1997) any reduction in the BIC greater than 10

lends “very strong” evidence that my model that incorporates legal considerations is preferred

over the original.

3.2.3 Discussion

The results above demonstrate that justices do not only bargain over policy but they

bargain over legal policy. This is a novel and significant finding because it demonstrates that

while justices will try to persuade one another to achieve their preferred outcomes (Epstein

and Knight 2000), negotiations extend past the political preferences to legal policy. A justice

will bargain to craft an opinion in the way that he believes the Court is producing the best

legal policy. Perhaps this bargaining on legal policy relates to the narrowness or broadness

in which the opinion is written, or it could have to do with implementation and execution

for the lower courts to follow. These results pave the way for future research to examine the

content and justification or reasoning for each type of response with regard to legal policy.

These findings, however, are not without limitations. First, my measure of case strength,

and therefore majority coalition-law agreement, is only based on the dispositional recom-

mendation at the end of the bench memos. It fails to take uncertainty of legal reasoning

into consideration. Take, for example, the following recommendation from United States v.

Donovan (1976):

I find the SG’s proposed rule too narrow and full of holes to adopt. The best
standard would be the normal probable cause standard that is applied in most
determinations. That has the advantage of consistency and predictability, since
government officials are presumably accustomed to judging their actions under
it. If you are troubled too much by the administrative difficulties that the SG
alleges would be encountered under a probable cause standard, then I would
opt for Judge Godbold’s solution. In either case, the suppression as to Donovan
and Robbins should be affirmed. If probable cause is the standard, you should
probably affirm for Buzzacco, too. If it is Godbold’s test, then a remand would
be better for him. Finally, the case should be remanded for a determination of

79

whether the Government was in good faith when it failed to include the names
of Merlo and Lauer in the inventory (20).10

All of the legal justifications point toward the same dispositional outcome, affirm in part and

remand. That being said the clerk offers three possible methods for reaching that decision:

(1) the OSG’s, (2) “the normal probable cause standard,” and (3) Judge Godbold’s solution.

My measure fails to take the variation in legal reasoning into account when they all point

towards the same outcome. If one of these justifications led to a straight affirmance and

another one a reverse, the measure would account for this discrepancy. This is feasibly what

the justices are bargaining over when the vote in the majority initially, which may explain

some of the less-intuitive results.

One other possible limitation is that I had to use the final majority decision from the

Supreme Court Database to determine the direction of the majority opinion (Epstein et al.

2019). As I mentioned previously, the justices’ initial conference votes are not set in stone and

in some cases the majority does not remain the majority if the coalition does not maintain

five votes. Therefore, by using the Supreme Court Database to determine the final majority

decision I am making the assumption that the final majority and initial majority are the

same. While this is not ideal, this is currently the only information that is available. Justices

place their initial votes during a secret conference and this data is currently unavailable to

researchers, however hopefully conference note data will be available soon thanks to ongoing

research by Black and Johnson (2017).

3.3 Conclusion

In this paper I demonstrated that law is an important consideration for justices’ decision

making, not just on the merits but at multiple stages of the judicial decision making process.

First, I showed that law as captured by the petitioner’s legal case strength, has a significant
10Bench memorandum in United States v. Donovan (75-212). Located in the Supreme
Court Case Files (Box 239, Folder #4) of Justice Harry A. Blackmun in the Library of
Congress, Washington D.C.

80

effect on case outcomes. A litigant with an abysmal oral argument performance can still

succeed before the Court if they have a strong case well supported by law and legal reasoning.

Furthermore, a strong case can compel justices who are ideologically opposed to a litigant’s

position to vote for them.

Next, I showed that law impacts Supreme Court justices’ response to the first draft of

the majority opinion. That is, the strength of the relationship between law the its support

of the majority’s position significantly influences the type of bargaining a justice attempts

to respond with after the first draft of the majority opinion circulates. More specifically,

as law’s support for the majority opinion increases, the likelihood of a threat significantly

increases. Conversely, the likelihood of a concurrence significantly decreases.

My findings contribute to the field’s growing understanding that law and legal consider-

ations impact justices’ decision making (Bartels 2009; Bartels and O’Geen 2014; Bailey and

Maltzman 2008; Black and Owens 2009; Hansford and Spriggs 2006; Richards and Kritzer

2002). This is both a normatively and positively important topic. Normatively, based on

the Constitution the Supreme Court is supposed to be a different type of branch of the fed-

eral government. While the majority of people are legal realists and recognize that justices’

preferences influence their decision making they believe they make these decisions in a prin-

cipled fashion based on laws and legal reasoning. They expect the justices to be more than

legislators in robes (Gibson et al. 2011). While research previously discussed demonstrated

that justices are more than their political counterparts in the other branches of government,

this paper helped expand this evidence by incorporating law directly into the models.

These findings and application are important for positive empirical judicial politics re-

search. As shown by both the Johnson, Wahlbeck and Spriggs (2006) and Maltzman, Spriggs

and Wahlbeck (2000) applications, law is a significant predictor in both merits outcomes and

bargaining decisions and not accounting for its influence could lead researchers to bias or

inaccurate inferences. Other areas ripe for testing law’s influence include justices’ question

81

asking behavior during oral arguments, how justices write their opinions, or how the public

perceives the Courts decisions.

82

CHAPTER 4

QUALITY VERSUS QUANTITY: AMICUS CURIAE BRIEF INFLUENCE
AND DECISION MAKING ON THE UNITED STATES SUPREME COURT

Not all of the Supreme Court’s cases are intriguing, page turning disputes. Most of their

cases fly under the radar of the average person, but this does not mean they are inconsequen-

tial. Take for example C & A Carbon Inc. v. Town of Clarkstown from the Court’s 1993

term. Clarkstown passed a local ordinance that required all waste flowing in or out of town

to be processed at their waste processing plant at a fee of $81 per ton. C & A Carbon Inc.,

a local waste management company, was caught transferring pre-processed waste directly

to the out of town landfill and then processing it to avoid Clarkstown’s fees. Ultimately,

Carbone sued the Town of Clarkstown alleging that the waste processing ordinance violated

the Commerce Clause by disrupting interstate commerce.

In addition to reviving briefs from the litigants, the Court accepted three amicus briefs:

two on behalf of the Town of Clarkston and one advocating for Carbon. Considering that

most cases the Court agrees to hear have no amicus briefs filed for either party, the three briefs

in this case demonstrate the broad implications of the Supreme Court’s decision (Collins

2008). The Chemical Manufacturers Association (CMA) and the National Solid Wastes

Management Association (NSWMA) filed on behalf of the town, however in relaying the

information conveyed in these briefs in a bench memo to Justice Harry Blackmun, his clerk

described NSWMA’s brief as largely repetitive of the petitioner’s arguments. This typically

suggests that the brief does not offer much help to the justices’ decision making. Conversely,

the National Association of Bond Lawyer’s (NABL) brief filed on behalf of Carbon was
described as “an excellent brief, worthy of substantial attention.”1 In the end, it is possible
1Bench memorandum in C & A Carbon Inc. v. Town of Clarkstown (92-1402). Located
in the Supreme Court Case Files (Box 637, Folder #6) of Justice Harry A. Blackmun in the
Library of Congress, Washington D.C.

83

that the NABL’s brief was effective with a 6-3 ideologically mixed majority ruling in favor

of Carbon.

Past research demonstrates that the number of amicus briefs filed in a case can increase

the likelihood of success of appearing before the Court (Black and Owens 2009; Caldeira and

Wright 1988; Caldeira, Wright and Zorn 1999) as well increase the likelihood of a positive

outcome (Collins 2008). Yet recent work suggests that prestige over quantity is important

when outside interests advocate before the Court. Powerful advocates, like the Office of

Solicitor General (OSG) or the American Civil Liberties Union (ACLU) are notorious for

experiencing policy success before the Supreme Court as amici (Box-Steffensmeier, Chris-

tenson and Hitt 2013; Lynch 2004). Is this because they are repeat players, have extensive

resources, or are simply better at constructing amicus briefs? The answer is likely all of the

above, but critical questions remain unclear: what makes an effective or ineffective amicus

brief and do effective briefs impact Supreme Court case outcomes?

I argue that effective amicus briefs are those that offer new and well constructed argu-

ments to augment the litigants’ merits briefs. That is, an effective amicus brief does not

simply reiterate their party’s merits brief, they should bring alternative legal justifications

and implications to the Court’s attention. It should also be concise. I will demonstrate that

amicus briefs that are poorly written and regurgitate the same argument as the litigant are

not particularly useful for the justices.

Additionally, I believe that effective briefs should have a significant impact on case out-

comes regardless of the number of additional briefs that are filed in the case. Archival

materials suggest that justices do not read all of the amicus briefs filed in a case and instead

focus their attention on certain briefs, therefore these briefs should have a stronger impact

on the justices’ decision making.

I find that there are significant differences between amicus briefs that are effective, briefs

that are neither effective or ineffective, and ineffective amicus briefs. Unsurprisingly, one of

84

the most effective brief writers is the Solicitor General. I also find that effective amicus briefs

have a significantly greater impact on justices’ voting behavior, even after controlling for the

presence of other briefs. Additionally, I find that quantity over quality isn’t necessarily

preferred in all cases as it is at the agenda stage (Caldeira and Wright 1988; Black and

Owens 2009). In fact, a poorly crafted amicus brief can significantly decrease a litigant’s

chances of obtaining a majority of the justices’ votes.

This research makes three important contributions. First, the importance of this topic

reaches far past academia. Judges and their clerks devote much of their time for each case

reviewing the material provided in these briefs. Furthermore understanding how parties

can write effective briefs is critical for legal practitioners. With a “dysfunctional Congress”

outside interests continue to turn to the judicial branch to effect policy change (Binder 2015).

This is evidenced by the growing number of amicus briefs submitted to the Supreme Court

in recent terms. The National Law Journal cites that the second decade of the 2000s saw

more than double the cases filed in the 1990s and the 2012 October term had a record 1,001

amicus briefs filed with an average of 14 per case (Franze and Anderson 2016). To file an

amicus brief requires friends of the Court to spend significant time and financial resources,

with some costing upwards of $100,000 (Ward 2007). Given the resources that are being

used to influence the Supreme Court it is important for outside interests to understand the

most effective way to achieve their goals. This research is the first step in understanding

how to do this.

Next, my findings suggest that justices use amicus briefs for both policy and legal reasons.

That is, my results suggest that effective or high caliber amicus briefs are not necessarily the

briefs that appeal to a justice’s policy preferences. Similarly, ineffective amicus briefs are not

simply briefs that clash with a justice’s preferences. This may suggest that outside interests

should focus their attention on the diversity of legal arguments to support a particular

litigant’s position instead of pandering to the justices political preferences. Oftentimes,

85

effective briefs are those that offer a fresh take on the legal dispute between the parties and

this array of legal arguments leads to successful outcomes.

