THE INFLUENCE OF THE FEDERALIST SOCIETY ON JUDICAL POLITICS AND

Chapter One:

Introduction

On May 2, 2022, an early draft of a majority opinion for the Supreme Court case Dobbs

v. Jackson Women’s Health was leaked. The leak made national headlines for several reasons.

First, Dobbs v. Jackson Women’s health was a significant and politically salient case concerning

the issue of abortion. Specifically in question was whether the landmark Supreme Court decision

Roe v. Wade (1973), which guaranteed the right to abortion for women in the United States,

should be overturned. The decision in Dobbs v. Jackson Women’s Health, which was issued on

June 24, 2022, overturned Roe v. Wade and allowed states to impose abortion bans previously

deemed by the Courts to be unconstitutional. In the time between the decision issued in Dobbs v.

Jackson Women’s Health and the publication of this dissertation less than two months later,

abortion policies in over a dozen states will have become dramatically more restrictive, with

several states going so far as to implement bans on abortion even in cases of rape and incest.

In addition to the impact the decision will have on women’s access to abortion in the

United States, the decision also highlights the power that the Federalist Society has over the legal

system in the United States. While the Federalist Society takes no official positions on specific

policy issues, the official website says that “the Federalist Society for Law and Public Policy

Studies is a group of conservatives and libertarians dedicated to reforming the current legal

order,” which they see as “strongly dominated by a form of orthodox liberal ideology which

advocates a centralized and uniform society” (About Us, 2022).

Simply put, the purpose of the Federalist Society is to make the United States legal

system more conservative. Members of the network influence court cases through a wide variety

of means—as litigants, as advocates presenting oral arguments before the Supreme Court, as

amicus curiae that support conservative case outcomes, as judges, as law clerks for justices of the

Supreme Court, and as justices themselves. In fact, six of the nine justices of the Supreme Court

are associated with the Federalist Society. This means that their influence over the Supreme Court

is such that, on any given case, justices associated with the Federalist Society can form a majority

without the support of a single justice who is not associated with the network. Dobbs v. Jackson

Women’s Health was one such case: the author of the leaked draft, Justice Alito, is associated

with the Federalist Society—as are all four other justices who ultimately voted to overturn Roe v.

Wade. Chief Justice John Roberts, who concurred in judgement, is also associated with the

Federalist Society.

The news of the leaked was draft was also significant because it highlighted the

willingness of justices associated with the Federalist Society to overturn a landmark precedent on

a highly salient political issue, and to reject the policy position supported by most Americans. In

the months prior to the decision in Dobbs v. Jackson Women’s Health, roughly 50% thought

should be upheld, while only 30% thought it should be overturned (Hubbard 2022; Romano

2022). The willingness of the Supreme Court to go against public opinion on this highly salient

issue has important implications for the legitimacy of the judicial branch and the Supreme Court,

especially if there are other salient policy issues on which the Supreme Court’s justices are

prepared to make decisions that reject public opinion. Outcomes such as the one in Dobbs v.

Jackson Women’s Health may erode support for the Supreme Court among liberals and those who

support precedents that the justices associated with the Federalist Society eventually overturn.

The fact that these justices from the Federalist Society had the power and the inclination

to overturn a generally popular and highly salient precedent begs several important questions.

Where does their organizational strength come from? Aside from abortion, what are other policy

issues on which Federalist Society members have particularly conservative views? Is the

decision-making of judges from the Federalist Society driven by legal principles, such as a

preference for states’ rights, or by ideological concerns? And what impact can we expect the

Federalist Society to have on public perception of judicial legitimacy now that they have a

dominant position atop the judicial hierarchy in the United States? My dissertation explores these

questions about the influence of the Federalist Society.

In the second chapter of this dissertation, I ask: where does the Federalist Society’s

organizational strength come from? Law school student chapters provide essential opportunities

for young, conservative law students to network with other conservatives. Events hosted by

student chapters often feature notable conservative lawyers, judges, and professors who advocate

and spread legal arguments to support conservative policy positions. Thus, Federalist Society

student chapters play a central role in the socialization process for young, conservative attorneys.

Leveraging an original dataset on Federalist Society’s law school student chapters, I use factor

analysis to create an indicator of organizational strength for Federalist Society student chapters

and then evaluate the sources of the Federalist Society’s organizational strength. Conceptualizing

law students as grassroots-level actors and law professors as elite-level actors, this study finds

that the Federalist Society’s organizational strength is linked to both grassroots-level and elite-

level factors. At the grassroots level, Federalist Society student chapters are stronger when the

law school’s alumni are more conservative. Student chapters are also stronger when law school

students are more efficacious. These same factors—ideology and efficacy—also matter at the

elite level. When law professors are more conservative or more efficacious, Federalist Society

student chapters are also stronger. Additionally, conservative professors’ ideological grievances

influence student chapter strength: chapters are stronger when there is a larger ideological gap

between conservative law professors and the average law professor at the law school, or when

there is a larger ideological gap between conservative law professors and the law school’s alumni.

In the third chapter, I evaluate the influence of the Federalist Society on the voting

behavior of justices of the Supreme Court who are associated with the Federalist Society.

Drawing from numerous sources, I identify eight policy issues on which, due to the process of

group polarization, Federalist Society members are expected to be especially conservative. The

results largely confirm these expectations, as Federalist Society members are especially likely to

cast conservative votes on cases involving abortion, affirmative action, campaign spending,

desegregation, religion, sex discrimination, and the Voting Rights Act of 1965.

Although the Federalist Society does not officially claim to be in favor of states’ rights,

support for states’ rights to determine their own policies is strongly implied by the network’s

official website, which claims that the “strongly dominated by a form of orthodox liberal

ideology which advocates a centralized and uniform society” and that the Federalist Society “is a

group of conservatives and libertarians dedicated to reforming the current legal order.” The

implication, of course, is that if liberals prefer a centralized and uniform society, then

conservatives must prefer decentralization and policy heterogeneity, which can be achieved by

leaving states to make their own policies rather than enforcing a uniform policy across all 50

states. Thus, I also evaluate whether Federalist Society members are more likely than their peers

to vote in favor of the state government’s position when a state is involved in a case, or whether

their propensity to vote in favor of the state is contingent upon the ideological position being

advocated by the state.

The results suggest that justices of the Supreme Court who are associated with the

Federalist Society care more about casting conservative votes than about ruling in favor of the

state. Although Federalist Society members are more likely than their Republican-nominated

peers to cast pro-state votes when the state’s position in a case is conservative, they are more

likely to cast anti-state votes when the state’s position is liberal. These results indicate that, when

their legal preference for states’ rights and ideological preferences come into conflict, they are

more likely than their Republican-nominated peers to prioritize casting conservative votes over

voting in a manner consistent with their preference for states’ rights.

The fourth chapter asks, how does the Federalist Society influence public perception of

judicial legitimacy and public support for the courts? To answer this question, I employed a

vignette experiment in which respondents were presented with articles concerning a (then)

forthcoming United States Supreme Court decision, Dobbs v. Jackson Women’s Health. These

vignettes provided four descriptions of the expected majority coalition. The vignettes also

provided either general information about the Court and its recent history in abortion cases or

information about the number of justices of the Supreme Court who are associated with the

Federalist Society and information about the very conservative voting behavior of these justices

in past cases concerning abortion.

I evaluated whether these cues about the composition of the majority opinion or the

informational cues influenced individuals’ support for broadly targeted and narrowly targeted

court curbing measures. I found tentative support for the expectation that cues describing the

majority coalition as consisting of Republican-nominated justices, Federalist Society members, or

both will increase support for broadly targeted court curbing among liberals, but that does not

appear to be the case for conservatives. However, I do not find support for the other hypotheses

concerning the cues in this experiment. In the additional analyses for this study in which I

evaluate the roles of several ideological factors in shaping individuals’ support for broadly and

narrowly targeted court curbing measures, I present novel findings suggesting that the most

important ideological factor influencing support for broadly targeted court curbing is a

respondent’s degree of ideological polarization.

In sum, my dissertation focuses on the influence of the Federalist Society and expands

our knowledge of the network’s influence. To my knowledge, the study in Chapter Two is the

first to offer a quantitative assessment of both the grassroots-level and elite-level of any network

or social movement, and it provides important insights into where the Federalist Society’s

organizational strength comes from. Meanwhile, Chapter Three and Chapter Four, will be

worrying to those concerned with the Court’s legitimacy. These results indicate that the Federalist

Society may ultimately diminish the Court’s perceived legitimacy.

Chapter Two:

The Sources of the Federalist Society’s Organizational Strength

The Federalist Society for Law and Public Policy Studies is a network of conservative

and libertarian legal scholars, attorneys, and judges. The influence of the network and its

members on the United States’ legal system is exceptional. They act as the de facto gatekeeper in

the selection of judges under Republican presidential administrations—since 1990, every single

Supreme Court justice confirmed under a Republican administration has been affiliated with the

network. Their influence in the Supreme Court nomination process is so extensive that, today, six

of the nine justices on the Supreme Court are current or former members of the network. The

network’s members also act as prolific sources of legal arguments in favor of conservative policy

positions in legal cases, which they often provide through amicus briefs. However, their

influence begins in law schools, where the Federalist Society both exposes conservative law

students to legal arguments supporting conservative policy positions and integrates them into a

network that can be invaluable to their future careers as legal scholars, attorneys, and judges. The

purpose of this paper is to evaluate the sources of the Federalist Society’s organizational strength.

The Federalist Society network was conceived by three law school students who had

been friends as undergraduates at Yale University—Steven Calabresi, Lee Liberman Otis, and

David McIntosh (Kruse 2018). Calabresi would go on to attend Yale Law School, while

Liberman Otis and McIntosh enrolled at the University of Chicago Law School. Arriving at Yale

Law school in the fall of 1980, Calabresi quickly felt ideologically alienated among his law

school peers. The law school student body, he felt, was just as overwhelmingly liberal as the

undergraduate student body, and he perceived the law school faculty as even more liberal than the

faculty at Yale University. Feeling alienated based on his ideological views and believing that

conservative views were unwelcome in the law school, he sought to create a forum in which

conservative and libertarian law students could share their political views (Hicks 2006). By the

fall of 1981, the Federalist Society was an official student organization at Yale Law School, and

in January 1982 the organization hosted its first event, which was a debate over Roe v. Wade, 410

U.S. 113 (1973).

Meanwhile, at the University of Chicago Law School, Liberman Otis and McIntosh

experienced a similar ideological alienation to that experienced by Calabresi at Yale (Hicks

2006). Having remained in touch with Calabresi during law school, Liberman Otis and McIntosh

decided to create their own chapter of the Federalist Society at the University of Chicago Law

School. For their faculty advisor, they recruited Chicago professor Antonin Scalia who, at the

time, was visiting Stanford (Kruse 2018).

These two Federalist Society organizations decided to hold a symposium on conservative

legal thought in the spring of 1982. Through Scalia, Liberman and McIntosh learned of a similar

organization at Stanford—The Stanford Foundation for Law and Economic Policy—and recruited

them as a cosponsor of the event (Hicks 2006). Liberman also read an article referencing the

conservative Harvard Journal of Law and Public Policy, and reaching out to them, gained their

sponsorship and their agreement to publish the symposium’s proceedings. Thus, the four

cosponsors of the Federalist Society’s first annual symposium were Yale Federalist Society, the

Chicago Federalist Society, the Stanford Foundation for Law and Economic Policy, and the

Harvard Journal of Law and Public Policy (Hicks 2006).

Their plan for the first symposium was ambitious, and the event would ultimately cost

nearly $25,000. Here, too, Scalia provided useful support for the fledgling organization. He spoke

with the conservative Institute for Educational Affairs (IEA) about helping to fund the event and

encouraged Calbresi to reach out to IEA for financial support. Ultimately, IEA would cover most

of the cost of the first symposium (Kruse 2018).

The Federalist Society’s first symposium was wildly successful. Two hundred students

from over twenty law schools attended the two-day event, which was entitled "A Symposium on

Federalism: Legal and Political Ramifications" and which examined the foundations of

federalism, the politics of returning power to the states, and using the federal government to

achieve conservative goals, among other issues (Hicks 2006). The event featured a roster of

prominent judges (e.g.; Robert Bork and Richard Posner), professors (e.g.; Antonin Scalia and

Paul Bator), and practitioners (e.g.; Theodore Olson and Michael McConnell). The Harvard

JLPP published the symposium’s proceedings in a special issue and described the symposium as

“an exceptionally important event” that “proclaimed the virtues of individual freedom and limited

government.” (Hicks 2006).

Following the success of the first symposium, the founding students decided to hold a

second annual symposium the following year and to build a national organization because,

according to Lee Liberman Otis, “It seemed like the obvious thing to do. The thing was as much

demand-driven as anything” (Hicks 2006). Five months after the first symposium at Yale, the

Federalist Society for Law & Public Policy Studies was officially formed as a non-profit

corporation (Kruse 2018). Currently, the Federalist Society has chapters at all 204 ABA-

accredited law schools and, in the 2017-2018 academic year alone, these chapters held over 1,100

events with a total attendance of over 60,000 students (“About Us,” 2022).

The Federalist Society is undoubtedly a powerful network that exerts substantial

influence on law and politics. What is much less clear is where that power comes from. In this

paper, I focus on the sources of strength for law schools’ Federalist Society student chapters for

several reasons. First, choosing Federalist Society student chapters was a natural choice because

the organization was started by law students and at law schools. Second, prominent conservatives

within the legal field must think that law schools’ student chapters are important, as they often

serve as speakers at the events held by Federalist Society student chapters at law schools. Third,

the events held by these student chapters contribute to conservative law students’ legal

socialization and provide them with strong legal arguments for conservative policy positions, and

thus likely have a strong influence on conservative law students’ legal thinking. And fourth, these

student chapters provide young, conservative law students with their first exposure to the

Federalist Society and a network of like-minded peers who can become part of their professional

network.

In this paper, I aim to evaluate the sources strength for Federalist Society student

chapters. I begin by discussing the scant prior research within political science pertaining to the

organization. Then, drawing on theories from social movement research, I derive testable

hypotheses about the sources of the strength of the Federalist Society student chapters at law

schools in the United States. Next, I introduce a novel dataset on Federalist Society student

chapters at the top 100 law schools in the United States. Then, I employ factor analysis to

generate a measure of organizational strength for law school student chapters of the Federalist

Society.

Finally, I evaluate my hypotheses concerning the sources of organizational strength. I

find that organizational strength is linked both to characteristics of law school students at the

grassroots level and to characteristics of law school faculty at the “elite” level. At the grassroots

level, law schools with more conservative students have stronger Federalist Society student

chapters, as do law schools with more efficacious students. These factors also matter at the elite

level: law schools with more conservative faculty and more efficacious faculty have stronger

Federalist Society student chapters. It also appears that grievance, operationalized as the

ideological conflict between conservative law faculty and a law school’s student body, impacts

organizational strength.

This paper makes important contributions to both the fields of judicial politics and the

social movement literature. It is important to understand where the Federalist Society’s strength

comes from, as the network plays an influential role in judicial politics both now and for the

foreseeable future. To that end, this paper introduces an original dataset on law schools’

Federalist Society student chapters, creates statistically valid measures of organizational strength,

and provides the first-ever systematic quantitative assessment of the sources of organizational

strength. This study also contributes substantially to the academic discussion of the role of

grassroots vs. elite actors in social movements, as social movement research has historically

focused largely on the role of elite actors while paying less attention to the role of grassroots

members of an organization. The findings in this study suggest that both grassroots members of a

movement (in this case, law students) and elite actors (law professors) influence the strength of a

social movement organization (SMO).

Research on the Federalist Society

Much has been written about the Federalist Society by legal scholars. Works on the

Federalist Society from the legal field typically discuss the history of the network and provide

discussions of its members’ activities. For example, Teles (2008) provides a comprehensive

biographical history of the network, including numerous interviews from key members. In

another biographical account of the Federalist Society, Avery and McLaughlin (2013) provide

both an overview of the organization’s history and more comprehensive discussion of the various

policy issues in which key members of the Federalist Society have been deeply involved. Much

of the remaining work on the Federalist Society within the legal community effectively comes

from the Federalist Society itself: the Federalist Society-associated Harvard Journal of Law and

Public Policy has published countless articles on the Federalist Society, including the one

referenced in the introduction of this paper (Hicks 2006), which discusses the history and

formation of the network. While these books and articles from the legal community provide

valuable background information on the network, none of them provide systematic quantitative

analysis of the organization.

Although six of the nine seats on the Supreme Court are now occupied by members of the

Federalist Society, there is surprisingly little research on the behavior of Federalist Society

members within the field of political science. One important article, Scherer and Miller (2009),

shows that not only are Federalist Society members on the U.S. Circuit Courts more conservative

than judges nominated by Democratic presidents, but they are also more conservative even than

other judges who were nominated by Republican presidents. There are also certain types of cases

on which they vote especially conservatively. Even after controlling for ideology, Federalist

Society members on the U.S. Circuit Courts vote more conservatively than do other Republican-

nominated judges on cases involving searches and seizure and on cases involving states’ rights.

Scherer and Miller (2009) is the only published article that focuses on the voting behavior of

Federalist Society members at any level within the judicial system.

Hollis-Brusky (2015) offers a useful way to think about the conservative legal network:

she conceives of the Federalist Society as a Political Epistemic Network (PEN), whose purpose is

to provide and spread policy-relevant legal knowledge to its members. In a series of case studies,

she reveals the network’s deep embeddedness in Supreme Court decision-making. Members of

the network play key roles in Supreme Court decision-making through their roles as litigants,

amici, Supreme Court law clerks, and even as Supreme Court justices themselves. Hollis-

Brusky’s work and her conception of the PEN is particularly relevant to the current study.

Because, as she argues, a critical activity of the Federalist Society is spreading politically-relevant

legal knowledge among conservatives, my indicator of organizational strength incorporates a

measure of the extent to which student chapters facilitate the spread of politically-relevant legal

knowledge among conservative law students, operationalized as the total events hosted by the

student chapter. These events are the critical means by which chapters share politically-relevant

legal knowledge, and they typically involve either a speaker providing conservative legal

arguments to law students or a debate in which both conservative and liberal legal arguments are

presented.

The Federalist Society as a Social Movement

An important question about the Federalist Society is: what is the source of the Federalist

Society’s strength? While the organization has three divisions (for law students, lawyers, and

faculty), the lifeblood of the organization is its student chapters that are embedded in every ABA-

accredited law school in the United States. The student chapters are, after all, where the

organization began. The Student Division, in the organization’s own words, “fosters a network of

conservative and libertarian students eager to challenge the legal establishment as lawyers,

faculty, judges, and policymakers.” (https://fedsoc.org/divisions). Student chapters are the

essential component for the Federalist Society network, providing conservative and libertarian

aspiring attorneys with the legal network and politically-relevant legal knowledge that they will

use in their careers as lawyers, law school professors, judges, and politicians in the conservative

legal movement. Because the Federalist Society’s student chapters introduce conservative and

libertarian law students to both the arguments they will use to support conservative policy

positions and a network of likeminded members of the legal field with whom to coordinate their

legal activism, it is important to learn where the organizational strength of Federalist Society

student chapters comes from.

To derive theoretical expectations about the sources of student chapters’ organizational

strength, I turn to the literature on social movements. Examining the Federalist Society through

the lens of research on social movements is a fruitful approach, as the concept of the social

movement aptly describes the Federalist Society and its goals. A widely accepted account of

social movements describes them as “networks of informal interactions between a plurality of

individuals, groups and/or organizations, engaged in political or cultural conflicts, on the basis of

shared collective identities” (Diani 1992).

The Federalist Society consists of a series of networks of students at law schools and

attorneys in cities around the United States, and members of these networks are deeply involved

in advocating conservative policy positions, often on an informal basis (e.g., members often write

amicus curiae briefs or serve as attorneys in important court cases, but they do so on their own or

through an organization other than the Federalist Society). The Federalist Society and its

members undoubtedly are “engaged in political or cultural conflicts, on the basis of shared

collective identities.” Indeed, the inspiration for the conception of the Federalist Society was the

shared belief of its three co-founders—Steven Calabresi, Lee Liberman Otis, and David

McIntosh—that the legal academy and the legal profession more broadly was dominated by

liberals, that the legal profession was substantially more liberal than the public at large, and that

the success of conservative electoral politics at implementing their policy goals would require

reshaping the legal field to be amenable to arguments for conservative policy positions.

While the Federalist Society claims not to support any specific policy positions, even a

glimpse of the list of events held at its student chapters reveals that policy—and not just law—is

clearly important to the organization and its members. For instance, on the website FedSoc.org,

which lists events for many law schools’ Federalist Society chapters, the subdomain for the

Student Chapter at Yale Law School (where the Federalist Society was conceived and where its

first annual symposium was hosted) includes events entitled “Litigating Obamacare and

Boumediene,” “The Future of Judicial Conservatism,” “What Offenses Rise to the Level of Being

‘High Crimes and Misdemeanors?’, “Corporate Constitutional Rights Ten Years After Citizens

United,” “Beyond Economic Piety: How to Focus Policymakers on Sustaining a Healthy

Society,” during the 2020-2021 school year. Meanwhile, the University of Pennsylvania Student

Chapter, which hosted the annual National Student Symposium on March 19, 2021, has hosted

six other events in 2021, including “The Future of the ACA,” “Religious Education and School

Choice,” “Impeaching a Former President,” and “Religious Liberty in the Age of COVID-19.”

While these titles do pertain to legal issues, they also clearly invoke policy issues—including

President Obama’s Affordable Care Act, the impeachment(s) of Donald Trump, campaign

finance, conservative religious positions, marijuana policy, and other important political issues on

which conservatives often have strong preferences. These events provide a forum where

prominent conservative legal scholars can transmit strong legal arguments to support conservative

policy positions to receptive conservative law students, giving those students the tools to hone

their legal arguments in support of conservative policy outcomes.

