ATTORNEYS, MERITS BRIEFS, AND U.S. SUPREME COURT DECISION MAKING

By

Jessica Ann Schoenherr

A DISSERTATION

Submitted to

Michigan State University

in partial fulfillment of the requirements

for the degree of

Political Science – Doctor of Philosophy

2020

ABSTRACT

ATTORNEYS, MERITS BRIEFS, AND U.S. SUPREME COURT DECISION MAKING

By

Jessica Ann Schoenherr

Attorneys are key players in the U.S. Supreme Court decision-making process who use

their legal arguments to influence the justices’ decisions. Tasked with preparing merits

briefs that make initial forays into the legal issues in a case, attorneys use that space to

limit the range of possible outcomes and direct the justices’ focus toward certain (favorable)

areas of the law. Yet scholars understand very little about this process because they have

struggled to empirically identify and examine legal arguments and their impact on the Court’s

decisions. In response to this problem, I use a combination of human-assisted coding and

machine learning techniques to develop a method of extracting the core components of a legal

argument for analysis. Using data collected from the legal arguments in more than three

thousand briefs from 1,509 different cases the Court reviewed between the 1984 and 2007

terms, I study attorneys’ use of the law, focusing on their decision to frame their arguments

using prevailing case law or a wholly new, entrepreneurial argument. I show that attorneys’

decisions regarding the framing of their arguments – whether to align their argument with

the Court’s prevailing approach or to use an entrepreneurial argument to radically change

an area of case law – can alter the justices’ decisions in a case. I use that same data to

show that attorneys are strategic in their deployment of these frames, looking for signs the

justices might respond positively to one type over the other if the argument is made well.

To my parents, for teaching me that Schoenherrs never say “can’t”

iii

ACKNOWLEDGMENTS

Writing a dissertation was the most solitary thing I did in graduate school. While political

science continues to move toward collaboration and coauthorship, I still had to complete this

one project that will bear only my name, and I am reasonably confident this will be the only

thing I ever write with that particular distinction. But it is absolutely incorrect to say that

I wrote this dissertation alone. It took an army of family, friends, and faculty to get me to

the finished product. And now, like in the liner notes to a great album, I get to thank all of

those people for everything they did. How great is that?

The first, largest, and most important thank you in this dissertation goes to my husband,

Ezra. I am not convinced I would have made it past the first week of graduate school, let

alone through all six years, without him. When I needed to learn how to program, he taught

me, even though he probably wanted to sit on the couch and read a good mystery novel

instead. When our dogs lost their minds, he took them for walks so I could focus. And

when things turned out well, he was the first one to celebrate with me. Ezra is my greatest

cheerleader and staunchest supporter, and I could not be more grateful that he is the partner

I have for this giant mess we call life. Thanks for being there with me every step of the way.

I have to thank my family – Mom, Dad, and Tracy – for making me the person that I

am today. My parents constantly sacrificed so that I could have a good education, and I

dedicate this dissertation to them as a (small) thank you for everything they did for me.

Dad taught me to stick up for myself, reminding me that “if someone punches you, punch

them back twice as hard.” Mom, who went back to college at 40 and got her D.B.A. at 60,

inspires me daily to fight for what I want and to never settle. And my little sister Tracy

reminds me to live life as fully as possible, even if that means eating eight pounds of tater

tots in two days. I’m grateful that she is around to keep me normal and grounded, despite

my childhood protestations that my parents should replace her with a puppy. Or a turtle.

iv

Or a cat. Or a fish. Or maybe a frog. I was obviously not very convincing.

I have so many family and friends to thank that I cannot possibly name them all here,

but a few people deserve special recognition: my father-in-law, Michael; Aunt Gayle and

Uncle Green Genes; Aunt Gene and Uncle Dave; Aunt Linda and Uncle Nick; my army of

cousins (Cathy and Lou, Lauren and James, Jason and Kristen); Bill and Jess, Carrie, Lily,

J2 and Nick, Debrah, Ellen, Kari, Vanessa, Jean, and Leslee. You all know what you did,

and know that I appreciated every phone call, text, visit, care package, and lunch. I also

have to explicitly acknowledge my two best friends, Stephanie and Joe, who somehow dealt

with my nonsense over the last six years and still love me at the end of this. Thanks guys.

In my six years at Michigan State, I met and worked with some of the most amazing

young scholars I know. I first and foremost need to thank my Women Hustling in Political

Science (WHiPS), the four women who pushed me to work hard every single day while still

accepting that some days my work would be accurately described as “trash.” I owe thanks

to Jamil Scott for the long talks about research and text analysis on our way to Starbucks

every day; to Emma Slonina for keeping me to task while also encouraging creativity and

alone time; to Lora DiBlasi for being my officemate and confidant for the last six years; and

to Elizabeth Lane, the Hamilton to my Lafayette (“no one matches my practical tactical

brilliance!”), for being my coauthor for life. You four are my favorite people in the discipline,

thanks for being my friends.

Next up is my TACO family, the judicial students at MSU who are now my core network

of colleagues and coauthors. Thank you to Miles Armaly and Rachel Schutte for offering

advice whenever I needed it; to Nick Waterbury for being a patient coauthor who calls me

out on my nonsense; and to Jonathan King for being a supportive and occasionally unruly

little brother who is all-too-willing to cause trouble with me. I am so grateful I had the four

of you (and Elizabeth) around! I also have to thank the current and former members of the

American Politics and Policy Research Group at MSU — Kesicia Dickinson, Adam Enders,

v

Marty Jordan, Bob Lupton, and Erika Rosebrook — for their support and help over the

years. You’ve all made my work better. Finally, thank you to Caleb Lucas and Nate Smith

for being truly phenomenal colleagues, even if they study things that make no sense to me.

I owe a world of thanks to the Michigan State University Department of Political Science

for financial and professional help over the last six years. Thank you to Karen Battin,

Rhonda Burns, Brian Egan, Sarah Krause, Kelly Washburn, and Krista Zeig for being the

true MVPs and organizing the chaos while answering every single stupid question I ever had.

I also need to thank Sarah Reckhow and Ani Sarkissian for their endless work as graduate

directors, including but not limited to making sure I had all the financial resources I needed

to successfully complete this project. Finally, thank you to the faculty who opened their

doors and offered mentorship, kindness, and humor when none were required, especially Eric

Chang, Eric Gonzales Juenke, and Mariana Medina. I am a little more sane today because

of all of you.

Ben Kleinerman was my first mentor in the discipline, and he was the one who convinced

me to go to graduate school when I had no idea what that even meant. Thank you for taking

me under your wing when I was a young undergrad who wanted to study the presidency, for

walking me through my first research project, for helping me get into graduate school twice,

and treating me like a member of your family. I wouldn’t be where I am today without you.

The chapters that follow here are the product of several workshops, conferences, and job

interviews. Thank you to every conference discussant who offered thoughts, to everyone who

asked a question, and to those who talked about my work over email or on the phone. Your

comments were truly invaluable in making this project better. I especially want to thank

Justin Wedeking for his help learning the text skills I needed to complete this project, Matt

Hitt and Doug Rice for their helpful comments on Chapter 3, and Kirk Randazzo for helping

me reframe legal entrepreneurship as a concept while driving me to the airport.

Which brings us to the end, where I say try to thank my committee for all of their help

vi

over the last six years. To be clear, these words will not even come close to explaining

how much these people have done for me. I would not be where I am today without my

committee, who are four of the smartest and best humans I know. Matt Grossmann read (I

think) almost every single piece of paper I wrote in graduate school and offered the kind of

pointed, perfect criticism that made everything I did better. Thank you for always having

time for me and constantly reminding me to look at the bigger academic picture! I tell Ian

Ostrander that he’s my academic big brother, the patient mentor who listened, explained,

and corrected, and occasionally let me shut his office door and scream into the void. Thank

you for always having an open door and a dark joke for me. Cory Smidt taught me that

it is better to be boldly wrong than meekly right, a lesson I’ve carried with me throughout

grad school, right down to the bold, innovative ideas presented in the pages that follow.

Thank you for teaching me methods, for walking through every single model with me until I

got it right, and for always being up for a conversation about anything, from pierogi to the

Cincinnatus.

And finally, I have to thank my advisor, Ryan Black. Ryan is, for all intents and purposes,

his students’ equivalent of Leo McGary, the brilliant and steadfast chief of staff on The

West Wing; he spends an inordinate amount of time getting his driven and occasionally lost

students to produce something worthwhile. Over countless hours, Ryan taught me everything

I needed to know about being a scholar -— how to write a snappy lit review, how to approach

a hard question with data, how to deal with coauthors, how to teach undergraduates, how

to mentor others, and how to say “no” to things (the hardest lesson of all for me!). He

pushed me to work harder, do more, and be better, and he did this while reminding me that

perspective is good, that family time is important, and that you can be a good academic

and a good person at the same time. Ryan, thank you for literally everything. I am a better

scholar and a better person because of you.

vii

TABLE OF CONTENTS

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

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

CHAPTER 1

INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . .
1.1 Attorneys, the Law, and Legal Entrepreneurship . . . . . . . . . . . . . . . .

CHAPTER 2 USING TEXT ANALYSIS OF MERITS BRIEFS TO STUDY AT-
TORNEYS’ ROLE IN THE U.S. SUPREME COURT DECISION-
MAKING PROCESS . . . . . . . . . . . . . . . . . . . . . . . . . .
2.1 Briefs and Judicial Behavior . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.2 Using Citations to Study Attorney Strategy . . . . . . . . . . . . . . . . . .
2.3 Data Collection . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.4 Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
2.5 Conclusion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

CHAPTER 3 CALL AND RESPONSE: LEGAL ENTREPRENEURSHIP AND

ATTORNEY SUCCESS AT THE U.S. SUPREME COURT . . . . .
3.1 Attorneys, Briefs, and the Law . . . . . . . . . . . . . . . . . . . . . . . . . .
3.2 Legal Entrepreneurship and Innovation . . . . . . . . . . . . . . . . . . . . .
3.3 Data and Measures . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
Identifying Entrepreneurship . . . . . . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . .
3.4 Methodology and Empirical Results . . . . . . . . . . . . . . . . . . . . . . .
3.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

3.3.1
3.3.2 Additional Variables

CHAPTER 4 PUSHING THE BOUNDARIES: DECIDING TO ENGAGE IN LE-

GAL ENTREPRENEURSHIP AT THE U.S. SUPREME COURT . .
4.1 Appealing to the Justices in the Merits Brief . . . . . . . . . . . . . . . . . .
4.2 Making the Decision to go Entrepreneurial
. . . . . . . . . . . . . . . . . . .
. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
4.3 Data and Methods
4.3.1 Dependent Variable: Legal Entrepreneurship . . . . . . . . . . . . . .
4.3.2
Independent Variables: Opportunity, Environment, and Resources . .
4.4 Methodology and Empirical Results . . . . . . . . . . . . . . . . . . . . . . .
4.5 Discussion . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

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

APPENDICES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .
CHAPTER 3 APPENDIX . . . . . . . . . . . . . . . . . . .
CHAPTER 4 APPENDIX . . . . . . . . . . . . . . . . . . .

APPENDIX A
APPENDIX B

viii

x

xi

1
3

7
10
12
14
18
28

31
33
36
40
41
46
48
57

60
63
66
69
70
75
77
85

88

90
91
95

BIBLIOGRAPHY . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

97

ix

LIST OF TABLES

Table 2.1: Summary Statistics for Citation Patterns in Briefs . . . . . . . . . . . . .

Table 3.1: Logistic Regression Results, Justice Votes in Favor of the Petitioner

. . .

Table 4.1: Logistic Regression Results, Attorney Decision to Engage in Legal En-

trepreneurship . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .

Table B.1: Alternative Logistic Regression Results, Attorney Decision to Engage in

Legal Entrepreneurship . . . . . . . . . . . . . . . . . . . . . . . . . . . .

26

49

78

96

x

LIST OF FIGURES

Figure 2.1: Citation Patterns in Supreme Court Merits Briefs Over Time –
Open dots indicate the average total number of citations in a merits brief
(left) and the average number of cases cited in a brief (right) between
the 1984 and 2007 terms. The lines are non-parametric lowess curves of
the overall trend.

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

Figure 2.2: Histogram of Mentions of a Case in a Brief – This histogram

shows the number of times a case got mentioned in a brief. . . . . . . . .

Figure 2.3: Trends in Citation Treatment Over Time – Hatch marks indi-
cate the average number of citations (left) of positively-discussed cases
(right) in a brief per term. Xs show the same for negatively-discussed
cases, and open dots represent the neutrally-discussed cases. The solid
(positive), dashed (negative), and dotted (neutral) lines are non-parametric
. . . . . . . . . . . . . . . . . . . . . .
lowess curves of the overall trend.

Figure 2.4: Trends in Citation Treatment for Petitioners, Respondents,
Winners, and Losers Over Time – Hatch marks indicate the average
number of positive citations in a brief per term. Xs show the same
for negatively-discussed cases, and open dots represent the neutrally-
discussed cases. The solid (positive), dashed (negative), and dotted
(neutral) lines are non-parametric lowess curves of the overall trend. . . .

Figure 2.5: Heat Map of the 20 Most-Cited Supreme Court Precedents –
Each box represents the average number of times attorneys mentioned
that precedent in their briefs that term; the darker the color, the higher
the average.

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

Figure 2.6: Heat Map of the 20 Most-Cited First Amendment Precedents
– Each box represents the average number of times attorneys mentioned
that precedent in their briefs that term; the darker the color, the more
they mentioned the case. . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 3.1: Probability a Supreme Court Justice Votes with the Petitioner
By Argument Type – Left side shows the probability a justice votes
with the petitioner when the petitioner uses a prevailing argument,
right side shows the probability a justice votes with the petitioner when
the petitioner uses an entrepreneurial argument. Vertical lines identify
95% confidence intervals. Predicted probabilities calculated using the
observed-value approach. . . . . . . . . . . . . . . . . . . . . . . . . . . .

19

20

21

22

23

28

50

xi

Figure 3.2: Probability a Supreme Court Justice Votes with the Respon-
dent By Argument Type – Left side shows the probability a justice
votes with the respondent when the respondent uses a prevailing argu-
ment, right side shows the probability a justice votes with the respon-
dent when the respondent uses an entrepreneurial argument. Vertical
lines identify 95% confidence intervals. Predicted probabilities calcu-
lated using the observed-value approach.

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

Figure 3.3: Probability a Supreme Court Justice Votes with the Petitioner
Based on Argument Type and Attorney Experience – Probabil-
ity a Supreme Court justices votes with the petitioner based on attorney
experience and the decision to use a prevailing (left) or entrepreneurial
(right) argument. Dashed lines are 95% confidence intervals around
those estimates. Predicted probabilities calculated using the observed-
value approach.

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

Figure 3.4: Probability a Supreme Court Justice Votes with the Respon-
dent Based on Argument Type and Attorney Experience –
Probability a Supreme Court justices votes with the respondent based
on attorney experience and the decision to use a prevailing (left) or
entrepreneurial (right) argument. Dashed lines are 95% confidence in-
tervals around those estimates. Predicted probabilities calculated using
the observed-value approach. . . . . . . . . . . . . . . . . . . . . . . . . .

51

52

54

Figure 3.5: Probability a Supreme Court Justice Votes with the Petitioner
Based on Argument Type and Attorney Status – Probability a
Supreme Court justice votes with the petitioner based an attorney sta-
tus and the decision to use a prevailing (left) or entrepreneurial (right)
argument. Dashed lines are 95% confidence intervals around those esti-
mates. Predicted probabilities calculated using the observed-value approach. 55

Figure 3.6: Probability a Supreme Court Justice Votes with the Respon-
dent Based on Argument Type and Attorney Status – Probabil-
ity a Supreme Court justice votes with the respondent based an attor-
ney status and the decision to use a prevailing (left) or entrepreneurial
(right) argument. Dashed lines are 95% confidence intervals around
those estimates. Predicted probabilities calculated using the observed-
value approach.

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

56

xii

Figure 4.1: Probability an Attorney Engages in Legal Entrepreneurship as
Ideological Congruence Increases - Probability an attorney engages
in legal entrepreneurship as the ideological congruence between the lower
court’s decision and the Supreme Court median increases. Dashed lines
are 95% confidence intervals around those estimates. Predicted proba-
bilities calculated using the observed-value approach.

. . . . . . . . . . .

Figure 4.2: Probability an Attorney Engages in Legal Entrepreneurship by
Issue Area – Probability an attorney engages in legal entrepreneurship
based on the issue area into which the case falls. Vertical lines are
95% confidence intervals. Predicted probabilities calculated using the
observed-value approach. . . . . . . . . . . . . . . . . . . . . . . . . . . .

Figure 4.3: Probability an Attorney Engages in Legal Entrepreneurship
When Lower Court Conflict Exists – Probability an attorney en-
gages in legal entrepreneurship when lower court conflict exists (right)
or does not (left). Vertical lines identify 95% confidence intervals. Pre-
dicted probabilities calculated using the observed-value approach.

. . . .

Figure 4.4: Probability an Attorney Engages in Legal Entrepreneurship
When a Lower Court Judge Dissents – Probability an attorney
engages in legal entrepreneurship when the Supreme Court does not
note a lower court dissent (left) or writes about a lower court dissent
in the eventual opinion (right). Vertical lines identify 95% confidence
intervals. Predicted probabilities calculated using the observed-value
approach.

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

Figure 4.5: Probability an Attorney Engages in Legal Entrepreneurship
Based on Attorney Experience – Probability an attorney engages
in legal entrepreneurship as the attorney’s past experience at oral argu-
ment increases. Dashed lines are 95% confidence intervals around those
estimates. Predicted probabilities calculated using the observed-value
approach.

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

80

81

82

83

84

Figure A.1: Difference in Probability a Justice Sides with the Petitioner
Based on Decision to Engage in Legal Entrepreneurship and
Attorney Experience – Probability a justice sides with the petitioner,
based on petitioner experience, when the petitioner engages in legal en-
trepreneurship minus the probability a justice sides with the petitioner,
based on petitioner experience, when the petitioner uses a prevailing
argument. Dashed lines are 95% confidence intervals around those esti-
mates. Predicted probabilities calculated using the observed-value approach. 91

xiii

Figure A.2: Difference in Probability a Justice Sides with the Respondent
Based on Decision to Engage in Legal Entrepreneurship and
Attorney Experience – Probability a justice sides with the respon-
dent, based on respondent experience, when the respondent engages in
legal entrepreneurship minus the probability a justice sides with the re-
spondent, based on respondent experience, when the respondent uses a
prevailing argument. Dashed lines are 95% confidence intervals around
those estimates. Predicted probabilities calculated using the observed-
value approach.

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

Figure A.3: Difference in Probability a Justice Sides with the Petitioner
Based on Decision to Engage in Legal Entrepreneurship and
Litigant Status – Probability a justice sides with the petitioner, based
on petitioner status, when the petitioner engages in legal entrepreneur-
ship minus the probability a justice sides with the petitioner, based
on petitioner status, when the petitioner uses a prevailing argument.
Dashed lines are 95% confidence intervals around those estimates. Pre-
dicted probabilities calculated using the observed-value approach.

. . . .

92

93

Figure A.4: Difference in Probability a Justice Sides with the Respondent
Based on Decision to Engage in Legal Entrepreneurship and
Litigant Status – Probability a justice sides with the respondent,
based on respondent status, when the respondent engages in legal en-
trepreneurship minus the probability a justice sides with the respondent,
based on respondent status, when the respondent uses a prevailing argu-
ment. Dashed lines are 95% confidence intervals around those estimates.
Predicted probabilities calculated using the observed-value approach.

. .

94

xiv

CHAPTER 1

INTRODUCTION

In the middle of The Brethren, Bob Woodward and Scott Armstrong tell a fascinating

story about the Supreme Court’s decision to overturn the death penalty in Furman v. Georgia

(1972). They explain that the justices had long struggled to reach an agreement regarding

the death penalty and finally decided to face the problem head-on during the 1971 term

(Woodward and Armstrong 1979). The justices took four radically different death penalty

cases, bundled them together, and agreed to use those cases to resolve the issue once and for

all. Seeing the justices’ struggle, the attorney tasked with convincing the justices to abolish

the death penalty, Anthony Amsterdam, decided to go big with his argument, appealing to

the justices using every argument he could reasonably employ. He said the death penalty

discriminated against minorities and the poor; that states imposed it in an arbitrary fashion;

that states imposed it so infrequently it could not possibly serve a deterrent purpose; and

that contemporary society rejected it. This multi-pronged approach got Amsterdam the

result he sought. Four of the justices (Marshall, Brennan, Douglas, and White) agreed with

some part of his argument, while a fifth (Stewart) refused to cast the deciding vote in favor

of the death penalty. For the next five years, no one in the United States was executed.

Woodward and Armstrong were quick to point out the legal reasoning behind that decision

was anyone’s guess.

In a literature that focuses almost exclusively on how Supreme Court justices’ ideological

preferences and personalities influence their decisions, scholars have long struggled to incor-

porate a story like this into their models. Scholars can utilize well-established measures of

a justice’s ideological proclivities to see how they influence decision making (Epstein et al.

2007; Martin and Quinn 2002; Segal and Cover 1989); they can use new measures of per-

sonality to see if their personal preferences and quirks played a role in the decision (Black

1

et al. 2020; Epstein and Knight 2013; Hall 2018); they can even look at how institutional

rules and norms, including the law, might have affected the justices’ behavior (Bailey and

Maltzman 2011; Hansford and Spriggs 2006; Hitt 2019; Maltzman, Spriggs and Wahlbeck

2000). Almost none of this literature considers the possibility that the attorneys – the very

people who brought these cases to the Court and argued them before the justices – might

influence judicial decision-making with their legal arguments (but see Corley 2008, Hazelton,

Hinkle and Spriggs 2019, and Wedeking 2010 for exceptions).

My goal over the next 100 pages is to correct this oversight by showing that attorneys are

key players in the Supreme Court decision-making process. In an adversarial legal system like

the one that exists in the United States, attorneys are the ones who make the initial foray

into a case’s legal issues and, in the process, place boundaries around where the justices

can legally roam (Epstein, Segal and Johnson 1996, but see McGuire and Palmer 1995).

Attorneys make decisions regarding the more technical aspects of a brief that can alter the

justices’ responses to a party’s argument (Black, Hall, Owens and Ringsmuth 2016; Corley

2008; Feldman 2016), and their decisions regarding framing can do the same (Wedeking

2010). Their experience and familiarity with the justices can even aid their ability to appeal

to the justices in a winning manner (McGuire 1995; Nelson and Epstein 2019). Attorneys

also make legal arguments that influence the justices’ understandings of a case, but scholars

have struggled to understand how the justices respond to these arguments. The struggle

stems from an empirical problem – studying the law is hard (Hansford and Spriggs 2006). I

develop a method of examining the legal arguments in briefs and use that method to collect

data on the arguments in more than three thousand briefs from 1,509 different cases. I then

employ that data to study how attorneys’ arguments influence the justices and, consequently,

the law.

Across the next three chapters, I address three different questions: (1) How can scholars

study the legal arguments in briefs and their potential influence over the justices? (2) Can

2

attorneys use their legal arguments to influence the justices’ approach to a case and the

law? and (3) What drives an attorney to use one legal argument over another? I suggest

that scholars can study briefs by breaking them down into their constituent parts, namely

mentions of the Court’s past opinions and attorneys’ application of those precedents to the

situation at hand, and then design an automated process to collect those arguments. I then

use that data to show that attorneys’ decisions regarding the framing of their arguments –

specifically, whether to use a prevailing legal argument or an entrepreneurial one – can alter

the justices’ decisions regarding a case. I finally examine attorneys’ strategy regarding the

employment of those arguments and find that attorneys look for signs that the justices might

respond positively to one type over the other. The last few pages contain a brief review and

some final thoughts about this project.

1.1 Attorneys, the Law, and Legal Entrepreneurship

When facing the Supreme Court, having a good attorney matters. Experienced attorneys

are more likely to win before the justices (McGuire 1993, 1995; Nelson and Epstein 2019)

and wealthy clients are all too happy to pay the otherwise-daunting $2000-an-hour fee if it

means they are more likely to win their cases (Biskupic, Roberts and Shiffman 2014; Rubino

2016). The assumption is that experienced attorneys have valuable skills, and anecdotal

evidence suggests that one of those skills is knowing how to frame a legal argument in the

most appealing manner possible (Garner and Roberts 2010). Good attorneys know how to

use the law to persuade the justices toward a preferred outcome (Garner 2003). They do this

in the merits brief, a written document in which attorneys make their only complete and

coherent appeal to the justices. Yet research on merits briefs focuses almost exclusively on

the more technical aspects of a brief, examining word choice and readability and not the law

(Black, Hall, Owens and Ringsmuth 2016; Feldman 2016). The work that does examine the

full brief focuses on overall language and not the law itself, let alone its persuasive capacity

(Black and Owens 2012c; Corley 2008; Feldman 2016; Wedeking 2010). This work is, I

3

suggest, the legal equivalent of studying scoring in the National Hockey League by looking

at the size of the goaltenders’ pads – important, yes, but maybe not where all of scholars’

analytical focus should go. Scholars need to focus on the legal arguments in the briefs, too.