Lastly, my findings shed new light on existing research. Spriggs and Wahlbeck (1997b)

find, contrary to conventional wisdom, amicus briefs primarily repeat parties arguments

instead of contributing unique arguments and when they do contribute unique arguments

the Court is unlikely to use them to for their opinions. Conversely, Collins Jr, Corley

and Hamner (2014) find that the majority of amici’s information differs from the litigants’

briefs. While I do not examine the arguments in each brief, I find that for an eight term

sample, most of the briefs that clerks suggest to their justices to spend more time reading

are those that either contribute unique information/arguments, or briefs that do a better

job of explaining the litigants’ arguments than the briefs on the merits. Although the Court

may not frame their opinions around this rationalization, I find they still influence their

final decision. Amici’s re-framed or unique perspectives on the legal arguments in a case can

demonstrate the broader impacts in which the justices’ decisions will impact society.

To explain the difference between effective and ineffective amicus briefs and how they im-

pact Supreme Court decision making, I begin by providing a discussion of existing literature

on amicus briefs. Then, using unique archival data, I describe how I developed my amicus

brief typology and validate it. Lastly, I examine how exactly these briefs impact Supreme

Court justices’ decision making on the merits.

4.1 The Influence of Amicus Briefs at the Supreme Court

Supreme Court justices are some of the most brilliant legal minds in the country. Yet,

despite this fact, they are not experts in every area of law that comes before the Court.

They may have practiced or specialized in one area in particular, like Justice Ruth Bader

Ginsburg who is famous for her advocacy of womens’ rights. Other justices were lifetime

judges or social servants whose attention was always spread across a wide variety of cases

but never an expert in on particular area, like Justice Brett Kavanaugh. In order for justices

86

gain the most information and make an educated decision Rule 37 was established. Rule

37 of the Supreme Court Rules allows outside interests to file briefs to influence both the
agenda and merits decisions. Rule 37 Section 1 reads as follows:2

An amicus curiae brief that brings to the attention of the Court relevant matter
not already brought to its attention by the parties may be of considerable help
to the Court. An amicus curiae brief that does not serve this purpose burdens
the Court, and its filing is not favored.

Amicus brief research at the agenda stage unequivocally agrees that when outside inter-

ests file briefs they provide information by signaling to the Court the broad implications of

the case. Whether the brief is filed in opposition to or in support of granting review, the

information provided signals the same thing: the case is important and thus the Court’s

likelihood of granting review increases (Black and Owens 2009; Caldeira and Wright 1988,

1990). Similarly, at the merits stage amici are supposed to provide unique information not

provided by the litigants, as the rule states. Yet research on whether friends of the Court

actually accomplish this objective is mixed.

Early research on amicus briefs failed to find a significant relationship between the number

of amicus briefs filed on the merits and successful outcomes (Songer and Sheehan 1993), or

it was limited to specific contexts (Kearney and Merrill 2000). Some suggested this was

because amici didn’t always provide unique information as required by Rule 37 and instead

reiterated the litigants’ arguments (Spriggs and Wahlbeck 1997b).

More recent work demonstrates that respondents benefit more than petitioners from

amicus brief filings and overall the ideological direction of the brief is a strong predictor

of positive outcomes before the Court (Collins Jr 2004). As more liberal (conservative)

briefs are filed, the likelihood of a liberal (conservative) outcome increases regardless of

the ideological preferences of the justices (Collins 2008). Furthermore, as the total number

of amicus briefs increase the variance surrounding the justices’ votes also increases. Put

2https://www.law.cornell.edu/rules/supct/rule_37

87

differently, as the briefs in a case increases it becomes significantly more difficult to predict

how a justice will vote (Collins 2008). Furthermore, amicus briefs do not solely influence

case dispositions. They also influence how the law is crafted via opinion content. Majority

opinion authors will borrow content verbatim from amicus briefs while writing their opinions

(Collins Jr, Corley and Hamner 2015). Using language directly from an amicus brief in a

Supreme Court majority opinion and thus the law of the land.

This recent work confirmed what outside interests already believed, that amicus briefs

influence Supreme Court justices. However, this research has maintained a narrow focus,

primarily on the number of briefs and the ideological direction they make their argument

relative to the justices’ preferences. Most empirical quantitative research has failed to con-

sider the heterogeneity among amicus briefs. The lone exception is the innovative work

by Box-Steffensmeier, Christenson and Hitt (2013) that demonstrates when powerful, well-

connected interest groups like the ACLU file amicus briefs, the justices notice and these

groups are more likely to receive favorable outcomes compared to smaller outside interests.

That is, powerful interest groups have prestigious reputations and are known for filing useful

briefs. These findings are supported by clerks’ personal statements. After briefs filed by the

OSG, they devoted their attention to briefs filed by other government entities, the ACLU,

followed by professional associations, and prominent academics or attorneys (Lynch 2004).

Clearly not all amicus briefs are created equal and as the number of amicus briefs filed

continue to grow, clerks’ time to spend on each brief decreases and they must prioritize.

They do not have time to read every brief so they skim most and devote time to a select few

(Lynch 2004). Clerks triage the briefs and all case documents for their justices, therefore if

a busy law clerk only devotes one to two minutes on a brief, it is likely their boss will do the

same, or less (Lynch 2004, 55). On first glance clerks claim they can uncover whether a brief

is repetitive, which annoys them for wasting their time. Briefs that repeat the parties’ claims

or one another’s argument not only violate Rule 37 but they are inefficient. Conversely, briefs

88

that warrant the clerks’ or even the justices’ attention are more efficient, relatively speaking.

I argue that these amicus briefs that garner more attention impact Supreme Court case

outcomes more than a general brief that is skimmed and discarded.

Unfortunately, the veil of secrecy that plagues all Supreme Court scholars prevents me

from examining which amicus briefs are read versus skimmed by the clerks as they are

received, and similarly, which briefs are read by the justices. To determine which briefs

are worthy of the justices’ time versus which ones are deemed unhelpful or repetitive I will

use the next best thing, bench memoranda. Bench memos are written by Supreme Court

law clerks after a case is granted review but before oral arguments.

In order to prepare

justices for their decision on the merits, bench memoranda summarize the facts of the case,

each litigant’s contentions, amici contentions, and provide an analysis of their claims. Some

briefs’ legal arguments comprise pages where other briefs are limited to the name of the

amicus filer followed by a colon and the word “repetitive.” And worse case scenario for an

outside interest, sometimes amicus briefs do not even receive a mention, the clerk will start

the amicus section by stating they will only discuss unique briefs and omit repetitive briefs.

Additionally, while clerks summarize each amicus brief and how it relates to the larger

issues in the case they will also add their own commentary. Typical examples that suggest

a quality brief worthy of their justice’s time include:

– This is an excellent brief, worthy of substantial attention.3
– This is a nice little amicus brief that you might want to look at.4

– This brief reiterates petitioner’s arguments, except far more clearly than petr itself.
3Brief filed on behalf of the petitioner by the National Association of Bond Lawyers
discussed on pg. 15 in the bench memorandum in C & A Carbon Inc. v. Town of Clarkstown
(92-1402). Located in the Supreme Court Case Files (Box 637, Folder #6) of Justice Harry
A. Blackmun in the Library of Congress, Washington D.C.

4Brief filed on behalf of the petitioners by two individuals who previously served as jurors
in a capital case discussed on pg. 15 in the bench memorandum in Jonathan D. Simmons v.
South Carolina (92-9059). Located in the Supreme Court Case Files (Box 643, Folder #3)
of Justice Harry A. Blackmun in the Library of Congress, Washington D.C.

89

The brief adds two important arguments as well.5

Some briefs, however, do not deserve such praise. Some clerks air their frustration of

their wasted time in reading certain briefs and will include commentary on unhelpful briefs.

The majority of the commentary is on the repetitive nature of the amicus briefs but also

commonly includes the following:

– The nurse’s brief is interesting, but not relevant to the question in front of the Court.6
– I found nothing helpful here.7
– This brief is disjointed and adds nothing new.8

These comments can be used as a proxy to understand which briefs are effective and useful

for the justices’ decision making and which briefs are ineffective and simply waste their time.

They provide for a way to determine the heterogeneity among amicus briefs and determine

if quality briefs influence justices more than sheer quantity.

4.2 Assessing Effective Versus Ineffective Amicus Briefs

I theorize that positive commentary regarding amicus briefs in bench memoranda repre-

sent effective amicus briefs and negative comments represent ineffective amicus briefs. Before
5Brief filed on behalf of the petitioner by the Office of the Solicitor General discussed on
pg. 15 in the bench memorandum in Estelle v. McGuire (90-1074). Located in the Supreme
Court Case Files (Box 590, Folder #1) of Justice Harry A. Blackmun in the Library of
Congress, Washington D.C.

6Brief filed on behalf of the respondent by the American Nurses Association discussed on
pg. 34 in the bench memorandum in Waters v. Churchill (92-1450). Located in the Supreme
Court Case Files (Box 638, Folder #2) of Justice Harry A. Blackmun in the Library of
Congress, Washington D.C.

7Brief filed on behalf of the petitioner by La Plata County discussed on pg. 18 in the bench
memorandum in County of Yakima v. Confederated Tribes & Bands of Yakima Indian Nation
(90-408). Located in the Supreme Court Case Files (Box 584, Folder #4) of Justice Harry
A. Blackmun in the Library of Congress, Washington D.C.

8Brief filed on behalf of the petitioner by the Association of the Bar of City of New
York discussed on pg. 12 in the bench memorandum in Medina v. State of California (90-
8370). Located in the Supreme Court Case Files (Box 596, Folders #6-8) of Justice Harry
A. Blackmun in the Library of Congress, Washington D.C.

90

I can examine the effect of accounting for heterogeneity in amicus briefs I must validate that

this commentary actually represents effective and ineffective briefs by examining if these

descriptions relate to other factors we associate with quality amici. To do this I use original

data from Justice Harry Blackmun’s bench memoranda located at the Library of Congress

in Washington, D.C. Each page of the bench memoranda were photographed and then pro-

cessed using optical character recognition software to convert the images to text files. Next,

each amicus was tagged using extensible markup language tags, commonly referred to as

XML tags to extract the discussion of each amicus brief. The commentary on each brief

was then hand coded as 1 for a positive comment, 0 for no comment, and -1 for a negative

comment. This coding of the type of comment will be used as the dependent variable in the

validation model.