Why People Protest

A large literature on social movements has stressed three broad sets of factors that

influence the emergence, development, and impact of social movements: the political

opportunities and constraints that the would-be movement faces, mobilizing structures, and

framing processes (McAdam 2017). However, this literature generally focuses on the how those

factors influence social movements, rather than the causes of individuals choosing to participate

in social movements.

Klandermans (1984) notes that social movement researchers “went too far in nearly

abandoning the social-psychological analyses of social movements” and made the case for

incorporating insights from psychology into the study of social movement participation. While

organization-centric perspectives persist as the major focus of much research on social

movements, a subfield focused on individual-level participation in social movements has emerged

over the last several decades. Borrowing from economic theory, Klandermans (2004) develops a

theory of social movement participation that incorporates the concepts of demand for a social

movement (among potential movement members), supply of social movement organizations, and

mobilization to bring demand and supply together.

This study focuses on demand-side causes of participation in a movement. Recent

research on the demand-side causes of participation focuses on the psychological factors of

individuals’ grievances, efficacy, identity, emotions, and embeddedness within a movement

network (Van Stekelenburg and Klandermans 2013; Van Stekelenburg, Klandermans, and

Walgrave 2018). In this study, I generate five testable hypotheses concerning the sources of

Federalist Society student chapter strength based on the roles of identity, efficacy, and grievances.

I turn to each of these next.

Group Identification

Group identification plays an important role in leading individuals to participate in a

social movement, and this has been illustrated in a variety of contexts. For example, women’s

collective identification as women made it more likely for women to participate in a women’s

movement (Kelly and Breinlinger 1995; 1996). Similarly, older people in Germany who

identified with the older people’s movement known as the Gray Panthers were more willing to

participate, and gay people in the United States who identified with the gay movement were also

more willing to participate (Simon et al. 1998).

Various studies have corroborated that identification with a group positively influences

willingness to participate, even after controlling for other factors (see Stürmer et al. 2003; Tindall

2004). Not only does identification with a group matter, but so too does the strength of that

identification (Kelly and Breinlinger 1996; Stryker, Owens, and White 2000). Politicization of

group identity influences willingness to participate, although findings are mixed. Some

researchers argue that the politicization of group identity amplifies the linkage between identity

and action (Stürmer and Simon 2004), while others argue that group identification must be

politicized in order to influence participation (van Stekelenburg and Klandermans 2010).

Regardless of which of these two perspectives is correct, in the context of the

conservative legal movement, the relevant identity—that is, as a conservative—is inherently

political. Where the conservative legal movement has a larger population of conservatives from

which to draw, it should be stronger. An important distinction that this study makes is delineation

between movement elites and grassroots members of the movement. Because professors are long-

term and quasi-institutional members of each law school’s student chapter and because they can

facilitate the organization’s activities through serving as legitimizers of the network and speakers

at events, they can be viewed as network elites. Meanwhile, law students are only at the law

school for three years and may not have as much long-term commitment, institutional knowledge,

or embeddedness in the network as professors. In the context of the conservative legal movement,

law students can be seen as grassroots members. I expect that both elites and grassroots members

influence the strength of a social movement.

At the grassroots level, because identity as a conservative will play an important role in

terms of motivating law students to join the Federalist Society, law schools with more

conservative student bodies should have stronger Federalist Society student chapters. Likewise, at

the elite level, law schools with more conservative professors should also have stronger Federalist

Society student chapters.

H1: Law schools with more conservative students should have stronger Federalist

Society student chapters.

H2: Law schools with more conservative professors should have stronger Federalist

Society student chapters.

Efficacy

Efficacy refers to whether an individual has the expectation that policies or conditions

can be changed through protest (Klandermans, Van der Toorn, and Van Stekelenburg 2008).

Much of the research concerning efficacy treats efficacy as an independent variable predicting

various forms of collective action. For instance, Klandermans (1997) shows that when people

believe that protest will address their grievances at an affordable cost, they are more likely to

participate. Furthermore, efficacy appears to be positively associated with low- and moderate-cost

collective action (Corcoran, Pettinicchio, and Young 2015). Even after controlling for other

individual-level determinants of participation (such as identity and emotion), efficacy appears to

be positively and significantly related to an individual’s propensity to participate in protests

(Chan 2016). While the literature pertaining to efficacy generally focuses on how efficacy

predicts behavior, it does provide some insight on the sources of efficacy.

Two studies are helpful here. Paulsen (1991) finds that academic success is a strong

predictor of an individual’s feelings of political efficacy, and a feeling of political efficacy

facilitates political participation, while Bowler and Donovan (2002) find that education is a strong

predictor of efficacy. Typically, studies that evaluate the effects of efficacy rely on measures

based on survey responses. Such data isn’t available here. However, since academic success

predicts efficacy, I can use indicators of academic success, or the quality of students and

professors, as measures to capture the average efficacy of students and professors at a law school.

This allows me to evaluate where efficacy matters: at the grassroots level, at the elite level, or at

both. I expect that law schools with higher-quality students will have stronger Federalist Society

student chapters because their law students at the grassroots level will be more efficacious.

Similarly, I expect that law schools with higher-quality faculty will also have stronger Federalist

Society student chapters because their faculty at the elite level are more efficacious.

H3: Law schools with more efficacious students should have stronger Federalist Society

student chapters.

H4: Law schools with more efficacious professors should have stronger Federalist

Society student chapters.

Grievance

Early work in social movements focused largely on relative deprivation and the role of

the perceived or relative severity of a group’s grievances (see Davies 1962, 1969; Smelser 1963;

Geschwender 1968; Gurr 1970). Although later researchers often claimed that the role of

grievance in influencing protest was either exaggerated or nonexistent (see McCarthy and Zald

1977; Tilly 1977; McAdam 1999), the role of grievance has been given more credence in recent

years (see Snow and Soule 2010; Simmons 2014; Almeida 2018). Indeed, Snow and Soule

(2010, p. 23) argue that “none of the various sets of conditions necessary for the emergence and

operation of social movements is more important that the generation of deeply felt shared

grievances.” While grievances have traditionally been framed as materially constituted,

grievances need not be material. Simmons (2014) argues that grievances can also be ideationally

constituted, while Swank et al. (2020) find that grievance in the form of heterosexist

discrimination makes college students more likely to engage in pro-LGBT protests.

In the context of the Federalist Society and its law school organizations, there are two

ways in which conservative law professors might experience relative deprivation and grievance

with respect to ideology. First, law professors might experience a feeling of grievance based on

the ideological differences between themselves and the student body at the law school. For

example, a conservative professor may observe that students are generally liberal and, feeling

deprived of a like-minded student body, choose to associate with and help strengthen the law

school’s Federalist Society student chapter. And second, they might experience ideological

grievance when comparing their own views to those of other law professors at the university

where they teach. For example, they might observe that other faculty are much more liberal than

themselves, and choose to associate with and help strengthen the student chapter in order to help

young conservative law students become successful in order to reduce the perceived lack of

conservatives among elites in the broader legal network. When there is a larger ideological gap

between a law school’s conservative faculty members and the law school’s average student,

conservative faculty should be more inclined to have feelings of relative deprivation that lead

them to become more active in the law school’s Federalist Society student chapter. Likewise,

when there is a larger ideological gap between a law school’s conservative faculty members and

the law school’s average professor, conservative faculty should also be more inclined to have

feelings of relative deprivation that lead them to become more active in the law school’s

Federalist Society student chapter. These theoretical explanations for the effects of grievance on

Federalist Society student chapter strength led to two hypotheses:

H5: As the ideological gap between a law school’s conservative faculty and the law

school’s average student increases, the strength of the law school’s Federalist Society

student chapter should increase.

H6: As the ideological gap between a law school’s conservative faculty and the law

school’s average law professor increases, the strength of the law school’s Federalist

Society student chapter should increase.

Research Design

Because there are no extant datasets on the characteristics of the Federalist Society

organization or on any of the individual Student Chapters, creating a measure of organizational

strength for the law school chapters required a substantial data collection effort. To make the

project more tractable, I focused my data collection efforts on the top 100 law schools in the US,

as judged by the US New and World Report rankings. Due to tied rankings, 101 law schools

were included in this list. I then collected data on Federalist Society student chapters, which I

hand-coded based on content analysis of a variety of sources. This data collection process took

place between April 1, 2020, and June 1, 2020, and was used to collect a wide variety of

information from Federalist Society student chapters. In total, I collected information on 73

variables. Of these, 33 variables indicated a type of officer position, with the other 40 variables

collecting other attributes of each chapter.

The data collection process involved a systematic, sequential series of online searches.

For each university, I began by searching the law school’s website for information on student

organizations. Most Federalist Society student chapters either had dedicated subdomains within

the law school website or appeared on lists of student organizations for the law school but did not

have their own subdomain in the law school’s website. Five law school student chapters had their

own standalone websites. Another 11 were not listed on the law school’s student organization

page, did not have their own subdomain, and did not have a standalone website. Additionally,

every law school’s student chapter had a subdomain listed on the Federalist Society’s official

website, fedsoc.org. These subdomains typically (but not always) list the names and positions of

officers for the law school’s student chapter.

I began by gathering information on the number of officers each Federalist Society

student chapter had. This information was often available on the law school website’s subdomain

for the Federalist Society student chapter or, less frequently, on the law school’s student

organizations page. In cases where neither of these two approaches worked, I was sometimes

able to find officer information on the chapter’s Facebook or Twitter page. When none of these

methods produced officer information, I relied on information from the fedsoc.org subdomain for

the Federalist Society student chapter. When information was available from multiple sources, I

relied on the source that provided the most comprehensive information. For instance, some

chapters’ Facebook pages or law school subdomains provided substantially more information on

officers than did the fedsoc.org subdomain, and in these cases I deferred to the Facebook page or

the subdomain.

From these sources, created the variables for 33 officer positions based on the types of

officer positions at each law school’s student chapter. Next, I created the variable Total Officers,

indicating the total number of officers listed for the student chapter. This value ranges from 0 to

33, with a mean of 5.45. Only two law schools’ Federalist Society student chapters had over 20

officers—Columbia Law School (with 21) and Yale Law School (33). The variable Total Officers

can roughly be seen as an indicator of both chapter size and level of organizational complexity.

After collecting information on the number and types of official positions for student

officers at the organization, I gathered numerous additional indicators of Federalist Society

student chapter strength. These measures were chosen because they provide information about the

student chapter’s size, level of organizational complexity, level of activity, and reputation, and

they were created by evaluating each student chapter’s social media presence, collecting

information on each chapter’s events, and collecting information about awards won by each

chapter.

First, I searched for student chapter social media accounts across five platforms:

Facebook, Twitter, Instagram, YouTube, and Soundcloud. Social media accounts for law schools’

student chapters were sometimes linked in the law school’s student organization page or the

Federalist Society subdomain on the law school website. When they were not linked, I searched

for them manually. For each law school, I collected the number of social media platforms and

created a count variable Platforms. The number of Platforms could theoretically range from 0 to

5, but the actual values for Platforms ranged from 0 to 4 since no school had all 5 platforms. The

mean for the variable Platforms was 1.68.

Next, I created the variable Total Followers, indicating the total number of followers the

student chapter had across all social media platforms. These values ranged from 0 (for student

chapters with no social media presence) to 2,309 in the case of Harvard Law School. The mean

value for Total Followers was 301.49. The variable Platforms is an indicator of organizational

complexity, while the variable Total Followers is an indicator of chapter size and level of

activity.

Because Hollis-Brusky (2015) argues that an important function of the Federalist Society

is to spread politically relevant legal knowledge among conservatives, my measure of

organizational strength should include an indicator of the extent to which Federalist Society

student chapters actively facilitate the spread of politically relevant legal knowledge among

conservative law students. At student chapters, the most critical way this information is spread is

through events hosted by student chapters. These events come largely in two forms: they most

frequently involve guests from the legal field—typically judges, law professors, or attorneys—

who are invited to speak about a specific topic, but they can also feature two (or more) guests

who debate a specific legal topic or policy issue while presenting liberal and conservative legal

arguments in favor of liberal and conservative policy positions. To capture the extent to which a

Federalist Society student chapter actively works to spread politically relevant knowledge among

its members, I generated a count variable Total Events, which indicates the number of events the

Federalist Society student chapter hosted between January 1, 2019, and April 1, 2020. This

information was available on fedsoc,org. On that website, each law school’s Federalist Society

student chapter has a subdomain with a list of events held by that chapter arranged in

chronological order, with the most recent events listed first. When an event had been scheduled

for the period between March 1, 2020, and April 1, 2020, but was cancelled due to the COVID-19

pandemic, I counted the event towards the law school’s Total Events. The values for Total Events

ranged from 0 to 40, with a mean of 7.62. Total Events is an indicator of the student chapter’s

level of activity. This measure does not distinguish between events in which speakers present

only conservative arguments and those in which a debate features both conservative and legal

arguments. I decided not to exclude debates from this indicator because although conservative

law students will hear liberal legal arguments at these events, they will still hear conservative

legal arguments.

Finally, I created two measures of each chapter’s reputation by counting the awards given

to each student chapter by the national organization and by counting the number of Joseph Story

Awards won by a law professor at each university. Unfortunately, information on the Federalist

Society student chapter awards is not kept on a single website or subdomain of the Federalist

Society. I gathered this information through extensive online searches for key terms, including

the names of the awards. Information was sometimes available on a law school’s Facebook page

when that law school was the recipient of an award. Other times, that information was available

as a press release or other article on the law school’s website. In all, I found 36 student chapter

awards given between 2011 and 2021. This variable, Student Chapter Award Total, ranges from

0 to 7, with a mean of 0.36. Information on the Joseph Story Award is available at the fedsoc.org

subdomain https://fedsoc.org/joseph-story-award. This award is given by the Federalist Society to

one law faculty member at one university each year and has been given out 26 times as of 2021. I

coded this variable based on where the law professor taught in the year that they won the award,

rather than where they teach now. The values for Story Award Total range from 0 to 4, with a

mean of 0.26. These two variables—Student Chapter Award Total and Story Award Total—are

indicators of student chapter reputation.

I merged this original dataset with data from four other sources. First, I merged it with

data from the Internet Legal Research Group, using their 2020 rankings (data available at

https://www.ilrg.com/rankings/law/). Specifically, I used this website’s information on each law

school’s undergraduate GPA scores (25th percentile, median, and 75th percentile), LSAT scores

(25th percentile, median, and 75th percentile), student: faculty ratio, and the number of volumes in

the law school’s library. Next, I merged the dataset with data on the political ideologies of alumni

at law schools (from Bonica et al. 2016) and data on the political ideologies of law professors at

each law school (from Bonica et al. 2018). The measure of alumni ideology is a proxy for a

measure of student ideology since those authors did not measure the ideologies of current law

students. While this measure is not perfect, it is reasonable to assume that the alumni ideology is

strongly associated with student ideology. Finally, I merged the dataset with data on the quality of

professors at 100 law schools (from Heald and Sichelman 2019). Their measure of professor

quality is based on two factors: the number of HEIN citations to papers published by each law

school’s professors and the number of SSRN downloads of those publications. For each law

school, the authors calculate Z-scores for both HEIN citations and SSRN downloads, and then

computes the average of those scores to generate the measure Average Z-score. This final

measure, therefore, incorporates components indicative of both faculty success in publishing and

the impact of their published work.

Chapter Strength

The dependent variable Chapter Strength is a measure of organizational strength, which

was created through exploratory factor analysis (EFA). Factor analysis has been employed

fruitfully as a data reduction method not only within political science but specifically within

judicial politics (see Ducat and Dudley 1987, Ostberg et al. 2002, Wedeking 2010; Black et al.

2016; Gibson and Nelson 2017). EPA is a useful method for my purposes because it allows for

the creation of a single variable based on multiple indicators of Chapter Strength.

The indicators used to generate the measure of Chapter Strength are as follows: (1) Total

Officers; (2) Total Followers; (3) Total Events; (4) Student Chapter Award Total; (5) Story

Award Total; and (6) Platforms. Table 2.1 provides summary statistics for these six indicators.

Table 2.1: Summary statistics for six indicators of Chapter Strength

Variable Observations Mean Std. Dev. Min Max

Total Officers 101 5.4455 4.7947 0 33

Total Followers 101 301.4851 364.8831 0 2209

Total Events 101 7.6238 8.6161 0 40

Student Chapter Award Total 101 0.3564 0.9755 0 7

Story Award Total 101 0.2574 0.6427 0 4

Platforms 101 1.6832 0.8937 0 4

Table 2.1 provides summary statistics for each of the six variables to be used in generating a

single measure of Chapter Strength.

These six variables were next subjected to factor analysis. Before estimating the factor

analysis routine, the first step for a researcher is data inspection to check for the violation of

statistical assumptions. Specifically, this entails checking for KMO, which tells us whether there

is sufficient variance within the dataset to identify latent factors and perform EFA. Howard

(2016) recommends that when using the KMO Measure of Sampling Adequacy—whose possible

values range between 0.00 and 1.00—researchers should seek values above 0.60 before

performing EFA. As Table A1 in the Appendix reports, the KMO Measure of Sampling

Adequacy for this sample is quite good, with each individual score above 0.81 and the overall

composite value at 0.8406.

After inspecting the data, the next step is to select the method of factor analysis.

Although the most used method of factor analysis is principal component analysis (PCA),

Howard (2016) recommends against using PCA for two major reasons. First, despite its wide use,

is not actually a form of factor analysis and it does not account for the structure of correlations

and instead only accounts for the variance of measured variables. As such, it is not a true form of

factor analysis (Howard 2016). Second, EFA methods account for both common and unique

variance within measured variables and the resultant factors are derived from common variance

alone, whereas PCA makes no distinction between common variance and unique variance,

leading to components that represent both common variance and unique variance. In other words,

PCA (unrealistically) assumes that the individual measures are without error. For these reasons, I

employ the iterated principal-factor method of factor analysis, which does not rely on the

unrealistic assumption that the individual measures are errorless. Results of this analysis are

reported in Table A2 in the Appendix. As it shows, the first factor has an eigenvalue of over 2.98,

while the highest eigenvalue for any other factor is less than 0.31.

Next, the researcher must determine a method of factor retention. Historically, the most

popular method has been to use the Kaiser criterion, according to which researchers should retain

all factors with eigenvalues above 1. Using the Kaiser criterion would lead to the retention of one

factor, as one factor has an eigenvalue of 2.98 and all others have eigenvalues below 1. This

method, however, has been denounced by numerous authors (see Costello and Osborne 2005;

Howard 2016) in favor of the visual scree plot (VSP). When using VSP analysis, researchers plot

each eigenvalue on a graph and identify a cutoff (or “elbow”) at which successive decreases in

eigenvalues become less noticeable. The VSP in Figure 2.1 indicates that there is one latent factor

that drives the variation in the six indicators of Chapter Strength. Velicer’s Minimum Average

Partial (MAP) test confirms that one factor should be retained.

Figure 2.1: Visual Scree Plot (VSP) Analysis for Chapter Strength Components

Figure 2.1 provides a Visual Scree Plot (VSP) for the latent factors identified through the iterated

principal-factor method of factor analysis. These results suggest that the measure Chapter

Strength has one latent factor.

The next decision for the researcher concerns the method of matrix rotation. This

decision is trivial since there is only one retained factor. Nonetheless, I follow the

recommendations of Howard (2016) and use a direct oblimin rotation with a delta of zero, which

is also called a direction quartimin. Finally, the research must evaluate the variables’ loading

factors to determine the extent to which each variable represents the factor or factors produced

through factor analysis. Following the recommendation of Howard (2016), I employ a primary

factor loading cutoff of 0.4. As Table 2.2 indicates, all six variables load well above the 0.4

cutoff. Factor loadings range from 0.5482 (for Story Award Total) to 0.8163 (for Total

Followers).

Table 2.2: Factor Loadings on Factor 1 (Chapter Strength)

Variable

Factor 1

Total Officers 0.7420

Total Followers

0.8163

Total Events 0.7220

Student Chapter Award Total

0.6788

Story Award Total

0.5482

Platforms

0.6168

Table 2.2 reports Factor Loadings for the six indicators used to generate Chapter Strength.

These analyses establish several important conclusions. First, the KMO Measure of

Sampling Adequacy indicates that the six variables used in this factor analysis satisfy the

assumptions of factor analysis. Second, factor analysis of the six indicators of chapter strength

clearly point to one latent factor driving these variables. Hereafter, that factor will be called

Chapter Strength. And third, all six variables load onto Chapter Strength sufficiently to merit

inclusion. Thus, the iterated principal-factor method of factor analysis provides a useful method

for collapsing multiple variables into a single, valid indicator of Chapter Strength. Table 2.3

reports the top 20 law schools in terms of Chapter Strength.

Table 2.3: Top 20 Law Schools by Chapter Strength

Law School Chapter Strength Law School Rank

University of Chicago 3.620 4

Harvard University 3.457 3

Yale University 3.059 1

University of Virginia 2.812 8

University of Pennsylvania 2.008 7

Columbia University 1.944 4

Brigham Young University 1.539 37

University of Kentucky 1.281 70

Georgetown University 1.277 14

Notre Dame University 1.271 22

Stanford University 1.126 2

Duke University 1.024 12

Northwestern University 0.928 9

George Washington University 0.913 23

Florida International University 0.823 90

George Mason University 0.779 42

Cornell University 0.745 13

University of Florida 0.725 24

Arizona State University 0.553 24

University of Minnesota 0.485 21

Table 2.3 list the top 20 law schools in terms of Federalist Society Chapter Strength.

Figure 2.2 graphs the relationship between law schools’ rank and Chapter Strength for all

law schools in the dataset and includes a locally weighted regression of the effect of rank on

Chapter Strength. Chapter Strength serves as the dependent variable for the study that follows.

As Figure 2.2 illustrates, high-ranking law schools typically relatively high values of Chapter

Strength, while lower-ranking law schools tend to have lower values of Chapter Strength.