Scholars have failed to study attorneys’ legal contributions to the decision-making pro-

cess because the law is empirically difficult to analyze in the best situations (Hansford and

Spriggs 2006) and even more difficult to examine when studying a document like a legal brief.

Empirical political scientists have successfully studied the justices’ use of the law in their

opinions by breaking a legal argument down into parts, specifically citations to the Court’s

past precedents and the justices’ discussions of those precedents (Black and Spriggs 2013;

Clark and Lauderdale 2010). But this work is entirely dependent on pre-existing data from

legal research services, namely the citation treatment data from Shepard’s Citations (Spriggs

and Hansford 2000). LexisNexis provides this service because it has value – attorneys need

to know what the Court thinks about a past precedent, and they typically need to identify

that information quickly. The same need does not exist for an attorney’s discussion of a past

precedent in an old brief, so legal research services do not offer the same information about

briefs. As a result, scholars have been unable to take advantage of these methodological

advances to study attorneys’ arguments.

Thankfully, advances in text analysis and machine learning have made it possible to

replicate the hand-coded process that LexisNexis utilizes and then apply it to Supreme

Court merits briefs. So I do that. After studying the “shepardization” process, I wrote a set

of programs in R that can take a brief and break it down into its argument’s constituent parts

for later analysis. This data offers political scientists their first look at how attorneys treat

the Court’s past precedents in their briefs and how attorneys construct a legal argument. I

walk through this process and the data I collect from it in Chapter 2.

Knowing that it is possible to identify the legal arguments in briefs, the next objective

is understanding how attorneys use them to appeal to the justices. When attorneys appear

4

before the Supreme Court, they do so with a mandate: win the case for their client (Garner

2003). But, by nature of appearing before the Supreme Court, attorneys are attempting

to do so by answering legal questions that even lower court judges failed to answer with

any level of satisfaction (Perry 1991). Every time an attorney appears before the Supreme

Court, then, she is essentially venturing into the legal unknown, trying to make her way

through the Court’s precedents toward an acceptable answer without knowing exactly what

the right answer is. I suggest that attorneys use one of two different approaches to structure

the process: they can use a prevailing argument and align their brief with the justices’

opinions in an area of case law, or they can engage in legal entrepreneurship and use a

new and innovative argument to push the law in radical new directions. Both have their

benefits and drawbacks, but legal entrepreneurship is an especially risky enterprise when

approaching justices who favor simplified decision making processes (Gennaioli and Shleifer

2007; Niblett, Posner and Shleifer 2010) and are notoriously averse to overturning their own

rulings (Hansford and Spriggs 2006).

Using the data collection process that I outlined in Chapter 2, I collect data on 3,018

briefs associated with 1,509 cases the Supreme Court heard between the 1984 and 2007

terms and identify attorneys’ use of prevailing and entrepreneurial arguments. I find that

entrepreneurial arguments are common but offer limited advantages to the attorneys who

use them. More specifically, my results suggest that entrepreneurial arguments can help

inexperienced and resource-poor petitioners win. They do not offer other attorneys any

added advantage for securing the justices’ votes, however. At the end of Chapter 3, I

conclude that attorneys can use their legal arguments to influence the justices’ approach to

a case and, consequently, to the law itself.

My last objective is to see when attorneys decide to employ one type of argument over

the other – that is, to see if attorneys are strategic in their use of entrepreneurial arguments.

Recall that the merits brief provides attorneys with their best opportunity to frame the

5

justices’ perceptions’ of a case and color their approach to its resolution (Black, Hall, Owens

and Ringsmuth 2016). Attorneys have to decide whether to go with the safe, prevailing

argument or to go with the messier entrepreneurial one. How do attorneys make that deci-

sion? Research suggests the justices look for signs of acceptance before making legal change,

including considering political opportunities, interest group advocacy work, and membership

changes (Clark 2019; Epstein and Kobylka 1992), so I suggest that attorneys do the same.

I suspect that attorneys consider large-scale political opportunities (new members, political

mood), the legal environment surrounding a case (the rigidity of the law in an issue area,

signals from the lower court that something in the law is problematic), and the attorneys’

own skills and resources before they decide to make an entrepreneurial argument.

Using the same data that I employed in Chapter 3, I find that attorneys are more likely

to engage in legal entrepreneurship when the legal environment surrounding their cases

suggests they should. The results also suggest that more experienced attorneys are more

likely to engage in legal entrepreneurship, as their experience aids their ability to identify

opportunities for change. Unlike the justices, attorneys do not pay attention to the political

environment surrounding a case, caring less about the politics and more about the case in

front of them. In my final chapter, then, I show that attorneys look for certain signals that

an entrepreneurial argument might be worth their time.

After three chapters of analysis, my conclusion is that scholarship can, in fact, offer a

fuller explanation of why Justices Marshall, Brennan, Douglas, White, and Stewart voted

to overturn the death penalty in 1972. Policy preferences can explain part of their decision,

as can personality, the Court’s internal rules, and their understanding of the law. But, as I

hope these pages show, Anthony Amsterdam’s decision to appeal to the justices using every

argument he could find mattered too, and now scholarship can integration that explanation

as well.

6

CHAPTER 2

USING TEXT ANALYSIS OF MERITS BRIEFS TO STUDY ATTORNEYS’

ROLE IN THE U.S. SUPREME COURT DECISION-MAKING PROCESS

In 1992, the Supreme Court agreed to hear arguments in Planned Parenthood of South-

eastern Pennsylvania v. Casey, a case challenging a restrictive Pennsylvania abortion law.

Tasked with presenting the justices with a merits brief in the case, Kathryn Kolbert, Planned

Parenthood’s lead attorney, had to decide how to use her brief to appeal to the justices

(Toobin 2008). She knew the Court did not particularly favor her client’s pro-choice stance,

as the justices had spent much of the 1980s chipping away at abortion rights and showed

no sign of curbing that trend (Mezey 2003). Kolbert needed the Court to reverse its course,

so she decided to explicitly remind the justices they had past rulings to uphold. She be-

gan her legal argument by asserting, in all capital letters, that “[t]his Court must reaffirm

the central holding of Roe v. Wade that the right to choose an abortion is a fundamental

right protected by the Constitution,” and went on to argue the justices could not uphold

Pennsylvania’s Abortion Control Act without overturning Roe’s promise of constitutional

protection for women seeking to terminate their pregnancies (Hull and Hoffer 2001). Her

brief reminded the justices they were duty-bound to uphold past rulings, and Roe could

not be the exclusion. Kolbert did not do this on accident – she knew the justices found

overturning precedent distasteful and hoped her decision to focus on precedent would turn

the Court in her favor (Toobin 2008). She had to do what was best for her client, and this

strategic argument was the only way to do that.

When Kathryn Kolbert wrote her brief, she, like every attorney who files a merits brief

with the Supreme Court, sought to direct a majority of justices toward her preferred outcome.

Attorneys actually have two opportunities to do this at the Supreme Court: first, in written

form via the merits brief and then again through the question-and-answer process of oral

7

argument (Johnson 2004). Because merits briefs come first, and because the justices control

the tempo, pace, and content of oral argument (Black, Johnson and Wedeking 2012), merits

briefs provide attorneys with their main (and sometimes only) opportunity to frame the

justices’ perceptions of a case (Black, Hall, Owens and Ringsmuth 2016). So after working

through a case’s background and lower court history in the first few pages, an attorney then

proceeds to construct complete and coherent legal argument that guides the justices through

the relevant case law and explains how the Court’s past decisions apply to the case at hand.

They offer suggestions about how the justices can use the law to reach a desired outcome,

essentially supplying valuable legal advice with a persuasive, self-serving spin. Briefs are

documents that can and do influence outcomes at the United States Supreme Court.

While briefs are clearly important, scholars have struggled to understand how the legal

arguments outlined in briefs influence the justices. A well-developed line of research shows

that technical factors, like word choice and readability, can influence outcomes at the Court

(Black, Hall, Owens and Ringsmuth 2016; Corley 2008; Feldman 2016; Wedeking 2010),

but that research almost never extends to the legal argument itself (see Hazelton, Hinkle

and Spriggs 2019 for an exception). This oversight exists because studying “the law” in an

empirically-rigorous manner remains a challenge for anyone studying the judiciary (Hansford

and Spriggs 2006). This paper changes that. I suggest that it is possible to break a brief

down into analyzable parts, namely citations of past precedent and attorneys’ discussions of

those citations. By looking at briefs in this manner, I can then use a combination of hand-

coding and machine learning techniques to collect data on the legal arguments contained

in briefs. After walking through the development of this tool, I use it to study the legal

arguments in 4,812 briefs from 2,232 cases the Supreme Court reviewed between the 1984

and 2007 terms. My analysis suggests that the legal arguments in briefs are worthy of study

and can be used to better understand attorney strategy at the Supreme Court.

In developing this tool and using it study briefs and attorney decision-making, I con-

8

tribute to the literature in three significant ways. First, I show that merits briefs offer a

comprehensive look at how Supreme Court opinions get used in practice. Scholars study

the legal arguments in Supreme Court opinions because opinions are supposed to be the

definitive legal authority on an issue (Clark and Lauderdale 2010). But Supreme Court

opinions can be maddeningly unclear (Black, Owens, Wedeking and Wohlfarth 2016; Hitt

2019; Maltzman, Spriggs and Wahlbeck 2000) and can sometimes create more confusion than

they relieve. Briefs, however, offer insight about how attorneys – people who work with the

law for a living – understand those opinions. Attorneys discuss the precedents that work

and the precedents that do not. They show the justices where opinions are misapplied or

need further clarification. Studying the legal content of briefs thus allows scholars to study

case law in its most practical, rather than its most idealistic, form.

Second, I incorporate attorneys into the judicial decision-making process as active players,

joining scholars like McGuire (1993, 1995), Nelson and Epstein (2019), and Wedeking (2010)

who examine attorneys’ contributions to the justices’ opinions. When scholars study Supreme

Court decision making, they frequently imply the justices are all-knowing legal oracles who

make decisions based on what they already know about the law. Yet the justices themselves

have frequently pointed out they are policy generalists who depend on the briefs to guide

them through certain issue areas (Garner and Roberts 2010; Garner and Thomas 2010), and

research confirms they will use attorneys’ own words in their opinions (Corley 2008; Black

and Owens 2012c). As Chief Justice Roberts explained “You may be the world’s leading

expert in a particular question in patent law, and more power to you; I’m not. And so if

you can’t translate that expertise down to my level, you’re not going to reach me, and it’s

not going to help you at the end of the day” (Garner and Roberts 2010, 27). The justices

admit they need attorneys to set up a case and help them think it through, so the literature

should include attorneys as active influencers as well.

Finally, the process I establish for studying briefs can be used on any legal document,

9

from cert petitions to merits briefs to amicus briefs to opinions. By automating the process

of extracting and analyzing citation patterns, I provide judicial scholars with a tool they

both need and seriously lack. Scholars have long avoided studying legal texts because they

are difficult to parse in an empirically-rigorous manner (Hansford and Spriggs 2006). The

ability to use tools like Shepard’s Citations made it easier to understand the legal arguments

in opinions (Clark and Lauderdale 2010; Spriggs and Hansford 2000), but those tools are not

available for studying every legal document. My tools are.

2.1 Briefs and Judicial Behavior

After granting certiorari and agreeing to review a case, the Supreme Court asks attorneys

to provide information about their cases through two different mediums: the written merits

briefs and oral argument. These two opportunities to discuss a case with the justices serve

very different purposes. Merits briefs offer information about the case record and legal argu-

ments surrounding review (Garner and Ginsburg 2010; Johnson 2004), while oral argument

is for addressing information in the briefs and asking about the policy implications of parties’

arguments (Black, Johnson and Wedeking 2012; Johnson 2001). Attorneys create a case’s

market of ideas with their briefs, and the justices use oral argument to separate the signal

from the noise. Both the merits briefs and oral argument help the justices reach decisions in

a case (Black et al. 2011; Corley 2008), but they do so in very different ways.

As the written record of each party’s position and argument, Supreme Court merits

briefs “make the court’s job easier” (Scalia and Garner 2008, 59) by reducing the legal

research each justice must complete before deciding a case. The justices clearly tell the

attorneys what to cover in their briefs; according to Supreme Court Rule 24, briefs must

include ten different pieces of information, including the questions presented, a list of the

parties involved in the case, a statement of the case facts, a recounting of the lower courts’

decisions, a presentation of legal arguments, and the suggested case outcome. The justices

expect these documents will be well-written (Garner and Roberts 2010), succinct (C-SPAN

10

2009; Garner and Ginsburg 2010; Totenberg 2011), and entertaining (Garner and Scalia

2010), and they respond positively to briefs that are readable (Feldman 2016), written in an

unemotional and detached manner (Black, Hall, Owens and Ringsmuth 2016), and presented

by experienced attorneys (McGuire 1995; Nelson and Epstein 2019). Attorneys must provide

this information in a clear, logical, and factual manner and avoid mischaracterizing their cases

or lying outright (Garner 2003). Their job is to provide the facts.

Briefs are in no way neutral, unbiased documents, however. They are an attorney’s only

opportunity to provide the justices with a complete, uninterrupted explanation of the case at

hand (Black, Hall, Owens and Ringsmuth 2016), which means attorneys purposefully turn

briefs into persuasive documents, providing the justices with one-sided logic and reasoning

that bolsters one argument while downplaying the other (Johnson 2004). Because the justices

observe the norm of sua sponte and restrict their decisions to the issues outlined by the briefs

(Black, Hall, Owens and Ringsmuth 2016 and Epstein, Segal and Johnson 1996, but see

McGuire and Palmer 1995), attorneys use the brief to set boundaries around the case and

limit the scope of the justices’ ruling. They walk the justices through complicated areas of

unfamiliar case law and put just the right spin on the precedents to ensure the law favors their

clients (Garner and Roberts 2010), and they offer the justices policy options that might align

with their preferences (and, incidentally, those of the attorney) (Johnson 2001) while giving

the justices legal reasoning to legitimize these more policy-oriented pursuits (Hansford and

Spriggs 2006). Attorneys provide information to the justices, but they do so in the manner

that will best help them secure a win for their clients.

The key to writing a good merits brief is constructing a well-reasoned legal argument

because, as noted legal scholar Bryan Garner points out, no amount of technical expertise

or great writing can overcome a weak or unconvincing argument (Garner 2003). These

arguments have to be solidly based in the law – the United States has a common law system,

after all, so the justices expect to see explanations of how their past decisions will lead them

11

to their current outcomes. The justices expect to see a legal argument that is both carefully

constructed and carefully explained (Garner and Stevens 2010; Garner and Thomas 2010),

and they will borrow wording from the brief itself for their opinions if this is done well (Black

and Owens 2012c; Corley 2008). The heart of the brief is the legal argument, and the justices

respond positively to briefs that offer a “good, clear explanation of what the law is” (Garner

and Roberts 2010, 5). Attorneys just need to align that explanation with a majority of the

justices’ preferences regarding both the law and policy. Their job is to write the kind of

argument that appeals to the justices and wins them over.

2.2 Using Citations to Study Attorney Strategy

While researchers and the justices alike acknowledge that the legal arguments contained

in briefs are influential, studying them is an empirically-challenging task. As Hansford

and Spriggs (2006) point out in their analysis of precedent, the law is “a difficult concept

to measure” because it is “a complex and multifaceted phenomenon that includes a wide

variety of components” (4-5). It is hard to put data-driven paramaters around something as

complicated as the law. As a result of this problem, the research on briefs typically takes

one of two tracks. In the first, scholars use historical and legal analyses to study a single

issue area, like abortion or the death penalty (see, for example, Cushman 1998; Epstein

and Kobylka 1992; Gillman 1993; Mezey 2003; Perry 2007). This type of high-depth, low-

breadth work is incredibly useful for scholars looking to better understand changes in case law

within a small area of the law, but it has limited generalizability to scholars’ comprehensive

understanding of how attorneys use briefs to influence the justices. Alternatively, scholars

focus on the more technical aspects of a brief, offering data-driven analyses of things like

readability and language (Black, Hall, Owens and Ringsmuth 2016; Corley 2008; Feldman

2016; Wedeking 2010). These analyses offer information about how to produce a well-written

brief while skirting around the more important issue of writing a well-reasoned brief. Neither

of these two approaches offer an answer about how to study the law in an empirically-rigorous

12

and comprehensive manner.

Thankfully, a well-developed literature on Supreme Court opinion-writing suggests a way

forward. By looking at the cases the justices cite in their opinions as well as their discussion of

those citations, scholars have successfully managed to study the law. They developed models

of the “legal importance” of cases using data on the justices’ discussion of past cases (Fowler

et al. 2007); they examined “precedent vitality,” or the rigidity with which precedent gets

followed and used (Black and Spriggs 2013; Hansford and Spriggs 2006); and they studied

how the justices use existing precedent to reach conclusions in their opinions (Clark and

Lauderdale 2010). By looking at citation-level data, scholars found a way to look at legal

arguments. Their work requires one assumption: that legal arguments are, most simply,

discussions about how the Court’s past decisions apply (or do not apply) to the situation at

hand.

To study the legal arguments in opinions, scholars use citation-level opinion data that

comes from legal research services like LexisNexis, which owns the Shepard’s Citations tool.

For almost 150 years, LexisNexis has kept a running list of the citations used in Supreme

Court opinions as well as their “treatment” by the justices in those opinions. The treatment

indicates whether the justice who wrote the opinion considered the cited precedent to be

applicable to the situation at hand (positive treatment), wholly different and inapplicable to

the situation at hand (negative treatment), or just a statement of fact (neutral treatment)

(Spriggs and Hansford 2000). Attorneys use this tool to quickly check the status of a prece-
dent before including it in a brief. Scholars use this data to study the law.1 So, for example,

Clark and Lauderdale (2010) use the treatment data to build a case proximity model that
1Westlaw also offers a shepardization service, which it calls KeyCite. The company has
its own approach to classifying citations, including not coding cases as positively treated
and providing information about the level of discussion given to a citation. I am ambivalent
about which system is objectively better at classifying treatment and I focus on LexisNexis
strictly because its coding process is well documented and well tested in political science
(Clark and Lauderdale 2010, 2012; Hansford and Spriggs 2006; Spriggs and Hansford 2000).

13

clusters positively cited cases on one side of a scale and groups negatively cited cases on

the other. The end result is a fine-grained examination of where justices place opinions in

relation to existing precedent and how their own preferences relate to their final opinions.

For the low cost of about $125 a month, this data is available for the taking, and scholars

have benefited greatly from using it.

While the literature on opinions offers a way forward for studying the legal arguments

in briefs, obstacles abound. It is easy enough to define the legal argument in a brief as a

discussion of how past precedents apply to the situation at hand. In fact, recent work by

Hazelton, Hinkle and Spriggs (2019) offers promising evidence that examining citations can

produce new information about attorneys’ legal strategies. But no legal research service

“shepardizes” the citations attorneys use in their briefs. LexisNexis offers the service because

attorneys need to look at opinions quickly and get the information they need; the same need

does not exist for reviewing a legal argument in a brief from 1984. In order to get data similar

to that which scholars use to study opinions, someone would need to shepardize briefs as well.

Advances in text analysis and machine learning make this possible (Schoenherr and Black

2019b), so I utilize a combination of human-assisted coding and machine learning techniques

to train a computer to collect this much-needed data.

2.3 Data Collection

To generate the citation and treatment data and get a first look at the legal arguments

in briefs, I studied LexisNexis’s shepardization process and then developed an automated

version of it. I followed a three-step process: (1) identify each mention of a precedent in

a brief; (2) develop and use a dictionary of legal sentiment to identify each precedent’s

treatment; (3) convert the sentence-by-sentence treatment into a summarized indicator of

how the attorney treated each precedent mentioned in the brief. I use this process to collect

data on the legal arguments contained in 4,812 merits briefs submitted to the Court in 2,232

14

cases the Court reviewed between the 1984 and 2007 terms.2

In the first step of the process, I downloaded the texts of the briefs from LexisNexis

and Westlaw and then broke those briefs down into their legal arguments, specifically the

citations and the attorney’s discussion of those citations. I began by writing a program that

utilized R’s tidyverse suite to identify the “Argument” section of each brief. This section,

which appears after the attorneys walk through the case facts and lower court history, is

typically the longest part of the document, and it contains the attorney’s legal analysis.

After locating the arguments section of each brief, I then broke that section down into
sentences and identified all sentences that clearly cited at least one Supreme Court case.3

My program then separated the citation from the sentence and matched that citation to

data in the Supreme Court Database (Spaeth et al. 2017). At the end of this step, I had a

list of every readily-identifiable mention of a precedent as well as the sentence from which
they came, amounting to 282,815 mentions of 12,172 different Supreme Court precedents.4

In the second step, I created my own dictionary of terms associated with positive and
2About 5% of the cases the Court reviewed during this time period are missing from my
analysis. This is almost exclusively due to the briefs not being available for download from
either LexisNexis or Westlaw. Per conversation with Westlaw’s product management team,
these briefs were never converted to a digital format and are simply missing as a result. The
other cases are missing because they are original jurisdiction cases and, given their unique
nature, I removed them from the analysis.

3I focus on Supreme Court cases and ignore references to state court decisions or lower
federal court decisions. Restricting my analysis to Supreme Court cases is crucial, because
those are the only precedents that bind the justices’ behavior; the justices do not have to
follow the lower federal courts or state courts. I also eliminate references to the federal code
(i.e., 18 U.S. Code §1657).

4By convention, attorneys typically cite a Supreme Court case using its name followed
by the U.S. Report citation and the year the justices published the opinion, e.g., Terry v.
Ohio, 392 U.S. 1 (1968).
Inevitably, however, attorneys turn to shorthand to save space
in their briefs, and the formal citation eventually gets reduced to Terry v. Ohio or 392
U.S. 1 and, at some point, all the way down to merely Terry. The program identifies full
citations and short hand and then uses the quanteda package (Benoit et al. 2017) and its
word frequency statistics to identify and include these less-formal citations as well. When
compared to hand-coded data, the program identifies more than 90% of the total number of
citations mentioned in a brief.

15

negative treatment and used it to identify the treatment of each citation. Dictionary-based

text analysis takes human-created dictionaries and uses word counts to “score” documents

based on their similarity to the dictionary (Monroe, Colaresi and Quinn 2008; Rice and Zorn

2019). I purposefully decided to use a dictionary-based approach rather than employing one

of the more popular black-box models, like a random forest. The dictionary-based approaches

do a good job of identifying sentiment as long as the dictionary accurately reflects the words

associated with those sentiments (Corley and Wedeking 2014; Grimmer and Stewart 2013). I

created this citation-treatment dictionary because no pre-existing sentiment dictionary could

accurately reflect the words used in legal documents – the law is, after all, its own specialized
language (Rice and Zorn 2019).5

To understand how the dictionary works in practice, consider that writers can “posi-

tively” cite cases that are similar to the case’s outcome, like Kathryn Kolbert did when she

suggested the guarantees outlined in Roe v. Wade applied to Casey. They can also treat

cases “negatively” and suggest a past precedent does not apply or should be overturned, as

Kolbert did when discussing Harris v. McRae (1980). Or they can discuss cases in a neutral,

fact-based manner, as Kolbert did when discussing Griswold v. Connecticut (1965). Essen-

tially, attorneys use certain words to indicate treatment – they “apply” a past precedent to

the current case, or they encourage the justices to “distance” the current case from a past

one. To find these sentiment-specific words and create a dictionary of legal sentiment, I

utilized data from three different sources: (1) LexisNexis’s list of shepardization terms and

their associated treatments; (2) the shepardization guides provided by Spriggs and Hansford
5As a thorough political scientist, I should note that I did attempt to classify sentiment
using three different supervised machine learning approaches: naive Bayes, support vector
machines, and random forest (Kuhn and Johnson 2016). After wasting two beautiful sum-
mer days at my laptop, the results suggested the dictionary did the best job of correctly
identifying attorneys’ treatments of citations. Importantly, the dictionary also did the best
job of directing mis-categorized treatments toward the neutral category, rather than putting
them in the opposite category (e.g., true negative treatments being categorized as positive
treatments).