Additional information was also extracted from the bench memos to use to determine

how this commentary relates to the effectiveness of the briefs. First is a count of how many

interests cooperated to file a brief. 90% of clerks agree that cooperation is preferred because

it reduces the amount of briefs filed for a particular case and because it reduces repetitive

briefs. Some clerks suggest that briefs filed collaboratively between multiple parties enhances

the value of a brief because it demonstrates a shared viewpoint, however most do not spend

additional time on these briefs (Lynch 2004, 57, 61). The perceived efficiency of multifilers,

but lack of extra attention suggests that multifiler briefs should be negatively related to

negative commentary and may garner positive commentary.

The Office of Solicitor General (OSG) is the federal government’s attorney at the Supreme

Court and is known as a reliable and trusted source of information. “The Tenth Justice,” as

the OSG is often referred, files amicus briefs in order to inform the Court of the government’s

position in a case and the potential implications for the federal government in light of the

Court’s decision (Bailey, Kamoie and Maltzman 2005). Occasionally, the justices even ask

the OSG to file a brief to obtain more information. In fact, parties supported by an OSG

91

amicus brief “exhibit spectacular success rates in cases before the Supreme Court” (Segal

1988, 142) (see also O’Connor (1982)). Clerks label the OSG as the most important amicus

filer, not only because of the significance of representing the federal government, but also

because his briefs are well-researched and excellently written (Lynch 2004, 47). The OSG is

included in the model as a dichotomous indicator coded as 1 if the OSG wrote an amicus

brief or joined an amicus brief and 0 otherwise. The presence of an OSG amicus should

be positively related to quality and effective amicus briefs or positive commentary in bench

memoranda and negatively related to critical comments.

The OSG is the federal government’s attorney before the Court, but sometimes other

U.S. government attorneys submit briefs. Departmental agencies will file briefs in cases that

they will be particularly effected. The federal government experiences success as an oral

advocate before the Court (Johnson, Wahlbeck and Spriggs 2006), and has great resources

to support a well researched brief that the Court is interested in. Following the OSG and state

governments, non-OSG federal government filers are the next most important amici (Lynch

2004, 49). Given the Court’s institutional concern, like the OSG, the federal government’s

brief should be more likely to receive favorable comments and less likely to receive unfavorable

comments. This variable is a dichotomous variable coded as 1 if the filer is part of the federal

government and 0 otherwise.

Like the federal government, the justices also take an interest and concern as to how

their potential decision will impact state governments (Lynch 2004). Unlike the federal

government, however, there is significant variation in the quality of amicus briefs from states.

States increase their likelihood of success as amici when there are ten or more together on

a brief, but states individually do not fare well as amici. This is likely because states like

California and New York have more experienced state attorneys general so they are able to

help less experienced states write an effective brief. Given the special attention yet wavering

quality it is unclear how effective or ineffective states’ amicus briefs are before the Supreme

92

Court. To examine this relationship I include a variable coded as 1 if a state government

appears on an amicus brief and 0 if no state is present.

To determine if effective amici are simply a reflection of Justice Blackmun’s ideological

alignment with their position, or if amici can provide a quality legal argument regardless of

their position, I include a measure of ideological compatibility between the litigant an amicus

filed in favor of and Justice Blackmun. To code this variable, first I determined whether the

petitioner (respondent) represented the liberal (conservative) position based on the direction

of the lower court’s ruling. If the litigant represents the liberal position, any amici filed on

behalf of the liberal litigant are coded as the negative value of Justice Blackmun’s Martin

and Quinn (MQ) score for that term. Any amici favoring the conservative litigant are then

coded as Justice Blackmun’s MQ score (Martin and Quinn 2002; Johnson, Wahlbeck and

Spriggs 2006). I expect that amici that Justice Blackmun ideologically agrees with will be

more likely to be perceived as effective and those he disagrees with as ineffective because

they provide additional legal justification to vote in his desired direction.

Case salience is another factor that could potentially impact the quality or effectiveness of

amici. Salient cases capture more attention, typically because they are interesting and they

have the potential to impact more people. As a result, more outside interests are compelled

to get involved. The greater number of amicus briefs should yield a greater likelihood of more

effective briefs and a lower likelihood of ineffective briefs. To capture case salience I utilize

Clark, Lax and Rice’s (2015) measure of early case salience. Unlike traditional measures of

salience that examine media coverage after a decision has been released by the Court, Clark,

Lax, and Rice’s (2015) early salience measure examines media coverage after a case has been

granted review but prior to oral arguments, which is the same period when amicus briefs are

being read and evaluated.

I also include an additional control variable in the model to determine whether an amicus

was filed on behalf of the petitioner. Justices are predisposed to favor the petitioner in most

93

cases because The Court overturns two-thirds of the lower court decisions it reviews (Epstein

et al. 2019). So to control for this favoritism and amici that file on behalf of that litigant,

I include a dichotomous indicator that equals 1 when the amicus argues for the petitioner

and 0 for the respondent.

4.3 Method and Results

Table 4.1: Multinomial Logistic Regression Model of Effective and Ineffective Amici

Multifiler

OSG

Federal Government

State Government

Ideological Compatibility

Case Salience

Petitioner

Number of Amici Filed for Each Litigant

Total Amici Filed

Constant

N
AIC
BIC
Log Likelihood
Chi-Squared
Standard errors in parentheses
∗ p < 0.05

Negative Positive
Comment Comment
-0.358∗
(0.144)
-1.181
(1.297)
-0.641
(0.939)
0.405
(0.430)
-0.117
(0.103)
-0.029
(0.173)
0.226
(0.307)
-0.219
(0.123)
0.048
(0.069)
-2.688∗
(0.327)
2634

0.003
(0.009)
1.065∗
(0.524)
-0.099
(0.483)
-0.911∗
(0.482)
0.103
(0.103)
0.190
(0.168)
-0.216
(0.189)
-0.080
(0.083)
-0.033
(0.053)
-2.496∗
(0.240)

1581.218
1698.743
-770.609
60.952

The dependent variable in this model is a nominal variable with three categories: neg-

ative comment, no comment, or positive comment. Because the coding scheme has three

categories, I model this using a multinomial logistic regression, which is best for nominal

94

dependent variables. Multinomial regression coefficients reveal the likelihood of a particular

outcome of the dependent variable relative to a base category. In Table 4.1 the base category

is no comment. The second column displays the likelihood of a negative comment for each

independent variable relative to the baseline, and the third column shows the likelihood of a

positive comment relative to no comment at all. The data consist of 2,634 amicus briefs filed

in all cases with a bench memo between the 1986-1993 terms. To account for similarities

that may occur at the case level, I cluster on the case.

The first row and second column of Table 4.1 reveals that as more groups join an amicus

brief the likelihood of a negative comment versus no comment at all significantly decreases.

To more clearly demonstrate this relationship the average marginal effects are shown in

Figure 4.1. Figure 4.1 shows the average marginal effect of multiple groups cooperating on

a brief on each amicus type while taking into account all relative odds among categories.

Multifilers are significantly less likely than solo group-filers (nearly 1%) of receiving a negative

or ineffective comment. Multifilers, however, do not gain a significant advantage as being an

effective amicus brief.

The next row of Table 4.1 represents the effectiveness of the OSG as an amicus. As

expected, when the OSG is an amicus a clerk is less likely to leave a negative comment relative

to the base category, but this result is not statistically significant. There is a statistically

significant increase in the likelihood of a positive comment when the OSG files or joins an

amicus brief. This relationship is further demonstrated by the left plot in Figure 4.2 that

shows the average marginal effect of the OSG filing an amicus brief on the effectiveness and

ineffectiveness of the brief. If an amicus brief is filed by the OSG, it is 7% more likely to

be a useful and effective brief for deciding the case than a non-OSG brief. The right plot in

Figure 4.2 displays the marginal effect of an amicus brief filed by the state government on

the different types of amicus briefs. As the clerks suggested, and research has shown, states

do not write effective briefs. A brief submitted by a state government is 3% less likely to

95

Figure 4.1: Average Marginal Effect of Multifiler on Amicus Type - This plot shows the average
marginal effect of multifiler on three types of amicus briefs. Vertical lines represent 95% confidence intervals

receive a positive comment or be deemed an effective brief.

The other predictors in the model do exert a significant effect on brief type. Briefs

filed by the federal government are not more or less effective than the typical non-federal

government amicus. In fact, negative values in both columns suggest the federal government

is less likely to receive a negative or positive comment making their briefs unworthy of

mentioning, however this relationship is not statistically significant.

Ideological compatibility and case salience do not significantly predict amicus commen-

tary but all coefficients are in the expected direction for positive and negative comments.

Relative to the base category, the likelihood of deeming an amicus effective increases as ide-

ological compatibility increases and case salience increases. Similarly the probability of an

ineffective brief is negatively related to both of these variables. This suggests that an amicus

does not have to support Justice Blackmun’s policy preferences in order to be effective. Also,

96

−0.010.000.01IneffectiveNoCommentEffectiveAmicus Brief TypeMarginal Effect of Multifiler BriefMulifiler BriefFigure 4.2: Average Marginal Effect of OSG and State Government on Amicus Type - The left
plot shows the average marginal effect of an amicus brief filed by the OSG on amicus type. The right plot
shows the average marginal effect of an amicus brief filed by a state government on amicus type. The vertical
lines are 95% confidence intervals.

non-salient cases can also yield effective amici. None of the control variables significantly

impact the likelihood of an amicus being effective or ineffective suggesting that the number

of amici in a case or simply being the favored party impacts the effectiveness of the briefs

filed.

The evidence for positive comments representing effective amici is supported with both

the OSG being favored and state governments not being favored. The evidence for negative

comments relating to ineffective amici is still supported, but somewhat less with mutlifiler

briefs being negatively related to a critical comment. This is not surprising, however. While

it is important to point out effective briefs for their justice to read, it is less consequential

that Justice Blackmun know which briefs are bad or to avoid since he is unlikely to spend

much time on them anyway. In the section that proceeds I will examine how effective and

ineffective briefs impact Supreme Court case outcomes.

97

−0.10−0.050.000.050.100.15IneffectiveNoCommentEffectiveAmicus Brief TypeMarginal Effect of an OSG BriefOSG Brief−0.050−0.0250.0000.0250.050IneffectiveNoCommentEffectiveAmicus Brief TypeMarginal Effect of a State Government BriefState Government Brief4.4 Effective and Ineffective Amicus Briefs Influence on Supreme

Court Case Outcomes

Now that I am able to account for some of the heterogeneity in amicus brief effectiveness

I must determine how they impact case outcomes. That is, do amicus briefs that are higher

quality or effective exert a greater impact on case outcomes compared to less effective amici?

To do this, I merge my unique data discussed in the previous section with Collins’s (2008)
data of amicus influence on individual justices’ voting behavior.9

I reexamine Collins’s

(2008) original model in which he finds that the number of liberal amicus briefs significantly

increases the likelihood of a justice placing a liberal vote, and similarly, the likelihood of a

justice placing a liberal vote decreases as the number of conservative amicus briefs increases,

regardless of a justice’s individual ideology.