Additionally, the curve of the locally weighted regression suggests that rank may have a greater

influence on Chapter Strength at high-ranking law schools that at lower-ranked law schools.

Figure 2.2: The Relationship between Law School Rank and Federalist Society Student

Chapter Strength

Figure 2.2 graphs the relationship between law school rank and Chapter Strength.

An indicator generated via factor analysis will have a mean value of 0, and thus the mean

value of Chapter Strength for the top 100 law schools is 0. Values range from -1.083 at the low

end (corresponding to the University of Oklahoma College of Law and the University of Hawai’i

at Mānoa William S. Richardson School of Law) to 3.620 at the high end (corresponding to the

University of Chicago Law School). The standard deviation in Chapter Strength for the top 100

law schools is 0.929.

There are eight key independent variables in this study. The first, Student Ideology, is

based on each law school’s alumni CFscores. CFscores stands for common-space campaign

finance scores. First developed by Bonica (2014), CFscores are similar to NOMINATE ideology

scores and Martin-Quinn ideology scores, which are often used as measures of legislative

ideology and judicial ideology, respectively. The crucial difference is that rather than being based

on voting behavior, CFscores are based on campaign contributions and work on the principle that

an individual who donates x amount of dollars to a Democrat is more liberal than another

individual who donates x amount of dollars to a Republican, and an individual who donates to

both Democrats and Republicans is more moderate than someone who donates only to candidates

from one party. CFscores take campaign contribution data and rescale them into a unidimensional

scale where negative scores indicate liberal ideology, and positive scores indicate conservative

ideology. Perhaps the greatest benefit of CFscores is that because they are based on campaign

contributions rather than voting behavior, CFscores can be used to evaluate the ideologies of

individuals who are not legislators or judges1. The measure Student Ideology, which is the mean

CFscores measure for a law school’s alumni, comes from Bonica et al. (2016), who graciously

provided me with the mean CFscore for each law school’s alumni in their dataset. The variable,

renamed as Student Ideology, is used to evaluate whether the average ideology of students

influences Federalist Society student Chapter Strength (H1).

I use two measures of professor ideology, which are both based on measures

from Bonica et al. (2018). These authors provided me with their measures of the mean law

professor ideology, the standard deviation for professor ideology, and the ideology for professors

at the 25th, 50th, and 75th percentiles in terms of conservatism. In the analyses that follow, I use

the mean law professor ideology at a university (Professor IdeologyMean) and the ideology of law

professors at the 75th percentile in terms of conservatism (Professor IdeologyP75) as indicators of

professor ideology. These measures are used to evaluate the influence of law professor ideology

on Chapter Strength (H2).

1 One downside of the CFscore measure is that for an individual to have a CFscore, they must first have

chosen to donate to a political candidate’s campaign. As such, this data is typically skewed towards older

and wealthier individuals.

Recall that efficacy is closely associated with academic success (Bowler and Donovan

2002). I use two indicators for law student efficacy, each of which either directly comes from the

ILRG data or is based on data from the ILRG. The first of these, LSATMedian, is a measure of the

median LSAT score for each law school’s students. The second measure, Entrance Scores, is also

based on the ILRG data, but it is a measure created through factor analysis combining six

components: the 25th percentile LSAT score for the law school (LSATLow), the median LSAT

score (LSATMedian), the 75th percentile LSAT score (LSATHigh), the 25th percentile

undergraduate GPA for the law school (GPALow), the median GPA (GPAMedian), and the 75th

percentile GPA (GPAHigh). The KMO Measure of Sampling Adequacy confirms the

appropriateness of using these measures for factor analysis, and the visual scree plot (VSP)

confirms that using these six factors for iterated principal-factor analysis reveals one latent factor.

I call this measure Entrance Scores. Results for the KMO Measure of Sampling Adequacy, a

visual scree plot for eigenvalues, and a table for primary factor loadings—all available in the

Appendix—reveal that there is sufficient variation in the dataset for the detection of latent factors,

that they load onto one factor, and that all six measures of undergraduate GPA and LSAT scores

load heavily onto that factor, Entrance Scores. Both LSATMedian and Entrance Scores are used

to evaluate the influence of student efficacy on Federalist Society student Chapter Strength (H3).

To evaluate the influence of law professor efficacy on Chapter Strength (H4), I use

measures from Heald and Sichelman (2019). Those authors created scores for professor quality

by based on an analysis of the published works of professors at 100 law schools (although some

of the top 100 law schools were not included in this analysis). As discussed earlier, their measure

has two components: HEIN Z-scores (indicating the publication success of the law school

faculty) and SSRN Z-scores (indicating the reputation of the law school faculty). Their final

measure, Average Z-score, is the average of the HEIN Z-score and the SSRN Z-score for each

law school’s professors. It should be noted here that although they generate measures for 100 law

schools, they do not generate measures for 11 of the 101 law schools included among top 100 law

schools according to US News and World Report (2021). For the purposes of this study, I rename

the variable Average Z-Score as Professor Quality.

There are two remaining key independent variables which are used to measure the

influence of professors’ ideological grievance on Chapter Strength. To evaluate conservative

professors’ ideological grievance with law students, I subtract the law school’s mean alumni

CFscore (Student Ideology) from the CFscore for the law school’s 75th percentile law professor in

terms of ideological conservatism (Professor IdeologyP75) such that Ideological ConflictStudents =

Professor IdeologyP75 – Student Ideology. This measure allows me to assess whether an increase

in the ideological gap (or ideological grievance) between conservative professors and students at

a law school predicts greater Chapter Strength. Ideological ConflictStudents is used to evaluate H5.

Alternatively, Chapter Strength may be driven by conservative professors’ ideological

grievances with their peers, rather than by their grievances with the student body. To evaluate the

influence of conservative professors’ ideological grievance with other professors at the law

school on Chapter Strength, I subtract the law school’s mean professor CFscore (Professor

IdeologyMean) from the CFscore for a professor at the 75th percentile in terms of conservatism at

that law school (Professor IdeologyP75) such that Ideological ConflictProfessors = Professor

IdeologyP75 – ProfessorIdeologyMean. Ideological ConflictProfessors allows me to assess whether an

increase in the ideological gap (or ideological grievance) between conservative professors and the

average professor at a law school predicts greater Chapter Strength. Ideological ConflictProfessors is

used to evaluate H6.

In addition to these key independent variables that are used to test the six major

hypotheses, I include a variety of control variables. To allay the potential concern that Chapter

Strength might be driven by the law school’s capacity for developing student organizations, I

generate the variable Law School Orgs, which indicates the number of student organizations at

each law school. Likewise, the number of journals at the law school is another indicator of the

law school’s capacity for developing organizations. The variable Journals is a count variable

indicating the number of law journals published by the law school. Law School Orgs and

Journals are valuable control variables because they provide indicators for the general level of

student engagement and activity at a law school. Without these variables, a law school whose

students are generally highly active and whose student organizations are generally highly

professionalized might appear to have particularly strong Federalist Society student chapters

when it may simply be the case that many or all student organizations at the law school are highly

active and professionalized.

The capacity for developing strong student organizations may also be linked to the

number of students at the law school. The US News (2021) rankings include the number of full-

time students at each law school, and I call this variable Full-Time Enrollment. Likewise, since

law schools often require student organizations to have faculty advisors and because for some

organizations (such as the Federalist Society) faculty may play an important role in generating

and maintaining organizational strength, I create a measure of the number of professors at the law

school based on Full-Time Enrollment and the Student: Faculty ratio (SFRatio), which is

available in the ILRG data, by dividing Full-Time Enrollment by SFRatio. Thus, the indicator

Professors = Full-Time Enrollment/SFRatio.

Another important consideration is the amount of physical and financial resources that a

law school provides to its students. I do not have a direct measure of either the law school’s

endowment or the buildings or physical space available at each law school, but I do have an

indirect measure of both financial resources and strength: Library, which is available in the ILRG

data, is a measure of the total number of volumes in each law school’s library. Because the

maintenance of a large library requires both financial resources and physical space, the variable

Library serves as an indirect indicator of financial and physical resources.

From a methodological standpoint (e.g., limited degrees of freedom, collinearity, etc.),

however, I cannot justify using all five of these indicators in the same model. Table A3 in the

Appendix reports the pairwise correlation coefficients for these five variables and reveals that

Full-Time Enrollment is highly correlated with Law School Orgs, Journals, and Professors, while

Journals is also highly correlated with Professors. To overcome this issue, I use iterated

principal-factor analysis to collapse Professors, Library, Full-Time Enrollment, Journals, and

Law School Orgs into a one-dimensional indicator Resources, which allows me to include all five

in my analysis while allaying potential concerns about collinearity. Validation for the measure

Resources may be found in the Appendix. Each model in the analysis that follows uses either

Resources or Full-Time Enrollment to control for the law school’s capacity for developing strong

student organizations.

It is also possible that differences in Chapter Strength might be influenced by whether the

law school is public or private. To control for this potential confounding factor, I create a binary

variable indicating whether the law school is public or private. Public is coded as 1 if the law

school is public, and 0 if it is private.

Summary statistics for the dependent variable Chapter Strength and the independent

variables discussed above are included in Table 2.42. Each of the analyses that follow use

ordinary least squares (OLS) models in which the dependent variable is Chapter Strength.

2 Data on professor ideology was missing for one law school (Texas A&M) and data on alumni ideology

was missing for seven law schools: Florida International University, Loyola Marymount, UNLV, UC-

Irivine, Texas A&M, Yeshiva, and Rutgers. Thus, while I collect variables for the 101 law schools ranked

in the top 100 by US News, regression results for student impact include only 94 law schools. For

Professor Quality, because Heald and Sichelman (2019) did not include all of the top 101 universities,

there is additional missing data for the University of South Carolina, St. Louis University, Lewis and Clark,

Drexel University, Northeastern University, the University of Louisville, the University of New

Hampshire, the University of New Hampshire, Louisiana State University, Florida International University,

and Texas A&M. The missing data is unlikely to introduce any substantial systematic bias in the results, as

none of the missing data comes from top 25 law schools, and it does not appear that missing data for

Professor Quality is systematically biased towards either conservative or liberal law school student bodies.

Table 2.4: Summary Statistics for Dependent Variable (Chapter Strength) and Independent

Variables

Variable Observations Mean Std. Dev Min Max

Chapter Strength

101

0.9

1.08

3.6

Student Ideology

0.51

0.36

1.1

0.8

Professor Ideology

Mean

100

0.85

0.3

1.29

0.6

Professor Ideology

P75

100

0.61

0.5

1.22

1.1

LSAT Median

100

161.16

4.9

153

173

Entrance Scores

100

1.00

1.7

2.51

Professor Quality

0.0

0.99

1.32

3.91

Ideological Conflict

Students

0.08

0.39

1.0

1.44

Ideological Conflict

Professors

100

0.46

0.07

1.70

Resources

100

0.97

1.1

3.79

Full

Time Enrollment

101

624.58

294.12

227

1788

Public

101

0.56

Table 2.4 provides summary statistics for the variables included in this study. Chapter Strength is

the dependent variable.

Before getting to the analysis, one research design issue must be addressed. A challenge

that complicates the evaluation of the sources of Chapter Strength is that several of the variables

used in this analysis are correlated with each other to such an extent that they cannot be used

simultaneously in the same model. For instance, LSAT Median, Entrance Scores, and Professor

Quality are all highly correlated with law school rank (with the absolute value of coefficients over

0.72 in each case), so no model can simultaneously include a measure of rank and either student

quality or professor quality. Similarly, correlation coefficients for Professor Quality and the two

measures of student quality (LSAT Median and Entrance Scores) are 0.82 and 0.80, respectively.

Thus, the effects of Professor Quality and student quality cannot be evaluated simultaneously. As

such, I must model the impacts of student ideology and student quality (H1 and H3) separately

from the impacts of professor ideology and professor quality (H2 and H4). Additionally, the

correlation coefficients between Professor IdeologyMean, Professor IdeologyP75, and Ideological

ConflictStudents are sufficiently high that I cannot evaluate the influence of professor ideology and

quality on Chapter Strength in a model that also evaluates the impact of professors’ ideological

grievance on Chapter Strength. For these reasons, the results that follow will be subdivided into

three parts.

Results

Student Impact

Table 2.5 reports the OLS regression results to assess the influence of Student Ideology

(H1) and student quality (H3) on Chapter Strength. Model 1 and Model 2 use LSATMedian as the

measure of student quality, whereas Model 3 and Model 4 use Entrance Scores. To control for the

law school’s capacity for developing student organizations, Models 1 and 3 use Full-Time

Enrollment, while Models 2 and 4 use Resources.

Table 2.5: OLS Regression Models Predicting Student Impact on Chapter Strength

Model 1

Model 2

Model

Model 4

Student Ideology

0.54*

0.5

0.4

0.50*

(0.22)

LSAT Median 0.12***

(0.0

)

0.12***

(0.0

)

Entrance Scores

0.57***

0.5

***

(0.0

)

(0.09)

Full

Time Enrollment

0.0

0.00

(0.00)

Resources

0.21*

(0.10)

0.24*

(0.

)

Public

0.00

0.0

0.06

(0.1

)

(0.15)

(0.1

)

(0.15)

Observations

Adjusted

0.4610

0.4708

0.4516

0.4649

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Models 1, 2, 3, and 4 evaluate H1 and H3.

The results indicate that student ideology is an important driver of Chapter Strength. In

all four models, Student Ideology is significant at p<0.05, and in all four the coefficient is

positive. These results support H1, that Federalist Society Chapter Strength is greater when a law

school has more conservative students. This is noteworthy given that ideology is significant even

in the presence of some important controls. In all four models, the measure of student quality is

also positive and significant and, in each model, the measure of student quality is significant at

p=0.001 regardless of whether LSATMedian or Entrance Scores is used. In Models 2 and 4, the

control variable Resources has a positive and statistically significant effect on Chapter Strength,

indicating that law schools with more resources may have greater Chapter Strength. On the other

hand, Full-Time Enrollment falls short of statistical significance in Models 1 and 3. Across all

four models, Public falls short of significance, suggesting that whether a law school is public or

private may not influence Chapter Strength.

Figure 2.3 graphs predictions for Chapter Strength at the full range of values for Student

Ideology and LSATMedian based on Model 2. The left panel graphs the influence of Student

Ideology, while the right panel graphs the influence of LSATMedian. The relationship between

Student Ideology and Chapter Strength merits illustration. At the median value for Student

Ideology, -0.5535, a law school’s Federalist Society student chapter is predicted to have a

Chapter Strength of 0.009, which would make its chapter the 33rd strongest in the dataset. If an

otherwise identical law school had a Student Ideology score of -0.118 (the 90th percentile in terms

of conservatism), it would be predicted to have a Chapter Strength of 0.265. This law school’s

Federalist Society student chapter would be ranked 24th in terms of Chapter Strength.

Figure 2.3: Student Impact: Influence of Student Ideology and Student Quality on

Federalist Society Student Chapter Strength

Figure 2.3 graphs the predicted effect of Student Ideology (H1) and student quality (H3) on

Chapter Strength based on Model 2. Model 2 uses LSATMedian as the measure of student

quality. Both graphs include predictions for the full range of values of Student Ideology and

LSATMedian at law schools ranked in the top 100 according to the 2021 US News rankings.

The graph in the right panel of Figure 2.3 provides support for the hypothesis that law

schools with more efficacious students will have greater Chapter Strength (H3). Among the law

schools in this dataset, the median LSATMedian is 161. A law school with a LSATMedian of 161

is predicted to have a Chapter Strength of -0.01, which would make its Federalist Society student

chapter the 34th strongest. If an otherwise identical law school had an LSATMedian of 169 (90th

percentile among law schools in the dataset), it would be predicted to have a Chapter Strength of

0.91, which would make its Federalist Society student chapter the 15th strongest.

Overall, I find strong support for the hypothesis that law schools with more conservative

alumni will have stronger Federalist Society student chapters (H1) and that law schools with more

efficacious students will also have stronger student chapters (H3). These results have important

implications for social movements, as they suggest that even in the case of an ideologically

oriented organization such at the Federalist Society, the aptitude of an organization’s members at

the grassroots level plays an important role in generating organizational strength.

Professor Impact

Table 2.6 reports four OLS regression models to evaluate the hypotheses that Chapter

Strength will be greater when law schools have more conservative professors (H2) and when they

have more efficacious professors (H4). In Models 5 and 6, the indicator for professor ideology is

Professor IdeologyMean, while in Models 7 and 8 the indicator for professor ideology is Professor

IdeologyP75. To control for a law school’s capacity for developing student organizations, Models

5 and 7 use Full-Time Enrollment while Models 6 and 8 use Resources. Measures of professor

ideology are significant at p=0.001 across all four models, as is Professor Quality.

Table 2.6: OLS Regression Models Predicting Professor Impact on Chapter Strength

Model 5

Model 6

Model 7

Model 8

Professor Ideology

Mean

0.7

0.8

(0.23)

Professor IdeologyP75

0.55***

(0.13)

0.57***

(0.13)

Professor Quality

0.6

***

0.5

***

0.64***

0.60***

(0.08)

(0.0

)

(0.08)

(0.0

)

Full

Time Enrollment

0.00

0.00*

(0.00)

Resources

0.20*

(0.09)

0.21*

(0.0

)

Public

0.00

0.0

0.01

(0.15)

(0.1504)

(0.1

)

Observations

Adjusted

0.5136

0.5185

0.5404

0.5468

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Models 1, 2, 3, and 4 evaluate H2 and H4.

In Models 5 and 6, Professor IdeologyMean is a positive and significant predictor of

Chapter Strength at p=0.001 while, in Models 7 and 8, Professor IdeologyP75 is positive and

significant at p=0.000. These results support for the hypothesis that Chapter Strength will be

greater at law schools with more conservative professors (H2). Meanwhile, Professor Quality is

positive and significant at p=0.000 in all four models, providing strong support for H4. As was

the case when evaluating the influence of law students at the grassroots level, this analysis

suggests that law schools’ Resources positively influences Chapter Strength, but that Full-Time

Enrollment and whether the law school is public or private do not.

Figure 2.4 graphs the results from Model 6. The left panel graphs the effect of Professor

IdeologyMean on Chapter Strength (H2). As with the grassroot-centered hypotheses involving

students, the impact of professor (elite) ideology and professor quality merits some exploration. A

law school whose mean professor ideology is the median value of -0.9366 is predicted to have a

Chapter Strength value of -0.009, which would make its Federalist Society student chapter the

33rd strongest. At the 90th percentile for CFscore, the prediction is 0.387. This chapter would be

the 23rd strongest.

Figure 2.4: Professor Impact: Influence of Professor Ideology and Professor Quality on

Federalist Society Student Chapter Strength

Figure 2.4 graphs the predicted effect of Professor IdeologyMean (H1) and Professor Quality (H3)

on Chapter Strength based on Model 6. Both graphs include predictions for the full range of

values of Professor IdeologyMean and Professor Quality at law schools ranked in the top 100

according to the 2021 US News rankings.

The right panel of Figure 2.4 graphs the effect of Professor Quality on Chapter Strength

(H4). A law school whose professors are at the median value of Professor Quality for the dataset

is predicted to have a Chapter Strength value of -0.077, which would make the chapter the 38th

strongest. An increase to the 90th percentile in Professor Quality predicts Chapter Strength of

0.834. This would make the law school’s Federalist Society student chapter the 15th strongest

overall.

Overall, Models 5 through 8 provide strong support for H2 and H4. They also indicate

that, as is the case at the grassroots level (e.g. law students), the aptitude of elites (in this case,

law professors) plays an important role in influencing Chapter Strength. Furthermore, as was the

case with the previous analysis of the impact of law students, the efficacy of law professors

appears to have a greater impact on Chapter Strength than does the average ideology of the law

school’s professors. One potential criticism of this finding is that the impact of the ideology of

conservative professors is downplayed in Model 6 because it relies on mean CFscores for a law

school rather than on an indicator for the ideologies of the law school’s more conservative

professors. However, analysis of the impact of school’s Professor IdeologyP75 (not shown) in

Model 7 and Model 8 produce the same results: professor efficacy appears to matter more than

ideology, whether Professor IdeologyP75 or Professor IdeologyMean is used.

Grievance Impact

Table 2.7 reports OLS regression models evaluating the influence of conservative

professors’ ideological grievance—both with respect to alumni of their law school and with

respect to their law professor at their own institution—on Chapter Strength. The results provide

strong support for the hypotheses that as the ideological gap between a law school’s conservative

faculty and the law school’s average almuni increases, Chapter Strength increases (H5), and that

as the ideological gap between a law school’s conservative faculty and the law school’s average

professor increases, Chapter Strength increases (H6).

Table 2.7: OLS Regression Models Predicting Impact of Conservative Professors’

Ideological Grievance on Chapter Strength

Model 9 Model 10 Model 11 Model 12

Ideological

Conflict

Students

0.53**

(0.1

)

0.52**

(0.1

)

Ideological

Conflict

Professors

0.87**

(0.2

)

0.90**

(0.2

)

Professor Quality

0.62***

0.6

***

0.64***

0.6

***

(0.08)

(0.09)

(0.08)

(0.0

)

Full

Time Enrollment

0.00

(0.00)

Resources

0.11

(0.09)

0.16

(0.0

)

Public

0.0

0.03

0.01

0.00

(0.1

)

(0.1

)

(0.15)

Observations

Adjusted

0.5443

0.5398

0.5163

0.5226

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Models 9 and 10 evaluate H5, while Models 11 and 12 evaluate H6.