16

(2000) and Hansford and Spriggs (2006); and (3) information gathered from 57 randomly-

selected search and seizure cases and 10 privacy cases that I, along with a team of research

assistants, hand coded for treatment. After engaging in an iterative process of creating a

dictionary, testing it against additional hand-coded data, correcting it, and beginning the

process again, I ended up with a dictionary containing 525 words associated with attorneys’

treatment of citations. With the help of the quanteda package in R (Benoit et al. 2017), I

applied the dictionary to the sentences associated with each citation to identify positive and

negative treatment of each citation. Anything not marked as positive or negative got treated

as a neutral, fact-based citation.

In the third step, I took the sentence-by-sentence treatment of each citation and created

an overall measure of precedent treatment. I shepardized at the sentence level in order to

avoid the dictionary identifying a citation as both positive and negative, something that

happened repeatedly when conducting this analysis at the paragraph level. As a result,

however, I had multiple mentions of the same precedent, which meant that I had to create

some sort of summary measure. Consider, for example, that the attorneys representing

the petitioners in Lawrence v. Texas (2003), which overturned sodomy laws in the United

States, mentioned Planned Parenthood v. Casey (1992) 21 times in their brief. To create

my summary measure, I summed the number of times the case got treated positively and

then subtracted from it the summed number of times the case got treated negatively. Cases

that maintained a positive number got treated as positive; cases that maintained a negative

number got treated as negative. If the summation process resulted in a zero (i.e., the attorney

treated the citation positively and negatively, and distributed that treatment equally), I

manually checked the citations to determine the correct direction of the treatment. So,

continuing with the Lawrence example, the attorney who wrote the brief treated Casey

positively six times and negatively once. Because five minus one is a positive number, the

summary measure indicates that Lawrence treated Casey positively overall. Anything with

17

no directionality got treated as a neutral citation.6 At the end of this step, I had a list of

122,299 unique brief-citation pairings, as well at the attorneys’ treatment of those citations.

2.4 Analysis

With the data collected, the question becomes: what can scholars do with it? I have

two main goals for my exploration of it: to better understand briefs’ content and to better

understand how briefs relate to each other. I aim, in short, to show this data offers valuable

insight into how attorneys use the law to appeal to the justices. To do this, I use a combina-

tion of lists and descriptive statistics culled from the citation and treatment data I collected.

For the most part, I analyze the briefs as a collective block of information, but I do also split
the data into issue areas in order to examine how attorneys work within an area of law.7

Beginning first with citation trends over time, the left panel of Figure 2.1 shows that

attorneys discuss significantly more precedents in their briefs over time. The right panel

of Figure 2.1 suggests attorneys are also citing significantly more cases while doing it. An

average brief submitted in during the 1984 term mentioned 21 cases 47 times. But an average

brief submitted during the 2007 term had 76 mentions of 31 cases. While the data suggest

attorneys maintain a constant average of just over two mentions of each case per brief, the

citation trend suggests that as the Court establishes more precedents, attorneys simply cite

more cases. Attorneys are not necessarily replacing old precedents with new, but rather are

adding more cases to the existing list.

6I manually checked the direction of any precedent that got mentioned 15 or more times.
This amounted to somewhere around 1-2 cases per brief. To be clear, the aggregation method
worked for almost every citation. I looked at these high-mention cases to be sure an attorney
did not repeatedly praise a case and then end the brief with “The Court should overturn Roe
v. Wade (1973).” This did happen occasionally.

7The Supreme Court Database separates each case into one of 14 distinct issue areas,
including civil rights, criminal procedure, and privacy law (Spaeth et al. 2017).
I follow
their categorization scheme for my own discussion, but should note that none of the cases
I examine fell under one issue area, so I ultimately only look at cases covering 13 different
areas of the law.

18

Figure 2.1: Citation Patterns in Supreme Court Merits Briefs Over Time – Open dots indicate the
average total number of citations in a merits brief (left) and the average number of cases cited in a brief
(right) between the 1984 and 2007 terms. The lines are non-parametric lowess curves of the overall trend.

Importantly, however, attorneys’ discussions of cases vary dramatically. Figure 2.2 is a

histogram of the number of times attorneys discuss a case in a brief, going from 1 mention
to 17 mentions, which covers 99% of the data under study here.8 According to Figure 2.2,

attorneys typically only mention a case once – they cite it and and then move on. There

are, however, several cases in each brief that get mentioned repeatedly, and occasionally ex-

haustedly. These, I suggest, are the central precedents, the cases worthy of deep examination

and explanation (Garner 2003), and the cases around which the justices focus their analysis

(Garner and Stevens 2010; Garner and Thomas 2010). The data make it easy for scholars

to identify these cases based on citation patterns and can help them better understand how

core arguments influence outcomes at the Court.

8There are briefs where attorneys mention the same case more than 100 times, like the
United States government did in Rasul v. Bush (2004) when it repeatedly brought up
Johnson v. Eisentrager (1950). I eliminate these outliers from the histogram to better show
the distribution of the data and offer a more useful plot.

19

4550556065707580198419861988199019921994199619982000200220042006Supreme Court TermAverage Number of Citations in a BriefAverage Citations Per Brief202326293235198419861988199019921994199619982000200220042006Supreme Court TermAverage Number of Cases Appearing in a BriefAverage Cases Mentioned Per BriefFigure 2.2: Histogram of Mentions of a Case in a Brief – This histogram shows the number of times
a case got mentioned in a brief.

Turning next to attorneys’ treatment of the citations they utilize, the data in Figure 2.3

suggest attorneys like to focus on the cases that apply and proceed carefully with the cases

they believe are inapplicable to the situation at hand. Figure 2.3 also shows that the increase

in citations noted in Figure 2.1 stems from the increased use of positive citations. Beginning

first with the hatch marks, which represent the average number of positive citations (left)

and cases mentioned (right) in a brief, the data suggest attorneys spend most of their brief

discussing the past precedents that apply to the situation at hand, and have leaned into that

role over time. In the 1984 term, an average brief put a positive spin on 14 cases that got

mentioned 34 times. By the 2007 term, that number increased to briefs discussing an average

of 20 positively-treated cases 57 times. Negative and neutral citations appear significantly

less often and do not exhibit the same growth trend over time. As the Xs on Figure 2.3

show, attorneys’ use of negative citations stayed relatively constant over time, shifting from

7 mentions of 3 inapplicable precedents to 10 mentions of 4. The open dots, which indicate

neutral citation patterns, show the same unmoving trend, going from 9 mentions of 6 neutral

20

020000400006000080000051015Total Citations of a Case in a Briefcases to 11 mentions of 8 neutral cases.

Figure 2.3: Trends in Citation Treatment Over Time – Hatch marks indicate the average number of
citations (left) of positively-discussed cases (right) in a brief per term. Xs show the same for negatively-
discussed cases, and open dots represent the neutrally-discussed cases. The solid (positive), dashed (nega-
tive), and dotted (neutral) lines are non-parametric lowess curves of the overall trend.

These trends do not change between briefs. The top two panels of Figure 2.4 show that

petitioners and respondents follow the same citation habits, and the bottom two panels show

that, on average, the winners and losers of a case do the same. This would seem to suggest

that writing a brief is a reasonably formulaic process in terms of basic construction, with

attorneys focusing on the cases that help their argument and saving pointed criticism for

the ones that do not.

21

0102030405060198419861988199019921994199619982000200220042006Supreme Court TermAverage Number of Citations Appearing in a BriefAverage Citations Per Brief0246810121416182022198419861988199019921994199619982000200220042006Supreme Court TermAverage Number of Cases Appearing in a BriefAverage Cases Mentioned Per BriefFigure 2.4: Trends in Citation Treatment for Petitioners, Respondents, Winners, and Losers
Over Time – Hatch marks indicate the average number of positive citations in a brief per term. Xs show
the same for negatively-discussed cases, and open dots represent the neutrally-discussed cases. The solid
(positive), dashed (negative), and dotted (neutral) lines are non-parametric lowess curves of the overall trend.

While looking at this over-time citation data can help scholars understand how attorneys

broadly approach their briefs, examining the same data on a case-by-case basis offers insight

into how attorneys use these citations to build their arguments. Figure 2.5 is a heat map

of the 20 most-cited cases between the 1984 and 2007 terms. The case names are listed on

the y-axis, and the Supreme Court term is listed on the x-axis. Each box represents the

average number of times a brief mentioned that precedent in that term, with the darker

colors signaling more frequent use of the citation. More simply, the heat map shows how

often attorneys treat these heavily-cited cases as a central part of their arguments. So, for

example, Figure 2.5 shows that Miranda v. Arizona (1966) was a well-utilized precedent

through much of the 1980s and early 1990s. Attorneys mentioned it 25 times in their briefs

22

05101520253035404550556065198419861988199019921994199619982000200220042006Supreme Court TermAverage Citations in a BriefPetitioner05101520253035404550556065198419861988199019921994199619982000200220042006Supreme Court TermAverage Citations in a BriefRespondent05101520253035404550556065198419861988199019921994199619982000200220042006Supreme Court TermAverage Citations in a BriefWinning Party05101520253035404550556065198419861988199019921994199619982000200220042006Supreme Court TermAverage Citations in a BriefLosing Partyin the 1984 term, and 20 times in each brief in the 1993 term. The heavy citation indicates

attorneys who brought up Miranda were actually building their arguments around it. Over

time, however, Miranda’s popularity tailed, and by the 2007 term, attorneys were giving it

a single, obligatory mention in their briefs before focusing on other precedents. The data in

Figure 2.5 also shows that precedent like Chevron U.S.A., Inc. v. National Resource Defense

Council (1984) is frequently utilized in briefs but rarely the central precedent around which

attorneys build their arguments.

Figure 2.5: Heat Map of the 20 Most-Cited Supreme Court Precedents – Each box represents the
average number of times attorneys mentioned that precedent in their briefs that term; the darker the color,
the higher the average.

Additionally, scholars can also use this data to identify new issues that are making their

way to the Court. Consider, for example, the citation trend of Regents of the University of

California v. Bakke (1978), which upheld affirmative action programs in higher education.

23

Boyde v. California (1990)Teague v. Lane (1989)Thornburg v. Gingles (1986)Batson v. Kentucy (1986)Caldwell v. Mississippi (1985)Bacchus v. Dias (1984)Chevron v. NRDC (1984)Strickland v. Washington (1984)Harlow v. Fitzgerald (1982)Eddings v. Oklahoma (1982)Parratt v. Taylor (1981)Lockett v. Ohio (1978)Regents of UC v. Bakke (1978)Monell v. Dept Social Services (1978)Gregg v. Georgia (1976)Buckley v. Valeo (1976)Lemon v. Kurtzman (1973)Bivens v. Six Unknown Named Agents (1971)Terry v. Ohio (1968)Miranda v. Arizona (1966)198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007Supreme Court Term20 Most-Cited Supreme Court CasesAs Figure 2.5 shows, attorneys used Bakke with some frequency in the mid-1980s, mentioning

it about five times a brief, but attorneys eventually stopped citing it, at least in part because

the Court was not reviewing affirmative action programs. But Bakke made a reappearance

in the 2000 term, around the time interest groups began putting affirmative action in college

admissions back on the lower courts’ dockets (Perry 2007). By the time the Court heard

oral arguments in the two University of Michigan affirmative action cases during the 2003

term, Bakke was again a central precedent in the briefs that used it. The data reflect these

changes.

Finally, Figure 2.5 also shows how quickly attorneys might begin using the justices’ new

decisions in their briefs. While the Court settled many of these cases in the 1970s and

early 1980s, 8 of the 20 most-cited cases under study here were decided during or after the

1984 term. The data in Figure 2.5 suggests that, with the understandable exception of

Thornburg v. Gingles (1986), a redistricting case that would not become important until

after the 1990 census, attorneys begin citing the Court’s recent opinions within a term of the

original decision. Batson v. Kentucky, the Court’s 1986 decision that attorneys could not use

peremptory challenges to strike jurors based on their race, was particularly well-cited during
the 1986 term.9 Attorneys begin using the Court’s decisions as soon as they relevantly can,

and they begin building their cases around the new precedents as well.

While Court-wide trends are important, the data can also offer an interesting look at

area-specific trends as well. Table 2.1 shows summary statistics for each of the thirteen issue

areas under study here. More than 70% of the cases come from four issue areas: criminal

procedure, economic activity, civil rights, and judicial power. These types of cases dominate

the Court’s docket, giving attorneys more opportunities to appeal to the justices in those
9This is not at typo, a fundamental misunderstanding of how time works, or some sort
of Hermione Granger Time Turner Escapade. Supreme Court terms run from October to
June and are identified by the year in which they start, e.g., the 1986 term began in October
1986. The Court decided Batson during the 1985 term but released the opinion in April of
1986, and attorneys began citing Batson in October 1986.

24

areas. There are three things worth noting in this table. First, note that the most-cited

cases in Table 2.1 are not always the most-mentioned cases; in privacy cases, for example,

the most-cited precedent is Planned Parenthood v. Casey (1992), but the citation that

appears in most briefs is Roe v. Wade (1973). Second, it is worth mentioning that while

attorneys are most likely to treat Miranda v. Arizona (1966) as a centralized precedent, they

are more likely to at least offer an obligatory mention of Chevron U.S.A., Inc. v. National

Resource Defense Council (1984). Finally, many of these cases are older precedents. This is

not unexpected – as Hansford and Spriggs (2006) point out, new cases have simply had less

time to make their appearances.

25

Table 2.1: Summary Statistics for Citation Patterns in Briefs

Issue Area

Cases Briefs

Most Mentioned Precedent

Precedent with Most Appearances

Criminal Procedure

Civil Rights

First Amendment

Due Process

Privacy

Attorneys

Unions

Economic Activity

Judicial Power

Federalism

Interstate Relations

Federal Taxation

Miscellaneous Cases

563

342

168

99

43

41

61

402

309

132

2

61

9

1,137

Miranda v. Arizona (1966)

Miranda v. Arizona (1966)

745

385

215

97

83

140

892

669

295

4

124

26

Miranda v. Arizona (1966)

Chevron U.S.A., Inc. v. N.R.D.C. (1984)

Buckley v. Valeo (1976)

Buckley v. Valeo (1976)

Loretto v. Teleprompter Manhattan (1982)

Penn Central Transportation Co. v. N.Y.C (1978)

Planned Parenthood v. Casey (1992)

Roe v. Wade (1973)

Hensley v. Eckerhart (1983)

Hensley v. Eckerhart (1983)

Abood v. Detroit Board of Education (1977)

Chevron U.S.A., Inc. v. N.R.D.C. (1984)

Bivens v. Six Unknown Fed. Agents (1971)

Chevron U.S.A., Inc. v. N.R.D.C. (1984)

Batson v. Kentucky (1986)

Chevron U.S.A., Inc. v. N.R.D.C. (1984)

Cipollone v. Liggett Group, Inc. (1992)

Hines v. Davidowitz (1941)

Nevada v. Hall (1979)

Pacific Ins. Co. v Industrial Acc. Comm. (1939)

National Carbide Corp.v. Commissioner (1949)

National Muffler Dealers Assn. v. U.S. (1979)

I.N.S. v. Chadha (1983)

I.N.S. v. Chadha (1983)

Overall

2,232

4,812

Miranda v. Arizona (1966)

Chevron U.S.A., Inc. v. N.R.D.C. (1984)

26

Finally, Figure 2.6 shows a heat map of the 20 most-cited First Amendment cases. Some

of the Court’s most-cited cases from Figure 2.5 make an appearance here as well, including

Lemon v. Kurtzman (1973) and Buckley v. Valeo (1976). Given the low number of First

Amendment cases the Court reviews, the high citation rates of these cases suggest their

relevance might transcend their issue area. But for the most part, the cases that appear in

Figure 2.6 are the cases one would expect to see cited in a First Amendment case, including

Tinker v. Des Moines (1969) and Lynch v. Donnelly (1984). This issue-area specific data

also makes it easier to identify cases that are typically cited together, like Buckley and

McConnell v. F.C.C. (2003), while also making it easy to identify when one precedent

replaces another, like when attorneys stopped using Sherbert v. Verner (1963) after the

Court replaced it with Employment Division of Oregon v. Smith in 1990. This particular

data makes it easier to understand the logistics of First Amendment briefs and perhaps

identify how one should look.

27

Figure 2.6: Heat Map of the 20 Most-Cited First Amendment Precedents – Each box represents
the average number of times attorneys mentioned that precedent in their briefs that term; the darker the
color, the more they mentioned the case.

These figures collectively show that the legal arguments in Supreme Court merits briefs

offer valuable insights into attorney strategy and influence. Having this kind of data opens

up the study of briefs specifically and the law more broadly. While I simply use descriptive

data here, scholars could use this data for more empirically-rigorous pursuits, including a

study of how attorneys use precedent to sway the justices.

2.5 Conclusion

When Kathryn Kolbert argued in her merits brief in Planned Parenthood v. Casey that

upholding the Pennsylvania Abortion Control Act would overturn Roe v. Wade, she ran the

risk of angering the justices. As journalists have documented, the justices purposefully tried

to move her argument away from Roe, and Kolbert’s refusal to do so frustrated the justices

28

McConnell v. FCC (2003)Riley v. National Federation of the Blind of NC (1988)City of Renton v. Playtime Theatres (1986)Witters v. Washington Services for the Blind (1986)Lynch v. Donnelly (1984)Perry Education Assoc. v. Perry Local Educators' Assoc. (1983)New York v. Ferber (1982)Widmar v. Vincent (1981)FCC v. Pacifica Foundation (1978)Abood v. Detroit Board of Education (1977)Buckley v. Valeo (1976)Gertz v. Welch (1974)Committee for Public Education v. Nyquist (1973)Lemon v. Kurtzman (1973)Walz v. Tax Commission (1970)Tinker v. Des Moines (1969)Pickering v. Board of Education (1968)Freedman v. Maryland (1965)New York Times Co. v. Sullivan (1964)Sherbert v. Verner (1963)198419851986198719881989199019911992199319941995199619971998199920002001200220032004200520062007Supreme Court Term20 Most-Cited First Amendment Casesduring oral argument (Toobin 2008). Indeed, the justices nearly decided that Casey was

the proper vehicle for overturning Roe and allowing states ban abortions completely (Toobin

2008). But Kolbert was savvy, and her decision to frame Casey as a referendum on Roe paid

off. Three justices – O’Connor, Souter, and Kennedy – were uncomfortable with the idea of

overturning such a long-lasting and controversial precedent, so they formed the now-famous

“troika” that saved the “essence” of Roe. In writing the argument the way she did, Kolbert

managed to persuade the justices toward her side; her brief influenced their approach toward

the case.

Engaging with Supreme Court justices and trying to persuade them toward a side is a

delicate task. Attorneys balance legal arguments with ideological appeals, carefully weighing

everything from their phrasing to their off-the-cuff responses in oral argument, all in the name

of trying to win. In this paper, I suggest that merits briefs are important and understudied

legal documents that can influence the justices’ decision-making processes. I then introduce

a new approach for empirically examining the legal arguments in briefs. Following work by

Hansford and Spriggs (2006) and Clark and Lauderdale (2010), I apply machine learning

techniques to a large corpus of data to extract information about the content of merits briefs

and their relationships to each other. I then walk through the data to show its value and

offer potential avenues for later analysis. Studying briefs is possible as long as scholars use

the right tools.

As with any text analysis project, there are limitations to what the data can accom-

plish. For one thing, text analysis requires clean data, something that is not always readily

available. Given the time period under examination here, digitization mistakes were not

unexpected – taking documents originally created on typewriters or early word processors

and turning them into searchable text files is a messy process (Lane and Schoenherr 2019),

even when done by professionals like LexisNexis or Westlaw. The text files are occasion-

ally missing words or they repeat sections, though they seem to do so in a mostly random

29

manner.

Additionally, attorneys sometimes make mistakes, and I am helpless against their errors.

Consider, for example, that an attorney’s typo regarding a U.S. Report volume number could

cause my program to identify a case as Californians v. California, 393 U.S. 1 (1968), rather

than Terry v. Ohio, 392 U.S. 1 (1966). People who are reading the briefs line by line will

catch these mistakes – in fact, Justice Blackmun would correct them in his copies of briefs,

occasionally adding a less-than-polite comment about the attorney who made them. I sadly

do not have Justice Blackmun correcting the briefs here. If an attorney put down the wrong

information, that case will simply be miscoded in my data. Both issues are things that

constant supervision and validation of the data can help eliminate (Grimmer and Stewart

2013) and I have worked had to keep these issues at bay.

Moving forward, my first goal is to apply this tool to merits briefs up to the 2018 term,

and to continue updating this data moving forward. When the data is completely collected,

cleaned, and ready to merge with the Supreme Court Database, researchers can use it (and

the method to get it) to provide fresh perspectives on existing and important debates. Con-

sider, for example, that scholars could update Johnson’s (2004) analysis of the information

overlap between briefs and oral argument. They could see if Johnson’s contention that the

justices use oral argument to reconcile information biases is still true today.

Additionally, I ultimately would like to follow Clark and Lauderdale (2010) and use

the citation patterns to place the briefs into ideological space. Early attempts suggest this

task is more difficult with briefs, as attorneys will never mention a brief again and it is

therefore difficult to reliably identify briefs’ ideological positions over time. I will continue

to push forward on this front, however, in hopes to eventually be able to see how the briefs

ideologically compare to the eventual Supreme Court opinions.

30

CHAPTER 3

CALL AND RESPONSE: LEGAL ENTREPRENEURSHIP AND

ATTORNEY SUCCESS AT THE U.S. SUPREME COURT

When Ruth Bader Ginsburg established the Women’s Rights Project at the American

Civil Liberties Union in 1971, she did so with a mandate: develop a legal strategy for

improving gender equality in the United States (Campbell 2003). She decided equality had

to come from the Supreme Court and set her sights on establishing that all laws separating

the sexes needed to pass a strict scrutiny test (Hirshman 2015). To get there, she would

have to approach the nine male justices and convince them that the “preferential” treatment

women received was discriminatory, something they had long refused to do (Hirshman 2015).

Ginsburg had to find a believable way to make a new and innovative argument. Starting

with her first Supreme Court brief in Reed v. Reed (1971), Ginsburg suggested the justices

needed to overturn their past decisions on sex-based discrimination and then drew a parallel

between the racial discrimination cases the Court reviewed in the 1950s and 1960s and her

cases (DeHart 2018). She suggested that sex discrimination and racial discrimination were

similar (though not equally oppressive) problems that required the same solution (DeHart

2018). The Court had already decided that discrimination on the basis of race was both

impermissible and worthy of the highest form of scrutiny, so why not take the argument one

step further? Ginsburg’s decision to align her radical argument with a familiar one worked,

too (DeHart 2018). She eventually got the justices to rule that any legislation that separated

the sexes deserved at least some level of scrutiny, if not the strictest level.

Like Ginsburg, attorneys who appear before the Supreme Court face the difficult task

of deciding how to use the Supreme Court’s past decisions to direct the justices toward a

desired outcome. By design, the Supreme Court uses its opinions to resolve legal conflict

and important, unanswered legal questions (Perry 1991), and parties appeal to the Court

31

precisely because they want the clarity and finality that lower courts failed to provide. But

while the justices can locate cases worthy of examination, they need help navigating to the

eventual result. Attorneys are the ones who put together initial thoughts on the relevant case

law and offer the justices legal arguments about the situation at hand (Garner 2003). They

compose merits briefs for the justices’ use and, in the process, establish the legal boundaries

of a case while shaping the justices’ responses to it (Epstein, Segal and Johnson 1996). Doing

this requires attorneys make a key strategic decision regarding the construction of the legal

argument: do they work within the prevailing legal framework and find a way to wedge their

issue into it, or do they, like Ginsburg, offer a wholly new and entrepreneurial argument

that redefines the justices’ thoughts about an area of case law? This paper addresses that

question.

To examine this facet of attorney strategy, I introduce the concept of legal entrepreneur-

ship, which occurs when attorneys attempt to push the law in new or radical directions in

order to upend the legal status quo. I develop an original measure of legal entrepreneurship

to systematically study how innovative legal arguments influence the justices’ votes in a case.

Using data from 3,018 merits briefs across 1,509 cases the Court heard between the 1984 and

2007 terms, I find that attorneys’ decision to engage in legal entrepreneurship can have very

real consequences on their ability to win their cases. More specifically, I find that attorneys

representing the petitioner are more likely to win a justice’s vote if they engage in legal

entrepreneurship, but respondents gain no advantage from doing the same. Additionally, I

find that engaging in legal entrepreneurship helps inexperienced petitioners or petitioners

representing resource-poor litigants gain a clear advantage. These results indicate that, in

certain circumstances, attorneys can benefit from using a more radical argument in their

briefs.