The dependent variable in the model is a justice’s propensity to vote liberally. It is a

dichotomous indicator that takes on the value of 1 if a justice places a liberal vote and

0 if the vote is conservative. This variable was coded using the Supreme Court Database

decision direction variable (Epstein et al. 2019). I follow Collins (2008) and cluster on the

case to account for case level similarities. Following others who have used archival data from

particular justices, I exclude Justice Blackmun’s votes from the analysis (Johnson, Wahlbeck

and Spriggs 2006). This prevents any endogeniety that may arise. Justice Blackmun is the

only justice who reads the bench memos crafted by his clerks. As a result, he could be

predisposed to favor the outcome advocated by a particular amicus brief that his clerks

called his attention to. Preventing possible endogeneity by removing Blackmun from the

model will also be a more conservative test of my hypothesis because it will be more difficult

to find an effect on the effective and ineffective amici.

Collins’s (2008) has five primary variables of interest that I also include in my model.
9These data were originally acquired form http://www.psci.unt.edu/~pmcollins/
data.htm and consist of data originally collected by Collins (2008) and Kearney and Merrill
(2000).

98

First, he includes a count of the number of liberal amicus curiae briefs (ACBs) filed in a

case. He also includes a count of the number of conservative amicus briefs filed in a case.

Ideological direction of the brief is determined by the lower court’s decision direction. If the

lower court decided a case liberally (conservatively), any briefs filed on behalf of the petitioner

are coded as conservative (liberal) and any briefs filed in favor of the respondent are coded as

liberal (conservative). The author hypothesizes that as the number of liberal briefs increases,

the likelihood of a justice voting liberally in that case also increases and as the number of

conservative amicus briefs increases, the likelihood of voting liberally decreases. While the

author finds support for this hypothesis, he originally hypothesized that justices will be more

effected by amici they are predisposed to favor. To account for this relationship the justices’

Martin and Quinn scores are also included in the model and interacted with the number of
conservative and liberal amicus briefs (Martin and Quinn 2002).10

Following the literature, Collins (2008) also includes an variable to indicate whether the

OSG filed a liberal or conservative amicus brief. As previously discussed, the OSG is a trusted

resource that is known for crafting well-written and well-researched merits and amicus briefs

(Bailey, Kamoie and Maltzman 2005; Lynch 2004; O’Connor 1982; Segal 1988). Liberal

OSG amicus briefs should increase the likelihood of a liberal vote and conservative OSG

amicus briefs should decrease the likelihood of a liberal vote. Additionally, Collins (2008)

also controls for the resource level of each litigant. Research on the “haves and have nots”

before the Court stems from the theory that litigants with more resources, like the U.S.

government or large businesses, have more resources to hire better attorneys to represent

them (Songer, Sheehan and Haire 1999). Lastly, he controls for the lower court’s decision

direction. As previously mentioned, the Supreme Court is prone to reversal, to account for
10Martin and Quinn scores typically take on negative values to represent liberal judicial
behavior and positive values indicate conservative voting. Collins (2008) transforms MQ
scores from the traditional scale to range from 0 to 10 with smaller values still relating to
liberal voting and larger values related to conservative behavior. These are the MQ scores I
also use.

99

this he includes an variable to represent whether the lower court’s decision was liberal (1)

or conservative (0) (Epstein et al. 2019).

To account for the number of effective and ineffective amicus briefs I include a count of the

number of effective liberal briefs, effective conservative briefs, ineffective liberal briefs, and

ineffective conservative briefs. I expect that effective liberal amicus briefs should increase the

likelihood of a liberal vote and ineffective amicus briefs decreases the likelihood of a liberal

vote. Conservative amicus briefs should have the opposite effect. Effective conservative

amicus briefs should decrease the likelihood of a liberal vote and ineffective conservative

amicus briefs should increase the likelihood of a liberal vote. Like Collins (2008), I believe

that effective briefs should have a greater effect on the justices who are predisposed to be

effected by them. Similarly, if justices read ineffective briefs advocating against their policy

preferences, they should be even less impacted by its filing. To account for this relationship I

also include an interactive term between my four effective and ineffective briefs and justices’

MQ scores.

I also include an additional control in the model. I control for liberal case strength. As

I have demonstrated throughout this dissertation, case strength has a significant impact on

multiple aspects of judicial decision making, including the outcomes on the merits. A justice

may vote liberally because the law more strongly supports that litigant’s case.

4.5 Method and Results

The dependent variable is dichotomous coded as 1 if a justice voted liberally and 0 if he

voted conservatively. To model this I use a logistic regression model with standard errors

clustered on the case. The results of the effectiveness model are shown in the last column in

Table 4.2. The middle column displays the results from Collins’s (2008) original model using

the same sample of data. Because these are maximum likelihood regression coefficients their

significance and direction is interpretable by the table but their effect size is not. Instead of

spending time interpreting these coefficients I will begin with primary results displayed in

100

Figure 4.3.

(a) Likelihood of a Liberal Vote Across MQ Scores
for an Effective vs. Regular Liberal Amicus Brief

(b) Likelihood of a Liberal Vote Across MQ Scores
for an Ineffective vs. Regular Liberal Amicus Brief

(c) Likelihood of a Liberal Vote Across MQ Scores
for an Effective vs. Regular Conservative Amicus
Brief

(d) Likelihood of a Liberal Vote Across MQ Scores
for an Ineffective vs. Regular Conservative Amicus
Brief

Figure 4.3: Likelihood of a Liberal Vote Across MQ Scores Comparing Effective and Ineffective
Amicus Briefs to Regular Amicus Briefs - Each plot has the 95th percentile of the transformed Martin
and Quinn scores along the x-axis an d the likelihood of a liberal vote on the y-axis. The orange line in each
plot represents a single conservative or liberal amicus brief that is neither effective no ineffective. Green lines
represent an effective liberal or conservative amicus brief and purple lines represent a single ineffective liberal
or conservative brief. The vertical lines around that estimates are 95% confidence intervals. All continuous
variables are held at their means and dichotomous variables are held at their modes.

101

0.00.20.40.60.81.023456789Transformed Martin & Quinn ScoreLikelihood of a Liberal VoteEffective Liberal Amicus Brief0.00.20.40.60.81.023456789Transformed Martin & Quinn ScoreLikelihood of a Liberal VoteIneffective Liberal Amicus Brief0.00.20.40.60.81.023456789Transformed Martin & Quinn ScoreLikelihood of a Liberal VoteEffective Amicus BriefRegular Amicus BriefEffective Conservative Amicus Brief0.00.20.40.60.81.023456789Transformed Martin & Quinn ScoreLikelihood of a Liberal VoteIneffective Amicus BriefRegular Amicus BriefIneffective Conservative Amicus BriefTable 4.2: Logistic Regression Model of the Justices’ Propensity to Vote Liberally

Collins’ Model Effectiveness Model

Number of Liberal ACBs

Number of Conservative ACBs

Transformed Martin & Quinn Score

MQ Score * Liberal ACBs

MQ Score * Conservative ACBs

Liberal OSG ACB

Conservative OSG ACB

Liberal Litigant Resource Level

Conservative Litigant Resource Level

Lower Court Decision Direction

Number of Effective Liberal ACBs

0.119
(0.068)
.034
(0.072)
-0.354∗
(0.025)
-0.006
(0.009)
-0.008
(0.010)
0.392
(0.205)
-0.626∗
(0.140)
0.025
(0.026)
-0.056
(0.029)
-0.449∗
(0.117)

0.109
(0.086)
0.027
(0.084)
-0.407∗
(0.028)
-0.012
(0.012)
-0.006
(0.012)
0.520∗
(0.223)
-0.478∗
(0.139)
0.057∗
(0.027)
-0.071∗
(0.028)
-0.228
(0.118)
1.512∗
(0.572)
-0.367
(0.493)
-0.659∗
(0.185)
0.178
(0.208)
-0.192∗
(0.080)
0.024
(0.079)
0.089∗
(0.025)
-0.018
(0.030)
0.436∗
(0.040)
1.340∗
(0.380)
6899

7790.959
7927.742
-3875.480
481.637
71.84%

Number of Effective Conservative ACBs

Number of Ineffective Liberal ACBs

Number of Ineffective Conservative ACBs

Number of Effective Liberal ACBs * MQ Score

Number of Effective Conservative ACBs * MQ Score

Number of Ineffective Liberal ACBs * MQ Score

Number of Ineffective Conservative ACBs * MQ Score

Liberal Case Strength

Constant

N
AIC
BIC
Log Likelihood
Chi-squared
Percent Correctly Classified
Standard errors in parentheses
∗ p < 0.05

2.754∗
(0.369)
6899

8458.755
8533.985
-4218.377
405.176
67.50%

102

Figure 4.3 shows the likelihood of a justice placing a liberal vote given a typical amicus

brief, an amicus that receives no positive or negative commentary, versus an effective or

ineffective amicus brief across the range of the justices’ ideological preferences. Starting

with the top left plot in the figure, the orange line represents a single liberal amicus brief

and the green line represents a single liberal amicus brief that was deemed to be effective.

Smaller values on the x-axis represent liberal justices and larger values are justices with

a conservative voting record. When a typical amicus brief is filed on behalf of the liberal

litigant, a liberal justice is predicted to vote liberally 81% of the time. When a liberal brief

is determined to be an effective amicus brief, the likelihood of the same liberal justice voting

liberally is 94%. This is a statistically significant increase of 13 percentage points.

This relationship persists across the ideological spectrum. For conservative justices rep-

resented by a transformed MQ score of 9, a regular liberal amicus brief only compels them

to vote liberally 27% of the time. Yet, if there is an effective liberal amicus brief filed for

the same case, the likelihood of a conservative justice voting liberally increases to 57%. This

is a significant increase of 30 percentage points, a 111% change. While a liberal justice is

predisposed to vote liberally regardless of the amici filed in the case, an effective liberal am-

icus brief significantly increases this likelihood. Contrary to my hypothesis, effective amici

are most influential on conservative justices. A single liberal effective amicus brief makes a

conservative justice more likely to vote liberally than conservatively. Thus, effective amicus

briefs actually have a greater impact on justices who are predisposed to disagree with them.

The bottom left plot in Figure 4.3 shows the same relationship as above, but compares an

effective conservative amicus brief to a regular conservative amicus brief. This relationship

is in the expected direction with an effective conservative amicus brief making justices less

likely to place a liberal vote than a regular conservative amicus. This effect, however, is not

statistically different from a regular conservative amicus.