Figure 2.5 graphs predictions for the influence of a law school’s relatively conservative

professors’ ideological conflict with the law school’s mean Student Ideology on Chapter

Strength, based on Model 10. In the graph, higher values for Ideological ConflictStudents indicate

that conservative professors are more conservative than the law school’s mean alum, whereas

negative values indicate that conservative professors are less conservative than the law school’s

mean alum. A law school with an Ideological ConflictStudents value at the median for all law

schools (0.13475) is predicted to have a Chapter Strength value of .062, which would make its

student chapter the 30th strongest. At the 90th percentile for Ideological ConflictStudent, a law school

is predicted to have a Chapter Strength value of 0.331. This law school’s chapter would be

ranked 23rd in Chapter Strength. The results support the hypothesis that when a law school’s

conservative professors are more ideologically distant from the law school’s average alum,

Chapter Strength should be greater (H5).

Figure 2.5: Grievance Impact: Effect of Conservative Professors’ Ideological Conflict with

Law School Alumni on Federalist Society Student Chapter Strength

Figure 2.5 graphs the predicted effect of Ideological Conflict (ProfessorIdeologyP75-StudentIdeologyMean) (H5)

on Chapter Strength based on Model 10. The graph includes predictions for the full range of

values of Ideological Conflict (ProfessorIdeologyP75-StudentIdeologyMean) at law schools ranked in the top 100

according to the 2021 US News rankings.

Figure 2.6 graphs predictions for the influence of a law school’s relatively conservative

professors’ ideological conflict with the law school’s average professor on Chapter Strength,

based on Model 12. When a law school has the median value for Ideological ConflictProfessors, it is

predicted to have a Chapter Strength of -0.014 (33rd overall). When an otherwise identical law

school has a value for at the 90th percentile, it is predicted to have a Chapter Strength of 0.371

(23rd overall). These results support the hypotheses that Chapter Strength should be greater when

a law school’s conservative professors are more ideologically distant from the law school’s mean

professor (H6).

Figure 2.6: Grievance Impact: Effect of Conservative Professors’ Ideological Conflict with

Average Law Professor Peer on Federalist Society Student Chapter Strength

Figure 2.6 graphs the predicted effect of Ideological Conflict (ProfessorIdeologyP75-ProfessorIdeologyMean) (H6)

on Chapter Strength based on Model 12. The graph includes predictions for the full range of

values of Ideological Conflict (ProfessorIdeologyP75-ProfessorIdeologyMean) at law schools ranked in the top 100

according to the 2021 US News rankings.

Overall, the ideological grievance of a law school’s conservative faculty predict Chapter

Strength, and two forms of ideological grievance impact Chapter Strength. The ideological

grievances of conservative faculty with respect the ideology of the law school’s students matter,

as do their ideological grievances with respect to the rest of the law school’s professors. Although

both of these forms of ideological grievance matter, I refrain from making any claims about

which form of ideological grievance is the stronger driver of Chapter Strength. While the

predicted Chapter Strength for a law school at the 90th percentile for Ideological ConflictProfessors is

slightly higher than the predicted Chapter Strength for a law school at the 90th percentile for

Ideological ConflictStudent, the difference is small and, in both cases, ideological grievance at the

90th percentile predicts a value of Chapter Strength that would rank as 23rd strongest overall.

Discussion and Conclusion

The results of this study indicate that both law students and professors play important

roles in the development of strong Federalist Society student chapters at law schools. Law

schools have greater Chapter Strength when they have more conservative students (H1) and when

they have more efficacious students (H3). Likewise, Chapter Strength is greater when a law

school has more conservative professors (H2) and when it has more efficacious professors (H4).

Finally, ideological conflict matters, too: Chapter Strength is greater when a law school’s

conservative professors are more ideologically distant from the law school’s mean alum (H5) and

when a law school’s conservative professors are more ideologically distant from the law school’s

mean professor (H6). Moreover, the results hold under multiple model specifications. These

findings are important to both the field of US judicial politics and to the US legal field because

the Federalist Society plays an important role in the judicial branch both now and for the

foreseeable future. Furthermore, this study provides the first-ever systematic quantitative

assessment of the strength of law school’s Federalist Society student chapters, as well as the

sources of Chapter Strength.

But this study makes several other important contributions. First, it provides an original

dataset on law schools’ Federalist Society student chapters. This is important for the field of

judicial politics because of how influential the Federalist Society has become within the US

judicial system, but the value of this dataset is interdisciplinary because it will be of interest to the

legal field and to sociologists interested in social movements. Second, this study is the first to

evaluate the organizational strength of any type of law school student organization, or to assess

the impact of student ideology, student efficacy, professor ideology, or professor efficacy on

organizational strength.

Third, and most importantly, this study contributes to the academic discussion of the role

of grassroots vs. elite actors in social movements. These findings suggest that both grassroots

members (in this case, law students) and elite actors (law professors) influence the strength of a

SMO. These findings are particularly instructive because rather than assessing how grassroots

members of an organization and the organization’s elites influence organizational strength, I have

effectively assessed the sources of organizational strength at over 80 organizations (totals

observations vary from 84 to 90, depending on the model and data availability). This gives the

findings greater weight as the results are aggregated across all chapters, rather than reflecting the

particular idiosyncrasies of one specific student chapter.

Finally, these findings will be instructive to social movement organizations and their

leaders, as the results highlight the important role of SMO organizers’ aptitudes, whether at the

grassroots level or at the elite level, even in the context of an ideologically driven SMO. If even a

highly conservative SMO such as a Federalist Society student chapter relies more on the aptitudes

of its members than on the ideologies of its members, then SMOs should focus on recruiting

high-quality grassroots members and elite allies, rather than focusing exclusively or primarily on

recruiting ideologically aligned members.

While this study’s contributions are numerous, there is much yet to learn. Future research

could evaluate whether different Federalist Society student chapters have different priorities in

terms of the types of policies they emphasize through their events. A second avenue of research

might assess whether stronger Federalist Society student chapters provide students with better

employment opportunities for law school graduates. Finally, future research could provide nuance

to our understanding of the role that students and professors play in generating organizational

strength at law schools or graduate schools by applying similar methods to other types of

organizations. While this study is by no means the last word on the sources of SMO strength, it

provides important insights into how organizational strength is influenced by the ideologies and

aptitudes of grassroots organizers and elite allies.

Chapter Three:

The Federalist Society’s Relationship with the Principles of Federalism and States’ Rights

What distinguishes the voting behavior of Federalist Society members from that of other

Republican-nominated justices? The network’s official website describes it as a group of

conservatives and libertarians focused on the promotion of legal principles such as the separation

of powers, federalism, limited government, and the rule of law. Members and allies of the

network often claim to have a jurisprudence focused on the “original public meaning” of the text

of the US Constitution and argue that their mode of constitutional interpretation is the “correct”

method (see Lawson 2012; Lawson 2013; Paulsen 2014). On the other hand, critics of the

“original public meaning” approach to adjudication claim that it is “little more than a lawyer's

version of a magician's parlor trick” (Cornell 2008) or, worse, “an ideal smokescreen behind

which judges may pursue their personal moral, political, or economic goals with relative

impunity” (Redish and Arnould 2011). Taking aim at the tension between ideological voting and

legal principles, this study addresses the central question of whether Federalist Society members

on the Supreme Court sincere and principled supporters of states’ rights or whether, as critics

suggest, that legal principle is a smokescreen behind which Federalist Society members hide their

ideological goals that are more conservative.

I begin with a brief overview of the jurisprudential philosophy espoused by Federalist

Society members. Then, after examining the very limited quantitative research on the behavior of

Federalist Society members in the federal court system, I discuss two competing explanations for

Federalist Society members’ especially conservative behavior: first, their behavior might reflect

ideology and group polarization; and second—as Federalist Society members would argue—their

behavior might be the result of their sincere adherence to legal principles.

I find support for the hypothesis based on the theory of group polarization: Federalist

Society members vote especially conservative on cases involving core policy issues for the

network. They also vote more conservatively than their Republican-nominated peers, even after

controlling for ideology, when one of the parties in a case is a state government—which fits with

both the ideological and “sincere federalism” explanations for their behavior. However, in two

key tests of their behavior when ideology and a preference for states’ rights come into conflict, a

Federalist Society member is substantially more likely to cast an anti-state vote than is an

ideologically identical Republican-nominated justice. These results undermine the notion that

justices from the Federalist Society sincerely apply the principles of federalism and states’ rights.

This study makes several important contributions. First, it offers the first systematic

quantitative analysis of the voting behavior of Federalist Society members on the Supreme Court,

which is particularly important given that Federalist Society members now hold a dominant role

on the Supreme Court and can be expected to hold a majority of Supreme Court seats for the

foreseeable future. Second, it has implications for the ongoing debate over whether Federalist

Society members are sincere supporters of the legal principles they purport to hold. This is

particularly important because Federalist Society members now play a central role in Supreme

Court decision-making. And third, it has implications for public confidence in the judicial system

and its broader legitimacy because sincere versus strategic behavior is an important factor

shaping public support for the courts.

The Federalist Society and Jurisprudence

In terms of judicial decision-making, Federalist Society members generally espouse

adherence to “originalism” and typically support one of two approaches to originalism: “original

intent theory” or “original meaning theory.” Original intent theory holds that judges should

interpret the constitution in a manner consistent with what was intended by those who drafted it,

while original meaning theory holds that judges should interpret the Constitution in a manner

consistent with how reasonable persons living at the time the Constitution was adopted would

have understood the original meaning of the text. Most originalists—including Clarence Thomas,

Amy Coney Barrett, and the late Antonin Scalia—favor the original meaning theory.

There are several other legal principles to which the Federalist Society and its members

generally claim to adhere. On the network’s official website, the “About Us” section

(https://fedsoc.org/about-us) claims, under the “Our Purpose” header, that the organization

believes in certain legal principles, among which are “separation of powers; federalism; limited,

constitutional government; and the rule of law in protecting individual freedom and traditional

values.” (“About Us,” 2021).

In addition to espousing these legal principles, the Federalist Society website also makes

several claims about the legal field and the Federalist Society’s purpose within the legal field.

First, it argues that the legal field is dominated by liberals, and that liberals prefer a centralized

and uniform society. Second, it describes itself as a group of conservatives and libertarians. And

third, it claims to be focused on promoting both the awareness of the legal principles mentioned

above and the application of them in practice (“About Us,” 2021).

Two important inferences can be made based on what the Federalist Society says about

itself. First, the organization’s claims about liberals imply that conservatives and libertarians

prefer a decentralized and heterogenous society. And second, its focus on federalism—which

refers to the distribution of power between the federal government and state government—

combined with its apparent preference for decentralization implies support for states’ rights rather

than allowing the federal government to dictate uniform legal policy across all 50 states.

While the connection between conservative and libertarian ideologies, decentralization,

federalism, and states’ rights is not fully explicit on the organization’s website, the inference of

this link is not only logical but also supported by the types of attorneys and professors who

frequently act as invited speakers to Federalist Society events. For example, libertarian law

professor Randy E. Barnett, who has written extensively in support of the original meaning theory

of constitutional interpretation, federalism, and states’ rights, has participated as a speaker in over

80 Federalist Society events since 2010.

While many Federalist Society professors, judges, activists, and attorneys are deeply

involved in sharing politically relevant legal knowledge with conservative law school students to

help them promote conservative policies, the training of young conservative attorneys is just one

facet of the network’s influence in the US legal system. Members have played important roles in

numerous important Supreme Court cases as litigants and through the filing of amicus curiae

briefs. For example, vice president of the Cato Institute Ilya Shapiro spoke at 26 Federalist

Society events in 2020 alone and has filed more than 400 amicus curiae briefs before the

Supreme Court in his career. Perhaps most significantly, the network—and especially the

organization’s co-chairman and former Executive Vice President, Leonard Leo—has acted as

gatekeeper in the selection of Republican Supreme Court justices for the last two decades

(O’Harrow and Boburg 2019; Zelizer 2018). Although the Federalist Society as an organization

appears to have legal preferences that should make their members who are judges behave

distinctly from other judges on cases involving states’ rights and despite the network’s dominant

position on the Supreme Court, systematic quantitative analysis of the behavior of Federalist

Society members is sparse.

Research on the Federalist Society

Legal scholars provide some insight on the history and goals of the Federalist Society, as

well as the activities of the more influential members of the organization. For example, Teles

(2008) provides a comprehensive biographical history of the network, including numerous

interviews from key members. In another biographical account of the network, Avery and

McLaughlin (2013) provide both an overview of their history and, importantly, a more

comprehensive discussion of the policy issues in which key members of the Federalist Society

have been deeply involved. Specifically, Avery and McLaughlin (2013) identify opposition to

affirmative action, opposition to limitations on campaign spending, opposition to sex

discrimination cases, and several other policy issues as important for the Federalist Society. Their

account is crucial for the analysis of whether Federalist Society members’ behavior reflects group

polarization or fidelity to their principles of federalism and states’ rights, as I draw, in large part,

from their discussion of the Federalist Society and its members’ involvement in court cases and

other activities to identify the “core issues” the network cares most about.

Hollis-Brusky (2015) introduces the concept of the Political Epistemic Network (PEN),

and argues convincingly that the Federalist Society is, first and foremost, a network whose

purpose is to share politically relevant legal knowledge among its conservative and libertarian

attorneys and judges. And, as we shall see, there are certain policy issues that members of the

organization have prioritized as important conservative policy issues about which to share

politically relevant legal knowledge.

Although six of the nine seats on the Supreme Court are now occupied by members of the

Federalist Society, there is surprisingly little quantitative research on the behavior of Federalist

Society members. One exception, Scherer and Miller (2009), shows that Federalist Society

members on the U.S. Circuit Courts vote more conservatively than do other Republican-

nominated judges in cases involving searches and seizures and states’ rights. While this study is

vital for understanding the organization, the dearth of research is remarkable given the

organization’s prominent place in American politics. To date, there is almost no systematic

analysis of the behavior of Federalist Society judges at the District Court level of the federal

judiciary, at the Supreme Court, or at any level of any state court system. Thus, we have almost

no understanding of whether the legal principles espoused by Federalist Society members match

their voting behavior.

The Voting Behavior of Federalist Society Justices

I evaluate two competing explanations for why Federalist Society members vote more

conservatively than their Republican-nominated peers. The first explanation centers on group

polarization, which might lead Federalist Society members to vote more conservatively than their

Republican-nominated peers on cases involving core policy issues for the Federalist Society. The

second explanation focuses on legal principles. While liberal court observers argue that Federalist

Society members use legal principles as a smokescreen for their conservative policy goals,

Federalist Society members claim that their behavior is driven by sincerely held legal principles.

If Federalist Society members are right, Federalist Society members’ prioritization of legal

principle should lead them to be more likely than their Republican-nominated peers to vote in

favor of the state’s position when voting in favor of a state government’s position requires them

to cast a liberal vote. On the other hand, if liberal court observers are right, then Federalist

Society members may be no more likely than their Republican-nominated peers to cast pro-state

votes when doing so requires them to cast a liberal vote.

Group Polarization.

Research in social and political psychology defines group polarization as the tendency for

members of a homogenous deliberating group to move toward a more extreme position in the

direction of members’ individual preferences (Sunstein 2002). Researchers have examined group

polarization in a variety of contexts. The effects of group discussion on attitudes, judgements,

jury decisions, perceptions of individuals, and ethical decisions are generally consistent with the

theory of group polarization (see Myers and Lamm 1976). Notably, the presence of like-minded

individuals who share a preference for one side of an issue is the crucial antecedent for group

polarization to occur (Baron 2005).

Group polarization is thought to work through three mechanisms: motivated reasoning,

the influence of persuasive arguments, and social comparison (Vinokur and Burnstein 1978;

Strandberg et al. 2019). First, humans are goal-oriented in their information processing, and they

tend to evaluate information with a bias towards reinforcing their preexisting views (Taber and

Lodge 2006; Nyhan and Reifler 2010). The second mechanism—the influence of persuasive

arguments—is the process by which individual members of a group hear the arguments proposed

by other members and adjust their positions to reflect the most convincing arguments presented

by other members (Isenberg 1986) Importantly, in the context of a relatively homogenous group,

strong arguments that conflict with individuals’ preexisting views are not only discounted due to

motivated reasoning but are also outnumbered by arguments that support their preexisting views

(Vinokur and Burnstein 1978; Sunstein 2009). Finally, social comparison is the process by which

members, after hearing what other members of the group believe, adjust their own position

towards the dominant position in order to be perceived favorably by other group members

(Brauer et al. 1995; Sunstein 2009).

There are several reasons why we should expect the behavior of Federalist Society

members to reflect group polarization. The first mechanism of group polarization—motivated

reasoning—should be activated because Federalist Society members are highly educated and

politically sophisticated, and because people who are politically sophisticated are more inclined

towards motivated reasoning (Tabor and Lodge 2006). The second mechanism of group

polarization—the influence of persuasive arguments—should also be activated because at

Federalist Society events, attendees are likely to hear especially strong legal arguments from

speakers on behalf of conservative policy outcomes. Federalist Society members should be

especially likely to be familiar with strong conservative arguments for what I will call “core

policy issues,” which are the types of policy issues in which Federalist Society members have

been deeply involved as attorneys and activists. The third mechanism of group polarization—

social comparison—should also be activated through the organization’s events, where relatively

moderate members are likely to update their views such that they are more conservative after

being exposed to the views of other members of the network.

Thus, the characteristics of the members of the network, the network’s ideological

composition, and the events hosted by the network all suggest that its members should be

susceptible to the process of group polarization. And I expect that because of the process of group

polarization, Federalist Society members will be especially likely to vote conservatively on cases

that involve core policy issues.

H1 – Group polarization and ideological voting: Federalist Society members will be

much more likely to vote conservatively on core policy issues.

While finding support for H1 would not provide sufficient evidence for the group

polarization argument, it does establish a necessary condition for that argument. And further

evidence will be gathered from some of the remaining hypotheses.

Federalism and States’ Rights.

The support of states’ rights is traditionally considered a conservative position (Sullivan

2006). Thus, if Federalist Society members are especially strong proponents of states’ rights

relative to their Republican-nominated peers, then they should also be expected to vote more

conservatively than their peers. While the Federalist Society’s does not explicitly claim to support

states’ rights, that support can be inferred based on what is said in the “About Us” section of their

website. They claim that the liberal legal orthodoxy prefers a centralized and homogenous

society, implying that, as conservatives and libertarians, they prefer a decentralized and

homogenous society. They also claim to support a “limited, constitutional government” and

federalism. Taken together, these facts imply that Federalist Society members generally support

states’ rights since federalism refers to the distribution of power between the federal government

and states. Furthermore, it also implies that granting states more authority to determine their own

legal and public policies would promote greater policy decentralization and heterogeneity.

Evidence of the organization’s support for states’ rights can be found in the fact that the

libertarian law professor Randy E. Barnett, who literally wrote the book on original meaning

theory (Barnett 2004) and who is an outspoken advocate for federalism and states’ rights, has

been an invited guest speaker at over 80 Federalist Society events since 2010. Because support

for states’ rights has traditionally been considered a conservative view and because, based on how

the Federalist Society characterizes itself on its official website, I hypothesize that Federalist

Society members should vote more conservatively than their Republican-nominated peers on

cases in which one party is a state.

H2 – Federalism and conservative voting: Federalist Society members will vote more

conservatively than other Republican-nominated justices in cases in which one party is a

state.

It is important to note that for the principles of federalism and states’ rights to explain the

especially conservative behavior of justices from the Federalist Society, it must be true that they

not only vote more conservatively on cases in which one party is a state, but that they are also

more likely to vote in favor of the state’s position in these cases. If they are more likely than their

Republican-nominated peers to vote conservatively but not more likely to vote in favor of the

state’s position, then this outcome would suggest that Federalist Society members use state

involvement in a case as a strategic means by which to justify a conservative outcome rather than

indicating that they truly prioritize federalism and states’ rights. Whether Federalist Society

members’ especially conservative voting behavior reflects a genuine prioritization of federalism

and states’ rights has been the subject of speculation, but that question has not yet been the

subject of a definitive systematic analysis. While the Federalist Society claims to adhere to the

principle of federalism and while there is a clear link between federalism, the preference for

decentralization, and states’ rights, liberal court observers have often questioned “how

suspiciously convenient it is that the jurisprudence advocated by society members so often yields

conservative results” and claim that “conservatives reverse-engineered their preferred

methodologies of interpreting the Constitution and legal statutes to reach the ends they want”

(Montgomery 2019). Other critics suggest that while Federalist Society members typically

espouse interpreting the US Constitution based on the “original meaning” of the text,

“conservatives are typically not so intent on following 'original meaning' in areas such as

affirmative action, executive powers, free speech and federalism” (Levin, DiSalvo, and Shapiro

2012).

I evaluate these competing views: on one hand, that Federalist Society members vote

especially conservatively when one of the parties in a Supreme Court case is a state government

and that they prioritize states’ autonomy in these cases; and on the other, that Federalist Society

members vote more conservatively when one party is a state, but not because they are more likely

than their peers to prioritize federalism and states’ rights.

If Federalist Society members are sincere supporters of the principles of federalism and

states’ rights, then they should be more likely than ideologically similar Republican-nominated

justices to vote in favor of the state government regardless of whether they are involved in the

case as the petitioner or the respondent and regardless of whether the lower court decision was

liberal or conservative. More explicitly, they should be more likely than ideologically similar

Republican-nominated peers to vote in favor of the petitioner when the petitioner is a state

government, and to be more likely to vote in favor of the respondent when the respondent is a

state government. And this should be true both when the lower court decision was conservative

(meaning that a vote in favor of the petitioner requires a liberal vote) and when the lower court

decision was liberal (meaning that a vote in favor of the petitioner requires a conservative vote).

This leads to the following hypothesis:

H3 – Sincere support for the principle of federalism: Regardless of the ideological

direction of the lower court decision, a Federalist Society member is more likely than an

ideologically similar Republican-nominated justice to vote in favor of the petitioner when the

petitioner is a state government and more likely to vote in favor of the respondent when the

respondent is a state government.