These findings make two new and important contributions to the literature. First, my

results suggest that learning how and when to make a certain type of argument is a crucial

32

piece of being a winning (good) attorney. Attorneys are more successful when they are

credible (Wedeking 2010). The well-developed literature on attorney experience shows that

experienced attorneys are more likely to win before the Supreme Court (McGuire 1993, 1995)

and see their arguments appear in the justices’ opinions (Corley 2008). The authors of these

works universally point to established credibility as an explanation for this success. My

research goes beyond the outcomes to show how attorneys appeal to the justices and build

credibility with them. It proposes that knowing when and how to make novel arguments

– that is, knowing when to use the prevailing frame and when to be entrepreneurial – is a

necessary tool in any winning advocate’s toolbox.

Secondly, I offer one of the first systematic analyses of attorney decision making. Much of

what political scientists understand about attorney strategy and legal change comes from the

two most famous examples of successful cause lawyering: Thurgood Marshall’s twenty-year

fight for racial equality (Haygood 2016) and Ruth Bader Ginsburg’s aforementioned battle

for women’s rights in the 1970s (DeHart 2018). Scholars are well-versed in the language of big

legal movements, but they understand much less about how members of the Supreme Court

bar generally build arguments and present them to the justices in order to nudge the law

in their preferred directions (though see Wedeking 2010 and Hazelton, Hinkle and Spriggs

2019 for exceptions). By studying legal entrepreneurship more broadly, I draw attention

away from the legal giants and their success stories and place it on the more typical attorney

practices that dominate the Court’s docket.

3.1 Attorneys, Briefs, and the Law

In the United States, legal change starts with attorneys, not with judges. The federal

judiciary’s adversarial structure empowers litigants and their attorneys to identify legal prob-

lems or conflicts, bring suit, and then go before judges and present arguments to resolve the

issues (Kagan 2003). Judges must wait for attorneys to first put cases on their dockets (Baird

2004) and later provide them with the information needed to rule on them (Garner 2003).

33

While the judges are the ones who ultimately decide the cases and offer the legal rationale

for getting there, the attorneys are the ones who put issues in front of them and offer the

initial arguments that eventually shape judges’ understandings of the law.

When attorneys present a case to the Supreme Court, they have the opportunity to

change the law and policy on a nationwide scale. Cases that go before the Court involve

legal questions that the lower courts failed to answer with any level of certainty (Perry 1991),

and by accepting these cases for review, the justices agree to provide an answer the federal

courts can uniformly implement. Yet before the justices can answer a question, they do what

all judges do: they ask the attorneys to take the first pass at an answer and provide a legal

argument that walks the justices from the conflict to a proposed resolution. In so doing, the

justices give attorneys the opportunity to shape the decision the Court will eventually make,

as the justices depend on attorneys’ information and expertise to get them through the case

(Johnson 2004; Garner and Roberts 2010; Garner and Thomas 2010). Attorneys get two

opportunities to do this: first in written merits briefs and then later during oral argument.

Merits briefs are supposed to make the justices’ jobs easier (Scalia and Garner 2008).

Each party submits a brief, which, per Supreme Court Rule 24, provides information the

justices need to review and understand the case, including procedural background, the legal

questions at hand, the party’s proposed solution, and accompanying legal reasoning. The

justices also expect attorneys to do the legal heavy lifting in a case; they want attorneys to,

as Justice Clarence Thomas explained, really “tee it up” and make a tough case look simple

and easy to resolve (Biskupic, Roberts and Shiffman 2014, Part 3). The justices are clear

about what they want in a brief – reliable information, good writing, and brevity (Garner

and Ginsburg 2010; Garner and Kennedy 2010; Garner and Roberts 2010; Totenberg 2011)

– and they are equally clear that good advocates do those things in their briefs (Garner and

Scalia 2010). Research also suggests the justices favor briefs that are easy to read and lack

affective language (Black, Hall, Owens and Ringsmuth 2016; Feldman 2016).

34

While merits briefs ostensibly exist to help the justices, they also provide attorneys with

their best opportunity to influence the justices’ approach to a case. By nature of being a

one-sided legal document, the merits brief gives an attorney space to make an uninterrupted

and coherent argument, which means she can put her most persuasive spin on the case,

bolstering her argument while downplaying the opposition’s (Black and Owens 2012c; Black,

Hall, Owens and Ringsmuth 2016; Johnson 2001). She has to be informative and cannot lie,

misrepresent, or overstate the issues in the case (Garner 2003), but she does this in a manner

that best helps her client. Importantly, the attorney has to do this in the brief and not at oral

argument, as the justices use oral argument to engage in their own information-gathering

activities and leave little time for attorneys to make full arguments (Black, Johnson and

Wedeking 2012; Johnson 2004).

The key to persuasion is a solid legal argument. Legal writing guides abound with

suggestions like Dorrill and Harwell’s (1987) comment that “to persuade [judges], you have

to offer them sound reasons for believing as you do” (as cited in Garner 2003, 105), or Garner’s

(2003) note that “When you write a brief, your implicit promise is that you’ll give the judge

good reasons for ruling as you request” (ix.) Attorneys need to take the justices from the

conflict at the center of the case to its proposed solution in a manner so elegant the justices

cannot help but accept the argument (Garner and Roberts 2010), perhaps even going so far

as to use the attorneys’ words as their own when writing their opinions (Corley 2008). Having

that winning argument anchoring a brief is crucial for success; as Garner (2003) points out,

technical excellence and great writing can enhance a good argument and make persuasion

easier, but those skills cannot overcome a brief with a weak and unconvincing argument.

Given the justices’ well-documented proclivity for siding with experienced attorneys over

novices (McGuire 1993, 1995; Nelson and Epstein 2019), it is not difficult to surmise that

learning how to write a solid legal argument takes time.

35

3.2 Legal Entrepreneurship and Innovation

Preparing an argument for review at the Supreme Court requires careful consideration

of a multitude of factors. Attorneys need a strategy for appealing to a majority of the

justices, nine legal elites whose personalities, policy preferences, and social circles help guide

their decision making (Baum 2006; Black et al. 2020; Epstein and Knight 1998; Hall 2018;

Segal and Spaeth 2002). Petitioners must decide how to respond to a lower-court loss,

while respondents decide how to approach a reverse-prone Court (Wedeking 2010), and both

sides have to produce briefs that sound credible and authoritative about the issues at hand.

To do this, attorneys have to position their arguments alongside existing precedent (Black

and Spriggs 2013; Hansford and Spriggs 2006), but the common-law system only offers

so much guidance to attorneys seeking to resolve cases that currently lack clear solutions.

Consequently, attorneys find themselves constructing arguments that venture into unknown

legal territory in what they hope is the most believable manner possible. Figuring out how to

use past Supreme Court precedent to suggest answers to otherwise-unanswerable questions

is an art, and attorneys do it in every case.

To see how this works in practice, consider Jan Walls Anderson’s brief for Charles

Acevedo, the respondent in the search and seizure case California v. Acevedo (1991). An-

derson wrote that,

The factor that distinguishes this case from United States v. Ross, 456 U.S. 798,
823 (1982) is that in Ross the police had probable cause to believe that the vehicle
and the trunk of the vehicle specifically contained narcotics. Here, as in United
States v. Chadwick, 433 U.S. 1 (1977), the only connection the contraband has
with the vehicle is the fact that the container carrying it was placed into the
trunk of the vehicle.

Anderson’s goal was to get the justices to side with Charles Acevedo, who claimed police

used an illegal search to find marijuana in a paper bag in his trunk. The Court did not have

a ruling that directly addressed searches of paper bags in car trunks (Acevedo would become

36

that precedent), but Anderson found other rulings to help her make her point. She identified

two precedents that were relevant to her argument, Ross and Chadwick, and discussed their

application to the case. The ruling in Ross, a well-cited search and seizure case in which the

justices upheld a trunk search for drugs, was not favorable to her argument, so she sought to

differentiate her case from it. But the Court’s decision in Chadwick was useful to her, given

the Court ruled that a warrantless search of locked luggage in a vehicle violated the Fourth

Amendment, so she talked about its relevancy to her case. Anderson, like other attorneys

who appear before the Court, identified relevant case law and then used her merits brief to

carefully walk through the precedents and explain their application (or lack thereof) to the

situation at hand. This is how attorneys credibly build an argument.

When putting together these arguments, attorneys can use one of two competing frames

to aid construction: they can use a prevailing legal argument, or they can engage in legal

entrepreneurship. When attorneys use a prevailing argument, they situate their arguments

on top of the Court’s recent rulings in an issue area, suggesting the justices already have the

answer to the case if they just take the argument one step further. This is what Anderson

is doing in her argument, trying to convince the justices that Acevedo is just like Chadwick

and not at all like Ross. Her goal is to show the justices their existing rulings placed easily-

applicable boundaries around the case. Briefs that use a prevailing argument should look

and read like the Court’s more recent opinions in an area of case law. They should cite the

same cases the Court did and they should discuss those cases in the same manner.

Attorneys should default to using the prevailing argument for one simple reason: they are

easier for the justices to consume and process. The justices want attorneys to make simple

arguments that are familiar, repetitive, and relatively easy to understand (Garner 2003;

Hazelton, Hinkle and Spriggs 2019). They like legal efficiency – that is, rules that are easy

to apply – and ideally aim to create “bright-line” rules that simplify decision making (Niblett,

Posner and Shleifer 2010). When the law is inefficient, the justices spend time and energy

37

identifying answers to legal questions (Gennaioli and Shleifer 2007), something they do in

place of pursuing their personal interests, which they also like to do (Epstein and Knight

2013). Beyond mere inefficiency, the justices are also, for the most part, institutionally averse

to overturning existing precedent (Hansford and Spriggs 2006) and are wary of anything

that suggests they distinguish or eliminate their own precedents.

Importantly, prevailing

arguments are also easier to identify and work with, which should help eliminate some of the

difficulty surrounding an attorney’s venture into the legal unknown.

Entrepreneurial arguments, on the other hand, are anything but familiar and easy. When

attorneys engage in legal entrepreneurship, they act like legal renegades. They introduce

new arguments that seem out of place in the Court’s recent jurisprudence – they can suggest

the Court overturn its existing approach to an area of case law, apply a different line of

jurisprudence, or build a case around a long-ignored precedent. Ruth Bader Ginsburg did this

in her sex-based discrimination arguments, asking the justices to overturn valid precedents

like Muller v. Oregon (1908) and Hoyt v. Florida (1961) while venerating the justices’

decision in Brown v. Board of Education (1954) and drawing parallels between two otherwise-

different areas of case law. She tried to win by changing the conversation and forcing the

justices to reexamine their approach to an area of case law.

Using an entrepreneurial argument is a high-risk, high-reward enterprise. When attorneys

write an entrepreneurial brief, they ask the justices to go through the mentally-taxing exercise

of looking at an established area of case law in a brand new light. Convincing experienced

legal minds they missed something in the law or, worse, made a mistake and need to correct

it, is not an easy task (Mauro 2019). Moreover, given the justices’ preferences for efficiency

and familiarity, entrepreneurial arguments would seem to guarantee failure. But Supreme

Court justices are also strategic seekers of policy who will occasionally invite inefficiency

for the sake of policy gains (Epstein and Knight 2013; Niblett, Posner and Shleifer 2010),

and the justices are willing to use the law to bolster their policy pursuits (Hansford and

38

Spriggs 2006). There is just enough uncertainty in the law and the justices’ decision-making

processes to make entrepreneurship an intriguing option for attorneys who, like Ginsburg,

could not possibly win using prevailing arguments. When you have nothing to lose, why not

try for the legal equivalent of a Hail Mary pass? If an attorney can get the argument just

right and persuade the justices toward her side, the decision to go entrepreneurial could pay

off handsomely.

Importantly, however, entrepreneurship should not work equally well in all situations.

For one thing, engaging in legal entrepreneurship should benefit the petitioner more than

the respondent. Recall that the petitioner appeals to the Court with a loss already in hand,

which would suggest the prevailing case law, as understood by the lower court judges, already

worked against her once (Wedeking 2010). Using a prevailing argument is consequently less

appealing, all else being equal. If the petitioning attorney can find a way to believably and

credibly appeal to the justices using an entrepreneurial argument, she might have a better

chance of securing their votes. On the other hand, the respondent does not have these same

advantages and therefore should not benefit from using an entrepreneurial argument. He

already won at the lower court, which would suggest the prevailing understanding of the

Court’s past rulings works in his desired direction. Moreover, because the respondent files

his brief after seeing the petitioner’s brief, he runs the risk of looking desperate (and perhaps

not credible) if he tries to respond using an entrepreneurial argument.

Additionally, engaging in legal entrepreneurship should help attorneys representing the

petitioner who are inexperienced or working for resource-poor clients, but not offer the same

boost to experienced attorneys. An attorney’s experience, his position as an attorney in the

Office of the Solicitor General, and a party’s economic status offer built-in advantages before

the Supreme Court (Black and Boyd 2012; Black and Owens 2012a; McGuire 1995; Nelson

and Epstein 2019). These traits signal to the justices that an expensive (and by proxy,

good) attorney is about to present an argument the justices know they can blindly trust to

39

be correct (Biskupic, Roberts and Shiffman 2014; Garner and Roberts 2010). These briefs

are easy to consume and they consequently make the justices’ jobs easier. When an unknown

attorney appears before the Court, however, she is going to make the justices’ jobs harder

no matter what she writes. The justices do not know her, they cannot trust her, and they

consequently have to study her argument with care. Given the justices’ focus, inexperienced

and resource-poor attorneys have a real opportunity to land an entrepreneurial argument in

this case; if they can construct the argument properly, the justices are listening.

3.3 Data and Measures

To better understand the value of employing prevailing and entrepreneurial arguments, I

create a new measure of legal entrepreneurship and use it alongside Black, Hall, Owens and

Ringsmuth’s (2016) data on merits briefs and judicial decision making. The data encompass

13,387 justice-votes cast between the 1984 and 2007 terms, covering 1,509 cases in which the
Court received only one brief from the petitioner and one brief from the respondent.1 My

dependent variable is the justice’s vote, specifically whether she voted for (1) or against (0)
1An astute observer will notice the Supreme Court reviewed almost 2,400 cases during this
time period, so I am only looking at 63% of the cases the Court reviewed. Cases are missing
for one of four different reasons: (1) the briefs were not available in LexisNexis or Westlaw; (2)
the cases had more than one brief for each party; (3) other datasets did not have observations
for these cases; or (4) the case was an original jurisdiction case. Beginning with the first,
both Westlaw and LexisNexis were missing briefs for about 5% of the Court’s cases during
this time period. These missing briefs were mostly (but not exclusively) from cases the Court
reviewed in the early 1980s. Per a conversation with Westlaw’s product management team,
the missing briefs were not converted to a digital format and were therefore never put into
their online repository. These briefs are simply missing and will never go into the dataset.
Regarding the second category of missing data, approximately 15% of cases are missing
because the parties submitted more than one brief for each side (e.g., two petitioner briefs
for one case). I purposefully omitted these cases in order to deal with potential endogeneity
problems associated with the justices receiving multiple arguments from the same side in a
case. About 10% of cases are missing because they were missing from the datasets I used for
control variables. The remaining cases were original jurisdiction cases, or cases in which the
Supreme Court acts as the trial court. Given the unique nature of these cases, I removed
them from the analysis.

40

the petitioner.

3.3.1

Identifying Entrepreneurship

The decision to use a prevailing argument or engage in legal entrepreneurship is the key

factor under analysis here, so I operationalize it using two dichotomous variables: one for

the petitioning attorney’s decision to engage in legal entrepreneurship (1) rather than use

the prevailing argument (0), and one for the responding attorney’s decision to do the same.

To identify these attempts to alter the legal status quo, I essentially need to replicate the

analysis I conducted on Jan Walls Anderson’s brief in California v. Acevedo: break each

brief down to its citations and the attorney’s discussion of them, and then compare each

brief’s arguments with the Court’s own approach to that particular area of the law. Doing

this by hand for the 1,509 cases under examination here would be inadvisable, however;

data suggests it would take an experienced coder more than two years’ worth of 40-hour

work weeks to read the briefs, break them down into citations, and identify the attorney’s

application of each citation (Schoenherr and Black 2019b).

I consequently automate the

process using a combination of human-assisted and machine learning techniques.

The automated analysis unfolds over four steps, which I outline here and discuss in more

detail below. In the first step, I take text files of the “Arguments” sections of the merits

briefs and use a computer program to identify each mention of a citation in the documents.

I then create and utilize a dictionary that identifies attorneys’ application of those citations

within the briefs, creating data I then use to identify each brief’s central arguments. Finally,

I use Hansford and Spriggs’s (2006) precedent vitality data to compare the briefs’ key legal

arguments with the Court’s and identify instances of legal entrepreneurship.

For the first step of the process, I wrote a computer program that utilizes R’s tidyverse

suite to turn text documents into sentence-level citation data whose content can be analyzed

using mechanized processes. Or, more simply, I extract the core pieces of an attorney’s legal

41

argument for later analysis. The program identifies the “Arguments” section of each brief,

breaks the section into sentences, and searches for mentions of a Supreme Court opinion
within each sentence.2 When the program finds a citation, it saves the citation as well as the

sentence that precedes it, which I use in the next step to identify the attorney’s application

of that precedent. At the end of this step, I have a list of every readily-identifiable mention
of a citation as well as the sentence from which it came.3 At this point, there are 187,764

mentions of just over 10,000 Supreme Court precedents across the 3,018 briefs under analysis,

an average of 62 mentions of 26 precedents per brief.

In the second step, I create a dictionary of terms associated with attorneys’ application

of citations and then use that dictionary to actually identify the attorneys’ treatment of

each citation in their briefs (applicable, not applicable, or a neutral statement of fact).

These applications map directly to the positive (applicable), negative (not applicable) and

neutral treatments identified by Shepard’s Citations (Hansford and Spriggs 2006; Spriggs

and Hansford 2000); I modify my language here for simple ease of explanation. My decision

to use a dictionary-based approach departs from the current trend of using more high-

powered, black-box models for sentiment classification (Grimmer and Stewart 2013), but

I argue the dictionary-based approach is more useful in this context. Both approaches

offer accuracy and efficiency at some cost; dictionaries require constant validation and are

context-dependent, while using a more complicated modeling technique like a random forest
2Because Supreme Court justices are only bound by Supreme Court precedent, I do not
use state court decisions or lower federal court decisions in my analysis. I also eliminate
references to the federal code (e.g., 18 U.S. Code §1657).

3By convention, attorneys typically cite a Supreme Court case using its name followed
by the U.S. Report citation and the year the justices published the opinion, e.g., Terry v.
Ohio, 392 U.S. 1 (1968).
Inevitably, however, attorneys turn to shorthand to save space
in their briefs, and the formal citation eventually gets reduced to Terry v. Ohio or 392
U.S. 1 and, at some point, all the way down to merely Terry. The program identifies full
citations and short hand and then uses the quanteda package (Benoit et al. 2017) and its
word frequency statistics to identify and include these less-formal citations as well. When
compared to hand-coded data, the program identifies more than 90% of the total number of
citations mentioned in a brief.

42

or support vector machines requires training data and offers less concrete information about

why the classification scheme is accurate (Corley and Wedeking 2014; Kuhn and Johnson

2016; Rice and Zorn 2019). Here, the dictionary-based approach offers information about

the words associated with application while simultaneously being created and validated to

identify sentiment in this specific context, making it an ideal tool for analysis and later use.

Logistically, I follow a process similar to that of LexisNexis’s Shepard’s Citations, which

identifies current opinions’ treatment of past precedent (Hansford and Spriggs 2006). Em-

ploying information from three different sources – LexisNexis’s list of shepardization terms

and treatments, Spriggs and Hansford’s (2000) overview of the shepardization process, and

hand-coded sentiment data gathered from 57 randomly-selected search and seizure cases and

10 privacy cases – I developed a list of 525 words and phrases that uniquely identify attor-

neys’ decisions to classify a precedent as relevant to the current case (“apply,” “establish,”

“mandate”) or dismiss a precedent as irrelevant (“distinguish,” “nullify,” “overrule”). All other

citations are considered neutral statements of fact. With the help of the quanteda package

(Benoit et al. 2017), I apply the dictionary to the sentence associated with a citation to

identify the attorney’s singular treatment of it, then aggregate that information to create

an overall measure of precedent treatment. At this point, I have a list of every case cited

in a brief as well as the attorneys’ overall treatment of that case, a list of almost 80,000

brief-citation pairs. The data now show, for example, that the attorney representing the

petitioner in Planned Parenthood v. Casey (1992) treated Roe v. Wade (1973) as relevant,

applicable precedent to the situation at hand, while the responding attorney suggested Roe

did not apply and should, in fact, be overturned.

After aggregating the data, I identify the parts of the argument most central to each brief

and restrict my analysis to those precedents. The average merits brief cites approximately

26 different Supreme Court precedents. But the distribution of those citations suggests

most of these cases receive a single, obligatory mention and only a few central precedents

43

receive repeated attention. According to the justices, their attention goes straight to those

central precedents (Garner and Stevens 2010; Garner and Thomas 2010), and legal scholars

encourage brief writers to focus their attention on these cases as much as possible (Garner

2003). I consequently restrict my analysis to these well-cited precedents, removing all cases

that get mentioned a below-average number of times. This resulted in looking at about 6
cases per brief rather than 26 while eliminating a substantial amount of noise from data.4

At this point, I have data on the key arguments in each of the 3,018 briefs under analysis

here.

In the fourth and final step, I examine the central precedents in order to identify en-

trepreneurial arguments in the petitioner and respondent briefs. To do this, I search the

arguments for instances of an attorney engaging in any one of three entrepreneurial acts: (1)

suggesting a reversal of the Court’s prevailing approach (and, by implication, replacing it

with something else); (2) applying an older, forgotten line of jurisprudence; or (3) building

the case around a precedent the Court has long since ignored.

Operationally, an attorney suggests a reversal of the Court’s current approach to the

law when she alone of the two attorneys in a case argues against the Court’s current treat-

ment of a past opinion. To identify instances of this occurring, I searched for situations

where attorneys cited the same cases but applied them differently – situations where one

attorney said the case applied and the other said the case did not. Every time one of the

attorneys argued against the Court’s prevailing approach (that is, whether the Court’s opin-

ions indicated the justices favored or disfavored the precedent), that attorney engaged in

legal entrepreneurship. To identify the Court’s prevailing approach, I used Hansford and

Spriggs’s (2006) measure of precedent vitality. Precedent vitality is essentially a running
4Looking at all the cases mentioned in each brief ultimately resulted in the suggestion
that every brief contained an entrepreneurial argument. Why? Because there was so little
consistency in approach to these one-off citations that keeping them in the analysis made
each brief’s argument appear more unique than it really was. By focusing on the important
cases, I am making an actual apples-to-apples comparison of only the crucial cases.

44

tally of the Court’s treatment of a case over time, providing term-by-term data on how the

Court applies its own past decisions. Positive precedent vitality scores indicate the Court

currently favors a precedent, while negative precedent vitality scores suggest the Court has

distinguished or even overturned the precedent in question. So, returning to the earlier ex-

ample from Planned Parenthood v. Casey (1992), the petitioning attorney suggested Roe

v. Wade (1973) was applicable and the responding attorney argued the opposite. At the

time the Court heard Casey, Roe had a precedent vitality score of +164. In this case, then,

the petitioner would align with the Court’s view, while the respondent would be making an

entrepreneurial argument.

Attorneys can also be entrepreneurial by bringing up different lines of jurisprudence or

by building their case around precedents that the Court has never discussed again. In the

first case, these entrepreneurial acts would manifest in attorneys citing older lines of case

law, appealing to the justices using arguments that are valid but unused. To identify these

instances of entrepreneurship, I again use the precedent vitality data, which also provide

information on the Court’s most recent use of a past precedent. I consider an attorney to

be engaging in legal entrepreneurship if she mentions a case the Court has not used in an

opinion in the last ten years.

In the second case, I identify instances in which attorneys

engage in legal entrepreneurship by centering their arguments around cases so obscure that

even the justices have never cited the opinions again. I again use the precedent vitality data

to identify these cases.

With the acts identified, I count an attorney as engaging in legal entrepreneurship in a

brief if she argued against the Court’s prevailing approach to a case, brought up an old area of

case law, built a case around a never-used precedent, or used some combination of these three

tactics. Any attorney who did not use these tactics used a prevailing argument instead. The

results of this analysis suggest that attorneys regularly engage in legal entrepreneurship,

making the decision to go for the entrepreneurial argument about 52% of the time. The

45

attorney representing the petitioner does so 51% of the time, while the attorney representing

the respondent goes entrepreneurial at the slightly higher rate of 53%.

3.3.2 Additional Variables

Given the justices’ well-documented preference for arguments made by experienced at-

torneys who either work for the government or have significant resources backing their work,

I include six different variables in my model to study the relationship between attorney sta-

tus, the decision to engage in legal entrepreneurship, and the justices’ votes. These variables

come from Black, Hall, Owens and Ringsmuth’s (2016) analysis of merits briefs.