The right side of Figure 4.3 displays the relationship for ineffective amici. The top right

103

plot shows the relationship between a regular liberal amicus brief (orange line) versus an

ineffective liberal amicus brief (purple line). An ineffective amicus brief makes all justices

significantly less likely to vote liberally than a regular amicus brief. For a liberal justice

with a transformed MQ score of two, a single amicus brief for leads to a 80% likelihood of

voting liberally, yet for the same case if the liberal brief is an ineffective brief this likelihood

drops to 70%. This same significant relationship persists for conservative justices, with an

ineffective amicus brief reducing the likelihood of a liberal vote from 27% to 17% compared

to a regular amicus brief. This means, that a poorly written, poorly-researched, or repetitive

brief can actually have the opposite effect of the interest groups’ intentions by hurting their

likelihood of policy success before the Court. The bottom figure shows the same relationship

as above but with an ineffective conservative amicus brief compared to a regular conservative

amicus brief. Again, like effective conservative briefs, the relationship between the two types

of briefs are not significant but in the expected direction.

It may seems surprising that the predicted relationship only exists for liberal amici and

not conservative interest groups, however during this time period the majority of the non-

government amicus filers that clerks from both liberal and conservative chambers cited as

submitting quality, well-written and well-researched briefs were from liberal interests (Lynch

2004). Specifically, the three primary non-governmental interest groups that clerks cited

said they considered and reviewed more carefully were the ACLU, NAACP, and the AFL-

CIO. The late 1980s and early 1990s preceded/were at the onset of the proliferation of the

conservative legal movement that the Court experiences today (Teles 2012).

Returning to table, I find some discrepancies between Collins’s (2008) original model and

my model. As hypothesized, a liberal OSG amicus brief significantly increases the likelihood

of a liberal vote, which fails to have a significant effect in the original model. Holding all

variables at their means and dichotomous variables at their modes, the presence of a liberal

OSG amicus brief increases the likelihood of a liberal vote from 46% to 56%. Similarly, a

104

conservative OSG amicus brief significantly reduces the likelihood of voting liberally from

46% to 36%.

Also as hypothesized, liberal litigants’ resource levels significantly increase the likelihood

of obtaining liberal votes. In the Collins (2008) model this variable is in the hypothesized

direction but fails to have a significant effect on the dependent variable. In both models

conservative litigant resources significantly decrease the likelihood of a liberal vote. Addi-

tionally, the original model finds that the lower court decision direction has a statistically

significant and negative effect on the likelihood of a liberal vote. That is, justices are sig-

nificantly less likely to vote liberally if the lower Court’s decision was liberal, however the

effectiveness model does not find this to be the case. Lastly, liberal case strength is a positive

and significant predictor of a justice’s liberal vote. Holding all continuous variables at their

mean and dichotomous variables at their modes, the likelihood of a liberal vote when the

conservative litigant’s argument is strongly supported by law and legal reasoning is 25%.

The likelihood increases to 55% when the liberal litigant is more strongly supported by law.

This is an increase of 30 percentage points or 120% change.

The bottom of the table shows the fit statistics of each model. There are reductions

in both the Aikake and Bayesian Information Criterion (AIC and BIC). Both fit statistics

penalize for the inclusion of additional variables, and even after this penalty the AIC and

BIC both are reduced greater than 10, which represents a significantly better fit (Long 1997).

Additionally, the effectiveness model correctly classifies 71% of the observations compared

to the original model that has a classification rate of 67.5%. These fit statistics demonstrate

that it is important to take into account not only the number of amicus briefs, but also the

different types of amicus briefs when examining their influence on judicial decision making.

4.6 Discussion and Conclusion

In this paper I use novel data to demonstrate that heterogeneity exists in the effectiveness

of amicus curiae briefs, and accounting for this heterogeneity is significant in understanding

105

how amicus briefs impact Supreme Court justices’ decision making. Supreme Court Rule 37

requires all amicus filers to bring justices’ attention to a relevant matter not already discussed

by the merits briefs. I find that when briefs genuinely bring a new legal perspective they are

more effective. More importantly, effective briefs can significantly influence how a justice

votes. With the current, more conservative Court today, it is critical that a high quality

liberal amicus brief carries the influence to make a conservative justice more likely to place

a liberal over a conservative vote in a case.

I also bring to light new findings on how amici can negatively impact a litigant’s chances

for success.

Ineffective liberal amicus briefs decrease the likelihood of liberal votes, and

therefore the ability to obtain a majority. Previous research always suggested that more is

more, but in actuality quality amicus briefs are much more important than sheer quantity

at the merits stage. Practitioners before the Court have revealed that in recent years coor-

dination between litigants and outside interests has grown to become much more prevalent

and systematic (Ward 2007), and this synchronization will likely continue to grow as the

length of litigants’ briefs continues to decrease (SCOTUS 2019). Litigants must consider if

a group wanting to file an amicus brief can be trusted to submit a quality brief that offers

information and legal arguments that are absent from their merits brief.

There are, of course, limitations to my findings. First, I only examine the Court from

1986-1993, which captures the last term of the Burger Court and the beginning of the

Rehnquist Court. While this analysis will eventually date back to 1972, this is still a limited

period. It really only captures the beginning of the rise of amicus participation before the

Court and the onset of groups working to file briefs together (Collins Jr 2004; Epstein et al.

2019). To put it into perspective, between the October term of 1986 and 1993 there were

over 2,500 amicus briefs discussed in Justice Blackmun’s bench memoranda; in the October

2015 term alone there were over 1,000 briefs filed. Strategies and the way outside interests

research and craft their briefs may have changed since the 1993 term. That being said, Rule

106

37, the rule governing amicus participation has not changed and therefore the characteristics

that make an effective amicus brief should remain the same. This research is the first step in

providing a better understanding of what makes an effective amicus brief and how outside

interests can best achieve their preferred outcomes before the United States Supreme Court.

Next, there are additional features related to amicus brief quality and effectiveness that

must be considered. I was able to test general features of amicus briefs, namely if the filer

is from the OSG, the federal government, or a state. This fails to account for experienced

members of the Supreme Court Bar. In interviews with clerks Lynch (2004) reveals that

clerks will spend extra time on briefs from well-known academics or attorneys. This is

evidenced in the bench memos as well. For example, in Turner Broadcasting Systems, Inc.,

et al. v. Federal Communications Commission (1993) a clerk commented, “The most talked-

about amicus brief in this case is the phone companies’ brief filed by Laurence Tribe —
with Ken Starr, Michael Kellogg, and Michael McConnell on brief.”11 This along with

accounting for the frequency in which other interest groups submit briefs, and how well

they are connected will provide more insight as to what is important in crafting an effective

amicus brief (Box-Steffensmeier, Christenson and Hitt 2013).

Moving forward, it is also important to also consider attributes of the amicus itself. For

example, is the amicus well written? This frequently appears as one of the comments of

an effective amicus brief. As hard-working law clerks grind through piles of briefs, it would

make sense that those that are easy to read would be effective. In fact, there are online

forums and books discussing how firms can make their briefs more readable and check their
brief’s readability (Guberman 2011).12 This fits with other research that demonstrates merit

brief quality, including readability, significantly influences opinion content (Feldman 2015),
11Brief filed on behalf of the petitioner by the Laurence Tribe discussed in the bench mem-
orandum in Turner Broadcasting Sys., Inc., et al. v. Federal Communications Commission
(93-44). Located in the Supreme Court Case Files (Box 643, Folder #6) of Justice Harry A.
Blackmun in the Library of Congress, Washington D.C.

12https://tinyurl.com/y2x6g4ef

107

and the Court’s opinion clarity can help enhance compliance among numerous groups (Black

et al. 2016). Clear, quality writing in both of these instances increases influence.

108

CHAPTER 5

CONCLUSION

Much of the work on the Supreme Court acknowledges the need for a systematic measure

to represent the law and legal doctrine, but this has been widely unavailable, and even seemed

unattainable. I attempt to remedy that deficit by gathering bench memoranda in more than

3,020 cases and using those data to create a novel measure of legal quality. In so doing,

I provide researchers with the opportunity to finally include a measure of legal doctrinal

quality in their work. As I demonstrate his will provide an opportunity to test the influence

of legal doctrine, update past research that has been done in its absence, and incorporate

a proper indicator of the law in future work. Despite our best efforts to understand what

goes into judicial decision-making, it is still unclear how the law and party’s legal arguments

factor into this calculus. With my measure of case strength and the data provided from

bench memos, scholars will finally be able to understand how legal doctrine can constrain

or encourage justices in pursuing favorable outcomes, fundamentally transforming the way

we view the Court and how justices arrive at their decisions that can have significant policy

consequences.

109

APPENDICES

110

APPENDIX A

CHAPTER 1 APPENDIX

111

Below are examples of the case strength coding scheme. All text comes from originally

collected bench memos from Justice Harry Blackmun’s archives located at the Library of

Congress in Washington, D.C.

Strong Affirm Recommendation

• Levitt v. Committee for Public Education & Religious Liberty (1972)

– I strongly feel that the decision below should be affirmed...

• PFZ Properties Inc. v. Rene Alberto Rodriguez, et al. (1991)

– Petr argues that CA1 erred in holding that an arbitrary, capricious or illegal denial
of a construction permit to a developer by officials acting under color of state law
does not present a substantive due process claim under §1983.
I recommend
that you DIG this case because petitioner has failed to allege any underlying
constitutional right.

Weak Affirm Recommendation

• David H. Lucas v. South Carolina Coastal Council (1991)

– My recommendation is that the Court affirm, finding that the Beachfront Man-
agement Act fell within the category of safety and health regulations that did not
require compensation. If not, the Court could easily DIG the case. The perma-
nent taking claim is not ripe, and the temporary taking claim was not decided
below. As a final alternative, the Court could remand to the SC CT to decide
whether this Act constitutes a nuisance, and if not was there a temporary taking.

• Board of Governors of the Federal Reserve System v. Dimension Financial Corp., et

al. (1985)

– Were it not for the deference due the Board’s interpretation of the BHCA, resp[ondent]s

would win this case hands down. Deference, especially with regard to what is
necessary in order to "prevent evasions" of the Act, makes the case close. Still,
I think the Board has gone too far in departing from the plain language of the
statute, especially since Congress seems purposely to have chosen precise rather
than open-ended terms in defining "bank." The BHCA may well need amendment
along the lines recommended by the Board, but that is a task for Congress, not
for the Board. I therefore recommend affirmance.

Uncertain Recommendation

112

• Browning-Ferris Industries et al v. Kelco Disposal, Inc. et al. (1988)

– I have changed my mind several times concerning a final recommenda-
tion on this case and will probably change it again before oral argument.
As things now stand, I think that the CA2 should be upheld, and that the Court
should not apply the 8A’s Excessive Fine Clause to punitive damages. As a mat-
ter of due process, I think that it should be up to the states to determine what
standards are appropriate in setting limits on punitive damages, without interfer-
ence from the federal courts, even if it means standards which have developed as
a matter of common law, and which provide little guidance to juries. the question
of punitive damages is one that should be decided in state legislatures, and that a
decision from this Court requiring that legislatures enact standards, would have
the effect of altering the legislative “playing field” to an improper degree. On the
other hand, it might force the legislatures into acting. My gut feeling is that if
this Court changes the basic rules, and says that punitive damages are not per-
missible unless awarded according to definite standards, the status quo will likely
remain, meaning punitive damages will be unavailable in many states. But I may
be completely wrong. I plan to continue thinking about this case, and will
provide you with a supplemental memo if appropriate.