If, however, Federalist Society members are strategic supporters of federalism and states’

rights, then their behavior in a case should be contingent upon whether a vote in favor of the state

government requires a liberal or a conservative vote. Evaluating how Federalist Society members

behave when their ideological preferences are aligned with their legal preferences reveals little

about whether they are sincere or strategic supporters of the principles of federalism and states’

rights. For instance, when the lower court decision was conservative and the respondent is a state

government, a vote in favor of the respondent satisfies both their ideological preferences and their

purported legal preferences. Likewise, when the lower court decision was liberal and the

petitioner is a state government, both preferences are fulfilled by a vote in favor of the petitioner.

Testing whether differences in the behavior of Federalist Society members and

ideologically similar Republican-nominated justices is driven by ideological preferences or legal

preferences requires evaluating their behavior when their ideological preferences and their legal

preferences collide such that a pro-state vote requires a liberal vote. A pro-state vote requires a

liberal vote under two circumstances: when the lower court decision was liberal and the

respondent is a state government, and when the lower court decision was conservative and the

petitioner is a state government. If Federalist Society members’ purported support for federalism

and states’ rights is strategic rather than sincere, then they will not be more likely than

ideologically similar Republican-nominated justices to vote in favor of the state when their

ideological and legal preferences collide. This leads to two hypotheses concerning strategic

behavior:

H4a – Strategic support for the principle of federalism: In cases in which the lower court

decision was conservative and the petitioner is a state government, a Federalist Society member

is either less likely or no more likely than an ideologically similar Republican-nominated justice

to cast a pro-state vote in favor of the petitioner.

H4b – Strategic support for the principle of federalism: In cases in which the lower court

decision was liberal and the respondent is a state government, a Federalist Society member is

either less likely or no more likely than an ideologically similar Republican-nominated justice to

cast a pro-state vote in favor of the respondent.

Note that the sincere federalism hypothesis (H3) is mutually exclusive with both H4a and

H4b because the former predicts that Federalist Society members will be more likely to cast pro-

state votes than will ideologically similar Republican-nominated justices, while the latter two

hypotheses provide the expectation that, under certain circumstances, Federalist Society members

will not be more likely to cast pro-state votes. Determining whether H3 or H4a and H4b are

supported will provide useful insight into the behavior of Federalist Society justices and how they

prioritize legal and ideological concerns relative to other Republican-nominated justices which, in

turn, has important implications for the public perception of judicial legitimacy.

Research Design

I analyze Supreme Court justices who cast votes after the confirmation of Justice Scalia

in 1986. This ensures that when comparing votes, they come from only the time when Federalist

Society members have been on the Court. It also limits the confounding effect of time by not

artificially inflating the number of Republican-nominated justices who are non-Federalist Society

members. Table 3.1 provides the names of all justices who have served on the Court since

Antonin Scalia was confirmed in 1986.

Table 3.1: Supreme Court Justices since 1986

Federalist Society Members Other Republican Nominees Democratic Nominees

Antonin Scalia William J. Brennan, Jr. Byron White

Clarence Thomas Harry Blackmun Thurgood Marshall

John Roberts Lewis F. Powell, Jr. Ruth Bader Ginsburg

Samuel Alito William Rehnquist Stephen Breyer

Neil Gorsuch John Paul Stevens Sonia Sotomayor

Brett Kavanaugh Sandra Day O'Connor Elena Kagan

Amy Coney Barrett Anthony Kennedy

David Souter

The analysis in this paper consists of two studies due to the nature of the questions asked.

The first and second hypothesis concern the conditions under which Federalist Society members

are more likely than their peers to vote conservatively, and thus the dependent variable indicates

whether the justice voted in a conservative direction (coded as 1) or a liberal direction (coded 0).

H3, H4a, and H4b concern the identity of the party for whom each justice votes. Thus, the

dependent variable used to test these hypotheses indicates whether the justice voted for the

petitioner (coded as 1) or the respondent (coded as 0). Both Study One and Study Two draw from

the Supreme Court Database (Spaeth et al. 2020), hereafter referred to as the SCDB, using the

justice-centered data. In terms of the data, the major difference between Study One and Study

Two is that Study Two is restricted to cases in which one party is a state.

Study One

The dependent variable Vote Direction is a binary variable indicating whether the justice

voted in a conservative direction (coded as 1) or in a liberal direction (coded as 0). The key

independent variable Federalist Society is binary, coded 1 if the justice is a member of the

Federalist Society and 0 otherwise.

Core Issue is a binary variable indicating whether a case involved an issue identified as a

core policy issue for the Federalist Society. Core Issue is coded as 1 if the case involves such an

issue, and as 0 otherwise. The issues coded as 1 for Core Issue include abortion, affirmative

action, campaign spending, conservative Christian values, desegregation, searches and seizures,

sex discrimination, the Voting Rights Act of 1965. A Core Issue sometimes includes multiple

issues in the SCDB (e.g., there are different codes for different types of searches and seizures).

Section 1 of Appendix B explains how cases pertaining to a Core Issue were identified and

includes the SCDB issue codes for each Core Issue.

State Involvement is a binary variable coded as 1 if one of the parties in the case is a state

government, and 0 otherwise. Section 1 of Appendix B conveys information about how cases

were coded based on state involvement.

The variable Federalist Society is used in two separate interactions that assess how

Federalist Society members differ in terms of their voting behavior from other Republican-

nominated justices. The first of these interactions, which tests H1, includes Federalist Society

membership and Core Issue. The second interaction, which evaluates H2, includes Federalist

Society and State Involvement.

In addition to the two interactions used to test H1 and H2, several control variables are

included. To control for each justice’s ideology, I use Martin-Quinn ideology scores lagged by

one year (Martin and Quinn 2002). Because Martin-Quinn scores for a given year are based on

the votes cast in that year, using them to predict votes in the same year introduces circularity—

which can be especially problematic when predicting voting behavior across multiple issues

(Martin and Quinn 2005). Because I am predicting votes across multiple issues, I use Martin-

Quinn scores lagged by one year to control for ideology. As such, the control variable Ideology

for a justice in the 2018 term is the justice’s Martin-Quinn score for the 2017 term3. Chief Justice

is a binary variable, coded as 1 if the justice is Chief Justice and 0 otherwise. Appointing Year

indicates the year in which the justice joined the Supreme Court. Appointing Party is a binary

variable indicating whether the justice was nominated by a Republican president (coded as 1) or a

Democratic president (coded as 0). Finally, to mitigate the possible influence of ideological

polarization between parties over time, I control for fixed effects of the term of the Supreme

Court.

I use logistic regression models, clustered by justice, to evaluate H1 (Group polarization

and ideological voting) and H2 (Federalism and conservative voting). I cluster by justice because

votes cast by the same justice are not independent observations.

Study One Results:

The results of three clustered logistic regression models predicting the probability that a

justice will cast a conservative vote are shown in Table 3.2. Model 1 includes all justices, while

Model 2 only includes Republican-nominated justices. Model 2 is the primary model of interest

because this study is chiefly concerned with differences in the behavior between Federalist

Society members and other Republican-nominated justices. The Model 3 robustness check

excludes three justices—Justices Brennan, Stevens, Souter—who developed liberal reputations.

Importantly, the results from all three models are largely consistent.

3 Using Segal Cover scores (1989) returns very similar results.

Table 3.2: Clustered Logistic Regression Model Predicting Conservative Votes

Model 1

Model 2

Model 3

All Justices Republican-

nominated

Justices

Republican-

nominated

Justices,

Excluding

Liberals

Federalist Society

0.34**

0.3

(0.11)

(0.13)

(0.18)

Core Issue -0.36

(0.2

)

-0.27

(0.27)

0.09

(0.28)

Federalist Society * Core Issue

1.27***

1.16***

0.7

(0.23)

(0.

)

(0.31)

State Involvement

0.0

0.2

(0.1

)

(0.1

)

(0.1

)

Federalist Society * State

Involvement

0.78***

(0.1

)

0.73***

(0.16)

0.52**

(0.1

)

Ideology

0.29***

***

0.27***

(0.02)

(0.03)

(0.0

)

Chief Justice

0.11

0.1

0.10

(0.07)

(0.0

)

Appointing Year

0.00

0.01*

(0.00)

(0.0

)

Appointing Party

0.03

(0.0

)

Fixed Year Effects

Yes

Observations

36,846

27,354

20,815

Pseudo

0.0716

0.0693

0.0412

*p<.05, ** p<.01, *** p<.001. Entries represent clustered logistic regression coefficients with

standard errors in parentheses, clustered by justice. Dependent variable is a binary: 1 = justice

casts a conservative vote, 0 otherwise. Models 1, 2, and 3 evaluate H1 and H2.

Because evaluating H1 and H2 relies on the interpretation of interactive terms, which are

notoriously difficult to interpret (Franzese, Kam, and Jamal 2001), I present the results of Model

2 from Table 2 graphically. The left panel of Figure 3.1 provides predictions for conservative

votes for both Federalist Society members and other Republican-nominated justices, both on

cases that involve a Core Issue and on those that do not. The right panel shows the predictions

both when a state is involved in the case, and when no state is involved in the case.

Figure 3.1: Adjusted Predictions for Voting Behavior Based on Issue and State Involvement

Figure 3.1 graphs predictions for the probability of a conservative vote for Federalist Society

members and other Republican-nominated justices. The left panel reports adjusted predictions

based on whether a case involves a Core Issue for the Federalist Society, and the right panel

reports adjusted predictions based on State Involvement in the case. These results are based on

Model 2 in Table 3.2.

The graphical results in the left panel of Figure 3.1 indicate that in cases involving a Core

Issue, the prediction for the probability of a conservative vote for Federalist Society members is

.71, with and a 95% confidence interval of [.68, .74]. Meanwhile, the predicted probability of a

conservative vote for other Republican-nominated in these cases is .51, with a confidence interval

of [.38, .64].

Figure 3.1 reveals two important points about voting on cases involving a Core Issue.

First, Federalist Society members appear more likely than their Republican-nominated peers to

vote conservatively in cases that involve a Core Issue. Second, while Federalist Society members

are substantially more likely to vote conservatively when a case involves a Core Issue than when

it does not, the difference in the behavior of other Republican-nominated justices based on the

involvement of a Core Issue is negligible.

The right panel of Figure 3.1 graphs predictions for both groups of justices based on

whether a state is involved in a case. When a state is involved is a case a Federalist Society

member is expected to vote conservatively with a probability of .67 and with a 95% confidence

interval of [.64, .70], while other Republican-nominated justices are expected to vote

conservatively with a probability of .57 and a 95% confidence interval of [.49, .64]. As was the

case with Core Issue, the behavior of Federalist Society members diverges substantially based on

State Involvement, whereas the behavior of other Republican-nominated justices does not.

Figure 3.2 graphs the critical test of the average marginal effect of Federalist Society

membership on casting a conservative vote based on whether a case involves a Core Issue and

State Involvement. The left panel, which graphs the average marginal effect of Federalist Society

membership on cases involving a Core Issue, provides support for H1. For cases involving a Core

Issue, Federalist Society membership predicts a roughly 20% increase in the probability of a

conservative vote relative to other Republican-nominated justices, even after controlling for

justice ideology.

Figure 3.2: Marginal Effect of Federalist Society Membership on Voting Behavior Based on

Issue and State Involvement

Figure 3.2 graphs the average marginal effect of Federalist Society membership on conservative

voting based on whether a case involves a Core Issue (left panel) and based on whether a state is

involved in a case (right panel). These results are based on Model 2 in Table 3.2. The left panel

evaluates H1, while the right panel evaluates H2.

The right panel of Figure 3.2, which graphs the average marginal effect of Federalist

Society membership in cases with and without State Involvement, provides the critical test of H2.

Federalist Society membership predicts a 10% increase in the probability of a conservative vote

relative to other Republican-nominated justices, even after controlling for justice ideology4. Thus,

H2—that Federalist Society members will vote especially conservatively when a state is involved

in a case—is supported. While Federalist Society members are more likely to vote conservatively

when a state is involved in a case, what is not yet known is whether they vote more

conservatively due to ideology or sincerely held legal principles. Study Two will answer this

4In the robustness check Model 3, included in Section 2 of the Appendix, the results still hold, although

they weaken slightly (p<.07) for cases involving a Core Issue.

question by examining how Federalist Society members behave when political ideology and legal

principle collide, and a pro-state vote requires a liberal vote.

Study Two

The dependent variable Vote For Petitioner is binary and indicates whether the justice

voted in favor of the petitioner (coded as 1) or the respondent (coded as 0). As before, the key

independent variable is the binary variable Federalist Society. To test H3 and the conditional

hypotheses in H4, Federalist Society is used in a three-way interaction that assesses how

Federalist Society members differ from other Republican-nominated justices in terms of their

voting behavior. This interaction includes Federalist Society, Petitioner Is State Gov, and Status

Quo. Petitioner Is State Gov is a binary variable indicating whether the state government is the

petitioner (coded as 1) or the respondent (coded as 0), while Status Quo indicates whether the

ideological direction of the lower court decision was conservative (coded as 1) or liberal (coded

as 0).5

I include the same variables as in Study One—Core Issue, Ideology, Chief Justice,

Appointing Year, Appointing Party, and fixed effects of term of the Supreme Court—but include

three others as well. Prior research has shown that the resources available to a litigant influence

case outcomes (see Sheehan, Mishler, and Songer 1992; Collins 2004 and 2007). Thus, I use the

coding scheme from Sheehan, Mishler, and Songer (1992), which assigns litigants numerical

values ranging from 1 (for poor individuals) to 10 (US government) to control for the relative

strength of the parties involved. Petitioner Resources is the measure of litigant strength, from 1-

10, assigned to the petitioner in each case. Meanwhile, Respondent Resources is the measure of

the respondent’s strength. Finally, I also include the variable State Ideology, which is drawn from

5 Because cases in which both parties are states are not instructive in terms of evaluating Federalist Society

members’ support for federalism, I exclude the eight Supreme Court cases in the dataset in which both

parties are states.

the 2018 update of the Berry et al. (1998) measures of state government ideology. I include this

control variable because the ideology of a state government might influence justices’ likelihood

of voting in favor of the state.

Study Two uses logistic regression models, clustered by justice, to evaluate whether

Federalist Society members’ support for the principles of federalism is sincere (H3) or strategic

(H4a and H4b). While the three-way interaction between Federalist Society, Petitioner Is State

Gov, and Status Quo is more complex than some other possible approaches (such as a two-way

interaction between Federalist Society and State Ideology), the three-way interaction used in this

study is necessary because it provides a more direct and case-specific measure of whether the

position advocated by the state is liberal or conservative. As with Study One, votes are clustered

by justice because votes cast by the same justice are not independent observations. Because each

hypothesis concerns how justices behave in cases in which one party is a state, the analysis is

restricted to cases in which one party is a state government, whereas Study One included all other

cases as well.

Study Two Results

Table 3.3 reports three logistic regression models, clustered by justice, that predict

whether the justice will vote for the petitioner or the respondent. Model 1 includes all justices.

Model 2 only includes Republican-nominated justices and is the primary model of interest. As

with Study One, Model 3 provides a robustness check that excludes Justices Brennan, Stevens,

and Souter.

Table 3.3: Clustered Logistic Regression Model Predicting Votes in Favor of the Petitioner

Model 1

Model 2

Model 3

All Justices Republican-

nominated

Justices

Republican-

nominated

Justices,

Excluding

Liberals

Federalist Society

0.5

***

0.56**

0.61**

(0.1

)

(0.1

)

(0.22)

Petitioner Is State Gov -0.89***

(0.2

)

-0.91**

(0.27)

-0.67

(0.37)

Federalist Society * Petitioner Is State Gov

0.70*

0.6

0.03

(0.2

)

(0.3

)

(0.2

)

Status Quo 0.50

(0.3

)

0.41

(0.46)

-0.12

(0.4

)

Federalist Society * Status Quo -2.40***

(0.37)

-2.24***

(0.

)

-1.64***

(0.4

)

Petitioner Is State Gov * Status Quo

-0.14

(0.22)

-0.29

(0.2

)

-0.64

(0.36)

Federalist Society * Petitioner Is State Gov

* Status Quo

0.34

(0.27)

0.31

(0.29)

0.72

(0.37)

Core Issue

0.0

0.12

(0.08)

(0.09)

(0.11)

Petitioner Resources

0.08

0.15***

0.20***

(0.04)

(0.0

)

Respondent Resources

0.1

0.0

0.1

(0.0

)

(0.0

)

(0.05)

State Ideology

0.0

0.01**

(0.00)

Ideology

0.02*

0.0

***

(0.01)

(0.0

)

(0.0

)

Chief Justice

0.1

0.13*

(0.07)

(0.08)

(0.0

)

Appointing Year

0.00

(0.00)

Appointing Party

0.13*

(0.05)

Fixed Year Effects

Yes

Observations

6,792

5,132

3,850

Pseudo

0.0910

0.1090

0.1755

*p<.05, ** p<.01, *** p<.001. Entries represent clustered logistic regression coefficients with

standard errors in parentheses, clustered by justice. Dependent variable is binary: 1=justice casts a

vote in favor of the petitioner, 0 otherwise. Models 1, 2, and 3 evaluate H3, H4a, and H4b.

The left panel of Figure 3.3 graphs predictions when a state is the petitioner in a case

before the Supreme Court. The predictions for Federalist Society members vary dramatically

based on whether the lower court decision was liberal or conservative. When the lower court

decision was liberal—and thus Federalist Society members’ ideological and legal preferences are

aligned—Federalist Society members are predicted to cast conservative, pro-state votes in favor

of the petitioner with a probability of .73 and a 95% confidence interval of [.67, .80]. But when

the lower court decision was conservative and the petitioner is a state government, the predicted

probability of a liberal, pro-state vote in favor of the petitioner falls to .36, with 95% confidence

interval of [.25, .47]. In other words, when the petitioner is a state government and the ideological

and legal preferences of Federalist Society members collide such that a pro-state vote is a liberal

vote, Federalist Society members are expected to cast anti-state votes 64% of the time.

Figure 3.3: Adjusted Predictions for Voting for Petitioner Based on State Role in Case and

LC Decision Direction

Figure 3.3 graphs predictions for voting in favor of the petitioner for Federalist Society members

and other Republican-nominated justices. The left panel reports adjusted predictions when the

state is the petitioner, and the right panel reports adjusted predictions when the state is the

respondent. These results are based on Model 2 in Table 3.3.

The right panel of Figure 3.3 reports predictions when a state government is the

respondent. Here again, the ideological direction of the lower court decision plays a much larger

role in influencing the likelihood that Federalist Society members will cast pro-state votes than it

does for other Republican-nominated justices. When the lower court decision was conservative

and the respondent is a state government—and thus Federalist Society members’ ideological and

legal preferences are aligned—the probability of a vote in favor of the petitioner is .42 with a

confidence interval of [.35, .50]. In other words, the probability of a pro-state vote in favor of the

respondent is .58. On the other hand, when Federalist Society members’ ideological and legal

preferences collide and the state is the respondent in a case with a liberal lower court decision,

they are predicted to cast anti-state votes in favor of the petitioner with a probability of .78 and a

95% confidence interval of [.76, .81].

Notably, while the propensity of Federalist Society members to cast pro-state votes

appears largely contingent upon whether the state’s position is liberal or conservative, that is not

the case for other Republican-nominated justices who are not members of the Federalist Society.

Their behavior does not vary substantially based on whether the state’s position is liberal or

conservative.

Figure 3.4, which provides the critical tests of the hypotheses regarding sincere (H3)

versus strategic (H4a and H4b) support for federalism and states’ rights, graphs the average

marginal effect of Federalist Society membership on the likelihood of voting for the petitioner

based on whether a state government is the petitioner or the respondent and whether the lower

court decision was liberal or conservative. The left panel of Figure 3.4 graphs the average

marginal effect of Federalist Society membership in cases with conservative lower court

decisions (meaning that a vote in favor of the respondent is a conservative vote) while the right

panel reports the effect when the lower court decision was liberal (meaning that a vote in favor of

the petitioner is a conservative vote). If Federalist Society members are sincere, then they should

be more likely than their peers to vote for the petitioner when the petitioner is a state, and more

likely to vote for the respondent when the respondent is a state—and that should be true

regardless of whether the lower court decision was liberal or conservative.

Figure 3.4: Marginal Effect of Federalist Society Membership on Voting for Petitioner

Based on State Role in Case and LC Decision Direction

Figure 3.4 graphs the average marginal effect of Federalist Society membership on a justice’s

propensity to vote in favor of the petitioner based on whether the state is the petitioner or the

respondent and whether the lower court decision was liberal or conservative. These results are

based on Model 2 in Table 3.3. Both panels evaluate H3. The left panel evaluates H4a, while the

right panel evaluates H4b.

As the left panel of Figure 3.4 indicates, Federalist Society members are less likely than

their peers, even after controlling for ideology, to cast liberal votes in favor of the petitioner when

the lower court decision was conservative regardless of the state’s role in the Supreme Court case.

These results undermine H3. For the critical test of H4a, which evaluates the behavior of

Federalist Society members when their ideological and legal preferences collide because the

petitioner is a state government advocating a liberal position, the average marginal effect of

Federalist Society membership is -.16 with a confidence interval of [-.31, -.01]. This indicates

that Federalist Society members are roughly 16% less likely than their peers to cast pro-state

votes in these cases, even after controlling for ideology. These results support H4a, that Federalist

Society members are no more likely than their peers to cast pro-state votes when the state’s

position is liberal. In fact, Federalist Society members are less likely than their peers to cast pro-

state votes in these circumstances.

The right panel of Figure 3.4 indicates that Federalist Society members are more likely to

vote in favor of the petitioner when the lower court decision was liberal regardless of the state’s

role in the Supreme Court case, which undermines H3. When the respondent is a state (and thus

Federalist Society members’ ideological and legal preferences collide), the average marginal

effect of Federalist Society membership is .10 with a confidence interval of [.03, .17], indicating

that Federalist Society members are roughly 10% more likely than their peers, even after

controlling for ideology, to cast anti-state votes in favor of the petitioner. These results support

H4b, that Federalist Society members are no more likely than their peers to cast pro-state votes in

favor the respondent when the respondent is a state government advocating a liberal position

before the Court. Here again, they are less likely than their Republican peers to cast pro-state

votes when their ideological and legal preferences collide.