Because experienced attorneys are more likely to win the justices’ votes (McGuire 1995;

Nelson and Epstein 2019), the first two variables I include are the petitioner and respondent’s

previous experience at oral argument. To create this variable, Black, Hall, Owens and

Ringsmuth (2016) take the natural logarithm of each attorney’s previous oral argument

experience, more specifically, ln(previous experience + 1). Additionally, given the Solicitor

General’s disproportionately large win rate before the justices (Black and Owens 2012c;

Wohlfarth 2009), I employ two dichotomous variables to control for his presence in a case as

the petitioner or respondent. Finally, I control for each party’s status, as parties with more

resources have advantages that less-wealthy parties do not (Black and Boyd 2013). Following

the procedure originally outlined by Collins (2004, 2007), Black, Hall, Owens and Ringsmuth

(2016) use the Supreme Court Database’s party codes to categorize each party into one of

10 groups, with the weakest parties – poor individuals – coded as 1 and the strongest party

– the U.S. government – coded as 10.

I also interact each of these six variables with their respective entrepreneurship variables

(e.g., petitioning attorney experience x petitioner decision to engage in legal entrepreneur-

ship) in order to study the conditional relationship between entrepreneurship and resources.

I additionally include 11 different variables in the model to control for factors that are

46

known to influence a justice’s decision to side with the petitioner. First, I include a variable

that identifies cases in which the Supreme Court’s opinion noted a lower-court dissent, which

comes from the Supreme Court Database (Spaeth et al. 2017). This variable is a proxy for

case quality, as Black, Hall, Owens and Ringsmuth (2016) suggest that the justices’ decision

to note a dissent in their majority opinion could indicate the petitioner had a particularly

strong legal argument. In the same vein, I include a dichotomous indicator of the presence

of lower court conflict, as its existence can alter the justices’ decision-making process (Perry
1991).5

Additionally, because the justices might modify their behavior when dealing with a more

salient case (Lax and Cameron 2007), I control for case salience. To do this, I employ Clark,

Lax and Rice’s (2015) measure of latent case salience. The higher the value, the higher

the salience. Following Black, Hall, Owens and Ringsmuth (2016), I also control for the

ideological congruence between the justices and the direction of the lower court’s decision. If

the lower court decision was liberal, then the ideological congruence variable takes the value

of the justice’s Segal-Cover score, and if the lower court decision was conservative, then the

ideological congruence variable is 0.

Next, I control for the readability of the petitioner and respondent briefs, as the justices

are more likely to side with attorneys who submit readable briefs (Feldman 2016). Black,

Hall, Owens and Ringsmuth (2016) measure readability using the Coleman-Liau Index, which

uses word and sentence length to calculate complexity. The higher the value of the Coleman-

Liau Index, the less readable the brief, and the less likely the justices are to side with that

party.

As Collins (2008) points out, parties can also gain an advantage through amicus par-
5The Supreme Court Database certReason variable identifies the reason the Supreme
Court gave for reviewing a case. The variable has thirteen categories, five of which deal with
conflict in the lower courts (categories 2-6). If the Court listed reasons two through six as
their reason for reviewing the case, I coded this variable as 1. Otherwise, it took the value
of zero.

47

ticipation in their case, so I include a count of the amicus briefs submitted in support of

the petitioner and the respondent. I also control for the Solicitor General’s presence as an

amicus favoring the petitioner or the respondent. Finally, I also include control variables for

the number of questions asked of the petitioner and respondent during oral argument, as the

side that receives more questions is more likely to lose (Johnson et al. 2009).

3.4 Methodology and Empirical Results

To reiterate, I expect to find that the justices are more likely to side with the petitioner

when the petitioning attorney engages in legal entrepreneurship. Additionally, I expect

the justices should be more likely to vote in favor of an inexperienced petitioning attorney

or an attorney representing a low-status petitioner when that attorney engages in legal

entrepreneurship while expecting that experienced and resource-rich petitioning attorneys

gain no advantage from doing to the same. I do not expect the justices to alter their voting

behaviors at all in response to a respondent engaging in legal entrepreneurship.

Because my dependent variable is dichotomous, I use a logistic regression model for the

analysis (Long 1997) and, following Black, Hall, Owens and Ringsmuth (2016), I estimate

the model using standard errors that are clustered by justice. Due to the non-linear nature

of the model, which makes interpretation of the coefficients difficult, I used predicted values

to address the results, which I calculate using the observed-value approach (Hanmer and

Kalkan 2013). Additionally, for ease of interpretation, I reversed the axis on all graphs

involving the respondent so that the graphs show the probability the justice votes with the

respondent, rather than the petitioner.

The results of the logistic regression model of the likelihood a justice votes with the

petitioner in a case are shown in Table 3.1. My analysis begins with Figure 3.1, which

addresses the probability that a justice sides with the petitioner based on her decision to use

a prevailing or entrepreneurial argument.

48

Table 3.1: Logistic Regression Results, Justice Votes in Favor of the Petitioner

Entrepreneurial Petitioner

Entrepreneurial Respondent

Petitioner Experience

Respondent Experience

OSG Petitioner

OSG Respondent

Petitioner Status

Respondent Status

Entrepreneurial Petitioner x Petitioner Experience

Entrepreneurial Petitioner x Petitioner OSG

Entrepreneurial Petitioner x Petitioner Status

Entrepreneurial Respondent x Respondent Experience

Entrepreneurial Respondent x Respondent OSG

Entrepreneurial Respondent x Respondent Status

Dissent Noted in Lower Court

Lower Court Conflict

Ideological Congruence

Latent Case Salience

Petitioner Readability

Respondent Readability

OSG Amicus for Petitioner

OSG Amicus for Respondent

Petitioner Amicus Support

Respondent Amicus Support

Questions for Petitioner

Questions for Respondent

Constant

Observations
AIC
BIC
Log Likelihood
Standard errors clustered by justice
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001
49

(Standard Errors)

Coefficient
0.370∗∗∗
(0.098)
0.190∗
(0.096)
0.123∗∗∗
(0.028)
−0.049
(0.030)
0.096
(0.111)
−0.155
(0.106)
0.047∗∗∗
(0.014)
−0.030∗
(0.013)
−0.102∗∗
(0.037)
0.118
(0.146)
−0.033∗
(0.016)
0.014
(0.040)
−0.206
(0.145)
−0.021
(0.016)
0.131∗∗
(0.043)
−0.297∗∗∗
(0.041)
−1.392∗∗
(0.083)
−0.057∗
(0.028)
−0.029∗
(0.013)
−0.011
(0.013)
0.757∗∗∗
(0.057)
−0.809∗∗∗
(0.065)
0.050∗∗∗
(0.008)
−0.044∗∗∗
(0.008)
−0.019∗∗∗
(0.001)
0.019∗∗∗
(0.001)
1.423∗∗∗
(0.340)

13, 387
16556.2
16758.76
−8251.1

Starting with the petitioning attorney’s decision to engage in legal entrepreneurship in

Figure 3.1, I find that Supreme Court justices are significantly more likely to side with

the petitioner when the petitioning attorney uses an entrepreneurial argument. When the

attorney representing the petitioner uses a prevailing argument, there is a 0.57 probability

the justice sides with the petitioner. This probability increases slightly but significantly to

0.59 if the petitioner engages in legal entrepreneurship. As I expected, going entrepreneurial

can benefit the petitioner.

Figure 3.1: Probability a Supreme Court Justice Votes with the Petitioner By Argument Type
– Left side shows the probability a justice votes with the petitioner when the petitioner uses a prevailing
argument, right side shows the probability a justice votes with the petitioner when the petitioner uses an
entrepreneurial argument. Vertical lines identify 95% confidence intervals. Predicted probabilities calculated
using the observed-value approach.

The respondent does not gain the same benefit. As the results in Figure 3.2 show, an

attorney who uses a prevailing argument to represent the respondent has a 0.43 probability of

securing the justice’s vote. If that same attorney uses an entrepreneurial argument instead,

the probability decreases slightly, to 0.42, but the difference is not statistically significant.

Just as I suggested, engaging in legal entrepreneurship neither helps nor hurts the attorney

50

0.540.550.560.570.580.590.600.61PrevailingArgumentEntrepreneurialArgumentProbability the Justice Votes with the Petitionerrepresenting the respondent in a case.

Figure 3.2: Probability a Supreme Court Justice Votes with the Respondent By Argument Type
– Left side shows the probability a justice votes with the respondent when the respondent uses a prevailing
argument, right side shows the probability a justice votes with the respondent when the respondent uses an
entrepreneurial argument. Vertical lines identify 95% confidence intervals. Predicted probabilities calculated
using the observed-value approach.

While the baseline figures suggest a petitioning attorney benefits from an entrepreneurial

argument and the responding attorney does not, the results suggest an attorney’s experience

before the Court complicates this relationship. Turning first to the left panel of Figure 3.3,

the results suggest that a justice is more likely to side with an experienced petitioning at-

torney when that attorney uses a prevailing argument. An inexperienced attorney with no

prior Supreme Court experience has a 0.55 probability of securing a justice’s vote when she

uses a prevailing argument, while a veteran Supreme Court advocate with 29 past appear-

ances under his belt (natural log value of 3.4) has a 0.63 probability of securing a justice’s

vote. This significant eight-percentage-point increase suggests that experienced petitioning

attorneys have an advantage before the justices when using a prevailing argument.

51

0.400.410.420.430.440.45PrevailingArgumentEntrepreneurialArgumentProbability the Justice Votes with the RespondentFigure 3.3: Probability a Supreme Court Justice Votes with the Petitioner Based on Argument
Type and Attorney Experience – Probability a Supreme Court justices votes with the petitioner based on
attorney experience and the decision to use a prevailing (left) or entrepreneurial (right) argument. Dashed
lines are 95% confidence intervals around those estimates. Predicted probabilities calculated using the
observed-value approach.

The right panel of Figure 3.3 shows that, as I expected, an inexperienced attorney who

makes an entrepreneurial argument is as likely as an experienced attorney to gain a justice’s

vote. An attorney representing the petitioner who has no prior experience before the Court

has a 0.59 probability of winning a justice’s vote after using an entrepreneurial argument,

while the experienced attorney has a 0.60 probability of doing the same. This difference is

not statistically significant. Importantly, a justice is significantly more likely to side with an

inexperienced attorney when she makes an entrepreneurial argument, while an inexperienced
attorney gains no significant advantage.6 In short, while Figure 3.1 suggests the petitioner

might benefit from engaging in legal entrepreneurship, the results in Figure 3.3 suggest

that inexperienced attorneys are the ones who truly benefit from using a radical, innovative

argument.

6See Figure A.1 in the appendix.

52

0.500.510.520.530.540.550.560.570.580.590.600.610.620.630.640.650.660.670.680.690.700.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.23.4Attorney Experience at Oral Argument (log)Probability the Justice Sides with the PetitionerPrevailing Argument0.500.510.520.530.540.550.560.570.580.590.600.610.620.630.640.650.660.670.680.690.700.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.23.4Attorney Experience at Oral Argument (log)Probability the Justice Sides with the PetitionerEntrepreneurial ArgumentInterestingly, the attorneys representing the respondent do not have an experiential ad-

vantage before the justices, regardless of their decision to use a prevailing or entrepreneurial

argument. As the left side of Figure 3.4 shows, an inexperienced attorney representing the

respondent who has never appeared before the Supreme Court has a 0.43 probability of

winning a justice’s vote when using a prevailing argument, while the experienced attorney

in the same situation has a 0.46 probability of winning a justice’s vote. The difference is not

statistically significant. The right side of Figure 3.4 tells the same story: an inexperienced

attorney representing the respondent who engages in legal entrepreneurship has a 0.42 prob-

ability of winning a justice’s vote, while an experienced attorney has a 0.44 probability of
doing the same. Again, the difference is not statistically significant.7 Experience offers no

advantage when making the responding argument, regardless of whether the attorney makes

a prevailing or entrepreneurial argument.

7See Figure A.2 in the appendix.

53

Figure 3.4: Probability a Supreme Court Justice Votes with the Respondent Based on Argument
Type and Attorney Experience – Probability a Supreme Court justices votes with the respondent
based on attorney experience and the decision to use a prevailing (left) or entrepreneurial (right) argument.
Dashed lines are 95% confidence intervals around those estimates. Predicted probabilities calculated using
the observed-value approach.

Turning next to attorney status, the results suggest that engaging in legal entrepreneur-

ship can help the economically-disadvantaged client as well. The left panel of Figure 3.5

shows that the justices are significantly more likely to vote in favor of high-status petition-

ers like the United States government when their attorneys use a prevailing argument. An

attorney using a prevailing argument to represent a low-status petitioner has a 0.52 proba-

bility of securing a justice’s vote, while an attorney using a prevailing argument to represent

a high-status petitioner has a 0.61 probability of doing the same. The right panel of Fig-

ure 3.5 shows that inexperienced attorneys receive major benefits from engaging in legal

entrepreneurship; an attorney using an entrepreneurial argument to represent a low-status

petitioner has a 0.57 probability of winning a justice’s vote, while an attorney doing the

same in representation of a high-status petitioner has a 0.60 probability of winning a jus-

tice’s vote. The difference between the low-status petitioner and the high-status petitioner is

54

0.400.410.420.430.440.450.460.470.480.490.500.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.2Attorney Experience at Oral Argument (log)Probability the Justice Sides with the RespondentPrevailing Argument0.400.410.420.430.440.450.460.470.480.490.500.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.2Attorney Experience at Oral Argument (log)Probability the Justice Sides with the RespondentEntrepreneurial Argumentnot statistically significant.8 The results do suggest the justices are significantly more likely

to vote in favor of a resource-poor litigant who uses an entrepreneurial argument than they

are to vote in favor of a similar litigant who uses a prevailing one. The same relationship

does not hold for a petitioning attorney representing a resource-rich litigant, who is no more

or less likely to win by going entrepreneurial.

Figure 3.5: Probability a Supreme Court Justice Votes with the Petitioner Based on Argument
Type and Attorney Status – Probability a Supreme Court justice votes with the petitioner based an
attorney status and the decision to use a prevailing (left) or entrepreneurial (right) argument. Dashed lines
are 95% confidence intervals around those estimates. Predicted probabilities calculated using the observed-
value approach.

Attorneys representing a resource-rich respondent maintain their advantage, however. As

both panels of Figure 3.6 show, the justices are more likely to vote with a well-resourced

respondent, regardless of whether the attorney uses a prevailing or entrepreneurial argument.

The results also suggest engaging in entrepreneurship never gives the respondent any added

advantage – he is as likely to win using a prevailing or entrepreneurial argument. A resource-

poor respondent has a 0.40 probability of securing a justice’s vote when he uses a prevailing

8See Figure A.3 in the appendix.

55

0.480.490.500.510.520.530.540.550.560.570.580.590.600.610.620.630.6412345678910Petitioner StatusProbability the Justice Votes with the PetitionerPrevailing Argument0.480.490.500.510.520.530.540.550.560.570.580.590.600.610.620.630.6412345678910Petitioner StatusProbability the Justice Votes with the PetitionerEntrepreneurial Argumentargument. This decreases slightly to 0.37 if he uses an entrepreneurial argument, but the

difference is not statistically significant. This holds for resource-rich respondents as well; an

attorney using a prevailing argument to represent a wealthy respondent has a 0.46 probability

of securing a justice’s vote, and that probability increases slightly, but not significantly, to
0.47 when the attorney uses an entrepreneurial argument.9 The results continue to show

that entrepreneurship is a useful strategy for petitioners but not respondents.

Figure 3.6: Probability a Supreme Court Justice Votes with the Respondent Based on Argument
Type and Attorney Status – Probability a Supreme Court justice votes with the respondent based an
attorney status and the decision to use a prevailing (left) or entrepreneurial (right) argument. Dashed lines
are 95% confidence intervals around those estimates. Predicted probabilities calculated using the observed-
value approach.

Turning away from the key independent variables and toward the controls, the results

suggest the justices are significantly more likely to side with the petitioner when they notice

a lower court dissent in the case and when there is lower court conflict. The justices are also

significantly less likely to side with the petitioner when a case is ideologically aligned with

them or when the case is salient. Additionally, the readability of the petitioner’s brief can

9See Figure A.4 in the appendix.

56

0.340.360.380.400.420.440.460.480.5012345678910Respondent StatusProbability the Justice Votes with the RespondentPrevailing Argument0.340.360.380.400.420.440.460.480.5012345678910Respondent StatusProbability the Justice Votes with the RespondentEntrepreneurial Argumentsignificantly influence the outcome in a case, though the readability of the respondent’s case

cannot.

Notably, while the Solicitor General’s presence as a party to the case does not significantly

influence the case outcome, filing an amicus brief in favor of the petitioner or respondent

does. Overall amicus support is also crucial for success for any party, and the more support

they have, the more likely they are to receive a justice’s vote. Finally, the number of questions

the attorneys receive during oral argument are statistically-significant factors in a justice’s

decision to side with the petitioner.

3.5 Discussion

Near the end of the Supreme Court’s 2018 term, the National Law Review interviewed

former Solicitor General Paul Clement about his strategy for asking the justices to overturn

precedent (Mauro 2019). Precedent vitality was a hot topic on the Court at the time: a five-

justice majority had just overturned a fifty-year-old precedent with its decision in Franchise

Tax Board of California v. Hyatt (2019), and Justice Stephen Breyer responded with a

blistering dissent that accused the majority of devaluing precedent and inviting litigants

to freely challenge long-standing decisions (Mauro 2019). Clement, an elite attorney who

had presented more than 90 cases to the justices (Kirkland and Ellis 2019), offered less-

experienced attorneys some advice: overturning precedent is a long and slow process in

which the justices “chip away at cases in various steps so that the day the case is actually

overruled it’s really not even news” (Mauro 2019). He suggested that trying to overturn

a problematic precedent in one fell swoop is probably not a successful strategy before the

current Court.

In this paper, I set out to determine whether making entrepreneurial arguments of any

kind is a useful strategy for attorneys. Using a new data processing technique, I developed

a method for identifying legal entrepreneurship and used it to see how an attempt to change

an area of case law influences the justices’ decisions regarding a case. My results suggest

57

that entrepreneurship can be a useful strategy for some attorneys, particularly those who

are petitioning the Court despite being inexperienced or representing resource-poor clients.

For experienced attorneys, attorneys representing wealthy clients, or attorneys representing

the respondent, engaging in legal entrepreneurship offers no added advantage. The results

suggest that for most attorneys, both argument types offer value, but neither one offers

an absolute advantage like the one provided to inexperienced and resource-poor petitioners

using entrepreneurial arguments. Clement suggested that slow innovation might be the

right answer, but for new attorneys who are making the first move in a case on behalf of

the petitioner, trying to change the world with their first crack might be exactly the right

strategy.

There are, of course, limitations to this paper’s approach. Because of data availability

limitations, my analysis ends at the 2007 term. I am currently working on collecting data

through the 2018 term and plan on updating this data at the end of every term. With this

data, I can gain more insight into entrepreneurship’s benefits and drawbacks, especially on

the Roberts Court, where wealthy litigants and experienced attorneys control much of the

docket (Biskupic, Roberts and Shiffman 2014). Adding more data to the analysis is crucial

for gaining a complete picture of what strategies are successful at the Supreme Court over a

long period of time. Having the data collection tools already developed and in place should

make adding data relatively easy.

Additionally, supervised machine learning is not a perfect process. It requires constant

monitoring, validation, and correction in order to function properly (Grimmer and Stewart

2013; Schoenherr and Black 2019b), and even then, it can still produce errors that human

coders would not make. Consider, for example, that an attorney’s typo regarding a U.S.

Report volume number could cause my program to identify a case as Californians v. Cal-

ifornia, 393 U.S. 1 (1968), rather than Terry v. Ohio, 392 U.S. 1 (1966). Continuing to

employ human-assisted coding alongside the automated process is ideal for ensuring the

58

results are valid and close to what human coders could get.

Moving forward, someone could use a more complicated measure of precedent treatment

to tease out innovation more fully.

I use a collapsed relevant-not relevant-neutral scale

(more commonly known as positive-negative-neutral in Shepard’s Citations parlance), but

LexisNexis classifies treatment using a six-point scale (warning, questions, caution, positive,

neutral, cited by). Introducing more complexity into the data collection process and, conse-

quently, the identification of entrepreneurial arguments could produce a different look into

innovation than the one presented here.

Another possible avenue for future research would involve an over-time analysis of attor-

neys’ decisions to innovate. Was entrepreneurship more common when during the Warren

Court, when the justices broke down standing rules to allow more people access? Or per-

haps it was more common during the Burger Court, when the Nixon four offered conservative

groups the chance to overturn hated precedents like Miranda? Are attorneys more likely to

engage in legal entrepreneurship when the Court gets a new member? These are all questions

that can be answered using the approach I outlined here.

59

CHAPTER 4

PUSHING THE BOUNDARIES: DECIDING TO ENGAGE IN LEGAL

ENTREPRENEURSHIP AT THE U.S. SUPREME COURT

In late 2002, attorneys representing two men accused of violating the Texas Homosexual

Conduct Law were preparing to challenge their clients’ convictions (and sodomy laws more

broadly) at the Supreme Court. They just had to figure out how to appeal to the justices.

The approach had to be just right; the Court had refused to overturn these types of sodomy

laws less than 20 years earlier in Bowers v. Hardwick (1986) and membership changes made

the Court even more conservative than it had been when it decided Bowers (Frank 2017). But

at the same time, the world was changing. Courts in several states had issued pro-gay rights

decisions, as had the Supreme Court itself in Romer v. Evans (1996) (Kaplan 2018). After

examining the state of the world, the attorneys opted to make what they later called “the

most conservative argument possible for a constitutional right to sex” (Carpenter 2012, 194).

Their merits brief presented an argument that carefully aligned with the Court’s more recent

jurisprudence regarding due process and privacy, praising Justice Anthony Kennedy’s opinion

in Romer and following the Court’s own lead in suggesting it was time to set Bowers aside

(Carpenter 2012). They did this while avoiding the more radical sex and sexual orientation

discrimination arguments that were popular at the time (Carpenter 2012). In a 6-3 ruling in

Lawrence v. Texas (2003), the Supreme Court ultimately validated these choices by vacating

the men’s convictions and declaring the Texas Homosexual Conduct Law unconstitutional;

a five-justice majority also voted to overturn Bowers v. Hardwick as well (Carpenter 2012).

When attorneys like the ones representing Tyron Garner and John Geddes Lawrence

present their cases to the Supreme Court in a merits brief, they make a thousand different

decisions about how best to appeal to the justices. They make choices about phrasing

(Garner and Kennedy 2010; Feldman 2016), about tone (Black, Hall, Owens and Ringsmuth

60

2016), and about ease of presentation (Garner and Roberts 2010). Beyond the more technical

aspects of a brief, attorneys also have decide how to lead the justices from the legal question

at hand to their proposed answer by way of the Court’s own precedents (Garner 2003). This

could mean providing the justices with a legal explanation for the ideological result they

seek (Hansford and Spriggs 2006), or it could mean presenting a legal argument so well

constructed the justices feel like it is the only possible answer (Garner and Roberts 2010).

But the key to writing a good brief is writing a persuasive legal argument, and attorneys

approach this task using one of two opposing frames. Their first option is going with what

I call the prevailing argument and aligning their argument as closely as possible with the

Court’s own words and thoughts, similar to what the petitioning attorneys did with their

brief in Lawrence. The alternative is making what I call an entrepreneurial argument and

offering the justices a fundamentally new way of looking at an area of case law, like those

attorneys could have done with the sex discrimination argument. Why do attorneys pick

one over the other?

My research suggests that attorneys are strategic in their employment of legal entrepreneur-

ship. Prevailing arguments are, all else equal, safer arguments to make because they appeal to

the justices’ preferences for simplified decision-making (Gennaioli and Shleifer 2007; Niblett,

Posner and Shleifer 2010). But attorneys also want to win, which can mean getting en-

trepreneurial when the prevailing approach to the law fails to aid their clients. Using a novel

dataset that identifies attorneys’ use of legal entrepreneurship between the 1984 and 2007

Supreme Court terms, I find that attorneys let the legal environment surrounding the case,

as well as their own skill, direct their decision to engage in legal entrepreneurship. External

political opportunities, including new justices on the bench and the public’s overall feelings

toward certain policies, do not play a role in this decision. I show, in short, that attorneys

are strategic in how they use the law to appeal to the justices in their merits briefs.