• Burrell, et al. v. McCray, et al. (1975)

– On the Eighth Amendment issue, I am up in the air. If the CA is correct, it

is not correct by much; if it is incorrect, it is not too far wrong.

• Santa Fe Industries v. Green (1976)

– I think very strong arguments can be made for both sides of the question
before the Court. Although I have some/preference for the resps, I recognize
that a decision going in favor of the petrs might well be correct. I might add that
I have absolutely no doubt that the majority of the Court will reverse the court
below.
I hope that the decision of the Court will rest on the language of rule
10b-5 itself, rather than on the language of § 10(b). If the statute is interpreted
to bar the suit, then I suspect that the Court will invalidate the SEC’s proposed
regulations -fcp deal with the problem.
If those regulations are particularized
and sufficiently focused, then perhaps many of the arguments raised against the
expansion of 10b-5 itself will be blunted. The SEC should have the power to
prevent the abuse of stockholders that "going private" allows.

Weak Reverse Recommendation

• Patterson v. Warner (1973)

113

– I don’t think that the statute in question runs afoul of due process, but the issue
is close. The proper equal protection analysis is the rational basis justification.
There is no rational basis for a double bond requirement, where the posting of a
single bond and/or court costs would suffice to protect the judgment awarded in
the JP court. I’m not sure why any bond requirement is necessary for appeals
from the JP court if they are unnecessary for appeals from a court of record, but
that inquiry need not be made here. The double bond provision of the statute is
violative of the Equal Protection Clause pursuant to Lindsey v Normet. in that
it penalizes the appellant rather than simply protecting the appellee. RECOM-
MENDATION: Reverse.

• Bryan v. Yellen (1979)

– I am troubled by giving res judicata effect to this collusive-looking, not-appealed
state court judgment on an important issue of federal policy in any event. No one
was in the suit to represent the federal interests or the interests e’j individuals
who might benefit from enforcement of the federal 33 law. I think a public policy
exception to res judicata should apply here and I would not give the state court
judgment preclusive effect.
In conclusion I think that the landowners should
succeed on the merits on the basis of §6 of the Project Act, although I would be
happy if the Court could find a way around §6. Section 6 is inconsistent with
the general federal policy of limiting concentration of land ownership where the
land is benefited by a federal water project. However, this inconsistency seems to
have been intended. I do not think the landowners should prevail on a fairness or
estoppel theory, and I do not think the Wilbur letter is entitled to much weight
in interpreting the statute. I think the resps have standing and are not barred by
res judicata. My weak recommendation is to reverse on the basis of §6.

Strong Reverse Recommendation

• James Kirkland Batson v. Kentucky (1985)

– I recommend that you vote to reverse the Kentucky Supreme Court’s holding,
but I recommend that you confront head on, and reject, the analysis in Swain
that suggests that the purposeful exclusion of black jurors within the confines of
a single trial cannot constitute a violation of the Equal Protection Clause. Since
Swain was decided, the Court has adopted a broader view of the function of the
jury and has come to see how entrenched and pernicious race-based attitudes are
in the justice system. The experience with Swain’s rule over the past generation
counsels its rejection.

• Ridgway, et al. v. Ridgway, et al. (1981)

114

– I conclude that the decision of the Maine Supreme Judicial Court should be
reversed. The Act gives servicemen complete freedom to name the beneficiaries
of their choice, and makes use of a broad anti-attachment provision to ensure that
those beneficiaries actually receive and enjoy the funds. This Court has held that
virtually identical provisions bar state law community property claims. Wissner
v. Wissner, 338 U.S. 655 (1950). The same analysis should be applied to the
state divorce decree at issue in this case.

115

APPENDIX B

CHAPTER 2 APPENDIX

116

Table B.1: Logistic Regression Model of the Likelihood of Bargaining

Legal Considerations
Weak Conflict

Uncertain

Weak Agreement

Strong Agreement

Policy Preferences
Author Distance

Coalition Distance

Strategic Interaction
Winning Margin

Cooperation

Contextual Controls
Political Salience

Legal Salience

Case Complexity

End of Term

Workload

Chief Justice

Freshman Author

Constant

Coefficient

(SE)
−0.277∗
(0.108)
.057
(0.099)
−0.173
(0.176)
−0.098
(0.147)
0.029∗
(0.003)
0.009∗
(0.002)
−0.008
(0.036)
−2.143∗
(0.607)

0.068
(0.044)
0.270
(0.155)
0.085∗
(0.023)
0.001∗
(0.000)
0.001
(0.008)
−0.371∗
(0.062)
−0.044
(0.082)
−2.367∗
(0.276)

N
AIC
BIC
Log Likelihood
Standard errors in parentheses
∗ p < 0.05

7563
7165.506
7227.885
−3573.753

117

Table B.2: Multinomial Logistic Regression Model of Justices’ Strategic Responses to the
Majority Opinion Author’s First Draft

0.010
(0.144)
0.285
(0.175)
−0.234
(0.249)
−0.004
(0.180)
0.035∗
(0.005)
0.010∗
(0.004)

0.053
(0.225)
−0.140
(0.356)
−0.346∗
(0.173)
0.010
(0.176)
0.027∗
(0.004)
−0.003
(0.012)

−0.683
(0.472)
0.288
(0.539)
0.672
(0.398)
0.523
(0.298)
0.042∗
(0.007)
−0.022
(0.013)
0.200∗
(0.056)
−1.689
(1.394)

Suggestion Threat Separate Concur Dissent
−0.809∗ −0.102
(0.250)
(0.252)
−0.092
−0.760
(0.343)
(0.406)
−0.532
0.006
(0.198)
(0.370)
−0.694∗
0.105
(0.180)
(0.296)
0.018∗
0.026∗
(0.005)
(0.005)
0.042∗
0.003
(0.007)
(0.007)
−0.159∗ −0.057
0.063
(0.056)
(0.054)
(0.038)
−2.583∗ −2.663∗
−1.520
(0.843)
(1.065)
(0.890)
−0.021
−0.068
0.188
0.146
(0.121)
(0.054)
(0.097)
(0.096)
0.569∗
−0.303
0.153
0.210
(0.366)
(0.180)
(0.243)
(0.317)
0.148∗
0.085
0.077
0.050
(0.100)
(0.043)
(0.079)
(0.044)
0.004∗
−0.001
0.001
0.077
(0.001)
(0.043)
(0.001)
(0.001)
0.031∗
−0.003
−0.006
−0.013
(0.025)
(0.016)
(0.010)
(0.018)
−0.876∗
−0.656∗ −0.639∗
0.158
(0.172)
(0.107)
(0.104)
(0.146)
0.543∗ −0.065
−1.160∗
0.359
(0.131)
(0.255)
(0.261)
(0.510)
0.099∗
0.137∗
0.036
0.053
(0.037)
(0.037)
(0.056)
(0.052)
−6.115∗ −3.330∗ −3.179∗ −5.313∗
(0.580)
(0.481)

0.071
(0.044)
−2.176∗
(0.508)
0.148∗
(0.075)
0.109
(0.142)
0.239∗
(0.065)
0.003
(0.001)
0.011
(0.029)
−0.811∗
(0.181)
0.030
(0.308)
−0.044
(0.054)
−4.945∗
(0.535)

(0.338)

(0.538)

Wait
−0.776∗
(0.225)
0.254
(0.178)
−0.126
(0.259)
−0.244
(0.161)
0.021∗
(0.005)
0.010
(0.014)
−0.145∗
(0.045)
−2.964∗
(0.973)

0.141
(0.097)
0.064
(0.267)
−0.115∗
(0.043)
0.002∗
(0.001)
0.001
(0.024)
−0.595∗
(0.180)
−0.301
(0.207)
−0.063
(0.047)
−4.203∗
(0.670)
7563
−6028.889
12075.777
12138.157

Legal Considerations
Weak Agreement

Uncertain

Weak Conflict

Strong Conflict

Policy Preferences
Author Distance

Coalition Distance

Strategic Interaction
Winning Margin

Cooperation

Contextual Controls
Political Salience

Legal Salience

Case Complexity

End of Term

Workload

Chief Justice

Freshman Author

Expertise

Constant

N
Log Likelihood
AIC
BIC
Standard errors in parentheses
∗ p < 0.05

118

Table B.3: Maltzman, Spriggs, and Wahlbeck’s Original Model

0.035∗
(0.005)
0.010∗
(0.004)

0.026∗
(0.004)
−0.003
(0.012)

0.042∗
(0.007)
−0.021
(0.012)
0.241∗
(0.053)
−1.734
(1.378)

Suggestion Threat Separate Concur Dissent
0.018∗
0.026∗
(0.005)
(0.005)
0.1∗
0.003
(0.007)
(0.007)
−0.197∗ −0.044
0.059
(0.048)
(0.038)
(0.051)
−2.572∗ −2.676∗
−1.517
(0.838)
(1.065)
(0.880)
−0.022
−0.078
0.198
0.147
(0.121)
(0.053)
(0.097)
(0.098)
0.563∗
−0.239
0.178
0.135
(0.363)
(0.185)
(0.251)
(0.319)
0.154∗
0.079
0.078
0.047
(0.100)
(0.043)
(0.081)
(0.044)
0.004∗
−0.001
0.001
0.001
(0.001)
(0.001)
(0.001)
(0.001)
0.030∗
−0.004
−0.006
−0.013
(0.025)
(0.016)
(0.010)
(0.017)
−0.891∗
−0.652∗ −0.644∗
0.159
(0.168)
(0.105)
(0.105)
(0.139)
0.539∗ −0.063
−1.109∗
0.355
(0.129)
(0.255)
(0.499)
(0.261)
0.101∗
0.137∗
0.036
0.052
(0.037)
(0.037)
(0.056)
(0.052)
−5.515∗ −3.373∗ −3.580∗ −5.152∗
(0.518)
(0.452)

0.070
(0.045)
−2.175∗
(0.508)
0.148∗
(0.075)
0.107
(0.139)
0.239∗
(0.065)
0.003
(0.001)
0.011
(0.029)
−0.811∗
(0.181)
0.030
(0.308)
−0.044
(0.054)
−4.945∗
(0.546)

(0.397)

(0.434)

Wait
0.020∗
(0.005)
0.009
(0.014)
−0.153∗
(0.043)
−2.959∗
(0.968)

0.141
(0.097)
0.049
(0.278)
−0.113∗
(0.043)
0.002∗
(0.001)
0.001
(0.024)
−0.593∗
(0.179)
−0.298
(0.206)
−0.063
(0.048)
−4.293∗
(0.603)
7563
−6038.711
12095.422
12157.802

Policy Preferences
Author Distance

Coalition Distance

Strategic Interaction
Winning Margin

Cooperation

Contextual Controls
Political Salience

Legal Salience

Case Complexity

End of Term

Workload

Chief Justice

Freshman Author

Expertise

Constant

N
Log Likelihood
AIC
BIC
Standard errors in parentheses
∗ p < 0.05

119

BIBLIOGRAPHY

120

BIBLIOGRAPHY

Bailey, Michael A. 2007. “Comparable Preference Estimates across Time and Institutions for
the Court, Congress, and Presidency.” American Journal of Political Science 51(3):433–
448.