The Model 3 robustness check in which liberal Republican-nominated justices are

excluded also leads to the rejection of H3 while providing additional support for H4a and H4b.

Taken together, the tests for sincere or strategic behavior when ideology and legal principle

collide provide strong support for the strategic account of the behavior of Federalist Society

members and undermine the view that their purported support for federalism and states’ rights is

sincere.

Discussion and Conclusion

This study offers the first systematic examination of the voting behavior of Federalist

Society members on the Supreme Court. More importantly, it is the first systematic quantitative

examination of whether Federalist Society members are sincerely principled or strategic actors

who use legal principle as the “smokescreen” behind which they pursue ideological goals. First, I

show that Federalist Society members are especially likely to vote conservatively on cases

involving a Core Issue (H1) for the network, and that they are more likely to vote conservatively

in cases with State Involvement (H2). But their behavior does not reflect a sincere preference for

states’ rights (H3). They do vote more conservatively than their peers when a pro-state vote also

implicates a conservative vote. But when ideology and the principles of federalism and states’

rights collide such that a pro-state vote is a liberal vote, Federalist Society members are not more

likely than their peers to vote in favor of the state (H4a and H4b). In fact, they are less likely to

do so. In other words, despite any claims that they are especially strong proponents of federalism

and states’ rights, their support is inconsistent and largely conditional upon whether a pro-state

vote implicates a liberal or conservative vote.

These findings are important because they indicate that Federalist Society members treat

state involvement strategically, and these findings will be concerning to observers of the Court

who are concerned about whether justices make decisions based on legal principles or policy

preferences. These results are also concerning because they show that there are major policy

implications for the Federalist Society’s dominant position in judicial politics on policy issues

such as abortion, affirmative action, campaign spending, sex discrimination, and other salient

political issues.

Although this study makes several important contributions, it only scratches the surface

of what should develop into an important subfield in judicial politics given that Republican

presidents reliably nominate members of the Federalist Society to fill vacancies on the Supreme

Court even though non-Federalist Society members are nominated by Republican presidents to

the lower federal judiciary. There are numerous avenues of research worth exploring.

While this paper provides the most comprehensive quantitative analysis of the voting

behavior of Federalist Society members to date, it only examines the behavior of Federalist

Society members at the Supreme Court. Although the behavior of Supreme Court justices is

highly consequential, more research on the behavior of judges at lower levels in the federal

judiciary and in the state court system is needed. Evaluating the voting behavior of Federalist

Society members on other courts will be useful because it would test whether Federalist Society

judges behave similarly when they must be concerned about being reviewed by a higher court.

The influence of the Federalist Society on the Supreme Court’s agenda-setting process

also deserves attention. Because this paper has shown that Federalist Society members have

especially conservative views on policy issues with which key members have been deeply

involved as advocates, it might be expected that as the number of Federalist Society members on

the Supreme Court increases—and especially when they constitute a majority of justices on the

Supreme Court—the Supreme Court’s docket might change in a way that reflects those

preferences.

While there is much more to learn about the behavior and influence of the Federalist

Society, this paper makes significant inroads into understanding the behavior of Federalist

Society justices on the Supreme Court. It not only identifies core issues for the Federalist Society,

but also systematically designs a test of the behavior of Federalist Society members on those core

issues to evaluate the Federalist Society’s arguments on their face. It thus provides the first

quantitative assessment of how the Federalist Society justices decide cases when their ideological

and purported legal principles collide. Understanding how Federalist Society justices make

decisions as well as the political and legal ramifications of those decisions is vitally important, as

the Federalist Society has a strong influence on the role of the courts in Americans’ lives and will

continue to do so for the foreseeable future.

Chapter Four:

The Influence of the Federalist Society on Support for the Supreme Court

Currently, six of the nine justices of the Supreme Court are (or have been) affiliated with

the Federalist Society. Despite the dominant position now held by members of the network on

the highest court in the United States, there is little quantitative research to date on the Federalist

Society. To my knowledge, no research to date takes aim at the Federalist Society’s influence on

public support for the Court. This study addresses that shortcoming in the literature. In this study,

I ask whether source cues associating members of the majority coalition with the Federalist

Society influence public support for the Court.

Drawing from the literature on source cues and court curbing, I generate several

expectations about citizens’ preferences regarding broadly targeted versus narrowly targeted court

curbing proposals. I test these expectations through a survey experiment, conducted on June 9,

2022, in which respondents were asked to read articles about the then-upcoming Supreme Court

decision in the case Dobbs v. Jackson Women’s Health (2022). Because of the importance of the

case—whose outcome would ultimately overturn the landmark decision in Roe v. Wade (1973)—

and the unprecedented leak of a draft opinion, when the survey experiment was conducted in June

2022, the average American was likely more aware of this case than of any pending case since

Bush v. Gore (2000). As such, this case provides an exceptionally difficult test for the influence

of source cues, as many Americans would have learned about the case and the leaked draft prior

to participation in the survey experiment.

Employing a split-sample analysis because I expect opposite effects of source cues for

liberals and conservatives, I find tentative (but inconclusive) evidence that source cues

associating members of the expected majority coalition with Republicans, the Federalist Society,

or both Republicans and the Federalist Society leads liberals to be more supportive of broadly

targeted court curbing measures. However, there is little evidence that these cues lead

conservatives to be less supportive of broadly targeted court curbing measures. This is important,

as it suggests an asymmetric effect of source cues that might erode aggregate support for the

Court and increase support for broadly targeted court curbing measures. Surprisingly, these

source cues do not have the predicted effects on support for narrowly targeted court curbing.

Next, I evaluate whether information frames influence individuals’ support for court

curbing. Respondents were randomly assigned to one of two informational frames. One frame

emphasized the especially conservative voting behavior of justices associated with the Federalist

Society in abortion cases, while the other provided more general information about the Court’s

recent history on abortion cases. The expectation was that the information frame emphasizing the

especially conservative voting behavior of Federalist Society members in abortion cases would

cause liberals to be more supportive of narrowly targeted court curbing measures, and to reduce

support for these measures among conservatives. The results, however, do not support these

expectations.

One difficulty with the split-sample analysis is that it makes distinguishing the effects of

a respondent’s left-right ideology and ideological polarization impossible, as the two are perfectly

correlated within each sample. In additional analyses, I untangle the effects of these ideological

factors by calculating models that include the full sample of respondents, rather than evaluating

liberals and conservatives separately. In these analyses, I find that, in line with Bartels and

Johnston (2020), general policy disagreement with the Court predicts support for both broadly

targeted court curbing and narrowly targeted court curbing measures. And, like those authors, I

find that specific policy disagreement (in this case, with respect to abortion) predicts support for

narrowly court curbing around the issue of abortion to a much greater extent than it does for

broadly court curbing.

Bartels and Johnston (2020) make important contributions to our understanding of how

individuals evaluate the Court by showing that, contrary to conventional wisdom, individuals’

ideological preferences do matter. But the additional analyses in this study move beyond those

findings and uncover two other ideological factors that influence support for court curbing

measures in important but nuanced ways.

The most critical of these findings concerns the role that ideological polarization plays in

shaping their support for broadly and narrowly targeted court curbing measures. Respondents

who are highly polarized—whether liberal or conservative—are more likely to support broadly

targeted court curbing measures, and less likely to support narrowly targeted court curbing

measures. Ideological polarization appears to play an especially central role in influencing

support for broadly targeted court curbing measures, as standardized beta weights indicate that a

respondent’s ideological polarization is at least twice as important an influence (with a beta

coefficient of 0.19) on support for broadly targeted court curbing as are general policy

disagreement (0.07) and specific policy disagreement over abortion (0.07). It also appears to be

the case that liberals are somewhat more supportive of both broadly and narrowly targeted court

curbing measures than are conservatives, but that might be due to case factors such as the

expectation, based on the news of the leaked majority opinion draft, that the Court would rule to

overturn Roe v. Wade in the case Dobbs v. Jackson Women’s Health.

These results have important implications for the Court. The Bartels and Johnston (2020)

findings might imply that, in the context of a conservative decision (or a series of conservative

decisions), liberals should be most likely to support broadly targeted court curbing measures and

conservatives should be least likely to do so, with moderates somewhere in between. But that

perspective does not consider the important role that ideological polarization plays in shaping

support for the Court. The results here suggest that support for broadly targeted court curbing

should be lowest among moderates and those who are only somewhat conservative, higher among

the most conservative individuals and those who are only somewhat liberal, and highest among

those who are most liberal. Given the long-term trend towards higher levels of ideological

polarization among political elites, the findings here point towards a future in which we may see

increased support for broadly targeted court curbing measures from both liberal legislators and

the most conservative legislators.

Research on the Federalist Society

Despite the power Federalist Society members now hold over the Supreme Court,

quantitative research on the network is sparse. We do, however, have some insights about how

the conservative legal movement is different from other social movements, how the network

influences judicial nominations, and how members of the Federalist Society who are judges

behave differently from other Republican-nominated judges.

Hollis-Brusky (2015) offers a way of thinking about the Federalist Society that

distinguishes it from other political networks. She argues that the Federalist Society network is a

political epistemic network (PEN) whose purpose is to spread politically relevant legal

knowledge among conservative attorneys. Tracing the involvement of Federalist Society

members at various positions within the legal field—as litigants, amicus curiae, advocates

presenting arguments before the Supreme Court, law clerks for Supreme Court justices, and as

Supreme Court justices themselves—she goes to great lengths to show how Federalist Society-

generated and promulgated legal arguments reach the Supreme Court. She also identifies some of

the key issues of concern for Federalist Society members.

While Hollis-Brusky (2015) makes the compelling case that the Federalist Society is

indeed a PEN, Bird, King, and McGee (2022) show that it is more than just a PEN. The Federalist

Society network also serves as a high-information partisan cue for US senators when voting on

judicial nominees. Those authors find consistent support for the hypotheses that Republicans are

more likely to support nominees who are affiliated with the Federalist Society, while Democrats

are less likely to support those nominees.

Two extant studies evaluate the behavior of judges who are members of the Federalist

Society. Federalist Society members on the United States courts of appeals are more conservative

than not only Democrat-nominated judges but also than other Republican-nominated judges

(Scherer and Miller 2009). On the Supreme Court, they also vote more conservatively than other

Republican-nominated justices on certain policy issues for which members of the network have

been deeply involved as advocates, and they are also less likely to cast votes in favor of states’

rights when the state’s position is liberal (see Chapter Three). While researchers have made

important contributions to our understanding of the Federalist Society and its influence, none of

the prior work, to my knowledge, evaluates its influence on public support for the Court.

Source Cues and Support for the Court

Citizens often rely on source cues when making political evaluations. The traditional

view is that do this when they lack high-quality information or well-reasoned beliefs (Popkin

1994; Lupia and McCubbins 1998). A more recent line of research suggests that cue-taking

reflects the motivation of individuals to support political identities that they value by expressing

views that have been cued to be consistent with support for those political identities (Kahan

2013). While source cues have been studied in a variety of contexts, they have less frequently

been studied in the judicial context. This is likely due to the traditional view that the courts are

“different” and less political than other political institutions.

Several studies in the last decade have, however, fruitfully applied the concept of source

cues to the study of the courts. One key finding relevant to the current study is that source cues

can have a strong effect on the acceptance of a Supreme Court decision, as Democrats are more

likely to support decisions handed down by a Democratic majority while Republicans are more

likely to support decisions handed down by a Republican majority (Nicholson and Hansford

2014). In addition to influencing respondents’ acceptance of individual Supreme Court decisions,

source cues can also influence support for at least some court-curbing measures, such as reducing

judges’ tenure from life to 18 years or allowing Congress (or the president) to reverse Supreme

Court decisions (Clark and Kastellec 2015). Likewise, Bartels and Johnston (2020) show that

polarized party cues can strongly influence support for narrowly targeted court curbing measures,

although they suggest that the effect of polarized party cues are a weaker influence on support for

broadly targeted court curbing measures.

While the evidence surrounding the influence of source cues on support for broadly

targeted court curbing measures appears somewhat mixed, with Clark and Kastellec (2015)

finding strong support and Bartels and Johnston (2020) finding tentative support, their findings

are largely consistent with each other. Thus, in the context of this vignette experiment, I expect

that source cues describing the expected majority coalition as either consisting of Republican-

nominated justices, Federalist Society members, or Republican-nominated justices who are also

Federalist Society members will influence support for broadly targeted court curbing measures.

Specifically, I expect that for liberals, respondents who receive an article describing justices in

the expected majority coalition as having been nominated by Republican presidents, as being

members of the Federalist Society, or both will report higher support for broadly targeted and

narrowly targeted court curbing measures. I expect these cues to have the opposite effect for

conservative respondents.

Coalition Information Hypothesis. When the expected majority coalition of an upcoming

case is described as consisting of Republican-nominated justices, Federalist Society members, or

Republican-nominated justices who are also Federalist Society members, liberals will report

higher levels of support for both broadly targeted and narrowly targeted Supreme Court curbing

measures, while conservatives will report lower levels of support for both broadly targeted and

narrowly targeted Supreme Court curbing measures.

While no study to date has, to my knowledge, evaluated the influence of the Federalist

Society as a source cue for individuals among the public, one study does explore the influence of

the Federalist Society as an informational cue for US Senators considering judicial nominations

(Bird, King, and McGee 2022). The authors of that study find that US senators use Federalist

Society membership as a high-information partisan cue when evaluating judicial candidates, with

Republicans being more likely to support the nominations of Federalist Society members and

Democrats being less likely to do so.

Does the public treat Federalist Society membership in a similar manner? Or does

Federalist Society membership carry little weight with the average citizen? If a cue provides

respondents with information that is likely to be relevant to a wide array of cases across various

issue areas, then it would likely influence respondents’ support for broadly targeted court curbing

measures. On the other hand, if the cue only provides information that pertains to a narrow subset

of cases the Supreme Court might see, then we should see changes in support for narrowly

targeted court curbing measures, but not for broadly targeted measures. Because the Federalist

Society information frame used in this study focuses on the voting behavior of Federalist Society

members in abortion cases (but not other types of cases), I do not expect the information frame to

influence support for broadly targeted court curbing measures. But I do expect this information

frame to influence respondents’ support for narrowly targeted court curbing measures, with

liberals reporting higher levels of support and conservatives reporting lower levels of support for

such measures.

Federalist Society Information Hypothesis: Liberal respondents who receive the

Federalist Society Information cue will report higher support for narrowly targeted court curbing

measures, while conservative respondents who receive the Federalist Society Information cue will

report lower support for narrowly targeted court curbing measures.

Policy Disagreement and Support for the Court

The Supreme Court must maintain a reservoir of support for its rulings to have

authoritative force, in part because its power of judicial review has no explicit enforcement

mechanism (Bartels and Johnston 2013). Historically, the public has been more supportive of the

Supreme Court than of the Presidency or Congress (Gibson 2007), and the conventional wisdom

is that individuals’ ideological preferences do not influence their support for the Supreme Court

(Caldeira and Gibson 1992; Gibson 2007). Instead, according to the conventional wisdom,

support for the Court is linked to factors such as individuals’ democratic values (Gibson and

Nelson 2015) and their political knowledge (Gibson and Caldeira 2009).

Another (and more recent) vein of research highlights the role of policy disagreement in

shaping individuals’ views of the Court. Bartels and Johnston (2013) employ a survey experiment

to assess how individuals’ political leanings and their perception of the Court’s ideology

influences their views on the Court’s legitimacy, showing that individuals with subjective

ideological disagreement with the Court report dimmer views of the Supreme Court’s legitimacy.

Christenson and Glick (2015) also find evidence that respondents’ views regarding the legitimacy

of the Court is influenced by their perceived ideological distance from the Court. Contrary to the

conventional wisdom, ideological preference do appear to influence individuals’ support for the

Court.

In their important book, Bartels and Johnston (2020) introduce an important conceptual

distinction concerning support for the Court. They note that previous researchers have used a

somewhat muddied concept of “legitimacy” that conflates support for court curbing with

procedural perceptions of judicial legitimacy. Rather than blending those two concepts, the

authors focus squarely on court curbing measures and create a conceptual distinction between

support for broadly targeted court curbing and narrowly targeted court curbing. Support for

broadly targeted court curbing refers to support for measures that would undermine the power of

justices individually or of the Court as a whole, while support for narrowly targeted court curbing

refers to support for measures that would undermine the power of the Court on a specific policy

issue (or on a few specific policy issues). The authors also distinguish between general and

specific policy disagreement. General policy disagreement refers to the degree to which an

individual feels like they disagree with the Court in general, while specific policy disagreement

refers to whether an individual disagrees with the Court on a specific issue (or a subset of issues).

Bartels and Johnston (2020) show that general policy disagreement influences support for

both broadly targeted and narrowly targeted court curbing measures. They also find that specific

policy disagreement influences support for both types of court curbing, but that its influence is

stronger on support for narrowly targeted court curbing. I draw two expectations from these

findings. First, I expect that respondents with higher levels of general policy disagreement with

the Court be more supportive of both broadly targeted and narrowly targeted forms of court

curbing. And second, I expect that specific policy disagreement will also influence individuals’

support for both broadly targeted and narrowly targeted forms of court curbing.

General Policy Disagreement Hypothesis. Respondents with higher levels of general

policy disagreement with the Court will report higher levels of support for both broadly and

narrowly targeted court curbing measures.

Specific Policy Disagreement Hypothesis. Respondents with higher levels of specific

policy disagreement with the Court will report higher levels of support for both broadly and

narrowly targeted court curbing measures.

Vignette Experiment: Dobbs v. Jackson Women’s Health

To evaluate these hypotheses, I conducted a survey experiment via Amazon’s Mechanical

Turk (mTurk). Although mTurk samples tend not to be representative (Berinsky, Huber, and Lenz

2012) with respondents skewing white and young (Ono and Zilis 2022), various researchers have

found that mTurk studies produce consistent results that can be replicated across national surveys

(Clifford, Jewell and Waggoner 2015; Mullinix et al. 2015).

The experiment was fielded with a sample of 1,100 participants6. To be eligible for

participation, respondents had to be in the United States, be at least 18 years of age, have

completed at least 500 prior tasks, and have at least 95% of their prior tasks accepted. Following

the recommendation of Kennedy et al. (2020), I ensured that each respondent only completed the

survey once. The survey was conducted on Thursday, June 9, 2022, and respondents were paid

$0.45 to complete the survey.7 The median response time was about 6 minutes and 40 seconds,

with most respondents taking between 4 and 10 minutes to complete the survey.

The embedded experiment presented realistic information about a then-pending case

before the United States Supreme Court—the landmark case Dobbs v. Jackson Women’s Health

(2022). At the time that respondents took the survey, a draft of the majority opinion written by

Associate Justice Samuel Alito had been made public more than a month earlier, but the official

announcement of the majority decision was still 15 days away. At the time, it was widely known

that the decision was expected to overturn Roe v. Wade (1973) and Planned Parenthood v. Casey

(1992), and that the decision would be divided largely along partisan lines, with at least five of

the six conservative justices (all six of whom, again, are Federalist Society affiliates) joining the

majority. At the time it was unclear how Chief Justice John Roberts would vote, although he

ultimate wrote an opinion concurring in judgement. Prior to the leaked draft, Dobbs v. Jackson

Women’s Health was already a highly salient case, but the highly unusual leak of the draft made

substantial national news before the decision was officially made.

6 The University of Kentucky IRB letter of Approval for Exemption Certification is available in Appendix

7 Following Arechar et al. (2017), I avoided conducting the online survey experiment over a weekend or

during an evening. The data was collected on the afternoon of Thursday, June 9, 2022.

This set of circumstances offers a particularly rigorous test for the influence of source

cues on support for court curbing because the leak was unusual, it made national news, it made

the likely outcome of the case known, and it revealed that the majority coalition would be largely

divided along partisan lines. Given these circumstances, one would expect source cues to have

little impact on respondents’ support for court curbing measures because many respondents

would already know the ideological direction of the outcome and the ideological composition of

the majority coalition. In this context, even modest findings in support of the influence of source

cues would be telling.

In the vignette experiment, respondents answered a series of questions regarding their

personal demographic information, ideology, partisanship, their preferences on abortion policy,

and two other policy issues (both of which were included to avoid signaling that the survey

experiment concerned abortion policy). Respondents were next asked questions evaluating their

political knowledge. Then, they were randomly assigned to a vignette discussing the then-

upcoming decision in Dobbs v. Jackson Women’s Health, or to a control group that did not

receive an article

Respondents who received an article were told that court observers “expect that Roe v.

Wade will likely be overturned this year” and that “It is expected that the majority coalition”

either (1) “will consist of moderate, libertarian, and conservative justices.”; (2) “will include only

Republican-nominated justices.”; (3) “will include only justices who are members of the

Federalist Society.”; or (4) “will include only Republican-nominated justices who are members of

the Federalist Society.” Respondents were also randomly assigned to one of two informational

cues providing either (1) general information about the Court’s recent history on abortion cases or

(2) specific information about the ideological orientation of the Federalist Society, the presence of

Federalist Society members on the Supreme Court, and the voting behavior of justices from the

Federalist Society on cases involving abortion. Respondents who were not assigned to the control

group receive one of the four coalition cues and one of the two informational cues. The text of

the articles is reproduced below, including all four coalition cues and both informational cues:

After hearing oral arguments in December, the US Supreme Court will decide another important

abortion case this year. In this case, the justices will be deciding whether to overturn Roe v. Wade

and allow states to pass more restrictive abortion laws. The Court is currently led by justices that

[Coalition Cue 1: form a coalition of moderates, libertarians, and conservatives./Coalition Cue 2:

were nominated by Republicans./Coalition Cue 3: are members of the Federalist

Society./Coalition Cue 4: were nominated by Republicans and are members of the Federalist

Society.]