In reaching this conclusion, I make two main contributions to the literature on Supreme

61

Court decision making. The first is offering one of the first examinations of attorney strat-

egy. Research shows that an attorney’s decisions regarding wording (Corley 2008), framing

(Wedeking 2010), and case citations (Hazelton, Hinkle and Spriggs 2019) in merits briefs in-

fluence the justices’ behavior. Yet scholars understand very little about why attorneys make

the decisions they do; research begins with attorneys’ product, not with attorneys’ initial

forays with the judicial decision-making process. My suggestion that attorneys strategically

engage in legal entrepreneurship is a logical conjecture that is well-backed by anecdotes (see,

for example, DeHart 2018 on the Women’s Rights Project or Kaplan 2018 on campaign fi-

nance), if not academic research. By introducing the concept of legal entrepreneurship and

examining attorneys’ decision to engage in it, I am one of the first to empirically show that

attorneys are strategic decision makers, just like the justices to whom they appeal.

By discussing attorneys’ role in shaping the justices’ understanding of the law, I am also

expanding scholars’ understanding of the factors that constrain the justices’ ability to pursue

policy outcomes. The research that builds on Epstein and Knight’s (1998) near-rule that

Supreme Court justices are strategic seekers of policy suggests the law can act as a constraint

on judicial behavior (Bailey and Maltzman 2011; Bartels 2009; Richards and Kritzer 2002).

Yet all too often, work on the law’s constraining capacity implicitly assumes the justices

independently decide the legal rationale for a case (see, for example, Hansford and Spriggs

2006 or Clark and Lauderdale 2010). In reality, by writing the briefs, the attorneys actually

set boundaries around the case and direct the justices toward certain legal arguments in the

process (Black, Hall, Owens and Ringsmuth 2016; Epstein, Segal and Johnson 1996, but

see McGuire and Palmer 1995). When working through an opinion, the justices’ decisions

are even more constrained than most research would suggest. My research incorporates

attorneys’ decisions into scholars’ understanding of the choices the justices make, adding

another layer of complexity to an already convoluted process.

62

4.1 Appealing to the Justices in the Merits Brief

An attorney’s job is to win the case. Whether motivated by payment, fame, ideological

idealism, or just a desire to see a case through, her goal is to get the Court to side with

her client. At the Supreme Court, attorneys have two opportunities to engage with the

justices and try to win them over: first, in the written merits brief, and then again during

oral argument. Merits briefs are informational documents in which each party provides the

justices with an overview of the case’s background and procedural history, as well as the

attorney’s proposed solution and the legal reasoning she used to get there (Schoenherr and

Black 2019b). Oral argument supplements the briefs, giving the justices the opportunity to

request clarifications, poke holes in arguments, and ask policy-oriented questions about a

case’s outcome (Johnson 2004). While attorneys present legal arguments in both situations,

the merits brief is the only place where they can make a complete, coherent, and uninter-

rupted argument in favor of their clients (Black, Hall, Owens and Ringsmuth 2016). This

is an attorney’s chance to “give the judge good reason for ruling as you request” (Garner

2003, IX) by presenting, organizing, and framing a case in just the right manner. Writing a

persuasive merits brief is a key to winning a case.

Writing a persuasive merits brief means leading the justices to an answer while avoiding

their irritants. At a minimum, the justices expect a brief that is technically correct (Gar-

ner and Kennedy 2010), well-written (Garner and Roberts 2010), and succinctly presented

(Scalia and Garner 2008; Totenberg 2011). They also favor briefs that are readable (Feldman

2016), detached and unemotional (Black, Hall, Owens and Ringsmuth 2016), and prepared

by experienced attorneys with multiple appearances before the Court (Nelson and Epstein

2019). Beyond the more mechanical aspects of the brief, the justices also expect attorneys

will identify the relevant legal issues and thoroughly explain their importance and applica-

tion to the situation at hand (Garner and Stevens 2010; Garner and Thomas 2010). As Chief

Justice John Roberts explained, “I, as a judge, have a responsibility to try to get the right

63

answer on the law. This brief is going to help me one way or another, and I want to get that

help out of it. And if you can’t express clearly what your position is, that’s not helping me”

(Garner and Roberts 2010, 28). The justices ask attorneys to be their legal escorts, leading

them from the case facts to the case disposition by way of the law. Attorneys are responsible

for finding the credible path to the mountain’s apex (Garner 2003). If they get the justices

lost or lead them into a crevasse, their client loses the case and they fail.

Credibly making it through the law requires skill in the face of the unknown. Federal

law is complicated at a minimum; even the justices periodically fail to reach an agreement

regarding the legal reasoning in a case and issue “unreasoned opinions” that simply state the

case disposition (Hitt 2019). Moreover, by nature of appealing to the Supreme Court, an

institution whose very function is to ensure uniform interpretation of federal law, attorneys

who approach the Court are trying to credibly answer legal questions that other brilliant
legal minds failed to answer to any level of satisfaction or uniformity.1 They do this knowing

the justices are strategic seekers of policy who may simply be using the law to bolster and

legitimize their pursuit of policy outcomes (Hansford and Spriggs 2006). In short, attorneys

have to figure out how to work through this complicated fray and appeal to the justices in

a manner that will secure a majority of their votes, and they do so without knowing exactly

which answer the justices want to find. Tasked with this duty, I suggest that attorneys work

through the fray by framing their brief using one of two different argument types: they can

use a prevailing legal argument, or they can engage in legal entrepreneurship. Each argument

type has its own rules and those rules help attorneys simplify the legal universe, identify the

relevant precedents, and make a credible appeal to the justices.

1As Alexander Hamilton explained in “Federalist 80,” the Founders created the Supreme
Court because somebody had to be responsible for “uniformity in interpretation of the na-
tional laws” as without one, “nothing but contradiction and confusion can proceed” (Hamilton
2001, 412). Additionally, Supreme Court Rule 10 explicitly states the justices are more likely
to review cases that involve some sort of lower court conflict (between federal courts of ap-
peal, between state courts of last resort, or some combination of the two) or an important
federal question (Perry 1991).

64

When attorneys use a prevailing argument, they commit to using the Court’s recent

jurisprudence to explain the situation at hand. Their goal is to build arguments that show

similarities between the current case and the Court’s past decisions. Attorneys traffic in the

familiar, citing the same cases the justices used in this area of case law before and discussing

them with the same reverence or disdain. Recall that in Lawrence v. Texas (2003), the

attorneys representing the petitioners wrote a brief that hailed Romer v. Evans (1996) and

advocated for the elimination of Bowers v. Hardwick (1986). In so doing, they celebrated
a recent privacy opinion while parroting its arguments.2 The attorneys wanted to show the

justices how similar the two cases were and thus suggest the Court should reach the same

result. This is what a prevailing argument does. These are conservative arguments, to be

sure, as they do not push the law too far, but simply ask the justices to uphold the legal

status quo. They appeal to the justices’ preferences for familiar arguments (Hazelton, Hinkle

and Spriggs 2019) and so-called “bright-line” rules that simplify the decision-making process

(Niblett, Posner and Shleifer 2010). Finding a prevailing argument should, consequently, be

a dependable, credible, relatively low-cost strategy for constructing a legal argument.

Alternatively, attorneys can engage in legal entrepreneurship and consciously abandon

the familiar in exchange for new and innovative arguments that, if accepted, would fun-

damentally change the Court’s entire approach to an area of case law. When attorneys

construct these types of arguments, they look completely out of place in the Court’s recent

jurisprudence. They suggest the justices generalize precedents they sought to limit, or they

suggest the justices limit the reach of a popular precedent, or they develop an argument

that looks completely out of place next to the Court’s recent decisions in that area of case

law. Consider, for example, Jay Sekulow’s argument in Board of Airport Commissioners

of the City of Los Angeles v. Jews for Jesus (1987), where he abandoned the commonly-
2According to Hansford and Spriggs’s (2006) measure of precedent vitality, Romer had
a score of +8 at the start of the 2002 term, indicating the justices cited the case positively,
while Bowers had a score of -6, suggesting the justices were already treating the precedent
with caution by the time Lawrence hit their docket.

65

used Establishment Clause framework and successfully convinced the justices that a ban

on handing out religious literature at an airport was an unconstitutional restriction of free

speech (Toobin 2008). The legal status quo did not appear to work for Sekulow, so he

sought to change the conversation entirely. When attorneys use entrepreneurial arguments,

they try to persuade the justices to their side by showing them a viable alternative to the

legal status quo. Making these arguments is costly on a Court that is both institutionally

averse to overturning precedent and looking for legal efficiency (Niblett, Posner and Shleifer

2010; Hansford and Spriggs 2006), but the justices are occasionally willing to accept change

when it is both credible and helps them reach desired outcomes (Gennaioli and Shleifer 2007;

Wedeking 2010). Attorneys consequently need to look for signs that the justices might be

willing to invest in an entrepreneurial argument.

4.2 Making the Decision to go Entrepreneurial

While neither prevailing nor entrepreneurial arguments are easy to construct, an en-

trepreneurial argument’s newness makes it more costly, as new arguments reach credibility

limits much sooner than prevailing arguments do (Wedeking 2010). Attorneys consequently

need some sort of sign that the justices might support their decision to abandon the pre-

vailing approach in favor of something different. I theorize that attorneys ultimately make

a decision about whether to engage in legal entrepreneurship or stick with the prevailing

argument by looking at the political and legal world surrounding a case. More specifically,

I suggest that attorneys make the decision to engage in entrepreneurship based on three

sets of conditions: (1) large-scale political opportunities, specifically the presence of a new

justice and the public’s overall policy mood; (2) the legal environment surrounding the case,

including the lower court’s understanding of the law and the issue area involved; and (3) the

attorneys’ own resources and skills.

Changes in the national political environment should create opportunities for legal change

and, consequently, signal to attorneys that an entrepreneurial argument might be worth it.

66

While legal change can be “hydra-headed” (Epstein and Kobylka 1992, 5), scholars have

identified a few factors that definitively drive legal change at the Supreme Court. Clark

(2019) finds that “social conditions met with clever lawyering” drive justices to reconsider

their approach to an area of case law (5), suggesting that lawyers who take advantage

of the electoral politics of the day can secure large-scale changes in the law. Epstein and

Kobylka (1992) suggest that, in addition to public mood, changes in the Court’s membership

contribute to legal change as well, as new justices can alter the Court’s approach to an area

of case law (Epstein and Kobylka 1992).

Knowing this, I develop two hypotheses regarding attorneys’ willingness to engage in legal

entrepreneurship based on the national political environment. To address the social condi-

tions of the era, I first theorize that attorneys will be more likely to engage in entrepreneurship

when public opinion favors one party over the other. I do not have any predictions about the

direction of the shift (e.g., attorneys are more likely to make an entrepreneurial argument

when public mood turns more liberal), but simply posit that attorneys should alter their

behavior regarding legal entrepreneurship based on so-called “public mood” (Stimson 2018).

In acknowledgment of the havoc that membership changes can cause, I also suggest that

attorneys will be more likely to engage in legal entrepreneurship when a new justice joins the

Court, as the uncertainty should create opportunity for an entrepreneurial argument to land.

Of course, attorneys have to consider the legal environment surrounding a case before

they can make a decision about how to proceed. The first consideration is the ideological

congruence between the lower court’s decision and the ideological tilt of the Supreme Court.

Research suggests the Court is more likely to review cases from ideologically-distant lower

courts and then overturn them (Black and Owens 2012b; Bryan and Owens 2017). When

the Supreme Court is aligned with the lower court’s decision, it is consequently more likely

to affirm. I suggest that if the justices are ideologically predisposed to favor the lower court’s

decision, then attorneys should be less likely to engage in legal entrepreneurship. After all,

67

why go for something new when the law already works in your favor?

The second consideration is the issue area under which the case falls. Some issue areas,

like criminal procedure or First Amendment law, have long-standing bright-line rules that

are difficult to eliminate or overturn (Graetz and Greenhouse 2016; Richards and Kritzer

2002). It is more difficult to make an entrepreneurial argument in those areas than it is to

make one in an area like privacy law, where the justices constantly reevaluate the law and are

willing to modify past precedent, if not overturn it outright (Hull and Hoffer 2001; Kaplan

2018). I thus theorize that attorneys are more likely to make entrepreneurial arguments in

some issue areas than they are in others.

I also believe that attorneys will be more likely to engage in legal entrepreneurship when a

case is salient. Attorneys know when they might have an important case before the justices

(Clark, Lax and Rice 2015) and might be more likely to try an entrepreneurial argument when

they know the justices will be carefully considering the ramifications of the case, especially

given that salience can alter the justices’ behavior (Lax and Cameron 2007).

The lower courts’ behavior is also part of the legal environment and should matter as

well. Lower court judges can send signals to the justices about legal activity in the rest of

the judiciary (Beim, Hirsch and Kastellec 2015) and the justices respond to those signals in

different ways (Black, Owens, Wedeking and Wohlfarth 2016; Perry 1991). Consequently,

I suggest that attorneys respond to those same lower court signals. First, I theorize that

attorneys are less likely to engage in legal entrepreneurship when there is lower court conflict.

Lower court conflict indicates the lower courts on the whole are struggling to use existing

precedent to answer certain legal questions (Perry 1991).

In these situations, the right

prevailing argument can answer the question, and an entrepreneurial argument would simply

add to the cacophony of legal arguments already being used.

Alternatively, I also suggest that attorneys are more likely to engage in legal entrepreneur-

ship when the lower court issues a dissent. When a lower court judge files a dissent in a

68

case, he sends a high-cost signal about problems with the majority decision. After all, lower

court norms suggest judges avoid writing separately whenever possible (Epstein, Landes and

Posner 2013). While large-scale legal conflict in the lower courts should scare attorneys away

from entrepreneurial arguments, a single dissent filed by a time-strapped lower court judge

should give attorneys a more individualized push to keep pointing out the issues with the ma-

jority’s approach to the law; if the judge took the time to point out issues with the majority’s

approach, then an enterprising attorney just might have a chance with an entrepreneurial

argument.

Finally, the last thing attorneys consider when deciding how to frame their argument

is their own ability to pursue a new line of legal reasoning. Attorneys should be more

entrepreneurial when they are experienced and well-funded. The justices are more likely

to vote in favor of an experienced attorney (McGuire 1995) and they are more likely to

borrow language from an experience attorney’s brief (Corley 2008). The Solicitor General

enjoys privileged status with the justices (Black and Owens 2012c) while also being one of

the winningest advocates before them (Wohlfarth 2009). Well-funded, high-status attorneys

are also advantaged at the Court (Black and Boyd 2012). These facts would suggest that

experienced and high-status attorneys know how to approach the justices and are thus better

able to employ entrepreneurial arguments. I thus theorize that attorneys will be more likely

to engage in legal entrepreneurship when they are experienced, in the Office of the Solicitor

General, or hold a high status.

4.3 Data and Methods

To better understand attorneys’ decision to engage in legal entrepreneurship, I develop an

original dataset of the legal arguments contained in merits briefs and identify attorneys’ use

of entrepreneurial arguments. I then employ this variable in concert with several measures

of political opportunity, legal environment, and attorney resources and skills so that I can

study attorney strategy regarding legal entrepreneurship. The data encompass 3,018 attor-

69

ney decisions regarding entrepreneurship that occurred between the 1984 and 2007 terms,

covering 1,509 cases in which the petitioner and respondent submitted one initial merits brief

each.

4.3.1 Dependent Variable: Legal Entrepreneurship

The dependent variable in this model is a dichotomous indicator of whether an attorney

did (1) or did not (0) utilize an entrepreneurial argument in his merits brief. Recall that

legal entrepreneurship is the decision to engage with the justices using new and innovative

legal arguments that would alter the justices’ current approach to an area of case law.

Operationally, then, legal entrepreneurship occurs when attorneys decide to either (1) suggest

the justices treat certain cases differently than they have in the past; (2) cite new cases that

look completely out of place in the Court’s more recent jurisprudence; or (3) employ some

combination of the two. To identify an attorney’s use of an entrepreneurial argument in a

brief, I follow a three-step process in which I use text analysis to identify an attorney’s use

of entrepreneurial arguments in a merits brief.

In the first step of the process, I used text analysis tools to identify the contents of the

legal arguments contained in each merits brief submitted to the Supreme Court between

the 1984 and 2007 terms. By the contents of the legal arguments, I explicitly mean the

cases the attorneys cited as well as their discussion of those cases, specifically the sentence

that preceded each case mention. To do this, I began by using a combination of Westlaw
and LexisNexis to locate the texts of the briefs associated with these cases.3 I limited my

data collection to cases that had only one petitioner brief and one respondent brief, which
3Both Westlaw and LexisNexis were missing briefs for about 5% of the Court’s cases
during this time period. These missing briefs are mostly (but not exclusively) from cases the
Court reviewed in the early 1980s. Per a conversation with several members of Westlaw’s
product management team, the missing briefs were difficult to convert to a digital format
and therefore not included in their online repository. Westlaw is currently working to rectify
the problem.

70

covers approximately 85% of the cases the Court heard during that time period (Black
et al. 2020).4 With the texts downloaded, I used a program I wrote in R to identify the

“Arguments” section of each brief, or the part of the brief that walks through a party’s legal

reasoning for moving from the legal question at hand to their proposed answer. The program

then worked through the argument section line by line to locate every mention an attorney
made of a Supreme Court precedent.5 Upon finding a citation, the program collected both
the citation and its context, specifically the sentence that preceded it.6 At the end of this

step, I had information about the citations contained in the 3,018 briefs associated with the

1,509 cases heard between the 1984 and 2007 terms that are under analysis here, totaling

almost 190,000 mentions of past precedent in the briefs.

With the citations located, the next step was to identify attorneys’ treatment of those

citations. To complete this part of the process, I used a combination of hand-coding and

machine-learning techniques to create a dictionary of terms associated with attorneys’ dis-

cussion of the Court’s past decisions, then employed that dictionary to identify attorneys’
treatment of the citations they mentioned.7 The dictionary separates discussions of prece-
4I did this to help eliminate potential endogeneity problems, particularly issues that might
stem from the justices hearing different arguments for the same side in a case (e.g., two
different arguments from the petitioner). Importantly, the Court prefers parties consolidate
their arguments into one; in fact, Supreme Court Rule 28 says that, with some exceptions,
“only one attorney will be heard for each side” at oral argument, essentially forcing the
parties to reach an agreement regarding their arguments well before the justices hear a word
from them.

5Because Supreme Court justices are only bound by Supreme Court precedent, I do not
use state court decisions or lower federal court decisions in my analysis. I also eliminate
references to the federal code (e.g., 18 U.S. Code §1657.)

6By convention, attorneys typically cite a Supreme Court case using its name, followed by
the United States Report citation, and then the year the justices published the opinion, e.g.,
Miranda v. Arizona, 384 U.S. 436 (1966). Inevitably, however, attorneys turn to shorthand
to save space in their briefs, and the formal citation gets reduced to Miranda. The program
I wrote identifies full citations and partial citations (e.g., references to either Miranda v.
Arizona or 384 U.S. 436), and then uses word frequencies to identify the shorthand citations
as well. When compared to hand-coded data, the program correctly identifies more than
90% of the total number of citations mentioned in a brief.

7As I discuss in Chapters 2 and 3, I purposefully use a dictionary here. While both

71

dent into one of three different categories: situations where the case applies to the situation

at hand, situations where the case does not apply, and simple statements of fact. These ap-

plications map to the positive, negative, and neutral treatments used by Shepard’s Citations,

respectively (Spriggs and Hansford 2000). The dictionary contains a list of 525 words and

phrases that uniquely identify attorneys’ decisions to treat a precedent as applicable to the

current case (“apply,” “establish,” “mandate”) or dismiss a precedent as inapplicable (“distin-
guish,” “nullify,” “overrule”).8 I used Benoit et al.’s (2017) quanteda package to apply this

dictionary to the attorney’s discussions of the citations, then aggregated that information to
create an overall measure of precedent treatment for each individual case in each brief.9 All

other citations were treated as neutral statements of fact. By the end of this step, I had data

on attorneys’ treatment of each case mentioned in the argument section of a brief, covering

almost 80,000 unique brief-citation pairings. So, for example, the data now explicitly reflect

the fact that the attorneys representing the petitioners in Lawrence said the Court’s ruling

in Romer applied to the situation at hand and that Bowers did not.

With the data on attorneys’ legal arguments collected, I moved to the final step, in

which I identify attorneys’ use of entrepreneurial arguments. To do this, I began by making

a necessary simplifying assumption: that entrepreneurship appears in the most-cited cases in

dictionary-based approaches and more high-powered black-box models offer accuracy and
efficiency (Grimmer and Stewart 2013) and I achieve similar results using either classifi-
cation scheme, I ultimately settled on the dictionary-based approach because it also offers
information about the words associated with the treatments, which is useful for scholars
seeking to better understand how attorneys make arguments.

8I developed this list using information from three different sources: LexisNexis’s list of
shepardization terms and treatments, Spriggs and Hansford’s (2000) overview of the shepar-
dization process, and hand-coded sentiment data gathered from 57 randomly-selected search
and seizure cases and 10 privacy cases. I outline this process more clearly in Chapter 2.

9For the most part, the aggregation process was a simple summation of treatment across
the mentions of a case. For any citation that got mentioned more than 15 times in a brief
(i.e., the one-to-two most-cited cases in each brief), I manually confirmed the attorney’s
treatment of the case by reading the brief. To be clear: simple summation would still get
the right answer almost every time, but validation is essential to successful text analysis
(Grimmer and Stewart 2013).

72

a brief. As I explained in Chapter 2, an average merits brief references about 26 past Supreme

Court decisions. The modal number of times a citation gets used is one, while the average

is almost three. This disparity suggests that attorneys mention most of their citations only

once but mention a few central cases repeatedly. Interviews with the justices confirm this is

both what they expect to see and what good attorneys tend to do (Garner and Stevens 2010;

Garner and Thomas 2010), and guides to legal writing are adamant that attorneys focus on

the important cases and try to avoid the one-off citations as much as possible (Garner 2003).

Given the evidence, I look for entrepreneurial behavior in the cases that get mentioned an

above-average number of times in a case. Typically, this resulted in looking at about 6 cases

per brief, rather than looking at 26. Making this decision eliminated a significant amount

of noise from the data; there was little consistency in the citation patterns of these one-off
cases and they essentially overwhelmed the analysis.10

With that simplifying assumption made, I turn to defining legal entrepreneurship. Recall

there are two ways to be entrepreneurial: first, to suggest the Supreme Court alter its

treatment of a certain precedent, and second, to use unique citations, or citations that

appear out of place in the Court’s jurisprudence. Beginning first with variation in treatment,

I proceed in three steps: (1) identify the cases that got cited in both briefs; (2) within that

set of citations, identify the situations where attorneys applied the cases differently; and (3)

find the attorney that argued against the Court’s treatment of the case. To do step 3, I

used Hansford and Spriggs’s (2006) measure of precedent vitality to identify the attorney

who argued against the Court’s current approach to the case. Their measure is essentially a

running tally of the Court’s treatment of its own decisions over time. Cases that the Court

repeatedly treats positively are considered high-vitality cases (e.g., the Court’s reverence for

Brown v. Board of Education [1954]), while rulings that the Court distinguishes, questions,
10More specifically, looking at all the cases mentioned in each brief ultimately suggested
that every brief had an entrepreneurial argument. The wide variation in cases cited and
attorneys’ discussions of them made each brief look like it had a more unique argument than
it did.

73

and overturns are considered low-vitality cases (e.g., the Court’s discussion of a case like

Korematsu v. United States [1944]). If an attorney suggested overturning a high-vitality

case or venerated a low-vitality case, he made an entrepreneurial argument.

The second way to be entrepreneurial is to bring a unique argument to the table, specif-

ically to cite precedents that have not been used in the Court’s recent jurisprudence. To

identify these instances of entrepreneurship, I again turn to Hansford and Spriggs’s (2006)

precedent vitality data, which also notes the frequency with which the justices cite past

precedents. I consider an attorney to be behaving in an entrepreneurial manner if he cites a

case the Court has never cited. If the Court has never felt the need to revisit the precedent,

then the attorney bringing it up is an entrepreneurial act. I also consider an attorney to be

engaging in legal entrepreneurship if he mentions a case the Court has not used in an opinion

in the last ten terms. I consider these cases to be “out of sight, out of mind” precedents, and

attorneys who mention them are bringing them back for a new viewing.

In summary, then, an attorney used an entrepreneurial argument in his brief if he did

any or all of the following three things in his brief: (1) he and another attorney based their

arguments on the same case but his discussion alone suggested the Court should reverse its

current approach; (2) he built his argument around a case the Court never cited in a later

opinion; or (3) one of his main citations is a case the Court has not cited in the last ten

years. Entrepreneurship is consequently a dichotomous variable that indicates the attorney
did (1) or did not (0) engage in legal entrepreneurship.11

11Given the way I measured legal entrepreneurship, it is possible to measure it on an
additive, continuous scale. There is no theoretically sound reason to do this, however, as I
am unwilling to assume that entrepreneurship is additive. But I did run this same model
using OLS and the continuous dependent variable. The results are in Model 1 in the appendix
and are substantively the same.