Bailey, Michael A., Brian Kamoie and Forrest Maltzman. 2005.

“Signals from the Tenth
Justice: The Political Role of the Solicitor General in Supreme Court Decision Making.”
American Journal of Political Science 49(1):72–85.

Bailey, Michael A. and Forrest Maltzman. 2008. “Does Legal Doctrine Matter? Unpacking
Law and Policy Preferences on the U.S. Supreme Court.” The American Political Science
Review 102(3):369–384.

Bartels, Brandon L. 2009. “The Constraining Capacity of Legal Doctrine on the U.S. Supreme

Court.” The American Political Science Review 103(3):474–495.

Bartels, Brandon L. and Andrew J. O’Geen. 2014. “The Nature of Legal Change on the U.S.
Supreme Court: Jurisprudential Regimes Theory and Its Alternatives.” American Journal
of Political Science .

Bartels, Brandon L and Andrew J O’Geen. 2015. “The Nature of Legal Change on the US
Supreme Court: Jurisprudential Regimes Theory and Its Alternatives.” American Journal
of Political Science 59(4):880–895.

Binder, Sarah. 2015.

18(1):85–101.

“The Dysfunctional Congress.” Annual Review of Political Science

Biskupic, Joan, Janet Roberts and John Shiffman. 2014. “The Echo Chamber.” Reuters .

Black, Ryan C. and Christina L. Boyd. 2012. “The Role of Law Clerks in the U.S. Supreme

Court’s Agenda-Setting Process.” American Politics Research 40(1):147–173.

Black, Ryan C. and Christina L. Boyd. 2013. “Selecting the Select Few: The Discuss List and
the U.S. Supreme Court’s Agenda-Setting Process.” Social Science Quarterly 94(4):1124–
1144.

Black, Ryan C, Christina L Boyd and Amanda C Bryan. 2014. “Revisiting the Influence of

Law Clerks on the US Supreme Court’s Agenda-Setting Process.” Marq. L. Rev. 98:75.

Black, Ryan C and James F Spriggs. 2013. “The Citation and Depreciation of US Supreme

Court Precedent.” Journal of Empirical Legal Studies 10(2):325–358.

Black, Ryan C., Maron W. Sorenson and Timothy R. Johnson. 2013.

“Toward an Actor-
Based Measure of Supreme Court Case Salience Information-Seeking and Engagement
during Oral Arguments.” Political Research Quarterly 66(4):804–818.

121

Black, Ryan C., Matthew E. K. Hall, Ryan J. Owens and Eve Ringsmuth. 2015. The Role
of Emotional Language in Briefs Before the U.S. Supreme Court. SSRN Scholarly Paper
ID 2703875 Social Science Research Network Rochester, NY: .

Black, Ryan C. and Ryan J. Owens. 2009.

“Agenda Setting in the Supreme Court: The

Collision of Policy and Jurisprudence.” The Journal of Politics 71(3):1062–1075.

Black, Ryan C. and Ryan J. Owens. 2011. “Solicitor General Influence and Agenda Setting

on the U.S. Supreme Court.” Political Research Quarterly 64(4):765–778.

Black, Ryan C. and Ryan J. Owens. 2012a. “Consider the Source (and the Message): Supreme
Court Justices and Strategic Audits of Lower Court Decisions.” Political Research Quar-
terly 65(2):385–395.

Black, Ryan C and Ryan J Owens. 2012b. The Solicitor General and the United States
Supreme Court: Executive Branch Influence and Judicial Decisions. Cambridge University
Press.

Black, Ryan C, Ryan J Owens, Justin P Wedeking and Patrick C Wohlfarth. 2019. The
Conscientious Justice: How Supreme Court Justices’ Personalities Influence the Law, the
High Court, and the Constitution. Cambridge University Press.

Black, Ryan C, Ryan J Owens, Justin Wedeking and Patrick C Wohlfarth. 2016. US Supreme

Court Opinions and their Audiences. Cambridge University Press.

Black, Ryan C and Timothy R Johnson. 2017. “Conference Notes, Transcription, and Crowd

Sourcing.” Law and Courts Newsletter pp. 2, 6–8.

Black, Ryan C, Timothy R Johnson and Justin Wedeking. 2012. Oral arguments and coalition
formation on the US Supreme Court: A deliberate dialogue. University of Michigan Press.

Bonica, Adam, Adam S Chilton, Jacob Goldin, Kyle Rozema and Maya Sen. 2016.

Political Ideologies of Law Clerks.” American Law and Economics Review p. ahw012.

“The

Box-Steffensmeier, Janet M., Dino P. Christenson and Matthew P. Hitt. 2013. “Quality Over
Quantity: Amici Influence and Judicial Decision Making.” American Political Science
Review 107(03):446–460.

Brisbin, Richard A. 1996. “Slaying the Dragon: Segal, Spaeth and the Function of Law in
Supreme Court Decision Making.” American Journal of Political Science 40(4):1004–1017.

Budziak, Jeffrey and Daniel Lempert. 2015. “Assessing Threats to Inference with Simulta-
neous Sensitivity Analysis: The Case of US Supreme Court Oral Arguments.” Political
Science Research and Methods pp. 1–24.

Caldeira, Gregory A. and John R. Wright. 1988. “Organized Interests and Agenda Setting

in the U.S. Supreme Court.” The American Political Science Review 82(4):1109–1127.

122

Caldeira, Gregory A. and John R. Wright. 1990. “Amici Curiae before the Supreme Court:

Who Participates, When, and How Much?” The Journal of Politics 52(3):782–806.

Caldeira, Gregory A., John R. Wright and Christopher J. W. Zorn. 1999. “Sophisticated Vot-
ing and Gate-Keeping in the Supreme Court.” Journal of Law, Economics, & Organization
15(3):549–572.

Casillas, Christopher J, Peter K Enns and Patrick C Wohlfarth. 2011a. “How public opinion

constrains the US Supreme Court.” American Journal of Political Science 55(1):74–88.

Casillas, Christopher J, Peter K Enns and Patrick C Wohlfarth. 2011b. “How public opinion

constrains the US Supreme Court.” American Journal of Political Science 55(1):74–88.

Clark, Tom S and Drew A Linzer. 2014. “Should I use fixed or random effects?” Political

Science Research and Methods 3(2):399–408.

Clark, Tom S, Jeffrey R Lax and Douglas Rice. 2015. “Measuring the political salience of

Supreme Court cases.” Journal of Law and Courts 3(1):37–65.

Collins Jr, Paul M. 2004. “Friends of the Court: Examining the Influence of Amicus Curiae

Participation in US Supreme Court Litigation.” Law & Society Review 38(4):807–832.

Collins Jr, Paul M. 2007. “Lobbyists before the US Supreme Court: Investigating the Influ-

ence of Amicus Curiae Briefs.” Political Research Quarterly 60(1):55–70.

Collins Jr, Paul M, Pamela C Corley and Jesse Hamner. 2014. “Me Too: An Investigation

of Repetition in US Supreme Court Amicus Curiae Briefs.” Judicature 97:228.

Collins Jr, Paul M, Pamela C Corley and Jesse Hamner. 2015. “The influence of amicus curiae

briefs on US Supreme Court opinion content.” Law & Society Review 49(4):917–944.

Collins, Paul M. 2008. “Amici Curiae and Dissensus on the U.S. Supreme Court.” Journal

of Empirical Legal Studies 5(1):143–170.

Collins, Paul M, Pamela C Corley and Jesse Hamner. 2015. “The Influence of Amicus Curiae

Briefs on US Supreme Court Opinion Content.” Law & Society Review 49(4):917–944.

Collins, Todd A and Christopher A Cooper. 2012.

“Case salience and media coverage of
Supreme Court decisions: Toward a new measure.” Political Research Quarterly 65(2):396–
407.

Corley, Pamela C. 2008. “The Supreme Court and Opinion Content The Influence of Parties’

Briefs.” Political Research Quarterly 61(3):468–478.

Corley, Pamela C. 2010. Concurring Opinion Writing on the US Supreme Court. SUNY

Press.

Ditslear, Corey and Lawrence Baum. 2001. “Selection of Law Clerks and Polarization in the

US Supreme Court.” The Journal of Politics 63(03):869–885.

123

Durr, Robert H., Andrew D. Martin and Christina Wolbrecht. 2000.

“Ideological Diver-
gence and Public Support for the Supreme Court.” American Journal of Political Science
44(4):768–76.

Dwyer, Colin. 2018.

“A Brief History Of Anthony Kennedy’s Swing Vote — And The

Landmark Cases It Swayed.” NPR .
URL: https://www.npr.org/2018/06/27/623943443/a-brief-history-of-anthony-kennedys-
swing-vote-and-the-landmark-cases-it-swayed

Epstein, Lee, Andrew D. Martin, Jeffrey A. Segal and Chad Westerland. 2007. “The Judicial

Common Space.” Journal of Law, Economics, & Organization 23(2):303–325.

Epstein, Lee, Andrew D Martin, Kevin M Quinn and Jeffrey A Segal. 2007.

“Ideological
Drift Among Supreme Court Justices: Who, When, and How Important.” Nw. UL Rev.
101:1483.

Epstein, Lee and Jack Knight. 1998. “The Choices Judges Make.” Washington, DC: Con-

gressional Quarterly .

Epstein, Lee and Jack Knight. 2000. “Toward a Strategic Revolution in Judicial Politics: A

Look Back, A Look Ahead.” Political Research Quarterly 53(3):625–661.

Epstein, Lee and Jack Knight. 2013. “Reconsidering Judicial Preferences.” Annual Review

of Political Science 16(1):11–31.

Epstein, Lee and Jeffrey A Segal. 2000.

Political Science pp. 66–83.

“Measuring issue salience.” American Journal of

Epstein, Lee, Thomas G. Walker, Nancy Staudt, Scott Hendrickson and Jason Roberts. 2019.