[Federalist Society Information Cue: Six of the nine Supreme Court justices are members of the

Federalist Society, which is a network of conservative attorneys and judges. Members of the

Federalist Society typically have very conservative views on abortion./General Information Cue:

Eight of justices have previously heard abortion cases while on the Supreme Court. The Supreme

Court’s most recent abortion case was June Medica Services v. Russo, which involved a

Louisiana law. That case was decided in 2020.]

Several of the justices asked tough questions during oral arguments, leading Court observers to

expect that Roe v. Wade will likely be overturned this year. It is expected that [Coalition Cue 1:

the majority coalition will consist of moderate, libertarian, and conservative justices./Coalition

Cue 2: the majority coalition will include only Republican-nominated justices./Coalition Cue 3:

the majority coalition will include only justices who are members of the Federalist

Society./Coalition Cue 4: the majority coalition will include only Republican-nominated justices

who are members of the Federalist Society.]

[Federalist Society Information Cue: Federalist Society members have cast anti-abortion votes

90% of the time in Supreme Court cases involving abortion./General Information Cue: Last

year’s Supreme Court docket did not include any cases involving abortion.] A decision is

expected in June.

Because the Coalition Information treatment has four possible conditions and the

Federalist Society Information treatment has two, there were a total of eight experimental

conditions in addition to the previously discussed control condition in which respondents did not

read an article about Dobbs v. Jackson Women’s Health. As such, each respondent was randomly

assigned to one of nine conditions.

After reading the vignette, respondents answered two manipulation check questions. The

first of these asked how the article described the composition of the expected majority coalition,

while the second asked respondents to evaluate the ideology of the Federalist Society on a scale

from 1 (extremely liberal) to 7 (extremely conservative). Finally, all respondents were asked a

series of questions about their support for various court curbing measures.

This study uses two dependent variables. The first dependent variable, Broadly Targeted

Court Curbing, is an index of three indicators. These three indicators reflect a respondent’s

support for removing judges who “who consistently make decisions at odds with what a majority

of the people,” reducing the Supreme Court’s jurisdiction on “certain types of controversial

issues,” and making the Court “less independent so that it listens a lot more to what the people

want.” For each of these items, possible responses ranged from “Strongly disagree” (coded as 0)

to “Strongly agree” (coded as 4). Thus, the dependent variable Broadly Targeted Court Curbing

may take on values from 0 to 12, where 12 is the highest level of support for broadly targeted

court curbing.

The second dependent variable, Narrowly Targeted Court Curbing, is also an index of

three indicators. These indicators reflect an individual’s support for curbing the Court’s power

specifically in cases involving abortion. The first question asked whether the Supreme Court

should be considered the “final word” on abortion. Possible responses range from “Definitely

yes” (coded as 0) to “Definitely no” (coded as 4). The second question asked respondents how

much authority the Supreme Court’s decision should have over the other branches of government

and the American people on the issue of abortion. Responses ranged from “Complete authority”

(coded as 0) to “No authority at all” (coded as 3). Finally, the third question asked respondents

whether they agree with the statement that “All Americans (including elected officials

nationwide) should comply with the Supreme Court's ruling on the issue of abortion.” Responses

ranged from “Strongly agree” (coded as 0) to “Strongly disagree” (coded as 4). The dependent

variable Broadly Targeted Court Curbing ranges from 0 to 11.

There are two key independent variables in this study: Coalition Information (as well as

its collapsed version) and Federalist Society Information. The variable Coalition Information,

which ranges from 0 to 3, indicates whether the respondent was told that the majority coalition in

Dobbs v. Jackson Women’s Health was expected to consist of “moderate, libertarian, and

conservative justices” (coded as 0), “Republican-nominated justices” (coded as 1), “only justices

who are members of the Federalist Society” (coded as 2), or “only Republican-nominated justices

who are members of the Federalist Society” (coded as 3). For the analyses that follow, the control

condition is also coded as 0. Coalition Information is treated as categorical, and I use this variable

to test the Coalition Information Hypothesis. I also created a collapsed version of this variable,

Coalition Information (Collapsed), which is a binary variable in which the Republican cue, the

Federalist Society cue, and the cue that combines both are all coded as 1. As with the previous

version of the variable, the two conditions in which the expected majority coalition is described

as consisting of “moderate, libertarian, and conservative justices” is coded as 0, while the control

group is also coded as 0.

Federalist Society Information is a binary variable indicating whether the respondent

received information about the voting behavior of justices who are members of the Federalist

Society on cases involving abortion (coded as 1) or whether they were given the more general

information (coded as 0). I use this variable to evaluate the Federalist Society Information

Hypothesis.

General Policy Disagreement is measured as the difference between the respondent’s

perception of the Supreme Court’s ideology and the respondent’s perception of their own

ideology. Because scores for both measures range from 1 (“extremely liberal”) to 7 (“extremely

conservative”), these values can theoretically range from -6 (when an “extremely conservative”

respondent views the Supreme Court as “extremely liberal”) to 6 (when an “extremely liberal”

respondent views the Supreme Court as a “extremely conservative”). Thus, negative values for

General Policy Disagreement indicate that the respondent sees the Supreme Court to the

ideological left of herself, while positive values indicate that the respondent sees the Supreme

Court to the ideological right.

Abortion is a binary variable coded as 0 if a respondent reports the view that by law

abortion should (1) should never be permitted or (2) only be permitted in case of rape, incest, or

when the woman's life is in danger. It is coded as 1 if the respondent reports the view that by law

abortion should (1) be permitted for reasons other than rape, incest, or danger to the woman's life,

but only after the need for abortion has been clearly established or (2) always be permitted as a

matter of personal choice. Because the case involved in this study concerns abortion, Abortion is

also an indicator of specific policy disagreement.

I also include a series of control variables that might be related to support for broadly

targeted or narrowly targeted court curbing measures. I control for the respondent’s left-right

ideology, using the self-reported Ideology measure ranging from 1 (“extremely liberal”) to 7

(“extremely conservative”). Sex is a binary variable coded as 0 if the respondent is male and 1 if

the respondent is female. Race is a binary variable coded as 0 if the respondent is white and 1

otherwise. Age is coded as 1 (if the respondent is 18-25 years old), 2 (26-35 years), 3 (36-45

years), 4 (46-55 years), 5 (56-65) years, or 6 (66 years old or older). Political Knowledge is a

measure of how many questions a respondent answered correctly out of four political knowledge

questions. Values range from 0 to 4, where 0 indicates that the respondent did not answer any

questions correctly and 4 indicates that the respondent answered 4 questions correctly.

One concern when conducting online survey experiments is whether respondents are

paying attention to the information being present or simply “clicking through” to complete the

survey. Two tables in the Appendix allay those concerns. Table C1 in the Appendix reports the

results of a two-sample t- test comparing respondents who received the Federalist Society

Information condition with those who did not. The results indicate that respondents who received

the Federalist Society Information condition reported an average ideology score for the Federalist

Society of 4.66, while respondents who received the general information reported an average

ideology score of 4.44. Table C2 in the Appendix reports the result of Pearson's chi-squared test,

which indicates that there is a strong relationship between the condition respondents were

assigned to and their responses to question about how the article describes the Federalist Society.

The results provide additional evidence that respondents largely both read and understood how

the article presented the Federalist Society.

Although there is clear evidence of respondents reading and understanding the

information they were given, many respondents—even those in the Federalist Society

Information condition—incorrectly identified the ideology of the Federalist Society. While the

Federalist Society describes itself as a group of conservatives and libertarians (About Us, 2022),

roughly one-third of respondents did not see them that way. It is not clear whether these

respondents were simply wrong, whether they were guessing, or whether they were intentionally

providing inaccurate responses. To account for potential systematic differences between these

respondents who view the Federalist Society as liberal and other respondents, I include the

variable Data Control, which is a control variable indicating that the respondent may not have

answered or honestly (or may simply be wrong). It is coded as 0 if the respondent reported that

the Federalist Society’s ideology is moderate, slightly conservative, conservative, or extremely

conservative, and it is coded as 1 if the respondent reported that the Federalist Society’s ideology

is slightly liberal, liberal, or extremely liberal.

Finally, I create a control variable Attention Control, which is used in additional

robustness checks in the Appendix. Attention Control has three components. The first two

components are binary indicators based on responses to attention check questions that were built

into the survey experiment. These two questions gave respondents specific directions about which

response(s) to select, regardless of their views. For each of these questions, respondents are coded

as 0 if they answered correctly, and 1 if they did not. The third component is a binary indicator

coded as 0 if the respondent completed the survey in under 15,000 seconds and 1 if the duration

was greater than 1,500 seconds (25 minutes). Attention Control is calculated as the sum of these

three indicators. Technically, these values may range from 0 to 3, but in practice the highest value

is 2 as no respondents answered both attention check questions incorrectly and took longer than

25 minutes to complete the survey. Table 4.1 reports summary statistics for the two dependent

variables and the control variables presented in the analysis that follows.

Table 4.1: Summary Statistics for Vignette Experiment

Variable Obs. Mean Std. Dev Min Max

Broadly Targeted Court Curbing

1074

7.28

3.36

Narrowly Targeted Court Curbing

1073

4.06

2.95

Coalition Information

1072

1.33

1.16

Coalition Information (Collapsed)

1072

0.66

0.47

Federalist Society Information

1072

0.44

0.50

Policy Disagreement

1073

0.58

2.24

Ideology

1073

3.56

1.91

Abortion

1070

0.50

Sex

1069

0.44

0.50

Race

1072

0.17

0.37

Age

1074

2.96

1.31

Political Knowledge

1074

2.09

1.17

Data Control

1074

0.34

0.47

Attention Control

1074

0.14

0.40

Before proceeding with the analysis, several idiosyncrasies about the data are worth

noting. First, this sample skews slightly male (56%) and white (83%, including Latinos who

identify as white), and especially Catholic (39%). Black Americans were under-represented (5%),

as were older respondents (only 4% were aged 66 or older). Respondents also skewed liberal,

with 55% reporting ideology scores of less than 4 and 32% reporting ideology scores of greater

than 4.

Additionally, it is worth noting the distribution of respondents among the Coalition

Information and Federalist Society Information conditions. Recall that respondents were

randomly assigned to one of nine conditions. For the purposes of this study, Coalition

Information and Federalist Society Information are both coded as 0 when the respondent is in the

control condition, so the group of respondents for whom both are coded as 0 is about twice as

large (241 respondents) as each other group, the rest of which included between 116 and 122

respondents. While some might prefer coding the control condition as missing data for these two

variables, the substantive differences in results produce by models calculated under those

conditions (not shown) are negligible. Table 4.2 provides information on the distribution of

respondents in each condition given the coding procedures for Coalition Information and

Federalist Society Information.

Table 4.2: Distribution of Treatment Conditions

Federalist Society Information

No (or control)

Yes

Total

Coalition

Cue

Broad (or control)

241

119

360

Republican

117

116

233

Federalist Society

118

122

240

Both (Fed. & Rep)

122

117

239

Total

598

474

1074

For both Broadly Targeted Court Curbing and Narrowly Targeted Court Curbing, I

calculate OLS regression models. Because Coalition Information and Federalist Society

Information are expected to influence liberals and conservatives in different ways, I conduct a

split sample analysis. This not only allows for the possibility that liberals and conservatives will

respond to the stimuli differently, but also for the possibility that other factors such as Policy

Disagreement and Ideology might influence support for court curbing among liberals and

conservatives in different ways.

Results

This study leverages a survey experiment conducted after a major news story in which a

draft opinion was leaked to the public prior to the announcement of the official opinion on a

highly salient case. Many respondents already knew the expected outcome of the case and would

have been aware that the majority coalition would include at least five Republican-nominated

justices, with all three Democrat-nominated justices voting in the minority. This makes a survey

experiment after the leak of the opinion draft in Dobbs v. Jackson Women’s Health an especially

difficult test for the influence of source cues on public support for broadly targeted and narrowly

targeted court curbing measures. I begin by evaluating the hypotheses concerning support for

broadly targeted court curbing and conclude with the hypotheses concerning support for narrowly

targeted court curbing. For each of those two dependent variables, I evaluate liberals and

conservatives separately. For each of the main four sets of analysis—support for broadly targeted

court curbing among (1) liberals and (2) conservatives and support for narrowly targeted court

curbing among (3) liberals and (4) conservatives—the models of central concern are Model 5 and

Model 6. Additional robustness checks based on Model 6 are available in the Appendix.

Broadly Targeted Court Curbing

Table 4.3 reports OLS regression results for the influence of the source cues on support

for broadly targeted court curbing measures among liberals. Model 1 only includes the

experimental conditions and the variable Data Control. Model 2 reports the same model, but with

one difference: in this model, Coalition Information is collapsed into a binary variable coded as 1

if the respondent received a vignette that described the majority coalition as consisting of

Republicans, Federalist Society members, or both, and 0 otherwise. Model 5 reports the full

model including all control variables, while Model 6 reports the full model with the collapsed

Coalition Information variable. Model 4 is identical to Model 6, except that it does not control for

respondent ideology. Likewise, Model 3 excludes the control variable for respondent ideology,

but otherwise reflects Model 5. For each of the models predicting support for broadly targeted

court curbing measures, additional robustness checks based on Model 6 are available in the

Appendix.

Table 4.3: OLS Regression Models Predicting Support for Broadly Targeted Court

Curbing (Liberals)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Republican 0.53

(0.31)

0.52

(0.29)

0.42

(0.29)

Federalist S. 0.72*

(0.33)

0.63*

(0.31)

0.52

(0.30)

Both Cues

0.55

0.34

0.33

(0.31)

(0.30)

(0.29)

All Cues

(Collapsed)

0.59*

(0.25)

0.49*

(0.23)

0.42

(0.23)

Federalist S.

Information

0.05

(0.23)

0.05

(0.23)

-0.01

(0.22)

0.00

(0.22)

-0.05

(0.22)

-0.04

(0.22)

Policy

Disagreement

0.25**

(0.07)

0.25**

(0.07)

0.07

(0.08)

0.07

(0.08)

Ideology

0.93***

0.94***

(0.17)

Abortion

0.19

0.18

0.33

(0.27)

(0.26)

Sex

0.49*

0.50*

0.53*

(0.22)

(0.21)

(0.26)

Race

0.29

0.30

0.25

(0.31)

(0.30)

Age

0.39***

0.40***

(0.09)

Political

Knowledge

-0.68***

(0.11)

-0.67***

(0.11)

-0.68***

(0.10)

-0.67***

(0.10)

Observations

584

577

Adjusted

0.1058

0.1084

0.2032

0.2050

0.2426

0.2449

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Dependent variable is the respondent’s support for Broadly Targeted Court

Curbing. The models are restricted to liberals, defined as respondents with self-reported ideology

scores of 3 or less. The coefficient and standard error for Data Control are omitted.

The results in Model 1 and Model 2, which include only the experimental conditions and

Data Control provide some tentative support for the Coalition Information Hypothesis. In Model

1, coefficient for the condition in which respondents are told that the majority coalition is

expected to consist of Federalist Society members is positive and significant at p=0.05, while the

Republican cue and the cue describing the expected majority coalition as consisting of

Republican-nominated justices who are also members of the Federalist Society fall just short of

statistical significance (p=0.09 and p=0.08, respectively). In Model 2, Coalition Information

(Collapsed) is significant at p=0.05, providing additional support for the Coalition Information

Hypothesis that source cues about the majority coalition will influence support for broadly

targeted court curbing measures among liberals, even in a highly salient case such as Dobbs v.

Jackson Women’s Health. Model 3 and Model 4 provide additional support for the Coalition

Information Hypothesis. In Model 3, the coefficient for the Federalist Society cue is positive

(0.63) and significant (p<0.05), while the Republican cue is positive (0.52) and approaches

significance (p<0.10). Surprisingly, the cue associating justices in the expected majority coalition

with both Republicans and the Federalist Society falls short of significance. Model 4 provides

additional support for the Coalition Information Hypothesis, as the coefficient Coalition

Information (Collapsed) is positive (0.49) and significant (p<0.05).

However, in the full models that control for respondent ideology, support is slightly

diminished. In Model 5, the coefficients for the Republican cue, the Federalist Society cue, and

the cue indicating that the coalition will be composed of Republican-nominated justices who are

also members of the Federalist Society all have positive coefficients ranging from 0.33 to 0.52,

suggesting that liberals who receive these cues may be somewhat more supportive of broadly

targeted court curbing measures. Surprisingly, only the Republican condition and the Federalist

Society condition approach significance (p=0.14 and p=0.09, respectively). In Model 6, the

variable Coalition Information (Collapsed) is positively signed with a coefficient of 0.42,

although it falls just short of traditional levels of acceptance (p=0.07). Nonetheless, these results

are important. Even with a relatively small sample of liberals (N= 577 in Model 5 and Model 6),

and in a highly salient case, source cues about the majority coalition exhibit some evidence of

influence over liberals’ support for broadly targeted court curbing measures.

This differences between the models that control for respondent ideology (Model 5 and

Model 6) and those that do not (Model 3 and Model 4) provide important insights about the

100

nature of support for broadly targeted court curbing measures, at least among liberals. Model 3

and Model 4 both support the now-mainstream view, most notably advocated in Bartels and

Johnston (2020), that individuals with higher levels of general policy disagreement with the Court

will be more supportive of broadly targeted court curbing measures. However, Model 5 and

Model 6 tell another story. When a respondent’s ideology is controlled for, liberals’ support for

broadly targeted court curbing is driven by how liberal they are, and not by how much they feel

like they disagree with the Court. Moreover, the results in Model 5 and Model 6 indicate that

respondent ideology is a very strong predictor (p<0.001) of support for broadly targeted forms of

court curbing. The negative coefficient for Ideology indicates that liberals are the expected to

report higher support for broadly targeted court curbing measures than are conservatives.

While Model 3 and Model 4 provide tentative support for the General Policy

Disagreement Hypothesis, indicating that liberals with higher levels of General Policy

Disagreement will be more supportive of broadly targeted court curbing measures, the full

models that control for ideology do not. The indicator Abortion, which is a measure of specific

policy disagreement, is not significant in any of the models presented, although the coefficient is

in the expected direction in each model. While the results here are surprising in that, at least

among liberals, general policy disagreement and specific policy disagreement do not appear to

have the expected effects, they also reveal what I believe to be a novel finding—that the

ideological factor that contributes most to support for broadly targeted court curbing appears to be

ideology, not policy disagreement. However, because of the split sample design, it is not yet clear

whether the most extreme liberals are more supportive of broadly targeted court curbing because

they are more liberal, more polarized, or both. Disentangling these influences would require

calculating regression models for the full sample.

Models predicting conservative support for broadly targeted court curbing are reported in

Table 4.4. The results here are surprising. I expected that respondents given the Republican cue,

101

the Federalist Society cue, or the cue associating the majority coalition with both Republicans and

the Federalist Society would report lower support for broadly targeted court curbing measures.

While the coefficients for the individual coalition cues were negative as expected with only one

exception (the Federalist Society coalition cue in Model 5), all feel short of traditionally accepted

levels of significance. In Model 3 and Model 5, the cue associating the majority coalition with

both Republicans and the Federalist Society approaches significance, but it falls short of that

threshold in both. Thus, I do not find support for the hypothesis that associating the majority

coalition with Republicans, the Federalist Society, or both will lead conservatives to be less

supportive of broadly targeted court curbing.

102

Table 4.4: OLS Regression Models Predicting Support for Broadly Targeted Court

Curbing (Conservatives)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Republican -0.03

(0.59)

-0.17

(0.49)

-0.17

(0.48)

Federalist S. -0.18

(0.55)

-0.13

(0.47)

0.05

(0.46)

Both Cues

0.65

0.79

0.67

(0.60)

(0.50)

All Cues

(Collapsed)

-0.27

(0.45)

-0.34

(0.38)

-0.24

(0.38)

Federalist S.

Information

0.90*

(0.43)

0.91*

(0.43)

0.06

(0.36)

0.06

(0.36)

0.11

(0.36)

0.11

(0.36)

Policy

Disagreement

0.00

(0.11)

-0.01

(0.11)

0.13

(0.12)

0.12

(0.12)

Ideology

0.86**

0.85**

(0.26)

Abortion

0.15

0.16

0.38

0.39

(0.37)

Sex

0.30

0.34

0.27

0.31

(0.36)

Race

0.11

0.14

0.07

0.10

(0.47)

(0.46)

Age

1.00***

1.01***

0.99***

1.00***

(0.13) (0.13) (0.13) (0.13)

Political

Knowledge

-1.20***

(0.15)

-1.19***

(0.15)

-1.15***

(0.15)

-1.14***

(0.15)

Observations

342

340

Adjusted

0.0283

0.0313

0.3316

0.3318

0.3516

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Dependent variable is the respondent’s support for Broadly Targeted Court

Curbing. The models are restricted to conservatives, defined as respondents with self-reported

ideology scores of 5 or more. The coefficient and standard error for Data Control are omitted.

Overall, the evidence for the Coalition Information Hypothesis is mixed: these cues

appear to influence liberal respondents to be more supportive of broadly targeted court curbing

(although the evidence is admittedly tentative), but they do not appear to have much effect on

conservatives. These are important findings. If cues about the composition of the majority

coalition increase support for broadly targeted court curbing among liberals but do not influence

conservatives to be less supportive of such measures, then source cues about the majority

103

coalition may lead to increased aggregate support for broadly targeted court curbing because the

increase in support among liberals is not offset by a similar decrease in support among

conservatives.