74

4.3.2

Independent Variables: Opportunity, Environment, and Resources

As I suggest earlier in the paper, attorneys, much like the justices themselves, should

look for an opportunity to change the current approach to an area of case law, and they

do so across three different sets of conditions:

large-scale political opportunities, the legal

environment surrounding the case, and the attorneys’ own resources and skills.

Beginning first with the large-scale political opportunities, there are two factors to con-

sider: the social conditions of the time period as well as membership changes on the bench.

To measure social conditions, I use Stimson’s (2018) measure of public mood. Using scaling

techniques on a series of public policy questions asked over time, Stimson’s measure iden-

tifies public support for government programs on a 0-to-100, liberal-to-conservative scale,

with larger values indicating support for more liberal policies (Yglesias 2019). In order to

study membership changes on the bench, I include a dichotomous indicator of whether or
not a new justice joined the bench that term.12

When examining the legal environment surrounding the case, I suggest that attorneys

consider five different factors: ideological congruence between the lower court decision and

the Supreme Court, the issue area under which the case falls, the salience of the case, the

presence of lower court conflict, and the presence of a lower court dissent. To measure

ideological congruence, I first identify the median justice’s Judicial Common Space (JCS)

score (Epstein et al. 2007). I use the median justice as a proxy for the Court’s ideology;

if the median is conservative, then so is the majority of the Court, and vice versa. JCS

scores measure a justice’s ideology on a -1 to 1 scale, with more positive scores correlating

to more conservative justices (or, in this case, more conservative medians). I then use the

Supreme Court Database to identify if the lower court decision was conservative or liberal
12Given the nature of the Supreme Court’s docket and calendar, the justices often grant
certiorari in cases in one term and then review it in another. As a result of this quirk, I also
ran these models with an indicator for a new justice appearing in one of the last two years.
These results, which are in Model 2 the appendix, remain substantively the same.

75

(Spaeth et al. 2017). Because the median on the Court is always conservative in this time

period, I code lower court decision congruence as the median justice’s JCS score if the lower

court decision was in a conservative direction, and if the lower court decision was liberal,

congruence is the negative value of the median justice’s JCS score.

To identify the second factor, issue area of a case, I use the Supreme Court Database

to identify the issue area into which each case falls and then create ten dichotomous issue
area variables, one for each area under study.13 To examine a case’s salience, the third legal

environment factor, I use Clark, Lax and Rice’s (2015) measure of latent case salience to

examine how an attorney’s decision to engage in legal entrepreneurship might change when

dealing with a salient case.

Finally, I also employ two variables to deal with behavior at the lower court level. To

identify instances of lower court conflict, I again turn to the Supreme Court Database, this

time using the certReason variable to find the cases in which the justices pointed to any

type of lower court conflict as a reason for reviewing the case. I then create a dichotomous
lower court conflict variable.14 To identify lower court dissents, I employ the Supreme Court

Database’s lower court disagreement dichotomous variable here.

The final set of variables that I include in my model deal with attorney resources, specifi-

cally their experience, position as a member of the Office of the Solicitor General, and status.

I use Black, Hall, Owens and Ringsmuth’s (2016) data for all of these measures. Attorney

experience is the natural log of (one plus) the attorney’s past oral argument experience, to
13The Supreme Court Database places each case into one of fourteen exclusive issue area
categories. None of the cases under study here fall into the eleventh, thirteenth, or fourteenth
categories (interstate relations, miscellaneous cases, and private action cases). That leaves
the cases in eleven issue areas for analysis. I treat economic activity cases (category eight)
as the baseline category simply because it is the largest category.

14The Supreme Court Database certReason variable has thirteen categories, five of which
involve conflict between lower courts (categories 2-6).
If the Court listened reasons two
through six as their reason for granting certiorari in the case, the lower court conflict variable
took the value of 1.

76

see if more experienced attorneys are more likely to engage in entrepreneurship.15 Solicitor

General’s presence is a dichotomous indicator of whether or not the Solicitor General is the

party putting together the merits brief. Finally, Black, Hall, Owens and Ringsmuth (2016)

created the measure of attorney status by following a procedure originally outlined by Collins

(2004, 2007) that uses the Supreme Court Database’s party codes to categorize each party

in a case into one of 10 groups. The weakest parties – poor individuals – are coded as 1

while the strongest party – the U.S. Government – is coded as 10.

The last variable I include in this model is a control for whether or not the attorney

making the entrepreneurial argument is the petitioner. As Wedeking (2010) explains, the

petitioner lost at the lower court and therefore might seek to employ a new argument before

the justices – after all, the last one did not work. When combined with the fact that the

petitioner presents her brief first, these factors might suggest the petitioner is more willing

to engage in legal entrepreneurship than the respondent might be.

4.4 Methodology and Empirical Results

Because my dependent variable is dichotomous, I use a logistic regression model for the

analysis (Long 1997), and I estimate the model using standard errors that are clustered by

case. Table 4.1 shows the results of the model of the likelihood an attorney engages in legal

entrepreneurship. Given the non-linear nature of the model, I use predicted values to address

the results rather than looking strictly at the coefficients, and I calculate these using the

observed-value approach best associated with Hanmer and Kalkan (2013).

15Experience at oral argument is an accepted proxy for overall attorney experience on a
brief (Black, Hall, Owens and Ringsmuth 2016; Black et al. 2020; Corley 2008). Multiple
attorneys work on a brief, but anecdotes suggest the most experienced attorney will be the
one who argues the case (Biskupic, Roberts and Shiffman 2014). If several novices write
the brief, then a novice is going to argue it as well; consider stories of new state attorneys
general who have catastrophic showings before the justices because they refused to accept
help from more experienced litigators (Toobin 2008). And while experienced attorneys may
take over cases they believe in, those attorneys are also going to help write the brief, as they
have to defend that brief during oral argument.

77

Table 4.1: Logistic Regression Results, Attorney Decision to Engage in Legal Entrepreneur-
ship

(Standard Errors)

Coefficient
−0.108
(0.080)
0.015
(0.011)

−0.325∗∗
0.300∗

(0.110)

(0.117)
0.104
(0.135)
0.078
(0.177)
0.196
(0.199)
0.299
(0.323)
−0.190
(0.286)
−0.579∗
0.466∗∗∗
(0.137)
−0.222
(0.181)
0.194
(0.245)
0.090
(0.054)

(0.269)

(0.080)

−0.187∗
0.187∗
0.098∗∗

(0.085)

(0.038)
0.020
(0.147)
−0.029
(0.016)
−0.093
(0.074)
−0.787
(0.695)

New Justice

Public Mood

Ideological Congruence

Criminal Procedure Case

Civil Rights Case

First Amendment Case

Due Process Case

Privacy Case

Attorneys Case

Unions Case

Judicial Powers Case

Federalism Case

Federal Taxation Case

Latent Case Salience

Lower Court Conflict

Dissent Noted in Lower Court

Attorney Experience

U.S. Government Attorney

Attorney Status

Petitioning Attorney

Constant

Observations
AIC
BIC
Log Likelihood
Standard errors, clustered by case, in parentheses
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

3, 018
4144.832
4271.091
−2051.416

78

I begin my analysis by first examining how large-scale political opportunities – a new

justice joining the Supreme Court and overall public mood – influence the probability an

attorney engages in legal entrepreneurship. Beginning with the new justice variable, I find

that attorneys are no more or less likely to make an entrepreneurial argument when a new

justice joins the Court. Attorneys who face a stable, familiar bench have a 0.53 [0.51, 0.55]

probability of engaging in entrepreneurship. When the justices lose an old friend and gains

a new colleague, that probability decreases to 0.50 [0.47, 0.53], but the difference is not

statistically significant. In short, despite my expectations, I do not find that attorneys treat
membership changes as a signal to be entrepreneurial.16

The results also suggest attorneys are no more or less likely to engage in legal en-

trepreneurship when public mood changes. As public mood moves from more conservative

to more liberal (specifically, from 56 to 66, covering 95% of the data under study here), the

probability an attorney engages in legal entrepreneurship in the brief increases slightly, from

0.50 [0.46, 0.54] when the mood is more conservative to 0.54 [0.51, 0.57] when the mood takes

a moral liberal turn. Again, the difference is not statistically significant, suggesting that at-

torneys do not change their approach based on public mood. This result goes against my

theory, but it does make sense, given recent work by Devins and Baum (2019) that suggests

the justices respond to elite-level opinion and not broad public mood. On the whole, these

results regarding membership changes and public mood suggest that large-scale political

opportunities are not related to an attorney’s decision to engage in legal entrepreneurship.

Turning next to the variables that identify the legal environment surrounding the case,
16One could argue that the truly important changes on the Court are the so-called “move-
the-median” confirmations (Krehbiel 2007), which occur when a new justice replaces an
ideologically-distant justice, like when Clarence Thomas replaced Thurgood Marshall. To
see if this understanding of membership changes is the true difference maker, I replace
the new justice variable with an indicator for a median-moving justice joining the bench
and run the same model. As Model 3 in the appendix shows, having a median-moving
justice join the bench does not significantly alter the probability an attorney engages in legal
entrepreneurship.

79

the results suggest attorneys are more likely to engage in legal entrepreneurship when the

environment signals they should. Starting with Figure 4.1, I find that the probability an

attorney engages in legal entrepreneurship decreases as ideological congruence increases. As

the left side of Figure 4.1 shows, the probability an attorney engages in legal entrepreneurship

when the lower court decision and the Supreme Court median are incongruent is 0.55 [0.53,

0.58]. That is, an attorney has a 55% likelihood of making an entrepreneurial argument in a

brief when the lower court decision is liberal and the Supreme Court median is conservative.

But, when the lower court decision aligns with the Supreme Court median (i.e., the lower

court decision is conservative and so is the median), the probability an attorney engages

in legal entrepreneurship decreases significantly, to 0.49 [0.46, 0.52]. This six-percentage-

point decrease would suggest that, as I expected, attorneys are less likely to engage in legal

entrepreneurship when they believe the median’s preferences, and therefore the preferences

of the majority of the Court, align with the lower court’s decision.

Figure 4.1: Probability an Attorney Engages in Legal Entrepreneurship as Ideological Congru-
ence Increases - Probability an attorney engages in legal entrepreneurship as the ideological congruence
between the lower court’s decision and the Supreme Court median increases. Dashed lines are 95% confidence
intervals around those estimates. Predicted probabilities calculated using the observed-value approach.

80

0.460.470.480.490.500.510.520.530.540.550.560.570.580.590.60-0.42-0.32-0.22-0.12-0.020.080.180.280.38Ideological CongruenceProbability an Attorney Engages in Legal EntrepreneurshipMy results also show that attorneys are more likely to engage in legal entrepreneurship

in some issue areas than others. As Figure 4.2 shows, attorneys modify their approach

based on the issue area. They are slightly more likely to engage in entrepreneurship in

privacy cases (probability of 0.56) or cases involving questions of judicial power (probability

of 0.60), while they are less likely to make an entrepreneurial argument when arguing in a

case involving unions (probability of 0.35) or federalism (probability of 0.43). In short, as I

suggested earlier, attorneys look for opportunities in certain issue areas and are less likely

to try entrepreneurial arguments in others.

Figure 4.2: Probability an Attorney Engages in Legal Entrepreneurship by Issue Area – Proba-
bility an attorney engages in legal entrepreneurship based on the issue area into which the case falls. Vertical
lines are 95% confidence intervals. Predicted probabilities calculated using the observed-value approach.

The results also suggest that attorneys are less likely to engage in legal entrepreneurship

when lower court conflict exists. Figure 4.3 shows that attorneys are significantly more

likely to engage in legal entrepreneurship when lower courts do not conflict on an issue

(probability of 0.54 [0.52, 0.56]) than they are when the lower courts produce conflicting

opinions (probability of 0.50 [0.46, 0.52]). Attorneys are equally likely to select a prevailing

81

0.200.250.300.350.400.450.500.550.600.650.70CriminalProcedureCivilRightsFirstAmendmentDueProcessPrivacyAttorneysUnionsEconomicActivityJudicialPowerFederalismFederalTaxationIssue AreaProbability an Attorney Engages in Legal Entrepreneurshipor entrepreneurial argument when the lower courts struggle to find a valid answer, which

suggests, as I expected, that wide-scale conflict might dampen an attorney’s willingness to

be entrepreneurial.

Figure 4.3: Probability an Attorney Engages in Legal Entrepreneurship When Lower Court
Conflict Exists – Probability an attorney engages in legal entrepreneurship when lower court conflict
exists (right) or does not (left). Vertical lines identify 95% confidence intervals. Predicted probabilities
calculated using the observed-value approach.

Additionally, attorneys are significantly more likely to make an entrepreneurial argument

when a lower court justice filed a dissent. The left side of Figure 4.4 shows that an attorney

has a 0.51 [0.49, 0.53] probability of using an entrepreneurial argument when the justices fail

to notice a lower court dissent. But, when the dissent is prominent enough that it draws the

justices’ attention, that probability increases by five percentage points to 0.56 [0.52, 0.59], as

seen on the right side of Figure 4.4. These results suggest attorneys recognize that a lower

court dissent might signal issues with the majority’s approach to the law and give attorneys

the push they need to make an entrepreneurial argument.

82

0.450.460.470.480.490.500.510.520.530.540.550.560.57No Lower Court ConflictLower Court ConflictProbability an Attorney Engages in Legal EntrepreneurshipFigure 4.4: Probability an Attorney Engages in Legal Entrepreneurship When a Lower Court
Judge Dissents – Probability an attorney engages in legal entrepreneurship when the Supreme Court does
not note a lower court dissent (left) or writes about a lower court dissent in the eventual opinion (right).
Vertical lines identify 95% confidence intervals. Predicted probabilities calculated using the observed-value
approach.

Reaching the final legal environment variable, I find that, contrary to my expectations,

attorneys are no more or less likely to make an entrepreneurial argument when a case is

salient. When a case is low salience, attorneys have a 0.50 [0.47, 0.53] probability of engaging

in legal entrepreneurship, while they have a 0.55 [0.51, 0.59] probability of doing the same

when the case is high salience. While the numbers would suggest attorneys are more likely to

use an entrepreneurial argument in salient cases, the difference is not statistically significant.

The last set of variables under study here are the ones concerning attorney resources.

Starting first with attorney experience, which is presented in Figure 4.5, I find that more

experienced attorneys are significantly more likely to engage in legal entrepreneurship than

are their less-experienced colleagues. Moving from the low value of no experience before the

justices (logged value of 0) to an otherwise extraordinary 29 past appearances before the

justices (logged value of 3.4), the inexperienced attorney has a 0.50 [0.48, 0.52] probabil-

83

0.470.480.490.500.510.520.530.540.550.560.570.580.59No Mention of Lower Court DissentMention of Lower Court DissentProbability an Attorney Engages in Legal Entrepreneurshipity of engaging in entrepreneurship, while the experienced attorney has a 0.58 [0.53, 0.64]

probability of doing the same. This eight-percentage-point increase suggests that attorneys

feel more confident making novel arguments when they are more familiar with the Supreme

Court’s ebbs and flows.

Figure 4.5: Probability an Attorney Engages in Legal Entrepreneurship Based on Attorney
Experience – Probability an attorney engages in legal entrepreneurship as the attorney’s past experience
at oral argument increases. Dashed lines are 95% confidence intervals around those estimates. Predicted
probabilities calculated using the observed-value approach.

Interestingly, the results suggest attorney experience is the only resource that mat-

ters. According to the results, the Solicitor General is no more or likely to make an en-

trepreneurial argument than his colleagues; he has a 0.54 [0.47, 0.60] probability of making

an entrepreneurial argument, while other attorneys are two percentage points less likely to do

the same. The difference is not statistically significant. Attorneys representing high-status

clients are less likely to make entrepreneurial arguments than are attorneys representing

economically-disadvantaged clients, with the probability of making an entrepreneurial argu-

ment decreasing from 0.57 to 0.49, but that difference is not statistically significant either.

Finally, the results suggest the petitioner and respondent are equally likely to make an en-

84

0.470.480.490.500.510.520.530.540.550.560.570.580.590.600.610.620.630.640.650.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.23.4Attorney Experience at Oral Argument (logged)Probability an Attorney Engages in Legal Entrepreneurshiptrepreneurial argument. Despite perhaps having more reason to go entrepreneurial, attorneys

representing the petitioner are not more willing to partake in entrepreneurial behavior.

4.5 Discussion

Twelve years after the Supreme Court’s decision in Lawrence v. Texas (2003), Justice

Anthony Kennedy delivered another majority opinion in support of same-sex rights, this time

ruling that the same Due Process Clause that protected same-sex couples from intrusion into

their bedrooms also guaranteed same-sex coupes the right to marry (Frank 2017). When

Kennedy announced the decision in Obergefell v. Hodges (2015), he did so in a world that

was dramatically different from the one in which Garner and Lawrence’s attorneys originally

approached the Court – public opinion had shifted in favor of gay rights, same-sex marriage

was already legal in 37 states and the District of Columbia, and the Obama administration

even helped the six same-sex couples argue their case before the Court (Carpenter 2012;

Kaplan 2018). But the attorneys who represented those couples did not approach the Court

simply because times had changed; they also did so because, after Lawrence, they knew they

could win (Frank 2017). Had the attorneys representing Tyron Garner and John Geddes

Lawrence misread the signs or misinterpreted the need to use a conservative argument in

Lawrence v. Texas, Obergefell could have taken much longer than twelve years to happen.

Every day, attorneys make decisions that can change the direction of future cases. They

decide which words to use, which appeals to make, which cases to cite, and how to discuss

them. And, importantly, they make a decision about how to frame their arguments, deciding

whether to stick with the familiar and prevailing or engage in entrepreneurial behavior. As

the results I present here show, attorneys are strategic about when to employ a new argument

and when to stick with the prevailing one, modifying their behavior based on greater forces

surrounding their cases. While membership changes and public mood may not influence

their decisions, attorneys really do pay attention to the legal environment surrounding their

cases. They look for signs from the lower courts that the justices might be willing to accept

85

entrepreneurial arguments, and they are more willing to try these arguments in some situ-

ations than in others. Experienced attorneys are more likely to engage in entrepreneurship

than novices, but other resources do not help them make decisions regarding their argument

framing. Attorneys, in short, make strategic decisions that can eventually change outcomes

on the U.S. Supreme Court.

While I am confident in the results presented in this paper, there are, as always, limits in

its approach. For one thing, I treat entrepreneurship as a singular activity – attorneys either

engage in it, or they do not. But the attorneys themselves suggest their entrepreneurial en-

deavors can be either gradual or dramatic, suggesting that a more complicated entrepreneur-

ship scale might be more appropriate at some point (Toobin 2008; Mauro 2019). In the future,

I could develop a more refined measure of engagement with entrepreneurial arguments to see

if attorneys work by degrees.

Additionally, the supervised machine learning processes I use here are not prefect. Reli-

able machine learning and text analysis requires constant monitoring, validation, and correc-

tion (Grimmer and Stewart 2013; Schoenherr and Black 2019b); to get data that is akin to

human coding, scholars have to invest significant time in checking the computerized output.

And even then, the computer can still produce errors that human coders would not make,

or find itself unable to correct errors that humans might find. Consider, for example, that

it is almost impossible for a computer to differentiate between casual mentions of any case
name shortened to Johnson17 or Smith.18 While a human coder could identify these casual

mentions based on context, the computer cannot, and these short-hand citations are con-

sequently left out of my analysis. Additionally, attorneys make mistakes in the text itself.
17A short list would include Johnson v. United States (2015), a criminal rights case, and
Texas v. Johnson (1989), the famous flag-burning cases, which should not be confused with
Johnson v. Texas (1993), another criminal rights case.

18A non-comprehensive list would include Employment Division v. Smith (1990), a reli-
gious rights case, Smith v. Texas (2007), on jury instruction, or Smith v. Berryhill (2019),
a Social Security case.

86

They occasionally mis-cite cases, which humans can find and computers cannot.19 Attorneys

also make typos; one slip of the finger could change a discussion of Terry v. Ohio, 392 U.S. 1

(1966) into a study of Californians v. California, 393 U.S. 1 (1968) in my data. Continuing

to employ human-assisted coding along the automated process would help mitigate some of

these problems, but they will probably always exist in some form.

My goal here was to present an exploratory analysis of what attorneys might consider

when deciding how to approach the justices. Moving forward, then, scholars should continue

to identify and test other factors that might influence attorneys’ decisions. Is it possible that

attorneys, like the justices, pay attention to how many amicus briefs get filed along their

petition for writ of certiorari (Caldeira and Wright 1988; Schoenherr and Black 2019a), and

therefore modify their approach to case based on anticipated levels of support? Anecdotal

evidence from the Women’s Rights Project suggests that Ruth Bader Ginsburg was able to

line up support and therefore expand her argument beyond the pages of her own brief (DeHart

2018); is it possible that other attorneys are more likely to engage in legal entrepreneurship

when they know they can do the same?

Another possible avenue for future research would involve an over-time analysis of at-

torneys’ decisions to innovate. Were attorneys more likely to make the decision to innovate

when certain justices were the median? Or were they more likely to use prevailing argu-

ments when worried the newer, more conservative Burger Court would overturn Warren

Court precedents? These are all questions worth greater investigation using the approach

that I outlined here.

19Hilariously, Justice Harry Blackmun would correct citations in his copies of briefs while

also noting his irritation that the attorney could not find the right citation.

87

CHAPTER 5

CONCLUSION

Most of the time, when people – consumers of news, reporters, judicial politics scholars,

or lawyers – talk about the Supreme Court, they tend to develop narratives about how the

justices’ personalities and ideological preferences drive outcomes at the Court. When the

Court upheld key provisions of the Affordable Care Act in National Federation of Independent

Businesses v. Sebelius (2012), journalists and academics alike told the tale of Chief Justice

Roberts’s heroic decision to side with the liberal justices in order to maintain the legitimacy

of the institution he steered (Biskupic 2019; Toobin 2012). For at least a ten-year period,

any discussion of an important case included commentary on how Justice Anthony Kennedy,

a man the New Yorker once branded “the agonizer,” would make the decisive vote in the

case (Kaplan 2018; Lithwick 2015; Rosen 1996; Tribe and Matz 2014). And before Kennedy,

there was Justice Sandra Day O’Connor, who, the reporters pointed out, was a no-nonsense

Goldwater Republican from rough-and-tumble Arizona whose judicial philosophy (or lack

thereof) perpetually kept the Court’s decisions from moving too far away from public opinion

(Biskupic 2005; Thomas 2019; Toobin 2008).

Academic research does this too. Segal and Spaeth (2002) looked at voting patterns

to establish that “Rehnquist votes the way he does because he is extremely conservative;

Marshall voted the way he did because he was extremely liberal” (86). Epstein and Knight

(1998) and later Bailey and Maltzman (2011) offered the correction that Supreme Court

justices are strategic seekers of policy who must deal with institutional and legal constraints

before they can vote in a policy-minded manner, a suggestion that led to a thousand pieces

of research on the constraints (for a quick sampling, see Baum 2006; Black and Owens

2009; Caldeira, Wright and Zorn 1999; Clark 2009; Hansford and Spriggs 2006; Johnson,

Wahlbeck and Spriggs 2006). And even more recent work by Black et al. (2020) and Hall

88

(2018) point out that the justices’ own personality profiles can alter their behavior on the

Court. Again, the emphasis is clear: the justices’ ideological preferences and personalities

are driving outcomes at the Supreme Court.

My goal in these pages was to show that the justices alone are not driving outcomes at

the United States Supreme Court. All too often, the research on the Court’s decision-making

process implicitly assumes that the justices independently identify the legal rationale for their

decisions – that is to say, that the justices’ knowledge of the law alone controls their decision-

making calculus. They do not. At the justices’ request, attorneys provide the justices

with their initial exposure to the legal arguments involved in a case and set boundaries

around it (Epstein, Segal and Johnson 1996). Attorneys and the legal arguments they make

consequently play a major role at the Supreme Court. While the justices’ personalities and

preferences undoubtedly influence outcomes at the Supreme Court, attorneys’ arguments

matter too. Attorneys help push the justices toward certain choices, and their decisions

matter and deserve study in dissertations well beyond this one.

89

APPENDICES

90

APPENDIX A

CHAPTER 3 APPENDIX

Figure A.1: Difference in Probability a Justice Sides with the Petitioner Based on Decision
to Engage in Legal Entrepreneurship and Attorney Experience – Probability a justice sides with
the petitioner, based on petitioner experience, when the petitioner engages in legal entrepreneurship minus
the probability a justice sides with the petitioner, based on petitioner experience, when the petitioner
uses a prevailing argument. Dashed lines are 95% confidence intervals around those estimates. Predicted
probabilities calculated using the observed-value approach.