“The U.S. Supreme Court Justices Database.”.

Feldman, Adam. 2015. Counting on Quality: The Effect of Merits Brief Quality on Supreme
Court Opinion Content. SSRN Scholarly Paper ID 2622603 Social Science Research Net-
work Rochester, NY: .

Franze, Anthony J. and R. Reeves Anderson. 2016. “In Unusual Term, Big Year for Amicus

Curiae at the supreme Court.” The National Law Journal .

George, Tracey E. and Lee Epstein. 1992.

“On the Nature of Supreme Court Decision

Making.” American Political Science Review 86(2):323–337.

Gibson, James L., Jeffrey A. Gottfried, Michael X. Delli Carpini and Kathleen Hall Jamieson.
2011. “The Effects of Judicial Campaign Activity on the Legitimacy of Courts: A Survey-
based Experiment.” Political Research Quarterly 64(3):545–558.

Gleason, Shane A. 2019. “Beyond Mere Presence: Gender Norms in Oral Arguments at the

US Supreme Court.” Political Research Quarterly p. 1065912919847001.

124

Guberman, Ross. 2011. Point Made: How to Write Like the Nation’s Top Advocates. Oxford

University Press.

Hall, Matthew EK. 2018. What Justices Want: Goals and Personality on the US Supreme

Court. Cambridge University Press.

Hansford, Thomas G. 2004.

“Information Provision, Organizational Constraints, and the
Decision to Submit an Amicus Curiae Brief in a U.S. Supreme Court Case.” Political
Research Quarterly 57(2):219–230.

Hansford, Thomas G and James F Spriggs. 2006. The Politics of Precedent on the U.S.

Supreme Court. Princeton University Press.

Hansford, Thomas G, James F Spriggs and Anthony A Stenger. 2013.

“The Information

Dynamics of Vertical Stare Decisis.” The Journal of Politics 75(4):894–906.

Harlan, John M. 1955. “What part does the oral argument play in the conduct of an appeal.”

Cornell LQ 41:6.

Harvey, Anna and Michael J Woodruff. 2013.

“Confirmation Bias in the United States
Supreme Court Judicial Database.” Journal of Law, Economics, and Organization
29(2):414–460.

Jacobi, Tonja and Dylan Schweers. 2017. “Justice, interrupted: The effect of gender, ideology,

and seniority at Supreme Court oral arguments.” Virginia Law Review pp. 1379–1485.

Johnson, Timothy R, David R Stras and Ryan C Black. 2014.

“Advice from the Bench
(Memo): Clerk Influence on Supreme Court Oral Arguments.” Marquette Law Review
98:21.

Johnson, Timothy R and James F Spriggs. 2007. “Oral Advocacy Before the United States

Supreme Court: Does it Affect the Justices’ Decisions.” Wash. UL Rev. 85:457.

Johnson, Timothy R., Paul J. Wahlbeck and James F. Spriggs, II. 2006.

“The Influence
of Oral Arguments on the U.S. Supreme Court.” The American Political Science Review
100(1):99–113.

Johnson, Timothy R, Ryan C Black, Jerry Goldman and Sarah A Treul. 2009. “Inquiring
minds want to know: Do justices tip their hands with questions at oral argument in the
US supreme court.” Wash. UJL & Pol’y 29:241.

Johnson, Timothy R, Ryan C Black and Justin Wedeking. 2009. “Pardon the Interruption:
An Empirical Analysis of Supreme Court Justices’ Behavior During Oral Arguments.”
Loy. L. Rev. 55:331.

Kearney, Joseph D and Thomas W Merrill. 2000. “Influence of Amicus Curiae Briefs on the

Supreme Court.” U. Pa. L. Rev. 148:743.

125

Kritzer, Herbert M. and Mark J. Richards. 2003.

“Jurisprudential Regimes and Supreme
Court Decisionmaking: The Lemon Regime and Establishment Clause Cases.” Law &
Society Review 37(4):827–840.

Lax, Jeffrey R. 2012. “More on the Persuasiveness of Oral Arguments.” The Monkey Cage .

Lax, Jeffrey R and Charles M Cameron. 2007. “Bargaining and opinion assignment on the

US Supreme Court.” The Journal of Law, Economics, & Organization 23(2):276–302.

Long, J Scott. 1997. Regression Models for Categorical and Limited Dependent Variables.

Sage Publications.

Luse, Jennifer K., Geoffrey McGovern, Wendy L. Martinek and Sara C. Benesh. 2009. ““Such
Inferior Courts . . .” Compliance by Circuits with Jurisprudential Regimes.” American
Politics Research 37(1):75–106.

Lynch, Kelly J. 2004. “Best Friends-Supreme Court Law Clerks on Effective Amicus Curiae

Briefs.” JL & Pol. 20:33.

Malphurs, Ryan A. 2010.

““People Did Sometimes Stick Things in my Underwear” The

Function of Laughter at the US Supreme Court.” People 10(2).

Maltzman, Forrest, James F Spriggs and Paul J Wahlbeck. 2000. Crafting Law on the

Supreme Court: The Collegial Game. Cambridge University Press.

Martin, Andrew D. and Kevin M. Quinn. 2002. “Dynamic Ideal Point Estimation via Markov
Chain Monte Carlo for the U.S. Supreme Court, 1953–1999.” Political Analysis 10(2):134–
153.

Martin, Andrew D, Kevin M Quinn and Lee Epstein. 2004.

united states supreme court.” NCL rev. 83:1275.

“The median justice on the

McGuire, Kevin T. 1995. “Repeat players in the Supreme Court: The Role of Experienced

Lawyers in Litigation Success.” The Journal of Politics 57(1):187–196.

McGuire, Kevin T. and James A. Stimson. 2004. “The Least Dangerous Branch Revisited:
New Evidence on Supreme Court Responsiveness to Public Preferences.” The Journal of
Politics 66(4):1018–1035.

O’Connor, Karen. 1982. “The Amicus Curiae Role of the US Solicitor General in Supreme

Court Litigation.” Judicature 66:256.

Owens, Ryan J and Justin Wedeking. 2012. “Predicting Drift on Politically Insulated Insti-
tutions: A Study of Ideological Drift on the United States Supreme Court.” The Journal
of Politics 74(2):487–500.

Peppers, Todd C. 2006. Courtiers of the Marble Palace: The Rise and Influence of the

Supreme Court Law Clerk. Stanford University Press.

126

Peppers, Todd C and Christopher Zorn. 2008.

“Law Clerk Influence on Supreme Court

Decision Making: An Empirical Assessment.” DePaul L. Rev. 58:51.

Richards, Mark J. and Herbert M. Kritzer. 2002. “Jurisprudential Regimes in Supreme Court

Decision Making.” The American Political Science Review 96(2):305–320.

Richards, Mark J., Joseph L. Smith and Herbert M. Kritzer. 2006. “Does Chevron Matter?”

Law & Policy 28(4):444–469.

Ringsmuth, Eve M, Amanda C Bryan and Timothy R Johnson. 2013. “Voting fluidity and

oral argument on the US Supreme Court.” Political research quarterly 66(2):429–440.

Roberts Jr, John G. 2005. “Oral Advocacy and the Re-emergence of a Supreme Court Bar.”

Journal of Supreme Court History 30(1):68–81.

Schoenherr, Jessica A and Ryan C Black. 2019. “Friends with benefits: Case significance,
amicus curiae, and agenda setting on the US Supreme Court.” International Review of
Law and Economics 58:43–53.

SCOTUS. 2019. “Revisions to Rules of the Supreme Court of the United States: Adopted

April 18, 2019 Effective July 1, 2019.”.

Segal, Jeffrey A. 1988. “Amicus curiae briefs by the solicitor general during the Warren and

Burger Courts: A research note.” Western Political Quarterly 41(1):135–144.

Segal, Jeffrey A and Harold J Spaeth. 1993. The Supreme Court and the attitudinal model

revisited. Cambridge University Press.

Segal, Jeffrey A. and Harold J. Spaeth. 1996. “The Influence of Stare Decisis on the Votes of
United States Supreme Court Justices.” American Journal of Political Science 40(4):971–
1003.

Segal, Jeffrey A and Harold J Spaeth. 2002. The Supreme Court and the Attitudinal Model

Revisited. Cambridge University Press.

Songer, Donald R and Reginald S Sheehan. 1993.

“Interest group success in the courts:
Amicus participation in the Supreme Court.” Political Research Quarterly 46(2):339–354.

Songer, Donald R, Reginald S Sheehan and Susan Brodie Haire. 1999. “Do the Haves Come
out Ahead Over Time-Applying Galanter’s Framework to Decisions of the US Courts of
Appeals, 1925-1988.” Law & Society Review 33:811.

Spaeth, Harold J and Jeffrey A Segal. 2001. Majority Rule or Minority will: Adherence to

Precedent on the US Supreme Court. Cambridge University Press.

Spaeth, Harold, Lee Epstein, Ted Ruger, Keith Whittington, Jeffrey Segal and Andrew D

Martin. 2018. “Supreme Court Database, Version 2018 Release 1.”.

127

Spriggs, II, James F. and Thomas G. Hansford. 2001. “Explaining the Overruling of U.S.

Supreme Court Precedent.” The Journal of Politics 63(4):1091–1111.

Spriggs, James F and Paul J Wahlbeck. 1997a. “Amicus Curiae and the Role of Information

at the Supreme Court.” Political Research Quarterly 50(2):365–386.

Spriggs, James F and Paul J Wahlbeck. 1997b. “Amicus Curiae and the Role of Information

at the Supreme Court.” Political Research Quarterly 50(2):365–386.

Teles, Steven M. 2012. The Rise of the Conservative Legal Movement: The Battle for Control

of the Law. Vol. 128 Princeton University Press.

Unah, Isaac and Ange-Marie Hancock. 2006.

“US Supreme Court decision making, case

salience, and the attitudinal model.” Law & Policy 28(3):295–320.

Wahlbeck, Paul J. 2006. “Strategy and Constraints on Supreme Court Opinion Assignment.”

University of Pennsylvania Law Review 154(6):1729–1755.

Ward, Artemus and David L Weiden. 2006. Sorcerers’ Apprentices: 100 Years of Law Clerks

at the United States Supreme Court. NYU Press.

Ward, Stephanie Francis. 2007. “Friends of the Court Are Friends of Mine.” Supreme Court

Report .

Wedeking, Justin. 2010. “Supreme Court Litigants and Strategic Framing.” American Journal

of Political Science 54(3):617–631.

Wohlfarth, Patrick C. 2009.

“The Tenth Justice? Consequences of Politicization in the

Solicitor General’s Office.” The Journal of Politics 71(1):224–237.

Woodward, Bob and Scott Armstrong. 1979. The Brethren: Inside the Supreme Court.

Simon and Schuster.

128