Turning to the General Policy Disagreement hypothesis, the results are once again

surprising given the current literature on policy disagreement and support for broadly targeted

court curbing. Even in Models 3 and 4, which control for General Policy Disagreement but not

Ideology, General Policy Disagreement does not predict support for broadly targeted court

curbing measures among conservatives. In the full models, Model 5 and Model 6, General Policy

Disagreement is positive signed but not significant. The split sample analysis also does not

support the hypothesis that specific policy disagreement will influence support for broadly

targeted court curbing. However, in Model 5 and Model 6, Ideology is both positive (0.86 and

0.85) and significant (p<0.01), indicating that, among conservatives, respondents who are more

conservative report higher levels of support for broadly targeted court curbing.

These findings appear to stand at odds with the current literature, given the now-standard

view is that policy disagreement is an important predictor of support for broadly targeted court

curbing measures. Among liberals, those who are most liberal are more supportive of broadly

targeted court curbing. And, among conservatives, those who are most conservative are more

supportive of broadly targeted court curbing measures. In other words, support for narrowly

targeted court curbing measures is driven, in large part, by respondents’ ideological polarization.

Respondents who are more polarized report higher levels of support for broadly targeted court

curbing.

These findings merit further attention. Because the data is split between two samples

(liberal and conservatives), the measure Ideology is both a measure of respondent ideology and a

measure of respondent polarization. To disentangle the influence of general policy disagreement,

specific policy disagreement, the individual’s left-right ideology, and the individual’s degree of

104

ideological polarization, I create a measures Respondent Polarization, which is the absolute value

of the difference between the respondent’s reported ideology score (from 1 to 7) and the midpoint

of that scale (4). Values for Respondent Polarization range from 0 (for moderates) to 3 (for

extreme liberals and extreme conservatives). Evaluating the influence of these three indicators

simultaneously in a split sample design is not possible because, in each subset of data, Ideology is

perfectly correlated with Respondent Polarization (-1.00 for liberals and 1.00 for conservatives).

To evaluate the influence of left-right ideology and respondent polarization simultaneously, I

recalculate Model 5 from Table 4.4, but include all respondents.

The results, reported in Table 4.5 show strong evidence that, when evaluating the

influence of an individual’s general policy disagreement, specific policy disagreement, left-right

ideology, and degree of ideological polarization (or distance from the midpoint of the scale)—one

factor is far more important than the others. And that factor is the individual respondent’s

polarization.

105

Table 4.5: OLS Regression Models Predicting Support for Broadly Targeted Court

Curbing (All Respondents)

Model 1 Beta

Republican 0.20

(0.24)

0.02

Federalist S. 0.38

(0.24)

0.05

Both Cues

0.00

(0.24)

Federalist S. Information 0.03

(0.18)

0.00

Policy Disagreement 0.11

(0.06)

0.07

Ideology

0.12

0.07

(0.07)

Respondent Polarization

0.68***

0.19

(0.09)

Abortion

0.45*

0.07

(0.20)

Sex

0.44*

0.07

(0.18)

Race

0.13

0.01

(0.24)

Age

0.61***

0.24

(0.07)

Political Knowledge -0.87***

(0.08)

-0.30

Observations

1060

Adjusted

0.2822

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Dependent variable is the respondent’s support for Broadly Targeted Court

Curbing. The model includes all respondents without missing data. The coefficient and standard

error for Data Control are omitted.

The coefficient for Respondent Polarization is both large (0.68) and significant

(p<0.001). The beta weights indicate that a one standard deviation change in individual

polarization predicts a 0.19 standard deviation increase in support for broadly targeted court

curbing. Coefficients for General Policy Disagreement and Abortion are both positively signed,

as expected, although the beta weights are much smaller (0.07 for each). And although Abortion

is statistically significant at traditional thresholds, General Policy Disagreement falls just short

106

(although p<0.10). Ideology is negatively signed, indicating that conservatives may be somewhat

less likely to support broadly targeted court curbing than are liberals. However, Ideology falls just

short of traditional significance thresholds (p<0.10). Although these results support the view that

policy disagreement influences support for broadly targeted court curbing, they also reveal that

respondent polarization is a stronger influence than either of those factors.

One final note is required here. Although respondent polarization appears to be the most

important ideological influence on an individual’s support for broadly targeted court curbing, an

individual’s age and political knowledge appear to be even larger influences on support for

broadly targeted court curbing. Respondents who are older are much likely to support broadly

targeted court curbing measures than are younger individuals, and those with higher levels of

political knowledge are also less likely to support broadly targeted court curbing.

Narrowly Targeted Court Curbing

Models predicting support for narrowly targeted court curbing among liberals are

reported in Table 4.6. For liberals, I expected that the Coalition Information cues in which the

expected majority coalition is associated with Republicans, the Federalist Society, or both would

increase liberals’ support for narrowly targeted court curbing. However, the results are only

significant in one condition—the Federalist Society cue—and even then, it’s in the opposite

direction from what was expected.

107

Table 4.6: OLS Regression Models Predicting Support for Narrowly Targeted Court

Curbing (Liberals)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Republican -0.22

(0.29)

-0.20

(0.26)

-0.18

(0.26)

Federalist S. -1.00**

(0.30)

-0.79**

(0.28)

-0.76**

(0.28)

Both Cues

0.00

0.11

(0.29)

(0.27)

All Cues

(Collapsed)

-0.38

(0.23)

-0.28

(0.21)

-0.25

(0.21)

Federalist S.

Information

-0.09

(0.22)

-0.13

(0.22)

-0.01

(0.20)

-0.05

(0.20)

-0.00

(0.20)

-0.03

(0.20)

Policy

Disagreement

0.51***

(0.06)

0.51***

(0.06)

0.56***

(0.07)

0.56***

(0.07)

Ideology

0.26

0.29

(0.15)

Abortion

1.32***

1.35***

1.28***

1.30***

(0.24)

Sex

0.86***

0.83***

0.85***

0.83***

(0.20)

(0.19)

(0.20)

Race

0.13

0.17

0.12

0.16

(0.27)

(0.28)

(0.27)

(0.28)

Age

0.05

0.07

0.06

0.07

(0.08) (0.08) (0.08) (0.08)

Political

Knowledge

-0.22*

(0.09)

-0.25**

(0.09)

-0.22*

(0.09)

-0.25**

(0.09)

Observations

583

576

Adjusted

0.3324

0.3230

0.4565

0.4494

0.4583

0.4518

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Dependent variable is the respondent’s support for Narrowly Targeted Court

Curbing. The models are restricted to liberals, defined as respondents with self-reported ideology

scores of 3 or less. The coefficient and standard error for Data Control are omitted.

I expected that liberals who received the Federalist Society Information condition—in

which respondents were told that there are Federalist Society members on the Supreme Court and

that these justices had conservatively 90% of the time in cases involving abortion—would report

higher levels of support for narrowly targeted court curbing measures concerning the issue of

abortion. The results do not bear this out, as the Federalist Society Information condition is not a

108

significant predictor of liberals’ support for narrowly targeted court curbing measures under any

of the reported model specifications.

I also expected higher levels of general and specific policy disagreement to predict higher

levels of support for broadly targeted court curbing measures. The results indicate that this is true.

General Policy Disagreement and Abortion are statistically and substantively significant across

all four models in which they are included. These results support the General Disagreement

Hypothesis for narrowly targeted court curbing.

Although the Models in Table 4.3 suggested that respondent ideology or polarization

plays a major role in shaping support for broadly targeted court curbing among liberals, that does

not appear to be the case for narrowly targeted court curbing—at least, not in the split-sample

analysis. It may be the case that left-right ideology and respondent polarization have either no

role or merely very small roles in influencing support for narrowly targeted court curbing

measures. Alternatively, it might be the case that both play roles but that they influence support in

opposite directions. Additional analysis after the models predicting support for narrowly targeted

court curbing among conservatives will provide clarity about which of these possibilities is true.

Table 4.7 reports models predicting conservatives’ support for narrowly targeted court

curbing measures. The vignette conditions are not significant across any of the reported model

specifications, although the coefficient for Federalist Society Information is consistently negative,

as expected. Remarkably, for conservatives, the coefficient for General Policy Disagreement is

negative across all four models in which it is included. The two models that control for

respondent ideology make it clear that the respondent’s left-right ideology, degree of polarization,

or both play important roles in shaping support for narrowly targeted court curbing. Those models

indicate that, among conservatives, those who are most conservative are least likely to support

narrowly targeted court curbing measures.

109

Table 4.7: OLS Regression Models Predicting Support for Narrowly Targeted Court

Curbing (Conservatives)

Model 1 Model 2 Model 3 Model 4 Model 5 Model 6

Republican 0.21

(0.36)

0.22

(0.36)

0.23

(0.35)

Federalist S. 0.12

(0.34)

0.21

(0.34)

0.05

(0.33)

Both Cues

0.29

0.16

0.26

(0.37)

(0.36)

All Cues

(Collapsed)

0.02

(0.28)

0.11

(0.28)

0.02

(0.27)

Federalist S.

Information

-0.25

(0.26)

-0.25

(0.26)

-0.31

(0.27)

-0.31

(0.27)

-0.36

(0.26)

-0.35

(0.26)

Policy

Disagreement

-0.09

(0.08)

-0.09

(0.08)

-0.20*

(0.08)

-0.20*

(0.08)

Ideology

0.75***

0.73***

(0.19)

(0.18)

Abortion

0.65*

0.66*

0.45

0.46

(0.27)

Sex

0.12

0.14

0.15

0.17

(0.27)

(0.26)

Race

0.64

0.62

0.60

0.59

(0.34)

Age

0.12

0.13

0.14

(0.10) (0.09) (0.09) (0.09)

Political

Knowledge

-0.02

(0.11)

-0.02

(0.11)

-0.06

(0.11)

-0.06

(0.11)

Observations

342

340

Adjusted

0.0073

0.0078

0.0282

0.0306

0.0708

0.0719

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Dependent variable is the respondent’s support for Narrowly Targeted Court

Curbing. The models are restricted to conservatives, defined as respondents with self-reported

ideology scores of 5 or more. The coefficient and standard error for Data Control are omitted.

The results here are interesting, both because they are at odds with the results for liberals

and because the effect of General Policy Disagreement appears to be the opposite of what was

expected here. Whereas the aggregate effects of left-right ideology and polarization fell short of

significance in the sample of liberals, the aggregate effects of those factors in the conservative

sample are strongly significant, with more conservative respondents being especially unlikely to

support narrowly targeted court curbing. Model 3 and Model 4 support the Specific

110

Disagreement Hypothesis, although Abortion falls just short of statistical significance in the full

models that include ideology.

As was the case when evaluating support broadly targeted court curbing measures, a full

model that includes both liberals and conservatives will help clarify how an individual’s general

policy disagreement, specific policy disagreement, left-right ideology, and degree of ideological

polarization shape support for narrowly targeted court curbing.

Table 4.8 reports results for the full sample. In this model, a respondent’s general policy

disagreement with the Court, specific policy disagreement on the issue of abortion, left-right

ideology, and degree of ideological polarization all influence support for narrowly targeted court

curbing measures. However, the dynamics are somewhat different for narrowly targeted court

curbing than they are for broadly targeted court curbing.

111

Table 4.8: OLS Regression Models Predicting Support for Narrowly Targeted Court

Curbing (All Respondents)

Model 1 Beta

Republican 0.01

(0.21)

0.00

Federalist S. -0.24

(0.21)

-0.03

Both Cues

0.00

(0.21)

Federalist S. Information -0.25

(0.15)

-0.04

Policy Disagreement 0.39**

(0.05)

0.29

Ideology

0.21***

0.13

(0.06)

Respondent Polarization

0.35***

0.11

(0.08)

Abortion

1.23***

0.21

(0.17)

Sex

0.51**

0.08

(0.15)

Race

0.25

0.03

(0.20)

Age

0.03

0.01

(0.06)

Political Knowledge -0.11

(0.07)

-0.04

Observations

1059

Adjusted

0.3248

*p<.05, ** p<.01, *** p<.001. Entries represent OLS regression coefficients with standard errors

in parentheses. Dependent variable is the respondent’s support for Narrowly Targeted Court

Curbing. The model includes all respondents without missing data. The coefficient and standard

error for Data Control are omitted.

Whereas respondents reporting high levels of ideological polarization were more likely to

support broadly targeted court curbing, they are less likely to support narrowly targeted court

curbing. And whereas the strongest ideological influence predicting support for broadly targeted

court curbing is respondent polarization, that is not the case for narrowly targeted court curbing.

The beta weights indicate that General Policy Disagreement (0.29) and Abortion (0.21) are the

two ideological factors with the greatest influence on support for narrowly targeted court curbing

112

measures. These results largely reflect the Johnston and Bartels (2020) assessment of the role of

policy disagreement in shaping support for court curbing.

As was the case for broadly targeted court curbing, those who are more conservative

appear to be less likely to support narrowly targeted court curbing. I hesitate, however, to

conclude that liberals are more likely to support both forms of court curbing than are

conservatives. It may simply be the case that liberals are more likely to support broadly and

narrowly targeted court curbing measures when the Court issues a salient conservative decision,

and that we would observe th4.e opposite relationship if the Court issued a liberal decision.

The results in Table 4.8 also provide insight into why Ideology was significant in the

models predicting conservative support for narrowly targeted court curbing, but not in the models

predicting liberal support. While the forces of left-right ideology and respondent polarization both

reduce support for narrowly targeted court curbing among conservatives, the effects of left-right

ideology and polarization largely counteract each other in terms of their influence on support for

narrowly targeted court curbing among liberals.

The additional analyses in this study predicting support for broadly and narrowly targeted

court curbing measures across the full sample provide important insights about the ideological

forces that influence support for court curbing. The model predicting support for narrowly

targeted court curbing largely reflects the Bartels and Johnston (2020) view that the most

important ideological factors driving support for court curbing are general and specific policy

disagreement. But the results reveal that at least two other ideological factors are important:

individuals who are more conservative are less likely to support narrowly targeted court curbing

measures, and individuals who with high degrees of ideological polarization—whether liberal or

conservative—are less likely to support narrowly targeted court curbing.

113

Most critically, these results upend the prevailing view concerning the ideological forces

that shape support for broadly targeted court curbing measures. Not only does ideological

polarization matter, but it seems to be an even stronger influence than general policy

disagreement on support for broadly targeted court curbing measures, as both very liberal and

very conservative individuals are more likely than moderates to support broadly targeted court

curbing.

Discussion and Conclusion

This study leverages a survey experiment to evaluate whether cues about the composition

of an upcoming Supreme Court decision influence respondents’ support for broadly targeted and

narrowly targeted court curbing measures, and whether informational cues about the voting

behavior of justices associated with the Federalist Society in abortion cases influence support for

narrowly targeted court curbing measures. The survey experiment used vignettes based on the

case Dobbs v. Jackson Women’s Health. This was a particularly difficult test for the influence of

source cues, as the survey experiment involved an exceptionally salient case and was conducted

after a draft of the majority opinion had been leaked to the public and made national news.

Nonetheless, I expected that cues about the expected majority coalition in the case of Dobbs v.

Jackson Women’s Health would influence respondents’ support for both broadly and narrowly

targeted court curbing measures. Specifically, the expectation was that these cues would make

liberals more supportive of both types of court curbing measures, and to make conservative less

supportive of both. I also expected that informational cues indicating that justices of the Supreme

Court who are associated with the Federalist Society have voted very conservatively on abortion

cases would increase liberal support for narrowly targeted court curbing, and to have the opposite

effect on conservatives.

The results here are mixed. For liberals, there is tentative support for the hypothesis that

cues associating justices expected to form the majority coalition with Republicans, the Federalist

114

Society, or both will increase support for broadly targeted court curbing measures, but these cues

do not positively influence support for narrowly targeted court curbing. For conservatives, these

cues do not appear to influence support for either broadly targeted or narrowly targeted forms of

court curbing. Cues providing information about the especially conservative voting behavior of

justices associated with the Federalist Society in abortion cases did not significantly influence

respondents’ support for narrowly targeted court curbing, although the coefficients were

consistently signed in the expected direction.

While only one of these expectations was largely supported—liberals are more likely to

support broadly targeted court curbing when they receive a cue indicating that the majority

coalition in a case consists of Republicans, Federalist Society members, or both—that fact itself

has implications for the Court. Media coverage of Supreme Court cases decided along ideological

lines, especially if that coverage invokes a Federalist Society cue, may increase support for

broadly targeted court curbing among liberals without dampening support for broadly targeted

court curbing among conservatives. In the aggregate, this may lead to more support for broadly

targeted court curbing over time.

In additional analysis of the relationship between policy disagreement and support for

court curbing, I also uncovered some interesting dynamics concerning the ideological factors that

influence support for court curbing. Bartels and Johnston (2020) reoriented the literature on

Supreme Court legitimacy by showing that general and specific policy disagreement shape

individuals’ support for both broadly and narrowly targeted court curbing measures. While the

analyses in this study reaffirm that general and specific policy disagreement influence support for

court curbing, they also indicate that researchers evaluating individual support for court curbing

must consider the important role played by ideological polarization.

The results indicate that ideological polarization is the most important ideological factor

driving support for broadly targeted court curbing, as individuals at both the liberal and the

115

conservative extremes are more likely to support broadly targeted court curbing than are

moderates. This result is particularly important because it reveals both that ideological

polarization influences support for court courting and that there may be more support for broadly

targeted court curbing among conservatives than the prevailing wisdom would expect. While an

individual’s degree of ideological polarization also influences support for narrowly targeted court

curbing, it has the opposite effect (and a weaker one as well), as more ideologically extreme

respondents appear to be less likely to support narrowly targeted court curbing measures.

There is also some evidence that liberals are more supportive of both broadly targeted

and narrowly targeted court curbing than are conservatives, but I refrain from advocating that

conclusion. It could simply be the case that liberals were more supportive of court curbing than

were conservatives because respondents anticipated a conservative decision. Perhaps, if

respondents had expected a liberal decision, then the relationship between left-right ideology and

court curbing would be reversed. Future research should evaluate whether it truly is the case that

liberals are generally more supportive of both broadly and targeted court curbing measures, or

whether the results found here are a consequence of Dobbs v. Jackson Women’s Health being a

highly salient case for which individuals anticipated a conservative decision.

This study contributes to the nascent literature on the Federalist Society and to a larger

literature on source cues by showing that cues associating the majority coalition in a decision

with Republicans, the Federalist Society, or both may increase support for broadly targeted court

curbing among liberals, but not conservatives. It also contributes to the growing body of research

suggesting that ideological factors do influence individuals’ support for the Court. Bartels and

Johnston (2013; 2020) reoriented the literature on support for the Supreme Court by showing that

policy disagreement is an important factor influencing individuals’ support for both broadly

targeted and narrowly targeted court curbing. This study adds important nuance to our

understanding of the ideological factors that influence individual support for the Court, showing

116

not only that ideological polarization is an important factor, but also that it is the most important

ideological factor driving an individual’s support for broadly targeted court curbing.

117

Chapter Five:

Conclusion

Since its creation in law schools four decades ago, the Federalist Society has come to

play a dominant role in the United States’ legal system. The network has chapters at every

accredited law school in the United States, and these chapters are the lifeblood of the Federalist

Society. Student chapters play an important role in the socialization of conservative attorneys.

While the network does not take official policy positions, its members have played an active role

in shaping law in the United States as litigants, attorneys, amicus curiae, law clerks, judges, and

even as justices of the Supreme Court. From these positions within the legal system, Federalist

Society members have achieved dramatic policy victories for conservatives in a variety of areas

of law—most notably, on controversial landmark Supreme Court cases involving the Voting

Rights Act of 1965 (Shelby County v. Holder), campaign spending (Citizens United v. FEC),

affirmative action (Schuette v. Coalition to Defend Affirmative Action), religion (Espinoza v.

Montana Department of Revenue) and, most recently, abortion (Dobbs v. Jackson Women’s

Health). Despite the magnitude of these policy victories for the Federalist Society network,

perhaps even more significant is the network’s success at promoting its members to positions in

the United States judicial system. The Federalist Society, which can claim six of the nine

Supreme Court justices as members, has a dominant position on the highest court in the United

States both now and for the foreseeable future.

The studies here tell us much about the power of the Federalist Society. The central

questions I ask are: Where does the Federalist Society’s power come from? How does its

influence manifest in Supreme Court decision-making? And, given the Federalist Society’s

dominant role on the Supreme Court, how can we expect it to influence public support for the

Court?

118

The second chapter of this dissertation examines law schools’ Federalist Society student

chapters and reveals that the organizational strength of the network, at the grassroots level, is

associated with the presence of conservative law students and the presence of highly efficacious

and successful law students. At the elite level, chapter strength is associated with the presence of

conservative law professors and the presence of highly efficacious and successful law professors.

The ideological grievances of conservative law professors—both with respect to the average law

professor and the average alum at a law school—also influences chapter strength: When a law

school’s conservative professors are more ideologically distant from that law school’s average

professor or average alum, Federalist Society chapters are stronger.

In the third chapter, I draw from the literature on group polarization, which is a well-

known topic in social psychology and political psychology. I hypothesize that Federalist Society

members should be expected to vote especially conservatively when a case concerns a “core

policy issue.” Drawing from biographical narratives about the Federalist Society and its place in

the conservative legal movement and extant quantitative research on the Federalist Society, I

identify eight policy issues on which justices from the Federalist Society should be expected to

vote especially conservatively: abortion, affirmative action, campaign spending, desegregation,

religion, searches and seizures, sex discrimination, and the Voting Right Act of 1965. The results

indicate that on these issues do vote more conservatively than their peers on these core policy

issues. I also examine whether the especially conservative voting behavior of Federalist Society

members on the Supreme Court can be attributed to higher levels of support for states’ rights

when compared to other Republican-nominated justices. The results suggest that Federalist

Society members do not prioritize states’ rights more than do other Republican-nominated

justices. On the contrary, when a pro-state vote requires a liberal vote, Federalist Society

members are less likely than their Republican-nominated peers to cast pro-state votes. In other

words, when justices’ preferences for states’ rights and casting conservative votes collide such

THE INFLUENCE OF THE FEDERALIST SOCIETY ON JUDICAL POLITICS AND

Students also viewed