91

-0.08-0.07-0.06-0.05-0.04-0.03-0.02-0.010.000.010.020.030.040.050.060.070.080.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.23.4Attorney Experience at Oral Argument (log)Difference in Probability the Justice Sides with the PetitionerFigure A.2: Difference in Probability a Justice Sides with the Respondent Based on Decision to
Engage in Legal Entrepreneurship and Attorney Experience – Probability a justice sides with the
respondent, based on respondent experience, when the respondent engages in legal entrepreneurship minus
the probability a justice sides with the respondent, based on respondent experience, when the respondent
uses a prevailing argument. Dashed lines are 95% confidence intervals around those estimates. Predicted
probabilities calculated using the observed-value approach.

92

-0.10-0.09-0.08-0.07-0.06-0.05-0.04-0.03-0.02-0.010.000.010.020.030.040.050.060.070.080.090.100.00.20.40.60.81.01.21.41.61.82.02.22.42.62.83.03.2Attorney Experience at Oral Argument (log)Difference in Probability the Justice Sides with the RespondentFigure A.3: Difference in Probability a Justice Sides with the Petitioner Based on Decision to
Engage in Legal Entrepreneurship and Litigant Status – Probability a justice sides with the petitioner,
based on petitioner status, when the petitioner engages in legal entrepreneurship minus the probability a
justice sides with the petitioner, based on petitioner status, when the petitioner uses a prevailing argument.
Dashed lines are 95% confidence intervals around those estimates. Predicted probabilities calculated using
the observed-value approach.

93

-0.05-0.04-0.03-0.02-0.010.000.010.020.030.040.050.060.070.080.090.1012345678910Petitioner StatusDifference in Probability Justice Votes with PetitionerFigure A.4: Difference in Probability a Justice Sides with the Respondent Based on Decision
to Engage in Legal Entrepreneurship and Litigant Status – Probability a justice sides with the
respondent, based on respondent status, when the respondent engages in legal entrepreneurship minus the
probability a justice sides with the respondent, based on respondent status, when the respondent uses a pre-
vailing argument. Dashed lines are 95% confidence intervals around those estimates. Predicted probabilities
calculated using the observed-value approach.

94

-0.05-0.04-0.03-0.02-0.010.000.010.020.030.040.050.060.0712345678910Respondent StatusDifference in Probability Justice Votes with PetitionerAPPENDIX B

CHAPTER 4 APPENDIX

I considered several alternate specifications when constructing my model of an attorney’s
decision to engage in legal entrepreneurship in Chapter 4. I present them here. Table B.1
contains three models. Model 1 is an ordinary least squares (OLS) regression on an alter-
native specification of the dependent variable, namely a continuous indicator of attorneys
engaging in 1, 2 or 3 different entrepreneurial acts. Model 2 is a logistic regression of an at-
torney’s decision to engage in legal entrepreneurship, but I replace the new justice indicator
with one that identifies if a new justice joined the Court over the last two terms. Finally,
Model 3 is also a logistic regression of an attorney’s decision to engage in legal entrepreneur-
ship, but I replace the new justice indicator with a variable that identifies when a justice
who moved the Court’s median joined it. Following the model in the paper in Table 4.1,
Models 2 and 3 have standard errors clustered by case.

95

Table B.1: Alternative Logistic Regression Results, Attorney Decision to Engage in Legal
Entrepreneurship

New Justice

New Justice In Last Two Terms

Model 1
−0.055
(0.029)

0.007
(0.004)

−0.099∗
0.150∗∗∗

(0.040)

Move the Median Justice

Public Mood

Ideological Congruence

Criminal Procedure Case

Civil Rights Case

First Amendment Case

Due Process Case

Privacy Case

Attorneys Case

Unions Case

Judicial Powers Case

Federalism Case

Federal Taxation Case

Latent Case Salience

Lower Court Conflict

Dissent Noted in Lower Court

Attorney Experience

U.S. Government Attorney

Attorney Status

Petitioning Attorney

Constant

Model 2

Model 3

-0.112
(0.076)

0.017
(0.011)
-0.322∗∗
(0.110)
0.294∗
(0.117)
0.100
(0.135)
0.078
(0.177)
0.191
(0.199)
0.295
(0.323)
-0.186
(0.286)
-0.574∗
(0.269)
0.465∗∗∗
(0.137)
-0.233
(0.181)
0.207
(0.245)
0.091
(0.054)
-0.190∗
(0.080)
0.188∗
(0.085)
0.098∗∗
(0.038)
0.019
(0.147)
-0.029
(0.016)
-0.093
(0.074)
-0.888
(0.703)
3018
4144.492
4270.751
2051.246

0.211
(0.143)
0.007
(0.012)

−0.331∗∗
0.306∗∗

(0.110)

(0.117)
0.109
(0.135)
0.071
(0.177)
0.191
(0.199)
0.316
(0.323)
−0.188
(0.286)
−0.584∗
0.457∗∗∗
(0.137)
−0.225
(0.181)
0.199
(0.245)
0.085
(0.054)

(0.269)

(0.080)

−0.195∗
0.193∗
0.099∗∗

(0.085)

(0.038)
0.014
(0.146)
−0.029
(0.016)
−0.093
(0.074)
−0.354
(0.737)

3018
4144.473
4270.733
−2051.237

(0.042)
0.055
(0.049)
0.075
(0.064)
0.058
(0.072)
0.119
(0.116)
0.005
(0.104)

(0.094)

−0.197∗
0.173∗∗∗
(0.049)
−0.075
(0.065)
0.004
(0.090)
0.021
(0.020)

−0.105∗∗∗
0.069∗
0.029∗

(0.029)

(0.031)

(0.014)
0.020
(0.053)
−0.009
(0.006)
−0.038
(0.027)
0.226
(0.252)

Observations
AIC
BIC
Log likelihood
Standard errors in parentheses
∗ p < 0.05, ∗∗ p < 0.01, ∗∗∗ p < 0.001

3018
6729.843
6862.115
−3342.922

96

BIBLIOGRAPHY

97

BIBLIOGRAPHY

Bailey, Michael A. and Forrest Maltzman. 2011. The Constrained Court: Law, Politics, and

the Decisions Justices Make. Princeton: Princeton University Press.

Baird, Vanessa A. 2004.

“The Effect of Politically Salient Decisions on the U.S. Supreme

Court’s Agenda.” Journal of Politics 66(3):755–772.

Bartels, Brandon L. 2009. “The Constraining Capacity of Legal Doctrine on the U.S. Supreme

Court.” American Political Science Review 103(3):474–495.

Baum, Lawrence. 2006. Judges and Their Audiences: A Perspective on Judicial Behavior.

Princeton: Princeton University Press.

Beim, Deborah, Alexander V. Hirsch and Jonathan P. Kastellec. 2015.

“Signaling and
Counter-Signaling in the Judicial Hierarchy: An Empirical Analysis of En Banc Review.”
American Journal of Political Science .

Benoit, Kenneth, Kohei Wantanabe, Paul Nulty, Adam Obeng, Haiyan Wang, Benjamin
Lauderdale and Will Lowe. 2017. “Quanteda: Qualitative Analysis of Textual Data.” R
Package .
URL: http://quanteda.io/

Biskupic, Joan. 2005. Sandra Day O’Connor: How the First Woman on the Supreme Court

Became Its Most Influential Justice. New York, NY: HarperCollins.

Biskupic, Joan. 2019. The Chief: The Life and Turbulent Times of Chief Justice John

Roberts. New York, NY: Basic Books.

Biskupic, Joan, Janet Roberts and John Shiffman. 2014. “The Echo Chamber: A small group

of lawayers and its outsized influence at the U.S. Supreme Court.” Reuters .
URL: https://www.reuters.com/investigates/special-report/scotus/

Black, Ryan C. and Christina L. Boyd. 2012. “U.S. Supreme Court Agenda Setting and the

Role of Litigant Status.” Journal of Law, Economics, and Organization 28(1):286–213.

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. and James F. Spriggs. 2013. “The Citation and Depreciation of U.S. Supreme

Court Precedent.” Journal of Empirical Legal Studies 10(2):325–358.

Black, Ryan C., Matthew E.K. Hall, Ryan J. Owens and Eve M. Ringsmuth. 2016. “The
Role of Emotional Language in Briefs Before the U.S. Supreme Court.” Journal of Law
and Courts 4(2):377–407.

98

Black, Ryan C. and Ryan J. Owens. 2009.

“Agenda Setting in the Supreme Court: The

Collision of Policy and Jurisprudence.” Journal of Politics 71(3):1062–1075.

Black, Ryan C. and Ryan J. Owens. 2012a. “A Built-In Advantage: The Office of the Solicitor

General and the U.S. Supreme Court.” Political Research Quarterly 66(2):454–466.

Black, Ryan C. and Ryan J. Owens. 2012b. “Looking Back to Move Forward: Quantifying
Policy Predictions in Political Decision Making.” American Journal of Political Science
56:802–816.

Black, Ryan C. and Ryan J. Owens. 2012c. The Solicitor General and the United States
Supreme Court: Executive Branch Influence and Judicial Decisions. Cambridge: Cam-
bridge University Press.

Black, Ryan C., Ryan J. Owens, Justin Wedeking and Patrick C. Wohlfarth. 2016. U.S.
Supreme Court Opinions and Their Audiences. New York: Cambridge University Press.

Black, Ryan C., Ryan J. Owens, Justin Wedeking and Patrick C. Wohlfarth. 2020. The
Conscientious Justice: How Supreme Court Justices’ Personalities Influence the Law, the
High Court, and the Constitution. Cambridge University Press.

Black, Ryan C., Sarah Treul, Timothy R. Johnson and Jerry Goldman. 2011. “Emotions,
Oral Arguments, and Supreme Court Decision Making.” Journal of Politics 73(2):572–581.

Black, Ryan C., Timothy R. Johnson and Justin Wedeking. 2012. Oral Arguments and
Coalition Formation on the U.S. Supreme Court: A Deliberate Dialogue. Ann Arbor, MI:
University of Michigan Press.

Bryan, Amanda C. and Ryan J. Owens. 2017.

“How Supreme Court Justices Supervise

Ideologically Distant States.” American Politics Research 45(3):435–456.

C-SPAN. 2009. “Supreme Court Term Review.” C-SPAN .

URL: http://www.c-span.org/video/?287321- 1/supreme-court-term-review

Caldeira, Gregory A. and John R. Wright. 1988. “Organized Interests and Agenda Setting

in the U.S. Supreme Court.” American Political Science Review 82(4):1109–1127.

Caldeira, Gregory A., John R. Wright and Christopher J.W. Zorn. 1999.

“Sophisticated
Voting and Gate-Kepping in the Supreme Court.” Journal of Law, Economics, and Orga-
nization 15(3):549–572.

Campbell, Amy Leigh. 2003. Raising the Bar: Ruth Bader Ginsburg and the ACLU Women’s

Rights Project. Xlibris.

Carpenter, Dale. 2012. Flagrant Conduct: The Story of Lawrence v. Texas. New York, NY:

WW Norton and Company.

Clark, Tom S. 2009. “The Separation of Powers, Court Curbing, and Judicial Legitimacy.”

American Journal of Political Science 53(4):971–989.

99

Clark, Tom S. 2019. The Supreme Court: An Analytic History of Constitutional Decision

Making. New York, NY: Cambridge University Press.

Clark, Tom S. and Benjamin E. Lauderdale. 2010. “Locating Supreme Court Opinions in

Doctrine Space.” American Journal of Political Science 54(4):871–890.

Clark, Tom S. and Benjamin E. Lauderdale. 2012. “The Genealogy of Law.” Political Analysis

20(3):329–350.

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 U.S. Supreme Court Litigation.” Law and Society Review 38(4):807–832.

Collins, Jr., Paul M. 2007.

“Lobbyists before the U.S. Supreme Court: Investigating the

Influence of Amicus Curiae Briefs.” Political Research Quarterly 60(1):55–70.

Collins, Paul M. 2008. Friends of the Court: Interest Groups and Judicial Decision Making.

New York: Oxford University Press.

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. and Justin Wedeking. 2014.

“The (Dis)Advantage of Certainty: The

Importance of Certainty in Language.” Law and Society Review 48(1):35–62.

Cushman, Barry. 1998. Rethinking the New Deal Court: The Structure of a Constitutional

Revolution. Oxford University Press.

DeHart, Jane Sherron. 2018. Ruth Bader Ginsburg: A Life. New York: Alfred A. Knopf.

Devins, Neal and Lawrence Baum. 2019. The Company They Keep: How Partisan Divisions

Came to the Supreme Court. New York, NY: Oxford University Press.

Epstein, Lee, Andrew D. Martin, Jeffrey A. Segal and Chad Westerland. 2007. “The Judicial

Common Space.” Journal of Law, Economics, and Organization 23(2):303–325.

Epstein, Lee and Jack Knight. 1998. The Choices Justices Make. Washington, DC: CQ

Press.

Epstein, Lee and Jack Knight. 2013. “Reconsidering Judicial Preferences.” Annual Review

of Political Science 16:11–31.

Epstein, Lee, Jeffrey A. Segal and Timothy R. Johnson. 1996. “The Claim of Issue Creation

on the U.S. Supreme Court.” American Political Science Review 90(4):845–852.

Epstein, Lee and Joseph F. Kobylka. 1992. The Supreme Court and Legal Change: Abortion

and the Death Penalty. Chapel Hill: University of North Carolina Press.

100

Epstein, Lee, William M. Landes and Richard A. Posner. 2013. The Behavior of Federal
Judges: A Theoretical and Empirical Study of Rational Choice. Cambridge: Harvard
University Press.

Feldman, Adam. 2016. “Counting on Quality: The Effect of Merits Brief Quality on Supreme

Court Opinion Content.” Denver Law Review 94(1):43–70.

Fowler, James H., Timothy R. Johnson, James F. Spriggs, Sangick Jeon and Paul J.
“Network Analysis and the Law: Measuring the Legal Importance of

Wahlbeck. 2007.
Precedents at the U.S. Supreme Court.” Political Analysis 15(3):324–346.

Frank, Nathaniel. 2017. Awakening: How Gays and Lesbians Brought Marriage Equality to

America. Cambridge, MA: Belknap Press.

Garner, Bryan A. 2003. The Winning Brief: 100 Tips for Persuasive Briefing in Trial and

Appellate Courts. 2nd ed. New York, New York: Oxford University Press.

Garner, Bryan A. and Anthony M. Kennedy. 2010. “Interviews with United States Supreme

Court Justices: Anthony M. Kennedy.” The Scribes Journal of Legal Writing 13:79–98.

Garner, Bryan A. and Antonin Scalia. 2010. “Interviews with United States Supreme Court

Justices: Antonin Scalia.” The Scribes Journal of Legal Writing 13(51-78).

Garner, Bryan A. and Clarence Thomas. 2010.

“Interviews with United States Supreme

Court Justices: Clarence Thomas.” The Scribes Journal of Legal Writing 13:99–132.

Garner, Bryan A. and John G. Roberts. 2010. “Interviews with United States Supreme Court

Justices: John G. Roberts Jr.” The Scribes Journal of Legal Writing 13:5–40.

Garner, Bryan A. and John Paul Stevens. 2010.

“Interviews with United States Supreme

Court Justices: John Paul Stevens.” The Scribes Journal of Legal Writing 13:41–50.

Garner, Bryan A. and Ruth Bader Ginsburg. 2010. “Interviews with United States Supreme
Court Justices: Ruth Bader Ginsburg.” The Scribes Journal of Legal Writing 13:133–144.

Gennaioli, Nicola and Andrei Shleifer. 2007. “The Evolution of Common Law.” Journal of

Political Economy 115(1):43–68.

Gillman, Howard. 1993. The Constitution Besieged: The Rise and Demise of Lochner Era

Police Powers Jurisprudence. Durham: Duke University Press.

Graetz, Michael J. and Linda Greenhouse. 2016. The Burger Court and the Rise of the

Judicial Right. New York: Simon and Schuster.

Grimmer, Justin and Brandon M. Stewart. 2013. “Text as Data: The Promise and Pitfalls
of Automatic Content Analysis Methods for Political Texts.” Political Analysis 21(3):267–
297.

101

Hall, Matthew E.K. 2018. What Justices Want: Goals and Personality on the U.S. Supreme

Court. Cambridge University Press.

Hamilton, Alexander. 2001. Federalist No. 80. In The Federalist Papers, ed. George W. Carey

and James McClellan. Indianapolis: Liberty Fund pp. 411–417.

Hanmer, Michael J. and Kerem Ozan Kalkan. 2013.

“Behind the Curve: Clarifying the
Best Approach to Calculating Predicted Probabilities and Marginal Effects from Limited
Dependent Variable Models.” American Journal of Political Science 57(1):263–277.

Hansford, Thomas G. and James F. Spriggs. 2006. The Politics of Precedent on the U.S.

Supreme Court. Princeton: Princeton University Press.

Haygood, Wil. 2016. Showdown: Thurgood Marshall and the Supreme Court Nomination

That Changed America. Vintage.

Hazelton, Morgan L.W., Rachel K. Hinkle and James F. Spriggs. 2019. “The Influence of
Unique Information in Briefs on Supreme Court Decision-Making.” Justice System Journal
40(2):126–157.

Hirshman, Linda. 2015. Sisters in Law: How Sandra Day O’Connor and Ruth Bader Gins-

burg Went to the Supreme Court and Changed the World. New York: HarperCollins.

Hitt, Matthew P. 2019. Inconsistency and Indecision in the United States Supreme Court.

Ann Arbor, MI: University of Michigan Press.

Hull, N.E.H. and Peter Charles Hoffer. 2001. Roe v. Wade: The Abortion Rights Controversy

in American History. Lawrence: University of Kansas Press.

Johnson, Timothy R. 2001.

“Information, Oral Arguments, and Supreme Court Decision

Making.” American Politics Research 29(4):331–351.

Johnson, Timothy R. 2004. Oral Arguments and Decision Making on the United States

Supreme Court. Albany: State University of New York Press.

Johnson, Timothy R., Paul J. Wahlbeck and James F. Spriggs. 2006. “The Influence of Oral
Argument Before the U.S. Supreme Court.” 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 Their Questions at Oral Ar-
guments in the U.S. Supreme Court?” Washington University Journal of Law and Policy
29:241–261.

Kagan, Robert A. 2003. Adversarial Legalism: The American Way of Law. Cambridge,

Masschusetts: Harvard University Press.

Kaplan, David A. 2018. The Most Dangerous Branch: Inside the Supreme Court’s Assault

on the Constitution. New York: Crown.

102

Kirkland and Ellis. 2019. “Paul D. Clement.” Official Biography .

URL: https://bit.ly/2zgiO6g

Krehbiel, Keith. 2007. “Supreme Court Appointments as a Move-the-Median Game.” Amer-

ican Journal of Political Science 51(2):231–240.

Kuhn, Max and Kjell Johnson. 2016. Applied Predictive Modeling. 5th ed. New York, New

York: Springer.

Lane, Elizabeth A. and Jessica A. Schoenherr. 2019.

“‘A Matter of Great Importance’:
Senator Preparation for Supreme Court Confirmation Hearings.” Arlen Specter Center for
Public Research Fellowship, Thomas Jefferson University .
URL: https://bit.ly/2KPbQxK

Lax, Jeffrey R. and Charles M. Cameron. 2007. “Bargaining and Opinion Assignment on the

U.S. Supreme Court.” Journal of Law, Economics, and Organization 23(2):276–302.

Lithwick, Dahlia. 2015. “Anthony Kennedy’s Right to Choose.” Slate .

URL: https://bit.ly/2Rbks2j

Long, J. Scott. 1997. Regression Models for Categorical and Limited Dependent Variables.
Advanced Quantitative Techniques in the Social Sciences Thousand Oaks, California:
Sage.

Maltzman, Forrest, James F. Spriggs and Paul J. Wahlbeck. 2000. Crafting Law on the

Supreme Court: The Collegial Game. New York: 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 10:134–153.

Mauro, Tony. 2019.

“Staring Down ‘Stare Decisis’: How to Ask SCOTUS to Overturn

Precedent.” The National Law Journal .
URL: https://bit.ly/2zk6TUN

McGuire, Kevin T. 1993. “Lawyers and the U.S. Supreme Court: The Washington Commu-

nity and Legal Elites.” American Journal of Political Science 37(2):365–390.

McGuire, Kevin T. 1995. “Repeat Players in the Supreme Court: The Role of Experienced

Lawyers in Litigation Success.” Journal of Politics 57(1):187–196.

McGuire, Kevin T. and Barbara Palmer. 1995. “Issue Fluidity on the U.S. Supreme Court.”

American Political Science Review 89(3):691–702.

Mezey, Susan Gluck. 2003. Elusive Equality: Women’s Rights, Public Policy, and the Law.

Boulder: Lynne Rienner Publishers.

Monroe, Burt L., Michael P. Colaresi and Kevin M. Quinn. 2008. “Fightin’ Words: Lexical
Feature Selection and Evaluation for Identifying the Content of Political Conflict.” Political
Analysis 16(4):372–403.

103

Nelson, Michael J. and Lee Epstein. 2019. “Lawyers with More Experience Obtain Better

Outcomes.” Working Paper .
URL: https://bit.ly/2HgzMpv

Niblett, Anthony, Richard A. Posner and Andrei Shleifer. 2010. “The Evolution of a Legal

Rule.” Journal of Legal Studies 39:325–358.

Perry, Barbara A. 2007. The Michigan Affirmative Action Cases. Lawrence: University of

Kansas Press.

Perry, H.W. 1991. Deciding to Decide: Agenda Setting in the United States Supreme Court.

Cambridge: Harvard University Press.

Rice, Douglas R. and Christopher Zorn. 2019. “Corpus-Based Dictionarties for Sentiment

Analysis of Specialized Vocabularies.” Political Science Research and Methods .

Richards, Mark J. and Herbert M. Kritzer. 2002. “Jurisprudential Regimes in Supreme Court

Decision Making.” American Political Science Review 96(2):305–320.

Rosen, Jeffrey. 1996. “The Agonizer.” The New Yorker pp. 82–90.

Rubino, Kathryn. 2016.

“$2000 An Hour Lawyers: That’s One Way To Fund Salary In-

creases.” Above the Law .
URL: https://bit.ly/2MsBo3C

Scalia, Anontin and Brian A. Garner. 2008. Making Your Case: The Art of Persuading

Judges. Saint Paul, Minnesota: Thomas/West.

Schoenherr, Jessica A. and Ryan C. Black. 2019a. “Friends with Benefits: Case Significance,
Amicus Curiae, and Agenda Setting on the U.S. Supreme Court.” International Review of
Law and Economics 58:43–53.

Schoenherr, Jessica A. and Ryan C. Black. 2019b. “The Use of Precedent in U.S. Supreme
Court Litigant Briefs”. In Elgar Research Handbook on Judicial Process, ed. Susan Sterrett
and Lee Demetrius Walker. Edward Elgar Publishing.

Segal, Jeffrey A. and Albert D. Cover. 1989.

“Ideological Values and the Votes of U.S.

Supreme Court Justices.” American Political Science Review 83(2):557–565.

Segal, Jeffrey A. and Harold J. Spaeth. 2002. The Supreme Court and the Attitudinal Model

Revisited. Cambridge: Cambridge University Press.

Spaeth, Harold J., Lee Epstein, Andrew D. Martin, Jeffrey A. Segal, Theodore J. Ruger
“2017 Supreme Court Database, Version 2017, Release 01.”

and Sara C. Benesh. 2017.
Washington University Law .

Spriggs, James F. and Thomas G. Hansford. 2000. “Measuring Legal Change: The Reliability

and Validity of Shepard’s Citations.” Political Research Quarterly 53(2):327–341.

104

Stimson, James A. 2018. Public Opinion in America: Moods, Cycles, and Swings. Second

ed. New York, NY: Perseus.

Thomas, Evan. 2019. First: Sandra Day O’Connor. New York, NY: Random House.

Toobin, Jeffrey. 2008. The Nine: Inside the Secret World of the Supreme Court. Anchor

Books.

Toobin, Jeffrey. 2012. The Oath: The Obama White House and the Supreme Court. New

York: Doubleday.

Totenberg, Nina. 2011. “Skip the Legalese And Keep It Short, Justices Say.” NPR .

http://www.npr.org/2011/06/13/137036622/skip-the-legalese-and-keep-it-short-

URL:
justices-say

Tribe, Laurence and Joshua Matz. 2014. Uncertain Justice: The Roberts Court and the

Constitution. New York: Henry Holt and Co.

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.” Journal of Politics 71(1):224–237.

Woodward, Bob and Scott Armstrong. 1979. The Brethren: Inside the Supreme Court. New

York: Simon and Schuster.

Yglesias, Matthew. 2019. “Public Support for Left-Wing Policymaking has Reached a 60-year

High.” Vox.com .
URL: https://bit.ly/2yHhlZw

105